MODERN ANTENNA DESIGN Second Edition
THOMAS A. MILLIGAN
IEEE PRESS
A JOHN WILEY & SONS, INC., PUBLICATION
MODERN A...

Author:
Thomas A. Milligan

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

MODERN ANTENNA DESIGN Second Edition

THOMAS A. MILLIGAN

IEEE PRESS

A JOHN WILEY & SONS, INC., PUBLICATION

MODERN ANTENNA DESIGN

MODERN ANTENNA DESIGN Second Edition

THOMAS A. MILLIGAN

IEEE PRESS

A JOHN WILEY & SONS, INC., PUBLICATION

Copyright 2005 by John Wiley & Sons, Inc. All rights reserved. Published by John Wiley & Sons, Inc., Hoboken, New Jersey. Published simultaneously in Canada. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning, or otherwise, except as permitted under Section 107 or 108 of the 1976 United States Copyright Act, without either the prior written permission of the Publisher, or authorization through payment of the appropriate per-copy fee to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, 978-750-8400, fax 978-646-8600, or on the web at www.copyright.com. Requests to the Publisher for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, (201) 748-6011, fax (201) 748-6008. Limit of Liability/Disclaimer of Warranty: While the publisher and author have used their best efforts in preparing this book, they make no representations or warranties with respect to the accuracy or completeness of the contents of this book and speciﬁcally disclaim any implied warranties of merchantability or ﬁtness for a particular purpose. No warranty may be created or extended by sales representatives or written sales materials. The advice and strategies contained herein may not be suitable for your situation. You should consult with a professional where appropriate. Neither the publisher nor author shall be liable for any loss of proﬁt or any other commercial damages, including but not limited to special, incidental, consequential, or other damages. For general information on our other products and services please contact our Customer Care Department within the U.S. at 877-762-2974, outside the U.S. at 317-572-3993 or fax 317-572-4002. Wiley also publishes its books in a variety of electronic formats. Some content that appears in print, however, may not be available in electronic format. Library of Congress Cataloging-in-Publication Data: Milligan, Thomas A. Modern antenna design / by Thomas A. Milligan.—2nd ed. p. cm. Includes bibliographical references and index. ISBN-13 978-0-471-45776-3 (cloth) ISBN-10 0-471-45776-0 (cloth) 1. Antennas (Electronics)–Design and construction. I. Title. TK7871.6.M54 2005 621.382 4—dc22 Printed in the United States of America. 10 9 8 7 6 5 4 3 2 1

2004059098

To Mary, Jane, and Margaret

CONTENTS

Preface 1 Properties of Antennas

xv 1

1-1 1-2 1-3 1-4 1-5 1-6 1-7 1-8

Antenna Radiation, 2 Gain, 3 Effective Area, 6 Path Loss, 6 Radar Range Equation and Cross Section, 7 Why Use an Antenna? 9 Directivity, 10 Directivity Estimates, 11 1-8.1 Pencil Beam, 11 1-8.2 Butterﬂy or Omnidirectional Pattern, 13 1-9 Beam Efﬁciency, 16 1-10 Input-Impedance Mismatch Loss, 17 1-11 Polarization, 18 1-11.1 Circular Polarization Components, 19 1-11.2 Huygens Source Polarization, 21 1-11.3 Relations Between Bases, 22 1-11.4 Antenna Polarization Response, 23 1-11.5 Phase Response of Rotating Antennas, 25 1-11.6 Partial Gain, 26 1-11.7 Measurement of Circular Polarization Using Amplitude Only, 26 1-12 Vector Effective Height, 27 1-13 Antenna Factor, 29 1-14 Mutual Coupling Between Antennas, 29 1.15 Antenna Noise Temperature, 30 vii

viii

CONTENTS

1-16 Communication Link Budget and Radar Range, 35 1-17 Multipath, 36 1-18 Propagation Over Soil, 37 1-19 Multipath Fading, 39 References, 40 2 Radiation Structures and Numerical Methods 2-1 Auxiliary Vector Potentials, 43 2-1.1 Radiation from Electric Currents, 44 2-1.2 Radiation from Magnetic Currents, 49 2-2 Apertures: Huygens Source Approximation, 51 2-2.1 Near- and Far-Field Regions, 55 2-2.2 Huygens Source, 57 2-3 Boundary Conditions, 57 2-4 Physical Optics, 59 2-4.1 Radiated Fields Given Currents, 59 2-4.2 Applying Physical Optics, 60 2-4.3 Equivalent Currents, 65 2-4.4 Reactance Theorem and Mutual Coupling, 66 2-5 Method of Moments, 67 2-5.1 Use of the Reactance Theorem for the Method of Moments, 68 2-5.2 General Moments Method Approach, 69 2-5.3 Thin-Wire Moment Method Codes, 71 2-5.4 Surface and Volume Moment Method Codes, 71 2-5.5 Examples of Moment Method Models, 72 2-6 Finite-Difference Time-Domain Method, 76 2-6.1 Implementation, 76 2-6.2 Central Difference Derivative, 77 2-6.3 Finite-Difference Maxwell’s Equations, 77 2-6.4 Time Step for Stability, 79 2-6.5 Numerical Dispersion and Stability, 80 2-6.6 Computer Storage and Execution Times, 80 2-6.7 Excitation, 81 2-6.8 Waveguide Horn Example, 83 2-7 Ray Optics and the Geometric Theory of Diffraction, 84 2-7.1 Fermat’s Principle, 85 2-7.2 H -Plane Pattern of a Dipole Located Over a Finite Strip, 85 2-7.3 E-Plane Pattern of a Rectangular Horn, 87 2-7.4 H -Plane Pattern of a Rectangular Horn, 89 2-7.5 Amplitude Variations Along a Ray, 90 2-7.6 Extra Phase Shift Through Caustics, 93 2-7.7 Snell’s Laws and Reﬂection, 93 2-7.8 Polarization Effects in Reﬂections, 94 2-7.9 Reﬂection from a Curved Surface, 94 2-7.10 Ray Tracing, 96

42

CONTENTS

2-7.11 2-7.12 2-7.13 2-7.14 2-7.15 References, 100

ix

Edge Diffraction, 96 Slope Diffraction, 98 Corner Diffraction, 99 Equivalent Currents, 99 Diffraction from Curved Surfaces, 99

3 Arrays

102

3-1 Two-Element Array, 104 3-2 Linear Array of N Elements, 109 3-3 Hansen and Woodyard End-Fire Array, 114 3-4 Phased Arrays, 115 3-5 Grating Lobes, 117 3-6 Multiple Beams, 118 3-7 Planar Array, 120 3-8 Grating Lobes in Planar Arrays, 125 3-9 Mutual Impedance, 127 3-10 Scan Blindness and Array Element Pattern, 127 3-11 Compensating Array Feeding for Mutual Coupling, 128 3-12 Array Gain, 129 3-13 Arrays Using Arbitrarily Oriented Elements, 133 References, 135 4 Aperture Distributions and Array Synthesis 4-1 Amplitude Taper and Phase Error Efﬁciencies, 137 4-1.1 Separable Rectangular Aperture Distributions, 139 4-1.2 Circularly Symmetrical Distributions, 140 4-2 Simple Linear Distributions, 140 4-3 Taylor One-Parameter Linear Distribution, 144 4-4 Taylor n Line Distribution, 147 4-5 Taylor Line Distribution with Edge Nulls, 152 4-6 Elliott’s Method for Modiﬁed Taylor Distribution and Arbitrary Sidelobes, 155 4-7 Bayliss Line-Source Distribution, 158 4-8 Woodward Line-Source Synthesis, 162 4-9 Schelkunoff’s Unit-Circle Method, 164 4-10 Dolph–Chebyshev Linear Array, 170 4-11 Villeneuve Array Synthesis, 172 4-12 Zero Sampling of Continuous Distributions, 173 4-13 Fourier Series Shaped-Beam Array Synthesis, 175 4-14 Orchard Method of Array Synthesis, 178 4-15 Series-Fed Array and Traveling-Wave Feed Synthesis, 188 4-16 Circular Apertures, 191 4-17 Circular Gaussian Distribution, 194 4-18 Hansen Single-Parameter Circular Distribution, 195 4-19 Taylor Circular-Aperture Distribution, 196 4-20 Bayliss Circular-Aperture Distribution, 200

136

x

CONTENTS

4-21 4-22 4-23 4-24 4-25

Planar Arrays, 202 Convolution Technique for Planar Arrays, 203 Aperture Blockage, 208 Quadratic Phase Error, 211 Beam Efﬁciency of Circular Apertures with Axisymmetric Distribution, 214 References, 215 5 Dipoles, Slots, and Loops 5-1 5-2 5-3 5-4 5-5 5-6 5-7 5-8 5-9 5-10 5-11 5-12 5-13 5-14 5-15

5-16 5-17 5-18 5-19 5-20 5-21 5-22 5-23 5-24 5-25 5-26

Standing-Wave Currents, 218 Radiation Resistance (Conductance), 220 Babinet–Booker Principle, 222 Dipoles Located Over a Ground Plane, 223 Dipole Mounted Over Finite Ground Planes, 225 Crossed Dipoles for Circular Polarization, 231 Super Turnstile or Batwing Antenna, 234 Corner Reﬂector, 237 Monopole, 242 Sleeve Antenna, 242 Cavity-Mounted Dipole Antenna, 245 Folded Dipole, 247 Shunt Feeding, 248 Discone Antenna, 249 Baluns, 251 5-15.1 Folded Balun, 252 5-15.2 Sleeve or Bazooka Baluns, 253 5-15.3 Split Coax Balun, 255 5-15.4 Half-Wavelength Balun, 256 5-15.5 Candelabra Balun, 256 5-15.6 Ferrite Core Baluns, 256 5-15.7 Ferrite Candelabra Balun, 258 5-15.8 Transformer Balun, 258 5-15.9 Split Tapered Coax Balun, 259 5-15.10 Natural Balun, 260 Small Loop, 260 Alford Loop, 261 Resonant Loop, 263 Quadriﬁlar Helix, 264 Cavity-Backed Slots, 266 Stripline Series Slots, 266 Shallow-Cavity Crossed-Slot Antenna, 269 Waveguide-Fed Slots, 270 Rectangular-Waveguide Wall Slots, 271 Circular-Waveguide Slots, 276 Waveguide Slot Arrays, 278 5-26.1 Nonresonant Array, 279 5-26.2 Resonant Array, 282

217

CONTENTS

xi

5-26.3 Improved Design Methods, 282 References, 283 6 Microstrip Antennas

285

6-1 Microstrip Antenna Patterns, 287 6-2 Microstrip Patch Bandwidth and Surface-Wave Efﬁciency, 293 6-3 Rectangular Microstrip Patch Antenna, 299 6-4 Quarter-Wave Patch Antenna, 310 6-5 Circular Microstrip Patch, 313 6-6 Circularly Polarized Patch Antennas, 316 6-7 Compact Patches, 319 6-8 Directly Fed Stacked Patches, 323 6-9 Aperture-Coupled Stacked Patches, 325 6-10 Patch Antenna Feed Networks, 327 6-11 Series-Fed Array, 329 6-12 Microstrip Dipole, 330 6-13 Microstrip Franklin Array, 332 6-14 Microstrip Antenna Mechanical Properties, 333 References, 334 7 Horn Antennas

336

7-1 Rectangular Horn (Pyramidal), 337 7-1.1 Beamwidth, 341 7-1.2 Optimum Rectangular Horn, 343 7-1.3 Designing to Given Beamwidths, 346 7-1.4 Phase Center, 347 7-2 Circular-Aperture Horn, 348 7-2.1 Beamwidth, 350 7-2.2 Phase Center, 352 7-3 Circular (Conical) Corrugated Horn, 353 7-3.1 Scalar Horn, 357 7-3.2 Corrugation Design, 357 7-3.3 Choke Horns, 358 7-3.4 Rectangular Corrugated Horns, 359 7-4 Corrugated Ground Plane, 359 7-5 Gaussian Beam, 362 7-6 Ridged Waveguide Horns, 365 7-7 Box Horn, 372 7-8 T-Bar-Fed Slot Antenna, 374 7-9 Multimode Circular Horn, 376 7-10 Biconical Horn, 376 References, 378 8 Reﬂector Antennas 8-1 Paraboloidal Reﬂector Geometry, 381 8-2 Paraboloidal Reﬂector Aperture Distribution Losses, 383

380

xii

CONTENTS

8-3 8-4 8-5 8-6 8-7 8-8 8-9 8-10 8-11 8-12 8-13

Approximate Spillover and Amplitude Taper Trade-offs, 385 Phase Error Losses and Axial Defocusing, 387 Astigmatism, 389 Feed Scanning, 390 Random Phase Errors, 393 Focal Plane Fields, 396 Feed Mismatch Due to the Reﬂector, 397 Front-to-Back Ratio, 399 Offset-Fed Reﬂector, 399 Reﬂections from Conic Sections, 405 Dual-Reﬂector Antennas, 408 8-13.1 Feed Blockage, 410 8-13.2 Diffraction Loss, 413 8-13.3 Cassegrain Tolerances, 414 8-14 Feed and Subreﬂector Support Strut Radiation, 416 8-15 Gain/Noise Temperature of a Dual Reﬂector, 421 8-16 Displaced-Axis Dual Reﬂector, 421 8-17 Offset-Fed Dual Reﬂector, 424 8-18 Horn Reﬂector and Dragonian Dual Reﬂector, 427 8-19 Spherical Reﬂector, 429 8-20 Shaped Reﬂectors, 432 8-20.1 Cylindrical Reﬂector Synthesis, 433 8-20.2 Circularly Symmetrical Reﬂector Synthesis, 434 8-20.3 Doubly Curved Reﬂector for Shaped Beams, 437 8-20.4 Dual Shaped Reﬂectors, 439 8-21 Optimization Synthesis of Shaped and Multiple-Beam Reﬂectors, 442 References, 443 9 Lens Antennas

447

9-1 9-2 9-3 9-4 9-5 9-6 9-7 9-8

Single Refracting Surface Lenses, 448 Zoned Lenses, 451 General Two-Surface Lenses, 454 Single-Surface or Contact Lenses, 459 Metal Plate Lenses, 461 Surface Mismatch and Dielectric Losses, 463 Feed Scanning of a Hyperboloidal Lens, 464 Dual-Surface Lenses, 465 9-8.1 Coma-Free Axisymmetric Dielectric Lens, 466 9-8.2 Speciﬁed Aperture Distribution Axisymmetric Dielectric Lens, 468 9-9 Bootlace Lens, 470 9-10 Luneburg Lens, 472 References, 472 10

Traveling-Wave Antennas 10-1 General Traveling Waves, 475

474

CONTENTS

xiii

10-1.1 Slow Wave, 478 10-1.2 Fast Waves (Leaky Wave Structure), 480 10-2 Long Wire Antennas, 481 10-2.1 Beverage Antenna, 481 10-2.2 V Antenna, 482 10-2.3 Rhombic Antenna, 483 10-3 Yagi–Uda Antennas, 485 10-3.1 Multiple-Feed Yagi–Uda Antennas, 492 10-3.2 Resonant Loop Yagi–Uda Antennas, 495 10-4 Corrugated Rod (Cigar) Antenna, 497 10-5 Dielectric Rod (Polyrod) Antenna, 499 10-6 Helical Wire Antenna, 502 10-6.1 Helical Modes, 503 10-6.2 Axial Mode, 504 10-6.3 Feed of a Helical Antenna, 506 10-6.4 Long Helical Antenna, 507 10-6.5 Short Helical Antenna, 508 10-7 Short Backﬁre Antenna, 509 10-8 Tapered Slot Antennas, 512 10-9 Leaky Wave Structures, 516 References, 518 11 Frequency-Independent Antennas Spiral 11-1 11-2 11-3 11-4 11-5

Antennas, 522 Modal Expansion of Antenna Patterns, 524 Archimedean Spiral, 526 Equiangular Spiral, 527 Pattern Analysis of Spiral Antennas, 530 Spiral Construction and Feeding, 535 11-5.1 Spiral Construction, 535 11-5.2 Balun Feed, 536 11-5.3 Inﬁnite Balun, 538 11-5.4 Beamformer and Coaxial Line Feed, 538 11-6 Spiral and Beamformer Measurements, 538 11-7 Feed Network and Antenna Interaction, 540 11-8 Modulated Arm Width Spiral, 541 11-9 Conical Log Spiral Antenna, 543 11-10 Mode 2 Conical Log Spiral Antenna, 549 11-11 Feeding Conical Log Spirals, 550 Log-Periodic Antennas, 550 11-12 Log-Periodic Dipole Antenna, 551 11-12.1 Feeding a Log-Periodic Dipole Antenna, 556 11-12.2 Phase Center, 558 11-12.3 Elevation Angle, 559 11-12.4 Arrays of Log-Periodic Dipole Antennas, 560 11-13 Other Log-Periodic Types, 561 11-14 Log-Periodic Antenna Feeding Paraboloidal Reﬂector, 563

521

xiv

CONTENTS

11-15 V Log-Periodic Array, 567 11-16 Cavity-Backed Planar Log-Periodic Antennas, 569 References, 571 12

Phased Arrays

573

12-1 12-2 12-3 12-4

Fixed Phase Shifters (Phasers), 574 Quantization Lobes, 578 Array Errors, 580 Nonuniform and Random Element Existence Arrays, 582 12-4.1 Linear Space Tapered Array, 582 12-4.2 Circular Space Tapered Array, 584 12-4.3 Statistically Thinned Array, 587 12-5 Array Element Pattern, 588 12-6 Feed Networks, 590 12-6.1 Corporate Feed, 590 12-6.2 Series Feed, 592 12-6.3 Variable Power Divider and Phase Shifter, 592 12-6.4 Butler Matrix, 594 12-6.5 Space Feeding, 596 12-6.6 Tapered Feed Network with Uniform-Amplitude Subarrays, 597 12-7 Pattern Null Formation in Arbitrary Array, 599 12-8 Phased Array Application to Communication Systems, 601 12-9 Near-Field Measurements on Phased Arrays, 602 References, 604 Index

607

PREFACE

I wrote this book from my perspective as a designer in industry, primarily for other designers and users of antennas. On occasion I have prepared and taught antenna courses, for which I developed a systematic approach to the subject. For the last decade I have edited the “Antenna Designer’s Notebook” column in the IEEE antenna magazine. This expanded edition adds a combination of my own design notebook and the many other ideas provided to me by others, leading to this collection of ideas that I think designers should know. The book contains a systematic approach to the subject. Every author would like to be read from front to back, but my own career assignments would have caused to me to jump around in this book. Nevertheless, Chapter 1 covers those topics that every user and designer should know. Because I deal with complete antenna design, which includes mounting the antenna, included are the effects of nearby structures and how they can be used to enhance the response. We all study ideal antennas ﬂoating in free space to help us understand the basics, but the real world is a little different. Instead of drawing single line graphs of common relationships between two parameters, I generated scales for calculations that I perform over and over. I did not supply a set of computer programs because I seldom use collections supplied by others, and younger engineers ﬁnd my programs quaint, as each generation learns a different computer language. You’ll learn by writing your own. IEEE Antennas and Propagation Society’s digital archive of all material published from 1952 to 2000 has changed our approach to research. I have not included extensive bibliographies, because I believe that it is no longer necessary. The search engine of the archive can supply an exhaustive list. I referred only to papers that I found particularly helpful. Complete sets of the transactions are available in libraries, whereas the wealth of information in the archive from conferences was not. I have started mining this information, which contains many useful design ideas, and have incorporated some of them in this book. In this ﬁeld, 40-year-old publications are still useful and we should not reinvent methods. Many clever ideas from industry are usually published xv

xvi

PREFACE

only once, if at all, and personally, I’ll be returning to this material again and again, because all books have limited space. As with the ﬁrst edition, I enjoyed writing this book because I wanted to express my point of view of a rewarding ﬁeld. Although the amount of information available is overwhelming and the mathematics describing it can cloud the ideas, I hope my explanations help you develop new products or use old ones. I would like to thank all the authors who taught me this subject by sharing their ideas, especially those working in industry. On a personal note I thank the designers at Lockheed Martin, who encouraged me and reviewed material while I wrote: in particular, Jeannette McDonnell, Thomas Cencich, Donald Huebner, and Julie Huffman. THOMAS A. MILLIGAN

1 PROPERTIES OF ANTENNAS

One approach to an antenna book starts with a discussion of how antennas radiate. Beginning with Maxwell’s equations, we derive electromagnetic waves. After that lengthy discussion, which contains a lot of mathematics, we discuss how these waves excite currents on conductors. The second half of the story is that currents radiate and produce electromagnetic waves. You may already have studied that subject, or if you wish to further your background, consult books on electromagnetics. The study of electromagnetics gives insight into the mathematics describing antenna radiation and provides the rigor to prevent mistakes. We skip the discussion of those equations and move directly to practical aspects. It is important to realize that antennas radiate from currents. Design consists of controlling currents to produce the desired radiation distribution, called its pattern. In many situations the problem is how to prevent radiation from currents, such as in circuits. Whenever a current becomes separated in distance from its return current, it radiates. Simply stated, we design to keep the two currents close together, to reduce radiation. Some discussions will ignore the current distribution and instead, consider derived quantities, such as ﬁelds in an aperture or magnetic currents in a slot or around the edges of a microstrip patch. You will discover that we use any concept that provides insight or simpliﬁes the mathematics. An antenna converts bound circuit ﬁelds into propagating electromagnetic waves and, by reciprocity, collects power from passing electromagnetic waves. Maxwell’s equations predict that any time-varying electric or magnetic ﬁeld produces the opposite ﬁeld and forms an electromagnetic wave. The wave has its two ﬁelds oriented orthogonally, and it propagates in the direction normal to the plane deﬁned by the perpendicular electric and magnetic ﬁelds. The electric ﬁeld, the magnetic ﬁeld, and the direction of propagation form a right-handed coordinate system. The propagating wave ﬁeld intensity decreases by 1/R away from the source, whereas a static ﬁeld Modern Antenna Design, Second Edition, By Thomas A. Milligan Copyright 2005 John Wiley & Sons, Inc.

1

2

PROPERTIES OF ANTENNAS

drops off by 1/R 2 . Any circuit with time-varying ﬁelds has the capability of radiating to some extent. We consider only time-harmonic ﬁelds and use phasor notation with time dependence ej ωt . An outward-propagating wave is given by e−j (kR−ωt) , where k, the wave number, is given by 2π/λ. λ is the wavelength of the wave given by c/f , where c is the velocity of light (3 × 108 m/s in free space) and f is the frequency. Increasing the distance from the source decreases the phase of the wave. Consider a two-wire transmission line with ﬁelds bound to it. The currents on a single wire will radiate, but as long as the ground return path is near, its radiation will nearly cancel the other line’s radiation because the two are 180◦ out of phase and the waves travel about the same distance. As the lines become farther and farther apart, in terms of wavelengths, the ﬁelds produced by the two currents will no longer cancel in all directions. In some directions the phase delay is different for radiation from the current on each line, and power escapes from the line. We keep circuits from radiating by providing close ground returns. Hence, high-speed logic requires ground planes to reduce radiation and its unwanted crosstalk.

1-1 ANTENNA RADIATION Antennas radiate spherical waves that propagate in the radial direction for a coordinate system centered on the antenna. At large distances, spherical waves can be approximated by plane waves. Plane waves are useful because they simplify the problem. They are not physical, however, because they require inﬁnite power. The Poynting vector describes both the direction of propagation and the power density of the electromagnetic wave. It is found from the vector cross product of the electric and magnetic ﬁelds and is denoted S: S = E × H∗

W/m2

Root mean square (RMS) values are used to express the magnitude of the ﬁelds. H∗ is the complex conjugate of the magnetic ﬁeld phasor. The magnetic ﬁeld is proportional to the electric ﬁeld in the far ﬁeld. The constant of proportion is η, the impedance of free space (η = 376.73 ): |S| = S =

|E|2 η

W/m2

(1-1)

Because the Poynting vector is the vector product of the two ﬁelds, it is orthogonal to both ﬁelds and the triplet deﬁnes a right-handed coordinate system: (E, H, S). Consider a pair of concentric spheres centered on the antenna. The ﬁelds around the antenna decrease as 1/R, 1/R 2 , 1/R 3 , and so on. Constant-order terms would require that the power radiated grow with distance and power would not be conserved. For ﬁeld terms proportional to 1/R 2 , 1/R 3 , and higher, the power density decreases with distance faster than the area increases. The energy on the inner sphere is larger than that on the outer sphere. The energies are not radiated but are instead concentrated around the antenna; they are near-ﬁeld terms. Only the 1/R 2 term of the Poynting vector (1/R ﬁeld terms) represents radiated power because the sphere area grows as R 2 and

GAIN

3

gives a constant product. All the radiated power ﬂowing through the inner sphere will propagate to the outer sphere. The sign of the input reactance depends on the near-ﬁeld predominance of ﬁeld type: electric (capacitive) or magnetic (inductive). At resonance (zero reactance) the stored energies due to the near ﬁelds are equal. Increasing the stored ﬁelds increases the circuit Q and narrows the impedance bandwidth. Far from the antenna we consider only the radiated ﬁelds and power density. The power ﬂow is the same through concentric spheres: 4πR12 S1,avg = 4πR22 S2,avg The average power density is proportional to 1/R 2 . Consider differential areas on the two spheres at the same coordinate angles. The antenna radiates only in the radial direction; therefore, no power may travel in the θ or φ direction. Power travels in ﬂux tubes between areas, and it follows that not only the average Poynting vector but also every part of the power density is proportional to 1/R 2 : S1 R12 sin θ dθ dφ = S2 R22 sin θ dθ dφ Since in a radiated wave S is proportional to 1/R 2 , E is proportional to 1/R. It is convenient to deﬁne radiation intensity to remove the 1/R 2 dependence: U (θ, φ) = S(R, θ, φ)R 2

W/solid angle

Radiation intensity depends only on the direction of radiation and remains the same at all distances. A probe antenna measures the relative radiation intensity (pattern) by moving in a circle (constant R) around the antenna. Often, of course, the antenna rotates and the probe is stationary. Some patterns have established names. Patterns along constant angles of the spherical coordinates are called either conical (constant θ ) or great circle (constant φ). The great circle cuts when φ = 0◦ or φ = 90◦ are the principal plane patterns. Other named cuts are also used, but their names depend on the particular measurement positioner, and it is necessary to annotate these patterns carefully to avoid confusion between people measuring patterns on different positioners. Patterns are measured by using three scales: (1) linear (power), (2) square root (ﬁeld intensity), and (3) decibels (dB). The dB scale is used the most because it reveals more of the low-level responses (sidelobes). Figure 1-1 demonstrates many characteristics of patterns. The half-power beamwidth is sometimes called just the beamwidth. The tenth-power and null beamwidths are used in some applications. This pattern comes from a parabolic reﬂector whose feed is moved off the axis. The vestigial lobe occurs when the ﬁrst sidelobe becomes joined to the main beam and forms a shoulder. For a feed located on the axis of the parabola, the ﬁrst sidelobes are equal.

1-2 GAIN Gain is a measure of the ability of the antenna to direct the input power into radiation in a particular direction and is measured at the peak radiation intensity. Consider the

4

PROPERTIES OF ANTENNAS

Angle 3 dB

Half-power beamwidth

Pattern level

10 dB

Tenth-power beamwidth

20 dB

Sidelobes

Vestigial lobe 30 dB

Null beamwidth

40 dB

FIGURE 1-1

Antenna pattern characteristics.

power density radiated by an isotropic antenna with input power P0 at a distance R: S = P0 /4πR 2 . An isotropic antenna radiates equally in all directions, and its radiated power density S is found by dividing the radiated power by the area of the sphere 4πR 2 . The isotropic radiator is considered to be 100% efﬁcient. The gain of an actual antenna increases the power density in the direction of the peak radiation: P0 G |E|2 1 P0 Gη S= = (1-2) or |E| = = Sη 4πR 2 η R 4π Gain is achieved by directing the radiation away from other parts of the radiation sphere. In general, gain is deﬁned as the gain-biased pattern of the antenna: P0 G(θ, φ) 4πR 2 P0 G(θ, φ) U (θ, φ) = 4π S(θ, φ) =

power density radiation intensity

(1-3)

The surface integral of the radiation intensity over the radiation sphere divided by the input power P0 is a measure of the relative power radiated by the antenna, or the antenna efﬁciency: 2π π Pr G(θ, φ) sin θ dθ dφ = ηe = efﬁciency P0 4π 0 0

GAIN

5

where Pr is the radiated power. Material losses in the antenna or reﬂected power due to poor impedance match reduce the radiated power. In this book, integrals in the equation above and those that follow express concepts more than operations we perform during design. Only for theoretical simpliﬁcations of the real world can we ﬁnd closed-form solutions that would call for actual integration. We solve most integrals by using numerical methods that involve breaking the integrand into small segments and performing a weighted sum. However, it is helpful that integrals using measured values reduce the random errors by averaging, which improves the result. In a system the transmitter output impedance or the receiver input impedance may not match the antenna input impedance. Peak gain occurs for a receiver impedance conjugate matched to the antenna, which means that the resistive parts are the same and the reactive parts are the same magnitude but have opposite signs. Precision gain measurements require a tuner between the antenna and receiver to conjugate-match the two. Alternatively, the mismatch loss must be removed by calculation after the measurement. Either the effect of mismatches is considered separately for a given system, or the antennas are measured into the system impedance and mismatch loss is considered to be part of the efﬁciency. Example Compute the peak power density at 10 km of an antenna with an input power of 3 W and a gain of 15 dB. First convert dB gain to a ratio: G = 1015/10 = 31.62. The power spreads over the sphere area with radius 10 km or an area of 4π(104 )2 m2 . The power density is S=

(3 W)(31.62) = 75.5 nW/m2 4π × 108 m2

We calculate the electric ﬁeld intensity using Eq. (1-2): |E| = Sη = (75.5 × 10−9 )(376.7) = 5333 µV/m Although gain is usually relative to an isotropic antenna, some antenna gains are referred to a λ/2 dipole with an isotropic gain of 2.14 dB. If we approximate the antenna as a point source, we compute the electric ﬁeld radiated by using Eq. (1-2): e−j kR P0 G(θ, φ)η E(θ, φ) = (1-4) R 4π This requires only that the antenna be small compared to the radial distance R. Equation (1-4) ignores the direction of the electric ﬁeld, which we deﬁne as polarization. The units of the electric ﬁeld are volts/meter. We determine the far-ﬁeld pattern by multiplying Eq. (1-4) by R and removing the phase term e−j kR since phase has meaning only when referred to another point in the far ﬁeld. The far-ﬁeld electric ﬁeld Eff unit is volts: 2 P0 G(θ, φ)η 4π 1 Eff (θ, φ) (1-5) Eff (θ, φ) = or G(θ, φ) = 4π P0 η During analysis, we often normalize input power to √ 1 W and can compute gain easily from the electric ﬁeld by multiplying by a constant 4π/η = 0.1826374.

6

PROPERTIES OF ANTENNAS

1-3 EFFECTIVE AREA Antennas capture power from passing waves and deliver some of it to the terminals. Given the power density of the incident wave and the effective area of the antenna, the power delivered to the terminals is the product Pd = SAeff

(1-6)

For an aperture antenna such as a horn, parabolic reﬂector, or ﬂat-plate array, effective area is physical area multiplied by aperture efﬁciency. In general, losses due to material, distribution, and mismatch reduce the ratio of the effective area to the physical area. Typical estimated aperture efﬁciency for a parabolic reﬂector is 55%. Even antennas with inﬁnitesimal physical areas, such as dipoles, have effective areas because they remove power from passing waves. 1-4 PATH LOSS [1, p. 183] We combine the gain of the transmitting antenna with the effective area of the receiving antenna to determine delivered power and path loss. The power density at the receiving antenna is given by Eq. (1-3), and the received power is given by Eq. (1-6). By combining the two, we obtain the path loss: Pd A2 G1 (θ, φ) = Pt 4πR 2 Antenna 1 transmits, and antenna 2 receives. If the materials in the antennas are linear and isotropic, the transmitting and receiving patterns are identical (reciprocal) [2, p. 116]. When we consider antenna 2 as the transmitting antenna and antenna 1 as the receiving antenna, the path loss is Pd A1 G2 (θ, φ) = Pt 4πR 2 Since the responses are reciprocal, the path losses are equal and we can gather and eliminate terms: G1 G2 = = constant A1 A2 Because the antennas were arbitrary, this quotient must equal a constant. This constant was found by considering the radiation between two large apertures [3]: 4π G = 2 A λ

(1-7)

We substitute this equation into path loss to express it in terms of the gains or effective areas: Pd A1 A2 λ 2 = G1 G2 = 2 2 (1-8) Pt 4πR λR We make quick evaluations of path loss for various units of distance R and for frequency f in megahertz using the formula path loss(dB) = KU + 20 log(f R) − G1 (dB) − G2 (dB)

(1-9)

RADAR RANGE EQUATION AND CROSS SECTION

7

where KU depends on the length units: KU

Unit

km 32.45 nm 37.80 miles 36.58 m −27.55 ft −37.87 Example Compute the gain of a 3-m-diameter parabolic reﬂector at 4 GHz assuming 55% aperture efﬁciency. Gain is related to effective area by Eq. (1-7): G=

4πA λ2

We calculate the area of a circular aperture by A = π(D/2)2 . By combining these equations, we have πD 2 πDf 2 G= ηa = ηa (1-10) λ c where D is the diameter and ηa is the aperture efﬁciency. On substituting the values above, we obtain the gain:

3π(4 × 109 ) G= 0.3 × 109

2 (0.55) = 8685

(39.4 dB)

Example Calculate the path loss of a 50-km communication link at 2.2 GHz using a transmitter antenna with a gain of 25 dB and a receiver antenna with a gain of 20 dB. Path loss = 32.45 + 20 log[2200(50)] − 25 − 20 = 88.3 dB What happens to transmission between two apertures as the frequency is increased? If we assume that the effective area remains constant, as in a parabolic reﬂector, the transmission increases as the square of frequency: Pd A1 A2 1 A1 A2 f 2 = = = Bf 2 Pt R 2 λ2 R2 c where B is a constant for a ﬁxed range. The receiving aperture captures the same power regardless of frequency, but the gain of the transmitting antenna increases as the square of frequency. Hence, the received power also increases as frequency squared. Only for antennas, whose gain is a ﬁxed value when frequency changes, does the path loss increase as the square of frequency. 1-5 RADAR RANGE EQUATION AND CROSS SECTION Radar operates using a double path loss. The radar transmitting antenna radiates a ﬁeld that illuminates a target. These incident ﬁelds excite surface currents that also radiate

8

PROPERTIES OF ANTENNAS

to produce a second ﬁeld. These ﬁelds propagate to the receiving antenna, where they are collected. Most radars use the same antenna both to transmit the ﬁeld and to collect the signal returned, called a monostatic system, whereas we use separate antennas for bistatic radar. The receiving system cannot be detected in a bistatic system because it does not transmit and has greater survivability in a military application. We determine the power density illuminating the target at a range RT by using Eq. (1-2): PT GT (θ, φ) Sinc = (1-11) 4πRT2 The target’s radar cross section (RCS), the scattering area of the object, is expressed in square meters or dBm2 : 10 log(square meters). The RCS depends on both the incident and reﬂected wave directions. We multiply the power collected by the target with its receiving pattern by the gain of the effective antenna due to the currents induced: RCS = σ =

powerreﬂected Ps (θr , φr , θi , φi ) = power density incident PT GT /4πRT2

(1-12)

In a communication system we call Ps the equivalent isotropic radiated power (EIRP), which equals the product of the input power and the antenna gain. The target becomes the transmitting source and we apply Eq. (1-2) to ﬁnd the power density at the receiving antenna at a range RR from the target. Finally, the receiving antenna collects the power density with an effective area AR . We combine these ideas to obtain the power delivered to the receiver: AR PT GT σ (θr , φr , θi , φi ) Prec = SR AR = (4πRT2 )(4πRR2 ) We apply Eq. (1-7) to eliminate the effective area of the receiving antenna and gather terms to determine the bistatic radar range equation: Prec GT GR λ2 σ (θr , φr , θi , φi ) = PT (4π)3 RT2 RR2

(1-13)

We reduce Eq. (1-13) and collect terms for monostatic radar, where the same antenna is used for both transmitting and receiving: Prec G2 λ2 σ = PT (4π)3 R 4 Radar received power is proportional to 1/R 4 and to G2 . We ﬁnd the approximate RCS of a ﬂat plate by considering the plate as an antenna with an effective area. Equation (1-11) gives the power density incident on the plate that collects this power over an area AR : PC =

PT GT (θ, φ) AR 4πRT2

The power scattered by the plate is the power collected, PC , times the gain of the plate as an antenna, GP : Ps = PC GP =

PT GT (θi , φi ) AR GP (θr , φr ) 4πRT2

WHY USE AN ANTENNA?

9

This scattered power is the effective radiated power in a particular direction, which in an antenna is the product of the input power and the gain in a particular direction. We calculate the plate gain by using the effective area and ﬁnd the scattered power in terms of area: PT GT 4πA2R Ps = 4πRT2 λ2 We determine the RCS σ by Eq. (1-12), the scattered power divided by the incident power density: σ =

Ps 4πA2R GR (θi , φi )GR (θr , φr )λ2 = = λ2 4π PT GT /4πRT2

(1-14)

The right expression of Eq. (1-14) divides the gain into two pieces for bistatic scattering, where the scattered direction is different from the incident direction. Monostatic scattering uses the same incident and reﬂected directions. We can substitute any object for the ﬂat plate and use the idea of an effective area and its associated antenna gain. An antenna is an object with a unique RCS characteristic because part of the power received will be delivered to the antenna terminals. If we provide a good impedance match to this signal, it will not reradiate and the RCS is reduced. When we illuminate an antenna from an arbitrary direction, some of the incident power density will be scattered by the structure and not delivered to the antenna terminals. This leads to the division of antenna RCS into the antenna mode of reradiated signals caused by terminal mismatch and the structural mode, the ﬁelds reﬂected off the structure for incident power density not delivered to the terminals. 1-6 WHY USE AN ANTENNA? We use antennas to transfer signals when no other way is possible, such as communication with a missile or over rugged mountain terrain. Cables are expensive and take a long time to install. Are there times when we would use antennas over level ground? The large path losses of antenna systems lead us to believe that cable runs are better. Example Suppose that we must choose between using a low-loss waveguide run and a pair of antennas at 3 GHz. Each antenna has 10 dB of gain. The low-loss waveguide has only 19.7 dB/km loss. Table 1-1 compares losses over various distances. The waveguide link starts out with lower loss, but the antenna system soon overtakes it. When the path length doubles, the cable link loss also doubles in decibels, but an antenna link TABLE 1-1 Losses Over Distance Distance (km)

Waveguide Loss (dB)

Antenna Path Loss (dB)

2 4 6 10

39.4 78.8 118.2 197

88 94 97.6 102

10

PROPERTIES OF ANTENNAS

increases by only 6 dB. As the distance is increased, radiating between two antennas eventually has lower losses than in any cable. Example A 200-m outside antenna range was set up to operate at 2 GHz using a 2-mdiameter reﬂector as a source. The receiver requires a sample of the transmitter signal to phase-lock the local oscillator and signal at a 45-MHz difference. It was proposed to run an RG/U 115 cable through the power and control cable conduit, since the run was short. The cable loss was 36 dB per 100 m, giving a total cable loss of 72 dB. A 10-dB coupler was used on the transmitter to pick off the reference signal, so the total loss was 82 dB. Since the source transmitted 100 mW (20 dBm), the signal was −62 dBm at the receiver, sufﬁcient for phase lock. A second proposed method was to place a standard-gain horn (15 dB of gain) within the beam of the source on a small stand out of the way of the measurement and next to the receiver. If we assume that the source antenna had only 30% aperture efﬁciency, we compute gain from Eq. (1-10) (λ = 0.15 m): G=

2π 0.15

2 (0.3) = 526

(27.2 dB)

The path loss is found from Eq. (1-9) for a range of 0.2 km: 32.45 + 20 log[2000(0.2)] − 27.2 − 15 = 42.3 dB The power delivered out of the horn is 20 dBm − 42.3 dB = −22.3 dBm. A 20-dB attenuator must be put on the horn to prevent saturation of the receiver (−30 dBm). Even with a short run, it is sometimes better to transmit the signal between two antennas instead of using cables. 1-7 DIRECTIVITY Directivity is a measure of the concentration of radiation in the direction of the maximum: maximum radiation intensity Umax directivity = = (1-15) average radiation intensity U0 Directivity and gain differ only by the efﬁciency, but directivity is easily estimated from patterns. Gain—directivity times efﬁciency—must be measured. The average radiation intensity can be found from a surface integral over the radiation sphere of the radiation intensity divided by 4π, the area of the sphere in steradians: 2π π 1 average radiation intensity = U (θ, φ) sin θ dθ dφ = U0 (1-16) 4π 0 0 This is the radiated power divided by the area of a unit sphere. The radiation intensity U (θ, φ) separates into a sum of co- and cross-polarization components: U0 =

1 4π

2π 0

π 0

[UC (θ, φ) + U× (θ, φ)] sin θ dθ dφ

(1-17)

DIRECTIVITY ESTIMATES

11

Both co- and cross-polarization directivities can be deﬁned: directivityC =

UC,max U0

directivity× =

U×,max U0

(1-18)

Directivity can also be deﬁned for an arbitrary direction D(θ, φ) as radiation intensity divided by the average radiation intensity, but when the coordinate angles are not speciﬁed, we calculate directivity at Umax . 1-8 DIRECTIVITY ESTIMATES Because a ratio of radiation intensities is used to calculate directivity, the pattern may be referred to any convenient level. The most accurate estimate is based on measurements at equal angle increments over the entire radiation sphere. The average may be found from coarse measurements by using numerical integration, but the directivity measured is affected directly by whether the maximum is found. The directivity of antennas with well-behaved patterns can be estimated from one or two patterns. Either the integral over the pattern is approximated or the pattern is approximated with a function whose integral is found exactly. 1-8.1 Pencil Beam By estimating the integral, Kraus [4] devised a method for pencil beam patterns with its peak at θ = 0◦ . Given the half-power beamwidths of the principal plane patterns, the integral is approximately the product of the beamwidths. This idea comes from circuit theory, where the integral of a time pulse is approximately the pulse width (3-dB points) times the pulse peak: U0 = θ1 θ2 /4π, where θ1 and θ2 are the 3-dB beamwidths, in radians, of the principal plane patterns: directivity =

4π 41,253 (rad) = (deg) θ1 θ2 θ1 θ2

(1-19)

Example Estimate the directivity of an antenna with E- and H -plane (principal plane) pattern beamwidths of 24◦ and 36◦ . Directivity =

41,253 = 47.75 (16.8 dB) 24(36)

An analytical function, cos2N (θ/2), approximates a broad pattern centered on θ = 0◦ with a null at θ = 180◦ : U (θ ) = cos2N (θ/2)

or

E = cosN (θ/2)

The directivity of this pattern can be computed exactly. The characteristics of the approximation are related to the beamwidth at a speciﬁed level, Lvl(dB): beamwidth [Lvl(dB)] = 4 cos−1 (10−Lvl(dB)/20N ) N=

−Lvl(dB) 20 log[cos(beamwidthLvl(dB) /4)]

directivity = N + 1 (ratio)

(1-20a) (1-20b) (1-20c)

12

PROPERTIES OF ANTENNAS Directivity, dB

3-dB Beamwidth

SCALE 1-1 3-dB beamwidth and directivity relationship for cos2N (θ /2) pattern.

Directivity, dB

10-dB Beamwidth

SCALE 1-2 10-dB beamwidth and directivity relationship for cos2N (θ /2) pattern.

Scales 1-1 and 1-2, which give the relationship between beamwidth and directivity using Eq. (1-20), are useful for quick conversion between the two properties. You can use the two scales to estimate the 10-dB beamwidth given the 3-dB beamwidth. For example, an antenna with a 90◦ 3-dB beamwidth has a directivity of about 7.3 dB. You read from the lower scale that an antenna with 7.3-dB directivity has a 159.5◦ 10-dB beamwidth. Another simple way to determine the beamwidths at different pattern levels is the square-root factor approximation: BW[Lvl 2(dB)] Lvl 2(dB) = BW[Lvl 1(dB)] Lvl 1(dB) By this factor, beamwidth10 dB = 1.826 beamwidth3 dB ; an antenna with a 90◦ 3-dB beamwidth has a (1.826)90◦ = 164.3◦ 10-dB beamwidth. This pattern approximation requires equal principal plane beamwidths, but we use an elliptical approximation with unequal beamwidths: U (θ, φ) = cos2Ne (θ/2) cos2 φ + cos2Nh (θ/2) sin2 φ

(1-21)

where Ne and Nh are found from the principal plane beamwidths. We combine the directivities calculated in the principal planes by the simple formula directivity (ratio) =

2 · directivitye · directivityh directivitye + directivityh

(1-22)

Example Estimate the directivity of an antenna with E- and H -plane pattern beamwidths of 98◦ and 140◦ . From the scale we read a directivity of 6.6 dB in the E-plane and 4.37 dB in the H -plane. We convert these to ratios and apply Eq. (1-22): directivity (ratio) =

2(4.57)(2.74) = 3.426 or 4.57 + 2.74

10 log(3.426) = 5.35 dB

DIRECTIVITY ESTIMATES

13

Many analyses of paraboloidal reﬂectors use a feed pattern approximation limited to the front hemisphere with a zero pattern in the back hemisphere: U (θ ) = cos2N θ

or

E = cosN θ

◦

for θ ≤ π/2(90 )

The directivity of this pattern can be found exactly, and the characteristics of the approximation are beamwidth [Lvl(dB)] = 2 cos−1 (10−Lvl(dB)/20N ) −Lvl(dB) N= 20 log[cos(beamwidthLvl(dB) /2)] directivity = 2(2N + 1) (ratio)

(1-23a) (1-23b) (1-23c)

We use the elliptical model [Eq. (1-21)] with this approximate pattern and use Eq. (1-22) to estimate the directivity when the E- and H -plane beamwidths are different. 1-8.2 Butterﬂy or Omnidirectional Pattern Many antennas have nulls at θ = 0◦ with rotational symmetry about the z-axis (Figure 1-2). Neither of the directivity estimates above can be used with these patterns because they require the beam peak to be at θ = 0◦ . We generate this type of antenna pattern by using mode 2 log-periodic conical spirals, shaped reﬂectors, some higherorder-mode waveguide horns, biconical horns, and traveling-wave antennas. A formula similar to Kraus’s can be found if we assume that all the power is between the 3-dB beamwidth angles θ1 and θ2 : 1 θ2 cos θ1 − cos θ2 sin θ dθ = U0 = 2 θ1 2 Rotational symmetry eliminates integration over φ: directivity =

2 Umax = U0 cos θ1 − cos θ2

(1-24)

FIGURE 1-2 Omnidirectional antenna pattern with sidelobes scanned above the horizon.

14

PROPERTIES OF ANTENNAS Directivity, dB of Omnipattern at q = 90°

3-dB Beamwidth

SCALE 1-3 Relationship between 3-dB beamwidth of omnidirectional pattern and directivity.

Example A pattern with rotational symmetry has half-power points at 35◦ and 75◦ . Estimate the directivity. Directivity =

2 = 3.57 (5.5 dB) cos 35◦ − cos 75◦

If the pattern also has symmetry about the θ = 90◦ plane, the integral for the average radiation intensity has limits from 0 to π/2. Equation (1-24) reduces to directivity = 1/ cos θ1 . Example A rotationally symmetric pattern with a maximum at 90◦ has a 45◦ beamwidth. Estimate the directivity. θ1 = 90◦ − 45◦ /2 = 67.5◦ , so directivity =

1 = 2.61 cos 67.5◦

(4.2 dB)

The pattern can be approximated by the function U (θ ) = B sin2M (θ/2) cos2N (θ/2) but the directivity estimates found by integrating this function show only minor improvements over Eq. (1-24). Nevertheless, we can use the expression for analytical patterns. Given beam edges θL and θU at a level Lvl(dB), we ﬁnd the exponential factors. √ ln[cos(θU /2)] − ln[cos(θL /2)] AA = and T M2 = tan−1 AA ln[sin(θL /2)] − ln[sin(θU /2)] −|Lvl(dB)|/8.68589 N= AA{ln[sin(θL /2)] − ln(sin T M2 )} + ln(cos(θL /2)) − ln(cos T M2 ) M = AA(N ) A second pattern model of an omnidirectional pattern based on the pattern function with minor sidelobes and a beam peak at θ0 measured from the symmetry axis is sin[b(θ0 − θ )] b(θ0 − θ ) We estimate the directivity from the half-power beamwidth (HPBW) and the beam peak θ0 [5]: directivity(dB) = 10 log

101 (HPBW − 0.0027HPBW2 ) sin θ0

(1-25)

DIRECTIVITY ESTIMATES

15

Scan Factor, dB

Beam Direction, q

SCALE 1-4 pattern.

Additional directivity of omnidirectional pattern when scanned into conical

Scale 1-3 evaluates this formula for a beam at θ0 = 90◦ given HPBW, and Scale 1-4 gives the additional gain when the beam peak scans toward the axis. The directivity of butterﬂy patterns with unequal beamwidths in the principal planes cannot be estimated directly from the foregoing formulas. Similarly, some pencil beam patterns have large sidelobes which decrease the directivity and cannot be estimated accurately from Eq. (1-19). Both problems are solved by considering the directivity as an estimate of the average radiation intensity. Example A butterﬂy pattern peak is at 50◦ in both principal planes, but the beamwidths are 20◦ and 50◦ . Estimate the directivity. The 3-dB pattern points are given by: Cut 1 (40◦ and 60◦ ): U01 =

cos 40◦ − cos 60◦ = 0.133 2

U02 =

cos 25◦ − cos 75◦ = 0.324 2

Cut 2 (25◦ and 75◦ ):

Average the two pattern integral estimates: 0.133 + 0.324 = 0.228 2 1 Umax = = 4.38 (6.4 dB) directivity = U0 0.228 U0 =

Suppose that the beams are at different levels on the same pattern. For example, the lobe on the right of the ﬁrst pattern is the peak and the left lobe is reduced by 3 dB. The peaks of the second pattern are reduced by 1 dB. We can average on one pattern alone. Each lobe contributes Umax (cos θ1 − cos θ2 )/4 to the integral. The integral of the ﬁrst pattern is approximated by 0.266 + 0.266 × 10−3/10 = 0.100 4 The integral of the second pattern is reduced 1 dB from the peak. The average radiation intensity is found by averaging the two pattern averages: 0.100 + 0.324 × 10−0.1 = 0.178 2 1 = 5.602 (7.5 dB) directivity = U0 U0 =

16

PROPERTIES OF ANTENNAS

Pencil beam patterns with large sidelobes can be averaged in a similar manner: Up = 1/directivity. By using Eq. (1-19) and assuming equal beamwidths, we have Up = HPBW2 /41, 253, where Up is the portion of the integral due to the pencil beam and HPBW is the beamwidth in degrees. Example Consider a pencil beam antenna with pattern beamwidths of 50◦ and 70◦ in the principal planes. The second pattern has a sidelobe at θ = 60◦ down 5 dB from the peak and a 30◦ beamwidth below the 5 dB. What is the effect of the sidelobe on the directivity estimate? Without the sidelobe the directivity estimate is directivity =

41253 = 11.79 50(70)

(10.7 dB)

Consider each pattern separately: UP1 =

502 = 0.0606 41,253

UP2 =

702 = 0.1188 41,253

The sidelobe adds to the second integral: UPS2 =

(cos 45◦ − cos 75◦ )10−5/10 = 0.0354 4

Averaging the integrals of the parts gives us 0.1074: directivity =

1 = 9.31 (9.7 dB) U0

If there had been a sidelobe on each side, each would have added to the integral. Estimating integrals in this manner has limited value. Remember that these are only approximations. More accurate results can be obtained by digitizing the pattern and performing numerical integration on each pattern by using Eq. (1-16) or (1-17). 1-9 BEAM EFFICIENCY Radiometer system designs [6, p. 31–6] specify the antenna in terms of beam efﬁciency. For a pencil beam antenna with the boresight at θ = 0, the beam efﬁciency is the ratio (or percent) of the pattern power within a speciﬁed cone centered on the boresight to the total radiated power. In terms of the radiation intensity U ,

θ1

beam efﬁciency =

0

0

π

2π

U (θ, φ) sin θ dθ dφ (1-26)

0 2π

U (θ, φ) sin θ dθ dφ

0

where U includes both polarizations if necessary. Extended noise sources, such as radiometry targets, radiate noise into sidelobes of the antenna. Beam efﬁciency measures the probability of the detected target being located within the main beam (θ ≤ θ1 ).

INPUT-IMPEDANCE MISMATCH LOSS

17

Sometimes we can calculate directivity more easily than the pattern everywhere required by the denominator of Eq. (1-26): for example, a paraboloidal reﬂector. We use Eqs. (1-15) and (1-16) to calculate the denominator integral: π 2π 4πUmax U (θ, φ) sin θ dθ dφ = directivity 0 0 This reduces Eq. (1-26) to

θ1

2π

directivity beam efﬁciency =

0

U (θ, φ) sin θ dθ dφ

0

4πUmax

(1-27)

Equation (1-27) greatly reduces the pattern calculation requirements to compute beam efﬁciency when the directivity can be found without pattern evaluation over the entire radiation sphere. 1-10 INPUT-IMPEDANCE MISMATCH LOSS When we fail to match the impedance of an antenna to its input transmission line leading from the transmitter or to the receiver, the system degrades due to reﬂected power. The input impedance is measured with respect to some transmission line or source characteristic impedance. When the two are not the same, a voltage wave is reﬂected, ρV , where ρ is the voltage reﬂection coefﬁcient: ρ=

ZA − Z0 ZA + Z0

(1-28)

ZA is the antenna impedance and Z0 is the measurement characteristic impedance. On a transmission line the two traveling waves, incident and reﬂected, produce a standing wave: Vmax = (1 + |ρ|)Vi Vmin = (1 − |ρ|)Vi 1 + |ρ| Vmax = VSWR = Vmin 1 − |ρ|

(1-29) (1-30)

VSWR is the voltage standing-wave ratio. We use the magnitude of ρ, a complex phasor, since all the terms in Eq. (1-28) are complex numbers. The reﬂected power is given by Vi2 |ρ|2 /Z0 . The incident power is Vi2 /Z0 . The ratio of the reﬂected power to the incident power is |ρ|2 . It is the returned power ratio. Scale 1-5 gives the conversion between return loss and VSWR: return loss = −20 log |ρ| VSWR

Return Loss, dB

SCALE 1-5

Relationship between return loss and VSWR.

(1-31)

18

PROPERTIES OF ANTENNAS Reflected Power Loss, dB

Return Loss, dB

SCALE 1-6 Reﬂected power loss due to antenna impedance mismatch.

The power delivered to the antenna is the difference between the incident and the reﬂected power. Normalized, it is expressed as 1 − |ρ|2 or reﬂected power loss(dB) = 10 log(1 − |ρ|2 )

(1-32)

The source impedance to achieve maximum power transfer is the complex conjugate of the antenna impedance [7, p. 94]. Scale 1-6 computes the power loss due to antenna impedance mismatch. If we open-circuit the antenna terminals, the reﬂected voltage equals the incident voltage. The standing wave doubles the voltage along the transmission line compared to the voltage present when the antenna is loaded with a matched load. We consider the effective height of an antenna, the ratio of the open-circuit voltage to the input ﬁeld strength. The open-circuit voltage is twice that which appears across a matched load for a given received power. We can either think of this as a transmission line with a mismatch that doubled the incident voltage or as a Th´evenin equivalent circuit with an open-circuit voltage source that splits equally between the internal resistor and the load when it is matched to the internal resistor. Path loss analysis predicts the power delivered to a matched load. The mathematical Th´evenin equivalent circuit containing the internal resistor does not say that half the power received by the antenna is either absorbed or reradiated; it only predicts the circuit characteristics of the antenna load under all conditions. Possible impedance mismatch of the antenna requires that we derate the feed cables. The analysis above shows that the maximum voltage that occurs on the cable is twice that present when the cable impedance is matched to the antenna. We compute the maximum voltage given the VSWR using Eq. (1-29) for the maximum voltage: Vmax =

1-11

2Vi 2 VSWR(Vi ) = VSWR + 1 1 + 1/VSWR

(1-33)

POLARIZATION

The polarization of a wave is the direction of the electric ﬁeld. We handle all polarization problems by using vector operations on a two-dimensional space using the far-ﬁeld radial vector as the normal to the plane. This method is systematic and reduces chance of error. The spherical wave in the far ﬁeld has only θ and φ components of the electric ˆ Eθ and Eφ are phasor components in the direction of the unit ﬁeld: E = Eθ θˆ + Eφ φ. ˆ We can also express the direction of the electric ﬁeld in terms of vectors θˆ and φ. a plane wave propagating along the z-axis: E = Ex xˆ + Ey yˆ . The direction of propagation conﬁnes the electric ﬁeld to a plane. Polarization is concerned with methods

POLARIZATION

19

f (y)

wt

E

t q(x)

FIGURE 1-3

Polarization ellipse.

of describing this two-dimensional space. Both of the above are linear polarization expansions. We can rewrite them as ˆ E = Eθ (θˆ + ρˆL φ) E = Ex (ˆx + ρˆL yˆ )

Eφ Eθ Ey ρˆL = Ex ρˆL =

(1-34)

where ρˆL is the linear polarization ratio, a complex constant. If time is inserted into the expansions, and the tip of the electric ﬁeld traced in space over time, it appears as an ellipse with the electric ﬁeld rotating either clockwise (CW) or counter clockwise (CCW) (Figure 1-3). τ is the tilt of the polarization ellipse measured from the x-axis (φ = 0) and the angle of maximum response. The ratio of the maximum to minimum linearly polarized responses on the ellipse is the axial ratio. If ρˆL = e±j π/2 , the ellipse expands to a circle and gives the special case of circular polarization. The electric ﬁeld is constant in magnitude but rotates either CW (left hand) or CCW (right hand) at the rate ωt for propagation perpendicular to the page. 1-11.1 Circular Polarization Components The two circular polarizations also span the two-dimensional space of polarization. The right- and left-handed orthogonal unit vectors deﬁned in terms of linear components are ˆ = √1 (θˆ − j φ) ˆ R 2 1 ˆ Lˆ = √ (θˆ + j φ) 2

or or

ˆ = √1 (ˆx − j yˆ ) R 2 1 Lˆ = √ (ˆx + j yˆ ) 2

(1-35a) (1-35b)

The electric ﬁeld in the polarization plane can be expressed in terms of these new unit vectors: ˆ + ER R ˆ E = EL L

20

PROPERTIES OF ANTENNAS

When projecting a vector onto one of these unit vectors, it is necessary to use the complex conjugate in the scalar (dot) product: EL = E · Lˆ

∗

ˆ ER = E · R

∗

ˆ onto itself, we obtain When we project R ˆ ·R ˆ ∗ = 1 (θˆ − j φ) ˆ · (θˆ + j φ) ˆ = 1 (1 − j · j ) = 1 R 2 2 Similarly,

ˆ ∗ = 1 (θˆ + j φ) ˆ · (θˆ + j φ) ˆ = 1 (1 + j · j ) = 0 Lˆ · R 2 2

The right- and left-handed circular (RHC and LHC) components are orthonormal. A circular polarization ratio can be deﬁned from the equation ρˆc =

ˆ E = EL (Lˆ + ρˆc R)

ER = ρc ej δc EL

Let us look at a predominately left-handed circularly polarized wave when time and space combine to a phase of zero for EL . We draw the polarization as two circles (Figure 1-4). The circles rotate at the rate ωt in opposite directions (Figure 1-5), with the center of the right-handed circular polarization circle moving on the end of the vector of the left-handed circular polarization circle. We calculate the phase of the circular polarization ratio ρˆc from the complex ratio of the right- and left-handed circular components. Maximum and minimum electric ﬁelds occur when the circles

f (y)

ER

L +

E

R

E

EL

E

L −

E

R

ER

EL EL

q (x)

E

ER dc t

FIGURE 1-4 Polarization ellipse LHC and RHC components. (After J. S. Hollis, T. J. Lyons, and L. Clayton, Microwave Antenna Measurements, Scientiﬁc Atlanta, 1969, pp. 3–6. Adapted by permission.)

POLARIZATION

21

RHC f

LHC f Wave Propagating Out of Paper

E

wt q

E

wt

q

r^L = e −jp/2

r^L = e jp/2

FIGURE 1-5 Circular polarization components. (After J. S. Hollis, T. J. Lyons, and L. Clayton, Microwave Antenna Measurements, Scientiﬁc Atlanta, 1969, pp. 3–5. Adapted by permission.)

Axial Ratio, dB

Circular Cross-polarization, dB

SCALE 1-7 Circular cross-polarization/axial ratio.

alternately add and subtract as shown in Figure 1-4. Scale 1-7 shows the relationship between circular cross-polarization and axial ratio: √ √ Emax = (|EL | + |ER |) / 2 Emin = (|EL | − |ER |) / 2 1 + |ρˆc | |EL | + |ER | Emax = LHC = |EL | − |ER | 1 − |ρˆc | axial ratio = Emin (1-36) |ρˆc | + 1 |ER | + |EL | Emax = RHC = Emin |ER | − |EL | |ρˆc | − 1 |ρˆc | < 1 LHC 0 ≤ 1 < 1 RHC ρˆc axial ratio(dB) = 20 log

Emax Emin

The tilt angle of the polarization ellipse τ is one-half δc , the phase of ρˆc . Imagine time moving forward in Figure 1-5. When the LHC vector has rotated δc /2 CW, the RHC vector has rotated δc /2 CCW and the two align for a maximum. 1-11.2 Huygens Source Polarization When we project the currents induced on a paraboloidal reﬂector to an aperture plane, Huygens source radiation induces aligned currents that radiate zero cross-polarization

22

PROPERTIES OF ANTENNAS

in the principal planes. We separate feed antenna radiation into orthogonal Huygens sources for this case. To calculate the far-ﬁeld pattern of a paraboloid reﬂector, we can skip the step involving currents and integrate over the Huygens source ﬁelds in the aperture plane directly. We transform the measured ﬁelds of the feed into orthogonal Huygens sources by

Ec Eθf cos φ − sin φ = (1-37) sin φ cos φ Ex Eφf where Ec is the φ = 0 direction of polarization in the feed pattern and Ex is the φ = 90◦ polarization. This division corresponds to Ludwig’s third deﬁnition of crosspolarization [8]. The following matrix converts the Huygens source polarizations to the normal far-ﬁeld components of spherical coordinates:

Ec cos φ sin φ Eθ = (1-38) − sin φ cos φ Eφ Ex 1-11.3 Relations Between Bases In problems with antennas at arbitrary orientations, circularly polarized components have an advantage over linear components. When the coordinate system is rotated, both the amplitude and phase change for ρˆL , the linear polarization ratio, whereas the circular polarization ratio ρˆc magnitude is constant under rotations and only the phase changes. In other words, the ratio of the diameters of the circles (Figure 1-4) is constant. The circular components can be found from linear polarization components by projection. ˆ ∗ = √1 (Eθ θˆ + Eφ φ) ˆ ·R ˆ · (θˆ + j φ) ˆ ER = (Eθ θˆ + Eφ φ) 2 1 ER = √ (Eθ + j Eφ ) 2

(1-39)

Similarly, 1 EL = √ (Eθ − j Eφ ) 2 The linear polarizations can be found in terms of the circular components in the same manner: 1 j Eθ = √ (EL + ER ) Eφ = √ (EL − ER ) 2 2 These relations enable the conversion between polarizations. Good circularly polarized antennas over a wide bandwidth are difﬁcult to build, but good linearly polarized antennas are obtained easily. After we measure the phase and amplitude of Eθ and Eφ component phasors, we compute the circular components from Eq. (1-39), the axial ratio by using Eq. (1-36), and the polarization ellipse tilt τ from one-half the phase of ER /EL . We employ a leveled phase-locked source to record two patterns with orthogonal linear sources (or the same linear source is rotated between patterns). Afterward, we use the equations given above to convert polarization

23

POLARIZATION

to any desired polarization components. We calculate the maximum and minimum linear components by projecting the linear components into the rotated coordinate system of the polarization ellipse: Emax = Eθ cos τ + Eφ sin τ Emin = −Eθ sin τ + Eφ cos τ 1-11.4 Antenna Polarization Response The path loss formulas assume that the two antennas have matched polarizations. Polarization mismatch adds an extra loss. We determine polarization efﬁciency by applying the scalar (dot) product between normalized polarization vectors. An antenna transmitting in the z-direction has the linear components Ea = E1 (ˆx + ρˆL1 yˆ ) The incident wave on the antenna is given by Ei = E2 (ˆx + ρˆL2 yˆ ) where the wave is expressed in the coordinates of the source antenna. The z-axis of the source is in the direction opposite that of the antenna. It is necessary to rotate the coordinates of the receiving antenna wave. Rotating about the x-axis is equivalent to changing the sign of the tilt angle or taking the complex conjugate of Ea . The measurement antenna projects the incident wave polarization onto the antenna polarization. The antenna measures the incident ﬁeld, but we need to normalize the antenna polarization to a unit vector to calculate polarization efﬁciency: E2 · E∗1 =

∗ E2 E1∗ (1 + ρˆL2 ρˆL1 ) 2 1 + |ρˆL1 |

We normalize both the incident wave and antenna responses to determine loss due to polarization mismatch: xˆ + ρˆL2 yˆ Ei = ∗ |Ei | 1 + ρˆL2 ρˆL2

∗ xˆ + ρˆL1 E∗a yˆ = ∗ |Ea | 1 + ρˆL1 ρˆL1

The normalized voltage response is ∗ ρˆL2 Ei · E∗a 1 + ρˆL1 = ∗ ∗ |Ei ||Ea | 1 + ρˆL1 ρˆL1 1 + ρˆL2 ρˆL2

(1-40)

When we express it as a power response, we obtain the polarization efﬁciency : =

1 + |ρˆL1 |2 |ρˆL2 |2 + 2|ρˆL1 ||ρˆL2 | cos(δ1 − δ2 ) |Ei · E∗a |2 = |Ei |2 |Ea |2 (1 + |ρˆL1 |2 )(1 + |ρˆL2 |2 )

(1-41)

This is the loss due to polarization mismatch. Given that δ1 and δ2 are the phases of the polarization ratios of the antenna and the incident wave. As expressed in terms of linear polarization ratios, the formula is awkward because when the antenna is rotated to determine the peak response, both the amplitudes and phases change. A formula

24

PROPERTIES OF ANTENNAS

using circular polarization ratios would be more useful, because only phase changes under rotation. Two arbitrary polarizations are orthogonal ( = 0) only if |ρˆ1 | =

1 |ρˆ2 |

◦

and δ1 − δ2 = ±180

(1-42)

This can be expressed as vectors by using unit vectors: a1 · a∗2 = 0; a1 and a2 are the orthonormal generalized basis vectors for polarization. We can deﬁne polarization in terms of this basis with a polarization ratio ρ. By paralleling the analysis above for linear polarizations, we obtain the polarization efﬁciency for an arbitrary orthonormal polarization basis: =

1 + |ρˆ1 |2 |ρˆ2 |2 + 2|ρˆ1 ||ρˆ2 | cos(δ1 − δ2 ) (1 + |ρˆ1 |2 )(1 + |ρˆ2 |2 )

(1-43)

It has the same form as Eq. (1-41) derived for linear polarizations. We can use Eq. (1-43) with circular polarizations whose polarization ratio ρc magnitudes are constant with rotations of the antenna. The maximum and minimum polarization efﬁciencies occur when δ1 − δ2 equals 0◦ and 180◦ , respectively. The polarization efﬁciency becomes (1 ± |ρˆ1 ||ρˆ2 |)2 (1-44) max / min = (1 + |ρˆ1 |2 )(1 + |ρˆ2 |2 ) In all other vector pair bases for polarization, the magnitude of the polarization ratio ρ changes under rotations. Figure 1-6 expresses Eq. (1-44) as a nomograph. If we have ﬁxed installations, we can rotate one antenna until the maximum response is obtained and realize minimum polarization loss. In transmission between mobile antennas such as those mounted on missiles or satellites, the orientation cannot be controlled and the maximum polarization loss must be used in the link analysis. Circularly polarized antennas are used in these cases. Example A satellite telemetry antenna is RHC with an axial ratio of 7 dB. The ground station is RHC with a 1.5-dB axial ratio. Determine the polarization loss. Because the orientation of the satellite is unknown, we must use the maximum polarization loss. To ﬁnd it, use the RHC ends of the scales in Figure 1-6. Draw a line from 7 on the leftmost scale to 1.5 on the center scale. Read the loss on the scale between: 0.9 dB. The measured cross-polarization response of a linearly polarized antenna is the reciprocal of the axial ratio, the same absolute magnitude in decibels. Example Suppose that the linear cross-polarization responses of two antennas in a stationary link are given as 10 and 20 dB. Compute the minimum polarization loss. We rotate one of the antennas until the maximum response is found. The speciﬁcation of cross-polarization response does not state whether an antenna is predominately leftor right-handed circularly polarized. It must be one or the other. Suppose that the 20dB cross-polarization antenna is LHC. If the other antenna also is LHC, we use a line drawn from the lower portion of the center scale in Figure 1-6 to the rightmost LHC scale and read 0.2 dB of loss on the scale between the two. The second possibility is

POLARIZATION

25

FIGURE 1-6 Maximum and minimum polarization loss. (After A. C. Ludwig, A simple graph for determining polarization loss, Microwave Journal, vol. 19, no. 9, September 1976, p. 63.)

that the antenna could be predominately RHC. On drawing a line to the RHC (lower) scale, we read 0.7 dB on the center scale. When polarization is expressed in terms of linearly polarized components, it is ambiguous to give only magnitudes and no information of the circular polarization sense. 1-11.5 Phase Response of Rotating Antennas The polarization sense of an antenna can be determined from the phase slope of a rotating antenna. Before starting the phase measurement, determine that the setup is proper. Some older phase–amplitude receivers are ambiguous, depending on whether the local oscillator frequency was above or below the signal frequency. We use the convention that increased distance between antennas gives decreased phase. Move the antenna away from the source and observe decreasing phase or correct the setup. A rotating linearly polarized source ﬁeld is given by Es = E2 (cos α xˆ + sin α yˆ ) where α is clockwise rotation viewed from the direction of propagation (forward). A horizontally polarized linear antenna has the response Ea = E1 xˆ . It responds to the

26

PROPERTIES OF ANTENNAS

rotating linear source ﬁeld, E1 E2 cos α. The phase is constant under rotation until the null is passed and it ﬂips 180◦ through the null. An RHC polarized antenna has the response E1 (ˆx − j yˆ ). It responds to the rotating linear source ﬁeld, E1 E2 (cos α − j sin α) = E1 E2 e−j α The magnitude remains constant, but the phase decreases with rotation. Phase increases when the antenna is LHC. By observing the phase slope, the sense of the predominant polarization can be determined: RHC = negative phase slope; LHC = positive phase slope. It is easily remembered by considering the basis vectors of circular polarization: ˆ = √1 (ˆx − j yˆ ) R 2 In rotation from the x-axis to the y-axis, the phase decreases 90◦ . 1-11.6 Partial Gain If we measure the antenna gain to one polarization (e.g., RHC) and operate it in a link with an antenna also measured to one polarization, Eq. (1-44) fails to predict the response. Polarization efﬁciency assumes that the antenna gain was measured using a source ﬁeld with matched polarization. Gains referred to a single polarization are partial gains. If we align the two polarization ellipses of the two antennas, the response increases. Similarly, when the ellipses are crossed, the link suffers polarization loss. To obtain the full gain, we add the factor 10 log(1 + |ρ|2 )

(1-45)

to the partial gain, an expression valid using ρ for either circular or linear polarization. In terms of axial ratio A for circular polarization, the conversion is 2(1 + A2 ) 20 log 1+A When using measured partial gains for both antennas, the range of polarization efﬁciency is given by polarization efﬁciency = 20 log(1 ± ρ1 ρ2 )

(1-46)

We can convert Eq. (1-46) to expressions that use the axial ratio of the two antennas: 2(A1 A2 + 1) (A1 + 1)(A2 + 1) 2(A1 + A2 ) minimum polarization efﬁciency = 20 log (A1 + 1)(A2 + 1)

maximum polarization efﬁciency = 20 log

1-11.7 Measurement of Circular Polarization Using Amplitude Only The analyses given above assume that you can measure both amplitude and phase response of antennas, whereas in some cases only amplitude can be measured. If

VECTOR EFFECTIVE HEIGHT

27

you do not know the sense of circular polarization, it will be necessary to build two antennas that are identical except for their circular polarization sense. For example, you can build two identically sized counter-wound helical wire antennas. You determine polarization sense by using both sources and comparing measured levels. Once you establish the polarization sense, mount a linearly polarized measurement antenna with low cross-polarization. For a given pointing direction of the antenna under test, rotate the source antenna and record the maximum and minimum levels. The ratio of the maximum to the minimum is the axial ratio. To measure gain, rotate the measurement linearly polarized antenna to determine the peak response. Replace the antenna under test with a linearly polarized gain standard (horn) and perform a gain comparison measurement. Given the antenna axial ratio A, you adjust the linearly polarized gain by the correction factor: A+1 gain correction factor(dB) = 20 log √ 2A

(1-47)

We obtain the RHC and LHC response from 1 ER = √ (Emax + Emin ) 2

1 and EL = √ (Emax − Emin ) 2

assuming that the antenna is predominately RHC. 1-12 VECTOR EFFECTIVE HEIGHT The vector effective height relates the open-circuit voltage response of an antenna to the incident electric ﬁeld. Although we normally think of applying effective height to a line antenna, such as a transmitting tower, the concept can be applied to any antenna. For a transmitting tower, effective height is the physical height multiplied by the ratio of the average current to the peak current: VOC = Ei · h∗

(1-48)

The vector includes the polarization properties of the antenna. Remember from our discussion of antenna impedance mismatch that the open-circuit voltage√VOC is twice that across a matched load ZL for a given received power: VOC = 2 Prec ZL . The received power is the product of the incident power density S and the effective area of the antenna, Aeff . Gathering terms, we determine the open-circuit voltage from the incident ﬁeld strength E and a polarization efﬁciency : ZL Aeff VOC = 2E η We calculate polarization efﬁciency by using the scalar product between the normalized incident electric ﬁeld and the normalized vector effective height: =

|Ei ·h∗ |2 |Ei |2 |h|2

(1-49)

28

PROPERTIES OF ANTENNAS

Equation (1-49) is equivalent to Eq. (1-41) because both involve the scalar product between the incident wave and the receiving polarization, but the expressions have different normalizations. You can substitute vector effective height of the transmitting antenna for the incident wave in Eq. (1-49) and calculate polarization efﬁciency between two antennas. When an antenna rotates, we rotate h. We could describe polarization calculations in terms of vector effective height, which would parallel and repeat the discussion given in Section 1-11. We relate the magnitude of the effective height h to the effective area Aeff and the load impedance ZL : ZL Aeff (1-50) h=2 η The mutual impedance in the far ﬁeld between two antennas can be found from the vector effective heights of both antennas [9, p. 6–9]. Given the input current I1 to the ﬁrst antenna, we ﬁnd the open-circuit voltage of the second antenna: Z12 =

(V2 )OC j kηe−j kr h1 ·h∗2 = I1 4πr

(1-51)

When we substitute Eq. (1-50) into Eq. (1-51) and gather terms, we obtain a general expression for the normalized mutual impedance of an arbitrary pair of antennas given the gain of each in the direction of the other antenna as a function of spacing r: √ Z12 j G1 G2 −j kr h1 ·h∗2 = e (1-52) √ kr |h1 ||h2 | ZL1 ZL2 The magnitude of mutual impedance increases when the gain increases or the distance decreases. Of course, Eq. (1-52) is based on a far-ﬁeld equation and gives only an approximate answer, but it produces good results for dipoles spaced as close as 1λ. Figure 1-7 gives a plot of Eq. (1-52) for isotropic gain antennas with matched polarizations which shows the 1/R amplitude decrease with distance and that resistance and

Normalized Mutual Impedance

Reactance

Resistance

Spacing, l

FIGURE 1-7 Normalized mutual impedance (admittance) from the vector effective length for two antennas with 0 dB gain along the line between them.

MUTUAL COUPLING BETWEEN ANTENNAS

29

reactance curves are shifted out of phase. The cosine and sine factors of the complex exponential produce this effect. We multiply these curves by the product of the antenna gains, but the increased gain from larger antennas means that it is a greater distance to the far ﬁeld. When we bring two antennas close together, the currents on each antenna radiate and excite additional currents on the other that modify the result given by Eq. (1-52). But as we increase the distance, these induced current effects fade. Equivalent height analysis can be repeated using magnetic currents (e.g., used with microstrip patches), and Eqs. (1-51) and (1-52) become mutual admittance. Figure 1-7 is also valid for these antennas when we substitute normalized mutual admittance for normalized mutual impedance. For antennas with pattern nulls directed toward each other, the mutual impedance decreases at the rate 1/R 2 , due to the polarization of current direction h. 1-13 ANTENNA FACTOR The EMC community uses an antenna connected to a receiver such as a spectrum analyzer, a network analyzer, or an RF voltmeter to measure ﬁeld strength E. Most of the time these devices have a load resistor ZL that matches the antenna impedance. The incident ﬁeld strength Ei equals antenna factor AF times the received voltage Vrec . We relate this to the antenna effective height: AF =

Ei 2 = Vrec h

(1-53)

AF has units meter−1 but is often given as dB(m−1 ). Sometimes, antenna factor is referred to the open-circuit voltage and it would be one-half the value given by Eq. (1-53). We assume that the antenna is aligned with the electric ﬁeld; in other words, the antenna polarization is the electric ﬁeld component measured: 1 η 4π AF = = ZL Aeff λ ZL G This measurement may be corrupted by a poor impedance match to the receiver and any cable loss between the antenna and receiver that reduces the voltage and reduces the calculated ﬁeld strength. 1-14 MUTUAL COUPLING BETWEEN ANTENNAS The simplest approach for coupling between antennas is to start with a far-ﬁeld approximation. We can modify Eq. (1-8) for path loss and add the phase term for the ﬁnite distance to determine the S-parameter coupling: S21 =

e−j kr E1 ·E∗2 G1 G2 2kr |E1 ||E2 |

(1-54)

Equation (1-54) includes the polarization efﬁciency when the transmitted polarization does not match the receiving antenna polarization. We have an additional phase term

30

PROPERTIES OF ANTENNAS

because the signal travels from the radiation phase center along equivalent transmission lines to the terminals of each antenna. Equations (1-52) and (1-54) have the same accuracy except that Eq. (1-54) eliminates the need to solve the two-port circuit matrix equation for transmission loss. These formulas assume that antenna size is insigniﬁcant compared to the distance between the antennas, and each produces approximately uniform amplitude and phase ﬁelds over the second element. We can improve on Eq. (1-54) when we use the current distribution on one of the two antennas and calculate the near-ﬁeld ﬁelds radiated by the second antenna at the location of these currents. Since currents vary across the receiving antenna, we use vector current densities to include direction: Jr electric and Mr magnetic. Although magnetic current densities are ﬁctitious, they simplify the representation of some antennas. We compute coupling from reactance, an integral across these currents [see Eq. (2-34)]: j S21 = √ (Et ·Jr − Ht · Mr ) dV (1-55) 2 Pr Pt The input power to the transmitting antenna Pt produces ﬁelds Et and Ht . The power Pr into the receiving antenna excites the currents. The scalar product between the incident ﬁelds and the currents includes polarization efﬁciency. If we know the currents on the transmitting antennas, we calculate the near-ﬁeld pattern response from them at the location of the receiving antenna. Similar to many integrals, Eq. (1-55) is notional because we perform the integral operations only where currents exist. The currents could be on wire segments or surfaces. A practical implementation of Eq. (1-55) divides the currents into patches or line segments and performs the scalar products between the currents and ﬁelds on each patch and sums the result. A second form of the reactance [see Eq. (2-35)] involves an integral over a surface surrounding the receiving antenna. In this case each antenna radiates its ﬁeld to this surface, which requires near-ﬁeld pattern calculations for both. Equation (1-55) requires adding the phase length between the input ports and the currents, similar to using Eq. (1-54). When we use Eq. (1-55), we assume that radiation between the two antennas excites insigniﬁcant additional currents on each other. We improve the answer by using a few iterations of physical optics, which ﬁnds induced currents from incident ﬁelds (Chapter 2). We improve on Eq. (1-55) by performing a moment method calculation between the two antennas. This involves subdividing each antenna into small elements excited with simple assumed current densities. Notice the similarity between Eqs. (1-52) and (1-54) and realize that Eq. (1-55) is a near-ﬁeld version of Eq. (1-54). We use reactance to compute the mutual impedance Z21 between the small elements as well as their selfimpedance. For the moment method we calculate a mutual impedance matrix with a row and column for each small current element. We formulate a matrix equation using the mutual impedance matrix and an excitation vector to reduce coupling to a circuit problem. This method includes the additional currents excited on each antenna due to the radiation of the other.

1.15 ANTENNA NOISE TEMPERATURE [10] To a communication or radar system, an antenna contributes noise from two sources. The antenna receives noise power because it looks out on the sky and ground. The

ANTENNA NOISE TEMPERATURE

31

FIGURE 1-8 Antenna sky temperature. Noise temperature of an idealized antenna (lossless, no Earth-directed sidelobes) located at the Earth’s surface, as a function of frequency, for a number of beam elevation angles. Solid curves are for geometric-mean galactic temperature, sun noise 10 times quiet level, sun in unity-gain sidelobe, cool temperate-zone troposphere, 2.7 K cosmic blackbody radiation, zero ground noise. The upper dashed curve is for maximum galactic noise (center of galaxy, narrow-beam antenna). Sun noise 100 times quiet level, zero elevation, other factors the same as solid curves. The lower dashed curve is for minimum galactic noise, zero sun noise, 90◦ elevation angle. (The bump in the curves at about 500 MHz is due to the sun-noise characteristic. The curves for low elevation angles lie below those for high angles at frequencies below 400 MHz because of reduction of galactic noise by atmospheric absorption. The maxima at 22.2 and 60 GHz are due to the water-vapor and oxygen absorption resonance.) (From L. V. Blake, A guide to basic pulse-radar maximum-range calculation, Naval Research Laboratory Report 5868, December 1962.)

ground generates noise because it is about 290 K and a portion of the antenna pattern falls on it. Similarly, the sky adds noise dependent on the elevation angle and the operating frequency. Figure 1-8 gives the sky temperature versus frequency and elevation angle. The frequency range of lowest noise occurs in the middle of microwave frequencies of 1 to 12 GHz. The graphs show a large variation between the dashed curves, which occurs because of antenna direction and the pointing relative to the galactic center. In the middle of microwaves the sky noise temperatures are around 50 K, whereas near zenith the temperature is under 10 K. Near the horizon it rises because of the noise from oxygen and water vapor. The exact value must be determined for each application. As frequency decreases below 400 MHz, the sky temperature rises rapidly and becomes independent of antenna pointing. The curve continues the rapid rise at the same slope for lower frequencies. Low-frequency sky temperatures are often given as decibels relative to 290 K.

32

PROPERTIES OF ANTENNAS

An antenna receives this blackbody noise from the environment, but the value that affects the communication system depends both on the pattern shape and the direction of the main beam. We determine the antenna noise temperature by integrating the pattern times the environmental noise temperature distribution: Ta =

1 4π

2π 0

π

G(θ, φ)Ts (θ, φ) sin θ dθ dφ

(1-56)

0

where G(θ, φ) is the antenna gain pattern and Ts (θ, φ) is the angle-dependent blackbody radiation of the environment. Changing the antenna pointing changes Ta . Equation (1-56) is a weighted average of the environment noise temperature, usually referred to as the sky temperature. The second source of noise in the antenna is that of components that have both dissipative losses and reﬂection losses that generate noise. A receiving system needs to maximize the signal-to-noise ratio for given resources. System considerations, such as bit error rate, establish the required S/N ratio. We determine the noise power from the product N = k0 B n T e

(1-57)

where k0 is Boltzmann’s constant (1.38 × 10−23 W/K · Hz = −228.6 dB) and Bn is the receiver bandwidth (Hz). Te is the effective noise temperature (K). When referring noise temperature to other parts of the network, we increase or decrease it by the gain or loss, since it represents power and not a true temperature. Antenna gain is a measure of the signal level, since we can increase gain independent of the noise temperature, although the gain pattern is a factor by Eq. (1-56). The antenna conductor losses have an equivalent noise temperature: Te = (L − 1)Tp

(1-58)

where Tp is the antenna physical temperature and L is the loss (a ratio > 1). From a systems point of view, we include the transmission line run to the ﬁrst ampliﬁer or mixer of the receiver. We do not include the current distribution losses (aperture efﬁciencies) that reduce gain in Eq. (1-58) because they are a loss of potential antenna gain and not noise-generating losses (random electrons). The antenna–receiver chain includes mismatch losses, but these do not generate random electrons, only reﬂected waves, and have a noise temperature of zero. We include them in a cascaded devices noise analysis as an element with loss only. Noise characteristics of some receiver components are speciﬁed as the noise ﬁgure FN (ratio), and cascaded devices’ noise analysis can be analyzed using the noise ﬁgure, but we will use noise temperature. Convert the noise ﬁgures to noise temperature using TE = (FN − 1)T0

(1-59)

T0 is the standard reference temperature 290 K. We calculate noise temperature for the entire receiver chain of devices at a particular point normally at the input to the ﬁrst device. To calculate the S/N ratio we use the transmitter power, path loss (including antenna gain and polarization efﬁciency), and the gains (losses) of any devices for signal to the location in the receiver chain where

ANTENNA NOISE TEMPERATURE

33

noise temperature is being calculated. We characterize a given antenna by the ratio G/T , a measure independent of transmitter power and path loss, but including the receiver noise characteristics. Using the input of the ﬁrst device as the noise reference point, we calculate the input noise temperature from component noise temperatures and gains: T2 T3 T4 T = T1 + + + + ··· (1-60) G1 G1 G2 G1 G2 G3 Equation (1-60) merely states that noise temperatures are powers that decrease when we pass backward through a device with gain G. Each noise term is referred to the input of the device, and we pass backward to all previous devices and reduce noise temperature by 1/G. If we decided to locate the noise reference point at the input to the second device, the noise initially referred to the chain input would increase by the gain of the ﬁrst device. The system noise temperature becomes T(2) : T(2) = T1 G1 + T2 +

T3 T4 + + ··· G2 G2 G3

The signal also passes through the ﬁrst device and the new gain at the input to the second device becomes GG1 . The gain and the noise temperature change by the same factor G1 and produce a constant ratio. By extending these operations to any location in the receiver chain, we show that G/T is constant through the receiver device chain. It is easiest to illustrate G/T noise calculations with an example. A ground station has a 5-m-diameter paraboloid reﬂector with 60% aperture efﬁciency with the system operating at 2.2 GHz (λ = 0.136 m). We compute antenna directivity using the physical area and aperture efﬁciency:

π · Dia directivity = 0.60 λ

2

5π = 0.60 0.136

2 = 7972

(39 dB)

The reﬂector feed loss is 0.2 dB and it has a VSWR of 1.5 : 1. The cable between the feed and the ﬁrst ampliﬁer (LNA) has a 0.5-dB loss. These are elements under control of the antenna designer. We calculate the noise temperature of these by using Eq. (1-58) when we use a physical antenna temperature of 37.7◦ C (100◦ F) (310.8 K). Feed loss: T1 = (100.2/10 − 1)310.8 = 14.65 K Feed mismatch: T2 = 0 K Cable: T3 = (100.5/10 − 1)310.8 = 37.92 K The gains of these devices are G1 = 10−0.2/10 = 0.955 (feed loss), G2 = 10−0.18/10 = 0.959 (reﬂected power loss for 1.5 : 1 VSWR), and G3 = 10−0.5/10 = 0.891 (cable loss). The antenna sees the environment that generates noise due to blackbody radiation from the sky and ground. A typical value for the antenna pointed at 5◦ elevation is 50 K. This is not a physical temperature but represents an equivalent received power. Remember that the 60% aperture efﬁciency has no noise or loss contribution, because it only represents the loss of potential gain, since no random electrons are generated. We must consider the rest of the receiver chain when calculating the total input noise temperature. For this example we assume that the LNA has a noise ﬁgure of

34

PROPERTIES OF ANTENNAS

2 dB with 20 dB gain. The ﬁnal portion of the receiver includes the mixer and IF of the receiver, which we assume has a 10-dB noise ﬁgure. We use Eq. (1-59) to convert noise ﬁgure to noise temperature. LNA noise temperature T4 = 290(102/10 − 1) = 162.62 K Receiver noise temperature T5 = 290(1010/10 − 1) = 2610 K The 20-dB (100) LNA gain greatly reduces the effect of the 2610-K receiver. We calculate the contribution of each device to the input noise temperature by applying Eq. (1-60) to each device. We pass the noise temperature of the receiver through the four devices, and its temperature is reduced by the gain of each device: Te5 =

T5 2610 = 31.98 K = G1 G2 G3 G4 0.955(0.959)(0.891)(100)

The gain of the LNA greatly reduced the effective noise of the receiver at the antenna input. This operation shows that cascading noise temperature involves passing each device’s noise temperature through the gains of all preceding devices to the input and reducing it by the product of their gains. Similarly, we perform this operation on all the other noise temperatures. T4 169.62 = 207.86 K = G1 G2 G3 0.955(0.959)(0.891) T3 37.92 = 41.40 K Te3 = = G1 G2 0.955(0.959) T2 0 =0 = Te2 = G1 0.955 Te1 = T1 = 14.65 Te4 =

These operations illustrate that the cascaded devices’ noise temperature equation (1-60) is easily derived by considering the passage of noise temperature (power) through devices with gain to a common point where we can add the contributions. The sky temperature is not an input noise temperature but the noise power delivered at the ﬁctitious point called the antenna directivity, where gain = directivity. Since noise temperature represents power, we convert it to decibels and subtract it from directivity to compute G/T : G/T (dB) = 39 − 10 log(345.9) = 13.6 dB This G/T is a measure of the antenna and receiver combined performance when the antenna is pointed to 5◦ elevation. Changing the pointing direction affects only the sky temperature added directly to the ﬁnal result. We use G/T in the link budget of the communication system. We can supply a single value for the antenna gain and noise temperature at the output port connected to the receiver. Recognize that the ﬁrst three noise temperatures and the sky temperature are associated with the antenna. We moved the noise reference of each device to the input by dividing by the gain of the preceding devices. To move

COMMUNICATION LINK BUDGET AND RADAR RANGE

35

to the output of the antenna, we increase the noise temperature and the antenna gain by the product of the gain for the devices: T = (Tsky + Te1 + Te2 + Te3 )G1 G2 G3 = (50 + 14.65 + 0 + 37.92)10−0.88/10 = 83.7 K gain(dB) = directivity(dB) − 0.88 dB = 39 − 0.88 = 38.12 dB This reduces the antenna to a single component similar to the directivity and sky temperature that started our analysis. 1-16 COMMUNICATION LINK BUDGET AND RADAR RANGE We illustrate communication system design and path loss by considering a sample link budget example. The 5-m-diameter reﬂector is pointing at a satellite in an orbit 370 km above the Earth with a telemetry antenna radiating 10 W at 2.2 GHz. Since the antenna pattern has to cover the visible Earth, its performance is compromised. Considering the orbit geometry and antenna pointing is beyond the scope of this discussion. The range from a satellite at 370 km to a ground station pointing at 5◦ is 1720 km. The satellite antenna pointing angle from nadir is 70.3◦ , and a typical antenna for this application would have gain = −2 dBiC (RHC gain relative to an isotropic antenna) and an axial ratio of 6 dB. Assume that the ground station antenna has a 2-dB axial ratio. We apply the nomograph of Figure 1-6 to read the maximum polarization loss of 0.85 dB since we cannot control the orientation of the polarization ellipses. The link budget needs to show margin in the system, so we take worst-case numbers. When we apply Eq. (1-9) for path loss, we leave out the antenna gains and add them as separate terms in the link budget (Table 1-2): free-space path loss = 32.45 + 20 log[2200(1720)] = 164 dB The link budget shows a 4.4-dB margin, which says that the communication link will be closed. This link budget is only one possible accounting scheme of the system parameters. Everyone who writes out a link budget will separate the parameters differently. This budget shows typical elements. Radar systems have similar link budgets or detection budgets that consider S/N : S/N =

Prec PT GT (directivity)λ2 σ (EIRP)λ2 (G/T )σ = = KT B (4π)3 R 4 KT B (4π)3 R 4 KB

The radar has a required S/N value to enable it to process the information required, which leads to the maximum range equation:

(EIRP)λ2 (G/T )σ R= (4π)3 (S/N )req KB

1/4 (1-61)

Equation (1-61) clearly shows the role of the transmitter, EIRP; the receiver and antenna noise; G/T ; and the requirement for signal quality, S/Nreq , on the radar range for a given target size σ . Equation (1-61) applies to CW radar, whereas most radars use pulses. We increase radar performance by adding many pulses. We ignore the aspects of pulse train encoding

36

PROPERTIES OF ANTENNAS

TABLE 1-2 Link Budget Frequency Transmit power Transmit antenna gain EIRP (effective isotropic radiated power) Free-space path loss Polarization loss Atmospheric loss Rain loss Pointing loss Receive antenna directivity G/T Boltzmann’s constant Carrier/noise (C/N) ratio (ignores bandwidth) Bit rate: 8 Mb/s Eb /N0 (energy per bit/noise density) Implementation loss Eb /N0 required Margin

2.2 GHz 10 dBW −2 dB 8 dBW 164 dB 0.85 0.30 0.00 0.00 39 dB 13.6 dB 228.6 dB 85 dB 69 dB 16 dB 2 dB 9.6 dB 4.4 dB

Information only 10 log(10) Transmit power dBW + antenna gain dB Isotropic antenna path loss Maximum for uncontrolled orientation 5◦ elevation at 2.2 GHz Little loss at this frequency Location in receiver chain for G/T calculation From preceding section EIRP + G/T − path loss − polarization loss −atmospheric loss − rain loss + 228.6 10 log(bit rate) bandwidth Eb /N0 = C/N − 10 log(bit rate) Groups extra system losses For bit error rate (BER) = 10−5 in QPSK Eb /N0 − required Eb /N0 − implementation loss

that allow coherent addition. Radar range is determined by the total energy contained in the pulses summed. We replace EIRP with GT (energy) since PT × time = energy. It is the total energy that illuminates the target that determines the maximum detection range. Using antennas in radar leads to speaking of the radiated energy correct for pulsed systems, but when we do not integrate pulse shape times time, the antenna radiates power. To be correct we should call radiation that we integrate over angular space to ﬁnd power, “power density.” To say “energy radiated in the sidelobes” is poor physics unless it is a radar system, because it is power. 1-17

MULTIPATH

Multipath means that the ﬁeld intensity at a particular point is the sum of a number of waves that arrive from different directions or from different sources. It arises from signal transmission paths such as edge reﬂections from the mounting structure around an antenna and general reﬂections from objects near the antenna. Nearby reﬂections only seem to modify the antenna pattern, while reﬂections from additional objects cause rapid ripple with changing pattern angle. In Section 3-1 we discuss how to use the ripple angular rate and pattern distribution to locate its source. Multipath causes degraded system performance or measurement errors. Of course, multipath can improve performance as well. In fact, we add nearby objects, such as ground planes, to improve antenna performance. We specify pattern response in terms of the power response, but we add ﬁelds. An extra signal −20 dB relative to the main signal is 0.01 in power but 0.1 in ﬁeld strength

PROPAGATION OVER SOIL

37

Peak-to-Peak Ripple, dB

Interference Signal Level, dB

SCALE 1-8 Signal peak-to-peak amplitude ripple due to multipath signal. Maximum Phase Error (degrees)

Interference Signal Level, dB

SCALE 1-9 Peak phase error due to multipath signal.

(voltage). Since the extra signal can have any phase relative to the main signal, it can add or subtract. Given an extra signal MP(dB), the pattern ripple is ripple(dB) = 20 log

1 + 10MP/20 1 − 10MP/20

(1-62)

where MP(dB) has a negative sign. Scale 1-8 gives the relationship between peakto-peak amplitude ripple and the level of the multipath signal. Equation (1-62) is numerically the same as the relationship between return loss and 20 log(VSWR). The multipath signal can change the phase when summed with the main signal over a range given by maximum phase error = ± tan−1 (10MP/20 ) (1-63) Scale 1-9 calculates the peak phase error due to a multipath signal. 1-18 PROPAGATION OVER SOIL When we position antennas over soil and propagate the signal any signiﬁcant distance, it will reﬂect from soil or water and produce a large multipath signal. Soil is a conductive dielectric that reﬂects horizontally and vertically polarized signals differently. Typical ground constants are listed in Table 1-3. Given the grazing angle ψ measured between the reﬂected ray and ground, the voltage reﬂection coefﬁcients are sin ψ − εr − j x − cos2 ψ (εr − j x) sin ψ − εr − j x − cos2 ψ ρh = and ρv = sin ψ + εr − j x − cos2 ψ (εr − j x) sin ψ + εr − j x − cos2 ψ (1-64) where x = σ/ωε0 = 17, 975σ /frequency(MHz). Figure 1-9 gives the reﬂection coefﬁcient for the two polarizations versus grazing angle. Horizontal polarization reﬂects from soil about the same as a metal surface. Vertical polarization reﬂection produces a more interesting curve. The graph shows that the reﬂection is low over a region of grazing angles. The minimum reﬂection direction is called the Brewster angle. At this angle the reﬂected wave is absorbed

38

PROPERTIES OF ANTENNAS

TABLE 1-3

Typical Ground Constants

Surface Dry ground Average ground Wet ground Fresh water Seawater

Dielectric Constant

Conductivity (S)

4–7 15 25–30 81 81

0.001 0.005 0.020 0.010 5.0

Reflection, dB

Horizontal Polarization

Vertical Polarization

Grazing Angle

FIGURE 1-9

Average soil reﬂection for horizontal and vertical polarization.

into the soil. At high grazing angles ρh has a phase near 180◦ and ρv a phase of 0◦ . When the grazing angle decreases and becomes less than the Brewster angle, the vertical polarization reﬂection changes from 0◦ to 180◦ . Remember that for most general response nulls, the signal phase changes by 180◦ when passing through the transition. As the grazing angle approaches zero both reﬂection coefﬁcients approach −1 and multipath is independent of polarization. The electric ﬁeld at the receiving antenna is the sum of the direct wave plus the reﬂected wave, which traveled along a longer path: E = Ed [1 − exp(−j φ)] = Ed (1 − cos φ + j sin φ) We compute the magnitude |E| = |Ed | 1 + cos2 φ + sin2 φ − 2 cos φ = 2|Ed | sin(φ/2) for the small phase difference between the two equal-amplitude signals. The received power Prec is proportional to E 2 . The path loss for this multipath link is modiﬁed from the free-space equation: λ 2 hT hR 2 2 2πhT hR Prec = 4PT → PT GT GR GT GR sin (1-65) 4πd λd d2

MULTIPATH FADING

39

Equation (1-65) states that the power received is proportional to 1/d 4 and increases by h2 for either antenna. We can approximate the propagation over soil by a region for closely spaced antennas when the results consist of the free-space transmission with 1/d 2 average transmission with signiﬁcant variation due to multipath and a second region proportional to 1/d 4 with small multipath variations. The breakpoint between the two models occurs at a distance d = 4hT hR /λ. Experiments at mobile telephone frequencies showed that Eq. (1-65) overestimates the received power when the receiving antenna height is less than 30 m and a more correct model modiﬁes the exponent of hR [11, p. 38]: h2T hCR (1-66) d4 Below 10 m, C = 1 and the exponent varies linearly between 10 and 30 m: C = hR /20 + 12 . On a narrow-beam terrestrial propagation path, scattering from an object along a path an odd multiple of λ/2 produces a signal that reduces the main path signal. Given an obstacle at a distance h radial from the direct ray path and located dT from the transmitter and a distance dR from the receiving antenna, we determine the differential path length as λ nλdT dR h2 dT + dR =n (1-67) or clearance height h = = 2 dT dR 2 dT + dR Prec = PT GT GR

We call these Fresnel clearance zones of order n. The direct path should clear obstacles by at least one clearance zone distance h to prevent the scattered signal from having a negative impact on the communication link. The ﬁrst Fresnel zone touches ground when dT = 4hT hR /λ is the breakpoint distance between 1/d 2 and 1/d 4 propagation models. 1-19 MULTIPATH FADING Most mobile communication occurs when there is no direct path between the base station antennas and the mobile user. The signal reﬂects off many objects along the path between the two. This propagation follows a Rayleigh probability distribution about the mean signal level: r r2 R2 pr (r) = 2 exp − 2 prob[r < R] = PR (R) = 1 − exp − 2 α 2α 2α √ R is the signal level,√ α the value of the peak in the distribution, with mean = α π/2 and median RM = α 2 ln(2) = 1.1774α. The median signal level is found by ﬁtting measured data for various localities (town, small town, open country, etc.) into a prediction model. The signal will have large signal fades where the level drops rapidly. The Rayleigh model can be solved for the average distance between fades given the level. As a designer it is important to realize the magnitude of the problem [12, pp. 125–130]: 2(R/RM ) average distance between fades = λ √ 2π ln(2)(R/RM ) 2

(1-68)

40

PROPERTIES OF ANTENNAS Distance between Fades, l

Rayleigh Multipath Signal Fade, dB

SCALE 1-10

Average distance between fades and depth of fade for a Rayleigh multipath.

Length of Fade Region, l

Rayleigh Multipath Signal Fade, dB

SCALE 1-11

Average fade length and depth of fade for a Rayleigh multipath.

R is the fade level (ratio) and RM is the median signal level found from a propagation model. Scale 1-10 shows the relationship between the average distance between fades and the depth of fade for Rayleigh multipath. A mobile channel operating at 1.85 GHz (λ = 16.2 cm) has a 15-dB fade every 2.75λ which equals 44.5 cm, while 10-dB fades occur every 1.62λ = 26.25 cm. The communication system must overcome these fades. Fortunately, the deep fades occur over a short distance: 2(R/RM ) − 1 average length of fade = λ √ 2π ln(2)(R/RM ) 2

The signal fades and then recovers quickly for a moving user. Scale 1-11 shows the average fade length along a path given the depth of fade. For the 1.85-GHz channel the 15-dB fade occurs only over 0.06(16.2) = 0.97 cm, and the 10-dB fade length is 0.109(16.2) = 1.76 cm. The solution to mobile communication multipath fading is found either in increasing the link margin with higher gain base station antennas or the application of diversity techniques. We use multiple paths between the user and the base station so that while one path experiences a fade, the other one does not. Diversity has no effect on the median signal level, but it reduces the effects of the nulls due to the Rayleigh distribution propagation. REFERENCES 1. S. A. Schelkunoff and H. Friis, Antenna Theory and Practice, Wiley, New York, 1952. 2. R. F. Harrington, Time-Harmonic Electromagnetic Fields, McGraw-Hill, New York, 1961. 3. H. Friis, A note on a simple transmission formula, Proceedings of IRE, vol. 34, May 1946, pp. 254–256.

REFERENCES

41

4. J. D. Kraus, Antennas, McGraw-Hill, New York, 1950. 5. N. McDonald, Omnidirectional pattern directivity in the presence of minor lobes: revisited, IEEE Antennas and Propagation Magazine, vol. 41, no. 2, April 1999, pp. 63–65. 6. W. F. Croswell and M. C. Bailey, in R. C. Johnson and H. Jasik, eds., Antenna Engineering Handbook, McGraw-Hill, New York, 1984. 7. C. G. Montgomery, R. H. Dicke, and E. M. Purcell, Principles of Microwave Circuits, McGraw-Hill, New York, 1948. 8. A. C. Ludwig, The deﬁnition of cross polarization, IEEE Transactions on Antennas and Propagation, vol. AP-21, no. 1, January 1973, pp. 116–119. 9. K. P. Park and C. T. Tai, Receiving antennas, Chapter 6 in Y. T. Lo and S. W. Lee, eds., Antenna Handbook, Van Nostrand Reinhold, New York, 1993. 10. L. V. Blake, Prediction of radar range, Chapter 2 in Radar Handbook, M. Skolnik, ed., McGraw-Hill, New York, 1970. 11. K. Fujimoto and J. R. James, Mobile Antenna Systems Handbook, 2nd ed., Artech House, Boston, 2001. 12. J. D. Parsons, The Mobile Radio Propagation Channel, Wiley, New York, 1992.

2 RADIATION STRUCTURES AND NUMERICAL METHODS

Antenna analysis, an important part of design, requires a compromise between extensive calculations and the fabrication and measurement of prototypes, which depends on your working environment. You should minimize cost, which means reducing the time from the start of a design to completion of a working model. In some cases you should not rush to build a prototype. For example, when designing large and expensive antennas, such as paraboloidal reﬂectors, the high fabrication cost justiﬁes the time required for analysis. Management will not let you proceed before knowing the design will work. You should develop a cost model for each design in which analysis is one factor. Analyses allow optimization of a design. You can design a number of antennas and adjust the dimensions until you ﬁnd the best one. Again, you should be considering the costs of your time. At some point the incremental improvements are not worth the extra time for further analyses. In any case, when you build the prototype, you can expect differences. You soon determine that you can achieve only limited knowledge about a design because fabrication and measurement errors mask the true response of the antenna. You are doing engineering, not a science project. Textbooks contain many analyses of ideal antennas, and this book is no exception. You need to consider the application and the ﬁnal antenna environment. The mounting structure has little effect on the pattern of a large antenna with narrow beamwidth because little radiation strikes it. The overall radiation characteristics of narrow- or wide-beam antennas depend signiﬁcantly on the shape of the vehicle and how the antenna is mounted. In later chapters we discuss how to use antenna mounting to improve performance, so you can take advantage of it. The size of the mounting structure limits the type of analysis used. In this chapter we discuss physical optics (PO) and geometric optics (GO) [geometric theory of diffraction (GTD)] for large structures. In physical optics we compute the current induced on the vehicle due to antenna radiation and include their radiation in the Modern Antenna Design, Second Edition, By Thomas A. Milligan Copyright 2005 John Wiley & Sons, Inc.

42

AUXILIARY VECTOR POTENTIALS

43

overall pattern. But the PO analysis cost rises rapidly as the number of small current patches increases for larger structures. PO analysis works well with large antennas, such as paraboloidal reﬂectors, that produce focused beams. Geometric optics uses ray optics techniques whose computation cost is independent of the size of the vehicle and whose accuracy improves as structure size increases. GO provides insight because we can visualize the combination of direct, reﬂected, and diffracted (GTD) rays to calculate the pattern, but it requires the solution of difﬁcult geometry problems. Smaller structures allow the use of multiple methods. For example, the moment method divides the surroundings into small patches and uses an expansion of the current in predetermined basis functions. This method uses integral equations of the boundary conditions to calculate a matrix equation involving coefﬁcients of the current expansion. Numerical methods invert the matrix to solve for the coefﬁcients, but it is a costly numerical operation and limits the size of the problem that can be handled to a few wavelengths. The ﬁnite-difference time-domain (FDTD) technique computes the ﬁelds on the structure in the time domain. This method handles moderate-sized structures and readily includes complex material properties such as biological features. FDTD divides the region into cubic cells and when excited by pulse feeding functions, it produces wide frequency bandwidth responses. Finite-element methods (FEMs) also divide the problem into cubic cells, but the analysis is performed in the frequency domain. FEM analysis must be repeated at every frequency of interest. FDTD and FEMs require a program to divide the structure into a mesh before starting the solution. Both methods calculate currents on a boundary surface by using the equivalence theorem with the incident ﬁelds and then calculate the far-ﬁeld radiation pattern from these boundary currents. Most methods start by assuming a current distribution on the antenna or, equivalently, a distribution of ﬁelds on an aperture. The ﬁelds on the aperture can be reduced to a current distribution. The moment method uses a summation of assumed basis function currents and solves for the coefﬁcients of the expansion, but it, too, starts with assumed currents over small regions. You will discover that the radiation pattern can be found with greater accuracy than the input impedance. For antennas constructed from wires, the moment method computes the input current for a given excitation voltage and we calculate impedance from the ratio. Interaction of an antenna with the currents induced on a structure has little effect on impedance for narrow-beam antennas. Even for wide-beam antennas, such as dipoles, the structure effect on impedance can be found by using source mutual coupling with its images. In the end, antenna impedance should be measured when mounted in the ﬁnal conﬁguration. An antenna has both a radiation pattern bandwidth and an impedance bandwidth, but you must give the pattern primary consideration. Too many designs concentrate on the wideband impedance characteristics of an antenna when, in fact, the antenna pattern has changed over the frequency range of the impedance bandwidth. Your primary task should be to design for the radiation pattern desired. In Chapter 1 we detailed the system aspects of impedance mismatch (Section 1-10), and you may determine the overall system impact of small impedance mismatch. 2-1 AUXILIARY VECTOR POTENTIALS We do not use vector potentials in design. It seems as though they would be useful, but only a few simple antennas ﬁt their direct use. You cannot measure them

44

RADIATION STRUCTURES AND NUMERICAL METHODS

because they are not physical entities, so they seem artiﬁcial. Physical optics (PO) calculates the radiation directly from currents using dyadic Green’s functions but uses long expressions. Nevertheless, many analysis techniques ﬁnd them more efﬁcient than PO expressions and you should be aware of them. We illustrate their use with a couple of simple antennas. We use vector potentials to introduce a few antenna concepts. In the ﬁrst example we apply the magnetic vector potential to calculate the radiation from a short-length current element (dipole) and show how to obtain the pattern. Integration of the radiation pattern power density (Section 1-2) determines the total power radiated. Because we know the input current and the total radiated power, the ratio of the power to the input current squared gives the radiation resistance. We combine the low radiation resistance with the material resistance to compute the antenna efﬁciency. Electric vector potentials used with ﬁctitious magnetic currents illustrate analysis by duality. We apply this to the analysis of a small loop and show that it has the same pattern as that of a small dipole. 2-1.1 Radiation from Electric Currents Normal electron currents radiate when time varying. The simplest example is a ﬁlamentary current on wire, but we include surface and volumetric current densities as well. We analyze them by using the magnetic vector potential. Far-ﬁeld electric ﬁelds are proportional to the magnetic vector potential A: E = −j ωA

(2-1)

We determine the magnetic ﬁeld from |E| = η|H|

(2-2)

and realizing the cross product of the electric ﬁeld with the magnetic ﬁeld points in the direction of power ﬂow, the Poynting vector. Since the electric ﬁeld direction deﬁnes polarization, we usually ignore the magnetic ﬁeld. We derive the magnetic vector potential from a retarded volume integral over the current density J: A=µ

J(r )e−j k|r−r | dV 4π|r − r |

(2-3)

where r is the ﬁeld measurement point radius vector, r the source-point radius vector, µ the permeability (4π × 10−7 A/m in free space), and k, the wave number, is 2π/λ. As written, Eq. (2-3) calculates the potential A everywhere: near and far ﬁeld. The vector potential can be written in terms of a free-space Green’s function: e−j kR where R = |r − r | 4πR A=µ g(R)J(r ) dV

g(R) =

(2-4)

Radiation Approximation When we are interested only in the far-ﬁeld response of an antenna, we can simplify the integral [Eq. (2-3)]. An antenna must be large in terms

AUXILIARY VECTOR POTENTIALS

45

of wavelengths before it can radiate efﬁciently with gain, but at great distances it still appears as a point source. Consider the radiation from two different parts of an antenna. Far away from the antenna, the ratio of the two distances to the different parts will be nearly 1. The phase shift from each part will go through many cycles before reaching the observation point, and when adding the response from each part, we need only the difference in phase shift. In the radiation approximation we pick a reference point on the antenna and use the distance from that point to the far-ﬁeld observation point for amplitudes, 1/R, for all parts of the antenna. The direction of radiation deﬁnes a plane through the reference point. This plane is deﬁned by the radius normal vector, given in rectangular coordinates by rˆ = sin θ cos φ xˆ + sin θ sin φ yˆ + cos θ zˆ We compute the phase difference to the far-ﬁeld point by dropping a normal to the reference plane from each point on the antenna. This distance multiplied by k, the propagation constant, is the phase difference. Given a point on the antenna r , the phase difference is kr · rˆ . When we substitute these ideas into Eq. (2-3), the equation becomes e−j kr µ J ej kr ·ˆr dV A= (2-5) 4πr In rectangular coordinates kr · rˆ becomes k(x sin θ cos φ + y sin θ sin φ + z cos θ ) We can combine k and rˆ to form a k-space vector: k = kˆr = k sin θ cos φ xˆ + k sin θ sin φ yˆ + k cos θ zˆ and the phase constant becomes k · r . Currents in ﬁlaments (wires) simplify Eq. (2-5) to a single line integral. Magnetic vector potentials and electric ﬁelds are in the same directions as the wires that limit the directions of current. For example, ﬁlamentary current along the z-axis produces z-directed electric ﬁelds. Spherical waves (far ﬁeld) have only θˆ and φˆ components found from the projection of Ez onto those axes. Filamentary currents on the z-axis produce only z-directed electric ﬁelds with a null from θˆ · zˆ = − sin θ at θ = 0. In turn, x- or y-directed currents produce electric ﬁelds depending on the scalar products (projections) of the xˆ and yˆ unit vectors onto the θˆ and φˆ vectors in the far ﬁeld: θˆ · xˆ = cos θ cos φ θˆ · yˆ = cos θ sin φ

φˆ · xˆ = − sin φ φˆ · yˆ = cos φ

By examining antenna structure you can discover some of its characteristics without calculations. Without knowing the exact pattern, we estimate the polarization of the waves by examining the directions of the wires that limit the current density. Consider various axes or planes of symmetry on an antenna: for example, a center-fed wire along the z-axis. If we rotate it about the z-axis, the problem remains the same, which means that all conical polar patterns (constant θ ) must be circles; in other words, all great

46

RADIATION STRUCTURES AND NUMERICAL METHODS

circle patterns must be the same. An antenna with the same structure above and below the x –y plane radiates the same pattern above and below the x –y plane. Always look for axes and planes of symmetry to simplify the problem. We can extend the magnetic vector potential [Eq. (2-1)] to determine near ﬁelds: E = −j ωA + H=

∇(∇ · A) j ωεµ

1 ∇ ×A µ

(2-6)

The electric ﬁeld separates into far- and near-ﬁeld terms, but the equation for the magnetic ﬁeld, the deﬁning equation of the potential, does not separate. If we substitute the free-space Green’s function from Eq. (2-4) into Eq. (2-6), expand, and gather terms, we can determine the ﬁelds directly from the electric currents and eliminate the use of a vector potential. 1 j j ηk 2 − + 3 3 J(r ) − E(r) = 4π kR k 2 R 2 k R V

3 j 3j + 2 2− 3 3 e−j kR dV kR k R k R k2 1 j ˆ H(r) = + 2 J(r ) × R e−j kR dV 4π kR k R ˆ R ˆ +[J(r ) · R]

V

ˆ = r−r = r−r R |r − r | R

since

(2-7) (2-8)

R = |r − r |

Terms with 1/R dependence are the far-ﬁeld terms. The radiative near-ﬁeld terms have 1/R 2 dependence and near-ﬁeld terms have 1/R 3 dependence. The impedance of free space, η, is 376.7 . We can rearrange Eqs. (2-7) and (2-8) so that they become the integral of the dot product of the current density J with dyadic Green’s functions [1]. It is only a notation difference that leads to a logic expression. Except for a few examples given below, we leave the use of these expressions to numerical methods when designing antennas. Example Use the magnetic vector potential to derive the far ﬁeld of a short-length current element. Assume a constant current on the wire. The current density is I lδ(r ), where δ(r ) is the Dirac delta distribution and l is the length over which the far-ﬁeld phase is constant. The integral in Eq. (2-4) easily reduces to µI le−j kr 4πr The current element is so short that the phase distances from all parts of the element are considered to be equal; e−j kr is the retarded potential phase term. The electric ﬁeld is found from Az using Eq. (2-1): I l −j kr Ez = −j ωµ e 4πr I l −j kr Eθ = Ez zˆ · θˆ = j ωµ sin θ e 4πr Az =

AUXILIARY VECTOR POTENTIALS

47

√ √ √ µ µ, and divide and multiply by ε: √ j I l2πf µε µ −j kr sin θ Eθ = e 4πr ε

Evaluate ω as 2πf , split µ in

The following terms can be recognized as l f = c λ

1 c= √ µε

η=

µ ε

The far-ﬁeld electric ﬁeld becomes j I lη −j kr sin θ e 2λr The magnetic ﬁeld is found from the electric ﬁeld using Eq. (2-2): Eθ =

Hφ =

j I l −j kr Eθ = e sin θ η 2λr

The term j can be evaluated as ej π/2 , a phase shift term. The power density Sr is Sr = Eθ Hφ∗ =

|I |2 l 2 η sin2 θ 4λ2 r 2

The normalized power pattern is equal to sin2 θ . Figure 2-1 gives the polar pattern of this antenna as a dashed plot. The dashed circle is the −3-dB pattern level. We measure the angular separation between the 3-dB points to determine the beamwidth (half-power beamwidth). For comparison, Figure 2-1 shows the pattern of a half-wavelength-long dipole as a solid curve. At a length about 5% shorter than a half wavelength, the reactive component of the impedance vanishes. The ﬁgure illustrates that a short dipole has about the same pattern as a long-resonant-length (reactance equals zero) dipole. We determine directivity (Section 1-7) by calculating the average radiation intensity, often normalized to the peak of the power pattern: π/2 2 Uavg = sin2 θ sin θ dθ = 3 0 Umax = 1 Umax directivity = = 1.5 (1.76 dB) Uavg The resonant-length dipole (≈ λ/2) has a directivity of 2.15 dB or only 0.39 dB more than that of the very short dipole. The total power radiated by the antenna is found by integrating the Poynting vector magnitude over a sphere: 2π π Pr = Sr r 2 sin θ dθ dφ

0

0 π

|I |l = 2λ 0 0 2 2π |I |l η = 3 2λ 2π

2 η sin θ dθ dφ

48

RADIATION STRUCTURES AND NUMERICAL METHODS

l/2 Dipole

Short Dipole

FIGURE 2-1 Pattern of a short current element and small loop (dashed curve) compared to a λ/2 -long dipole (solid curve) located along the 0 to 180◦ axis.

We represent the radiated power as a radiation resistance at the input of the antenna: 2 l Pr 2π η = RR = |I |2 3 λ While a short dipole with a length λ/20 has a radiation resistance of about 2 , a resonant-length dipole has about a 50- radiation resistance and is more efﬁcient because the relative material resistance is low. The input resistance of the antenna is the sum of the radiation resistance and the resistance due to material losses: Pin = (RR + RL )|I |2 The gain of an antenna is the ratio of the peak radiation intensity to the input power averaged over the radiation sphere: gain =

Sr,peak r 2 Umax = Pin Pin 4π 4π

By using the idea of radiation resistance, we rewrite this as gain =

4πUmax (RR + RL )|I |2

AUXILIARY VECTOR POTENTIALS

49

Efﬁciency is the ratio of radiated power to input power: ηe =

Pr Rr |I |2 Rr = = Pin (Rr + RL )|I |2 Rr + RL

Instead of integrating the pattern to calculate the total power radiated, we sometimes compute the input power of the antenna from currents induced on the antenna elements by given voltage sources on various terminals of the antenna in analysis: Pin = Re(V1 I1∗ ) + Re(V2 I2∗ ) + · · · + Re(VN IN∗ ) The gain can be found from gain =

Sr (θ, φ)r 2 U (θ, φ) = Pin Pin 4π 4π

This method is considerably easier than integrating the radiation intensity to compute directivity. By integrating the pattern, we found only the input resistance of the short antenna, not the reactive component. A short antenna has a large capacitive reactance term that limits the impedance bandwidth when combined with a match network. The short antenna has a large pattern bandwidth but a narrow impedance bandwidth. Of course, an active network could be designed to impedance-match the antenna at any frequency, but the instantaneous bandwidth is narrow. The moment method of analysis gives us the currents for given input voltages and calculates the complete input impedance. 2-1.2 Radiation from Magnetic Currents Magnetic currents are ﬁctitious, but they enable slot radiation to be solved by the same methods as electric currents on dipoles by using duality. Slot radiation could be calculated from the surface currents around it, but it is easier to use magnetic currents to replace the electric ﬁeld in the slot. Magnetic currents along the long axis of slots in ground planes replace the electric ﬁelds across the slots by application of the equivalence theorem. Similarly, current loops can be replaced by magnetic dipole elements to calculate radiation. We use the electric vector potential F with magnetic currents. The far-ﬁeld magnetic ﬁeld is proportional to the electric vector potential: H = −j ωF

(2-9)

We determine the magnitude of the electric ﬁeld from Eq. (2-2); it is perpendicular to H. The electric vector potential is found from a retarded volume integral over the magnetic current density M. Applying the radiation approximation, it is e−j kr F= (2-10) ε M ej k·r dV 4πr where ε is the permittivity (8.854 × 10−12 F/m in free space). Equation (2-9) is the dual of Eq. (2-1), and Eq. (2-10) is the dual of Eq. (2-5). The dual of Eq. (2-3) is valid in both the near- and far-ﬁeld regions.

50

RADIATION STRUCTURES AND NUMERICAL METHODS

The magnetic currents in a slot are perpendicular to the slot electric ﬁelds: M = E × n, ˆ where nˆ is the normal to the plane with the slot. The ﬁlamentary currents of thin slots reduce Eq. (2-10) to a line integral, and magnetic current direction limits the direction of the electric vector potential and the magnetic ﬁeld. Since the electric ﬁeld (far ﬁeld) is orthogonal to the magnetic ﬁeld, the electric ﬁeld is in the same direction as the ﬁeld across the slots. We use the direction of the electric ﬁeld across the slots to estimate the polarization of the far ﬁeld. As with ﬁlamentary electric currents, the far ﬁeld is zero along the axis of the magnetic current. The electric vector potential can also be used to derive the near ﬁeld: H = −j ωF +

∇(∇ · F) j ωµε

1 E = − ∇ ×F ε The magnetic ﬁeld separates into near- and far-ﬁeld terms in the electric vector potential; the electric ﬁeld does not. We can determine the radiated ﬁelds directly in terms of the magnetic currents and avoid using the vector potential: E(r) = −

k2 4π

ˆ M(r ) × R

V

1 j + 2 e−j kR dV kR k R

1 j j − + 3 3 M(r ) − kR k 2 R 2 k R V j 3j 3 ˆ R ˆ e−j kR dV +[M(r ) · R] + 2 2− 3 3 kR k R k R

k H(r) = 4πη 2

(2-11)

(2-12)

Equations (2-11) and (2-12) can be rearranged to ﬁnd the dyadic Green’s functions for magnetic currents. These differ from the dyadic Green’s functions for electric currents by only constants. Example Derive the ﬁelds radiated from a small constant-current loop. We could use the magnetic vector potential and calculate over the currents in the wire but must account for changing current direction around the loop. Place the loop in the x –y plane. The electric ﬁeld radiated by the loop is in the φ direction because the currents in the loop can only be in the φˆ direction. When solving the integral for the magnetic vector potential, note that the direction of the current on the loop, φˆ at a ˆ unit vector. The integral general point is not in the same direction as the ﬁeld point, φ, must be solved with a constant vector direction, one component at a time. Although the magnetic vector potential can be computed, it is easier to replace the current loop with an incremental magnetic current element. The equivalent magnetic current element is Im l = j ωµI A where A is the area of the loop. The magnetic current density is M = Im lδ(r )ˆz = j ωµI Aδ(r )ˆz

APERTURES: HUYGENS SOURCE APPROXIMATION

51

The electric vector potential is found using Eq. (2-10): Fz =

j ωµεI A −j kr e 4πr

The magnetic ﬁeld is found from this electric vector potential using Eq. (2-9): Hz = −j ωFz =

ω2 µεI A −j kr e 4πr

We calculate Hθ by projection: ω2 µεI A −j kr e sin θ Hθ = Hz zˆ · θˆ = − 4πr Eφ and Hθ are related in the far ﬁeld because the wave propagates in the r direction: Eφ = −ηHθ =

ω2 µεI Aη −j kr e sin θ 4πr

The small current loop and small current element have the same pattern shape, sin θ , but opposite polarizations. The directivity is 1.5 (1.76 dB). Figure 2-1 uses a dashed curve to plot the response of the small loop, while the solid curve gives the pattern of a half-wavelength slot that radiates on both sides of the ground sheet. 2-2 APERTURES: HUYGENS SOURCE APPROXIMATION Many antennas, such as horns or paraboloid reﬂectors, can be analyzed simply as apertures. We replace the incident ﬁelds in the aperture with a combination of equivalent electric and magnetic currents. We calculate radiation as a superposition of each source by using the vector potentials. Most of the time, we assume that the incident ﬁeld is a propagating free-space wave whose electric and magnetic ﬁelds are proportional to one another. This gives us the Huygens source approximation and allows the use of integrals over the electric ﬁeld in the aperture. Each point in the aperture is considered to be a source of radiation. The far ﬁeld is given by a Fourier transform of the aperture ﬁeld: f(kx , ky ) = Eej k·r ds (2-13) S

This uses the vector propagation constant k = kx xˆ + ky yˆ + kz zˆ kx = k sin θ cos φ

ky = k sin θ sin φ

kz = k cos θ

where f(kx , ky ) is the pattern in k-space. We multiply the Fourier transform far ﬁeld by the pattern of the Huygens source: j e−j kr (1 + cos θ ) 2λr

(2-14)

When apertures are large, we can ignore this pattern factor. In Eq. (2-13), f(kx , ky ) is a vector in the same direction as the electric ﬁeld in the aperture. Each component

52

RADIATION STRUCTURES AND NUMERICAL METHODS

is transformed separately. The far-ﬁeld components Eθ and Eφ are found by projection (scalar products) from f(kx , ky ) times the pattern factor of the Huygens source [Eq. (2-14)]. If we have a rectangular aperture in which the electric ﬁeld is expressed as a product of functions of x and y only, the integral reduces to the product of two single integrals along each coordinate. The Fourier transform relationships provide insight into pattern shape along the two axes. Large apertures radiate patterns with small beamwidths. An antenna with long and short axes has a narrow-beamwidth pattern in the plane containing the long dimension and a wide beamwidth in the plane containing the short dimension. This is the same as the time and frequency dual normally associated with the Fourier transform. We draw on our familiarity with signal processing to help us visualize the relationship between aperture distributions and patterns. Large apertures give small beamwidths, just as long time pulses relate to low-frequency bandwidths in normal time–frequency transforms. The sidelobes of the pattern correspond to the frequency harmonics of an equivalent time waveform under the Fourier transform and rapid transitions in the time response lead to high levels of harmonics in the frequency domain. Rapid amplitude transitions in the aperture plane produce high sidelobes (harmonics) in the far-ﬁeld response (Fourier transform). Step transitions on the aperture edges produce high sidelobes, while tapering the edge reduces sidelobes and we relate the sidelobe envelope of peaks to the derivative of the distributions at the edges. To produce equal-level sidelobes, we need Dirac delta functions in the aperture that transform to a constant level in the pattern domain. Another example is periodic aperture errors that produce single high sidelobes. When we discuss aperture distribution synthesis, we see that the aperture extent in wavelengths limits our ability to control the pattern. A uniform amplitude and phase aperture distribution produces the maximum aperture efﬁciency and gain that we determine from the following argument. An aperture collects power from a passing electromagnetic wave and maximum collectible power occurs at its peak amplitude response. If the amplitude response somewhere else in the aperture is reduced from the maximum, that portion will collect less power. The amplitude response can be reduced only by adding loss or reﬂecting power in reradiation. The antenna with the highest aperture efﬁciency reﬂects the least amount of power when illuminated by a plane wave. Similarly, if the phase shift from the collecting aperture to the antenna connector is different for different parts of the aperture, the voltages from the various parts will not add in phase. Gain is directly proportional to aperture efﬁciency [Eq. (1-10)]. Therefore, a uniform amplitude and phase aperture distribution has maximum gain. All this assumes that the input match on various aperture distribution antennas is the same. For example, consider the pattern of a uniform aperture distribution in a rectangular aperture a × b. We use the Fourier transform and ignore the polarization of the electric ﬁeld in the aperture. (This assumes that the ﬁeld has a constant polarization or direction.) b/2 a/2 ej k·r dx dy f (kx , ky ) = E0 = E0

−b/2 b/2 −b/2

−a/2 a/2 −a/2

ej kx x ej ky y dx dy

APERTURES: HUYGENS SOURCE APPROXIMATION

53

We separate the integral into a product of two integrals each with the form a/2 ej kx a/2 − e−j kx a/2 a sin(kx a/2) ej kx x dx = = j k kx a/2 x −a/2 On combining the two similar integrals, we have f (kx , ky ) =

ab sin(kx a/2) sin(ky b/2) kx a/2 ky b/2

where kx = k sin θ cos φ, ky = k sin θ sin φ, and kz = k cos θ and k = 2π/λ. The pattern in both planes is given by a k-space function, sin u/u. Figure 2-2 plots this pattern function as a solid curve using kx -space [(ka/2) sin θ ] as the abscissa to produce a universal curve independent of aperture size a. The half-power points occur when sin u 1 =√ u 2

or

u = 1.39156

When we substitute for u, we have in the principal planes πa sin θ = 1.39156 λ By solving for θ , we compute the half-power beamwidth (HPBW): HPBW = 2 sin−1

0.4429λ a

By using the approximation u = sin u for small angles, the half-power beamwidth can be estimated as ◦λ HPBW = 50.76 a

Amplitude, dB

Cosine Uniform

ka/2 sin q

FIGURE 2-2 Universal k-space pattern for the radiation from uniform (solid curve) and cosine (dashed curve) aperture distributions.

54

RADIATION STRUCTURES AND NUMERICAL METHODS

Note that we have ignored the (1 + cos θ )/2 pattern of the Huygens source, which reduces the beamwidth for radiation from small apertures. We discover on Figure 2-2 the pattern nulls occur at integer multiples of π and the ﬁrst sidelobe amplitude is 13.2 dB below the peak. The gain of a uniform amplitude and phase aperture distribution is given by Eq. (1-7), where A is the area of the aperture. Chapter 4 develops amplitude taper efﬁciency for nonuniform aperture amplitude distributions to calculate the gain reduction. Phase error efﬁciency gives the gain reduction due to phase anomalies. Each of these is found from the distribution of aperture ﬁelds. Figure 2-2 plots the pattern of a half cosine aperture distribution as a dashed curve. The distribution peaks in the center and tapers linearly to zero at the edges. Tapering the aperture distribution widens the beamwidth and reduces both gain and sidelobe levels. The pattern beamwidth is 1.342 times wider than the uniform distribution beamwidth. A cosine distribution produces a −0.91-dB amplitude taper loss, and the distribution edge taper causes the sidelobes to fall off at a faster rate. Example Compute the length of the aperture with a uniform distribution that will give a 10◦ beamwidth. 50.76◦ a = 5 wavelengths λ 10◦ We can calculate radiated power by integrating the Poynting vector magnitude over the radiation sphere, but there is an easier way. We assumed that the aperture ﬁelds are free-space waves. The total power radiated is in the aperture, Pr |E|2 Pr = ds Pavg = Uavg = η 4π aperture

where η is the impedance of free space. The radiated electric ﬁeld is e−j kr (1 + cos θ ) f(kx , ky ) 2λr The Poynting vector magnitude is E=j

Sr =

(1 + cos θ )2 |E|2 = |f(kx , ky )|2 η 4λ2 r 2

(2-15)

By combining Eqs. (2-14) and (2-15), we determine directivity: Sr r 2 U (θ, φ) = Uavg Pr /4π 2 j k·r Ee ds π(1 + cos θ ) = λ2 |E|2 ds

directivity(θ, φ) =

(2-16)

By considering electric and magnetic ﬁelds separately in the aperture, we eliminate the requirement that the ratio electric and magnetic ﬁelds are the same as free space used in the Huygens source approximation. Given the ﬁelds in an aperture, we can equate

APERTURES: HUYGENS SOURCE APPROXIMATION

55

them to magnetic and electric currents: Ms = E a × n

Js = n × Ha

(2-17)

where Ea and Ha are the aperture ﬁelds and n is the outward normal. The equivalence theorem [2, p. 113] results in exact solutions by using the total aperture ﬁeld, incident and reﬂected. When using the equivalence theorem, we replace the total ﬁelds present with equivalent currents. The induction theorem equates currents only to the incident ﬁelds on the aperture, which ignores wave reﬂection and results in approximate solutions: Ms e−j k|r−r | Js e−j k|r−r | ds ds A=µ (2-18) F=ε 4π|r − r | 4π|r − r | s

s

We derive the radiated ﬁelds from each distribution of currents by using vector potentials where r is the ﬁeld point and r is the source point in the aperture. These expressions are valid in the near and far ﬁelds. By integrating over only a ﬁnite aperture, we assume zero ﬁelds outside the aperture, while rigorous expressions require integrals over closed boundaries. A planar aperture must extend to inﬁnity, but the ﬁelds outside the aperture are nearly zero and contribute little. 2-2.1 Near- and Far-Field Regions The radiative near- and far-ﬁeld regions are characterized by the approximations made to the integrals [Eq. (2-18)]. The radiative near-ﬁeld region lies between the near ﬁeld, with no approximations, and the far-ﬁeld region. In both approximations the ﬁeld (observation) distance r is substituted for |r − r | in the amplitude term. The vector potentials reduce to µ ε −j k|r−r | Ms e ds A= Js e−j k|r−r | ds (2-19) F= 4πr 4πr s

s

We handle the phase term differently in the two regions. First, we expand the phase term in a Taylor series, 1 |r − r | = r 2 + r 2 − 2r · r = r − rˆ · r + [r 2 − (ˆr · r )2 ] · · · 2r where rˆ is the unit vector in the ﬁeld point direction. We retain the ﬁrst two terms for the far-ﬁeld approximation and the vector potentials become e−j kr ε F= Ms ej k·r ds , etc. (2-20) 4πr s

where we have combined k, the propagation constant, with the unit vector rˆ : k = kˆr = k(sin θ cos φ xˆ + sin θ sin φ yˆ + cos θ zˆ ) The magnetic vector potential integral parallels Eq. (2-20) as in Eq. (2-19). In the radiative near-ﬁeld zone approximation the terms in r 2 are retained and we obtain the following integral for the electric vector potential: kr 2 (k · r )2 e−j kr ε Ms exp j (k · r ) + − (2-21) F= ds 4πr 2rk 2r

56

RADIATION STRUCTURES AND NUMERICAL METHODS

No clear boundary between the three regions exists because the ﬁelds are continuous. Common boundaries are r far ﬁeld L λ where L is the maximum dimension on the aperture. Example Determine the maximum difference between the radiative near- and far-ﬁeld approximations at a point normal to the maximum aperture dimension when r = L2 /λ and r = 2L2 /λ. Normal to the maximum dimension, rˆ · r = 0. The phase difference is

2 kr max 2r

= where rmax

phase difference φ =

L 2

2πL2 8λr

φ = π/4 at r = L2 /λ

and φ = π/8 at r = 2L2 /λ

The usual minimum distance used for antenna patterns is 2L2 /λ, where L is the maximum dimension of the antenna. At that distance, the phase error across the aperture from a point source antenna is π/8. The distance is not sufﬁcient for low-sidelobe antennas [3] because quadratic phase error raises the measured sidelobes. We can use vector potentials in the aperture after determining equivalent currents, but we will ﬁnd it more convenient to use the ﬁelds directly. Deﬁne the following integrals: j k·r Ea e ds g= Ma ej k·r ds (2-22) f= s

s

using the far-ﬁeld approximation. Near-ﬁeld integrals require additional phase terms. Given an aperture, we calculate the vector potentials in terms of Ea and Ha through the currents by using either the equivalence or inductance theorems, and we use the integrals of Eq. (2-22) in the vector potentials. We combine the ﬁelds in the far ﬁeld due to each partial source: E = −j ωA − j ηωF × rˆ For an aperture in the x –y plane, we carry out these steps by using the inductance theorem and obtain the following far ﬁelds from the incident aperture ﬁelds j ke−j kr [fx cos φ + fy sin φ + η cos θ (−gx sin φ + gy cos φ)] 4πr −j ke−j kr Eφ = [(fx sin φ − fy cos φ) cos θ + η(gx cos φ + gy sin φ)] 4πr Eθ =

(2-23)

BOUNDARY CONDITIONS

57

where f and g have been expanded in terms of their x- and y-components and η is the impedance of free space. 2-2.2 Huygens Source The Huygens source approximation is based on the assumption that the magnetic and electric ﬁelds are related as in a plane wave in the aperture: ηgy = fx

and

− ηgx = fy

ηHy = Ex

and

− ηHx = Ey

since

With this approximation, the far ﬁeld [Eq. (2-23)] becomes j ke−j kr (1 + cos θ )(fx cos φ + fy sin φ) 4πr −j ke−j kr Eφ = (1 + cos θ )(fx sin φ − fy cos φ) 4πr Eθ =

(2-24)

The two-dimensional vector Fourier transform f = (fx , fy ) of the aperture electric ﬁeld in the x –y plane determines the far-ﬁeld components. We derive the radiated components by projecting (vector scalar product) this ﬁeld onto the vectors θˆ / cos θ ˆ The transform f expands the ﬁeld in k-space [usually, (kx , ky )]. This normalizes and φ. the pattern and removes the direct dependence on aperture length. We separate out all but f when we consider aperture distributions. We drop the terms for the radiation from a point source and the pattern of a Huygens point source [Eq. (2-14)] and limit our discussions to Huygens sources and far ﬁelds. General aperture ﬁelds require Eq. (2-23), and for any region other than the far ﬁeld, additional phase terms are needed in the transforms [Eq. 2-21)]. 2-3 BOUNDARY CONDITIONS Material boundaries cause discontinuities in the electric and magnetic ﬁelds. The effects can be found by considering either vanishing small pillboxes or loops that span the boundary between the two regions. By using the integral form of Maxwell’s equations on these differential structures, the integrals reduce to simple algebraic expressions. These arguments can be found in most electromagnetic texts and we give only the results. Conversely, we will discover that artiﬁcial boundaries such as shadow and reﬂection boundaries used in geometric optics (ray optics) cannot cause a discontinuity in the ﬁelds because they are not material boundaries. The idea that the ﬁelds remain continuous across the boundary leads to the necessity of adding terms to extend ray optics methods. We discuss these ideas when considering the uniform theory of diffraction (UTD) method used with ray optics. Suppose that we have a locally plane boundary in space described by a point and a unit normal vector nˆ that points from region 1 to region 2. We compute the tangential ﬁelds from the vector (cross) product of the ﬁelds and the normal vector. The ﬁelds

58

RADIATION STRUCTURES AND NUMERICAL METHODS

can be discontinuous at the interface between the two regions if surface magnetic MS or electric current JS densities exist on the surface. nˆ × (E2 − E1 ) = −MS

nˆ × (H2 − H1 ) = JS

(2-25a,b)

The normal components of the ﬁelds change due to the differing dielectric and magnetic properties of the materials and the charges induced on the surface: nˆ · (ε2 E2 − ε1 E1 ) = ρS

nˆ · (µ2 H2 − µ1 H1 ) = τS

(2-26)

with ρS and τS given as electric and magnetic surface charge densities, respectively. Perfect dielectric and magnetic materials can have no currents, which reduces Eq. (2-25) to nˆ × (E2 − E1 ) = 0 nˆ × (H2 − H1 ) = 0 (2-27) Equation (2-27) means that the tangential ﬁelds are continuous across the boundary. These boundary conditions are used in the method of moment analyses to determine currents. The method applies the boundary condition in integral equations to determine the coefﬁcients of the expansion of currents in the sum of basis functions. The currents described as these sums do not satisfy the boundary conditions at all points but do when integrated over a region. This method leads to approximations that will converge as more terms are included in the expansions. When doing analysis we ﬁnd two types of surfaces convenient. We use these surfaces to reduce analysis effort by using planes of symmetry. The ﬁrst one is the perfect electric conductor (PEC). A PEC surface causes the ﬁelds to vanish inside and to have electric currents induced on it: nˆ × E2 = 0

nˆ × H2 = JS

(PEC)

(2-28a,b)

A PEC surface is also called an electric wall. The second surface is the perfect magnetic conductor (PMC) and is a hypothetical surface. Whereas good conductors approximate PEC, there are no PMC materials. The PMC has no internal ﬁelds like the PEC and forces the tangential magnetic ﬁeld to be zero: nˆ × E2 = −MS

nˆ × H2 = 0

(PMC)

(2-29)

A PMC surface supports the hypothetical magnetic current density MS . We ﬁnd that the magnetic wall (PMC) concept simpliﬁes analysis.

I

I

I

M V

I

I

I Electric Currents

FIGURE 2-3

M

M

+ − + −

V Voltage Source

M

M

M

Magnetic Currents

Ground-plane images.

PHYSICAL OPTICS

59

We use images of currents to include material boundaries in analysis. Figure 2-3 illustrates ground-plane images. When we analyze radiation from currents in the presence of a boundary, we include the actual antenna and its image to compute the ﬁelds. The ﬁgure shows an inﬁnite ground plane, but a ﬁnite ground-plane image can be used in the angular region where a reﬂected wave occurs in the ﬁnite plane. We consider this idea further when discussing geometric optics. We can use images in dielectric boundaries provided that we calculate the polarization sensitive reﬂection coefﬁcients to adjust the magnitude and phase of the image. 2-4 PHYSICAL OPTICS Physical optics uses things that can be measured. We can measure both currents and ﬁelds, but auxiliary vector potentials have no physical reality, only mathematical artifacts that simplify Maxwell’s equations. Nevertheless, the auxiliary vector potentials provide simple models for problems that enable simple mental pictures, as shown earlier, but we cannot easily formulate them into a systematic analysis tool for antenna problems. The physical optics analysis method combines the use of Green’s functions to calculate ﬁelds radiated by a given distribution of currents and then uses boundary conditions to determine the currents induced on objects due to incident ﬁelds. We compute the effects of a mounting structure by inducing currents on it and adding their radiation to the antenna pattern. The method assumes that radiation from the induced currents on the structure does not change the initial currents. We start analyses from either currents or incident ﬁelds and work from those. The resonant structure of many antennas determines the approximate current distribution that we normalize to the radiated power. We calculate the ﬁelds from these currents. Physical optics can use an iterative technique to calculate incremental currents induced on the original radiators and improve the solution, but we usually just sum the radiation from the original currents to the radiation from the induced currents. The second starting point for physical optics can be incident ﬁelds. These could be plane waves or could be ﬁelds found from the measured radiation patterns of antennas: for example, the pattern of a reﬂector feed. We add the radiation from the induced currents to the incident waves. 2-4.1 Radiated Fields Given Currents The radiated ﬁelds can be found from distribution of the electric and magnetic currents by the use of dyadic Green’s functions that contain source and ﬁeld coordinates. We sometimes refer to the Green’s functions as vector propagators or transfer functions between currents and ﬁelds. We calculate the ﬁelds from integrals over the source points of the dot (scalar) product between the dyadic and current densities. The dyadic Green’s function contains both near- and far-ﬁeld terms and requires slightly different expressions for the electric and magnetic ﬁelds. The general propagator from electric and magnetic currents has separate terms for electric and magnetic currents, which when used with surface patch currents can be reduced to short subroutines or procedures easily programmed [1]: (2-30) E(r) = GEJ (r, r ) · J(r ) dV + GEM (r, r ) · M(r ) dV H(r) = GHJ (r, r ) · J(r ) dV + GHM (r, r ) · M(r ) dV (2-31)

60

RADIATION STRUCTURES AND NUMERICAL METHODS

These expressions integrate over the currents located at source points r for a dyadic Green’s function that changes at each ﬁeld point r and source point r . Although these Green’s functions are valid at all ﬁeld points in space both near and far ﬁeld, they are singular at a source point. Only retaining terms with 1/R dependence for the far ﬁeld greatly simpliﬁes the expressions. When ﬁelds are incident on a perfect electric conductor (PEC), the combination of incident and reﬂected tangential magnetic ﬁelds induces an electric current density on the surface. The ﬁelds inside the conductor are zero. We assume locally plane surfaces on patches and compute currents that satisfy the boundary condition. Given the local unit normal nˆ to the surface, the induced current density is given by JS = nˆ × (Hincident + Hreﬂected ) Hincident = Hreﬂected JS = 2nˆ × Hincident

(2-32)

The reﬂected magnetic ﬁeld equals the incident magnetic ﬁeld because the ﬁeld reﬂects from the conductive surface. The sum of the tangential electric ﬁelds must be zero. Because the reﬂected wave changes direction, the vector (cross) product of the electric and magnetic ﬁelds must change direction. The reﬂected tangential electric ﬁeld changes direction by 180◦ , so the tangential magnetic ﬁeld must not change direction because the Poynting vector changed its direction. Equation (2-32) is the magnetic ﬁeld equation applied on a PEC. Equation (2-25b) is the general magnetic ﬁeld equation at a boundary. Physical optics starts with a given current distribution that radiates, or the measured pattern of an antenna. When an object is placed in the radiated ﬁeld, the method calculates induced current on the object to satisfy the internal ﬁeld condition. For example, PEC or PMC have zero ﬁelds inside. When we use simple functions such as constant-current surface patches, the sum of the radiation from the incident wave and the scattered ﬁelds from induced surface currents produces only approximately zero ﬁelds inside. As the patch size decreases, the method converges to the correct solution. To obtain the radiated ﬁeld everywhere, we sum the incident wave and scattered waves. The ﬁelds radiated by the induced currents produce the shadow caused by the object. With geometric optics techniques such as UTD, the object blocks the incident wave and we determine the ﬁelds in the shadow regions from separate diffraction waves. In physical optics the incident wave continues as though the object were not present. Only geometric optics techniques use blockage. We can calculate the ﬁelds radiated from antennas in free space or measure them in an anechoic chamber that simulates free space, but we mount the antenna on ﬁnite ground planes, handsets, vehicles, over soil, and so on, when we use them. Physical optics is one method of accounting for the scattering. We show in later chapters that the mounting conﬁguration can enhance the patterns. 2-4.2 Applying Physical Optics In this book we do not discuss how to develop numerical techniques, but it is important to understand how to apply methods. Whether you develop your own codes or use commercial codes, certain rules should be applied. Consider Eq. (2-32). The normal to

PHYSICAL OPTICS

61

the surface points in the direction of the incident wave: outward. If the normal pointed inward, the sign of induced electric current density would change. Most codes have made the assumption that the normal points outward, but some codes may check on the direction of the normal relative to the incident wave and make the necessary sign change. We must keep track of the direction of the normal, and it may be necessary to rotate the normal depending on the expected direction of the incident wave. If an object can have radiation from both sides, it may be necessary to use two objects in the analysis. Many codes store each object as a separate entity in a disk ﬁle. In some cases we need to store an object multiple times. Take, for example, a Cassegrain dual reﬂector. The feed antenna illuminates the subreﬂector and induces currents on it. These currents radiate and excite currents on the main reﬂector. When the main reﬂector–induced currents radiate, the subreﬂector intercepts or blocks part of the ﬁelds. We account for this blockage by using a second subreﬂector object on which the code calculates a new set of induced currents by using the main reﬂector currents as the source. We could add these currents to the existing disk ﬁle object or merely keep the second object. We want to keep the second object separate so that we can calculate additional currents induced on the main reﬂector using these currents as sources. These currents will be reduced from the initial set, but they are an important contribution to the ﬁelds radiated behind the reﬂector. This example illustrates iterative PO. When objects face each other signiﬁcantly, iterative PO is necessary to calculate correct patterns. The method converges rapidly in most cases. Figure 2-4 illustrates the geometry of a corner reﬂector. A half-wavelength-long dipole is placed between two metal plates usually bent to form a 90◦ angle. We can use other angular orientations between the plates, but this is the usual design. The ﬁgure does not show the feed line to the dipole, which usually starts at the juncture of the two plates and runs up to the dipole. This feed line contains the balun discussed in Section 5-15. Although the ﬁgure shows the plates as solid, many implementations use metal rods to reduce weight and wind loading. The analysis starts with assumed currents on the dipole. We divide the plates analytically into small rectangular patches, which can be small (≈ λ/8 to λ/4) on a side since it takes only a few to cover the plates. You should repeat the analysis with

FIGURE 2-4

Corner reﬂector with a dipole located between two ﬂat plates.

62

RADIATION STRUCTURES AND NUMERICAL METHODS

different-sized current patches to determine if the analysis has converged. In a similar manner, we break down the current on the dipole into short linear segments, each with constant amplitude. By using a near-ﬁeld version of Eq. (2-31), we calculate the magnetic ﬁeld incident on each patch on the plates. This ﬁeld induces electric currents on the plates calculated from Eq. (2-32). Remember that we combine radiation from the source dipole with that radiated from the induced currents to reduce the radiation behind the antenna. The currents were induced to satisfy the boundary condition of the plate, but only with both radiations present. Figure 2-5a illustrates this process of inducing currents. Figure 2-6 shows the antenna pattern calculated using these currents. The E-plane pattern drawn as a solid line produces a null at 90◦ because the dipole pattern has this null. The plates cause the narrowing of the beam in the H -plane. The plates reduced the back radiation to −22 dB relative to the forward radiation, called the front-to-back ratio (F/B). The gain has increased from the 2.1 dB expected from a dipole to 9.3 dB. An equivalent geometric optics analysis uses two images in the plates, as shown in Figure 2-5b, for the analysis. If you look at Figure 2-4 or 2-5, you should notice that the two plates face each other. Currents on one plate will radiate toward the other plate and induce another set of currents on it. We could ignore these induced currents if the radiation was insigniﬁcant, but to produce correct patterns we must include them. The solution to this problem calls for an iterative technique where we calculate the radiation from the currents on the ﬁrst plate and induce incremental currents on the second plate. These incremental currents produce further radiation that induces additional currents on the other plate. The method converges rapidly. Figure 2-7 gives the antenna pattern after the iterations have been completed and we include radiation from all currents. The actual F/B ratio of the antenna is 29 dB, and the additional currents increased the gain by 0.7 dB to 10 dB. Adding the two plates in the original analysis increased the gain by 7.2 dB, whereas the iterative technique had a much smaller effect. Figure 2-8 illustrates the iterative technique and shows that the equivalent geometric optics analysis adds a third image to represent the reﬂection between the plates. Remember when you mount the antenna in an application, the structure will change the realized pattern, but the high F/B ratio reduces this effect. The mounting structure used when measuring the antenna changes the pattern as well, which limits our knowledge of the real pattern.

Dipole Currents

Image (a)

Image (b )

FIGURE 2-5 Cross-sectional view of a corner reﬂector: (a) magnetic ﬁeld radiated from a dipole induces currents on plates; (b) plate currents replaced with image dipoles.

PHYSICAL OPTICS

E-plane

63

H-Plane

FIGURE 2-6 Pattern calculated from a combination of dipole and plate currents in a corner reﬂector with 1 × 0.9λ plates without induced current iteration.

Physical optics can determine the impedance effects of the limited images in the ground planes, such as the corner reﬂector. The local nature of impedance effects allows the use of images to calculate the mutual impedance effects of ground planes. We use impedance calculations not only to determine the bounds of ground-plane effects on input impedance, but to calculate the total power radiated by the antenna. The images (excited currents on ground planes) radiate but do not receive input power. A ground plane at least λ/2 on a side located about λ/4 away from the antenna produces nearly the same impedance effects as an inﬁnite ground plane, but the ground plane alters the radiation pattern greatly because it restricts possible radiation directions. It has commonly been thought that physical optics could compute the ﬁeld only in the main beam pattern direction of a paraboloidal reﬂector. The method can determine this pattern region accurately by using only a few patches, each one being many wavelengths on a side. As the processing power of computers increases, the patch size can be shrunk until PO can calculate the pattern in every direction, including behind the reﬂector. It is important to remember to include the feed pattern behind the reﬂector even though its radiation is obviously blocked by the main reﬂector. Physical optics uses induced currents to cancel the ﬁelds inside objects when the incident ﬁelds and the radiation from the induced currents are added. We can calculate the pattern behind a reﬂector using UTD (GTD), the uniform (geometric) theory of diffraction. This

64

RADIATION STRUCTURES AND NUMERICAL METHODS

E-Plane

H-plane

FIGURE 2-7 Pattern of corner reﬂector with 1 × 0.9λ plates with induced current iteration equivalent to multiple reﬂectors between the plates.

Dipole

Image

Image

( a)

Image (b)

FIGURE 2-8 (a) Wall currents on plates radiate magnetic ﬁelds that induce additional currents on facing plates; (b) added induced currents equivalent to additional image dipole.

geometric optics-based method blocks the radiation from the feed and uses diffractions from the rim edge to calculate the pattern behind the reﬂector. We discuss UTD in Section 2-7. A comparison of UTD and physical optics calculations [1,4] of the pattern behind shows that the two methods match. The dashed curve of Figure 2-9 plots the results of the PO analysis of a 20λ-diameter centrally fed paraboloidal reﬂector. The feed antenna radiation tapers to −12 dB at the

65

Gain, dB

PHYSICAL OPTICS

Pattern Angle

FIGURE 2-9 Physical optics analysis of a 20λ-aperture-diameter paraboloidal reﬂector (dashed curve) compared to analysis that includes PTD (solid curve).

reﬂector rim. Figure 2-9 shows the feed power spillover peaking at angles off the boresight near 100◦ . PO analysis computes the currents on a patch by assuming that it is embedded in an inﬁnite plate. The reﬂector rim violates this assumption and we need extra terms to calculate the pattern behind the reﬂector accurately. Adding PTD (the physical theory of diffraction) to PO improves the match between the two methods behind the reﬂector as shown by the solid curve on Figure 2-9. PTD handles caustic regions of PO in a manner similar to the equivalent current method based on diffraction coefﬁcients of UTD with geometric optics for shadow and reﬂection boundaries. For this example, the additional PTD currents add with the same phase because of the symmetry of the reﬂector geometry and produce the maximum effect. The PTD currents on the rim of an offset reﬂector will not add and produce a peak effect behind the reﬂector but will produce a more defuse effect. We only need PTD over a limited pattern angular range to reduce error, and the cost of implementing the ﬁx may exceed the necessity of knowing the pattern in these regions. Similarly, UTD needs the addition of edge currents for accurate calculation of the radiation near 180◦ , behind the reﬂector. Although any model for the feed pattern can be used with PO, results matching UTD exactly occur only over all regions of the back radiation when the feed satisﬁes Maxwell’s equations in the near and far ﬁelds [4, p. 212]. One such feed is the Gaussian beam approximation. Again, like PTD ﬁxes, the small errors when using other feed antenna approximations occur only at limited pattern regions that may be unimportant. 2-4.3 Equivalent Currents We can relate the concept of equivalent currents to physical optics. In this case we generate an artiﬁcial surface that covers a source of radiation. The incident ﬁelds generate surface electric and magnetic current whose radiation cancels the internal

66

RADIATION STRUCTURES AND NUMERICAL METHODS

ﬁelds and generates the external pattern. We use these at the apertures of antennas such as horns. By using the dyadic Green’s functions we can calculate the near-ﬁeld patterns and the coupling between antennas when the assumption is made that the presence of a second antenna does not alter the aperture ﬁelds. Given the outward normal n, ˆ we calculate the equivalent currents by nˆ × Eincident = −MS

nˆ × Hincident = JS

(2-33)

We must use both electric and magnetic current densities on the surface to replace the internal ﬁelds. If the ratio of the electric ﬁeld to the magnetic ﬁeld equals the impedance of free space (376.7 ), the combination of the two currents produces the radiation of the Huygens aperture source when used with the dyadic Green’s function. We use equivalent currents for a variety of analyses over ﬂat apertures such as horns and paraboloidal reﬂectors, but they can also be used with curved structures or apertures. We can, for example, use equivalent currents for calculation of the effects of radomes. Locally, we assume that the incident waves are plane waves and use boundary conditions to calculate reﬂected and transmitted waves. It is necessary to separate the incident wave into parallel and perpendicular polarizations, the ray-ﬁx representation discussed in Section 2-7.8. These polarizations have differing reﬂection and transmission coefﬁcients. We generate one surface on the inside of the radome and another on the outside. We use locally free-space waves for the reﬂected and transmitted waves lying outside the radome. Both these waves can be replaced with equivalent currents. The equivalent currents produce null ﬁelds inside the radome when combined with the incident wave radiation [4, p. 155]. Including these equivalent currents in a PO analysis, we add the effect of the radome. Equivalent currents can also be used with lenses. We use the incident waves combined with the idea of locally plane waves to calculate reﬂected and transmitted waves at each surface and replace them with equivalent currents. We include the dielectric constant of the lens in the dyadic Green’s functions for the internal radiation of the lens to calculate the ﬁelds at the second surface. We apply locally plane waves at the second surface to determine the transmitted and reﬂected rays and then replace them with equivalent currents. Because the lens has internal reﬂections, we need to apply an iterative PO analysis to calculate the multiple reﬂections between the two surfaces. The method converges rapidly because the internal reﬂections are small. 2-4.4 Reactance Theorem and Mutual Coupling In Section 1-14 we discussed how the coupling between two antennas can be found from reactance. Given a transmitting antenna that generates a ﬁeld at the receiving antenna, the reactance is described by an integral equation [5]: reactance =

(Et · Jr − Ht · Mr ) dV = t, r

(2-34)

The volume integral is over the receiving antenna currents, but it is often reduced to a surface or line integral. A second form of Eq. (2-34) uses the ﬁelds radiated by

METHOD OF MOMENTS

67

both antennas. Given a surface that surrounds the receiving antenna, the integral for reactance is taken over this surface: (Er × Ht − Et × Hr ) · ds = t, r

reactance =

(2-35)

Sr

The differential normal ds is pointed away from the receiving antenna. When we represent the two antennas and the transmission between them as an impedance matrix, it implies that we know the input currents to both antennas. By expressing the coupling as an impedance matrix, we compute mutual impedance from the reactance integral: −1 t, r (2-36) Z12 = I1 I2 Antennas that we describe by input currents only have electric current densities excited on their surfaces. The mutual impedance formula using reactance reduces to Z12 =

−1 I1 I2

Et · Jr dV

(2-37)

Vr

The volume integral reduces to a line integral in most cases. Antennas with given input voltages such as slots can be described using magnetic currents and we use a mutual admittance matrix for the antenna pair: Y12 =

1 −1 · reactance = Vt Vr Vt Vr

Ht · Mr dV

(2-38)

Vr

By using reciprocity antennas made of linear, isotropic materials, we have equal crossmatrix terms: Z12 = Z21 and Y12 = Y21 (2-39) We calculate self-impedance terms by integrating over the surface of the antenna: for example, the radius of a dipole with the source of the ﬁeld located at the center of wires or slots. 2-5 METHOD OF MOMENTS The method of moments (MOM) [6] expands the currents on an antenna (or scattering object) in a linear sum of simple basis functions. The approximate solution is a ﬁnite series of these basis functions: N fa = ai fi (2-40) i=1

We compute the coefﬁcients by solving integral equations to satisfy boundary conditions on the surface of the antenna (or object). The integral equation can be expressed in the form Lfa = g, where L is a linear operator, usually a scalar product using an integral, fa the unknown currents given by Eq. (2-40), and g the known excitation

68

RADIATION STRUCTURES AND NUMERICAL METHODS

or source function. We substitute the summation of Eq. (2-40) into the linear operator equation and use the scalar product integral to calculate the terms in a matrix equation. The solution of the matrix equation determines the coefﬁcients of current expansion. The MOM produces ﬁlled matrices that require time-consuming numerical methods for inversion. The art of the MOM is in choosing basis functions and deriving efﬁcient expressions for evaluating the ﬁelds using the basis function currents. Common basis functions are simple staircase pulses, overlapping triangles, trigonometric functions, or polynomials. The method does not satisfy boundary conditions at every point, only over an integral average on the boundaries. By increasing the number of basis functions, the method will converge to the correct solution. We need to judge how many terms are required for an adequate engineering evaluation. Spending excessive time on the solution cannot be justiﬁed if it greatly exceeds our ability to measure antenna performance accurately using real hardware. 2-5.1 Use of the Reactance Theorem for the Method of Moments We can use the reactance theorem to generate a moment method solution to the currents on a thin-wire antenna. Thin-wire solutions assume that there are no circumferential currents and reduces the problem to ﬁlamentary currents. An electric ﬁeld integral equation (EFIE) satisﬁes the boundary condition of Eq. (2-25a), a zero tangential ﬁeld at the surface of the wires, but it does not seem explicit in the derivation. The reactance theorem produces an impedance matrix whose inversion yields the coefﬁcients of the current expansion [7]. Similar to many other methods, the Green’s function has been solved explicitly to reduce run time. This method [7] uses overlapping sinusoidal currents on V-dipoles as basis function currents and uses the Green’s function to calculate the radiation from one V-dipole at the location of a second V-dipole. Both the radiating and receiving dipoles use the same expansion function. Galerkin’s method uses the same weighting (or testing) function as the basis function and yields the most stable solutions. The reactance equation (2-37) calculates the mutual impedance between the two dipoles when each has unity current. We compute self-impedance by spacing a second V-dipole one radius away and by using the reactance theorem to calculate mutual impedance, a technique equivalent to the induced EMF method. The scalar (dot) product between the incident vector electric ﬁeld and the current density along the dipole reduces the vectors to scalars that can be integrated. The current density acts as the testing or weighting function for the method of moments. Performing the integration means that the current density only satisﬁes the zero tangential electric ﬁeld boundary condition in an average sense. If series impedances are placed in the V-dipole, their impedance is added to the diagonal elements of the mutual impedance matrix. To excite the structure, we place a delta voltage source in series with the Vdipole terminals. The solution for the currents can be found by inverting the matrix equation and using the voltage excitation vector starting with the matrix equation [Zmn ][Im ] = [Vn ]

(2-41)

After computing the matrix inverse and specifying the input voltage vector, the complex current values are found on the structure: [In ] = [Zmn ]−1 [Vm ]

(2-42)

METHOD OF MOMENTS

69

Given the input voltage and the solution for the currents, the input impedance can be calculated. Similarly, the far- and near-ﬁeld patterns can be calculated by using Eqs. (2-30) and (2-31) of the dyadic Green’s function. The code must satisfy Kirchhoff’s current law at the junction between groups of Vdipoles, which adds a constraint to the currents. Because an overlapping sinusoidal basis function closely follows the actual currents normally excited on dipoles, the segments can be on the order of a quarter-wavelength long or more and yield acceptable results. Basis functions that closely follow expected current distributions are sometimes called entire domain functions. These reduce the size of the matrix to be inverted but require more complicated calculations for matrix terms and radiation. Although the concept of a V-dipole was expanded to a V rectangular plate [8], the method is only a subset of general integral equation solutions. This approach generates a simple impedance matrix formulation easily understood from an engineering point of view. 2-5.2 General Moments Method Approach The method of moments can solve other types of electromagnetic problems: for example, electrostatic problems involving charges and dielectrics [9]. These solutions can determine the characteristic impedance of transmission lines useful in the design of antenna feeders. All moment method solutions are found from the solution of integral equations over boundary conditions. The boundary conditions can be either the tangential electric ﬁeld (EFIE) or magnetic ﬁeld (MFIE) conditions given by Eq. (2-25a,b) or a combination applied using an integral scalar product. We need a combination for closed bodies near an internal resonance frequency (resonant cavity) because the solutions exhibit resonances that make the solution invalid over a narrow frequency range. The method of moments can be applied to dielectric bodies when we use the constitutive relations of Eqs. (2-25) and (2-26), where the formulations for dielectric bodies use either volume or surface integrals [9]. Consider the use of the electric ﬁeld integral equation (EFIE) with metal surfaces. We expand the currents on the objects using basis functions Bm (r ) with coefﬁcients Im : J(r ) =

Im Bm (r )

(2-43)

The basis functions can be applied over a limited range of the structure in piecewise linear functions, which can be staircase pulses, overlapping triangular functions, or sinusoidal basis functions, whereas multiple functions can be applied over the whole or part of the structure for entire domain basis functions. For example, these could be a sum of sinusoidal functions which form a Fourier series representation. On a PEC surface the tangential electric ﬁeld vanishes [Eq. (2-28a)]. At ﬁeld point r along the surface S, nˆ × [Eincident (r) + Escattered (r)] = 0 M Im Bm (r ) · G(r, r ) ds Escattered = m=1

(2-44)

s

We can only satisfy Eq. (2-44) using a ﬁnite sum in the average sense of an integral. Since the integral and summation operate on a linear function, we can interchange them.

70

RADIATION STRUCTURES AND NUMERICAL METHODS

We introduce weighting (or testing) vector functions tangent to the surface Wn (r) and take the scalar (dot) product of this vector with the sum of electric ﬁelds. This limits the result to the tangential component of the electric ﬁeld: [Eincident (r) · Wn (r) + Escattered (r) · Wn (r)] ds = 0

(2-45)

s

We identify the weighted integral of the incident ﬁeld with the source and weighted integral of the ﬁeld radiated by the basis functions (scattered ﬁeld) as the impedance matrix terms. The integrals over the boundaries are one form of scalar product represented by · notation. Using unity current on each basis function, we calculate the matrix terms by using the scalar product:

Bm · G(r, r ) ds , Wn (r) =

Zmn = s

Vn = −Eincident (r), Wn (r) = −

Bm · G(r, r ) · Wn (r) ds ds (2-46)

s

s

Eincident (r) · Wn (r) ds

(2-47)

s

The combination of Eqs. (2-46) and (2-47) when integrated over each portion of the source gives a matrix equation: [Zmn ][Im ] = [Vn ]

(2-48)

The weighting functions could be as simple as pulse functions, overlapping triangular functions on lines or surfaces (rooftop), piecewise sinusoidal functions, or others. The type of basis functions determines the convergence more than the weighting (testing) functions, which only determine the averaging. Realize that the moment method converges to the exact solution when we increase the number of basis functions, but it is a matter of engineering judgment to determine how many terms give acceptable answers. Equation (2-47) deﬁnes the source voltage occurring over a segment when the formulation uses a piecewise function expansion. The incident voltage is the weighted integral of the incident electric ﬁeld. For example, the NEC formulation applies an excitation voltage across one segment. The reaction integral formulation of Section 25.1 applies a voltage source at the end of a segment. The modeling of sources is an important part of the art in the method of moments. The expansion of Eq. (2-44) is only one possible moment method solution. We could use the boundary condition on the magnetic ﬁeld, a combination of the electric and magnetic ﬁeld conditions on a PEC. If the surface has ﬁnite conductivity, the boundary conditions are modiﬁed. The moment method is a general method that computes approximate solutions to the currents. Unlike physical optics, the currents do not have to be assumed beforehand but are found as a ﬁnite series approximation. Antenna designers discover that adequate codes are available for most problems. Moment method solutions are typically limited to objects only one or two wavelengths in size, although any method can be stretched. Analysis of large structures becomes intractable because of the large amount of computer memory required and the length

METHOD OF MOMENTS

71

of time needed to calculate the solution. Coarse models may not give totally accurate results but can be useful in determining trends. Given these ideas, remember that physical models can be built that solve the electromagnetic problem instantaneously. We found that it takes considerable time to learn any code, and a new code has to offer considerable advantages or solve problems that the present one cannot solve before we invest our time. 2-5.3 Thin-Wire Moment Method Codes Thin-wire codes that assume only ﬁlamentary currents are readily available. We have experience with NEC, the Richmond code (ASAP), and AWAS [10], a commercial code. All have advantages, but they take time to learn. A commercial code with a graphical interface makes the input and output easier: for example, for NEC. These pay for themselves quickly by saving time. NEC can include plates, but since it uses a MFIE (magnetic ﬁeld integral equation) for them, it is limited to closed bodies. When accuracy becomes important, it is necessary to decrease the segment length and increase their number. These codes use matrix inversion with calculation time proportional to N 3 and a matrix ﬁll time proportional to N 2 . Run time increases enormously as the number of segments increases. The commercial code AWAS determines the segmentation, while the user of NEC must specify it. The rule is to use at least 10 segments per wavelength, but initial analysis can tolerate the errors due to using fewer segments. The segments should be longer than the diameter, and care must be taken that the segments do not overlap because the radius of the wires is too large. Solid objects, such as plates, can be modeled as wire frames, with the rule that the perimeter of the wire equal the spacing between the wires [11]. This rule can be violated, but a test of the convergence should be made. When we model slots in a solid object, we cannot apply the perimeter equalspacing rule because the slot will disappear. These codes compute the radiation pattern more accurately than the input impedance due to simplistic source models, and we may have to build the antenna to determine the true input impedance. Of course, an antenna with a good input impedance response that does not have the required pattern is useless. We can reduce NEC run time if the antenna has symmetry with multiple inputs. The code reduces input by allowing the user to specify symmetry. For example, a multiarm spiral analysis requires only the input of one arm. The various mode voltages are entered after the basic structure impedance matrix has been solved. If an object has M-way symmetry, the matrix ﬁll time is reduced by M 2 and the solution time by M 3 . The various voltage modes can be applied afterward. If we add another wire segment after specifying symmetry, the symmetry is destroyed and the program uses the full matrix. The only advantage we gain is in specifying the model because the program solves the full matrix instead of the reduced matrix. 2-5.4 Surface and Volume Moment Method Codes Antennas made of plates or containing ﬁnite plate ground planes can be solved by using wire meshing of a thin-wire code. The method of moments code has been extended to plates [12,13] using a rooftop basis function on both rectangular and triangular patches. The number of basis functions (i.e., matrix size) grows rapidly. One solution is to use entire domain basis functions. These require more complicated integrals, but

72

RADIATION STRUCTURES AND NUMERICAL METHODS

they reduce the matrix size. Dielectric portions of the problem lead to either volumetric integrals or various forms of surface integrals that use equivalent currents to replace the internal ﬁelds [9,14]. These problems lead to a variety of boundary conditions solved using a ﬁnite series of basis function and integral equations to satisfy those boundary conditions approximately. MOM analysis of antennas mounted on dielectric substrates requires special techniques. Commercial codes determine the currents ﬂowing on these antennas while accounting for the dielectric. Often, Green’s functions are found numerically, which increases the execution time. Since the currents are located on the surface and the integrals of the boundary conditions are over the same surface, the singularity of the Green’s function causes a numerical problem. For example, the free-space Green’s function has the term 1/|r − r |, which becomes inﬁnite on the surface. Spectral domain methods remove the singularity by using a sum of current sheets on the surface as an entire domain basis function. A uniform plane wave propagating at an angle to the surface excites the current sheet. The actual current ﬂowing on the metal portions is expanded as a sum of these current sheets [15, p. 208ff; 16]. The uniform current sheets are expanded in a spatial Fourier transform as well as the Green’s function, and the MOM problem is solved. The Fourier-transformed Green’s function no longer has the singularity. When the metallization can be expressed as an inﬁnite periodic structure, the current is expanded as a Fourier series. The inﬁnite periodic structure is used with frequency-selective surfaces and inﬁnite arrays. In this case the ﬁelds and currents are expanded in Floquet modes (harmonics). 2-5.5 Examples of Moment Method Models Figure 2-10 demonstrates the use of a wire mesh to replace a solid plate. We located a resonant (≈ λ/2) dipole λ/4 distance over a λ-wide ground plane in the H -plane and offset 3/8λ from one edge. This is repeated in Figure 2-20 using GTD analysis. The rods only run parallel to the dipole because cross wires do not have currents induced on them in the ideal world of analysis. The circumference of the rods equals the spacing between the rods and forms an equivalent solid plate. An actual antenna could use smaller-diameter rods and work as effectively as the solid plate and would reduce weight and wind loading. NEC analysis produces the same pattern as the GTD analysis of Section 2-7.2, except that the E-plane size of the rods alters the backlobe predicted by GTD to some extent, because that analysis assumes inﬁnite-length rods.

FIGURE 2-10 Use of a wire mesh to replace a solid plate for dipole over a ground plane in a MOM calculation.

METHOD OF MOMENTS

73

Over most of the pattern angles the two analyses produce identical results. The NEC analysis accounts for the mutual impedance between the dipole and its image in the ﬁnite ground plane. For impedance calculations a small ground plane gives almost the same reaction to the antenna as an inﬁnite ground plane. Figure 2-11 shows a wire frame model of a cell phone. The model contains more wires than necessary for λ/10 spacing, but more wires improve the geometry match. When using crossed wires that shield both polarizations, we reduce the wire circumference in half since the wires approach the squares from four sides. The small wire antenna must be connected to the wire grid of the model to generate proper currents on the box. Either we restrict possible locations of the antenna or we must distort the wire grid locally. You should write an automatic grid generator if you use this analysis often. Consider that you need to specify whether an edge wire should be generated when two plates share the same edge. The hand holding the cell phone and the head nearby have signiﬁcant effect on the antenna performance. The model given in Figure 2-11 has limited use. We need either a moment method analysis, such as WIPL-D, which includes volume dielectric structures, or FDTD, which can include complex material structures to model the head and produce good results. Figure 2-12 illustrates a wire frame model of an airplane used for low-frequency analysis. Antennas mounted on free-ﬂying models such as airplanes or spacecraft will excite the structure. Electrically, small antennas can excite the entire vehicle as an antenna. For example, a small antenna mounted on a large ground plane that would produce vertical polarization can excite the wings or fuselage and the entire system will

FIGURE 2-11 Wire frame MOM model of a cellular telephone handset with an antenna connected to the mesh.

74

RADIATION STRUCTURES AND NUMERICAL METHODS

FIGURE 2-12 Wire frame MOM model of an airplane.

radiate horizontal polarization. Models similar to Figure 2-12 can eliminate surprises. The model restricts antenna mounting locations to the wire positions and may require local distortions of the grid. Moment methods can include solid plates. Figure 2-13 shows an open waveguide horn analysis that uses a combination of plates and a single-feed wire monopole [12]. Locating the monopole or a small dipole inside the waveguide produces excitation of the waveguide mode that feeds the horn. Even though the model does not necessarily produce accurate impedance information, the model accurately calculates the pattern generated by the currents excited in the walls. We can either use an aperture method

FIGURE 2-13 MOM model of a pyramidal horn using ﬂat plates fed by a small dipole. (From [14, p. 229].)

METHOD OF MOMENTS

75

PEC (a)

PMC

(b)

FIGURE 2-14 Use of electric and magnetic walls to reduce the model size in MOM analysis of a pyramidal horn: (a) PEC wall divides the horn; (b) PMC wall divides the horn.

for the horn that replaces the aperture ﬁelds or use the currents excited in the walls to calculate the pattern. Either method works for the front lobe. The moment method calculation requires signiﬁcantly greater calculation time but produces results that better match measurements in all directions. Figure 2-14 demonstrates how to reduce calculation time by using planes of symmetry in a moment method analysis. In this case the small dipole feed is separated by two equally fed closely spaced dipoles. The right–left symmetry of the antenna allows reduction of the model by half. A vertical PMC wall divides the antenna into two parts, with only one remaining in the analysis. A horizontal PEC conductor divides the remaining model in half because halfway between the dipole feed is a virtual short circuit. Figure 2-14 contains only one-fourth the size of the original problem. Since matrix inversion requires N 3 calculations for an N × N matrix, dividing the analytical model down to one-fourth size reduces this calculation by a 64 : 1 factor. This also reduces the matrix element (ﬁll time) calculations by 16 : 1. Reducing the model by using symmetry planes enables the solution of larger problems and reduces calculation time. Analyses in later chapters use the moment method to predict antenna performance. Wire frame and plate analyses determine vehicle and mounting structure pattern effects. The moment method produces excellent analyses because it determines the approximate

76

RADIATION STRUCTURES AND NUMERICAL METHODS

current distribution as a sum of simple basis functions and we need not start with an assumed current distribution on the antenna.

2-6 FINITE-DIFFERENCE TIME-DOMAIN METHOD The ﬁnite-difference time-domain (FDTD) method solves the coupled Maxwell’s curl equations directly in the time domain by using ﬁnite time steps over small cells in space. The method reduces the differential equations to difference equations that can be solved by sets of simple equations. The method alternates between the electric and magnetic ﬁelds solved at locations a half-step apart because central differences are used to approximate derivatives. A 1966 paper by Yee [17] described the basic method that many authors have improved upon, but the original method remains the approach of choice. FDTD can solve many types of electromagnetic problems, of which antenna analyses are only one type. Computer memory and speed limit the size of problems that can be solved, but larger and larger problems can be solved as the cost of computing keeps reducing. Besides antenna problems, the method is applied to microwave circuits, biological interaction with electromagnetic waves, optics, and radar cross-section problems. The number of uses expands every day. The method allows each cell to be made of different materials, leading to the solution of volumetric complex structures. The solution of the equations is robust and the errors are well understood. Currently, the method solves moderately small antenna problems on the order of a few wavelengths. Of course, faster and larger computers can solve larger problems, especially if the analyst has patience. FDTD handles microstrip antennas with their complex layering of dielectrics, including a ﬁnite ground plane without the use of complex Green’s functions required of frequency-domain solutions. The interaction of antennas with the near environment, such as the effect of the head on cellular telephone handsets, can be solved. In this case the complex electromagnetic properties of the head can be described as cells each with different electrical properties. In addition to giving a solution to the radiation pattern and allowing characterization of the communication system, it can provide insight into the radiation safety concerns of users. The method handles the solution of the interaction of antennas with the human body in a straightforward manner for prediction of biomedical applications, such as electromagnetic heating for cancer treatment. Learning to apply the technique, whether formulating your own routines or using a commercial code, will yield insight for design. The method can produce time-domain animated displays of the ﬁelds that show radiation centers and where the ﬁelds propagate, but the user must learn to interpret these new displays. It will be worth your effort to learn this task. The time-domain responses using impulse signals can produce solutions over a wide band of frequencies when converted to the frequency domain using the discrete Fourier transform (DFT). The only drawback is the computer run time required. 2-6.1 Implementation By using a direct implementation of Maxwell’s curl equations in the time domain, you do little analytical processing of the equations. No vector potential or Green’s

77

FINITE-DIFFERENCE TIME-DOMAIN METHOD

functions are developed as in frequency-domain methods. Although the antenna may be volumetrically complex and contain many different materials, the method yields sparse matrices rather than the dense matrices produced by moment methods. It is a direct solution that does not require the inversion of large matrices and includes only nearest-neighbor interactions. Having only nearest-neighbor interactions means that it is possible to run problems on parallel machines. You need to embed the antenna in a rectangular region and divide it into rectangular cubical cells with sizes ranging from 10 to 20 samples per wavelength at the highest frequency where analysis is desired. The outer surfaces contain absorbing boundaries to eliminate reﬂections that would produce errors. Formulating absorbing boundary conditions has been a signiﬁcant part of the method. You need to locate a solution surface between the absorbing boundaries and the antenna outer surface where we compute currents by using the equivalence theorem. The DFT of the time response determines the radiation pattern at a given frequency after the equivalent currents are found. If you need the pattern amplitude in only a few directions, the time-domain radiation can be found directly: for example, the gain in one direction. We can formulate some problems in one or two dimensions if they possess symmetry instead of the three-dimensional rectangular cube. The solution time is reduced dramatically, and the time-animated presentation may provide sufﬁcient insight when the radiation pattern is found in two dimensions. Because this is a time-domain analysis, we need to excite the structure with a pulse. You use the pulse frequency power response to normalize the patterns and compute gain. When the formulation includes the material losses, the efﬁciency of the antenna can be found since the dissipation in the inner cells prevents the radiation from reaching the outer surface. 2-6.2 Central Difference Derivative Numerical derivatives have greater potential for errors than integrals, but FDTD uses them to reduce Maxwell’s differential curl equations to simple difference equations. A second-order accurate formula for a derivative can be found by using central differences instead of using the difference between the value at a location or time and the value at the next point in a sequence of evenly spaced points: f (u0 + u/2) − f (u0 − u/2) ∂f = + O( u)2 ∂u u

(2-49)

We can use ﬁnite differences to solve the curl equations provided that we use electric and magnetic ﬁelds spaced at half intervals because each is related to the derivatives of the other ﬁeld and we want to use central differences to reduce error. Because Maxwell’s equations involve time derivatives, we need to calculate the electric and magnetic ﬁelds at interspersed half time intervals. 2-6.3 Finite-Difference Maxwell’s Equations Consider Maxwell’s curl equations in the time domain, including lossy materials: ∂H 1 = − (∇×E − M + σ ∗ H) ∂t µ ∂E 1 = − (∇×H − J + σ E) ∂t ε

(2-50)

78

RADIATION STRUCTURES AND NUMERICAL METHODS

Equations (2-50) contain the source currents J and M and include losses due to conducting dielectric material σ and magnetic material losses σ ∗ . Both equations have the same form, with only an interchange of symbols. Expanding the curl operator, we get the following equation for the x-component of the magnetic ﬁeld: 1 ∂Ey ∂Ez ∂Hx ∗ = − − Mx − σ Hx (2-51) ∂t µ ∂z ∂y The x-component of the electric ﬁeld has the same form but with the interchanges H → E, E → H , M → J , and σ ∗ → σ . You obtain the equations for the y- and z-components by a cyclic variation (repeating pattern of interchanges) x → y → z → x → y, and so on. For example, the equations are reduced to two dimensions by leaving out the y-component. FDTD calculates the ﬁeld at discrete times and locations on a grid. The ﬁelds can be represented as an indexed function using integers: f (i x, j y, k z, n t) = f (i, j, k, n) Because we use central differences [Eq. (2-49)], for derivatives, and the magnetic (electric) ﬁeld is found from the space derivative of the electric (magnetic) ﬁeld, the magnetic and electric ﬁelds need to be spaced a half-space interval apart. The time derivative becomes f (i, j, k, n + 12 ) − f (i, j, k, n − 12 ) ∂f (i, j, k, n) = ∂t t and means that the electric and magnetic components are interspersed at t/2 times that which produces a leapfrog algorithm. We substitute these ideas into Eq. (2-51) to derive the time-stepping equation for one component: Hx (i − 12 , j, k, n + 1) =

1 − σ ∗ (i − 12 , j, k) t/2µ(i − 12 , j, k) 1 + σ ∗ (i − 12 , j, k) t/2µ(i − 12 , j, k) +

Hx (i − 12 , j, k, n)

t/µ(i − 12 , j, k) 1 + σ ∗ (i − 12 , j, k) t/2µ(i − 12 , j, k)

Ey (i − 12 , j, k + 12 , n + 12 ) − Ey (i − 12 , j, k − 12 , n + 12 ) z −

Ez (i − 12 , j + 12 , k, n + 12 ) − Ez (i − 12 , j − 12 , k, n + 12 ) y

− Mx (i − 12 , j, k, n + 12 )

(2-52)

FDTD uses similar equations for the other components [18,19]. Yee’s Cell Figure 2-15 shows one cubic cell and the components of the ﬁelds. When we consider the upper face, we see that the magnetic ﬁeld components are spaced a

FINITE-DIFFERENCE TIME-DOMAIN METHOD

79

Hy

z Ez Hx

Hx Hz Hy Hz

Ey

Hz Ex Hx (i,j,k)

Hy y

x

FIGURE 2-15 Unit cell of a Yee space lattice showing time and space separation of electric and magnetic ﬁelds in a cell. (From [15], Fig. 1, 1966 IEEE.)

half space interval from the central electrical ﬁeld and the arrows show the direction of ﬁelds. Although it would appear that the electric ﬁeld is different on the upper and lower face along the z-axis, the method assumes that the ﬁeld is constant throughout the cell. The magnetic ﬁelds shown are at the center of adjoining cells. A leapfrog solution uses stored values of the electric ﬁelds to calculate the magnetic ﬁelds at a half time interval later and stores these values. In the second step the solution takes another half time step and uses the stored values of the magnetic ﬁelds to calculate the electric ﬁelds. The method gains stability by using the half time steps and by solving for both electric and magnetic ﬁelds. Although the ﬁelds are a half time step out of synch, we can average between the two half time steps to produce simultaneous ﬁelds at a point, but we only need to do this when calculating equivalent currents on the surface used for far-ﬁeld pattern calculations. 2-6.4 Time Step for Stability You need to pick the time step to produce a stable solution. Consider a plane wave traveling through the cubes. If the time step is too large, the wave can pass through more than one cell for each time step. At that point the solution cannot follow the actual wave propagation and fails. We must reduce the time step until it is less than the Courant condition or the wave propagation rate. Consider the fastest-moving wave in the problem, usually free space, and for equal sides to the cube, we compute the time step from the velocity and cell length: x v t ≤ √ d

(2-53)

The cell length is x and the number of dimensions is d. The time step must be lower for conducting materials (σ > 0) to produce a stable solution. The magic step uses

80

RADIATION STRUCTURES AND NUMERICAL METHODS

the equality and produces the most stable solutions. If you pick unequal sides to the rectangular cell, Eq. (2-53) is modiﬁed. 2-6.5 Numerical Dispersion and Stability FDTD analyses produce solutions that fail to propagate through the cells at the proper phase velocity in all directions. The propagation velocity depends on the cell size in wavelengths; it has a frequency-dependent component. You need to consider this numerical dispersion because it affects accuracy. Because the waves travel at different velocities in different directions, the dispersion problem increases for large structures where many time steps must be taken. After many steps, signals disperse because they have taken different routes and fail to add together in the correct phase. Finer cells solve the problem, but the computation requirements grow rapidly. The equation for the propagation constant can be found from considering the FDTD formulation to produce the following equation for three-dimensional problems: ky y 2 kz z 2 1 ω t 2 kx x 2 1 1 1 = + + sin sin sin sin c t 2 x 2 y 2 z 2 (2-54) The factor kx is the FDTD propagation constant in the cells along the x-axis, only approximately the same as kx , the actual propagation constant in the structure. The yand z-axes have similar problems. If you take the limit as cell length approaches zero, u → 0, and so on, then sin(au)/u → a. Because t → 0 as the cell size shrinks for the solution still to satisfy the Courant limit, Eq. (2-54) reduces to the expression ω 2 = kx2 + ky2 + kz2 (2-55) c Equation (2-55) is the normal propagation constant equation for a plane wave in space and shows that the cell propagation constants converge to the correct values as the cell size shrinks. If you formulate a problem in one or two dimensions, you remove terms from the right side of Eq. (2-54) to determine the dispersion relationship. Absorbing boundary conditions (ABCs) can cause numerical instabilities. ABCs approximate inﬁnite space to simulate radiation by the antenna into space. FDTD problems must be placed in a ﬁnite number of cells because each cell requires computer storage. Every FDTD problem uses a ﬁnite number of cells for the ABCs with more cells required in the directions of maximum radiation. ABCs degrade as the number of time steps increases and eventually leads to numerical instabilities. A lively research on ABCs has produced good ones, but be aware that most have been found to produce problems at some point. If you write your own analyses, you will need to ﬁnd appropriate ones. Commercial codes will give their limitations. At one time, ABCs limited solution dynamic range, but ABCs are now available that produce reﬂection coefﬁcients from 10−4 to 10−6 . Numerical dispersion limits the dynamic range as well. Remember that the antenna will be modeled with small cubes that limit the resolution of the results. The errors of modeling lead to solution errors that limit the dynamic range. 2-6.6 Computer Storage and Execution Times The antenna to be analyzed is modeled by a set of cubic cells. Choosing an appropriate number is an art. Similarly, it will be necessary to have a meshing program. Using a

FINITE-DIFFERENCE TIME-DOMAIN METHOD

81

two-dimensional model will greatly reduce computer storage and run time. Remember that our purpose should be to gain insight unavailable from measurements. The calculations require the storage of three components of both the electric and magnetic ﬁelds in each Yee cell. Because we solve the problem in the time domain, the components are only real numbers, unlike frequency responses, which use complex numbers for each component. The material properties of the cells can be indicated with short 1byte integers provided that there are no more than 256 different ones. Single-precision storage of the components requires 30 bytes for each cell; double-precision storage requires 54 bytes. A three-dimensional problem with 200 cells on a side contains 8 M cells and would need 240 Mbytes of storage for single-precision and 432 Mbytes for double-precision components. At each time step approximately 10 ﬂoating-point operations (ﬂops) are needed for each component in each cell. We must run the time steps until the input pulse has peaked and died out in each cell. This takes about 10 times the number of cells in the longest direction (maximum number along one axis). The three-dimensional problem with 200 cells on a side runs for 2000 time steps and requires 60 ﬂops times the number of cells. The solution needs 2000 × 8 M × 60 ﬂops = 960 Gﬂops for completion. 2-6.7 Excitation We specify the excitation of an antenna in the time domain since FDTD operates in the time domain. If all we need is a single-frequency solution, a ramped sinusoidal waveform can be applied. The waveform is tapered from zero in about three cycles and the FDTD solution steps continue until a steady state is reached. It is more efﬁcient to use a waveform that gives a wide-frequency-range response after performing a discrete Fourier transform on the radiating boundary to compute equivalent currents used at a given frequency. The computer storage and run times are the same for the wideband response as the single-frequency response. A suitable wide-bandwidth excitation is the differentiated Gaussian pulse shown in Figure 2-16: (t/τp )2 − 1 t (2-56) Vinc (t) = −V0 exp − τp 2 We calculate the frequency response of the differentiated Gaussian pulse from the Fourier transform of Eq. (2-56): √ (ωτp )2 − 1 (2-57) Vinc (ω) = −j ω 2π τp2 V0 exp − 2 The spectrum of Eq. (2-57) peaks for ωp = 1/τp . Figure 2-17 gives the normalized frequency response and shows that the −20-dB-level normalized frequency extends from 0.06 to 2.75. For example, if we wanted to center the frequency response at 10 GHz, the normalizing pulse time is easily found: τp =

1 = 1.592 × 10−11 s = 15.92 ps 2π(10 × 109 )

A check of Figure 2-17 shows that the antenna frequency response could be found from 2 to 22 GHz with only a 10-dB loss in dynamic range compared to the response at 10 GHz. A single time response computation yields a wide-frequency-range response.

82

RADIATION STRUCTURES AND NUMERICAL METHODS

Normalized Voltage

Gaussian Pulse

Differentiated Gaussian Pulse

Normalized Time

Amplitude, dB

FIGURE 2-16 Differentiated Gaussian pulse time response used in FDTD analysis.

Gaussian Pulse Differentiated Gaussian Pulse

Normalized Frequency

FIGURE 2-17

Differentiated Gaussian pulse normalized frequency response.

A sinusoidal modulated Gaussian pulse produces a narrow-bandwidth excitation useful in visualization because the bandwidth of the pulse can be controlled:

(t/τp )2 Vinc (t) = V0 exp − sin ω0 t 2

(2-58)

FINITE-DIFFERENCE TIME-DOMAIN METHOD

83

The unmodulated Gaussian pulse shown in Figure 2-16 has a low-pass frequency response: √ (ωτp )2 Vinc (ω) = 2πτp V0 exp − (2-59) 2 Figure 2-17 gives the low-pass frequency response of the Gaussian pulse with a −4.37dB response at ωp = 1/τp . The sinusoidal modulation centers the frequency response of the Gaussian pulse around ω0 , and the convolution of the two frequency responses produces a two-sided response of the Gaussian pulse.

2-6.8 Waveguide Horn Example [19] The literature contains solutions for the patterns of a number of antennas. Figure 2-18 shows the meshing of a commercial standard gain horn analyzed and compared to measurement. The horn operates from 8.2 to 12.4 GHz. The horn has a radiating aperture that is 110 mm wide and 79 mm high and a bell length of 228 mm. The 51-mm length of the input waveguide and the details of the feed probe were included in the model. Placing a perfectly magnetic conductor through the midsection of the horn uses symmetry to halve the number of cells to a uniform mesh of 519 × 116 × 183 Yee cells. Ten cells were used on the outside for the ABCs around the sides of the horn and 40 cells for the front ABCs in the maximum radiation direction. The model placed 20 cells between the edge of the horn and the equivalent current surface used for pattern calculations. The longest side of the grid determined the number of time steps at 10 times the number of cells = 5190 time steps. The model contains approximately 11 M Yee cells that require 330 Mbytes of computer storage. Assuming that the problem

Probe

WR-90 Waveguide

FDTD Cells Shown 2X

Coaxial Line a

b a

lw D b b

a

FIGURE 2-18 FDTD model of a standard gain horn. (From [17], Fig. 7.17, 1998 Artech House, Inc.)

84

RADIATION STRUCTURES AND NUMERICAL METHODS

(a)

(b)

FIGURE 2-19 FDTD calculated electric ﬁeld in the vertical symmetrical plane of a standard gain horn: (a) early time with a pulse in the throat; (b) pulse leaving the mouth of the horn. (From [17], Fig. 7.20, 1998 Artech House, Inc.)

takes 60 ﬂops per cell for each time step, the solution required 3.43 Tﬂops of computer calculations. The initial calculation used a differentiated Gaussian pulse excitation with τp = 15.9 ps that centered the response at 10 GHz. The calculation produced patterns that matched measurements. A second calculation used a sinusoidal modulated Gaussian pulse with the time constant 79.6 ps. This pulse time constant gives a normalized frequency of 2 GHz for the Gaussian pulse. The −3-dB frequency is 0.83 times the normalizing frequency. The pulse is centered at 10 GHz with a 3-dB bandwidth of 3.32 GHz. Figure 2-19 shows the ﬁelds when the pulse reached the horn aperture. Note the high ﬁelds in front of the horn and the amount of ﬁelds still radiating beyond and behind the aperture. By using a sinusoidal modulated pulse, the visual display contains nulls that improve its clarity. 2-7 RAY OPTICS AND THE GEOMETRIC THEORY OF DIFFRACTION Ray optics can give you a good physical feel for radiation and spur design ideas, but we need to question the accuracy of their use. Ray optics or geometric optics (GO) methods come from the design of lens and optical reﬂectors where the wavelength is very short compared to the size of the object being analyzed, whereas we may be interested in analyzing or designing an antenna on a structure only a few wavelengths in size. Below we show that GO is essentially correct over most of the radiation sphere and that by using elements of the geometric theory of diffraction [GTD (UTD)], the pattern prediction can be improved. In this case improvement means that we will increase the area of the radiation pattern that becomes more accurate. You will discover that it takes

RAY OPTICS AND THE GEOMETRIC THEORY OF DIFFRACTION

85

an increasing amount of effort to improve small areas of the pattern prediction, and at some point you should decide that it fails to give enough improvement to justify the work. Your real design purpose is to determine antenna dimensions that produce the desired antenna response. Of course, as the expense of the antenna increases, your customer may demand better predictions of the ﬁnal result, and then the cost of a better analysis is justiﬁed. You need to accept a new approach. Even though a part of the pattern prediction shows errors, obvious discontinuities, it only means that the pattern is inaccurate in directions near them and that over most of the radiation sphere the prediction is essentially correct. Discussion of this method begins with simple examples given in two-dimensional space that introduce the ideas behind GO and GTD. These examples can ignore the details of rotation of polarization directions because the waves are either polarized with the electric ﬁeld normal to the page or located in the plane of the page. We consider radiation blockage by objects, the reﬂection of rays by the objects, and the diffraction of rays around edges that ﬁlls in the pattern in the shadow regions and across the boundary of the last reﬂected ray. After the discussion of simple examples, the key points of GTD will be given for use in three-dimensional problems. This involves the rotation of coordinate systems so that ray polarizations line up with planes of incidence for reﬂections, with edges for diffraction and curvature directions on curved surfaces that shed rays around the object into the shadow. You will need to investigate the references if you want to develop your own routines, but this discussion will introduce you to the topic and give you an appreciation of the method so that you can use available computer programs and understand their limitations. GO uses ray methods to approximate electromagnetics. It is exact only in the limit of zero wavelength (inﬁnite frequency), but we gain useful insight from it at any frequency. It will not give good results close to physical boundaries; but when we include the GTD, the results are accurate down to one-wavelength sizes and are useful at λ/4 sizes. GO gives us physical insight when we deal with reﬂectors. We must consider three aspects to use GO fully: (1) ray reﬂections, (2) polarization, and (3) amplitude variations along the ray path and through reﬂections. 2-7.1 Fermat’s Principle Rays travel through a medium at the speed of light determined by the index of refrac √ tion: n = εr µr . We deﬁne the optical path length as C n dl, where C is a prescribed path in space. Fermat’s principle determines the paths of rays between two points. It states that the optical path length is stationary along a valid ray path. An expression is stationary when its ﬁrst derivatives are zero and the optical path is a minimum (or maximum). We use Fermat’s principle to trace ray paths through reﬂection or refraction by searching for the minimum optical path lengths. We can ﬁnd more than one possible ray between points because Fermat’s principle requires only a local minimum. When we exclude the boundaries of lenses, regions of homogeneous medium, rays travel in straight lines. 2-7.2 H -Plane Pattern of a Dipole Located Over a Finite Strip Figure 2-20 illustrates the geometry of this problem and the various regions of the analysis. The diagram shows the end of the dipole rod with the two rods located

86

RADIATION STRUCTURES AND NUMERICAL METHODS

I RB

RB

l/4 II

II 3l/8

5l/8 SB

SB III

FIGURE 2-20 GTD example using a two-dimensional model of a dipole located over an asymmetrical ground plane.

normal to the page. The dipole pattern is omnidirectional in the page with the electric ﬁeld directed normal to the page. When we trace rays from the dipole to the ﬁnite strip, we discover two signiﬁcant directions on both sides of the strip. The dashed boundaries labeled RB (reﬂection boundary) are the directions of the last rays reﬂected from the strip. Similarly, the dashed boundaries labeled SB (shadow boundary) are the last rays of radiation not blocked by the strip. The radiation in region I results from the sum of the direct radiation from the dipole plus the radiation reﬂected by the strip. Only direction radiation from the dipole occurs in the two parts of region II. Finally, region III is totally blocked from any radiation by a direct or reﬂected ray. This region receives rays diffracted around the edges. If we add the direct and reﬂected rays in an analysis, we obtain the pattern given in Figure 2-21, which also traces the actual pattern. The pattern, using only the direct and reﬂected rays, accounts for the phasing between the direct radiation from the dipole and an image dipole located below the strip. If you compare the two traces on Figure 2-21, you see that the two patterns are similar near θ = 0, but the direct plus reﬂected ray pattern has discontinuities at the SBs and RBs. Figure 2-22 gives the results for the same analysis, but using a 5λ-wide ground-plane strip. When using the larger strip, the two patterns match to about 80◦ , and in the second case the simple analysis is correct over most of the forward semicircle. Simple geometric optics gives good results for large objects provided that you realize the patterns will contain discontinuities. Removing the discontinuities requires extra effort. A discontinuity in the pattern cannot exist because shadow and reﬂection boundaries occur in free space. It takes a material boundary to produce a discontinuous ﬁeld. But, for example, the tangential electric ﬁeld must be continuous across even material boundaries. Edge diffraction solves the discontinuity problem. Figure 2-23 gives the pattern of the edge diffraction for both edges normalized to the total pattern. The edge diffraction has matching discontinuities to the sum of the direct and reﬂected rays at the SBs and RBs. The UTD (uniform theory of diffraction) technique [20, p. 55] calculated these diffractions. When these diffractions are added to the direct and reﬂected ray radiation, the total pattern given in Figure 2-21 is obtained. The dipole, its image in the ground plane, and the two edge diffractions form a four-element array where each element has a unique pattern. Adding edge diffractions to the geometric optics ﬁelds removes the discontinuities and allows calculation of the pattern behind the strip ground plane.

RAY OPTICS AND THE GEOMETRIC THEORY OF DIFFRACTION

87

Total Pattern

Direct + Reflected

FIGURE 2-21 H -plane pattern of a dipole over asymmetrical ground using direct and reﬂected rays compared only to a full solution for the 1λ ground plane of Figure 2-20.

Total

Direct + Reflected

FIGURE 2-22 H -plane pattern of a dipole over symmetrical ground using direct and reﬂected rays compared only to a full solution for a 5λ ground plane.

2-7.3 E -Plane Pattern of a Rectangular Horn Figure 2-24 illustrates the cross section of a horn or, in this case, a two-dimensional approximation to a horn. The waveguide feeds the horn and produces a uniform aperture distribution in the E-plane. In this model the direct GO radiation is a constant wedge signal as shown in Figure 2-25 ranging between −15◦ and +15◦ . The reﬂected pattern

88

RADIATION STRUCTURES AND NUMERICAL METHODS

FIGURE 2-23 GTD edge diffraction of an H -plane pattern for an asymmetrical 1λ ground plane under a dipole.

combines with the direct radiation and produces the same pattern. Figure 2-25 also shows the diffraction patterns from the two edges. These peak along the plates and exhibit a discontinuity at the same angle as the GO ﬁeld. Each diffraction pattern has a discontinuity on one side at 90◦ because the mouth of the horn blocks the diffraction from the opposite edge. When we add the diffracted ﬁelds to the GO ﬁeld, the pattern shown in Figure 2-26 is obtained. By just adding the three components, we obtain an accurate pattern of the horn over most of the angles of the plot. At 90◦ we see discontinuities in the pattern caused by not considering enough terms in the GTD calculation. You need to realize that these discontinuities only cause pattern errors at nearby angles. The majority of the pattern is correct. We need another term to correct the pattern near 90◦ . The blockage of the diffraction from one edge by the mouth of the horn causes a secondary diffraction at that edge. We call this double diffraction. Some available programs do not implement double diffraction because the general three-dimensional double diffraction takes considerable calculation due to the extensive ray tracing required. In these cases you must accept the pattern discontinuities. Some programs calculate double diffraction as an option, but turning on this option will slow the calculations. Figure 2-27 gives the pattern when double diffraction is included. Double diffraction reduces the discontinuity at 90◦ , but a small discontinuity remains. Adding triple diffraction would reduce this further, but the pattern area affected by the small discontinuity has shrunk. A new discontinuity near 60◦ appeared in the pattern after adding double diffraction at the

RAY OPTICS AND THE GEOMETRIC THEORY OF DIFFRACTION

89

FIGURE 2-24 Geometry of a two-dimensional model of a rectangular horn used for GTD analysis.

mouth of the horn. We could continue to add another term to remove this one or just accept it. 2-7.4 H -Plane Pattern of a Rectangular Horn The tangential electric ﬁelds vanish at the walls of the two-dimensional horn in the H -plane. This affects the GO ﬁeld and produces the following equation for them: E GO = cos

π tan θ e−j kR √ 2 tan α R

(2-60)

Equation (2-60) includes the phasing term and square-root spreading factor of a twodimensional ﬁeld. The horn walls tilt from the centerline by the angle α. Figure 2-28 plots the GO ﬁeld and shows that it vanishes at the walls. We do not expect edge diffraction because the ﬁeld vanishes at the edges, but Figure 2-28 shows diffraction patterns that peak in the direction of the walls. We call this new term slope diffraction. This new type requires another set of coefﬁcients not identical to the edge (or wedge) diffraction coefﬁcients. While the amplitude of the edge diffraction is proportion to the ﬁeld incident on the edge, the amplitude of slope diffraction is proportional to the derivative of the ﬁeld in the direction normal

90

RADIATION STRUCTURES AND NUMERICAL METHODS

FIGURE 2-25 E-plane pattern of a rectangular horn with a GO term (solid curve) and edge diffractions (dashed curves).

FIGURE 2-26 Combination of GO and edge diffractions in the E-plane pattern of a rectangular horn.

to the edge. We ﬁgure the same geometric factors for both edge and slope diffraction but now must calculate the normal derivative of the incident electric ﬁeld. Figure 2-29 plots the H -plane pattern of the horn. The pattern fails to predict a pattern behind it. The E-plane diffraction produces a back hemisphere pattern for a real horn, but our two-dimensional model does not include the E-plane. 2-7.5 Amplitude Variations Along a Ray Power decreases in a general ray as the distance from the source increases. If we expand the constant-phase surface (eikonal) about the ray in a Taylor series, we obtain a surface described by its radii of curvature [20, p. 55]. The maximum and minimum

RAY OPTICS AND THE GEOMETRIC THEORY OF DIFFRACTION

91

FIGURE 2-27 E-plane pattern of a rectangular horn combining GTD terms of direct GO, edge diffractions, and double diffractions between edges.

FIGURE 2-28 H -plane pattern of a rectangular horn with a GO term (solid curve) and edge slope diffractions (dashed curves).

values lie in the orthogonal principal planes. These radii of curvature determine the amplitude spread of the wave from point to point on the ray. We compute the ratio of differential areas about the ray at two locations as dA2 ρ1 ρ2 = dA1 (ρ1 + d)(ρ2 + d)

(2-61)

92

RADIATION STRUCTURES AND NUMERICAL METHODS

FIGURE 2-29 H -plane pattern of a rectangular horn by GTD analysis by combining direct GO ﬁeld and edge slope diffraction.

r1 dA1 r2

dA2

d

Principal Planes S

FIGURE 2-30 Astigmatic ray.

where ρ1 and ρ2 are the principal radii of curvature and d is the distance between two points on the ray (Figure 2-30). The electric ﬁeld variation along the ray becomes E0 e

−j kd

ρ1 ρ2 (ρ1 + d)(ρ2 + d)

(2-62)

for the astigmatic ray spreading from unequal radii of curvature. When d = −ρ1 or d = −ρ2 , GO fails because it predicts an inﬁnite power density. We call these locations caustics. Remember that the ray always has differential area and never has any real area as implied by Figure 2-30. We have three special cases of the astigmatic ray: 1. Spherical wave, ρ1 = ρ: E0 e−j kd

ρ ρ+d

(2-63)

93

RAY OPTICS AND THE GEOMETRIC THEORY OF DIFFRACTION

2. Cylindrical wave, ρ1 = ∞: E0 e−j kd 3. Plane wave, ρ1 = ρ2 = ∞:

ρ ρ+d

E0 e−j kd

(2-64)

(2-65)

The plane wave does not spread but has constant amplitude as distance changes. Both cylindrical and plane waves require inﬁnite power, and they are therefore nonphysical, but we ﬁnd them convenient mathematically. 2-7.6 Extra Phase Shift Through Caustics We cannot determine the ray amplitude at a caustic but can determine its amplitude and phase on either side. Passage through a caustic causes an extra phase shift to the ray [21, p. 31]. The denominator factors in the square root of Eq. (2-62) produce a 180◦ sign change when the ray distance factor d passes through either ρ1 or ρ2 . The square root changes 180◦ to +90◦ (ej π/2 ) or −90◦ (e−j π/2 ), depending on the direction of movement along the ray. When tracing a ray moving through a caustic in the direction of propagation, you multiply by ej π/2 . The ﬁeld is multiplied by e−j π/2 for a ray traced in the opposite direction of propagation. 2-7.7 Snell’s Laws and Reﬂection We derive Snell’s laws of reﬂection and refraction from Fermat’s principle. The two laws of reﬂection are given as: 1. The incident ray, the reﬂected ray, and the normal of the reﬂecting surface at the point of reﬂection lie in the same plane. 2. The incident and reﬂected rays make equal angles with the surface normal. Implicit in Snell’s laws is the idea that locally the wavefront behaves like a plane wave and that the reﬂector can be treated as a plane. Given the direction of the incident ray S1 , reﬂected ray S2 , and the reﬂector normal n, Snell’s laws of reﬂection can be expressed vectorially [22]: n×(S2 − S1 ) = 0

n·(S1 + S1 ) = 0

(2-66)

We combine Eq. (2-66) to determine the ray directions before or after reﬂection: S1 = S2 − 2(S2 · n)n

S2 = S1 − 2(S1 · n)n

(2-67)

Snell’s law of refraction can also be expressed vectorially as n×(n2 S2 − n1 S1 ) = 0 where n1 and n2 are the index of refractions in the two mediums.

(2-68)

94

RADIATION STRUCTURES AND NUMERICAL METHODS

2-7.8 Polarization Effects in Reﬂections The electric ﬁeld is orthogonal to the ray direction (a free-space wave) and is described by a two-dimensional polarization space (Section 1-11). We can describe polarization in any conveniently rotated two-dimensional basis vectors in the plane with the ray vector as its normal. We will use a ray-ﬁxed coordinate system that changes direction after a reﬂection: (2-69) Ei = ai|| Ei|| + ai⊥ Ei⊥ where ai|| is a unit vector in the plane of incidence and ai⊥ is perpendicular to the plane of incidence. We compute ai⊥ from the normal to the plane n at the reﬂection point and the incident ray unit vector Si : ai⊥ =

Si × n |Si × n|

ai||

ai⊥

=

(2-70)

× Si

After reﬂection, we calculate the output ray-ﬁxed polarization vectors using the output ray Sr : ar⊥ = ai⊥ and ar|| = ar⊥ × Sr Ei|| is the incident electric ﬁeld in the direction of ai|| and Ei⊥ is in the direction of ai⊥ . Of course, the direction of unit vector a|| changes from incident to reﬂected rays. The electric ﬁeld parallel to the reﬂector surface must vanish on the conductor surface: Er⊥ = −Ei⊥

(2-71)

where Er⊥ is the reﬂected ﬁeld along ai⊥ . We calculate the reﬂection properties of E|| from the corresponding magnetic ﬁelds parallel to the surface: Hr|| = Hi||

(2-72)

By combining Eqs. (2-71) and (2-72), we obtain the dyadic relation for the ray-ﬁxed coordinate system: Ei|| 1 0 Er|| = (2-73) 0 −1 Er⊥ Ei⊥ where Er|| and Ei⊥ are the reﬂected ﬁeld components. At each reﬂection we rotate the polarizations to align ai⊥ with the normal to the plane of incidence. We can express Eq. (2-73) as a dyadic in terms of the incident and reﬂected wave polarization vectors = ai ar − ai ar . Of course, the alternative method is to describe polarizations in a R || || ⊥ ⊥ ﬁxed three-dimensional coordinate system, but it requires a 3 × 3 reﬂection matrix. 2-7.9 Reﬂection from a Curved Surface A wave reﬂected from a curved surface changes its radii of curvature and principal planes. The ﬁeld along the reﬂected ray is given by ρ1r ρ2r Er (s) = Ei0 ·R e−j ks (2-74) r (ρ1 + s)(ρ2r + s)

RAY OPTICS AND THE GEOMETRIC THEORY OF DIFFRACTION

95

where s is the distance along the ray from the reﬂection, ρ1 and ρ2 the reﬂected ray the reﬂection dyadic. Ei0 the incident ray electric ﬁeld. For radii of curvature, and R a ﬂat surface we use images of the incident ray caustics for ρ1r and ρ2r , but in general, ρ1r and ρ2r become 1 1 = ρ1r 2

1 1 + i ρ1i ρ2

+

1 1 = ρ2r 2

1 f1

1 1 + i ρ1i ρ2

+

1 f2

(2-75)

where f1 and f2 are generalized focal lengths of the surface. The spreading factor of Eq. (2-74) simpliﬁes in the far ﬁeld:

ρ1r ρ2r ≈ r (ρ1 + s)(ρ2r + s)

r r ρ1 ρ2 s

Kouyoumjian and Pathak [23] derived formulas for the focal lengths of a surface. We start with a surface with principal radii of curvature R1 and R2 with directions u1 and u2 at the point of reﬂection. For an incident ray with principal axes deﬁned by unit vectors xi1 and xi2 , we deﬁne a matrix relation between the incident ray and surface principal curvature directions:

xi · u θ = 1i 1 x2 · u1

xi1 · u2 xi2 · u2

(2-76)

where the determinant is |θ | = (xi1 · u1 )(xi2 · u2 ) − (xi2 · u1 )(xi1 · u2 ). Given the angle of incidence θ i , the following are the focal lengths: 2 2 2 2 θ21 + θ12 + θ11 1 cos θ i θ22 = + f1,2 |θ |2 R1 R2 2 2 2 2 2 θ21 − θ12 − θ11 1 1 1 4 cos θ i θ22 1 1 − i + − i + ± 2 |θ |2 R1 R2 ρ1i ρ2 ρ1i ρ2

1/2 2 2 2 2 2 θ21 + θ12 + θ11 θ22 4 cos2 θ i 4|θ |2 + + − (2-77) |θ |4 R1 R2 R1 R2 With a single reﬂection, we need not compute the direction of the principal axes. We need only the focal lengths. Multiple reﬂections require knowledge of the reﬂected-ray principal plane directions. Deﬁne the following matrices to determine the directions of the principal axes after reﬂection:

1 ρ1i Qi0 = 0

0 1 ρ2i

1 C0 = R1 0

0

1 R2

Qr = Qi0 + 2(θ −1 )T C0 θ −1 cos θi br1 = xi1 − 2(n · xi1 )n

br2 = xi2 − 2(n · xi2 )n

96

RADIATION STRUCTURES AND NUMERICAL METHODS

where n is the surface normal at the reﬂection point. One principal-axis direction is xr1

r Q22 − 1/ρ1r br1 − Qr12 br2 = 2 Qr22 − 1/ρ1r + (Qr12 )2

(2-78)

We derive the other from the cross product of Eq. (2-78) and the reﬂected ray unit vector: xr2 = −Sr × xr1 (2-79) We must reapply Eqs. (2-75) through (2-79) for every reﬂection. We use Eqs. (2-75) through (2-79) for analysis, but except for computer optimizations, they cannot be applied directly to synthesis. If we limit the reﬂectors to ﬁgures of rotation, the radii of curvature are given by the meridians and parallels and these problems reduce to two dimensions. Similarly, a cylindrical reﬂector fed with a cylindrical wave [Eq. (2-64)] reduces the problem to two dimensions. The incident and reﬂected waves remain in the single plane chosen for the reﬂector analysis. 2-7.10 Ray Tracing Tracing rays through a reﬂector system is conceptually straightforward. Where a ray strikes a reﬂector, we compute the normal to the surface. By using Eq. (2-67), we solve for the reﬂected-ray direction. Equation (2-73) determines the polarization effects when we express the incident and reﬂected rays in the ray-ﬁxed coordinates. We use geometric arguments to determine the amplitude variation along the ray through the reﬂection instead of the general expressions given above. We experience difﬁculty when we try to discover the reﬂection points for given ﬁeld and source points. No analytical expressions exist for calculating the reﬂection point of a general surface. The usual computer routines search for the minimum optical path length (Fermat’s principle) without using Eq. (2-67), since a local minimum will satisfy this equation. 2-7.11 Edge Diffraction Keller [24] extended the idea of reﬂection to edge diffraction by applying a generalized Fermat’s principle to the rays. Figure 2-31 illustrates the rays in edge diffraction and the associated polarization directions. The ﬁgure shows the edge vector at the diffraction point. The vector cross product between the edge vector and the incident ray points in the direction of the incident plane normal. We measure the diffraction angle of incidence in this plane between the incident ray and the edge normal. Because diffraction obeys a generalized Fermat’s principle, the diffracted ray exits at the same angle, similar to the reﬂected ray angles. The diffracted rays lie in a cone with the edge vector as its axis. The diffracted rays spread the incident power into a cone. Figure 2-31 shows a particular diffracted ray and how we determine the diffracted ray exiting plane. We deﬁne diffracted ray polarization in terms of the incident and diffracted planes. The vectors are parallel and perpendicular to the two planes. Given the edge unit vector e, you compute the incident perpendicular polarization vector: aφ =

e × S sin β0

(2-80)

RAY OPTICS AND THE GEOMETRIC THEORY OF DIFFRACTION

97

e^

s^

Efd

Ed

Plane of ^ ^ Diffraction (s, e)

b0

QE

E if′

Edge-Fixed Plane of Incidence (s^′, e^)

E ib′0

b0

^ s′

Edge (a)

f

^

f

^

f′

^

n

Re f Bo lectio und n ary

f′ O

QE w ado y Sh undar Bo

Plane ⊥ e^ at QE

np

(b)

FIGURE 2-31 Ray-ﬁxed coordinates related to edge-ﬁxed coordinates at the edge diffraction point on a curved edge by showing planes of incidence and diffraction. (From [25], Fig. 5, 1974 IEEE.)

where S is the incident ray and β0 is the angle between the edge tangent and the incident ray. The diffracted ray perpendicular polarization is similar to the incident ray aφ = −

e×S sin β0

(2-81)

where S is the diffracted ray unit vector. We have the following vector relations for diffraction: |e × S| = |e × S | and e · S = e · S (2-82)

98

RADIATION STRUCTURES AND NUMERICAL METHODS

We determine the parallel polarization vector along the ray-ﬁxed coordinates by the following cross products: aφ × aβ0 = S

and aφ × aβ0 = S

(2-83)

By using ray-ﬁxed coordinates, the diffraction matrix reduces to 2 × 2. When β0 = π/2, the parallel polarization components are parallel to the edge and the electric ﬁeld vanishes: Eβ0 + Eβ0 = 0. Acoustics calls this the soft boundary condition (Dirichlet); it operates on the parallel polarization components. The perpendicular components satisfy the hard boundary condition (Neumann). At a diffraction point Qe we describe diffraction by the matrix equation

i d Eβ (Qe ) Eβ0 (s) ρ −Ds 0 0 e−j ks (2-84) = i 0 −Dh Eφd (s) s(s + ρ) Eφ (Qe ) where s is the distance from the diffraction point. Diffraction locates one caustic on the diffraction point. We compute the second caustic distance ρ from the incident ray radius of curvature in the plane of incidence ρei and the edge curvature unit vector nˆ e : 1 nˆ e · (ˆs − sˆ ) 1 = i − ρ ρe a sin2 β0

(2-85)

where a is the edge radius of curvature. When a → ∞ (straight edge), the second term of Eq. (2-85) vanishes. A number of factors determine the wedge diffraction coefﬁcients. The diffracting edge factors include (1) the angle between the faces, (2) the edge curvature, and (3) the curvature of the faces. The ray angle factors are (1) the incident angle relative to the edge tangent, (2) the diffraction angle to the shadow boundary, and (3) the angle to the reﬂection boundary. The diffraction coefﬁcients peak at the shadow and reﬂection boundaries. UTD formulation uses characteristic lengths associated with incident and diffracted ray caustics. These many factors are beyond the current discussion. 2-7.12 Slope Diffraction The spatial rate of change of the ﬁeld normal to the edge produces slope diffraction, an added ﬁeld component. This ray optics term also satisﬁes the generalized Fermat’s principle with geometry determined by Eqs. (2-80) through (2-83), and (2-85). The slope diffraction equation has the same form as Eq. (2-84):

d i Eβ (Qe ) Eβ0 (s) ρ −es 0 0 = (2-86) e−j ks 0 −eh Eφd (s) s(s + ρ) Eφi (Qe ) where the diffraction coefﬁcients es,h are related to the ﬁeld derivative normal to the surface: 1 ∂Ds,h ∂ es,h = (2-87) j k sin β0 ∂φ ∂n The term ∂/∂n of Eq. (2-87) indicates the derivative of the incident ﬁelds given in the vector of Eq. (2-86). Equation (2-87) has the term ∂Ds,h /dφ for the soft and hard slope diffraction terms returned from a subroutine; it is only a notational derivative.

RAY OPTICS AND THE GEOMETRIC THEORY OF DIFFRACTION

99

2-7.13 Corner Diffraction Every structural discontinuity diffracts waves. We derive edge diffraction from an inﬁnite wedge where the wedge terminations (corners) produce diffracted rays. Recall from Section 2-4.2 that PTD added currents at edges to handle the effect of not having an inﬁnite surface; the formulation for corner diffraction uses equivalent currents to derive these coefﬁcients. We handle edge diffraction from each edge as always. Since each corner arises from two edges, we compute separate corner diffraction for each edge, two terms per corner. Whereas edge diffraction is bound to a cone, corner diffraction radiates in all directions. The edge must be visible from both the source and receive points before corner diffraction contributes. We must include corner diffraction in any three-dimensional problem. As the source and receiver become farther and farther away from the object, corner diffraction contributions dominate over edge diffractions since it is derived from equivalent currents. 2-7.14 Equivalent Currents GTD fails to predict ﬁelds at caustics. In many cases we consider these points unimportant, but for those cases where we need the ﬁelds, equivalent currents provide the answer. We derive equivalent currents from edge diffraction, which then replaces it and we use them instead of edge diffraction for all pattern points. The use of currents reduces the problem to a PO solution and line integrals are required. We relate the incident ﬁelds expressed in the ray ﬁxed to equivalent currents: √ 2j i Eβ Ds 2πk ej π/4 ηk 0 √ 2j i Eφ Dh 2πk ej π/4 M= 0 k I =

(2-88) (2-89)

The soft and hard diffraction coefﬁcients Ds,h depend on the source and receiver positions. Since we calculate the ﬁelds using vector potentials or dyadic Green’s functions, the formulation has no caustics. They are only associated with a geometric optics solution. Equivalent currents allow the calculation of the ﬁelds directly behind a reﬂector near the axis. The GTD solution produces a caustic as all points along the rim “light up” for an axisymmetrical design. PTD uses equivalent currents in a similar but different way to calculate correct ﬁelds in the same region. Equivalent currents derived from the diffraction coefﬁcients produce the entire solution, since the reﬂector blocks the incident ﬁeld. In PO we continue to include the direct ﬁeld and the induced current radiation on the reﬂector, but add the PTD current radiation. Realize that slope diffraction also adds to the equivalent currents. 2-7.15 Diffraction from Curved Surfaces [26, 27] In one analytical approach to surface-wave radiation we postulate waves bound to a surface that radiate only from discontinuities. Surface waves on inﬁnite structure do not radiate but attenuate exponentially away from the surface, because they are bound to it. We can formulate GTD as radiating from discontinuities, and this produces an approach

100

RADIATION STRUCTURES AND NUMERICAL METHODS

for ﬁelds radiated on the shadowed side of a curved body. The continuous discontinuity of the curved surface causes power to be radiated at every point in the shadow region. These waves radiate tangentially from a wave traveling along a geodesic and bound to the surface. Surface waves require a dielectric coating or a corrugated surface to slow and bind the wave to the surface. The surface curvature slows and binds the wave to the surface without the need for a dielectric or corrugated surface coating. The wave that propagates along the surface sheds power in rays tangentially to it. The rays travel along a surface geodesic from the attachment point to the radiation point. The geodesic curve is a minimum distance path on the surface between two points. In differential geometry it has a broader meaning, but for our purpose, the minimum distance deﬁnition will serve. The curved surface diffraction satisﬁes a generalized Fermat’s principle (minimum distance) as do all other terms of GTD. The best approach uses another ray-ﬁxed coordinate along the surface where the vectors are normal and tangential to the surface at both the attachment and radiation (shedding) points. Curved surface diffraction considers three types of problems with different formulations. Two of them start with an antenna mounted on the surface. We either calculate the pattern in the presence of the curved object or calculate the coupling to a second antenna also mounted on the curved object. The third case determines the ﬁeld scattered for a source located off the surface. All three use the ray-ﬁxed coordinates. We start with the surface normal nˆ and the tangent vector ˆt directed along the geodesic path. A vector cross product deﬁnes the third direction of the local coordinate system. We ˆ and the three vectors form a triad: nˆ × bˆ = ˆt. On a general use the surface binormal b, surface all three vectors change direction as the wave moves along the geodesic. We use the term torsion for a path with a changing binormal. A soft dyadic diffraction coefﬁcient is used with ﬁelds aligned with the attachment point binormal and the tangential shedding point binormal. We apply the hard dyadic diffraction coefﬁcient ﬁelds aligned along the normal vectors. No formulas exist for computing the attachment and shedding points on a general curved surface given the source and receive points. We usually start with a known diffraction and ﬁnd other points by incrementing along the curve by small steps.

REFERENCES 1. L. Diaz and T. A. Milligan, Antenna Engineering Using Physical Optics, Artech House, Boston, 1996. 2. B. F. Harrington, Time-Harmonic Electromagnetic Fields, McGraw-Hill, New York, 1961. 3. P. S. Hacker and H. E. Schrank, Range distance requirements for measuring low and ultralow sidelobe antenna patterns, IEEE Transactions on Antennas and Propagation, vol. AP-30, no. 5, September 1982, pp. 956–966. 4. K. Pontoppidan, ed., Technical Description of Grasp 8, Ticra, Copenhagen, 2000 (selfpublished and available at www.ticra.com). 5. J. H. Richmond, A reaction theorem and its application to antenna impedance calculations, IRE Transactions on Antennas and Propagation, vol. AP-9, no. 6, November 1961, pp. 515–520. 6. R. F. Harrington, Field Computation by Moment Methods, Macmillan, New York, 1968; reprinted by IEEE Press, New York, 1993.

REFERENCES

101

7. J. H. Richmond, Radiation and scattering by thin-wire structures in homogeneous conducting medium, IEEE Transactions on Antennas and Propagation, vol. AP-22, no. 2, March 1974, p. 365 (see also ASAP wire code). 8. N. N. Wang, J. H. Richmond, and M. C. Gilreath, Sinusoidal reactance formulation for radiation from conducting structures, IEEE Transactions on Antennas and Propagation, vol. AP-23, no. 3, May 1975, pp. 376–382. 9. B. M. Kolundzija and A. R. Djordjevic, Electromagnetic Modeling of Composite Metallic and Dielectric Structures, Artech House, Boston, 2002. 10. A. R. Djordjevic et al., AWAS for Windows Version 2.0: Analysis of Wire Antennas and Scatterers, Artech House, Boston, 2002. 11. A. C. Ludwig, Wire grid modeling of surface, IEEE Transactions on Antennas and Propagation, vol. AP-35, no. 9, September 1987, pp. 1045–1048. 12. A. W. Glisson and D. R. Wilton, Simple and efﬁcient numerical methods for problems of electromagnetic radiation and scattering from surfaces, IEEE Transactions on Antennas and Propagation, vol. AP-28, no. 5, September 1980, pp. 563–603. 13. S. M. Rao, D. R. Wilton, and A. W. Glisson, Electromagnetic scattering by surfaces of arbitrary shape, IEEE Transactions on Antennas and Propagation, vol. AP-30, no. 3, May 1982, pp. 409–418. 14. B. M. Kolundzija et al., WIPL-D: Electromagnetic Modeling of Composite Metallic and Dielectric Structures, Artech House, Boston, 2001. 15. P. -S. Kildal, Foundations of Antennas, Studentlitteratur, Lund, Sweden, 2000. 16. C. Scott, The Spectral Domain Method in Electromagnetics, Artech House, Boston, 1989. 17. K. S. Yee, Numerical solution of initial boundary value problems involving Maxwell’s equations in isotropic media, IEEE Transactions on Antennas and Propagation, vol. 14, no. 3, May 1966, pp. 302–307. 18. K. S. Kunz and R. J. Luebbers, The Finite Difference Time Domain Method for Electromagnetics, CRC Press, Boca Raton, FL, 1993. 19. J. G. Maloney and G. S. Smith, Modeling of antennas, Chapter 7 in A. Taﬂove, ed., Advances in Computational Electrodynamics: The Finite-Difference Time-Domain Method, Artech House, Boston, 1998. 20. D. J. Struik, Differential Geometry, Addison-Wesley, Reading, MA, 1950. 21. D. A. McNamara, C. W. I. Pistorius, and J. A. G. Malherbe, Introduction to the Uniform Geometrical Theory of Diffraction, Artech House, Boston, 1990. 22. F. S. Holt, in R. E. Collin and F. J. Zucker, eds., Antenna Theory, Part 2, McGraw-Hill, New York, 1969. 23. R. Kouyoumjian and P. Pathak, The dyadic diffraction coefﬁcient for a curved edge, NASA CR-2401, June 1974. 24. J. B. Keller, Geometrical theory of diffraction, Journal of the Optical Society of America, vol. 52, 1962, pp. 116–130. 25. R. G. Kouyoumjian and P. H. Pathak, A uniform geometrical theory of diffraction for an edge in a perfectly conducting surface, Proceedings of IEEE, vol. 62, no. 11, November 1974, pp. 1448–1461. 26. P. H. Pathak, W. D. Burnside, and R. J. Marhefka, A uniform GTD analysis of the diffraction of electromagnetic waves by a smooth convex surface, IEEE Transactions on Antennas and Propagation, vol. AP-28, no. 5, September 1980. 27. P. H. Pathak, N. Wang, W. D. Burnside, and R. G. Kouyoumjian, A uniform GTD solution for the radiation from sources on a convex surface, IEEE Transactions on Antennas and Propagation, vol. AP-29, no. 4, July 1981.

3 ARRAYS

We begin with arrays of antennas before discussing particular antenna elements to show the relationship between antenna size and shape and the resulting pattern characteristics. We ignore the feed network design initially and assume that the proper array feed distribution will be obtained. At ﬁrst, we assume a distribution of point sources and compute the approximate array pattern. Working with simple models provides insight rather than accuracy, and later we consider element pattern and interaction. In Chapter 12 we discuss feed network design and analysis in the discussion of phased arrays. The chapter begins with a mathematic description of an array and gives various assumptions used to simplify the expressions. We analyze a simple two-element array to gain insight into the radiation phenomenon and how far-ﬁeld patterns can be found with simple arguments. The discussion of a uniformly spaced linear array shows the Fourier series relationship between array layout and the pattern space given in sin(angle) space. The principal idea is that pattern beamwidth shrinks as the array length increases. If we space the elements too far apart, multiple beam peaks or grating lobes form in the pattern, and we show how to control these grating lobes and their relationship to maximum scan angle, array layout, and element spacing. Phased arrays scan the beam by controlling the relative phasing between the elements. We extend the linear array to planar layouts that produce narrow beams in both principal planes. The planar array design is unchanged from the methods for linear arrays, but the grating lobe analysis shows their unique properties, as they sometimes form outside the plane of scan. We can divide the phased array into pieces to form multiple scanning beams, but the beam shape is determined by the segment size and shape used for each beam. By adding amplitude control the phased array can form multiple beams with beamwidths determined by the entire size of the array. Each element in an array receives a portion of the power radiated by the other elements on transmittal, or scatters power into neighboring elements in reception. The Modern Antenna Design, Second Edition, By Thomas A. Milligan Copyright 2005 John Wiley & Sons, Inc.

102

ARRAYS

103

radiation from each antenna excites currents on its neighboring elements that also radiate, and we associate the total pattern with the antenna input. In an array the effective element patterns change due to this scattering. Because of reciprocity, which says that transmit and received patterns are identical, we can analyze the problem either way. This leads to mutual coupling, which we describe and analyze by mutual impedance (admittance, or scattering) matrices. This phenomenon causes the input impedance of the elements to change as we scan the array. The mutual coupling can lead to scan blindness when the feed reﬂection coefﬁcient grows due to mutual coupling, and the array totally reﬂects the signal into the feed network. If we want the exact pattern designed for, we must compensate the feeding coefﬁcients for the mutual coupling. A discussion of array gain gives two methods of calculation. First, the effective area and the associated gain of a planar array cannot exceed its area when we include the extra half-element spacing area provided by the edge elements. When we space the elements so that their individual effective areas no longer overlap, array gain is the element gain multiplied by the number of elements. We can calculate gain by adding up the input power instead of integrating the pattern to compute total radiated power. We relate input power of elements to the self- and mutual resistances to determine gain of linear and planar arrays using realistic elements. The chapter ends with a discussion of three-dimensional arrays using arbitrarily oriented elements. We add this analysis to the simple array formula to handle the polarization of rotated antennas. Related to this problem is the pointing of an antenna on a positioner. We apply rotation matrices to both problems. An array radiates or receives from two or more antennas at the same frequency. To calculate the ﬁeld radiated from arrays we add the electric ﬁelds radiated from each element. The amplitudes and phases of each antenna, determined by the feed network, give us extra degrees of freedom to shape the pattern and design shifts from radiating elements to the feed network. A single antenna radiates an electric ﬁeld with both polarization components: E = Eθ (θ, φ)θˆ + Eφ (θ, φ)φˆ where Eθ and Eφ are the two complex components (amplitude and phase) referred to some point on the antenna. If we move the antenna or the phase reference point, we only change the antenna radiated phase. We assume that the movement is small enough that the radiation approximation can still be used. Given r as the location of the antenna relative to the phase reference point, the added phase component is ej k·r , where k = 2π/λ(sin θ cos φ xˆ + sin θ sin φ yˆ + cos θ zˆ ) and r = x xˆ + y yˆ + z zˆ is the location of the antenna; k · r is the phase distance from the antenna to the reference plane through the reference point and is deﬁned by the radiation (receiving) direction. The electric ﬁeld radiated from the moved antenna becomes ˆ j k·r [Eθ (θ, φ)θˆ + Eφ (θ, φ)φ]e We assume that nearby objects do not alter the patterns in the movement, but we can alter element patterns if necessary. Suppose that we have an array of antennas located at points r 1 , r 2 , and so on. We obtain the total pattern by adding the electric ﬁelds radiated from each: E=

N i=1

ˆ j k·r i [Eθi (θ, φ)θˆ + Eφi (θ, φ)φ]e

(3-1)

104

ARRAYS

Bringing the antennas close together will change the patterns of each because every antenna will block the radiation of the others and the distribution of currents on the elements may be changed. The shape of small resonant antennas limits the possible distribution of currents, but the magnitude and phase may be changed due to the coupling. We make various approximations to Eq. (3-1). Changes in the patterns due to nearby antennas are ignored, and isolated element patterns are used. We assume initially a certain amplitude and phase distribution on the elements and ignore the problem of the feed network. Polarization reduces to a single term for equally polarized elements, such as dipoles, slots, or horns. If the antennas have identical element patterns, we can separate Eq. (3-1) into a product. ˆ E = [Eθ (θ, φ)θˆ + Eφ (θ, φ)φ] Ei ej k·r i (3-2) where Eθ and Eφ are the normalized patterns of the single element. Ei is the electric ﬁeld of the ith element, including the amplitude and phase of the feed distribution. Equation (3-2) describes pattern multiplication that separates the pattern into an element pattern and an array factor. The method requires that all antennas have the same pattern and be orientated in the same direction. The array factor represents the pattern from an array of isotropic pattern antennas. Because array factors can be calculated by hand, we ﬁnd them useful for gaining insight. We leave calculations using Eqs. (31) and (3-2) to the computer. The element patterns themselves could be arrays and we could use pattern multiplication to synthesize planar and volumetric arrays from linear arrays.

3-1 TWO-ELEMENT ARRAY Consider two elements lying on the z-axis and spaced a distance d centered on the origin (Figure 3-1). If we rotate the isotropic pattern antennas around the z-axis, the problem remains unchanged, which means that all great-circle (constant φ) patterns are identical. On the z-axis, the element phase constant becomes ej kz cos θ . For simple line arrays we can locate pattern nulls and peaks by simple arguments. Example Two elements are spaced λ/2 and have equal amplitudes and phases. Locate the nulls and peaks. The phase reference planes can be placed at any convenient point. Consider the pattern at θ = 90◦ . We place the reference plane through the axis of the array. The added phase factor is zero for both elements and we just add components. The equal element phases add to give a beam peak. If we place a second reference plane through the top element, the wave radiated from the bottom element travels across the array λ/2 to the reference plane. Increasing the distance propagated decreases phase and it changes by −180◦ . The two out-of-phase signals cancel to produce a pattern null. The array has symmetry about the x –y plane, which means that the array will have the same pattern above and below the symmetry plane. We denote this conﬁguration an even-mode array. Figure 3-2 plots this pattern with a solid line. You should repeat the example for an odd-mode array (phases 0◦ and 180◦ ) and convince yourself that the null occurs at θ = 90◦ and the beam peak occurs at θ = 0◦ (180◦ ), plotted in

TWO-ELEMENT ARRAY

105

z d/2 d cos q 2 Radiation

q

Reference plane

−d cos q 2 −d/2

FIGURE 3-1

Two-element array on a z-axis.

Figure 3-2 as a short-dashed curve. The solid and short-dashed curves have the same directivity. Example Suppose that the two elements are spaced λ/4, with the top element phase −90◦ and the bottom element phase 0◦ . Locate the beam peak and pattern null. We start by placing a reference plane through the top element. The wave radiated from the bottom element travels across the array, and its phase decreases by 90◦ . Both radiated waves have the same phase (−90◦ ) at the reference plane and add in phase for a beam peak. Consider a second plane through the bottom element. The wave from the top element loses 90◦ propagating across the array and the two waves are 180◦ out of phase and cancel for a null. The second example is an end-ﬁre array. Figure 3-2 illustrates the end-ﬁre pattern with a long-dashed curve. All three patterns on the ﬁgure have the same directivity. The phase distribution of an end-ﬁre array matches those of a wave traveling in the direction of the maximum. In these examples unequal amplitudes would limit the null depth to the difference. Varying the element phases while maintaining equal amplitudes changes the null directions. Consider a general two-element array with equal amplitudes and a phase difference between them. We split the phase shift into equal parts. The top-element phase is −δ/2 and the bottom-element phase is δ/2. When we apply Eq. (3-2) with an isotropic element pattern, we obtain the following electric ﬁeld using Euler’s identity:

δ πd E(θ ) = 2E0 cos cos θ − λ 2

e−j kr r

(3-3)

106

ARRAYS

Even Mode l/2 Spacing

Odd Mode l/2 Spacing

Endfire l/4 Spacing

FIGURE 3-2 Two isotropic element array pattern: even-mode λ/2 spacing (solid curve); odd-mode λ/2 spacing (short-dashed curve); end-ﬁre λ/4 spacing (long-dashed curve).

θ is measured from the z-axis. If we spaced the elements along the x-axis and found the pattern in the x –z plane, we substitute sin θ for cos θ in Eq. (3-3). In Chapter 4 we sample continuous distributions and position the elements along the x- or y-axis. Pattern peaks occur when the argument of the cosine is nπ, the nulls when it is (2n − 1)π/2. δ λ cos θmax = nπ + (3-4) 2 πd π δ λ cos θnull = (2n − 1) + (3-5) 2 2 πd If we subtract either Eq. (3-4) or (3-5) evaluated at two peaks or nulls, we get the same equation: λ cos θ1 − cos θ2 = (n1 − n2 ) (3-6) d Figure 3-3 illustrates the pattern of an equally phased two-isotropic-pattern-element array spaced 5λ along the z-axis. Because array symmetry makes the patterns on the right and left sides the same, we consider only one side. The wide element spacing

TWO-ELEMENT ARRAY

107

5l Spaced Array

Array - 10 dB Added to Isotropic

FIGURE 3-3 Two-isotropic-element array-spaced 5λ pattern (solid curve); added central element 10 dB higher power than array (dashed curve).

allows six solutions to Eq. (3-4) from 0 to 90◦ for the pattern peaks and ﬁve solutions for Eq. (3-5) over the same range for the nulls because the magnitude of cos θ is limited to 1. We call the multiple beams grating lobes. We usually choose the main beam and call the others grating lobes, but they are just all lobes of the array. Figure 3-3 shows that we must space the elements close together to prevent grating lobes. With a greater number of elements in the array, the amount of beam movement due to element phasing adds another factor to the prediction of when grating lobes form. The amount of phase scanning determines the maximum spacing allowed without the formation of grating lobes. The n = 0 lobe forms at θ = 90◦ and we compute the n = 1 mode direction from Eq. (3-4): θ = cos−1 ( 15 ) = 78.46◦ . When we substitute these angles into Eq. (3-3), we calculate a relative phase of 180◦ between them. The lobes have a phase of zero for n even and 180◦ phase for n odd in the far-ﬁeld approximation. Remember we remove the exponential and 1/R factors from Eq. (3-3) for the far-ﬁeld pattern. The actual phase of any real point depends on the distance from the center of the array. The dashed curve in Figure 3-3 shows what happens if we add the array pattern to an isotropic radiator in the center. For a peak response of the array −10 dB relative to the isotropic antenna, we get the 5.7-dB peak-to-peak ripple shown by using Scale 1-8. The array pattern either adds or subtracts from the isotropic radiator pattern. The angular ripple rate is half that of the array lobes. Below we see that a two-element array

108

ARRAYS

spaced at an integer multiple of λ/2 has a 3-dB greater gain than a single element. We feed half the power of the array into each element. By adding these factors we calculate the array element level to be −16 dB below the main central radiating antenna. When we mount an antenna over a ﬁnite ground plane, the diffraction from the edges creates a two-element array. A 5λ-wide ground plane would produce the same pattern ripple angular rate as shown in Figure 3-3. You will often observe a similar-amplitude ripple in measured antenna patterns. Note the minimum angular distance between the peak and minimum responses in the pattern. The extra signals occur along the line in the pattern plane perpendicular to this direction. Use Eq. (3-6) to determine the distance between the array elements and you should be able to identify the structure causing the ripple. The scattering point could be on the test ﬁxture. Consider whether the mounting structure will be different in the ﬁnal conﬁguration. You can calculate the effect from a single diffraction point by forming an array using the baseline of the primary radiator and the diffraction point. Both conﬁgurations produce the same angular ripple rate. The ripple peak occurs along that array axis, but Figure 3-3 shows that the angular ripple rate will be reduced along this end-ﬁre direction of the θ = 0 axis. If you make a careful consideration of the angular rates, in various pattern planes, you should be able to discover the cause. Always consider unexpected sources of diffraction. You can consider the ripple using its beamwidth. To produce a symmetrical pattern about zero, we use sin θ instead of cos θ in Eq. (3-3), which means that the array lies along the x-axis. The −3-dB angle for the two-element uniform amplitude array can be found from Eq. (3-3): πd π sin θ3 dB = λ 4

θ3 dB = sin−1

λ 4d

(3-7)

The beamwidth is twice Eq. (3-7). For large d we can approximate sin X ≈ X and beamwidth = λ/2d. The 5λ spaced array has a beamwidth of 5.7◦ (0.1 rad). We can look at a 5λ-wavelength ground-plane example that has a large-amplitude element compared to the edge diffraction as two 2.5λ-spaced two-element arrays where one element has a high amplitude. Each two-element array produces a pattern with an 11.4◦ beamwidth the value of the composite pattern in Figure 3-3. We often mount an antenna in the center of a ground plane for measurement and observe patterns similar to Figure 3-3. If in the actual application the antenna is mounted off center, we need to add the patterns of arrays formed on both sides of the ﬁnite ground plane. The ﬁnal pattern will be the composite pattern from each array and be more complicated than the simple case given above. We calculate average radiation intensity by an integral: Uavg = The directivity is

4E02 η

π/2 0

cos2

δ πd cos θ − sin θ dθ λ 2

|2Emax |2 Umax = Uavg 1 + sin(2πd/λ) cos δ/(2πd/λ)

(3-8)

where Emax = cos[(πd/λ) cos θmax − δ/2]. When d ≥ λ/2, Emax = 1. Figure 3-4 shows the directivity versus spacing for the special cases δ = 0◦ and δ = 180◦ (even and odd modes). The directivity varies because each antenna receives power from the other. The

LINEAR ARRAY OF N ELEMENTS

109

Odd-mode array (0°, 180°)

4

Directivity, dB

3

2

Even-mode array (0°, 0°)

1

0.5

1.0

1.5 Spacing, l

2.0

2.5

FIGURE 3-4 Directivity of even- and odd-mode two-isotropic-element arrays.

combination of the input power and the power transferred between elements changes with spacing. 3-2 LINEAR ARRAY OF N ELEMENTS Suppose that there are N isotropic radiators equally spaced along the z-axis and fed with equal amplitudes. We assign a ﬁxed phase shift δ between progressive elements. The array factor ﬁeld is sin(N ψ/2) (3-9) N sin(ψ/2) where ψ = kd cos θ + δ [1, p. 258]. We use this to plot a universal radiation pattern for the array (Figure 3-5) for two to 10 elements. The abscissa ψ is plotted in degrees (360◦ is substituted for 2π in k). Both ends of the plot are lines of symmetry. The plot is periodic (period 360◦ ). We see that the level of the ﬁrst sidelobe (N = 2 has no sidelobe) decreases as N decreases but approaches a limit of 13.3 dB of the continuous aperture. Figure 3-6 demonstrates the periodic pattern for N = 6 and shows a projection to a polar pattern when the progressive phase between elements is zero and the elements are spaced λ/2. We can plot similar curves for other array distributions; all have a period of 360◦ . Figure 3-6 illustrates the use of a circle diagram, a method of constructing a polar pattern from the universal pattern such as Eq. (3-9) for the uniform-amplitude distribution. An array can be analyzed as a sampling of the continuous distribution that produces a Fourier series of the distribution. A Fourier series has multiple responses. In Chapter 4 we design large arrays by sampling continuous distributions. The pattern angle of an array is measured either from the axis using cosine of pattern angle or is measured broadside using sine. You should become comfortable with either notation since the sine and cosine of angles involves only a complementary operation of the angles.

110

ARRAYS

−2 −4

3

4

N=2

−6 Amplitude, dB

−8 6

−10

3 4

8

−12

8 10

−14

6

5

10

5 7

−16 8

−18 −20

FIGURE 3-5

6

0

20

40

60

80 100 120 kd cos q + d

140

160

180

ψ-space pattern of linear arrays with a uniform amplitude distribution.

360 sin q Spacing/l Visible Region

FIGURE 3-6 Circle diagram of a six-element uniform-amplitude array with λ/2 spacing.

Since cos θ (or sin θ ) is limited to ±1, the region along the abscissa of the universal pattern used (the visible region) is found from the range of ψ: −360◦ d +δ λ

to

360◦ d +δ λ

LINEAR ARRAY OF N ELEMENTS

111

The circle diagram is constructed by ﬁrst drawing a circle the same diameter as the visible region below the universal diagram centered at δ, the progressive phase shift between elements. Figure 3-6 has a δ = 0. Since the element spacing is λ/2, the range is ±180◦ . The polar pattern radius equals the amplitude of the universal pattern. Both the universal pattern and the polar pattern use a logarithmic (dB) scale from 0 to −40 dB. Projecting points vertically from the universal pattern to the visible region performs the cosine or sine operation, and the polar pattern becomes the real pattern in space. We project each point vertically until it intersects the dashed visible region circle in two places and then draw lines from these points to the center. After you project the nulls and peaks of the universal pattern to the dashed circle, it is easy to sketch the polar pattern. The circle diagram helps us visualize patterns and the effects of scanning, but no one would do serious design with it. Second, it is useful only for small arrays because large arrays produce unwieldy diagrams. When the spacing between elements is greater than λ/2, the visible region widens to include more than one periodic main lobe and the array has multiple beams. To have a beam centered at θ1 , set the progressive phase difference between elements: δ=

−360◦ d cos θ1 λ

(3-10)

−360◦ d λ

(3-11)

End ﬁre (θ1 = 0) occurs when δ=

We can use Figure 3-5 to compute beamwidth angles of arrays. Table 3-1 is a list of the ψ-space angles of the 3- and 10-dB levels. Example A six-element equally spaced uniform array has spacings of λ/2 and zero progressive phase shift between elements (δ = 0◦ ). Calculate the 3-dB beamwidth. We read from Table 3-1 the value ψ3 dB = 26.90◦ . Because the pattern is symmetrical in ψ space (Figure 3-6), the second ψ3 dB is −26.9◦ . kd cos θ1,2 + δ = ±ψ3 dB ±26.9◦ 360◦ λ ◦ cos θ1,2 = cos θ1,2 = ±26.90 λ 2 180◦ ◦ ◦ θ1 = 81.4 θ2 = 98.6 TABLE 3-1 ψ-Space Angles of 3- and 10-dB Levels of an Equal-Amplitude Distribution Array (deg) N

3 dB

10 dB

N

3 dB

10 dB

N

3 dB

10 dB

2 3 4 5 6 7 8 9 10

90.00 55.90 40.98 32.46 26.90 22.98 20.07 17.81 16.02

143.13 91.47 67.63 53.75 44.63 38.18 33.36 29.62 26.64

11 12 13 14 15 16 17 18 19

14.55 13.33 12.30 11.42 10.65 9.98 9.39 8.87 8.40

24.21 22.18 20.47 19.00 17.74 16.62 15.64 14.77 14.00

20 24 28 32 36 40 50 64 100

7.980 6.649 5.698 4.985 4.431 3.988 3.190 2.492 1.595

13.29 11.08 9.492 8.305 7.382 6.643 5.314 4.152 2.657

112

ARRAYS

Remember that θ is measured from the axis of the array (z-axis) and the 3-dB beamwidth is the difference (17.2◦ ). On Figure 3-6 the visible region ranges between −180◦ and +180◦ . There are four sidelobes in the visible region (Figure 3-6). Since an array samples a continuous aperture distribution, the continuous distribution is Nd long. We can estimate beamwidth by using a uniform amplitude distribution: ◦

HPBW = 50.76

λ ◦ = 16.92 Nd

This formula approximates the array beamwidth reasonably. Example A six-element array has a progressive phase shift δ of 90◦ between elements. Compute the 10-dB beam edge angles for λ/2 spacing. Figure 3-7 shows the circle diagram analysis of this example. The line to the center of the polar pattern has been shifted to 90◦ and the pattern spans 360◦ of the linear scale. By projecting the nulls and peaks to the circle below, the pattern can easily be sketched. ◦

ψ10 dB = ±44.63

(Table 3-1)

kd cos θ1,2 = ±ψ10 dB − δ

360 sin q Spacing/l Phase Shift

Visible Region

FIGURE 3-7 Six-element uniform-amplitude array with λ/2 spacing scanned with 90◦ progressive phase shift between elements.

LINEAR ARRAY OF N ELEMENTS

113

Solving for cos θ1,2 , we have ±ψ10 dB − δ ±44.63◦ − 90◦ = kd 180◦ ◦ ◦ ◦ θ1 = 104.6 θ2 = 138.4 beamwidth = 33.8

cos θ1,2 =

There are ﬁve sidelobes in the visible region (Figure 3-7). Equation (3-9) gives the beam maximum direction: cos θ0 =

−δ −90◦ = kd 180◦

◦

θ0 = 120

The main beam is no longer symmetrical about the beam peak. The 3-dB pattern angles are 110.5◦ and 130.5◦ . The beamwidth (3-dB beamwidth = 20◦ ) increases with scan angle. What element spacing would result in this beamwidth for broadside radiation (δ = 0◦ )? 360◦ ◦ d cos θ1 = 26.90 (Table 3-1) λ On solving for spacing, we have d 26.9◦ = λ 360◦ cos θ1 Remember that the beam is centered on θ = 90◦ , so that θ1 = 90 − 20/2 = 80◦ . 26.9◦ d = = 0.431 λ 360◦ cos 80◦ The effective spacing has been reduced by approximately the cosine of the scan angle from θ = 90◦ , broadside: d ◦ cos 30 = 0.433 λ The accuracy of the cosine relation increases with more elements. Example Determine the progressive phase shift between elements for an end-ﬁre array with 0.3λ element spacing and compute beamwidth for a uniform distribution array with ﬁve elements. Figure 3-8 illustrates this example using the circle diagram. End ﬁre occurs when [Eq. (3-11)] −360◦ (0.3λ) ◦ δ= = −108 λ This is the progressive phase shift for all distributions with 0.3λ element spacing for an end-ﬁre pattern. Table 3-1 gives the ψ-space angle, ψ3 dB = ±32.46◦ . Substituting in the expression for ψ, we have 360◦ (0.3λ) ◦ cos θ1,2 = ±32.46 λ ±32.46◦ + 108◦ cos θ1,2 = 360◦ (0.3) 140.46 cos θ1 = = 1.301 108

cos θ2 =

75.46 = 0.699 108

114

ARRAYS

Visible Region

360 sin q Spacing/l Phase Shift

FIGURE 3-8 Five-element uniform-amplitude array scanned to end ﬁre.

θ1 is in invisible space, since | cos θ | ≤ 1; θ2 = 45.6◦ . Symmetry about the z-axis supplies us with the second angle θ1 = −45.6◦ and beamwidth is the difference: 91.2◦ . The end-ﬁre array samples a traveling-wave distribution. The continuous uniform distribution phased for end ﬁre with the same length has a 90.4◦ beamwidth. Remember that we have been dealing with isotropic pattern antennas. For example, broadcast towers, seen from above, approximate isotropic antennas in the horizontal plane. The patterns of the individual antennas modify the results of isotropic antenna arrays. In small arrays the element pattern is quite signiﬁcant, but the beamwidths of large arrays are determined mainly by the array factor. The beamwidths calculated for array factors approximate the actual beamwidths only when the elements have signiﬁcant patterns. We must rely on computer solutions of speciﬁc cases, including the element pattern, for better results. 3-3 HANSEN AND WOODYARD END-FIRE ARRAY [2] The end-ﬁre array directivity increases if the sum of the progressive phase shifts between elements is decreased by approximately π. The equivalent traveling-wave velocity slows in the structure relative to free space. The progressive phase shift between elements becomes δ = −kd −

2.94 π −kd − N N

rad

(3-12)

PHASED ARRAYS q

115

q Hansen and Woodyard end fire

Normal end fire

10 dB

20 dB

30 dB

FIGURE 3-9 Patterns of a normal end ﬁre and a Hansen and Woodyard end-ﬁre array of isotropic elements.

where N is the number of elements in the array. The normal end-ﬁre progressive phase shift between elements, δ = −kd, places one edge of the visible region at the origin of ψ space. This method shifts the edge to a lower portion of the curve. The universal radiation curve peak (Figure 3-5) shifts into invisible space and the sidelobes rise in proportion to the new beam peak, but the beamwidth narrows. Equation (3-12) holds strictly only for large arrays, but the directivity increases for all arrays when it is applied. Example Suppose that eight elements are spaced λ/4 apart with a uniform amplitude distribution. Compare the two endﬁre designs. The two patterns are compared in Figure 3-9. The results are as follows: The beamwidth decreases, and the directivity increases by 2.5 dB. The sidelobes rise to 9 dB from 13 dB.

3-4 PHASED ARRAYS Suppose that a wave approaches at an angle to the axis of an array located on the z axis (Figure 3-10). The wave reaches the top element ﬁrst and progresses down the array in succession. If the signals are added directly, they will cancel each other to some extent because they have a progression of phases. Figure 3-10 shows the results of adding a series of time delays to equalize the path lengths in the lines where the position zi along the axis determines the time delay τi for incident angle θ0 : τi =

zi cos θ0 + τ0 c

116

ARRAYS

t1 Time delay t2

t3

t4

t5

Incident wave

t6

FIGURE 3-10 Linear array scanned with time-delay networks.

and velocity of light c. We add an arbitrary time delay τ0 to keep all time delays, τi , positive. This feed network is frequency independent, as we vary the progression of time delays to scan the beam. Phase shifters replace the time-delay networks in phased arrays. They provide equivalent beam scanning at a single frequency. To scan to an angle θ0 , the required phase shift is 2π − z cos θ0 modulo 2π (rad) λ −360◦ ◦ z cos θ0 modulo 360 (deg) λ for elements located on the z-axis. For a general space array we must counteract the phase difference to the reference plane, ej k·r , for the direction of scan so that the phases of all elements are zero. To scan in the direction (θ0 , φ0 ), we must add a phase factor to every element, depending on its position. The phase factor on each element of a general space array is e−j k0 · r (3-13) where k0 =

2π (sin θ0 cos φ0 xˆ + sin θ0 sin φ0 yˆ + cos θ0 zˆ ) λ

GRATING LOBES

117

is the vector propagation constant in the direction of the beam and r is the element location. Adding this phase factor to the element phases causes the product of the exponential factors [Eq. (3-2)] to be 1 at the scan angle, and the components Ei add in the scan direction. Using phase shifters limits the frequency bandwidth. Given a ﬁxed phase shift over a small frequency range, increasing the frequency scans the beam toward broadside: θ =

π f2 − f1 − θ0 tan f2 2

(3-14)

rad

where θ0 is the scan angle [3]. Limiting the allowable scanning with frequency to plus or minus one-fourth of the local beamwidth deﬁnes the bandwidth of the array. When the beam is scanned to 30◦ off the axis, the bandwidth is related directly to the beamwidth at broadside (θ = 90◦ ): ◦

bandwidth(%) beamwidth (deg) at θ0 = 30

The beam shifts less with frequency near broadside, since the tangent factor in Eq. (314) approaches zero. A general estimate is given by bandwidth(%)

beamwidth (deg) 2 cos θ0

(3-15)

where the broadside beamwidth is used. Example Given an array with 100 elements spaced at λ/2, determine the bandwidth when scanned to 45◦ . The beamwidth is estimated from the aperture width: HPBW =

50.76◦ ◦ 1 100( 12 )

bandwidth(%)

1 = 0.7% 2 cos 45◦

Any radar antenna would have a broader beamwidth because the sidelobes need to be reduced, but this is a good ﬁrst estimate. The bandwidth can be increased by feeding subarrays with time-delay networks. The subarrays continue to be scanned with phase shifters. Only a few time-delay networks are needed, and the subarray beamwidth determines the bandwidth. In Chapter 12 we discuss the problems caused by using subarrays. 3-5 GRATING LOBES Phased arrays vary the progressive phase by Eq. (3-13) to scan the beam. When the array element spacing is greater than λ/2, the appearance of secondary beam peaks (grating lobes) limits the scan angle. The grating lobe attains full amplitude when d −1 λ (1 + cos θgr ) = 1 −1 (3-16) θgr = cos λ d

118

ARRAYS

360 sin q Spacing/l Phase Shift

Visible Region

FIGURE 3-11 Ten-element array with 3λ/4 spacing scanned to 26◦ , showing the onset of a grating lobe.

Example The spacing of the elements of an array is 0.75λ. Determine the scan angle when the grating lobe is full amplitude. 4 ◦ − 1 = 70.5 θgr = cos−1 3 At this point the grating lobe is the same amplitude as the main beam. The lobe does not appear suddenly, but it grows as the visible region shifts and starts including the second periodic main lobe. Figure 3-11 shows the grating lobe formation for an array with 0.75λ element spacings on a circle diagram. The dashed circle of the visible region spans more than one beam of the universal radiation pattern of the uniform amplitude array. Arrays with element spacing greater than λ always have grating lobes (multiple main beams), but the pattern of the antenna elements may reduce the grating lobes to acceptable levels and allow a wide element spacing. 3-6 MULTIPLE BEAMS An array can form multiple beams. Equation (3-13) gives the phase coefﬁcients to multiply each element feed voltage Ei to scan it to a given angle. The array will form

MULTIPLE BEAMS

119

a second beam if we add a second distribution: Ei e−j k2 ·ri . The distribution Ei remains constant for both beams. We add the two distributions to obtain both beams:

Ei (e−j k1 ·ri + e−j k2 ·ri )

(3-17)

This multiplies the distribution Ei by a second distribution whose amplitudes and phases are functions of the antenna position and the scan angles of the two beams. Each beam uses the entire array to form its beam. In a phased array both phase and amplitude must be varied to achieve multiple beams. An array, which can only vary phase, must be divided into subarrays to form multiple beams, but its beamwidths will depend on the subarray widths. We can produce unequal beams with different amplitude distributions and pattern shapes if needed. We can add as many beams as necessary by including the distribution element factors with the scanning phase coefﬁcients in Eq. (3-17). The element feeding coefﬁcients become the sum. Example Compute the feed coefﬁcients of a 15-element array with λ/2 spacings and a uniform distribution scanned to 45◦ and 120◦ from the z-axis. First center the array on the z-axis. The elements are located at zi =

(−8 + i)λ 2

To scan to 45◦ , the element phase factors are 360◦ 1 (−8 + i)λ exp(−j kzi cos 45 ) = exp −j √ λ 2 2 ◦

◦

phase factors are ej 90 (−8+i) . The ninth-element (z9 = λ/2) To scan to 120◦ , the element ◦ ◦ −j 127.3 j 90 phase factors are e and e . Assume a voltage magnitude of one-half for each uniform-amplitude-distribution beam so that the center element has a magnitude of 1. We sum◦ the distributions to calculate the feeding coefﬁcient of the ninth element:0.32ej 161.4 . When converted to decibel ratios, Table 3-2 lists the feeding coefﬁcients for the array. We can estimate both beamwidths from Table 3-1. ψ3 dB = 10.65◦ :

cos θ1,2

±10.65◦ + 127.28◦ 180◦ = ±10.65◦ − 90◦ 180◦

θ1 = 40◦ , θ2 = 49.6◦ θ1 = 116.2◦ , θ2 = 124◦

The pattern (Figure 3-12) has these beams. The gain of each beam depends on the feed network. If a single input supplies power to two beams, each beam can receive only half the input power and gain reduced 3 dB for both beams. Butler matrices [4] and Blass beamforming networks [5] supply an input for each beam. The inputs are isolated from each other and the transmitter power in each port feeds only one beam, and therefore the full array gain is available to each input. Similarly, we can place a receiver on each port and use the full effective area for each.

120

ARRAYS

TABLE 3-2 Feeding Coefﬁcients for a Dual-Beam 15-Element Array Beamwidth (dB)

Angle (deg)

Element

Element

Beamwidth (dB)

Angle (deg)

1 2 3 4 5 6 7 8

−2.38 −8.59 −0.01 −11.49 −1.64 −1.99 −9.91 0.00

130.48 111.84 −86.80 74.56 55.92 −142.72 161.36 0.0

9 10 11 12 13 14 15

−9.91 −1.99 −1.64 −11.49 −0.01 −8.59 −2.38

161.36 142.72 −55.92 −74.56 86.80 −111.84 −130.48

FIGURE 3-12 Fifteen-element linear array pattern with simultaneous beams at θ = 45◦ and 120◦ .

We will delay the important topics of array synthesis and sidelobe reduction until after we have discussed aperture distributions. A trade-off is made between the beamwidth and the sidelobe levels. The beamwidth narrows only by putting more power into the sidelobes. 3-7 PLANAR ARRAY The linear array only controls the pattern in one plane; it depends on the element pattern to control the beam in the other plane. Planar arrays can control the beam shape in both planes and form pencil beams. Whereas a linear array can only scan in

PLANAR ARRAY

121

a single plane, a planar array can scan to any angle in the upper hemisphere. Most planar arrays rely on the element pattern or ground plane to eliminate the backlobe on the opposite side of the plane. The planar array has N − 1 nulls that can be used to control the pattern where N is the total number of elements. A simple feed distribution uses the product of two linear arrays. This eliminates many degrees of freedom of the array because an M × N array would be determined by M − 1 + N − 1 nulls when we could have used M × N − 1 possible nulls. Figure 3-13 shows the spherical pattern of a uniformly spaced 8 × 8 planar array where all elements are fed the same amplitude where a 90◦ beamwidth element eliminates the backlobe. The pattern along either principal axis shows the steady sidelobe reduction that starts with −13.2 dB. Diagonal plane sidelobes are the product of the sidelobes in the principal planes. The ﬁrst sidelobe in the diagonal plane is down 26.4 dB. An array feed distribution, not a product of two linear arrays, can yield more equal sidelobes in all pattern planes. Figure 3-14 illustrates the pattern of the rectangular array when the element feeding coefﬁcients are phased to scan the beam along one principal plane. The main beam broadens in the plane of scan as the effective array length is reduced but stays narrow in the orthogonal plane. More sidelobes appear behind the main beam. We see a large sidelobe growing on the horizon that will become a grating lobe when the array is scanned further. The sidelobes in the plane orthogonal to the plane of scan move with the main beam but roll into a cone that becomes tighter with increased scan.

FIGURE 3-13 Spherical radiation pattern of an 8 × 8-element uniform-amplitude and spaced square planar array.

122

ARRAYS

FIGURE 3-14 Spherical radiation pattern of an 8 × 8-element square-planar array scanned along a principal plane. 2.5 2 1.5 1

ky-space

0.5 0 −0.5 −1 −1.5 −2 −2.5 −2.5

−2

−1.5

−1

−0.5

0

0.5

1

1.5

2

2.5

kx-space

FIGURE 3-15 Contour plot of the pattern of a 4 × 4-element square array in kx ky -space showing multiple beams and sidelobes.

Figure 3-15 shows a contour plot of the universal pattern of a 4 × 4 element rectangular array. We denote this universal pattern kx ky space because the principal axes have sin θ factors similar to the universal pattern of a linear array. The array for Figure 3-15 has its y-axis element spacing 1.5 times wider than the x-axis spacing. The diagram axes extend until multiple beams show on the ﬁgure. The main beams correspond

PLANAR ARRAY

123

to the center of the large “squares.” The visible region on the ﬁgure is a unit circle with its center at the negative scan direction (−kx0 , −ky0 ). This technique mirrors the circle diagram of the linear array where the visible region is given by a linear region centered at the negative scan direction. You should notice that the diagonal sidelobes have smaller amplitudes than the principal plane sidelobes. We move the unit circle as the array scans and the diagram shows those locations of scan that have multiple beams (grating lobes). A grating analysis simpliﬁes the diagram of Figure 3-15 to the main beam locations. When we place the two axes of the planar array at an angle instead of orthogonal, we form a triangular array. Figure 3-16 gives the positions of a hexagon array made with equilateral triangles. We derive the characteristics of this array from a linear transformation of the rectangular array [6, p. 11-23ff]. Because the array has six-way symmetry, Figure 3-17 the pattern of a uniform-amplitude 61-element hexagon array shows the same six-way symmetry in the ring sidelobe around the main beam. If we collapsed the hexagonal distribution to a line in one plane, the distribution has a taper that reduces the sidelobes. The sidelobe amplitudes of the uniform hexagonal array are lower that the principal-plane sidelobes of the rectangular uniform array. Figure 3-18 plots the spherical pattern of the hexagon array when scanned to 36◦ . The ﬁrst ring sidelobe has a distorted six-fold symmetry. Similar to the scanned rectangular array (Figure 3-14), the hexagon array moves more sidelobes into visible space in the area opposite the scanned main beam. Figure 3-14 showed a grating lobe entering visible space, but the hexagon array pattern in Figure 3-18 does not. The grating lobes of a rectangular array can be found from a linear array when it is scanned along one of the principal axes, but the hexagon array requires a more elaborate analysis. When we scan the rectangular array off the principal axes, we can no longer use the grating lobe analysis of linear arrays.

FIGURE 3-16 Position of elements in a hexagonal planar array.

124

ARRAYS

FIGURE 3-17 Spherical radiation pattern of a 61-element hexagonal array.

FIGURE 3-18 Spherical radiation pattern of a 61-element hexagonal array scanned along a principal plane.

GRATING LOBES IN PLANAR ARRAYS

125

3-8 GRATING LOBES IN PLANAR ARRAYS The circle diagrams of the linear array can be used in principal planes of a rectangular array to compute grating lobes. For the planar array we use a sin θ pattern space to see the periodicity of the grating lobes and to analyze scanning in planes other than the principal axes. The visible region is now limited to a unit circle in kx ky -space where kx = sin θ cos φ and ky = sin θ sin φ. kx is the pattern in the x –z plane, and ky is the pattern in the y –z plane. Figure 3-19a shows the array layout, and Figure 3-19b shows the corresponding kx ky -plane grating lobe diagram. We reduce a contour plot of the pattern response similar to Figure 3-15 to only the main beams for analysis of grating lobes. The full contour plot is too busy. The narrower x-axis array spacing compared to the y-axis spacing leads to wider-spaced grating lobes in the kx -plane than in the ky -plane. Beam scanning corresponds to movement of the unit circle in kx ky -space. Each small circle in kx ky -space is a main beam in pattern space. When the unit circle encloses more than one kx ky -circle, the pattern has multiple main beams or grating lobes. The kx ky -plane diagram could also include sidelobe peaks or the contour plot of the array pattern and illustrate the pattern change with scan. The kx ky -plane is the two-dimensional Fourier transform of the distribution that becomes the periodic twodimensional Fourier series because the distribution is discrete. Increasing frequency or relative spacing between elements increases the unit circle diameter on an existing kx ky -diagram in a manner similar to the circle diagram. When we scan the beam, we move the unit-circle center in kx ky -space. We use Eq. (3-13) to locate the unit-circle center on the diagram: k(sin θ0 cos φ0 , sin θ0 sin φ0 ). The off-center circle on Figure 3-19 corresponds to a scanned beam and encloses two main beams. In this case the grating lobe does not lie in the scan plane and would fail to show in a simple pattern cut through the scan plane. A rectangular array produces a rectangular grating lobe diagram, while other periodic arrays lead to more complicated grating lobe diagrams. Figure 3-20 shows the layout of the hexagonal array and the corresponding grating lobe diagram. The hexagonal array (or equilateral triangular array) can be found from a linear transformation of the rectangular array. The grating lobe diagram can be found from the transformation as well. The spacing along the x-axis A1 corresponds to the y

k-y Main Beam l/A2 Visible Region, Broadside

A2

x λ/A1

A1 Scanned Beam (a)

k-x

Grating Lobe

(b)

FIGURE 3-19 (a) Grating lobe diagram of a rectangular array; (b) distribution in k-space.

126

ARRAYS

l/(a1 sin α) Grating Lobe a2

a a1

Visible Region Broadside Beam λ/(a2 sin a) Scanned Beam

(a)

(b)

FIGURE 3-20 (a) Grating lobe diagram of a hexagonal array; (b) distribution diagram.

vertical spacing B2 on the grating lobe diagram, and the spacing along the diagonal of each diagram are related. In both cases the corresponding axes on the two diagrams are perpendicular: λ λ B2 = and B1 = A1 sin α A2 sin α The angle between the triangular axes α is 60◦ for the hexagonal array. By allowing a grating lobe when the beam is scanned to 90◦ , we can determine the maximum element spacing without grating lobes: λ = 2 or A1 sin 60◦

1 A1 = = 0.577 λ 2 sin 60◦

Figure 3-20 shows the visible region unit circle with the beam broadside to the plane and then scanned to 36◦ for an element spacing of λ. When scanned, the unit circle encloses three lobes. The three lobes do not lie in a plane. Figure 3-21 gives the

FIGURE 3-21 Spherical radiation pattern of hexagon-array grating lobes.

SCAN BLINDNESS AND ARRAY ELEMENT PATTERN

127

spherical pattern of the array when scanned to 36◦ and shows the three lobes in the pattern. The array sidelobes have been reduced by sampling a circular Taylor distribution with the array so that the lobes show clearly. In Chapter 4 we discuss the use of continuous aperture distributions to determine the feed amplitudes of planar arrays. 3-9 MUTUAL IMPEDANCE Antennas in an array couple to each other because they receive a portion of the power radiated from nearby elements. This affects the input impedance seen by each element, which depends on the array excitation. We scan a phased array by changing the feeding coefﬁcients, and this changes the element input impedance called the scan impedance. To ﬁrst order, the coupling or mutual impedance is proportional to the element pattern level along the array face, and we reduce coupling by using narrower-beamwidth elements. Mutual coupling can be represented by an impedance, admittance, or scattering parameter matrix. The ﬁrst element of an N -element array has the impedance equation V1 = Z11 I1 + Z12 I2 + Z13 I3 + · · · + Z1N IN If we know the radiation amplitudes, we calculate the ratio of the currents: I2 I3 IN V1 = I1 Z11 + Z12 + Z13 + · · · + Z1N I1 I1 I1 The effective or scan impedance of the ﬁrst element is Z1 =

V1 I2 I3 IN = Z11 + Z12 + Z13 + · · · + Z1N I1 I1 I1 I1

(3-18)

It depends on the self-impedance and the excitation of all the other antennas. Scan impedance was formerly called active impedance, but this led to confusion. The power into the ﬁrst element is I2 I3 IN ∗ ∗ P1 = Re(V1 I1 ) = I1 I1 Re Z11 + Z12 + Z13 + · · · + Z1N (3-19) I1 I1 I1 By knowing the feeding coefﬁcients and the mutual impedances, we can compute the total input power and gain. In general, every antenna in the array has different input impedances. As the feeding coefﬁcients change in a phased array to scan the beam, so will the impedance of elements. The scan impedance change with scan angle causes problems with the feed network. We can repeat the same arguments for slots using mutual conductance, since magnetic currents are proportional to the voltage across each slot. 3-10 SCAN BLINDNESS AND ARRAY ELEMENT PATTERN [7, pp. 339–355; 8, pp. 365–366] Large arrays made from elements with wide beamwidths can exhibit scan blindness. When a phased array is scanned, at certain angles the input reﬂection coefﬁcient of every element rapidly increases to 1. The array fails to radiate and forms a pattern

128

ARRAYS

null. Mutual coupling between elements causes the change of the scan impedance, which leads to scan blindness, which is complicated and difﬁcult to predict accurately except where the array structure supports a surface wave. One approach says that scan blindness occurs when a grating lobe ﬁrst enters from invisible space and radiates along the surface of the array. In this case we solve Eq. (3-16) for the scan angle of the grating lobe: λ | cos θgr | = − 1 (3-20) d The angle in Eq. (3-20) is measured from the array plane (or axis). Scan blindness occurs approximately at this angle, but it can be reduced to only a dip in the pattern if the array is small or the mutual coupling between the elements is small, because they have narrow beams. The grating lobe causes a large increase in mutual coupling. Arrays made with antennas that can support surface waves, such as microstrip patches on dielectric substrates, can exhibit scan blindness when the electrical distance between the elements equals the surface-wave propagation phase shift: | cos θgr | =

λ ksw λ − = −P d k d

(3-21)

P is the relative propagation constant with a value > 1 for a surface wave (Section 101). Scan blindness will occur at an angle near this value because of the complicated nature of the coupling addition in the array. We can build a small portion of the array and determine where scan blindness will occur. Feed the center element and load all others with the feeder resistance. Each element in an array will couple to its neighboring elements and we can associate the combination of the element radiation and the coupled radiation of the neighbors when loaded to the element. We call this the array element pattern or scan element pattern (formerly called the active pattern). Elements near the edges will have different effective patterns, but in a ﬁrst-order solution we assume the pattern of the center element for all and calculate the total pattern as the product of the element pattern and the array factor. The array element pattern will exhibit dips where scan blindness will occur in the full array. Because it is only a small portion of the array, the full scan blindness will not occur. You should build a small array and test for scan blindness whenever it is a possibility. For example, arrays that scan to large angles off broadside using broad-beamwidth elements need to be tested with a small array before building the complete array. 3-11

COMPENSATING ARRAY FEEDING FOR MUTUAL COUPLING

Mutual coupling (impedance) is a measure of how much one antenna receives radiation from its neighbors in the array. Each element radiation changes the effective excitation on its neighbors. In a large array not requiring exact patterns, the effects average out. But when the array is small or you try to achieve low sidelobes, mutual coupling must be compensated for in the array. Small antenna elements such as dipoles or slots are resonant structures that radiate in only one mode. Mutual coupling only changes the element excitation, not the shape of the current distribution on the element. In this case we measure or calculate the mutual coupling matrix and use it to compute element

ARRAY GAIN

129

excitation to achieve the desired excitation [9]. Find the coupling matrix by adding the identity matrix to the S-parameter matrix of the antenna coupling: C=I+S

(3-22)

We compute the new feed excitation from the desired excitation and matrix inverse of Eq. (3-22): Vrequired = C−1 Vdesired (3-23) Because we assumed a single mode distribution on the antenna elements, S is independent of scanning and Eq. (3-23) gives the compensation for all scan angles. The compensation can be applied to the received signal in an adaptive array by matrix multiplication in digital signal processing. The effects of these operations have been illustrated [10]. Without compensation adaptive arrays, such as the MUSIC algorithm, only generate small peaks, whereas compensation produces the expected large peaks. Compensation for multimode elements starts with a moment solution [11] and uses the pattern characteristics to solve for the feeding coefﬁcients. We use the pattern desired to compensate the feeding coefﬁcients. We start with a matrix between the pattern response and the currents on all the antenna elements found from a moment method solution, where each array element has multiple current segments: A(k) = FI

(3-24)

A(ki ) is an element of the column matrix that gives the pattern response at an angle given by ki = xˆ sin θi cos φi + yˆ sin θi sin φi + zˆ cos θi or a given pattern angle (θi , φi ). The elements of the matrix F are the isotropic element phase terms, ej ki ·rj , and I is the column vector of the currents on the segments. We calculate excitation voltages by inverting the mutual impedance matrix: I = Z−1 V

(3-25)

We substitute Eq. (3-25) into Eq. (3-24) and note that the matrix V has only q nonzero terms corresponding to the feed points. We specify q pattern points, which reduces F to q × M for M current segments. The vector V has M − q zero elements and we delete the corresponding columns in the matrix product FZ−1 . This reduces the matrix to q × q, denoted B: A(k) = BV This uses the nonzero element V . We solve for the feeding coefﬁcients by inverting the matrix B found from q pattern points: V = B−1 A(k)

(3-26)

Choosing good pattern points is an art that requires pattern evaluation to verify whether the ﬁnal pattern is acceptable. 3-12 ARRAY GAIN We can use the mutual impedance concept to determine the effective input power of every element and thereby avoid having to integrate the pattern to calculate average

130

ARRAYS

radiation intensity. We represent the circuit relations of two antennas by a two-port impedance matrix: V1 Z11 Z12 I1 = V2 Z21 Z22 I2 Symmetrical elements across the diagonal of the matrix are equal for antennas satisfying reciprocity. The total input power is given by Pin = Re(V1 I1∗ ) + Re(V2 I2∗ ) The general N -element array has an N × N matrix and N terms in the input power sum. Given the feed coefﬁcients, we have a relation between different Ii . For our two elements, I2 = I1 ej δ and V1 = (Z11 + Z12 ej δ )I1 The power into the ﬁrst element is Re(Z11 + Z12 ej δ )I1 I1∗ By symmetry, the power into the second antenna is the same. The total input power to the array is Re(Z12 ej δ ) ∗ Pin = 2 Re(Z11 )I1 I1 1 + Re(Z11 ) The factor Re(Z11 )I1 I1∗ is the power into an isolated element: 4πE02 /η. The average radiation intensity (100% efﬁcient antenna) is Pin /4π: gain = directivity =

|2E(θmax )|2 |2E(θmax )|2 /η = Pin /4π 1 + [Re(Z12 ej δ )]/[Re(Z11 )]

(3-27)

By comparing Eqs. (3-27) and (3-8), we can identify sin(2πd/λ) cos δ R12 cos δ Re(Z12 ej δ ) = = Re(Z11 ) R11 2πd/λ sin(2πd/λ) R12 (d) = R11 2πd/λ We can use this mutual impedance ratio to compute directivity of arrays of isotropic elements of any number. Example Calculate the directivity of a linear array of three equally spaced isotropic elements with equal amplitudes and phases. The powers into the elements are 4πE02 R12 (d) R12 (2d) P1 = P3 = + 1+ η R11 R11 2 4πE0 32 E02 2R12 (d) Umax = P2 = 1+ η R11 η

ARRAY GAIN

131

The total power into the array is found from the sum: 4πE02 4R12 (d) 2R12 (2d) Pt = P1 + P2 + P3 = + 1+ η R11 R11 directivity =

Umax 9 = Pt /4π 3 + [4R12 (d)/R11 ] + [2R12 (2d)/R11 ]

The directivity of the general N -element equally spaced linear array, excited by equal-amplitude and equal-phase signals, is easily found by extending the development: directivity =

N 2 (element directivity) N−1 N +2 (N − M)[R12 (Md)/R11 ]

(3-28)

M=1

The directivity attained in an array depends on the particular mutual impedance terms of the radiators. The equation above only handles uniform-amplitude linear arrays. We can extend the idea of mutual resistance to calculate input power to a general planar array consisting of identical elements and determine gain. By using a two-element array spaced along the x-axis we can integrate the pattern to compute directivity and from that determine the ratio of mutual resistance to selfresistance of the elements versus element spacing: R12 (x) element directivity = R11 2π

2π 0

π 0

Ee2 (θ, φ) cos2

πx λ

cos φ sin θ sin θ dθ dφ − 1

(3-29) Equation (3-29) uses the normalized element pattern in the integral. By using an axisymmetrical element pattern, we calculate the ratio of resistances at a number of different distances and interpolate on the table for the directivity (gain) analysis of a planar (linear) array. If the element pattern is not symmetrical, the normalized resistance must be calculated for a number of φ. Given the element excitations Ei with elements located at the vector locations xi , we can derive an equation similar to Eq. (3-28) for directivity of a planar array:

directivity = N i=1

2

N

Ei (element directivity)

i=1 N [R12 (|xi − xj |)/R11 ]Re[Ej /Ei ]|Ei |2

(3-30)

j =1

Figure 3-22 illustrates the directivity calculated from Eq. (3-30) for linear arrays with realistic elements, such as a microstrip patch with 90◦ beamwidths, as the element spacing is varied. The graph shows directivity reduction when the element spacing exceeds λ and grating lobes form a more pronounced characteristic as the number of elements increases. When the second grating lobe occurs for wider element spacing, the directivity exhibits only minor variations. Increasing the element directivity (decreased beamwidth) reduces variation because the element pattern reduces the grating lobe. We use Eq. (3-30) with a planar array to obtain Figure 3-23. This array consists of 217 elements arranged in a hexagonal pattern, with amplitudes found from sampling a circular Taylor distribution (Sections 4-18 and 4-19) to lower the sidelobes. The 30-dB circular Taylor distribution reduces the gain by 0.6 dB relative to a uniform distribution

132

ARRAYS

32 elements, 60° Beamwidth 32 elements, 90° Beamwidth

Directivity, dB

16 elements, 60° Beamwidth 16 elements, 90° Beamwidth 8 elements, 60° Beamwidth 8 elements, 90° Beamwidth 4 elements, 60° Beamwidth 4 elements, 90° Beamwidth 2 elements, 60° Beamwidth 2 elements, 90° Beamwidth

Element Spacing, l

FIGURE 3-22 Directivity of a uniform-amplitude line array versus element spacing for 60◦ and 90◦ beamwidth elements.

16.2 dB Gain, 30° Beamwidths

Directivity, dB

12.8 dB Gain, 45° Beamwidths

10.4 dB Gain, 60° Beamwidths

8.6 dB Gain, 75° Beamwidths 7.3 dB Gain, 90° Beamwidths

Element Spacing Normalized to 0.6l

FIGURE 3-23 Directivity of a 217-element uniform-amplitude hexagonal array for various element beamwidths versus spacing.

due to the amplitude taper across the array. Initially, the element spacing is 0.6λ, but the element spacing has been allowed to grow in Figure 3-23 to show the effect. Figure 3-23 also illustrates the effect of increasing the element gain on the gain of a planar array. When we space the elements less than λ, increasing the element gain has no effect on array gain because the effective area of an antenna with a 90◦ beamwidth exceeds the area between elements and collects all power incident on the array. If we increase the element gain, the effective areas of the elements overlap and they share the incident power. On Figure 3-23 the curves overlap for element spacing less than

ARRAYS USING ARBITRARILY ORIENTED ELEMENTS

133

about λ. At the lower end of Figure 3-23 the gain increases by 6 dB as the element spacing doubles. This shows that increasing the element gain will have no effect on array gain when the present element covers the area associated with it. When the grating lobe enters from invisible space as the element spacing increases, those arrays with narrower-beam elements suppress the lobes and continue the general gain increase. The array directivity (gain) drops as the element spacing increases for the wide-beamwidth element with 90◦ beamwidth because of the grating lobe. At a large element spacing the array gain becomes N times the element gain. We determine array gain from the array area and the amplitude taper for closely spaced elements. For wide element spacing, we calculate gain from the product of the number of elements and the element gain. Figure 3-23 shows a smooth transition between the two regions. 3-13 ARRAYS USING ARBITRARILY ORIENTED ELEMENTS When we mount arrays on vehicles, the elements are pointed in arbitrary directions. Although Eq. (3-1) will calculate the pattern of any array, the element patterns are usually measured in a coordinate system in a different orientation than in the array. The idea of an array factor times the element pattern collapses and an analysis must rotate the pattern direction into the coordinates of each element. We will use coordinate rotations on the elements not only to specify them, but to calculate the pattern of the array. In a later chapter we use the same concept to point a feed antenna at a reﬂector. We rotate the pointing direction into the coordinate system of the orientated antenna to determine what direction angles to use for the element pattern. We do this by using a 3 × 3 rotation matrix on rectangular components: X (3-31) [Xrotated Yrotated Zrotated ] = [rotation matrix] Y Z A similar problem is rotating an object. Both cases use the same matrix. To rotate an object we multiply the vector by the rotation matrix from the left to compute the rotated coordinates. Rotating a position is given by the equation Xrotated [Xold Yold Zold ][rotation matrix] = Yrotated (3-32) Zrotated The rotation matrix can be found from the directions of the unit vectors when rotated. It is given by rotated X-axis (3-33) rotation matrix = rotated Y -axis rotated Z-axis The method uses 3 × 3 matrices to perform the rotation by a multiplication with a position or direction vector. Rotation about the X-axis is given by 1 0 0 0 cos A sin A 0 − sin A cos A

134

ARRAYS

rotation about the Y -axis is given by cos B 0 sin B

0 1 0

− sin B 0 cos B

and rotation about the Z-axis is given by cos C sin C − sin C cos C 0 0

0 0 1

We use products of these axis rotations to reorient an object or pointing direction. Consider the rotation of a position by the product of three rotation matrices: Xrotated [Xold Yold Zold ]R1 R2 R3 = Yrotated Zrotated The logical approach is to multiply the 3 × 3 matrices, R1 , R2 , and R3 , before multiplying by the position vector. When we postmultiply R1 by R2 , it rotates the axis of rotation of R1 . The postmultiplication by R3 rotates the rotation axis R2 and R1 is rotated once more. We can take the rotations one by one from left to right and use the rotation matrices about each of the principal axes provided that we convert the column vector back to a row vector after each multiplication. A convenient way to deﬁne the orientation of objects in space is to use spherical coordinate angles, since they are the same as pattern angles. We line up the matrices from right to left in this case. When rotating the coordinate system about an axis, the other axes change direction. The next rotations are about these new axes. The three rotations are often called the Euler angles. We use the following three rotations for spherical coordinate pointing: 1. Z-axis rotation = φ 2. New Y -axis rotation = θ 3. New Z-axis rotation: aligns the polarization of the antenna The last rotation takes some thought because the ﬁrst two rotations have altered the orientation of the antenna. When calculating the pattern of the array for a particular direction, ﬁrst compute rectangular components of the direction vector and the two polarization vectors. Multiply the direction vector by k(2π/λ) and take the dot (scalar) product with the position vector to calculate phasing of a particular element. You need to determine the pattern direction in the rotated antenna’s coordinate system found by using Eq. (3-31). Multiply the rotation matrix by the unit direction vector placed to the right. When you convert the output vector to spherical coordinates, you obtain pattern coordinates of the rotated antenna. Both the pattern components of the rotated element and the unit polarization vectors are needed. In the next operation you rotate the prime coordinate polarization unit vectors into the rotated element coordinate system using the same operation as the direction vector.

REFERENCES

135

You calculate ﬁnal radiated components by projecting the rotated prime coordinate polarization vectors on the element pattern unit polarization vectors: Eθ = Eθ,element θˆelement · θˆrotated + Eφ,element φˆ element · θˆrotated Eφ = Eθ,element θˆelement · φˆ rotated + Eφ,element φˆ element · φˆ rotated

(3-34)

Since we measure element patterns on antenna positioners, it is convenient to consider positioners as a series of coordinate system rotations. REFERENCES 1. C. A. Balanis, Antenna Theory, Analysis and Design, 2nd ed., Wiley, New York, 1997. 2. W. W. Hansen and J. R. Woodyard, A new principle in directional antenna design, Proceedings of IRE, vol. 26, March 1938, pp. 333–345. 3. T. C. Cheston and J. Frank, Array antennas, Chapter 11 in M. I. Skolnik, ed., Radar Handbook, McGraw-Hill, New York, 1970. 4. J. L. Butler, Digital, matrix, and intermediate frequency scanning, Section 3 in R. C. Hansen, ed., Microwave Scanning Antennas, Vol. III, Academic Press, New York, 1966. 5. J. Blass, The multidirectional antenna: a new approach to stacked beams, IRE International Convention Record, vol. 8, pt. 1, 1960, pp. 48–50. 6. Y. T. Lo and S. W. Lee, eds., Antenna Handbook, Vol. II, Van Nostrand Reinhold, New York, 1993. 7. R. J. Mailloux, Phased Array Antenna Handbook, Artech House, Boston, 1994. 8. P.-S. Kildal, Foundations of Antennas, Studentlitteratur, Lund, Sweden, 2000. 9. H. Steyskal and J. S. Herd, Mutual coupling compensation in small array antennas, IEEE Transactions on Antennas and Propagation, vol. AP-38, no. 12, December 1990, pp. 1971–1975. 10. A. G. Derneryd, Compensation of mutual coupling effects in array antennas, IEEE Antennas and Propagation Symposium, 1996, pp. 1122–1125. 11. B. J. Strait and K. Hirasawa, Array design for a speciﬁed pattern by matrix methods, IEEE Transactions on Antennas and Propagation, vol. AP-18, no. 1, January 1971, pp. 237–239.

4 APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

Continuous apertures and arrays share similar characteristics. We compute the radiation pattern of the aperture by using the Fourier transform. Array sampling of an aperture distribution leads to a Fourier series analysis for its pattern. We rely on our familiarity with signal processing to give us insights into these processes and their characteristics. We apply aperture theory to the analysis of horns, lens, and reﬂector antennas, but it also describes array antennas. Since we can design antennas only approximately to produce particular aperture distributions, we often realize them by sampling with an array. We start with aperture efﬁciencies developed from the Huygens source approximation of Section 2-2. We apply this method to horns, lens, and reﬂector antennas for both synthesis and tolerance analysis. The uniform and cosine distributions occur naturally in horns and simple resonant antennas. We use aperture distributions to realize bounds on antenna characteristics given size and excitation distribution. Taylor developed an aperture distribution based on Dolph’s use of the Chebyshev polynomials to produce the narrowest beamwidth for a speciﬁed sidelobe level for an array. The Chebyshev array design produces equal-amplitude sidelobes that we discover to be undesirable for large arrays because the equivalent aperture distribution peaks at the ends and the average value of the sidelobes limits the directivity to 3 dB above the sidelobe level. Large edge peaking of the distribution requires a feed network containing a large ratio of coupling values. Mutual coupling between elements causes unwanted excitation for a large ratio of element amplitudes and we lose control. Our usual practice is to sample a Taylor distribution for large arrays. The distribution has limited edge peaking, and large arrays can realize high gains. Aperture distribution synthesis involves manipulating pattern nulls to achieve desired characteristics. Taylor used the zeros of the Chebyshev array to alter the positions of the inner nulls of the uniform distribution to lower sidelobe levels. Elliott extended this Modern Antenna Design, Second Edition, By Thomas A. Milligan Copyright 2005 John Wiley & Sons, Inc.

136

AMPLITUDE TAPER AND PHASE ERROR EFFICIENCIES

137

idea to iterate the positions of these nulls to produce a linear aperture that radiates individually speciﬁed sidelobes. Schelkunoff developed a transformation between the pattern of an array and a polynomial where we combine the roots (or zeros) of the array polynomial in the complex plane with a mapped pattern variable that traverses the unit circle to analyze array patterns. We synthesize arrays by manipulating these polynomial zeros in the complex plane. Similar to Elliott’s method of null positioning for the continuous linear aperture, Orchard (and Elliott) developed an iterative method applied to array polynomial zeros to synthesize arrays. The method allows us to specify sidelobes individually and to shape the main beam pattern by moving some zeros off the unit circle. When designing shaped beams, improved synthesis by the Orchard method reduces our use of both array sampling of the Woodward continuous aperture method and direct Fourier series synthesis for linear arrays, but both earlier methods give us insight. We consider the design of series feeding where elements are fed directly from a transmission line for a linear array or continuous linear aperture. This requires speciﬁcation of the couplers or loading of the transmission line along the array because a portion of the power is extracted at each position with the remaining power dissipated in a load. We repeat aperture analysis for circular apertures to show limitations of large reﬂector antennas and for use in sampling with an array. For planar arrays, we reduce many rectangular apertures to the product of two linear distributions. A Chebyshev-type planar array with equal sidelobes can be designed so that the sidelobes in the diagonal planes are not reduced unnecessarily. Convolution synthesis of planar arrays allows manipulation of the pattern zeros in groups of smaller arrays similar to the Schelkunoff method. Finally, we consider aperture blockage and phase errors that lead to gain reduction and increased sidelobes.

4-1 AMPLITUDE TAPER AND PHASE ERROR EFFICIENCIES When we use the Huygens source approximation, we calculate power radiated by summing (integrating) the magnitude squared of the electric ﬁeld in the aperture and dividing by the impedance of free space. The average radiation intensity is the radiated power divided by the area of a unit sphere, 4π. To complete the calculation, we compute the maximum radiation intensity by dividing the maximum of the magnitude squared of Eq. (2-24) by the impedance of free space and directivity (Umax /Uavg ) becomes 2 j k · r Ee ds π(1 + cos θ )2 s max λ2 |E|2 ds

(4-1)

s

Equation (4-1) can be used for directivity in any pattern direction, including the maximum of the numerator integral. An aperture with a uniform amplitude and phase distribution has directivity 4πA/λ2 , where A is the area. We separate directivity reductions into individual terms due to

138

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

aperture ﬁeld amplitude and phase variations, and we express the general aperture directivity as 4πA directivity = 2 · ATL · PEL λ where ATL is the amplitude taper efﬁciency (loss) and PEL is the phase error efﬁciency (loss). Only amplitude variations contribute to ATL, and only phase variations determine PEL. We start with a uniform phase distribution in the aperture where the beam peak occurs normal to the aperture (θ = 0◦ ) and PEL = 1. We obtain uniform phase ﬁelds by using |E| in Eq. (4-1): directivity =

4π λ2

2

|E| ds

s

=

|E|2 ds

4πA · ATL λ2

s

where kx = ky = 0 on the boresight (θ = 0◦ ). On solving for ATL, we derive

2 |E| ds

s

ATL =

(4-2)

A

|E| ds 2

We have forced a constant phase everywhere in the aperture to separate out the amplitude taper effects. We account for nonuniform phase with PEL. The phase error efﬁciency can be found from PEL(θ, φ) =

directivity(θ, φ) (4πA/λ2 ) · ATL

where we use directivity (θ , φ) and PEL (θ , φ) depends on the pattern direction (θ, φ): 2 j k · r Ee ds (1 + cos θ )2 s PEL(θ, φ) = 2 4 |E| ds s

k = k(sin θ cos φ xˆ + sin θ sin φ yˆ + cos θ zˆ ) For an aperture in the x –y plane, k · r = k(x sin θ cos φ + y sin θ sin φ)

(4-3)

AMPLITUDE TAPER AND PHASE ERROR EFFICIENCIES

139

We determine maximum PEL to relate it and ATL to directivity. Traditionally, we use the boresight value (θ = 0◦ ) and Eq. (4-3) reduces to 2 E ds s PEL = 2 |E| ds

(4-4)

s

Unless speciﬁed, PEL will be Eq. (4-4) and we use Eq. (4-3) for scanned apertures. Equations (4-2) and (4-4) separate the effects of amplitude and phase variations in the aperture on the directivity at the boresight. If these efﬁciencies are expressed in decibels, the directivity becomes directivity(dB) = 10 log

4πA + ATLdB + PELdB λ2

Expressed in decibels, the efﬁciencies are called losses: amplitude taper loss (ATL) and phase error loss (PEL). It is important to remember that these are the losses at the boresight. A linear phase taper across the aperture scans the beam, but Eq. (4-4) predicts the boresight loss, which could be a null of the pattern. ATL is independent of phase variations that cause squinting of the beam. 4-1.1 Separable Rectangular Aperture Distributions If the distribution in a rectangular aperture is separable, E(x, y) = E1 (x)E2 (y) the efﬁciencies also are separable. and PEL = PELx PELy

ATL = ATLx ATLy

(4-5)

Given a rectangular aperture with an x-axis excursion of ±a/2, ATLx = a

a/2

2 |E1 (x)| dx

−a/2 a/2

−a/2

(4-6) |E1 (x)| dx 2

2 a/2 E1 (x) dx −a/2 PELx = 2 a/2 |E1 (x)| dx −a/2

The formulas for the y-axis are the same except for the substitution of y for x.

(4-7)

140

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

4-1.2 Circularly Symmetrical Distributions If a circular aperture has a circularly symmetrical distribution, we easily reduce Eqs. (4-2) and (4-4) to

a

2 0

ATL = a2

a

2 |E(r)|r dr (4-8) |E(r)| r dr 2

0

2 a E(r)r dr PEL = 0a 2 |E(r)|r dr

(4-9)

0

where a is the radius. We need a short word on formulas using integrals. They look formidable and seem to have little immediate practical use. In the catalog of distributions to follow, results will be given. A general distribution must be solved by numerical integration. One of the Newton–Cotes methods, such as Simpson’s rule or the Rhomberg integration, can be used when evenly spaced values are known. With a known function for the distribution, we use the Gauss–Legendre technique, whereby the method selects the required function values. It is sometimes easier to calculate the integrals numerically instead of writing routines for special functions that arise with circular apertures. Exact expressions are ideal; but unless a distribution is forced by a mode on the structure, it is difﬁcult to achieve the exact distribution. We need only approximations to the accuracy of practical interest.

4-2 SIMPLE LINEAR DISTRIBUTIONS We assume that rectangular apertures have separable distributions so that we can deal with one coordinate at a time. We compute the pattern in the plane containing the line. By drawing the pattern in kx (or ky )-space, we can calculate patterns independent of the aperture size in a way similar to that used for arrays in Chapter 3. In Chapter 2 we derived the kx -space pattern for a uniform distribution: a sin(kx a/2) kx a/2

(4-10)

where a is the aperture width and kx = k sin θ cos φ. We suppress cos φ and consider only patterns in the φ = 0◦ plane. Figure 4-1 shows the k sin θ space pattern of a uniform distribution. The pattern does not repeat at 2π intervals (radians) as the array does, but the sidelobes continue to decrease at a rate of 1/x. The ﬁrst sidelobe is 13.2 dB below the peak. The aperture size a, along with the scanning variable sin θ0 , determines the visible region in Figure 4-1. It ranges between ±ka/2 centered on ka/2 sin θ0 , since the maximum value of sin θ = 1.

SIMPLE LINEAR DISTRIBUTIONS

141

10

Pattern level, dB

20

30

40

50

60 −5 −25 −100

−4 −20 −80

−3 −15 −60

−2 −10 −40

−1 −5 −20

1 5 20

0

2 10 40

3 15 60

4 20 80

5 25 100

ka (sin q − sin ) 0 2

FIGURE 4-1 kx -space pattern of uniform line-source distribution.

Example An aperture is four wavelengths long. Determine the number of sidelobes between θ = ±90◦ when sin θ0 = 0 (boresight). The maximum value in (k sin θ )-space is 2π 4λ = 4π or λ 2

12.57

There are three sidelobes on each side of the main beam (Figure 4-1) in the visible region. The ﬁrst sidelobe occurs when ka/2 sin θ1 = 4.5, or θ1 = sin−1

4.5λ ◦ = 21 aπ

We found the half-power beamwidth in Chapter 2: HPBW = sin−1

0.4429λ a

(4-11)

valid when we ignore the obliquity factor, (1 + cos θ )/2. When we approximate x = sin x (radians) for small angles, we obtain ◦λ

HPBW = 50.76

a

(4-12)

We use this as the standard and describe other HPBW by their beamwidth factors. The beamwidth factor of the uniform distribution is 1.00. We also consider the null

142

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

beamwidth (BWnull ) of the distribution. The ﬁrst null occurs at ±π in the k sin θ pattern: λ ◦λ BWnull = 2 sin−1 ≈ 114.59 a a We also establish a beamwidth factor for the null beamwidth. When we scan the beam to a direction θ0 , the visible region centers at πa/λ sin θ0 in (k sin θ )-space. Example

Compute beam edges when θ0 = 30◦ and a = 6λ for a uniform distribution. a (sin θ1,2 − sin θ0 ) = ±0.4429 λ ±0.4429 + 0.5 sin θ1,2 = 6 ◦ ◦ θ1 = 35.02 θ2 = 25.23

The beamwidth is the difference, 9.79◦ . If we take the beam center as the average between the 3-dB beam edges, we get 30.12◦ for the beam center. By using the cosine of the beam center times the aperture size, we get 5.19λ, the projected aperture dimension. On substituting this in Eq. (4-11), we calculate HPBW = 9.79◦ . The actual beam peak is at θ = 30◦ , but the pattern is asymmetrical about θ0 . Other simple geometrical distributions on a linear aperture follow the same Fourier transform relation as the uniform distributions with differing transforms in (k sin θ )space. Table 4-1 lists the properties of some common distributions. Example Compute the beamwidth of a 7λ aperture with a cosine distribution. From Table 4-1, the beamwidth factor = 1.342. The taper increases the beamwidth over that of a uniform distribution:

50.76λ 0.4429λ ◦ ◦ HPBW = 1.342 = 9.73 or HPBW = 2 sin−1 1.342 = 9.74 a a We can add distributions and calculate the pattern from the sum of the transforms. Adding a pedestal (uniform distribution) to the cosine-squared distribution decreases TABLE 4-1 Common Linear Distribution Characteristics Distribution Uniform Triangular Cosine Cosine2

fx sin(kx a/2) kx a/2 sin(kx a/4) 2 kx a/4 cos(kx a/2) 2π 2 π − (kx a)2 sin(kx a/2) (kx a/2)[1 − (a/2λ)]2

First Sidelobe (dB)

HPBW Factor

BWnull Factor

ATR (dB)

13.2

1.000

1.000

0

26.5

1.439

2.000

1.25

23.0

1.342

1.5

0.91

31.5

1.625

2.000

1.76

SIMPLE LINEAR DISTRIBUTIONS

143

the beamwidth and the sidelobes of the cosine-squared distribution. The aperture distribution is given by E(x) = PD + (1 − PD) cos2

πx a

|x| ≤

a 2

where PD is the voltage pedestal level. The ﬁrst sidelobe of the uniform distribution lies within the null beamwidth of the cosine-squared distribution. The phase of sidelobes with respect to the main beam alternates between 180◦ and 0◦ , and the sidelobe of the pedestal subtracts from the main lobe. The second sidelobe of the pedestal occurs in almost exactly the same k-space location as the ﬁrst sidelobe of the cosine-squared distribution. These lobes cancel each other to some extent. Table 4-2 gives the required pedestal measured relative to the peak of the distribution for a given maximum sidelobe level. The minimum sidelobes (43.2 dB) occur for a pedestal level of −22.3 dB. At lower pedestal levels, the sidelobes rise and the beamwidth factor increases at a constant rate as the pedestal level decreases. The amplitude taper efﬁciency of the cosine squared on a pedestal is ATL =

2(1 + PD)2 3 + 2PD + 3PD2

(4-13)

ratio

Amplitude distributions based on simple functions have limited use. The uniform and cosine distributions or close approximations occur naturally, but the others must be forced on an aperture. An array can sample a distribution to achieve results similar to those for an aperture. A sampled cosine squared on a pedestal is handy for quick tolerance studies of array feed networks, but is far from optimum. Table 4-2 lists the pedestal to achieve a given sidelobe level for this distribution. We consider distributions that allow close control of sidelobes and achieve minimum beamwidths. The rate of decrease of the far-out sidelobe depends on the functional relation of the distribution at the edges [1]. If α is the exponent of the distribution approximation xeα , where xe is the distance from the edge, then the sidelobes decay as U −(1+α) , where U TABLE 4-2 Pedestal Level to Achieve a Given Maximum Sidelobe Level for a Cosine Squared on a Pedestal Distribution Sidelobe (dB)

Pedestal (dB)

Beamwidth Factor

ATL (dB)

30 32 34 36 38 40 42 42.7a 43.2b

−12.9 −14.2 −15.7 −17.3 −18.7 −20.0 −21.4 −21.9 −22.3

1.295 1.325 1.357 1.390 1.416 1.439 1.463 1.471 1.476

0.79 0.89 0.99 1.10 1.18 1.25 1.32 1.34 1.36

a b

Hamming distribution. Minimum sidelobe level.

144

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

is a linear function of the k-space variable. Both the triangular and cosine distributions have α = 1, and the far-out sidelobes decay as 1/U 2 . The cosine-squared distribution sidelobes decay as 1/U 3 , since α = 2. In the case of a cosine squared on a pedestal, the edge functional relation is a step (pedestal, α = 0) and the far-out sidelobes decay as 1/U . The sidelobes of the pedestal eventually overtake the cosine-squared distribution sidelobes, decreasing as 1/U 3 . To achieve uniform sidelobes, α must be −1, which occurs only when the distribution edges are Dirac delta functions, which requires inﬁnite energy in the aperture or design reduction to discrete sources (an array). We must accept a trade-off between radiated power in the main beam and in the sidelobes. When we narrow the main beam in a ﬁxed size aperture, more power radiates in the sidelobes. We achieve minimum beamwidth in the main beam when all the sidelobes radiate the same power (maximum radiated power in the sidelobes for a given level) and all sidelobes are at the same level. This case leads to the Dolph–Chebyshev array [2], impossible to duplicate in a continuous aperture. 4-3 TAYLOR ONE-PARAMETER LINEAR DISTRIBUTION [3] The uniform distribution has k-space zeros at (Figure 4-1) ±nπ, n = 1, 2, 3, . . . . Taylor deﬁnes a new variable U to replace k sin θ : sin πU πU

(4-14)

where U = (a/λ)(sin θ − sin θ0 ) and a is the aperture width. The nulls (zeros) are then located at integer values of U . Taylor adjusted the inner zeros of the uniform distribution to lower the sidelobes while retaining the outer zeros at their locations in the uniform distribution. The zeros are modiﬁed by a parameter B the boundary between the two regions in U -space: Un = n 2 + B 2 (4-15) The pattern has different expressions in two regions: √ B2 − U 2 sinh π √ |U | ≤ B π B2 − U 2 F (U ) = √ sin π U 2 − B 2 |U | ≥ B √ π U 2 − B2

(4-16a) (4-16b)

The high value of Eq. (4-16a) at the boresight depresses the sidelobes of the uniform distribution, and the parameter B controls all the parameters of the distribution. We compute B from the desired sidelobe level (SLR) by an iterative solution of the equation SLR = 13.26 + 20 log

sinh πB πB

(4-17)

Scale 4-1 gives the Taylor single-parameter distribution B for a given sidelobe level. The aperture distribution over the range −0.5 to 0.5 is given by the equation I0 [πB 1 − (2x)2 ] (4-18) I0 (πB)

TAYLOR ONE-PARAMETER LINEAR DISTRIBUTION

145

Design Sidelobe Level, dB

Taylor Single Parameter, B

SCALE 4-1 Taylor single-parameter B for a sidelobe level.

Design Sidelobe Level, dB

Taylor Single Parameter Edge Taper, dB

SCALE 4-2 Taylor single-parameter edge taper for a given sidelobe level.

Design Sidelobe Level, dB

Taylor Single Parameter Taper Loss, dB

SCALE 4-3 Taylor single-parameter amplitude taper loss for a given sidelobe level.

Design Sidelobe Level, dB

Taylor Single Parameter HPBW Factor

SCALE 4-4

Taylor single-parameter HPBW factor for a given sidelobe level.

Using Eq. (4-18), we calculate aperture edge taper as a function of sidelobe level, given by Scale 4-2. By inserting the expression for the aperture distribution [Eq. (4-18)] into Eq. (4-6), we calculate amplitude taper loss as a function of sidelobe level (Scale 4-3). The HPBW factor can be found from Eq. (4-16) or read easily from Scale 4-4. Figure 4-2 compares the U -space patterns of the Taylor one-parameter and uniform distributions. Synthesizing aperture distributions and arrays concentrates on the placement of pattern nulls. The one-parameter distribution scaled the locations of the nulls (zeros) by using Eq. (4-15). You should notice that the nulls approach those of the

146

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

30-dB Taylor One-Parameter Line Distribution

Pattern, dB

Uniform Distribution

U = a/l sin q

FIGURE 4-2 U -space pattern of 30-dB Taylor one-parameter linear distribution versus uniform distribution.

uniform distribution as U increases. Except for a shift near U = 0, the pattern falls off at a 1/U rate for far-out sidelobes. You can use the one-parameter Taylor distribution to estimate the characteristics of a linear distribution for a given sidelobe level. A comparison of this distribution to the cosine squared on a pedestal (Table 4-3) shows that it is not as efﬁcient for moderate sidelobe levels. The cosine squared on a pedestal distribution achieves low sidelobes by canceling sidelobes in two distributions and cannot be extended to any sidelobe level, whereas the one-parameter distribution can produce designs for any sidelobe level. More important, it demonstrates the systematic use of U -space pattern null placement for design. Taylor improved on this distribution by considering the zeros of the Dolph–Chebyshev array to ﬂatten the ﬁrst few sidelobes of the pattern response and achieved a more efﬁcient distribution.

TABLE 4-3 Comparison Between the Taylor One-Parameter Distribution and Cosine Squared on Pedestal Linear Distribution for Selected Sidelobe Levels Distribution

Pedestal (dB)

ATR (dB)

HPBW Factor

30-dB 30-dB 36-dB 36-dB 40-dB

−21.13 −12.9 −28.49 −17.3 −32.38

0.96 0.79 1.30 1.10 1.49

1.355 1.295 1.460 1.390 1.524

one-parameter cos2 + pedestal one-parameter cos2 + pedestal one-parameter

TAYLOR n LINE DISTRIBUTION

147

4-4 TAYLOR n LINE DISTRIBUTION [1] The Taylor n line-source distribution modiﬁes the location of the inner pattern zeros (nulls) of a uniform distribution to approximate the Dolph–Chebyshev array. The distribution contains a pedestal α = 0 and retains the 1/U fall-off for the far-out sidelobes. We can modify any number of inner zeros of the pattern to approximate the uniform sidelobe-level array, but we force the aperture voltage to peak at the ends in approximating the Dirac delta functions. We limit the number of altered zeros to keep the distribution practical. After a point, shifting more zeros reduces beamwidth negligibly. We manipulate the location of pattern zeros to obtain desired patterns. Both aperture and array syntheses depend on zero locations. The number of array elements determines the number of independent zeros (n − 1), but a continuous aperture has an inﬁnite number of independent zeros. Practical consideration of the distribution edge shape limits the number, but we are free to move zeros. For a given aperture size, we can move zeros out of the invisible region into the visible region and narrow the main beam as much as we want while maintaining low sidelobes. The invisible region represents stored energy in the aperture. When a zero moves out of the invisible region, the amount of stored energy and the Q of the antenna increase. The overall efﬁciency of the antenna decreases while the antenna becomes more and more narrowband. We call these arrays superdirective because their directivity exceeds that of a uniform distribution. The Taylor line-source distribution retains the zeros in the invisible region and prevents superdirectivity. There is no limit to the directivity achievable on paper for a given aperture, but the theoretical distributions are unrealizable except for very small levels of superdirectivity. The costs of superdirectivity are decreased bandwidth and efﬁciency. We will modify the location of the ﬁrst n − 1 pairs of inner nulls to lower the sidelobes. Choosing the zeros symmetrically about the origin of U -space gives us a constant phase distribution. We remove the inner zeros by dividing them out of the uniform distribution U -space pattern: sin πU n−1 πU 1 − U 2 /N 2 N=1

We then add new nulls Un without becoming superdirective: n−1

U2 1− 2 sin πU N=1 UN F (U ) =

n−1 U2 πU 1− 2 N=1 N

(4-19)

Because we want to approximate the DoIph–Chebyshev response, we choose the inner zeros from the array: A2 + (N − 12 )2 UN = n A2 + (n − 12 )2

N = 1, 2, . . . , n − 1

(4-20)

148

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

where A relates to the maximum sidelobe level: cosh πA = b

(4-21)

in which 20 log b = sidelobe level. Equation (4-19) gives us the U -space (k-space) pattern of the distribution with modiﬁed zeros. We determine the aperture distribution by expanding it in a Fourier cosine series: E(x) =

∞

Bm cos 2mπx

|x| ≤ 0.5

(4-22)

m=0

where the aperture size has been normalized. We calculate the pattern of the distribution from the Fourier transform: 1/2 1/2 j kx x f (kx ) = E(x)e dx or f (U ) = E(x)ej 2πU x dx −1/2

−1/2

We substitute Eq. (4-22) for E(x) and reverse the order of summation and integration: ∞

f (U ) =

Bm

m=0

1/2

−1/2

cos 2mπx cos 2πU x dx

(4-23)

Since the aperture function is an even function, the odd-function part of the integral is zero, as reﬂected in Eq. (4-23). We calculate coefﬁcients Bm by matching the patterns at integer values of U . The integral [Eq. (4-23)] is zero unless U = m: B0 = f (0)

Bm = f (m) 2

m = 1, 2, . . . , n − 1

Since we have only modiﬁed the location of the ﬁrst n − 1 zeros of the U -space pattern, f (m) = 0 for m ≥ n and the Fourier cosine series has only n components: E(x) = f (0) + 2

n−1

f (m) cos 2mπx

(4-24)

m=1

The coefﬁcients are given by f (0) = 1

n−1 (−1)m 1 − m2 /UN2 N=1 f (m) = n−1 −2 1 − m2 /N 2

m = 1, 2, . . . , n − 1

(4-25)

N=1,N =m

Equation (4-19) computes the U -space pattern of the Taylor distribution but requires L’Hospital’s rule at integer values of U . The ﬁnite number of coefﬁcients Bm makes Eq. (4-23) more convenient since the integral is easily solved: n−1 sin(πU ) 1 sin[π(U − i)] sin[π(U + i)] Bi + + f (U ) = B0 πU 2 i=1 π(U − i) π(U + i)

(4-26)

TAYLOR n LINE DISTRIBUTION

149

Example Design the Taylor line-source distribution with 30-dB maximum sidelobes and n = 6. We use Eq. (4-21) to calculate A: b = 1030/20 = 31.6228 A=

cosh−1 b = 1.3200 π

We substitute this constant into Eq. (4-20) to compute the ﬁve (n − 1) nulls: No. Null UN

1

2

3

4

5

1.4973

2.1195

2.9989

3.9680

4.9747

The ﬁrst null value gives us the BWnull factor (1.4973). The null beamwidth has been increased almost 50% relative to the uniform distribution. The coefﬁcients of the Fourier cosine aperture distribution are found from Eqs. (4-24) and (4-25) (Table 4-4). Coefﬁcients of the series are normalized so that the distribution is 1 at x = 0, and the amplitude distribution is found by plotting the Fourier cosine series. We calculate the U -space pattern by using Eq. (4-26). We calculate the half-power point and compare it to the uniform distribution to determine the HPBW factor, 1.2611. By using Eq. (4-6), we calculate ATL = 0.66 dB for the distribution. The U -space plot (Figure 4-3) of the example above shows the 30-dB sidelobe level. The ﬁrst sidelobe is at 30 dB, and lobes after that fall away from 30 dB. With a higher value of n, the ﬁrst unchanged zero, more sidelobes would be nearer 30 dB. The dashed curve gives the pattern of a uniform distribution. Notice that the inner ﬁve nulls have been shifted to lower the sidelobes. At the sixth null and higher, the Taylor distribution has the same nulls as the uniform distribution. The n distribution has a narrower beamwidth than the one-parameter distribution (Figure 4-2) and a higher taper efﬁciency of 0.66 dB versus 0.96 dB. Figure 4-4 shows the normalized aperture voltage for 30-dB-maximum sidelobe Taylor distributions. The one-parameter design produces a lower pedestal than the two n designs. The n = 20 design voltage peaks as it approaches the edge. This peaks because the Taylor n distribution approximates the Dolph–Chebyshev array that peaks at the edge of the array. The amplitude taper efﬁciency was calculated for a number of designs and is given in Table 4-5. The corresponding beamwidth factors are listed in Table 4-6 together with TABLE 4-4 Fourier Cosine Series Coefﬁcients for Taylor Distribution: 30 dB, n = 6 No. 1 2 3 4 5 6

Bm

Bm Normalized

Function

1.0000 0.5733 −0.0284 −0.000213 0.005561 −0.003929

0.64672 0.37074 −0.01838 −0.000138 0.003597 −0.002541

1 cos 2πx cos 4πx cos 6πx cos 8πx cos 10πx

150

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

30-dB Taylor Line Distribution

Pattern, dB

Uniform Distribution

U = a/l sin q

Normalized Voltage

FIGURE 4-3 U -space pattern of 30-dB Taylor n = 6 linear distribution versus uniform distribution.

12 One Parameter

6

Normalized Aperture

FIGURE 4-4 A 30-dB Taylor linear aperture distribution comparison.

the null beamwidth factors (location of ﬁrst zero in U -space) in Table 4-7. ATL depends on the sidelobe level (Table 4-5) more than the number of modiﬁed zeros. Both the 20and 25-dB sidelobe levels show that there is an optimum number of zeros. The edge of the distribution peaks toward the Dirac delta function and reduces the amplitude taper efﬁciency. More than three modiﬁed zeros are needed to reduce the sidelobes

TAYLOR n LINE DISTRIBUTION

151

TABLE 4-5 Amplitude Taper Losses of Taylor Line-Source Distributions Sidelobe Level (dB)

TABLE 4-6

n

20

25

30

35

40

45

50

4 5 6 7 8 10 12 16 20

0.17 0.15 0.15 0.15 0.16 0.19 0.24 0.35 0.46

0.43 0.41 0.39 0.37 0.36 0.34 0.34 0.35 0.27

0.69 0.68 0.66 0.65 0.63 0.61 0.59 0.57 0.56

0.95 0.93 0.92 0.91 0.90 0.88 0.86 0.84 0.82

1.16 1.15 1.15 1.14 1.13 1.11 1.09 1.07

1.37 1.36 1.36 1.35 1.34 1.32 1.30

1.56 1.55 1.55 1.54 1.53 1.51

Beamwidth Factor of Taylor Line-Source Distribution Sidelobe Level (dB)

n

20

25

30

35

40

45

50

4 5 6 7 8 10 12 16 20

1.1043 1.0908 1.0800 1.0715 1.0646 1.0545 1.0474 1.0381 1.0324

1.1925 1.1837 1.1752 1.1679 1.1617 1.1521 1.1452 1.1358 1.1299

1.2696 1.2665 1.2611 1.2555 1.2504 1.2419 1.2353 1.2262 1.2203

1.3367 1.3404 1.3388 1.3355 1.3317 1.3247 1.3189 1.3103 1.3044

1.4065 1.4092 1.4086 1.4066 1.4015 1.3967 1.3889 1.3833

1.4733 1.4758 1.4758 1.4731 1.4695 1.4628 1.4576

1.5377 1.5400 1.5401 1.5379 1.5326 1.5280

TABLE 4-7

Null Beamwidth Factor of Taylor Line-Source Distributions Sidelobe Level (dB)

n

20

25

30

35

40

45

50

4 5 6 7 8 10 12 16 20

1.1865 1.1696 1.1566 1.1465 1.1386 1.1270 1.1189 1.1086 1.1023

1.3497 1.3376 1.3265 1.3172 1.3095 1.2978 1.2894 1.2783 1.2714

1.5094 1.5049 1.4973 1.4897 1.4828 1.4716 1.4632 1.4518 1.4444

1.6636 1.6696 1.6671 1.6632 1.6569 1.6471 1.6392 1.6277 1.6200

1.8302 1.8347 1.8337 1.8306 1.8231 1.8161 1.8051 1.7975

1.9990 2.0031 2.0032 1.9990 1.9934 1.9835 1.9760

2.1699 2.1739 2.1740 2.1705 2.1623 2.1553

below 40 dB; hence, the blanks represent unrealizable designs. The beamwidth factor (Table 4-6) reduces with increasing n, but it depends mainly on the sidelobe level. Example Compute beamwidths and ATL of an 8λ-wide aperture with n = 8, 40-dB sidelobes, and a Taylor line-source distribution design.

152

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

From Table 4-6, HPBW = From Table 4-7, BWnull =

1.4066(50.76◦ ) ◦ = 8.92 8

1.8306(114.59◦ ) ◦ = 26.22 8

From Table 4-5, ATL = 1.14 dB. A square aperture with the same distribution in both directions has directivity = 10 log

4πA − 2ATL = 26.77 dB λ2

4-5 TAYLOR LINE DISTRIBUTION WITH EDGE NULLS Rhodes [4] has shown that it is impossible to have a step discontinuity of the ﬁelds at the edge of a physical aperture. Given the radius of curvature of the edge, ρ, the ﬁeld varies as C2 d ρ C1 2 d Es ∼ ρ

Ed ∼

polarized perpendicular to the aperture edge polarized parallel to the aperture edge

where C1 and C2 are constants and d is the distance from the edge. Without the possibility of an edge pedestal, a traditional Taylor line source cannot be realized with a physical aperture. We can sample the distribution with an array or closely approximate it, but we cannot achieve the exact distribution. A Taylor distribution with a null at the edge can be realized in an aperture. Rhodes [5] extended the Taylor line source by modifying the U -space pattern zeros of the cosine distribution. Since α = 1, the far-out sidelobes drop off as 1/U 2 and the distribution is zero on the edges. The zeros of the cosine distribution occur at (N + 1/2)π

N = 1, 2, 3, . . .

k space

When the Taylor U -space variable is used, the modiﬁed U -space pattern becomes n−1 1 − U 2 /UN2 N=1 f (U ) = n−1 2 [1 − (2U ) ] (1 − U 2 /(N + 12 )2 ) cos πU

(4-27)

N=1

We remove the inner n − 1 zeros at N +

1 2

and substitute new ones given by

UN = ±(n +

1 ) 2

A2 + (N − 12 )2 A2 + (n − 12 )2

N = 1, 2, . . . , n − 1

(4-28)

TAYLOR LINE DISTRIBUTION WITH EDGE NULLS

153

When we compare Eqs. (4-28) and (4-20), we see that the nulls are shifted by (n + 12 )/n between the two Taylor distributions. When n is large, the nulls are close to the same for the two distributions. To determine the amplitude distribution in the aperture, we expand the aperture ﬁelds in a Fourier cosine series, E(x) =

Bm cos(2m + 1)πx

|x| ≤ 0.5

(4-29)

Like the Taylor line source, there are only n terms in the series whose coefﬁcients are found by equating the pattern from the Fourier transform of Eq. (4-29) to Eq. (4-27). The coefﬁcients are given by n−1 2 1 − 14 /UN2 B0 = n−1 N=1 1 − 14 /(N + 12 )2 N=1

n−1 (−1)m (m + 12 ) 1 − (m + 12 )2 /UN2 N=1 Bm = n−1 [1 − (2m + 1)2 ] 1 − (m + 12 )2 /(N + 12 )2

m = 1, 2, . . . , n − 1

N=1,N =m

(4-30)

The U -space pattern can be found using the coefﬁcients Bm : f (U ) = C0

n−1

Bi

sin[π(U − i − 12 )] π(U − i − 12 )

i=0

C0 = 2

n−1 i=0

Bi

+

sin[π(U + i + 12 )]

π(U + i + 12 )

(4-31)

sin[π(i + 12 )] π(i + 12 )

Example Design the Taylor line-source distribution with edge nulls for 30-dB maximum sidelobes and n = 6. We use Eq. (4-21) to calculate A: b = 1030/20 = 31.6228 A=

cosh−1 b = 1.32 π

the same as the pedestal edge Taylor line-source distribution. We substitute this constant into Eq. (4-28) to compute the ﬁve modiﬁed nulls: No. Null UN

1

2

3

4

5

1.6221

2.2962

3.2488

4.2987

5.3892

The null locations have increased by (n + 12 )/n = 6.5/6 = 1.0833 from the pedestal Taylor line-source design. The null beamwidth factor has also increased by this factor

154

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

as well. The coefﬁcients of the Fourier cosine aperture distribution are found from Eq. (4-30) Table 4-8. The normalized coefﬁcients sum to 1 at x = 0. Equation (4-27) determines the U space pattern given the nulls. On ﬁnding the half-power point and comparing it with the uniform distribution half-power point, we compute the beamwidth factor: 1.3581. Tables 4-9 to 4-11 give results for this Taylor line source. As n increases, the results approach the result of the pedestal Taylor line source. Since the maximum sidelobe of the cosine distribution is 23 dB, a distribution must have peaking toward the edges to raise the sidelobes above that level. In all distributions the voltage approaches zero linearly at the edges. TABLE 4-8 Fourier Cosine Series Coefﬁcients for Taylor Distribution with Edge Nulls: 30 dB, n = 6 Bm

Bm Normalized

Function

0.50265 0.023087 0.017828 −0.010101 0.007374 −0.003245

0.94725 0.04351 0.02220 −0.02075 0.01390 −0.006116

cos πx cos 3πx cos 5πx cos 7πx cos 9πx cos 11πx

No. 1 2 3 4 5 6

TABLE 4-9 Amplitude Taper Losses of a Taylor Line Source with Edge Null Distribution Sidelobe Level (dB) n

25

30

35

40

45

50

4 6 8 12 16 20

0.86 0.67 0.56 0.45 0.41 0.39

1.13 0.97 0.87 0.74 0.68 0.64

1.36 1.24 1.14 1.02 0.96 0.92

1.55 1.47 1.39 1.28 1.22 1.17

1.71 1.66 1.60 1.51 1.45 1.41

1.84 1.84 1.79 1.71 1.66 1.62

TABLE 4-10 Beamwidth Factor of a Taylor Line Source with Edge Null Distribution Sidelobe Level (dB) n

25

30

35

40

45

50

4 6 8 12 16 20

1.3559 1.2666 1.2308 1.1914 1.1705 1.1576

1.4092 1.3581 1.3242 1.2850 1.2635 1.2502

1.4815 1.4407 1.4097 1.3716 1.3500 1.3363

1.5443 1.5153 1.4882 1.4522 1.4308 1.4170

1.5991 1.5831 1.5608 1.5276 1.5068 1.4930

1.6470 1.6448 1.6280 1.5984 1.5785 1.5649

155

ELLIOTT’S METHOD FOR MODIFIED TAYLOR DISTRIBUTION

TABLE 4-11 Null Beamwidth Factor of a Taylor Line Source with Edge Null Distribution Sidelobe Level (dB) n

25

30

35

40

45

50

4 6 8 12 16 20

1.5184 1.4371 1.3913 1.3431 1.3182 1.3031

1.6980 1.6221 1.5755 1.5242 1.4971 1.4805

1.8715 1.8060 1.7604 1.7075 1.6786 1.6605

2.0374 1.9875 1.9450 1.8918 1.8616 1.8424

2.1949 2.1656 2.1284 2.0765 2.0455 2.0254

2.3433 2.3395 2.3097 2.2610 2.2298 2.2091

4-6 ELLIOTT’S METHOD FOR MODIFIED TAYLOR DISTRIBUTION AND ARBITRARY SIDELOBES [6, pp. 162–165] Elliott’s method separates the distribution nulls into right- and left-hand values in U space that allows different sidelobe levels in the two regions. By applying a differential expression, the null positions in U -space can be found from the solution of a set of linear equations to produce designs with arbitrary sidelobes. Consider Eq. (4-19) and factor the null location term:

U2 U U U U 1− 2 = 1+ 1− = 1+ 1− UN UN UNL UNR UN We associate UNL with a nulls on the left side of the origin and UNR with the right side or a positive pattern angle. If we also separate the term in the denominator of Eq. (4-19), we can independently pick the number of nulls to be moved on either side of the pattern:

F (U ) = C0

nL −1

nR −1 (1 + U/UN ) (1 − U/UN ) N=1 N=1 nL −1 nR −1 πU (1 + U/N ) (1 − U/N )

sin πU

N=1

(4-32)

N=1

Equation (4-32) allows different Taylor distributions on the two sides. We add a normalization factor C0 when we use different distributions. The pattern peak will shift off zero for unbalanced distributions. Since the two sides are not independent, a simple selection of the two levels will not produce the desired sidelobes. Table 4-12 lists the U -space locations of the pattern peaks and sidelobe level for a design with 35- and 30-dB sidelobes. The left distribution lowered the sidelobes on the right and the right one raised the left sidelobes. A few manual iterations produced suitable left and right distributions to give the desired sidelobes. The main beam shifts a little bit. A linear progressive phase shift across the aperture can shift the pattern to broadside. We expand the aperture distribution in a complex exponential series similar to Eq. (4-22): n R −1 E(x) = Bi e−j 2πix |x| ≤ 0.5 (4-33) i=−nL +1

156

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

TABLE 4-12 Modiﬁed Taylor Distribution Sidelobes for Independent Left and Right Sidelobe Design Using n = 6 for Both Sides Left-Side−35-dB U -Space

Right Side−30-dB Sidelobe (dB)

Left-Side−36-dB U -Space

Right-Side−28.6-dB Sidelobe (dB)

−5.4849 −4.4905 −3.5275 −2.6313 −1.8997 −0.0511 1.7546 2.5372 3.4697 4.4584 5.4711

−35.70 −35.06 −34.66 −34.44 −34.39 0 −31.05 −31.45 −32.00 −32.74 −33.75

−5.4874 −4.4976 −3.5399 −2.6510 −1.9293 −0.0758 1.7141 2.5119 3.4544 4.4498 5.4676

−36.00 −35.44 −35.11 −34.97 −34.99 0 −30.05 −30.55 −31.19 −32.02 −33.13

We calculate the coefﬁcients by the same method used for Eq. (4-25): B(0) = 1

nR −1 nL −1 (−1)|m| N=1 (1 + m/UN ) (1 − m/UN ) N=1 B(m) = n −1 nR −1 L − (1 + m/N ) (1 − m/N ) N=1,N =m

N=1,N =m

m = −nL + 1, . . . , −1, 1, 2, . . . , nR − 1

(4-34)

We derive the pattern from the integral of the ﬁnite complex exponential: f (U ) = C0

n R −1 i=−nL +1

Bi

sin[π(U − i)] π(U − i)

(4-35)

We include the normalization factor C0 for unequal left and right sidelobes. We control the sidelobes by adjusting the location of the nulls in the U -space pattern. We can iterate the null positions to produce individually selected sidelobes. The peak of each sidelobe given in Table 4-12 was found through a one-dimensional search between pairs of nulls. A search based on the Fibonacci numbers [7, p. 280] computes the peak with the minimum number of evaluations of the pattern using Eq. (4-35). p We denote the pattern peaks by Um starting with the peak between −nL and U−nL +1 , those between nulls, the peak near 0, and the last peak between UnR −1 and nR for nL + nR − 1. We adjust the U -space nulls by the differentials δUN found from the solution of a matrix equation. The terms of the matrix are the differential term of a Taylor series expansion of the numerator of Eq. (4-32) evaluated at the pattern peaks: p

am,n

Um /UN2 = p 1 − Um /UN

am,0 = 1

N = −nL + 1, . . . , −1, 1, . . . , nR − 1 (4-36)

157

ELLIOTT’S METHOD FOR MODIFIED TAYLOR DISTRIBUTION

The vector of differential nulls is δU = [δU−nL +1 , . . . , δU−1 , δC/C0 , δU1 . . . , δUnR −1 ]T where δC/C0 is the change in the pattern normalization. We form a vector using the ratio of the desired pattern peak fd (U ) to the actual pattern fa (U ) with terms p p [fd (Um )/fa (Um )] − 1. We solve the matrix equation for the null shifts: [am,n ]

−1

T p fd (Um ) = [δUN ] p −1 fa (Um )

(4-37)

We calculate new distribution nulls UN + δUN , substitute the new nulls into Eq. (4-34) to determine the new expansion coefﬁcients Bm , and evaluate the pattern using Eq. (4-35) between the new nulls to compute new pattern peaks. We iterate the process until the sidelobe levels are satisfactory. Notice that f (U ) is a voltage. The Taylor linear distribution produces a pattern with only approximately equal sidelobes. Table 4-13 lists the iteration to produce a distribution with a pattern that has ﬁve 30-dB sidelobes. The solution starts with a 30-dB, n = 6 Taylor distribution. In two iterations the method found a distribution with exactly the desired sidelobes. Table 4-14 gives the results of repeating the example of Table 4-12 of the design for 35and 30-dB sidelobes. This method can produce a linear distribution with individually TABLE 4-13

Iteration of Distribution Nulls for a Pattern with 30-dB Sidelobes

Taylor Distribution Null 1.4973 2.1195 2.9989 3.9680 4.9747

First Iteration

Second Iteration

U -Space

Sidelobe

Null

U -Space

Sidelobe

Null

U -Space

Sidelobe

1.7557 2.5387 3.4709 4.4591 5.4718

−30.22 −30.46 −30.89 −31.53 −32.48

1.4708 2.0827 2.9490 3.9075 4.9145

1.7258 2.4987 3.4215 4.4072 5.4424

−30.00 −29.99 −29.96 −29.86 −29.63

1.4729 2.0859 2.9541 3.9152 4.9242

1.7284 2.5027 3.4274 4.4147 5.4471

−30.00 −30.00 −30.00 −30.00 −29.99

TABLE 4-14

Iteration for 35- and 30-dB Sidelobes in Linear Distribution

Left-Side−36-dB U -Space −5.4874 −4.4976 −3.5399 −2.6510 −1.9293 −0.0758 1.7141 2.5119 3.4544 4.4498 5.4676

Right-Side−28.6-dB Sidelobe (dB) −36.00 −35.44 −35.11 −34.97 −34.99 0 −30.05 −30.55 −31.19 −32.02 −33.13

Second Iteration Null

U -Space

Sidelobe (dB)

Null

−4.9964 −4.0169 −3.0845 −2.2583 −1.7008

−5.4798 −4.4859 −3.5348 −2.6555 −1.9410 −0.1037 1.6615 2.4475 3.3839 4.3839 5.4313

−35.00 −35.00 −35.00 −35.00 −35.00 0 −30.00 −30.00 −30.00 −30.00 −29.98

−4.9778 −4.0043 −3.0829 −2.2670 −1.7143

1.4495 2.0883 2.9801 3.9574 4.9699

1.4023 2.0249 2.9049 3.8780 4.9001

158

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

selected sidelobes. It may be necessary to design an intermediate distribution if the change in sidelobes is too great for the simple iteration scheme to converge. We will design linear arrays by a similar technique of manipulating pattern nulls to produce arbitrary sidelobes.

4-7 BAYLISS LINE-SOURCE DISTRIBUTION [8] The Bayliss distribution produces a pattern null on a boresight while controlling the height of the sidelobes. The second dashed curve in Figure 4-5 below is a Bayliss difference pattern also designed to give 30-dB sidelobes when combined with the Taylor distribution. As in the Taylor distribution, the ﬁrst few sidelobes are nearly the same height, to minimize the beamwidth of the two beams split about the boresight. A monopulse tracking system uses an auxiliary pattern with a boresight null coincident with the beam peak of the main pattern. The tracking system drives the antenna positioner until the signal in this difference channel nulls so that the main channel (sum) points at the emitter or radar target. The accuracy of the pointing angle is improved, since a null is a more exact direction than the broad sum pattern peak. Noise and receiver sensitivity, along with the slope of the difference pattern, limit the tracking accuracy. Stronger signals can be tracked farther into the null. Because the phase of a pattern shifts by 180◦ when passing through a null, phase relative to the sum pattern (a reference signal) can be used to give direction. Without monopulse or some other sequential lobing technique, such as conical scan, radar cannot track effectively.

10

20

Pattern, dB

Taylor 30 dB

Bayliss 28 dB

30

40

50

60 −20

0 ka (sin q − sin ) 0 2

FIGURE 4-5

Taylor and Bayliss line distributions to give 30-dB sidelobes (n = 6).

20

BAYLISS LINE-SOURCE DISTRIBUTION

159

Any odd-function distribution produces a null on a boresight. A uniform distribution that switches phase by 180◦ in the center has the best amplitude taper efﬁciency but high sidelobes (10 dB). These high sidelobes allow interfering or noise signals to enter the receiver. The Bayliss distribution adjusts the inner nulls of the U -space pattern to lower the sidelobes. Adjusting the zeros to correspond to the Dolph–Chebyshev array does not lower the sidelobes to the same level as it did in the Taylor distribution. Further adjustments of the four inner zeros are required. Bayliss found the proper location through a computer search. We locate the zeros by ξN2 (n + 12 ) N = 1, 2, 3, 4 A2 + n2 UN = (4-38) 2 2 A + N 1 N = 5, 6, . . . , n − 1 (n + 2 ) A2 + n2 By using the U -space pattern, we have n−1

1 − U 2 /UN2

f (U ) = U cos πU n−1N=1 1 − U 2 /(N + 12 )2

(4-39)

N=0

The coefﬁcients were ﬁtted to polynomials depending on the sidelobe level. Given S = |sidelobe level(dB)|: A = 0.3038753 + S(0.05042922 + S(−0.00027989 + S(0.343 × 10−5 − S(0.2 × 10−7 ))))

(4-40a)

ξ1 = 0.9858302 + S(0.0333885 + S(0.00014064 + S(−0.19 × 10−5 + S(0.1 × 10−7 ))))

(4-40b)

ξ2 = 2.00337487 + S(0.01141548 + S(0.0004159 + S(−0.373 × 10−5 + S(0.1 × 10−7 ))))

(4-40c)

ξ3 = 3.00636321 + S(0.00683394 + S(0.00029281 + S(−0.161 × 10−5 )))

(4-40d)

ξ4 = 4.00518423 + S(0.00501795 + S(0.00021735 + S(−0.88 × 10−6 )))

(4-40e)

The location of the pattern peak was also ﬁtted to a polynomial: Umax = 0.4797212 + S(0.01456692 + S(−0.00018739 + S(0.218 × 10−5 + S(−0.1 × 10−7 ))))

(4-41)

We obtain the aperture distribution by a Fourier sine series having only n terms: |x| ≤ 0.5 (4-42) E(x) = Bm sin(m + 12 )2πx

160

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

where

n−1 1 − (m + 12 )2 /UN2

(−1)m (m + 12 )2 Bm = 2j

N=1

n−1

(4-43)

1 − (m + 12 )2 /(N + 12 )2

N=0,N =m

The phase constant (−j ) has little effect on the coefﬁcient Bm except to balance the phase ±90◦ about the null. Example Design a Bayliss distribution with 30-dB sidelobes and n = 6. Use Eq. (4-40) to compute the coefﬁcients. A = 1.64126

ξ1 = 2.07086

ξ2 = 2.62754

ξ3 = 3.43144

ξ4 = 4.32758

We substitute these constants into Eq. (4-38) to calculate the ﬁve (n − 1) nulls: No. Null UN

1

2

3

4

5

2.1639

2.7456

3.5857

4.5221

5.4990

Equation (4-41) computes the beam peak of the split-beam pattern in U space: Umax = 0.7988

ka sin θmax = πUmax = 2.5096 2

where a is the aperture width. We substitute these zeros into Eq. (4-43) to determine the coefﬁcients of the Fourier sine series of the aperture distribution (Table 4-15). By evaluating the series across the aperture, the coefﬁcients can be normalized to give a maximum aperture voltage of 1. We use Eq. (4-39), after substituting the zeros, to evaluate the pattern. The 3-dB pattern points can be found by searching the pattern: ka sin θ1 = 1.27232 2

ka sin θ2 = 4.10145 2

Figure 4-5 contains the plot of a Bayliss distribution (n = 6) designed to have sidelobes 30 dB below the Taylor distribution with 30-dB sidelobes. The losses to the difference pattern are about 2 dB higher than the sum pattern. We design the Bayliss TABLE 4-15 Fourier Cosine Series Coefﬁcients for Bayliss Distribution: 30 dB, n = 6 No. 0 1 2 3 4 5

Bm

Bm Normalized

Function

0.13421 0.081025 −0.0044151 0.001447 −0.0003393 −0.000014077

0.85753 0.51769 −0.028209 0.0092453 −0.0021679 −0.00008994

sin sin sin sin sin sin

πx 3πx 5πx 7πx 9πx 11πx

BAYLISS LINE-SOURCE DISTRIBUTION

161

distribution to have 28-dB sidelobes. If designed for 30-dB sidelobes as in the example above, then, relative to the sum Taylor distribution, the sidelobes would be 32 dB down from the sum pattern peak. The last nulls show that the unmodiﬁed zeros of the Taylor distribution occur at ±nπ, whereas the unmodiﬁed zeros of the Bayliss distribution occur at ±(n + 12 )π. By using Eq. (4-6), we calculate amplitude taper efﬁciency of the pattern at the beam peak. When we evaluate the phase error efﬁciency by using Eq. (4-7), the result is zero because of the boresight null. We use Eq. (4-3) to evaluate the phase error efﬁciency at the beam peak:

PEL =

a/2 −a/2

2 E(x)ej k sin θmax x dx

a/2 −a/2

(4-44)

2 |E(x)| dx

Table 4-16 lists results of calculations on Bayliss distributions with n = 10 for various sidelobe levels. Lower sidelobe levels produce higher distribution losses and push the beam peak out. The position of the beam peak is independent of n, since the ﬁrst four zeros are ﬁxed by Eq. (4-40). Like the Taylor distribution, the sidelobe level determines most of the parameters of the Bayliss distribution. Changing n has less effect than it has for the Taylor distribution. The values of parameters for distribution with n = 10 will differ little from those in Table 4-16. Example Compute the beam peak and beam edges for an 8λ-wide aperture excited in a Bayliss distribution with n = 10 and 30-dB sidelobes. 8λ 2π sin θmax = 2.5096 λ 2 2.5096 sin θmax = 8π 1.263 sin θ1 = 8π ◦ θmax = 5.73

TABLE 4-16 Parameters

4.071 8π ◦ ◦ θ1 = 2.88 θ2 = 9.32 sin θ2 =

Characteristics of a Bayliss Line-Source Distribution with n = 10 3-dB Edge

Sidelobe Level (dB)

Beam Peak, ka/2 sin θmax

ka/2 sin θ1

ka/2 sin θ2

ATL (dB)

PEL (dB)

20 25 30 35 40

2.2366 2.3780 2.5096 2.6341 2.7536

1.140 1.204 1.263 1.318 1.369

3.620 3.855 4.071 4.270 4.455

0.50 0.54 0.69 0.85 1.00

1.81 1.90 1.96 2.01 2.04

162

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

4-8 WOODWARD LINE-SOURCE SYNTHESIS [9] In the preceding sections, methods to determine distributions that give the minimum beamwidth for speciﬁed sidelobe levels were discussed. Some applications require shaped beams extending over a range of angles. The Woodward synthesis samples the desired k-space pattern at even intervals to determine the aperture distribution. No integrals are required to compute coefﬁcients. The technique is based on the scanned pattern of a uniform amplitude distribution. Express the pattern in terms of U -space so that when scanned to U0 , it becomes sin π(U − U0 ) π(U − U0 ) with the nulls of the pattern occurring at integer values of U − U0 . U=

a sin θ λ

U0 =

a sin θ0 λ

The visible region extends between +a and −a, centered about U0 . Figure 4-6 shows two patterns, scanned to U0 = 1 and U0 = 2. The peak of the curve scanned at U0 = 2 occurs at one of the nulls of the pattern scanned to U0 = 1. If we allow only integer values of U0 , the pattern scanned to U0 solely determines the pattern at the point U0 in U -space. The two curves (Figure 4-6) in the regions below U = 0 and above U = 3 cancel each other to some extent when the distributions are

0.9 0.8 0.7

Voltage amplitude

0.6 0.5 0.4 0.3 0.2 0.1 0 −0.1 −0.2 −0.3 −10

−8

−6

FIGURE 4-6

−4

−2

0 a sin q l

2

4

6

Scanned uniform distributions: U0 = 1 and U0 = 2.

8

163

WOODWARD LINE-SOURCE SYNTHESIS

added. We form the aperture distribution from a sum of 2a/λ + 1 independent sample points of scanned apertures: N

E(x) =

Ei e−j (i/a)x

(4-45)

i=−N

where N = integer part (a/λ). Each term is a uniform amplitude distribution scanned to an integer value of U . The amplitudes Ei are determined by the sample values of the U -space pattern at those points. Example Design a 10λ aperture with a constant beam between θ = 0◦ and θ = 30◦ . The nonzero portion of the U -space pattern extends from U1 = 10 sin 0◦ = 0 to U2 = 10 sin 30◦ = 5. When we sample the U -space pattern, we discover six nonzero terms: i Ei

0

1

2

3

4

5

0.5

1.0

1.0

1.0

1.0

0.5

At U1 = 0 and U2 = 5, we use the average value. The aperture distribution is 0.5 + e−j x/a + e−j 2x/a + e−j 3x/a + e−j 4x/a + 0.5e−j 5x/a The U -space pattern of this distribution (Figure 4-7) shows some ripple in the beam and the reduction to 6 dB at the beam edges. If we increased the sample level at the edges, U = 0 and U = 5, the pattern would increase to that level.

−10

Level, dB

−20

−30

−40

−50

−60 −10

−8

−6

−4

−2

0 a sin q l

2

4

6

8

10

FIGURE 4-7 U -space pattern of Woodward–Lawson sampling for constant beam from 0 to 30◦ (10λ aperture).

164

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

A cosecant-squared power pattern can be designed by the same method as in the preceding example. When an antenna with this pattern on the ground points its maximum toward the horizon, it delivers a constant signal to an aircraft that maintains a constant altitude. The pattern falloff matches the range decrease as the aircraft ﬂies toward the antenna. The voltage pattern is given by E = E0

sin θmax sin θ

where θmax is the angle of the pattern maximum. In U -space this becomes E(U ) = E0

Um U

The amplitudes of the scanned apertures decrease as 1/U . Example Design a 10λ aperture with a cosecant-squared pattern from θ = 5◦ to θ = 70◦ with the maximum at 5◦ . There are 2a/λ + 1 possible sample points (21). The nonzero portion of the U -space pattern extends from Umin = 10 sin 5◦ = 0.87 to Umax = 10 sin 70◦ = 9.4. We sample only at integer values of U , which gives us nine nonzero terms: Um = 0.8716. The coefﬁcients are given in Table 4-17. The sum [Eq. (4-45)] for this distribution contains nine terms. E(x) =

9

Ei e−j (i/a)x

i=1

Figure 4-8 shows the amplitude and phase of this aperture distribution. The pattern obtained by summing the scanned aperture distributions (Figure 4-9) shows ripple about the desired pattern. Increasing the aperture size increases the number but does not change the level of ripples. The aperture distribution (Figure 4-8) has a negative phase slope to scan the beam off broadside. 4-9 SCHELKUNOFF’S UNIT-CIRCLE METHOD [10] Schelkunoff’s unit-circle method consists of the manipulation of the zeros (nulls) of the array pattern to achieve a desired pattern for a line array. The method is similar to designing networks by specifying the placement of poles and zeros in the complex plane, but the array has only zeros to manipulate. We can use the representation to describe any uniformly spaced array. TABLE 4-17 Woodward Synthesis Coefﬁcients of 10λ Cosecant-Squared Pattern i

Ei

i

Ei

i

Ei

1 2 3

0.8716 0.4358 0.2905

4 5 6

0.2179 0.1743 0.1453

7 8 9

0.1245 0.1089 0.0968

165

SCHELKUNOFF’S UNIT-CIRCLE METHOD

0.9 0.8 0.7 Amplitude (V)

0.6 0.5 0.4 0.3 0.2 0.1 0 −5

−4

−3

−2

1 −1 0 Aperture position (a)

2

3

4

5

−180 −5

−4

−3

−2

−1 0 1 Aperture position (b)

2

3

4

5

180 150 120 90

Phase (deg)

60 30 0 −30 −60 −90 −120 −150

FIGURE 4-8 Aperture distribution of Woodward–Lawson sampling for cosecant-squared pattern (10λ aperture): (a) aperture amplitude distribution; (b) aperture phase distribution.

Consider a uniformly spaced array along the z-axis with the pattern angle θ measured from the axis. The array response will be symmetrical about the z-axis. If we deﬁne the variable ψ = kd cos θ + δ, where δ is a ﬁxed progressive phase shift between elements, d the element spacing, and k the wave number (2π/λ), the pattern of the array is given by E = I0 + I1 ej ψ + I2 ej 2ψ + I3 ej 3ψ · · · (4-46)

166

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

−10

Pattern levels, dB

−20

−30

−40

−50

−60 −10

−8

−6

−4

−2

0 a sin q l

2

4

6

8

10

FIGURE 4-9 U -space pattern of 10λ aperture Woodward–Lawson sampling for a cosecant-squared pattern.

where Ii , a phasor, is the excitation of the ith element in the array. We simplify the notation further by deﬁning W = ej ψ (4-47) We then write Eq. (4-46) as E = I0 + I1 W + I2 W 2 + I3 W 3 + · · · + IN−1 W N−1

(4-48)

where N is the number of elements in the array. We use the ﬁrst element as our phase reference point. This array factor (isotropic elements) is a polynomial with N − 1 roots (zeros) for N elements. We denote the roots as Wi and rewrite Eq. (4-48) as E = E0 (W − W1 )(W − W2 ) · · · (W − WN−1 ) We can ignore the normalization E0 and compute array pattern magnitude as |E(W )| = |W − W1 ||W − W2 | · · · |W − WN−1 | where |W − Wi | is the distance from the root Wi to W in the complex plane. W is limited to the unit circle [Eq. (4-47)] because it always has unit value. Both the spacing of the elements and the progressive phase shift δ determine the limits of the phase of W : ◦

θ =0

◦

θ = 180

ψs = kd + δ

start

ψf = −kd + δ

ﬁnish

(4-49)

SCHELKUNOFF’S UNIT-CIRCLE METHOD

167

q q = 0°

w = ejy

ys y

yf q = 180°

FIGURE 4-10 Unit circle in the W -plane.

As θ increases, ψ decreases and W progresses in a clockwise rotation along the unit circle (Figure 4-10). We have no 2π limitation on either ψs or ψf . The element spacing determines the number of times W cycles the unit circle as θ varies from 0 to 180◦ . If ψs − ψf , 2kd, exceeds 2π, there is a possibility of more than one main beam (grating lobes). The zeros Wi , suppress the pattern when W moves close to one or more of them. The pattern rises to form a lobe when W is far from the zeros. The main-beam peak occurs at the point with the maximum product of the distances from the zeros. Whenever W passes through that point, another main beam forms. A uniformly fed array has the W -space polynomial f (W ) =

1 − WN 1−W

for N elements

The zeros of f (W ) are the N zeros of W N = 1 with the zero at W = 1 removed: Wi = ej 2πi/N . These are spaced uniformly on the unit circle. Figure 4-11 shows the unit circle diagram of a 10-element array fed with uniform phase and amplitude. W starts at −1 since d = λ/2, and it progresses clockwise around the unit circle one revolution to the same point as θ varies from 0 to 180◦ . At θ = 90◦ , the product of the distances from the zeros is a maximum. A lobe forms within the space between each pair of zeros. As W moves from the start to the main beam at W = 1, it starts at a zero and passes through four additional zeros. These zeros Wi correspond to the nulls in the pattern from θ = 0◦ to θ = 90◦ . An equal number of nulls occur as W moves through the range θ = 90 to 180◦ . A uniform-amplitude end-ﬁre array can be represented on the same unit-circle diagram. With antenna elements spaced λ/4, the excursion from start ψs to ﬁnish ψf is only π(2kd). A progressive phase shift δ of −kd through the array forms an end-ﬁre pattern. From Eq. (4-49), ψs = 0◦ and ψf = 180◦ . The end-ﬁre array pattern has only ﬁve nulls, including the null at θ = 180◦ as θ ranges from 0 to 180◦ , since only ﬁve zeros occur in the visible region.

168

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

FIGURE 4-11 Unit-circle representation of a 10-element array with λ/2 spacings.

The Hansen and Woodyard increased-directivity end-ﬁre array corresponds to a shift in the start and stop locations on the unit circle. The excursion from start to ﬁnish remains π determined by element spacing. Equation (4-49) calculates the start: ψs = 90◦ − 108◦ = −18◦ . The pattern has ﬁve nulls from θ = 0 to 180◦ . A binomial array has all its zeros at W = −1 and its pattern has no sidelobes, since they occur for points on the unit circle between zeros. Only one beam forms as W traverses the unit circle. The W -space polynomial is f (W ) = (W + 1)N−1 . For an array of given size we can manipulate the location of the nulls either to reduce sidelobes or to place pattern nulls. We reduce a sidelobe by moving the zeros on both sides of it closer together, but either the main-lobe beamwidth increases or the other sidelobes rise. We form a null in the array pattern by moving one of the zeros to that point on the unit circle corresponding to W at the null angle. Given a desired null θn , Wi = ej (kd cos θn +δ)

(4-50)

Equation (4-50) gives the phase angle kd cos θn + δ of the zero required on the unit circle in W -space. In the case of an end-ﬁre array in which the spacing between elements is less than λ/2, we can shift zeros from invisible space into visible space to narrow the beam and reduce sidelobes. We thereby form large lobes in invisible space that represent energy storage in the array. The large energy storage reduces the bandwidth and efﬁciency of the array. This super-directivity method has limited success, although we can produce beautiful patterns on paper. Example Design a four-element array of broadcast towers to give nearly uniform coverage for θ = ±45◦ with nulls at θ = 270◦ and 135◦ [11, p. 69]. We will align the array with θ = 0◦ to obtain symmetry for the ±45◦ requirement. We actually need only three elements, since only two nulls are speciﬁed. Using λ/4 spacings, we set δ = −90◦ to get an end-ﬁre array. Equation (4-50) gives the zeros of

SCHELKUNOFF’S UNIT-CIRCLE METHOD

169

the polynomial required for the pattern nulls. W1 : W2 :

360◦ λ ◦ ◦ ◦ cos(270 ) − 90 = −90 λ 4 360◦ λ ◦ ◦ ◦ cos(135 ) − 90 = −153.64 ψ= λ 4 ψ=

We determine the polynomial from the roots: ◦

◦

f (W ) = (W − e−j 90 )(W − e−j 153.64 ) ◦

= W 2 + 1.6994W ej 53.18 + ej 116.36

◦

We normalize the phase to the ﬁrst element of the array [constant term of f (W )]: ◦

◦

f (W ) = W 2 e−j 116.36 + 1.6994W e−j 148.18 + 1 At this point the polynomial representation of the array f (W ) does not include the progressive phase factor δ = −90◦ . We add the factor to the polynomial by adding −90◦ to the phase of the second element (W term) and −180◦ to the third element (W 2 term): ◦ ◦ f (W ) = W 2 e−296.36 + 1.6994W e−j 148.18 + 1 The coefﬁcients of the polynomial are the voltage (or current) components of the array. No null develops at θ = 180◦ because the two available nulls (N − 1) were used. Adding the fourth element gives us the freedom to improve the response ﬂatness in the ±45◦ region of θ . Figure 4-12 shows a unit-circle representation and pattern to give a nearly equal ripple response between ±45◦ and the required nulls. We increase

FIGURE 4-12 Four-element linear array with pattern nulls at θ = 90, 135, and 180◦ . The elements are spaced at 0.35λ to give a ﬂat response ±45◦ .

170

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

TABLE 4-18 Four-Element 0.35λ Spaced-Array Coefﬁcients for Uniform Beam ±45◦ No. 1 2 3 4

Amplitude (dB)

Phase (deg)

−9.50 −4.11 −4.11 −9.11

0.0 −103.3 138.4 35.1

the spacing to 0.35λ and place the pattern nulls at 90◦ , 135◦ , and 180◦ . W starts at +1 on the unit circle or ψs = 0◦ and determines δ: ψs = 0 = kd + δ

or

δ = −kd = −

360◦ ◦ 0.35λ = −126 λ

We compute the phase of the zeros from Eq. (4-50): ◦

◦

◦

◦

ψ1 = 360 (0.35) cos(90 ) − 126 = −126 ◦

◦

◦

◦

◦

◦

◦

◦

ψ2 = 360 (0.35) cos(135 ) − 126 = −215.1 ◦

ψ3 = 360 (0.35) cos(180 ) − 126 = −252

(144.9 ) ◦

(108 )

By following the same steps as above, we compute the phase and amplitude of the array elements (Table 4-18). 4-10

DOLPH–CHEBYSHEV LINEAR ARRAY [2]

The Chebyshev polynomials have equal ripples in the region x = ±1, and the amplitude varies between +1 and −1. Outside that region the polynomial value rises exponentially: −1 (−1)m cosh(m cosh |x|) x < −1 −1 Tm (x) = cos(m cos x) −1 ≤ x ≤ 1 −1 x>1 cosh(m cosh x) The order of the polynomial m equals the number of roots. Dolph devised a method of relating the Chebyshev polynomials to the array factor polynomial for a broadside array. We scale the polynomial to make the equal-ripple portion the sidelobes and the exponential increase beyond x = 1 becomes the main beam. Take an array fed symmetrically about the centerline that has either 2N + 1 or 2N elements. We expand the array factor in a polynomial with factors cos(ψ/2), where ψ = kd cos θ + δ. The beam peak occurs when ψ = 0. If we make this correspond to a value x0 , where the Chebyshev polynomial has a value R, the sidelobes will be equal to the ripple at the level 1/R. By substituting x = x0 cos(ψ/2), we use the Chebyshev polynomial for the array polynomial with Tm (x0 ) = R

or

x0 = cosh

cosh−1 R m

(4-51)

DOLPH–CHEBYSHEV LINEAR ARRAY

171

where 20 log R is the desired sidelobe level in decibels. The zeros of Tm (x) are given by (2p − 1)π xp = ± cos (4-52) 2m By using the equation xp = x0 cos(ψ/2) = x0 (ej ψ/2 + e−j ψ/2 ), we calculate the angles of the symmetrical zeros in the W -plane: ψp = ±2 cos−1

xp x0

(4-53)

Both values of xp [Eq. (4-52)] give the same ψp pair. Given the zeros in the W -plane, we multiply out the root form of the polynomial to calculate feeding coefﬁcients of the array. Example Design a 10-element array with 25-dB sidelobes. The array has nine nulls, so we pick m = 9 for the Chebyshev polynomial. [Eq. (4-51)]

R = 1025/20 = 17.7828

x0 = 1.0797

We need only the ﬁrst ﬁve zeros, since they are symmetrical about zero. We calculate them from Eq. (4-52), divide them by x0 , and use Eq. (4-53) for their angles on the unit circle of the W -plane (Table 4-19). We multiply out the root form of the polynomial for the voltage (current) feeding coefﬁcients of the array. Because the roots are symmetrical about the real axis, all phase angles are zero. We obtain the following coefﬁcients: Nos.

1, 10

2, 9

3, 8

4, 7

5, 6

Coefﬁcient (dB)

−8.07

−5.92

−2.84

−0.92

0.0

Figure 4-13 shows the unit-circle representation and pattern of the array with λ/2 spacing. We can estimate the beamwidth of a Chebyshev array by using a beamwidth broadening factor and the beamwidth of a same-length uniformly fed array [12]. The beamwidth broadening factor is given by

2 2 −1 2 2 f = 1 + 0.632 cosh (cosh R) − π R TABLE 4-19 Chebyshev Polynomial Roots and W -Plane Roots for 10-Element 25-dB Sidelobe Array p

Xp

ψp (deg)

1 2 3 4

0.9848 0.6428 0.3420 0.0

±48.41 ±106.93 ±143.06 180.00

(4-54)

172

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

FIGURE 4-13 Ten-element Chebyshev array designed for 25-dB sidelobes.

Equation (4-54) is valid in the range of sidelobe levels from 20 to 60 dB and for scanning near broadside. Example Compute the broadside beamwidth of a Dolph–Chebyshev array with 61 elements, a 30-dB sidelobe level, and λ/2 spacings. Equation (4-54) gives the value 1.144 for f using R = 1030/20 . We estimate the beamwidth of the uniform array from HPBW = 50.76◦ λ/N d = 1.66◦ , where d is the element spacing: ◦

◦

HPBWarray = (f )HPBWuniform = 1.144(1.66 ) = 1.90

We use the beam-broadening factor to estimate the array directivity: D=

2R 2 1 + (R 2 − 1)f λ/N d

(4-55)

Example We calculate the directivity of the 61-element array above from Eq. (4-55). D = 52.0 (17.2 dB). If we take its limit as N d → ∞, Eq. (4-55) becomes 2R 2 . An inﬁnite Dolph– Chebyshev array has a gain 3 dB more than the sidelobe level.

4-11

VILLENEUVE ARRAY SYNTHESIS [13]

Villeneuve devised a method similar to the Taylor distribution that modiﬁes the n − 1 inner zeros of a uniform amplitude array to lower sidelobes. Since the positions of the outer zeros remain ﬁxed, the outer pattern sidelobes decrease as 1/U . The uniform distribution W -plane zeros are located uniformly around the unit circle except for W = 1: 2πp ψp = (4-56) Ne

ZERO SAMPLING OF CONTINUOUS DISTRIBUTIONS

173

The inner zeros correspond to the Chebyshev zeros [Eq. (4-53)] except that we multiply them by a constant factor α dependent on the number of elements, the sidelobe level, and n: nπ (4-57) α= −1 Ne cos (1/x0 ) cos (2n − 1)π/2m The order of the Chebyshev polynomial m = Ne /2. We use Eq. (4-51) to compute x0 . Example Design a 10-element Villeneuve array containing 10 elements for 25-dB sidelobes and n = 4. We determine α = 1.00653 using Eq. (4-57). The inner three W -plane zeros are found by multiplying the Chebyshev zeros by α, which occur in pairs, and the next three zeros are found from the uniform amplitude array using Eq. (4-56): ψp :

±48.73

± 73.82

± 107.63

± 144

180

Figure 4-14 illustrates the W -plane and pattern of the 10-element Villeneuve array: Nos.

1, 10

2, 9

3, 8

4, 7

5, 6

Coefﬁcient (dB)

−8.44

−5.85

−2.91

−0.91

0.0

The sidelobes drop off instead of staying constant: −25.08, −25.19, −25.43, −26.14. 4-12 ZERO SAMPLING OF CONTINUOUS DISTRIBUTIONS [14] We sample continuous distributions, such as the Taylor line source, for large arrays. By using that method, we avoid the numerical difﬁculties of multiplying out long polynomials. When a small array samples an aperture distribution, its pattern fails to follow the pattern of the distribution. We improve the pattern by matching the zeros of the array to the distribution nulls. The array ψ-space pattern repeats at 2π intervals, but the k-space pattern of the aperture has no repeat. We space elements by λ/2 to

FIGURE 4-14 Ten-element Villeneuve array designed for 25-dB sidelobes, n = 4.

174

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

span the total ψ-space nonrepeating region. We then equate an array with λ/2 spacings to an aperture of the same length regardless of the actual spacings between elements. Since the array samples a continuous distribution, the aperture is N d long, where d is the distance between array elements and we consider the array element to be sampling d/2 on both sides of its location. Consider the U -space pattern of a uniform aperture distribution: sin πU/πU . The aperture zeros occur at integer values of U . The corresponding zeros of the uniformly fed array are Wi = ej 2πi/N , where i = 1, 2, . . . , N − 1. The Taylor distribution modiﬁes the location of the zeros of the uniform distribution to Ui , and the sampled zeros of the array must move to follow this pattern: Wi = ej 2πUi /N

(4-58)

Example Given a Taylor line source with 30-dB sidelobes and n = 6, compute the zeros of an array with 12 elements to sample the distribution. The array spans 12/2 in U -space. We calculate zeros of the distribution from Section 4-4 and the angles of the array zeros from Eq. (4-58): Ui

±1.473

±2.1195

±2.9989

±3.9680

±4.9747

6

ψi

±44.19

±63.58

±89.97

±119.04

±149.24

180

We multiply out the root form of the polynomial to compute the array feeding coefﬁcients. The array has 30-dB ﬁrst sidelobes. A straight sampling of the distribution gives an array whose sidelobes exceed 30 dB. Figure 4-15 shows the unit-circle diagram of a zero-sampled Taylor line source with 25-dB sidelobes and n = 5. The method places the zeros on the unit circle close enough together to limit the sidelobe peaks to less than 25 dB when W for a given pattern direction lies between the zeros. The array has higher sidelobes than the equivalent

FIGURE 4-15 Twelve-element array designed by zero-sampling 25-dB Taylor distribution: pattern of normal array (solid curve); pattern with null ﬁlling by moving three zeros off unit circle (dashed curve and triangles).

FOURIER SERIES SHAPED-BEAM ARRAY SYNTHESIS

175

aperture, but closer to the speciﬁed 25 dB, because the ﬁnite array cannot control sidelobes as well as the continuous aperture. Aperture

25.29

25.68

26.39

27.51

29.63

12-Element Array

25.03

25.07

25.18

25.44

26.41

The dashed plot of Figure 4-15 illustrates pattern behavior when W -space zeros are moved off the unit circle. We can ﬁll pattern nulls and generally shape the pattern. When we place all zeros on the unit circle in the complex plane, it can be proved that the array excitations will have amplitude symmetry about the centerline. Moving the zeros off the unit circle disturbs this symmetry. We can eliminate all pattern nulls by moving all the W -plane zeros off the unit circle. If we start with a uniformly fed array and move all the zeros to the same radius, the distribution taper across the array will be linear in decibels. In the next two sections we explore techniques for moving the zeros systematically to produce shaped patterns from an array. 4-13 FOURIER SERIES SHAPED-BEAM ARRAY SYNTHESIS The preceding methods seek the narrowest beamwidths for a given sidelobe level. Arrays can also produce shaped beams. We discussed the Woodward line-source method for shaped beams in Section 4-8. We obtain good approximations by sampling the line-source distribution with an array. Beyond sampling a line source, we can apply Fourier series to design an array directly. An array for a shaped beam must be much larger than is required for the beamwidth. The extra size of the array gives us the degrees of freedom necessary for beam shaping. Increasing the array size increases the match between the speciﬁed and the actual beam shape. Because the array pattern is periodic in k-space, we can expand the pattern in a Fourier series. The array pattern for a symmetrically fed array is given by either f (ψ) = 1 + 2

m In n=1

or f (ψ) = 2

m In n=1

I0

I0

cos

cos

2nψ 2

(2n − 1)ψ 2

N odd

(4-59)

N even

(4-60)

where m = (N − 1)/2 (odd) or m = N/2 (even) with ψ = kd cos θ + δ. Equations (459) and (4-60) are Fourier series expansions of the pattern in ψ-space. The elements farthest from the centerline produce the highest harmonics in the series. In an asymmetrically fed array, we express Eqs. (4-59) and (4-60) as a sum of exponential terms: m an ej nψ N odd n=−m f (ψ) = (4-61) m j (2n−1)ψ/2 −j (2n−1)ψ/2 an e + a−n e N even n=1

176

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

Suppose that we have a desired pattern in k-space given by fd (ψ). We expand it in an inﬁnite Fourier series of the same form as Eq. (4-61) with m = ∞. We equate the ﬁrst m coefﬁcients of the two Fourier series to approximate the desired pattern. As in any Fourier series method, we solve for the coefﬁcients by using the orthogonality of the expansion functions when integrated over a period: π 1 fd (ψ)e−j nψ dψ N odd 2π −π an = (4-62) π 1 −j (2n−1)ψ/2 fd (ψ)e dψ N even 2π −π π 1 fd (ψ)ej (2n−1)ψ/2 dψ N even (4-63) a−n = 2π −π We determine the array coefﬁcients directly from the Fourier series coefﬁcients. Example Design a 21-element array with λ/2 element spacing with a constant beam 2b wide centered in ψ-space. We use Eq. (4-62) to compute coefﬁcients an : b 1 sin nb an = e−j nψ dψ = 2π −b πn Suppose that the constant beam is 45◦ at broadside: 67.5◦ ≤ θ ≤ 112.5◦ . Then b=

360◦ λ ◦ ◦ cos 67.5 = 68.88 λ 2

We can ignore the constant factor i/π and expand to compute the array coefﬁcients (Table 4-20). The method fails to some extent when we try it on arrays with spacings greater than λ/2. The integral does not cover the total visible region. We can, however, use it with TABLE 4-20 Fourier Series Synthesis Coefﬁcients of 21-Element Array for Pattern of Figure 4-16 n 0 ±1 ±2 ±3 ±4 ±5 ±6 ±7 ±8 ±9 ±10

an

Amplitude (dB)

Phase (deg)

1.0000 0.9328 0.3361 −0.1495 −0.2488 −0.0537 0.1336 0.1209 −0.0240 −0.1094 −0.0518

0.00 −0.60 −9.47 −16.50 −12.08 −25.40 −17.48 −18.35 −32.40 −19.22 −25.72

0 0 0 180 180 180 0 0 180 180 180

FOURIER SERIES SHAPED-BEAM ARRAY SYNTHESIS

177

spacings less than λ/2 with good results. As we increase the number of elements in the array, the match to the desired pattern improves. Of course, tapering the desired pattern reduces the higher harmonics and the subsequent need for more elements. Example Suppose that we want to scan the beam of the 21-element array with λ/2 element spacing to 60◦ with a 45◦ beamwidth. The beam edges are 37.5◦ and 82.5◦ . We could calculate coefﬁcients by integrating Eq. (4-62) directly with this requirement, but we can use δ, the progressive phase shift between elements, to simplify the problem. The beam edges in ψ-space are 180◦ cos(37.5◦ ) + δ 142.8◦ + δ

and 180◦ cos(82.5◦ ) + δ 23.49◦ + δ

We pick δ to center the beam in ψ-space: b = 142.8◦ + δ, −b = 23.49◦ + δ. On solving, we have δ = −83.15◦ and b = 59.65◦ . We use the formula sin(nb)/πn to compute coefﬁcients of the array and then add the progressive phase shift through the array. Figure 4-16 shows the array pattern. When we scan the beam to end ﬁre, we must account for the symmetry about θ = 0◦ . Because we limit the spacings to less than λ/2 to prevent grating lobes, we have an unspeciﬁed region of ψ-space that we can choose in any convenient manner. Example

Design a 21-element end-ﬁre array with a 90◦ beamwidth and 0.30λ spacings.

FIGURE 4-16 Twenty-one-element array designed by the Fourier series method to scan to 60◦ with a 45◦ beamwidth.

178

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

For an end-ﬁre array we pick δ = −kd = −108◦ . This places the edge of the visible region on the ψ-space origin. We are free to specify the invisible region that will be included in the integral [Eq. (4-62)]. We specify the invisible region as the mirror image of the portion in the visible region and solve for b: ◦

◦

◦

◦

−b = 360 (0.3) cos(45 ) − 108 = −31.63

We use the sin(nb)/πn formula to calculate array coefﬁcients and then apply the progressive phase shift δ to the coefﬁcients obtained to get the proper phase to scan to end ﬁre. We cannot control the sidelobes of an array designed using Fourier series expansion. The initial speciﬁcation calls for no sidelobes. Sampling a Woodward linear aperture with an array also fails to give control of the sidelobes. The Woodward linear distribution cannot control sidelobes; it provides only ease of design. In the next section we explore a method with direct control of sidelobes of an array.

4-14

ORCHARD METHOD OF ARRAY SYNTHESIS [15]

In Section 4-6 we manipulate the nulls of a continuous linear distribution to control the sidelobes of the radiated pattern individually. In Section 4-9 we show that the nulls of the linear aperture pattern can be related directly to the roots of Schelkunoff polynomial representation of the linear array pattern in W -space. The unit circle method gives us a tool for array synthesis expanded in the Orchard method for the design of arrays with arbitrary patterns. We apply an iterative technique on the W -space zeros to produce the pattern desired. We control all the sidelobes individually and produce shaped patterns for the main beam. The ﬁnite size of the array limits the control of the main beam shape as we saw in the Fourier series expansion method. Each array element corresponds to a term in the Fourier series expansion. We start with the Schelkunoff transformation of the array pattern: f (W ) = C0

N

(W − Wn )

(4-64)

n=1

A normalization constant C0 has been added. We write Wn = exp(an + j bn ). Expansion of Eq. (4-64) produces the feeding coefﬁcients of an array with N + 1 elements: W = ej ψ

with ψ = kd cos θ + δ

θ is measured from the array axis. The effect of δ on the unit circle method is to rotate the starting and ﬁnishing points when varying W to calculate the pattern using Eq. (4-64). An equally valid method is to rotate the zeros about the origin of the complex plane, which leaves the ψ-space pattern shape unchanged. When designing a shaped beam, we need to rotate the main beam peak to the proper location to calculate the amplitudes because our speciﬁcation will be in terms of the pattern angle θ relative to the peak.

ORCHARD METHOD OF ARRAY SYNTHESIS

179

Figure 4-11 illustrates that the pattern amplitude is the product of distances from each zero to the pattern W point. Expansion of Eq. (4-64) in terms of the product of distances to W gives |f (W )|2 = C02

N

[1 − 2ean cos(ψ − bn ) + e2an ]

(4-65)

n=1

The Orchard method requires the speciﬁcation of each sidelobe and additional values located at the minimum ripple points in the shaped region. For a single-beam unshaped pattern, we only specify sidelobes, and all an will be zero since all zeros Wn will be on the unit circle. We restrict the array to λ/2 spaced elements when applying the method so that the entire unit circle is used in the pattern. An array with N zeros has N pattern peaks which lie between the zeros in the W -plane. When we include the normalization constant C0 to specify the main beam peak and all the zeros, we have N + 1 unknowns to ﬁnd. Without loss of generality we specify the last zero as WN = −1 or ψ = π to reduce the number of unknowns to N . Since we rotate the zeros after we determine the proper zero spacing for speciﬁed sidelobes, we place the main beam between WN−1 and WN = −1. Before starting the iteration technique, we generate a list of sidelobe levels with the main beam as the last one. The method expands the pattern in a multiple-variable Taylor series using bn , an , and the normalization constant as variables. To facilitate calculating the partial derivatives, we express Eq. (4-65) in decibels: N−1

10 ln[1 − 2ean cos(ψ − bn ) + e2an ] + 10 log10 [2(1 + cos ψ)] + C ln(10) n=1 (4-66) The second term of Eq. (4-66) is due to the zero WN = −1 and C is the normalization constant of the main beam. The logarithm to the base 10 has been expressed as a natural logarithm for the calculation of derivatives: G=

∂G Mean [ean − cos(ψ − bn )] = ∂an 1 − 2ean cos(ψ − bn ) + e2an

(4-67)

∂G Mean sin(ψ − bn ) =− ∂bn 1 − 2ean cos(ψ − bn ) + e2an

(4-68)

∂G =1 ∂C

(4-69)

The variable M = 20/ln(10). The multiple-variable Taylor series involves three types of terms: G(bn , an , C) = G0 (bn0 , an0 , C0 ) +

N−1 n=1

+

N−1 n=1

∂G (bn − bn0 ) ∂bn

∂G (an − an0 ) + (C − C0 ) ∂an

(4-70)

180

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

Every nonzero value of an ﬁlls in the pattern null at ψ = bn . If we specify the desired pattern amplitude at every sidelobe peak, the main beam, and at points between the sidelobe peaks equal to the number of nonzero an , we form a square matrix equation. The solution gives the changes in bn , an , and C. Since we expanded Eq. (4-66) as a linear approximation, the solution of Eq. (4-70) gives only an approximate solution. In a few iterations the method converges and we obtain an acceptable pattern. Suppose that the shaped pattern is limited to a range in W -space so that there are only L nonzero an . Given the desired pattern Sm (ψm ) at ψm and the current pattern G0 (ψm ), one row of the matrix is

∂G(ψm ) ∂G(ψm ) ∂G(ψm ) ∂G(ψm ) ,..., , ,..., ,1 ∂b1 ∂bN−1 ∂a1 ∂aL

We need N + L rows or pattern points to solve Eq. (4-70) for changes in bn , an , and C: [δb1 , . . . , δbN , δa1 , . . . , δaL , δC]T We require a search routine to locate the pattern peaks between the pattern nulls or minima between peaks in the shaped region for given values of bn and an after we normalize to the current pattern peak. We subtract these from the levels desired: [S(ψ1 ) − G0 (ψ1 ), . . . , S(ψN+L ) − G0 (ψN+L )]T After solving the square matrix equation, we update the W -plane zeros: b1 = b1 + δb1 ... bN−1 = bN−1 + δbN−1 .. . a1 = a1 + δa1 .. . aL = aL + δaL C = C0 + δC The iteration alters the beam peak and its location. The pattern peak is normalized after iteration, and for a shaped pattern a new zero rotation is found to line up the beam peak for the pattern-shaping function. Example Design an eight-element array with its beam peak at 90◦ and speciﬁed sidelobes before the peak of 25, 30, and 25 dB and 20, 25, and 30 dB after the peak. The sidelobes values begin with the ﬁrst sidelobe after the peak and rotate to the peak: −20 − 25 − 30 − 25 − 30 − 25 0

181

ORCHARD METHOD OF ARRAY SYNTHESIS

FIGURE 4-17 Eight-element array designed using Orchard synthesis for individually speciﬁed sidelobes, λ/2 spacings.

The solution converges in four iterations after starting with uniformly spaced zeros on the unit circle. Figure 4-17 shows the unit-circle zeros on the left and the corresponding pattern on the right with λ/2 element spacing: W -Space Zero (deg)

178.14

142.72

99.26

58.01 −62.18

−96.89 −130.37

Pattern Null (deg)

8.25

37.54

56.53

71.20

122.57

110.21

136.41

The feeding coefﬁcients for the ﬁnal design are given in Table 4-21. Although the Orchard method requires the elements to be spaced λ/2 during synthesis, the completed design can be used at another element spacing. Figure 4-18 gives the unit-circle diagram of the same array with a 0.7λ element spacing. The range of W now exceeds 2π and the sidelobe regions of the unit circle have been used more than once. Sidelobes 3 and 4 occur twice in the pattern. Of course, if we scan the array too far, the pattern would have grating lobes. Figure 4-19 plots the pattern of an end-ﬁre array with λ/4-element spacing using the same zeros. Only a portion of the unit circle is used, and not all sidelobes are realized. Figure 4-20 illustrates the end-ﬁre case with the elements spaced so that the ﬁnal position of W occurs at a null. The pattern contains all six sidelobes. The unit-circle analysis mirrors that of TABLE 4-21 Synthesis

Coefﬁcients of Eight-Element Array of Figure 4-17 Designed by Orchard Magnitude (dB)

Phase (deg)

Element

Element

Magnitude (dB)

Phase (deg)

1 2 3 4

−8.69 −3.90 −1.06 0

8.70 3.22 1.29 4.91

5 6 7 8

0 −1.06 −3.90 −8.69

3.79 7.41 5.48 0

182

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

FIGURE 4-18 Eight-element array designed using Orchard synthesis for individually speciﬁed sidelobes, 0.7λ spacings.

FIGURE 4-19 Eight-element array designed using Orchard synthesis for individually speciﬁed sidelobes, λ/4 spacings scanned to end ﬁre.

FIGURE 4-20 Eight-element array designed using Orchard synthesis for individually speciﬁed sidelobes, 0.42λ spacings scanned to end ﬁre.

183

ORCHARD METHOD OF ARRAY SYNTHESIS

the circle diagram in Chapter 3, where increasing the element spacing increases the visible region. In this case the visible region corresponds to rotation about the unit circle. Expansion of Eq. (4-64) produces the array feeding coefﬁcients independent of element spacing, and the progressive phase shift between elements δ affects phase but not amplitude. The four examples given in Figures 4-17 to 4-20 have the same sequence of feed magnitudes. We can use Orchard synthesis to generate a difference pattern similar to the Bayliss line distribution and control all the sidelobes. A difference pattern has two main beams. Using the same example of an eight-element array, we modify the sidelobe list to include side-by-side main beams. We eliminate the −25-dB lobe next to the original main beam from the values above: −20

− 25

− 30

− 25

− 30

0

0

When we apply the synthesis by placing the last main beam at 90◦ , we obtain a pattern with two main beams with the null between them at 101.6◦ , corresponding to a W plane null at −36.3◦ . We rotate all W -plane zeros by 36.3◦ to place the null between the two main beams at 90◦ . Figure 4-21 shows the W -plane and polar pattern for the ﬁnal design. Note the placement of the W -plane zero at W = +1. Table 4-22 lists the feeding coefﬁcients.

FIGURE 4-21 Difference pattern array using eight-elements designed by Orchard synthesis.

TABLE 4-22 Coefﬁcients of Eight-Element Difference Pattern Array of Figure 4-21 Designed by Orchard Synthesis Element

Magnitude (dB)

Phase (deg)

Element

Magnitude (dB)

Phase (deg)

1 2 3 4

−6.32 −0.35 0.0 −6.91

5.39 1.28 0.8 7.05

5 6 7 8

−6.91 0.0 −0.35 −6.32

178.35 184.6 184.12 180

184

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

We ﬁll in the null between the different lobes to form a ﬂat-topped beam for the eight-element array and use a constant-amplitude shaping function for the pattern desired. The beamwidth of the ﬂat lobe is determined by the lobe spacing, and only certain sizes are possible. Remember that an array is a Fourier series approximation to the pattern desired. With only eight elements the match is poor between the pattern desired and the approximate pattern. We use one nonzero an to move the W -plane zero off the unit circle that forms the pattern null between the two beams and add another pattern speciﬁcation: −20

− 25

− 30

− 25

− 30

0

0

−1

The last number gives the pattern level at the null relative to the shaped pattern level. This last term uses Eq. (4-67) for its columns. The constant beam design uses a 22◦ wide beam centered at 90◦ for the pattern shape function. We start with an = 0.01 before iterating. The iteration using the matrix equation computes a1 = 0.4435, which can be either positive or negative without changing the pattern. Rotation of the W -plane zeros placed the zero for minimum ripple along the positive real axis and produced a symmetrical pattern about θ = 90◦ . Figure 4-22 contains the ﬁnal design W -space zeros and polar pattern. The iterations produced the sidelobe levels speciﬁed (Table 4-23).

FIGURE 4-22 Flat-topped beam eight-element array designed by Orchard synthesis. TABLE 4-23 W -Plane Zeros of Eight-Element Flat-Topped Beam of Figure 4-22 Designed by Orchard Synthesis W -Space Zero (deg)

W -Space Radius

Pattern Null (deg)

165.51 123.99 84.33 0 −91.65 −115.39 −161.18

1.0 1.0 1.0 0.6418 1.0 1.0 1.0

23.15 46.46 62.06 90 120.61 129.87 153.57

ORCHARD METHOD OF ARRAY SYNTHESIS

185

TABLE 4-24 Coefﬁcients of Eight-Element Array for Flat-Topped Beam of Figure 4-22 Designed by Orchard Synthesis Element

Magnitude (dB)

Phase (deg)

Element

Magnitude (dB)

Phase (deg)

1 2 3 4

−12.95 −10.78 −24.69 −7.90

−174.39 178.85 167.92 −1.47

5 6 7 8

−1.15 0.0 −2.26 −9.10

1.54 4.37 3.95 0.0

The radius of the fourth term could be 1/0.6418 = 1.5581 without affecting the pattern result. Inserting the zeros into Eq. (4-64) and expanding the polynomial produces the feeding coefﬁcients (Table 4-24). The Fourier series example for a constant beam centered at 60◦ with a 45◦ beamwidth using 21 elements spaced λ/2 (Figure 4-16) was repeated using Orchard synthesis. Fourier series synthesis could not control the sidelobes. First, we need to ﬁgure out how many array lobes cover the shaped pattern region. Place the zeros uniformly around the unit circle in the W -plane and determine how many of the roots are within the beam. For a 21-element array six beams and ﬁve zeros lie in the ψ = π cos θ angular region of the constant beam found using Eq. (4-71): beams =

N (cos θmin − cos θmax ) 2

(4-71)

The solution to Eq. (4-71) is an integer given N as the number of W -plane zeros. All sidelobes were set at −30 dB and the ripple at −0.9 dB below the constant beam: Lobes

1–14

15–20

21–25

Sidelobe (dB)

−30

0.0

−0.9

Figure 4-23 gives the ﬁnal result of the synthesis, an improvement over Figure 4-16, with its uncontrolled sidelobes.

FIGURE 4-23 Twenty-one-element array designed by Orchard synthesis to scan to 60◦ with a 45◦ beamwidth.

186

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

We must consider the element excitations. The zeros not lying on the unit circle can be either inside or outside the circle and produce the same pattern. Different combinations of zero locations lead to different element amplitudes. The last example has ﬁve zeros displaced from the unit circle, which produces 25 = 32 combinations. We need to check the amplitude distribution that results from each case (Table 4-25). Arrays with a large range of amplitudes are difﬁcult to produce. In some cases the range of amplitude available is limited, such as waveguide slot arrays. Mutual coupling between elements also makes it difﬁcult to achieve the desired low amplitudes on some elements because nearby elements will excite them, and compensation for mutual coupling may prove difﬁcult. Figure 4-23 shows one of the combinations of root placements that produced the minimum amplitude variation in the array. The Fourier series synthesis gave an amplitude variation of 32.4 dB, whereas the Orchard synthesis variation is 13.29 dB. This synthesis produced better patterns with less amplitude variation. Decreasing the ripple depth increases the amplitude variation of the array. Csc2 θ cos θ Pattern This pattern produces constant round-trip signals versus the elevation angle for radar. The pattern from the array axis is given by csc2 (θ − 90◦ ) cos(θ − 90◦ ). The peak occurs beyond 90◦ and decreases for greater angles. The shaped pattern function requires the rotation of the W -plane zeros at each step so that the pattern peak calculated from the zeros occurs at the proper angle. The changing zero locations move the beam peak location at each iteration. Example Design a 16-element csc2 (θ − 90◦ ) cos(θ − 90◦ ) beam array to operate from 100 to 140◦ and have 30-dB sidelobes. Equation (4-71) determines that ﬁve beams cover the pattern region and sets the number of nonzero an as 4. The 16-element array has 15 zeros, with the ﬁrst 10 speciﬁed as −30 dB, ﬁve for the shaped-beam region, and four for the minima between the shaped-beam peaks. We specify the shaped-beam lobes relative to the shape levels. The last lobe is the beam peak.

TABLE 4-25 Coefﬁcients of 21-Element Array for Flat-Topped Beam of Figure 4-23 Designed by Orchard Synthesis Element 1 2 3 4 5 6 7 8 9 10 11

Amplitude (dB) −11.85 −6.89 −6.23 −11.34 −5.12 0.0 −0.01 −6.33 −9.73 −2.09 −1.75

Phase (deg) −65.47 −146.85 126.55 4.64 −158.19 108.08 23.97 −65.70 86.93 −4.11 −70.12

Element 12 13 14 15 16 17 18 19 20 21

Amplitude (dB) −3.48 −3.01 −2.97 −4.93 −7.51 −9.64 −9.44 −8.06 −9.26 −13.29

Phase (deg) −115.31 −158.73 134.01 49.59 −40.53 −111.80 −159.41 142.99 73.66 0.0

187

ORCHARD METHOD OF ARRAY SYNTHESIS

Lobe

11

12

13

14

15

16

17

18

19

Amplitude (dB)

1.0

0.8

0.6

0.4

0.2 −0.9 −0.7 −0.5 −0.2

Allowing the ripple to increase in the lower levels of the shaped pattern region decreases the range of element amplitudes. The method converged in 11 iterations to the design given in Figure 4-24. All 24 = 16 combinations of an placements inside and outside the unit circle were checked (Table 4-26). The amplitude variation ranged from 11.47 to 25.47 dB. Figure 4-25 illustrates the design repeated with eight elements (Table 4-27). Although the sidelobes could be controlled at −30 dB, the shaped pattern region shows less pattern control than with 16 elements. Extensions to the Orchard method make various improvements. By adding balancing zeros inside and outside the unit circle in the W -plane, the feeding coefﬁcients of the array can be made real with only 0 or 180◦ phases [16]. This adds elements to the array and changes the shape of the beam somewhat. The coefﬁcients are real only if the pattern is symmetrical about θ = 90◦ . To implement the method you add a term

FIGURE 4-24 Sixteen-element array with csc 2 θ cos θ pattern designed by Orchard synthesis. TABLE 4-26 Coefﬁcients of 16-Element Array for csc2 θ cos θ Beam of Figure 4-24 Designed by Orchard Synthesis Element

Magnitude (dB)

Phase (deg)

Element

Magnitude (dB)

Phase (deg)

1 2 3 4 5 6 7 8

−11.47 −9.84 −8.07 −4.79 −2.65 −2.04 −0.82 0.0

−149.27 −100.16 −69.72 −40.25 −0.56 34.20 65.82 107.07

9 10 11 12 13 14 15 16

−1.19 −2.71 −3.53 −7.82 −9.46 −4.30 −4.55 −8.72

149.09 −177.54 −131.64 −52.50 100.34 −157.29 −76.99 0.0

188

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

FIGURE 4-25

Eight-element array with csc 2 θ cos θ pattern designed by Orchard synthesis.

TABLE 4-27 Coefﬁcients of Eight-Element Array for csc2 θ cos θ Beam of Figure 4-24 Designed by Orchard Synthesis Element

Magnitude (dB)

Phase (deg)

Element

Magnitude (dB)

Phase (deg)

1 2 3 4

−11.62 −8.61 −12.62 −2.61

130.02 −170.78 179.21 −173.92

5 6 7 8

0.0 −2.05 −5.73 −11.54

−122.98 −79.05 −44.44 0.00

to Eqs. (4-67) and (4-68) for the extra elements located off the unit circle. A design of a ﬂat-topped beam centered at 90◦ using the balanced zeros produced a design with more than 30 dB of variation between the elements similar to a Fourier series expansion that had about the same range of amplitudes. The range of amplitudes in the array can be reduced by placing all the zeros off the unit circle [17, p. 124]. We give up the nulls between the lobes and must now search a large set of possible solutions to select a design with the least amplitude variation. A genetic algorithm sorts through the large set of zero combinations inside/outside all that satisfy the pattern requirements to discover the best design.

4-15

SERIES-FED ARRAY AND TRAVELING-WAVE FEED SYNTHESIS

A series-fed array uses couplers along a line that distribute power to the elements from a single transmission line. A single wave travels along the line with each element removing a portion of the power. A matched load absorbs the remaining power at the end to prevent the reﬂection of a wave traveling toward the source end. A second backward traveling wave would produce another beam with reduced amplitude indistinguishable from a sidelobe. The coupling could be a physical coupler or it could be just a series or shunt load across the transmission line. Waveguide slots are an example of loads on a transmission line. An array using couplers can have phase shifters between

SERIES-FED ARRAY AND TRAVELING-WAVE FEED SYNTHESIS

189

the couplers and the elements to form a phased array. A second conﬁguration for a phased array places phase shifters in the transmission line between the couplers. This case uses the simple control of identical phase shifters set to the same value to scan the beam. The phase shifters are the progressive phase δ along the array used for scanning. The array distribution is given by the sequence of radiated powers, Pi . The traveling wave or nonresonant array dissipates a ratio of the input power R in the load: N

Pi = Pin (1 − R)

i=1

N We normalize the distribution to the sum of radiated power: P0 = Pi . The elei=1 ment power becomes Pi (1 − R)/P0 and we use the normalized power distribution to calculate coupling values: C1 = P1

remaining power = 1 − P1

The coupling to the second element removes power from the remaining power: C2 =

P2 1 − P1

remaining power = 1 − P1 − P2

The general expression is Ci =

1−

Pi i−1 j =1

Pj

(4-72)

If the element electrical model consists of a shunt conductance on a transmission line, such as waveguide slots, the power radiated by each slot = |Vinc |2 gi and the normalized gi = Ci . Similarly, an electrical model of an element as a series resistance on a transmission line can be solved in a similar manner. Power radiated = |Iinc |2 ri and the normalized ri = Ci . Some array feeders have signiﬁcant losses between the elements and we must account for these losses when designing the couplers. Suppose that the feeder has identical losses Lf = 1 − 10−attenuation/10 between couplers. The power balance equation becomes N N (j − 1)Pj Lf Pi Pin RPin (N − 1)Lf RPin j =2 i=1 = + + + total load losses to load antennas losses to antennas N N Lf (j − 1)Pj + Pi j =2 i=1 Pin = 1 − R − (N − 1)Lf R As before, we must normalize the power at each element to the input power Pi /Pin . The coupling to the ﬁrst element is C1 = P1 and the power left is 1 − P1 . The transmission medium attenuates the signal between the ﬁrst and second elements and we compute the power at the second element = (1 − P1 )(1 − Lf ). We determine the coupling value from the ratio P2 C2 = (1 − P1 )(1 − Lf )

190

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

and the element removes P2 power. The remaining power travels to the next element but is attenuated by (1 − Lf ). The power removed, P2 , is subtracted from the power at that point and the remaining power is attenuated before reaching the next extraction: P3 [(1 − P1 )(1 − Lf ) − P2 ](1 − Lf ) P4 C4 = {[(1 − P1 )(1 − Lf ) − P2 ](1 − Lf ) − P3 }(1 − Lf ) C3 =

etc.

The total loss due to attenuation is found from the sum of the normalized powers: loss(dB) = 10 log

N

Pi

i=1

Continuous Traveling Wave As the wave propagates along the antenna, it loses power continuously. The slots or holes must radiate more and more of the remaining power if the distribution is to be uniform. In general, the holes or slots must load the waveguide increasingly as the wave travels to the termination. The power at any point in the guide is z P (z) = P0 exp −2 α(z) dz (4-73) 0

where P0 is the power at z = 0 and α(z) is the attenuation distribution (nepers/length). Suppose that we have a desired amplitude distribution, A(z) (voltage):

L

Pin =

L

|A(z)|2 dz +

0

ρL (z) dz + Pload

(4-74)

0

where Pload is the power lost in the termination, |A(z)|2 the radiated power distribution, and ρL (z) the ohmic loss in the walls. Let the power into the termination be a ratio of the input power Pload = RPin ; then Pin =

1 1−R

L

|A(z)|2 + ρL (z) dz

(4-75)

0

The power anywhere along the leaky wave antenna is

L

P (z) = Pin −

|A(z)|2 + ρL (z) dz

(4-76)

0

We differentiate this to get dP (z) = −[|A(z)|2 + ρL (z)] dz

(4-77)

We differentiate Eq. (4-73) to relate α(z) to P (z): 1 dP (z) = −2α(z) P (z) dz

(4-78)

CIRCULAR APERTURES

191

We substitute Eq. (4-75) into Eq. (4-76) for Pin . By combining Eqs. (4-76) and (4-77) into Eq. (4-78), we derive the required attenuation distribution [18, p. 153]: α(z) =

L

[1/(1 − R)]

1 |A(z)|2 2

z

|A(z)|2 + ρL (z)dz −

0

(4-79) |A(z)|2 + ρL (z)dz

0

If we assume a lossless transmission line, ρL (z) = 0 and Eq. (4-79) simpliﬁes. Example Design the attenuation distribution for a uniform distribution along a lossless transmission-line leaky wave antenna. Substitute A(z) = 1 and ρL (z) = 0 into Eq. (4-79) and perform the integrations: α(z) =

1 2

[L/(1 − R)] − z

=

1 (1 2

− R) L[1 − z(1 − R)/L]

Given R = 0.05 (5% of the power into the load) for a structure with length 10λ. The initial and ﬁnal attenuation constants are 0.95 = 0.0475 Np/λ or 0.413 dB/λ 20 0.95 αf (L) = = 0.95 Np/λ or 8.25 dB/λ 2LR αi (0) =

We reduce the variation between the initial and ﬁnal values by dissipating more power in the termination. Given R = 0.1, αi (0) = 0.045 Np/λ or αf (L) = 0.45 Np/λ or

0.39 dB/λ 3.9 dB/λ

If we take the ratio of the attenuations at the ends, we have α(L)/α(0) = 1/R. We can normalize Eq. (4-79) to the interval ±2 and use the linear distributions given above where x = z/L and ρL (z) = 0. Figure 4-26 shows the attenuation distribution for a Taylor distribution with 30-dB sidelobes and n = 8 for various levels of power dissipation in the load. Table 4-28 lists the bounds on α(x)L for various Taylor distributions. Changing the number of modiﬁed zeros has only a minor effect on the bounds. A cos2 on a pedestal distribution with a 30-dB sidelobe level has very similar bounds on the attenuation. The 40-dB sidelobe level design requires a greater variation of attenuation than the 30-dB cases. Long structures may not be able to provide the low levels of radiation above the ohmic losses for an effective design. In all cases we decrease the attenuation range on an antenna by decreasing the antenna efﬁciency though absorbing more power in the termination. 4-16 CIRCULAR APERTURES Many common apertures conform to circles. The two-dimensional Fourier transform relation for the pattern holds for any aperture rim shape and becomes for the circle 2π a E(r , φ )ej kr sin θ cos(φ−φ ) r dr dφ (4-80) f (θ, φ) = 0

0

192

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

Input power dissipated: 5% 25

10%

Attenuation × length, dB

20

15% 15

20% 25%

10

5

0

0.2

0.4

0.6

0.8

1

Normalized length

FIGURE 4-26 Leaky wave attenuation distribution for Taylor distribution with 30-dB sidelobes, n = 8. TABLE 4-28 Maximum and Minimum Normalized Attenuation α(z )L of a Leaky Wave Taylor Distribution 30 dB n=6

n = 12

40 dB, n = 8

Termination Power (%)

Maximum

Minimum

Maximum

Minimum

Maximum

Minimum

5 6 8 10 12 15 20 25

27.08 25.38 22.70 20.66 19.00 17.00 14.42 12.42

0.59 0.58 0.57 0.56 0.55 0.53 0.50 0.46

26.04 24.54 22.06 20.08 18.46 16.52 14.04 12.09

0.63 0.63 0.61 0.60 0.59 0.57 0.53 0.50

31.61 29.63 26.52 24.12 22.18 19.81 16.78 14.44

0.12 0.12 0.12 0.11 0.11 0.11 0.10 0.10

CIRCULAR APERTURES

193

where a is the radius, r the radial coordinate, and φ the angle coordinate of the aperture point. The integral leads to a kr -space. When the distribution has circular symmetry, the φ integral can be evaluated easily, which reduces Eq. (4-80) to

a

f (kr ) = 2π

E(r )J0 (kr sin θ )r dr

(4-81)

0

where J0 (x) is the zeroth-order Bessel function of the ﬁrst kind. All great-circle patterns (constant φ) are identical. For a uniform distribution, f (kr ) =

2J1 (ka sin θ ) ka sin θ

plotted in Figure 4-27. The zeros occur at the zeros of J1 (x). The 3-dB pattern point of the uniform distribution is ka sin θ1 = 1.6162 −1

HPBW = 2 sin

sin θ1 = 0.5145

λ D

(4-82)

0.5145λ D

−10

Pattern level, dB

−20

−30

−40

−50

−60 −5 −25 −100

−4 −20 −80

−3 −15 −40

−2 −10 −20

−1 −5 −10

0

1 5 20

2 10 40

3 15 60

4 20 80

ka (sin q − sin0)

FIGURE 4-27 kr -space pattern of uniform circular aperture distribution.

5 25 100

194

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

where D is the diameter. For large apertures we can approximate sin θ by θ (rad). Converted to degrees, the half-power beamwidth becomes ◦

HPBW = 58.95

λ D

(4-83)

Example Compute the beamwidth of a uniform distribution circular aperture with 10.5λ diameter. The beamwidths are found from Eqs. (4-82) and (4-83): HPBW = 2 sin−1 HPBW =

0.5145 ◦ = 5.62 10.5

58.95◦ ◦ = 5.61 10.5

The ﬁrst zero of J1 (x) gives the k-space pattern null point. ka sin θnull = 3.8317 BWnull = 2 sin−1

1.2197λ ◦ λ 139.76 D D

(4-84)

We can also deﬁne a null beamwidth factor and relate the beams of other distributions to the uniform circular distribution beamwidth [Eq. (4-84)]. All other circular distributions relate to Eq. (4-82) or (4-83) through a beamwidth factor. The uniform distribution has a unity beamwidth factor.

4-17

CIRCULAR GAUSSIAN DISTRIBUTION [19]

A truncated Gaussian distribution has a simple functional relation: E(r) = e−ρr

2

|r| ≤ 1

(4-85)

We can easily calculate the edge taper through the conversion between logarithms: edge taper(dB) = 8.686ρ

(4-86)

We determine amplitude taper efﬁciency by substituting Eq. (4-85) into Eq. (4-8) and carrying out the integrations: ATL =

2(1 − e−ρ )2 ρ(1 − e−2ρ )

(4-87)

Table 4-29 lists designs for various sidelobe levels in terms of the single parameter: edge taper. Equation (4-86) relates the parameter ρ to the edge taper. Example Estimate the beamwidth of the pattern radiated from a circular distribution with a 13-dB edge taper and radius of three wavelengths.

HANSEN SINGLE-PARAMETER CIRCULAR DISTRIBUTION

TABLE 4-29

195

Circular-Aperture Gaussian Distribution, e −ρr (|r| < 1) 2

Sidelobe Level (dB)

Edge Taper (dB)

ATL (dB)

Beamwidth Factor

20 22 24 25 26 28 30 32 34 35 36 38 40

4.30 7.18 9.60 10.67 11.67 13.42 14.93 16.23 17.32 17.81 18.75 21.43 24.42

0.09 0.24 0.41 0.50 0.59 0.76 0.92 1.06 1.18 1.23 1.34 1.65 2.00

1.0466 1.0800 1.1109 1.1147 1.1385 1.1626 1.1839 1.2028 1.2188 1.2263 1.2405 1.2820 1.3296

We use linear interpolation in Table 4-29 to determine the beamwidth factor. From Eq. (4-82), 1.1568(0.5145) ◦ = 11.38 HPBW = 2 sin−1 6 From Eq. (4-83), ◦

HPBW = 58.95

1.1568 6

◦

= 11.36

The amplitude taper efﬁciency is calculated from Eq. (4-87): 13 = 1.497 8.686 2(1 − e−1.497 )2 = 0.847 ATL = 1.497(1 − e−2.993 ) ρ=

(−0.72 dB)

We obtain the same value by interpolating in Table 4-29. Sidelobes below 40 dB are difﬁcult to obtain with this distribution. The inner sidelobes continue to decrease with a decreasing edge level, but the outer lobes fail to reduce and dominate over the ﬁrst few sidelobes. Table 4-29 results from a search because no direct method exists for computing the edge taper for a speciﬁed sidelobe level.

4-18 HANSEN SINGLE-PARAMETER CIRCULAR DISTRIBUTION [20, 21] This distribution leads directly from sidelobe level to a single parameter H that relates through closed-form expressions to all other distribution parameters. The pattern of a uniform distribution is modiﬁed close in to the main beam. By using the U -space

196

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

variable of Taylor, we have U = (2a/λ) sin θ , where a is the radius. The pattern has different expressions in two regions: √ 2I1 (π H 2 − U 2 ) π√H 2 − U 2 f (U ) = √ 2J1 (π U 2 − H 2 ) √ π U2 − H2

|U | ≤ H

(4-88a)

|U | ≥ H

(4-88b)

I1 (x) is the ﬁrst-order modiﬁed Bessel function of the ﬁrst kind. The high function value of Eq. (4-88a) at the boresight reduces the sidelobes of the uniform distribution [Eq. (4-88b)], 17.57 dB, below the level at U = H . The sidelobe level is 2I1 (πH ) SLR = 17.57 + 20 log (4-89) πH Given the sidelobe level [positive (dB)], we use Eq. (4-89) in an iteration scheme to determine H . The aperture distribution is given by E(r) = I0 (πH 1 − r 2 )

|r| ≤ 1

(4-90)

where I0 is the zeroth-order modiﬁed Bessel function of the ﬁrst kind. Equation (4-8) can be integrated for this circularly symmetrical distribution [Eq. (4-90)] to derive the amplitude taper efﬁciency: ATL =

4I12 (πH ) π2 H 2 [I02 (πH ) − I12 (πH )]

(4-91)

Table 4-30 lists the parameters of the Hansen distribution for various sidelobe levels. At the top, Tables 4-29 and 4-30 are very similar. Any sidelobe level can be achieved with this distribution, subject to tolerance problems generated by any low-sidelobe design. The distribution is not optimum, but it is convenient.

4-19

TAYLOR CIRCULAR-APERTURE DISTRIBUTION [22]

Similar to the line source, the Taylor circular-aperture distribution modiﬁes inner zeros of the uniform amplitude and phase circular-aperture k-space pattern to approximate the Dolph–Chebyshev distribution. By use of the variable πU = ka sin θ the uniform distribution pattern is found to be J1 (πU )/πU . We remove n − 1 inner zeros and add those of the Dolph–Chebyshev distribution: n−1 J1 (πU ) (1 − U 2 /UN2 ) N=1 f (U ) = n−1 πU (1 − U 2 /SN2 ) N=1

(4-92)

TAYLOR CIRCULAR-APERTURE DISTRIBUTION

TABLE 4-30

197

Hansen Single-Parameter Circular-Aperture Distribution

Sidelobe Level (dB)

H

Edge Taper (dB)

ATL (dB)

Beamwidth Factor

20 22 24 25 26 28 30 32 34 35 36 38 40 45 50

0.48717 0.66971 0.82091 0.88989 0.95573 1.08027 1.19770 1.30988 1.41802 1.47084 1.52295 1.62525 1.72536 1.96809 2.20262

4.49 7.79 10.87 12.35 13.79 16.59 19.29 21.93 24.51 25.78 27.04 29.53 31.98 38.00 43.89

0.09 0.27 0.48 0.60 0.72 0.96 1.19 1.42 1.64 1.75 1.85 2.05 2.24 2.68 3.08

1.0484 1.0865 1.1231 1.1409 1.1584 1.1924 1.2252 1.2570 1.2876 1.3026 1.3174 1.3462 1.3742 1.4410 1.5039

Given a zero of J1 (x), J1 (x1N ) = 0, let x1N = πSN . By retaining approximately the same number of zeros in the visible region as in the uniform distribution, we avoid superdirectivity. The new zeros UN are modiﬁed zeros of the uniform distribution: A2 + (N − 12 )2 UN = Sn A2 + (n − 12 )2

(4-93)

where A relates to the maximum sidelobe level, cosh πA = b and 20 log b = sidelobe level(dB). Equation (4-93) is the same as Eq. (4-20) except for the scaling constant Sn , the nth zero of J1 (x) divided by π. Equation (4-92) gives the U -space pattern of the new distribution. We expand the aperture distribution in a Fourier–Bessel series: E(r) =

n−1

Bm J0 (πSm r)

r≤1

(4-94)

m=0

We compute coefﬁcients Bm by transforming the Fourier–Bessel series [Eq. (4-94)] into U -space and comparing the far-ﬁeld pattern with Eq. (4-92). As indicated in Eq. (4-94), the series contains only n nonzero terms: B0 = 1 Bm =

−

n−1

J0 (πSm )

(1 − Sm2 /UN2 )

N=1 n−1

N=1,N =m

(1 − Sm2 /SN2 )

m = 1, 2, . . . , n − 1

(4-95)

198

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

Example Design a Taylor circular-aperture distribution with 30 dB maximum sidelobes and n = 6. We use Eq. (4-21) to calculate the constant A: b = 1030/20 = 31.6228 cosh−1 b = 1.32 A= π We substitute this value into Eq. (4-93) to compute the ﬁve nulls: No. Null UN

1

2

3

4

5

1.5582

2.2057

3.1208

4.1293

5.1769

The ﬁrst null of the uniform distribution occurs at x11 x11 = 3.83171 S1 = = 1.2197 π We use this with the location of the ﬁrst zero to determine the null beamwidth factor: BWnull =

U1 1.5582 = = 1.2775 S1 1.2197

The coefﬁcients of the Fourier series [Eq. (4-95)] are given in Table 4-31. Figure 4-28 contains the k-space pattern.

10 Taylor 30 dB

Bayliss 27.4 dB

Pattern level, dB

20

30

40

50

60 −20

0 ka (sin q − sin q0)

20

FIGURE 4-28 Taylor and Bayliss circular aperture distributions to give 30-dB sidelobes (n = 6).

TAYLOR CIRCULAR-APERTURE DISTRIBUTION

199

Tables 4-32 to 4-34 list the characteristics for a few designs of the circular Taylor distribution. Table 4-32 shows that for each sidelobe level there is an optimum n. As the sidelobes are lowered, the optimum value of n increases. The blanks are unsuitable designs. The beamwidth factor (Table 4-33) and the null beamwidth factor (Table 4-34) continue to decrease as n increases at a given sidelobe level. In all three tables the values depend primarily on the sidelobe level. TABLE 4-31 Fourier–Bessel Series Coefﬁcients for Taylor Distribution: 30 dB, n = 6 Bm

Bm Normalized

Function

1.0000 0.93326 0.038467 −0.16048 0.16917 −0.10331

0.53405 0.49841 0.01808 −0.08570 0.09035 −0.05517

1 J0 (x11 r) J0 (x12 r) J0 (x13 r) J0 (x14 r) J0 (x15 r)

No. 0 1 2 3 4 5

TABLE 4-32 Amplitude Taper Losses of Taylor Circular-Aperture Distribution (dB) Sidelobe Level (dB) n

25

30

35

40

45

50

4 6 8 12 16 20

0.30 0.28 0.43 1.03 1.85

0.71 0.59 0.54 0.62 0.86 1.20

1.14 1.03 0.94 0.86 0.87 0.94

1.51 1.48 1.40 1.28 1.22 1.20

1.84 1.88 1.82 1.71 1.64 1.60

2.23 2.21 2.12 2.05 2.01

TABLE 4-33 Beamwidth Factor of Taylor Circular-Aperture Distribution Sidelobe Level (dB) n

25

30

35

40

45

50

4 6 8 12 16 20

1.0825 1.0504 1.0295 1.0057 0.9927

1.1515 1.1267 1.1079 1.0847 1.0717 1.0634

1.2115 1.1957 1.1796 1.1580 1.1451 1.1367

1.2638 1.2581 1.2457 1.2262 1.2137 1.2054

1.3095 1.3149 1.3067 1.2899 1.2782 1.2701

1.3666 1.3632 1.3499 1.3391 1.3314

200

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

TABLE 4-34 Null Beamwidth Factor of Taylor Circular-Aperture Distribution Sidelobe Level (dB)

4-20

n

25

30

35

40

45

50

4 6 8 12 16 20

1.1733 1.1318 1.1066 1.0789 1.0643

1.3121 1.2775 1.2530 1.2244 1.2087 1.1989

1.4462 1.4224 1.4001 1.3716 1.3552 1.3442

1.5744 1.5654 1.5470 1.5197 1.5029 1.4920

1.6960 1.7056 1.6928 1.6680 1.6514 1.6402

1.8426 1.8370 1.8162 1.8003 1.7890

BAYLISS CIRCULAR-APERTURE DISTRIBUTION [8]

We can also design a Bayliss distribution (difference pattern) for circular apertures. This gives us the pattern necessary for monopulse tracking along one axis. The U space pattern has modiﬁed zeros to produce nearly equal sidelobes close in to the main lobes: n−1 (1 − U 2 /UN2 ) (4-96) f (U, φ) = cos φπU J1 (πU ) N=1 n−1 (1 − U 2 /µ2N ) N=0

where UN are the new zeros and πµN are zeros of J1 (πU ). Bayliss lists those zeros µN (Table 4-35). The inner zeros have been removed and replaced by new ones, UN . We compute the zeros in a manner similar to that used for a linear distribution (Section 4-7): ξN2 µ n A2 + n2 UN = A2 + N 2 µn A2 + n2

N = 1, 2, 3, 4 (4-97) N = 5, 6, . . . , n − 1

The four inner zeros had to be adjusted to achieve the desired sidelobe level. Bayliss found these through a computer search. The values for ξN and A can be found through the polynomial approximations [Eq. (4-40)]. TABLE 4-35 Bessel Function Zeros, J1 (πµN ) N 0 1 2 3 4

µN 0.5860670 1.6970509 2.7171939 3.7261370 4.7312271

N

µN

N

µN

N

µN

5 6 7 8 9

5.7345205 6.7368281 7.7385356 8.7398505 9.7408945

10 11 12 13 14

10.7417435 11.7424475 12.7430408 13.7435477 14.7439856

15 16 17 18 19

15.7443679 16.7447044 17.7450030 18.7452697 19.7455093

BAYLISS CIRCULAR-APERTURE DISTRIBUTION

201

Like the Taylor circular aperture distribution, the aperture distribution is expanded in a ﬁnite-length Fourier–Bessel series: E(r, φ) = cos φ

n−1

Bm J1 (πµm r)

r ≤1

(4-98)

m=0

where the coefﬁcients are found by transforming Eq. (4-98) and comparing it with a U -space pattern [Eq. (4-96)]. The coefﬁcients are given by n−1 (1 − µ2m /UN2 ) µ2m N=1 Bm = j J1 (πµm ) n−1 (1 − µ2m /µ2N )

(4-99)

N=0,N =m

Example Design a Bayliss circular-aperture distribution with 30-dB sidelobes and n = 6. We start with Eq. (4-40) to compute coefﬁcients A and ξN : A = 1.64126

ξ1 = 2.07086

ξ2 = 2.62754

ξ3 = 3.43144

ξ4 = 4.32758

We substitute these constants into Eq. (4-97) along with the zeros from Table 4-35 to calculate the modiﬁed zeros: N

1

2

3

4

5

UN

2.2428

2.8457

3.7163

4.6868

5.6994

We use the zeros in Eq. (4-96) to calculate the pattern. The U -space pattern peak can be found by using Eq. (4-41): Umax = 0.7988

ka sin θmax = πUmax = 2.5096

where a is the aperture radius. The coefﬁcients of the Fourier–Bessel series are found from Eq. (4-99) (Table 4-36). The normalized coefﬁcients give an aperture distribution peak of 1. The 3-dB pattern points can be found by searching the pattern: ka sin θ1 = 1.3138

ka sin θ2 = 4.2384

TABLE 4-36 Fourier–Bessel Series Coefﬁcients for Bayliss Distribution: 30 dB, n = 6 No. 0 1 2 3 4 5

Bm

Bm Normalized

Function

0.62680 0.50605 −0.06854 −0.0028703 0.014004 −0.011509

1.2580 1.0157 −0.03415 −0.005761 0.028106 −0.02310

J1 (πµ0 r) J1 (πµ1 r) J1 (πµ2 r) J1 (πµ3 r) J1 (πµ4 r) J1 (πµ5 r)

202

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

TABLE 4-37 Characteristics of a Bayliss Circular-Aperture Distribution, n = 10 3-dB Beam Edge

Sidelobe Level (dB)

Beam Peak, ka/2 sin θmax

ka/2 sin θ1

ka/2 sin θ2

ATL (dB)

PEL (dB)

20 25 30 35 40

2.2366 2.3780 2.5096 2.6341 2.7536

1.165 1.230 1.290 1.346 1.399

3.700 3.940 4.160 4.363 4.551

1.47 1.15 1.32 1.62 1.95

1.80 1.89 1.96 2.01 2.05

Figure 4-28 contains a plot of a Bayliss circular-aperture distribution (n = 6) designed to have sidelobes 30 dB below those of the Taylor distribution with 30-dB sidelobes. The losses to the difference pattern are about 2.6 dB higher than the sum pattern. The amplitude taper efﬁciency is calculated from 2 n−1 4 Bm J1 (πµm r) r dr m=0 0 AT L = 2 1 n−1 2 π 4 Bm J1 (πµm r) r dr

1

(4-100)

m=0

0

where the integrals over φ have been separated and evaluated. An integral expression for the phase error efﬁciency can be found similarly by evaluating the separable cos φ integrals along the coordinate φ = 0, the peak:

PEL(U ) =

2 Bm J1 (πµm r)J1 (πU r)r dr m=0 0 2 1 n−1 4 Bm J1 (πµm r) r dr

2π

1

0

n−1

(4-101)

m=0

Table 4-37 lists the parameters of Bayliss circular-aperture distributions with n = 10 and various sidelobe levels. The optimum design for n = 10 occurs for 25-dB sidelobes.

4-21

PLANAR ARRAYS

We design planar arrays with nearly circular boundaries by sampling circular distributions. Given enough sample points in the array, a distribution such as the circular Taylor will be modeled adequately to produce a similar pattern. We can use pattern multiplication to combine the designs for linear arrays into a planar array, but in the special case of a square array, a true Chebyshev design can be obtained in all planes. A technique has been developed to allow the synthesis from pattern nulls provided that some of the possible nulls are not speciﬁed. We are still left with the problem of specifying the numerous nulls possible with a planar array. Chebyshev Array [23] When we combine two Dolph–Chebyshev linear arrays through pattern multiplication, it produces a pattern that has lower sidelobes than

203

CONVOLUTION TECHNIQUE FOR PLANAR ARRAYS

those speciﬁed in all planes except the principal ones along the axes. These designs give beamwidths in the diagonal planes that are wider than necessary. The pattern deviates from the optimum because sidelobes are suppressed more than necessary. We use a technique on a square array to produce equal sidelobes in all constant φ cuts around the array. The array is square in the number of elements, but different spacings along the axes can produce a rectangular array. We expand the pattern in a single Chebyshev polynomial: TL−1 (x0 cos ψ1 cos ψ2 )

(4-102)

where ψ1 = kdx cos θ cos φ + δ1 and ψ2 = kdy cos θ sin φ + δ2 L = 2N or L = 2N + 1 for L2 elements in the array. We compute x0 from Eq. (4-51) for a given sidelobe level. The pattern for an odd number of elements in each row and column is E(θ, φ) =

N+1 N+1

εm εn Imn cos 2(m − 1)ψ1 cos 2(n − 1)ψ2

L = 2n + 1

m=1 n=1

where εm = 1 for m = 1 and εm = 2 for m = 1. Similarly, E(θ, φ) = 4

N N

Imn cos(2m − 1)ψ1 cos(2n − 1)ψ2

L = 2n

m=1 n=1

The element excitations Imn are given by Imn

2 N+1 N+1 (q − 1)π 2 (p − 1)π cos = εp εq TL−1 x0 cos L p=1 q=1 L L × cos

2π(m − 1)(p − 1) 2π(n − 1)(q − 1) cos L L

L = 2N + 1 (4-103)

or

Imn

1 1 2 π π p− q− N N 4 2 2 = TL−1 x0 cos cos L p=1 q=1 L L

1 1 p− 2π n − 2π m − 2 2 cos × cos L

1 1 q− 2 2 L

L = 2N (4-104)

4-22 CONVOLUTION TECHNIQUE FOR PLANAR ARRAYS We may synthesize a desired pattern through multiplication of two or more simpler patterns. Because patterns derive from Fourier transforms of distributions in space, the distribution to produce the product of two simpler patterns is the convolution of the

204

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

simpler distributions [24, p. 30]. We ﬁnd it easier to synthesize by using a few elements and then build up patterns through multiplication. Consider the convolution of a linear array with another linear array on the same axis. We describe the array as a distribution consisting of weighted impulse functions, δ(x − xi ): N1 a1i δ(x − xi ) A1 (x) = i=1

where a1i are the feeding coefﬁcients and xi are the locations for an N1 -element array. To determine the array that gives the product of two array patterns, we convolve the second array, A2 (x), with the ﬁrst: A1 (x) ∗ A2 (x) =

A1 (τ )A2 (x − τ ) dτ

(4-105)

We evaluate a function at the argument of the impulse function when we convolve the two arrays [25, p. 237]. Equation (4-105) reduces to A1 (x) ∗ A2 (x) =

N2 N1

a1i a2i δ(x − xi − xj )

(4-106)

i=1 j =1

Example Consider the two 2-element arrays in Figure 4-29 and the graphical solution of the convolution. Figure 4-29a shows the location of the elements in the arrays on the x-axis. To perform the convolution, we reﬂect the x-axis of one array and move it across the other array while performing the integral at each location x, the coordinate of the convolution. We have a net result to the integral only when two impulse functions are aligned, x = xi + xj . We have four elements in the resulting array (Figure 4-29c). If the elements are equally spaced in the two arrays, two elements will sum into one. Patterns are the result of a three-dimensional Fourier transform. For a general array with element locations ri , we must perform a convolution along all three axes to ﬁnd the distribution that gives the product of the patterns of two simpler distributions. For the general array, Eq. (4-106) becomes A1 (r) ∗ A2 (r) =

a1i a2j δ(r − ri − rj )

(4-107)

where r is the location vector and ri and rj are the locations of elements in the two arrays. A rectangular array can be described as the convolution of a linear array on the xaxis with a linear array on the y-axis. When y = yj there is a string of values x = xi that satisfy the impulse argument [Eq. (4-107)]. These are the locations of the elements. We step through all values of yj until the entire array is formed. Equation (4-107) gives the feeding coefﬁcient of each element a1i a2j since no two elements of the convolution are in the same place. The pattern of the rectangular array is the product of the linear array patterns along the axes.

CONVOLUTION TECHNIQUE FOR PLANAR ARRAYS

205

a11 A1 (x) x11

a21

a12

A2 (x)

x21

x12

a22

x22

(a)

A1 (t) t

x

A2 (x − t) t

a11a22

x12 + x21 a12a21

x21 + x22

a11a21

x11 + x22

x11 + x21

(b)

a21a22 (c)

FIGURE 4-29 Convolution of two linear arrays: (a) separate arrays; (b) graphical convolution; (c) convolution.

Given an array, we compute the pattern from a Fourier transform containing N terms each of which corresponds to one element. Ignoring the element pattern, we have N E= ai ej k·ri (4-108) i=1

The array has N − 1 independent nulls (zeros) in the pattern. Given the set of nulls kj we can substitute them into Eq. (4-108) to form a matrix equation in N − 1 unknowns ai . We must normalize one coefﬁcient, ai = 1, to solve the set of equations for the

206

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

feeding coefﬁcients:

j k1 ·r1 e a2 ej k2 ·r1 a3 [B] . = − .. .. . j k ·r aN e N −1 1

(4-109)

where bij = ej ki ·rj +1 . We ﬁnd the direct solution of Eq. (4-109) unwieldy for a large array. We can subdivide the array into smaller arrays whose convolution is the total array and use pattern multiplication. We reduce the number of nulls we need to specify in the synthesis of an array. Convolution can be used in the synthesis of planar arrays by using a rhombic array (four elements) as the basic building block [26] (Figure 4-30). If we convolve two identically shaped rhombic arrays, we obtain a nine-element (three on a side) array (Figure 4-30b). By continuing to convolve the resulting array with other rhombic arrays, we can build up a large array in the shape of the rhombus. Each rhombic array has three pattern nulls without the symmetry of the linear array about some axis. The rhombic array has symmetry only about the plane of the rhombus. We build up an array of N + 1 by N + 1 elements through the convolution of N rhombic arrays. The original array has (N + 1)(N + 1) − 1 independent nulls. The convolution of N rhombic arrays reduces the number of independent nulls to 3N . Similarly, when we use the convolution of two linear arrays to form a square array, N + 1 by N + 1, the number of independent nulls is 2N , or N for each array. We denote a single rhombic array as RA1 and the convolution of two rhombic arrays as RA2 . The number of elements on each side of an RAN array is N + 1. We can convolute a rhombic array with a linear array to form an M × N array (M > N ). Denote the linear array by LN , where the array has N + 1 elements. The planar array P AM,N becomes P AM,N = LM−N ∗ RAN−1 (4-110) We specify 3(N − 1) nulls in space for the rhombic arrays and M − N nulls about the axis of the array. Like all convolutions, the pattern is the product of the individual array patterns.

(a)

(b)

FIGURE 4-30 Rhombic array with its convolution: (a) rhombic array RA1 ; (b) convolution of two rhombic arrays RA2 .

CONVOLUTION TECHNIQUE FOR PLANAR ARRAYS

207

This method allows the speciﬁcation of nulls in space with other than linear symmetry. Second, it reduces the required speciﬁcation of nulls. Third, it provides a method for synthesis of triangularly or hexagonally spaced elements. Example Consider the six-element rectangular array shown in Figure 4-31a. It can be broken down into the convolution of a four-element rectangular array (rhombic) and a two-element linear array from Eq. (4-110): P A3,2 = L1 ∗ RA1 Pick the three nulls of the rhombic array at θ

90◦

90◦

90◦

φ

110◦ −60◦ 180◦

We measure the pattern nulls from the normal of the plane containing the rhombus and the x-axis (φ). For an array of broadcast towers, the nulls point toward the horizon. We restrict θ to less than or equal to 90◦ . We substitute the positions of the elements (Figure 4-31b) and the nulls into Eq. (4-109) to solve for the feeding coefﬁcients of the rhombic array (Table 4-38).

(a)

(b)

(c)

FIGURE 4-31 Rectangular array from convolution of rhombic and linear arrays: (a) six-element rectangular array; (b) rhombic array; (c) linear array. TABLE 4-38 Coefﬁcients of Rhombic Array for Horizon (θ = 90◦ ) Nulls at φ = 100◦ , −60◦ , and 180◦ Element

Amplitude (dB)

Phase (deg)

1 2 3 4

0.00 4.12 0.00 4.12

0.0 −79.2 −109.2 −30.2

208

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

TABLE 4-39 Coefﬁcients of Six-Element Rectangular Array with Pattern of Figure 4-32b Element

Amplitude (dB)

Phase (deg)

1 2 3 4 5 6

0.00 8.13 4.12 0.00 8.13 4.12

0.0 −88.5 177.2 147.2 −124.3 −30.1

We pick the single null of the two-element array at 135◦ . This null has symmetry about the axis of the array. With the ﬁrst element at zero phase, we pick the element phase to cancel the ﬁrst element voltage when θ = 135◦ : ◦

◦

◦

◦

phase = 180 − 360 (0.3)cos 135 = 256.37

When we convolute the two arrays, we obtain the feeding coefﬁcients from Eq. (4-107) (Table 4-39). The elements in the center that result from two convolutions have summed feeding coefﬁcients producing a six-element array. Figure 4-32 shows the patterns of the convolution. We obtain the six-element array pattern (Figure 4-32b) by multiplication of the patterns of the individual subarrays (Figure 4-32a).

4-23

APERTURE BLOCKAGE

Blocking an aperture reduces the gain and raises the sidelobes. The blockage either scatters the aperture power in unwanted directions in a broad pattern or is just an area without ﬁelds. Scattered blockage causes higher sidelobes and greater loss than the

10 dB

Rhombic array

10 dB

20 dB

20 dB

30 dB

30 dB

Pattern of convolution

Linear array

(a)

(b)

FIGURE 4-32 Patterns of the convolution of a rhombic and a linear array to form the six-element rectangular array of Figure 4-31.

APERTURE BLOCKAGE

209

nonexcitation blockage. Scattered blockage has the same power input as the unblocked aperture, but ﬁelds scattered off the blockage do not contribute signiﬁcantly to the maximum ﬁeld. Compared to the unblocked aperture, the blockage efﬁciency becomes 2 j k·r Ee ds blocked max blockage efﬁciency = 2 j k·r Ee ds unblocked

scattered

(4-111)

max

We use Eq. (2-16) to compute the directivity of each distribution by using the total power radiated from the unblocked aperture [denominator of Eq. (2-16)] for the blocked aperture. A centrally blocked circular aperture with a uniform distribution has the blockage efﬁciency (1 − b2 )2 (scattered), where b is the normalized blockage radius. The blockage of a circular Gaussian distribution has a simple blockage function: blockage efﬁciency =

e−ρb − e−ρ 1 − e−ρ

2

The second type of blockage is an area without ﬁelds. The blockage does not waste power in the aperture. When we take the ratio of the two directivities, we must account for the power in each aperture: 2 j k·r Ee ds |E|2 ds blocked unblocked max blockage efﬁciency = 2 Eej k·r ds |E|2 ds unblocked

max

blocked

nonexcitation

(4-112) The blockage of a uniformly excited centrally blocked circular aperture where the center is not excited reduces the directivity only by the area lost from the aperture, 1 − b2 (nonexcitation). In a sense, nonexcitation blockage is not a true loss; it is a loss of potential radiation aperture. Table 4-40 lists the blockage losses of centrally blocked circular apertures calculated by Eq. (4-111), the more severe case. The uniformly excited aperture is affected least by blockage. All points are equally important. The tapered distributions suffer more loss with increased taper toward the edge. The lists for different tapered distributions track each other fairly closely, and any one of them gives a good estimate of the blockage loss. Blockage causes sidelobes. In the case of scattered blockage the exact sidelobes cannot be found without an analysis of the scatterer. A Cassegrain reﬂector would need a geometric theory of diffraction (GTD) analysis to locate the directions of scattering from the subreﬂector. We can handle the nonexcitation blockage in a general fashion. Consider the aperture to be broken into two radiating apertures. The ﬁrst is the unblocked aperture; the second is the blockage. If we take the blockage aperture to be 180◦ out of phase with respect to the unblocked aperture distribution, the sum gives us the blocked distribution. We use this analysis as an approximation to scattered blockage with the realization that scattering may produce unpredicted lobes.

210

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

TABLE 4-40 Blockage Losses of Circular-Aperture Distributions (dB) Central Blockage (%) 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

Taylor

Hansen

Uniform

Gaussian 12-dB Edge

30 dB, n = 6

40 dB, n = 6

30 dB

40 dB

0.02 0.03 0.04 0.06 0.07 0.09 0.11 0.13 0.15 0.17 0.20 0.22 0.26 0.29 0.32 0.36 0.39 0.43 0.47 0.52 0.56

0.04 0.06 0.08 0.10 0.13 0.16 0.19 0.23 0.27 0.32 0.36 0.41 0.47 0.52 0.58 0.65 0.71 0.78 0.86 0.94 1.02

0.04 0.06 0.08 0.10 0.13 0.16 0.20 0.24 0.28 0.32 0.37 0.42 0.48 0.54 0.60 0.67 0.74 0.81 0.88 0.96 1.05

0.05 0.08 0.11 0.14 0.18 0.22 0.26 0.31 0.37 0.43 0.49 0.56 0.63 0.71 0.79 0.88 0.97 1.07 1.17 1.27 1.38

0.05 0.07 0.09 0.12 0.16 0.19 0.23 0.28 0.33 0.38 0.43 0.49 0.56 0.63 0.70 0.77 0.86 0.94 1.03 1.12 1.22

0.07 0.09 0.13 0.17 0.21 0.26 0.32 0.38 0.44 0.51 0.59 0.67 0.76 0.85 0.95 1.06 1.17 1.28 1.40 1.53 1.66

We can calculate an upper bound on the sidelobes easily. Assume that the blockage distribution is uniform and compared to the main aperture, produces a broad, ﬂat beam. Since the blockage aperture ﬁelds are 180◦ out of phase from the unblocked aperture ﬁelds, their radiation subtracts from the main beam and adds sidelobes 180◦ out of phase with respect to the main lobe. The sidelobe due to the blockage is proportional to the area: sidelobe level = 20 log b. This formula estimates values much higher than are realized. Table 4-41 lists the sidelobes of a centrally blocked Taylor circular aperture distribution with 40-dB design sidelobes. They are far less than predicted by the upper bound.

TABLE 4-41 Sidelobe Level Due to Central Blockage of a Circular Aperture with Taylor Distribution (40 dB, n = 6) Blockage (% of Diameter)

Sidelobe Level (dB)

Blockage (% of Diameter)

Sidelobe Level (dB)

Blockage (% of Diameter)

Sidelobe Level (dB)

7 8 9 10 11 12

34.5 32.8 31.3 29.8 28.5 27.3

13 14 15 16 17 18

26.1 25.6 24.2 23.3 21.7 21.7

19 20 21 22 23 24

21.1 20.4 19.7 19.1 18.5 18.0

QUADRATIC PHASE ERROR

211

Ludwig [27] has found distributions to reduce the sidelobes of blocked apertures. The ﬁrst sidelobe can be reduced only a little, but the outer sidelobe levels can be controlled. In many applications one high sidelobe next to the main beam is acceptable. A Taylor distribution for circular apertures with a zero edges value, like Section 4-5 for linear apertures, reduces the far-out sidelobes. A second aperture function with a doughnut distribution also reduces all but the ﬁrst sidelobe. Reducing the edge taper of the blockage distribution lowers the blockage-caused sidelobes. Sachidananda and Ramakrishna [28] use a numerical optimization technique to reduce the sidelobes of a blocked aperture for both the sum and difference patterns of a monopulse excitation. They start with the Taylor and Bayliss circular-aperture distribution functions [Eqs. (4-94) and (4-98)]. The coefﬁcients Bm are determined through the numerical optimization, which restrains the sidelobes while optimizing the monopulse tracking coefﬁcients and sum pattern gain. 4-24 QUADRATIC PHASE ERROR A linear phase error function scans the aperture beam with some loss of gain because of the shrinking of the projected aperture in the direction of the main beam. Quadratic phase error (order 2) does not scan the beam but causes loss and a change in the sidelobe levels and the depth of the nulls between them. This phase error arises mainly from defocusing when the source of radiation appears as a point source. A feed axially displaced from the focus of a parabolic reﬂector produces quadratic phase error in the aperture. The ﬂare angle of a horn changes the distance from the assumed point source in the throat to different points in the aperture at the end of the ﬂare. We can approximate the phase distribution as quadratic. We express the quadratic phase error in a line-source aperture as e−j 2πS(2x/a)

2

linear:

|x/a| ≤ 0.5

(4-113a)

where S is a dimensionless constant, cycles and a is the aperture width. Similarly, the circular-aperture phase is circular: e−j 2πSr

2

r ≤1

(4-113b)

where r is the normalized radius. We use Eq. (4-7) with the linear-source aperture phase error [Eq. (4-113a)] and use Eq. (4-9) with the quadratic phase error [Eq. (4-113b)] in a circularly symmetrical aperture distribution to compute phase error loss:

PELx =

PEL =

2 2 E(x)e−j 2πS(2x/a) dx −a/2 2 a/2 |E(x)| dx a/2

linear

(4-114)

−a/2

2 E(r)e r dr 0 2 1 |E(r)|r dr 1

−j 2πSr 2

0

circular

(4-115)

212

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

A few distributions have simple formulas for the phase error efﬁciency when excited with quadratic phase error [29]: PELx =

uniform linear:

√ & 1 % 2 √ C (2 S) + S 2 (2 S) 2S

(4-116)

where C(t) and S(t) are the Fresnel integrals, tabulated functions: uniform circular: PEL = Circular Gaussian(e−ρr ) : 2

PEL =

sin πS πS

2 (4-117)

ρ 2 [1 − 2e−ρ cos(2πS) + e−2ρ ] (4-118) [ρ 2 + (2πS)2 ](1 − e−ρ )2

We use numerical integration for the general distribution. Table 4-42 lists quadratic phase error losses for various linear-aperture distributions. We will use the lists for uniform and cosine distributions to evaluate the gains of rectangular horns. The effect of quadratic phase error decreases as the distribution taper increases. Table 4-43 lists results for a few circularly symmetrical aperture distributions. Quadratic phase error raises the sidelobes of low-sidelobe antennas. Figure 4-33 shows the effects on a circular Taylor distribution with 35-dB design sidelobes. The ﬁrst lobe increases, and the null between the main lobe and the ﬁrst sidelobe disappears as the quadratic phase error increases. A source antenna spaced a ﬁnite distance, as on

TABLE 4-42 Quadratic Phase Error Loss of Linear-Aperture Distributions (dB) Cycles, S 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00

Uniform

Cosine

Cosine

Cosine2 + 19.9-dB Pedestal

0.04 0.15 0.34 0.62 0.97 1.40 1.92 2.54 3.24 4.04 4.93 5.91 6.96 8.04 9.08 9.98 10.60 10.87 10.80 10.50

0.02 0.07 0.16 0.29 0.45 0.65 0.88 1.14 1.43 1.75 2.09 2.44 2.82 3.20 3.58 3.95 4.31 4.65 4.97 5.25

0.01 0.04 0.09 0.16 0.25 0.36 0.49 0.64 0.80 0.97 1.16 1.36 1.57 1.79 2.01 2.23 2.46 2.69 2.91 3.13

0.02 0.07 0.16 0.28 0.44 0.63 0.84 1.08 1.34 1.62 1.90 2.19 2.48 2.76 3.04 3.29 3.52 3.73 3.92 4.09

2

213

QUADRATIC PHASE ERROR

TABLE 4-43

Cycles, S

Quadratic Phase Error Loss of Circular-Aperture Distributions (dB)

Uniform

0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00

0.04 0.14 0.32 0.58 0.91 1.33 1.83 2.42 3.12 3.92 4.86 5.94 7.20 8.69 10.46 12.62 15.39 19.23

Gaussian 12-dB Edge 0.03 0.13 0.29 0.53 0.82 1.20 1.64 2.16 2.76 3.44 4.22 5.08 6.04 7.10 8.24 9.44 10.66 11.81 12.75 13.36

Taylor

Hansen

30 dB

40 dB

30 dB

40 dB

0.04 0.15 0.33 0.59 0.93 1.36 1.86 2.46 3.16 3.95 4.86 5.88 7.01 8.25 9.56 10.87 12.01 12.80

0.03 0.11 0.26 0.46 0.72 1.03 1.41 1.84 2.33 2.87 3.47 4.11 4.79 5.50 6.21 6.91 7.56 8.14 8.62 8.99

0.03 0.11 0.25 0.45 0.70 1.01 1.38 1.81 2.30 2.85 3.46 4.16 4.85 5.63 6.43 7.26 8.09 8.88 9.60 10.20

0.02 0.08 0.19 0.34 0.53 0.76 1.03 1.34 1.69 2.08 2.50 2.95 3.43 3.94 4.46 4.98 5.51 6.03 6.53 6.99

25

30

Level, dB

35

40

45

l/16 Maximum - aperture phase error distance =

2D l

2

l/32

l/64

0

50 6

7

8

9

10

2pa sin q q

FIGURE 4-33 Effects of quadratic phase error on 35-dB circular Taylor distribution (n = 6).

214

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

an antenna measurement range, feeds the aperture with a quadratic phase error. The source would have to be spaced 8D 2 /λ to measure the sidelobe level within 0.5 dB. Low-sidelobe antennas require greater distances than the usual 2D 2 /λ for accurate sidelobe measurement [30]. 4-25 BEAM EFFICIENCY OF CIRCULAR APERTURES WITH AXISYMMETRIC DISTRIBUTION From Eq. (1-27) we can derive an approximate formula for axisymmetric distributions that depends on the normalized variable kr (or U ). For large apertures we can approximate cos θ ≈ 1 in the main beam, integrate the φ integral to obtain 2π, and incorporate the (ka)2 directivity factor into the integral:

kr1

ATL · PEL beam efﬁciency =

|f (kr )|2 kr dkr

(4-119)

0

2|f (0)|2 u1 |f (U )|2 U dU ATL · PEL

=

(4-120)

0

2|f (0)|2

where kr is the factor (2πa sin θ )/λ, U (the Taylor distribution factor) is (2a sin θ )/λ, and a is the aperture radius. U1 and kr1 correspond to the beam edge θ1 . The integrals of Eqs. (4-119) and (4-120) cause underestimations of beam efﬁciency for small apertures when we ignore the cos θ factor, which should divide the argument of the integral. Table 4-44 lists beam edges in kr -space (2πa sin θ )/λ for various distributions along with the beam efﬁciency at the null beam edge. We can use it to determine the aperture size required for a given beam efﬁciency beamwidth speciﬁcation. Example Calculate the aperture radius to give a 90% beam efﬁcient beamwidth of 5◦ for the distribution: parabolic on 12-dB pedestal. TABLE 4-44 Beam Efﬁciencies of Axisymmetric Circular-Aperture Distributions

Distribution Uniform Parabolic Parabolic + 12-dB pedestal Taylor 30 dB, n = 6 30 dB, n = 10 40 dB, n = 6 Hansen 30 dB 40 dB

kr = 2πa sin θ/λ Speciﬁed Beam Efﬁciency (%)

Null, kr

Beam Efﬁciency at Null (%)

80

85

90

95

3.83 5.14 4.58

83.7 98.2 96.4

2.82 2.81 2.60

4.71 3.03 2.81

5.98 3.31 3.10

3.75 3.64

4.90 4.74 6.00

96.2 91.4 99.5

2.65 2.76 2.90

2.88 3.06 3.13

3.19 3.65 3.42

3.82

5.37 6.64

99.3 99.9

2.79 3.17

3.01 3.42

3.28 3.73

3.69 4.19

3.85

REFERENCES

215

From Table 4-43, 5◦ 2πa sin λ 2 3.10 a = = 11.31 λ 2π sin(5◦ /2)

kr (90%) = 3.10 =

The beam edge has cos 2.5◦ = 0.999, which justiﬁes the approximation in Eq. (4-119). REFERENCES 1. T. T. Taylor, Design of line source antennas for narrow beamwidth and low sidelobes, IEEE Transactions on Antennas and Propagation, vol. AP-4, no. 1, January 1955, pp. 16–28. 2. C. L. Dolph, A current distribution for broadside arrays which optimizes the relationship between beamwidth and sidelobe level, Proceedings of IEEE, vol. 34, June 1946, pp. 335–348. 3. R. C. Hansen, Linear arrays, Chapter 9 in A. W. Rudge et al., eds., The Handbook of Antenna Design, Vol. 2, Peter Peregrinus, London, 1982. 4. D. R. Rhodes, On a new condition for physical realizability of planar antennas, IEEE Transactions on Antennas and Propagation, vol. AP-19, no. 2, March 1971, pp. 162–166. 5. D. R. Rhodes, On the Taylor distribution, IEEE Transactions on Antennas and Propagation, vol. AP-20, no. 2, March 1972, pp. 143–145. 6. R. S. Elliott, Antenna Theory and Design, Prentice-Hall, Englewood Cliffs, NJ, 1981. 7. D. A. Pierre, Optimization Theory with Applications, Wiley, New York, 1969. 8. E. T. Bayliss, Design of monopulse antenna difference patterns with low sidelobes, Bell System Technical Journal, vol. 47, May–June 1968, pp. 623–650. 9. P. M. Woodward, A method of calculating the ﬁeld over a plane aperture required to produce a given polar diagram, Proceedings of IEE, vol. 93, pt. IIIA, 1947, pp. 1554–1558. 10. S. A. Schelkunoff, A mathematical theory of linear arrays, Bell System Technical Journal, vol. 22, 1943, pp. 80–107. 11. J. Kraus, Antennas, McGraw-Hill, New York, 1950. 12. R. S. Elliott, Beamwidth and directivity of large scanning arrays, Microwave Journal, vol. 6, no. 12, December 1963, pp. 53–60. 13. A. T. Villeneuve, Taylor patterns for discrete arrays, IEEE Transactions on Antennas and Propagation, vol. AP-32, no. 10, October 1984, pp. 1089–1092. 14. R. S. Elliott, On discretizing continuous aperture distributions, IEEE Transactions on Antennas and Propagation, vol. AP-25, no. 5, September 1977, pp. 617–621. 15. H. J. Orchard, R. S. Elliott, and G. J. Stern, Optimising the synthesis of shaped beam antenna patterns, IEE Proceedings, vol. 132, pt. H, no. 1, February 1985, pp. 63–68. 16. Y. U. Kim and R. S. Elliott, Shaped-pattern synthesis using pure real distributions, IEEE Transactions on Antennas and Propagation, vol. AP-36, no. 11, November 1988, pp. 1645–1648. 17. F. Ares-Pena, Application of genetic algorithms and simulated annealing to some antenna problems, Chapter 5 in Y. Rahamat-Samii and E. Michielssen, eds., Electromagnetic Optimization by Genetic Algorithms, Wiley, New York, 1999. 18. C. H. Walters, Traveling Wave Antennas, Dover, New York, 1970. 19. G. Doundoulakis and S. Gethin, Far ﬁeld patterns of circular paraboloidal reﬂectors, IRE National Convention Record, pt. 1, 1959, pp. 155–173.

216

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

20. R. C. Hansen, Circular aperture distribution with one parameter, Electronic Letters, vol. 11, no. 8, April 17, 1975, p. 184. 21. R. C. Hansen, A one-parameter circular aperture with narrow beamwidth and low sidelobes, IEEE Transactions on Antennas and Propagation, vol. AP-24, no. 4, July 1976, pp. 477–480. 22. T. T. Taylor, Design of circular apertures for narrow beamwidth and low side lobes, IEEE Transactions on Antennas and Propagation, vol. AP-8, no. 1, January 1960, pp. 17–22. 23. F. I. Tseng and D. K. Cheng, Optimum scannable planar arrays with an invariant side-lobe level, Proceedings of IEEE, vol. 56, no. 11, November 1968, pp. 1771–1778. 24. D. Steinberg, Principles of Aperture and Array System Design, Wiley, New York, 1976. 25. D. K. Cheng, Analysis of Linear Systems, Addison-Wesley, Reading, MA., 1959. 26. S. R. Laxpati, Planar array synthesis with prescribed pattern nulls, IEEE Transactions on Antennas and Propagation, vol. AP-30, no. 6, November 1982, pp. 1176–1183. 27. A. C. Ludwig, Low sidelobe aperture distributions for blocked and unblocked circular apertures, IEEE Transactions on Antennas and Propagation, vol. AP-30, no. 5, September 1982, pp. 933–946. 28. M. Sachidananda and S. Ramakrishna, Constrained optimization of monopulse circular aperture distribution in the presence of blockage, IEEE Transactions on Antennas and Propagation, vol. AP-31, no. 2, March 1983, pp. 286–293. 29. A. W. Love, Quadratic phase error loss in circular apertures, Electronics Letters, vol. 15, no. 10, May 10, 1979, pp. 276, 277. 30. P. S. Hacker and H. E. Schrank, Range requirements for measuring low and ultralow sidelobe antenna patterns, IEEE Transactions on Antennas and Propagation, vol. AP-30, no. 5, September 1982, pp. 956–966.

5 DIPOLES, SLOTS, AND LOOPS

A dipole is a conductive rod usually split in the center and fed from a balanced transmission line that carries equal and oppositely ﬂowing currents. Not all dipoles are split and fed in the center because currents can be excited on it electromagnetically or it can be shunt fed. The dipole length determines possible current distributions in modes, and when we place a continuous rod near an antenna radiating a linear polarization component directed along the rod, it excites a standing-wave current on the rod. The amount excited on the rod depends on how close its length is to resonance and the antenna spacing. Of course, the continuous rod loads the fed antenna through mutual coupling. We can feed the continuous rod from a coax line by attaching the outer conductor to the center and then connecting the center conductor away from the center in a shunt feed. A slot is a narrow-width opening in a conductive sheet. When excited by a voltage across the narrow dimension it appears to radiate from an equivalent magnetic current ﬂowing along the long dimension that replaces the voltage (or electric ﬁeld) across it. Most slots, similar to dipoles, have a ﬁnite length with either short or open circuits at both ends. The voltage along the slot forms a standing wave. Of course, magnetic currents are ﬁctitious, and real electric currents ﬂow in the conductive sheet around the slot. These currents do not have a simple distribution and are difﬁcult to use for analysis, so we use simpler magnetic currents, although when analyzing a slot using the method of moments, we model the conductors around the slot and calculate patterns, reaction, and so on, from these real currents. Initial slot calculations assume that the conductive sheet is inﬁnite, similar to the analysis of dipoles situated in free space. Complete analysis of the dipole requires analysis in the presence of the mounting conﬁguration. Similarly, full analysis of slots includes the effects of the ﬁnite sheet and scattering from the objects around it.

Modern Antenna Design, Second Edition, By Thomas A. Milligan Copyright 2005 John Wiley & Sons, Inc.

217

218

DIPOLES, SLOTS, AND LOOPS

After considering ideal cases, we analyze the effects of ﬁnite ground planes, nearby scatterers, and the interaction between dipoles and slots. The batwing antenna presents an unusual case where the antenna at ﬁrst glance looks like a dipole but actually radiates from a combination of a slot and a ﬁnite dipole structure. Another interesting case is the waveguide slot. Currents ﬂow on the inside surfaces of a waveguide, and the ﬁnite current skin depth prevents it from reaching the outside. The metal walls shield the currents and prevent the loss of power by radiation. When we cut a slot in the wall, the internal currents ﬂow out the slot and onto the outside of the waveguide and radiate. The excitation and length of the slot relative to the internal currents determine the amount radiated. Similarly, the slots load the waveguide as a transmission line because of the loss of power. Our analysis starts with a dipole in free space or a slot on an inﬁnite conductive sheet. The two problems are duals. Dipoles radiate from a standing-wave electric (real) current, whereas the slot radiates from a standing-wave magnetic current. We use the same mathematics for both patterns. By the Babinet–Booker principle of complementary structures, we relate the input impedance of one to the other. Both structures radiate the same pattern but differ in polarization. Dipoles and slots share the same analysis through duality, so we develop them together. Singly and in arrays, they satisfy many antenna needs. Although they share a dual analysis, they have unique feeding requirements. We discuss baluns for dipoles and waveguide slot excitations as practical implementations. In Chapter 2 we presented the analysis of a small loop excited with a uniform current (Section 2-1.2). The loop current was replaced with a small magnetic current element ﬂowing along the normal to the plane of the loop. Multiple turns and ferrite loading increase the efﬁciency of loops and produce a more useful antenna. Exciting a uniform current on a loop is a difﬁcult task that offers little practical beneﬁt. The loops discussed will have standing-wave electric currents excited on them determined by feeding methods. The natural balun used to excite a small loop produces a standingwave current with zero current at the point where the two sides are connected to form the loop. A resonant length loop of about one wavelength perimeter radiates a dipole pattern from a standing-wave current. The quadriﬁlar helix consists of two loops twisted around a common axis. The twist produces currents that radiate circular polarization from each loop. Analysis shows that the currents are standing wave. Feeding a dipole or loop requires a balun to prevent current ﬂow either along the outside of a coaxial feeder or excitation of unbalanced currents along a two-wire line. The current ﬂowing along the outside of the coax or unbalanced currents on the twowire line radiate in unwanted directions or radiate undesired polarization. When we design an antenna without considering or knowing its ﬁnal mounting, we produce an uncontrolled situation without a balun. Our initial conﬁguration may work without a balun, but the antenna may fail to produce the desired pattern in the ﬁnal location. If you control the installation completely, you can reduce your design effort and may be able to eliminate the balun.

5-1 STANDING-WAVE CURRENTS Think of a dipole as a diverging two-wire transmission line. The characteristic impedance increases as the wave approaches the open-circuited ends. The slot is the

STANDING-WAVE CURRENTS

219

Current S Voltage

Open circuit

FIGURE 5-1 Standing wave.

dual of a strip dipole. A voltage excited across the slot propagates along a slotline toward short-circuited ends. Each type of transmission line reﬂects the incident wave from the terminations. The combination of two waves traveling in opposite directions creates a standing wave on the line. The current and voltage are 90◦ out of phase and 90◦ out of space phase (Figure 5-1). Current and voltage change places on the short-circuited termination of the slot. The dipole is not a uniform transmission line, but we can approximate the current as a standing wave with the current vanishing on the ends. The slot voltage is a standing wave also vanishing on the ends. The standing waves for a center-fed dipole or slot are expressed as follows: Dipole L I = I0 sin k −z 2 L I = I0 sin k +z 2

Slot L V = V0 sin k −z 2 L V = V0 sin k +z 2

z≥0

(5-1)

z≤0

The voltage distribution on the slot is equivalent to a magnetic current. We calculate radiation from the linear sinusoidal current distributions by the vector potentials: electric (slot) (Section 2-1.2) and magnetic (dipole) (Section 2-1.1). Figure 5-2 gives typical sinusoidal distributions for various lengths. The currents match at the feed point and vanish on the ends. Consider the pattern of the 2λ dipole at θ = 90◦ . We can assume that it is a continuous array and sum the ﬁelds from each z q I (V)

l/2

Dipole

Slot

I (V)

l

I (V)

3l/2

Electrical length

FIGURE 5-2 Sinusoidal distributions.

2l

220

DIPOLES, SLOTS, AND LOOPS 100

5

80

4

60

3

40

Beam peak, q

Directivity, dB

Beam peak

Directivity 2

20

1

0.5

1.0

1.5

2.0

0 3

2.5

Dipole length, l

FIGURE 5-3 Dipole (slot) directivity and beam peak versus length.

portion along the axis. The equal positive and negative portions of the standing-wave current sum to zero and produce a pattern null normal to the axis. By integrating Eqs. (2-5) and (2-10), we compute far ﬁelds for radiators centered on the z-axis through the far-ﬁeld conversion [Eqs. (2-1) and (2-9)] [1, p. 82]: Eθ = j η

I0 −j kr cos(kL/2 cos θ ) − cos(kL/2) e 2πr sin θ

dipole

(5-2)

where L is the total dipole length. Using the Y = 0 plane as the slot ground plane, the far-ﬁeld magnetic ﬁeld is found as Hθ =

±j V0 −j kr cos(kL/2 cos θ ) − cos(kL/2) e η2πr sin θ

slot

(5-3)

where L is the total slot length. We apply the upper sign for Y > 0 and the lower sign for Y < 0. The electric ﬁeld of the slot is found from Eφ = −ηHθ . Equations (5-2) and (5-3) have the same pattern shape and directivity. We integrate the magnitude squared of Eqs. (5-2) and (5-3) to determine the average radiation intensity. Joined with the maximum radiation intensity, we calculate directivity (Figure 5-3) versus length. 5-2 RADIATION RESISTANCE (CONDUCTANCE) The far-ﬁeld power densities, Poynting vectors, are given by 2 |Eθ | dipole Sr = η slot |Hθ |2 η

221

RADIATION RESISTANCE (CONDUCTANCE)

where η is the impedance of free space (376.7 ). When these are integrated over the radiation sphere to compute the power radiated, the results contain either |I0 |2 (dipole) or |V0 |2 (slot), the maximum sinusoidal current (voltage). We deﬁne the radiation resistance (conductance) as Pr |I0 |2 Pr Gr = |V0 |2 Rr =

dipole (5-4) slot

Figure 5-4 is a plot of the radiation resistance of each versus length [2, p. 157]. The input resistance differs from the radiation resistance because it is the ratio of the input current (voltage) to the power radiated: kL 2 kL Vi = V0 sin 2 Ii = I0 sin

dipole (5-5) slot

Combining Eqs. (5-4) and (5-5), we ﬁnd that Rr sin (kL/2)

Ri =

2

Gr Gi = 2 sin (kL/2)

dipole (5-6) slot

500 Dipole input

Resistance, Ω

400

Slot

300 Dipole

200

100

0

Slot input

0.5

1

1.5 Length, λ

2

2.5

FIGURE 5-4 Dipole and slot radiation and center-fed input resistances.

3

222

DIPOLES, SLOTS, AND LOOPS

The input resistances (Figure 5-4) differ from the radiation resistances by Eq. (5-6). The input resistance of a one-wavelength dipole is large but not inﬁnite, as shown; it depends greatly on the diameter and input region. If we take the product of the radiation or input resistances, we determine that Rdipole Rslot =

η2 4

(5-7)

one of the consequences of the Babinet–Booker principle [3]. The input resistance depends on the current at the input [Eq. (5-6)]. When the standing-wave current is high and the voltage is low, the input resistance is moderate. A center-fed half-wavelength dipole has the same input resistance as radiation resistance, since the current maximum occurs as the input. On the other hand, a center-fed halfwavelength slot has a current minimum (voltage maximum) at its input, which gives it high input resistance. When both are a full wavelength long, the dipole standingwave current is at a minimum and the slot standing-wave current is at a maximum (Figure 5-2). The dipole has a high input resistance and the slot has a low input resistance. We can lower the input resistance by feeding at a high current point, but we may excite a distribution different from that expected. A short dipole looks like a capacitor at the input. As the length increases, the radiation resistance grows and the capacitance decreases. Just before the length reaches λ/2, the capacitance becomes zero. The exact length at which the antenna resonates (zero reactance) depends on the diameter of the elements and the input gap. A good starting point is 95% of a half wavelength. Beyond the resonant length, the dipole becomes inductive. The impedance of a thin half-wavelength dipole is 73 + j 42.2 , whereas the resonant-length dipole resistance is about 67 . The slot looks like an inductor when short. Think of it as a short-length short-circuited shunt slotline stub. The inductance increases as its length increases and the slot resonates like the dipole, just short of λ/2. Additional resonances occur at longer lengths. Increasing the frequency is equivalent to increasing the length for the thin dipole.

5-3 BABINET–BOOKER PRINCIPLE [3; 4, p. 337] A strip dipole and a slot are complementary antennas. The solution for the slot can be found from the solution to an equivalent dipole by an interchange of the electric and magnetic ﬁelds. Not only the pattern but also the input impedance can be found. Figure 5-5 shows two such complementary structures. Babinet’s principle of optical screens (scalar ﬁelds) states that given the solutions to the diffraction patterns of a screen, Fi , and the screen’s complement, Fc , the sum equals the pattern without the screen. Booker extended Babinet’s principle to vector electromagnetic ﬁelds. Strict complementation of an electric conductor requires a nonexistent magnetic conductor. Booker solved this problem by using only perfectly conducting inﬁnitesimally thin screens and by interchanging the electric and magnetic ﬁelds between the screen and its complement. If we take two such complementary screens and perform line integrals over identical paths to compute the impedance of each, we obtain the result Z1 Zc =

η2 4

(5-8)

DIPOLES LOCATED OVER A GROUND PLANE

223

Slot

FIGURE 5-5

Complementary screens.

where Z1 is the input impedance of the structure, Zc the input impedance of the complementary structure, and η the impedance of free space (376.7 ). Equation (5-8) extends Eq. (5-7) to the total impedance and includes mutual impedances as well as self-impedances. Certain antennas, such as ﬂat spirals, are self-complementary—an exchange of the spaces and conductors leaves the structure unchanged except for rotation. For a twoarm structure, η2 Z02 = or Z0 = 188 4 Rumsey [5, p. 28] extended these ideas to antennas with more than two conductors to determine the input impedances in various feeding modes. We must relate ﬂat-strip dipoles to normal round-rod dipoles to use the available results for round dipoles. The diameter of an equivalent round rod equals one-half the strip width of the ﬂat structure. Consider a thin dipole with its near λ/2 resonance of 67 . We calculate equivalent slot impedance from Eq. (5-8): Zslot =

376.72 = 530 4(67)

A half-wavelength slot impedance is Zslot =

376.72 = 363 − j 211 4(73 + j 42.5)

The λ/2 dipole is inductive when it is longer than a resonant length, whereas the slot is capacitive.

5-4 DIPOLES LOCATED OVER A GROUND PLANE We analyze a dipole over a ground plane as a two-element array of the dipole and its image. The ground plane more than doubles the gain of the element by limiting the radiation directions. We can expect a change in the input impedance as the dipole interacts with its image. A vertical dipole excites currents in the ground plane, when transmitting, equivalent to its image. The image is vertical (Figure 5-6) and has the same phase as the dipole (even mode). The impedance of the dipole becomes

224

DIPOLES, SLOTS, AND LOOPS

FIGURE 5-6

Ground-plane images.

Z = Z11 + Z12 . Z12 is the mutual impedance between the dipole and its image spaced 2H , where H is the center height of the dipole over the ground plane. The array radiates its maximum in the direction of the ground plane. The dipole also radiates its maximum pattern along the ground plane given by Umax

η|I0 |2 = (2π)2

kL 1 − cos 2

2 (5-9)

where L is the dipole length. The radiated power of the single dipole is R12 Pin = R11 |I0 |2 1 + R11 The two-element array increases the ﬁeld over a single element by 2 and the radiation intensity by 4: 4Ud,max 4η[1 − cos(kL/2)]2 directivity = = Pin /4π (R11 + R12 )π We used only the power into the dipole, since no source is connected to the image. Figure 5-7 is a plot of the directivity of a vertical dipole versus height over the ground plane. A horizontal dipole and its image (Figure 5-6) form an odd-mode two-element array (Section 3-1). The input impedance of the dipole becomes Z11 − Z12 for the odd-mode array. The value of the mutual impedance Z12 approaches that of the self-impedance Z11 as the two dipoles move close together. The input impedance approaches zero as the distance from the dipole to ground plane shrinks. The input impedance of all oddmode array elements decreases as the elements approach each other. The two-element odd-mode array produces a null along the ground plane. The beam peak occurs normal to the ground plane (θ = 0◦ ) when the distance between the dipole and its image is less than λ/2 or H ≤ λ/4. The pattern bifurcates after that height is exceeded. The maximum radiation from the array is

UA,max

2πH 4 sin2 λ = 4

λ 4 λ H ≥ 4

H ≤

θmax = cos−1

λ 4H

DIPOLE MOUNTED OVER FINITE GROUND PLANES

225

9 Horizontal

Directivity, dB

8.5 Vertical

8.0

7.5

7.0

6.5

0

FIGURE 5-7

0.5

1.0 Height above ground plane, l

1.5

2

Directivity of half-wavelength dipoles over a ground plane.

The dipole pattern [Eq. (5-2)] increases the radiation intensity. The total input power into the single dipole becomes Pin = |I0 |2 (R11 − R12 ) UA,max Ud,max directivity = Pin /4π After inserting the various terms, we obtain the directivity of a horizontal dipole over ground: 4η sin2 (2πH /λ)[1 − cos(kL/2)]2 π(R11 − R12 ) directivity = 2 4η[1 − cos(kL/2)] π(R11 − R12 )

λ 4 λ H ≥ 4

H ≤

Its plot is included in Figure 5-7.

5-5 DIPOLE MOUNTED OVER FINITE GROUND PLANES Most conﬁgurations have a dipole mounted over a ﬁnite ground plane. You can calculate the ﬁnal pattern by using GTD, PO, or MOM, or you can measure the pattern using the actual ground plane. Analyses produce idealized patterns, and measurements contain errors due to the presence of the positioner mounting. If the ﬁnal system requires

226

DIPOLES, SLOTS, AND LOOPS

E-Plane

10l 2l Dia H-Plane

l Diameter Disk

FIGURE 5-8 Dipole spaced λ/4 over disk ground planes with 1λ, 2λ, and 10λ diameters.

exacting patterns, it has no margin and will fail. In this section we consider dipoles with idealized ground planes and give you ideas about the ﬁnal performance or spur you to use the ground plane purposely as a design parameter. Figure 5-8 shows the result of a PO analysis of a dipole mounted λ/4 above ﬁnite disks 1, 2, and 10λ in diameter. The E-plane pattern contains a pattern null at 90◦ due to the dipole pattern. The ground plane restricts the broad H -plane pattern as pattern angles approach 90◦ and reduces the backlobe more and more as it increases in size. At one wavelength the disk increases the gain of the antenna from the 7.5 dB given in Figure 5-7 to 8.1 dB. We can size the ground plane to produce small gain increases. We can analyze ﬂat-plate reﬂectors from three perspectives. In the ﬁrst, plates restrict radiation directions and thereby increase directivity. Waves polarized parallel with the surface must vanish on the reﬂector surface and cause a greater restriction of the beam. We see this effect in Figure 5-7, which shows horizontal dipoles having greater directivities than vertical dipoles for close spacing over a ground plane. In the second method we use aperture theory to analyze the reﬂector by using an aperture plane and integrate the ﬁelds or evaluate illumination losses. If the phase of the ﬁelds on the aperture varies rapidly, we must either take ﬁne increments in numerical integration or evaluate only around areas of stationary phase. Third, we can replace the reﬂector with images and restrict the valid pattern region. In GTD this method is combined with diffractions to smooth the ﬁeld across shadow and reﬂection boundaries. In Section 5-4 we analyzed the pattern and gain of a dipole mounted over an inﬁnite ground plane by the method of images. The antenna and its image formed a two-element

DIPOLE MOUNTED OVER FINITE GROUND PLANES

227

TABLE 5-1 Results of a GTD Analysis of a Horizontal λ/2 Dipole λ/4 Over a Limited Square Ground Plane (H -Plane) Ground-Plane Size (λ)

Directivity (dB)

Front-to-Back Ratio (dB)

H -Plane Pattern Level at 90◦ (dB)

H -Plane Beamwidth (deg)

Phase Center (λ)

0.5 0.6 0.7 0.8 0.9 1.0 1.2 1.4 1.6 1.8 2.0 2.5 3.0 4.0 5.0 10

5.37 6.32 7.08 7.68 8.14 8.34 8.65 8.45 7.96 7.39 6.95 7.13 7.74 7.28 7.56 7.41

8.4 10.3 12.0 13.5 14.8 16.0 17.8 19.1 20.0 21.1 22.3 25.0 28.3 32.8 35.4 36

−6.3 −7.6 −8.8 −9.8 −10.6 −11.2 −12.0 −12.3 −12.2 −12.3 −12.4 −12.7 −13.8 −14.8 −16.2 −19.1

108.5 104.0 100.9 97.8 95.1 93.2 93.3 99.4 108.4 112.4 113.1 115.8 111.4 116.1 118.0 121.3

0.18 0.15 0.14 0.12 0.10 0.08 0.04 0.01 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0

array, but with real power into only one element. The imaging method gives limited information that can be ﬁlled with GTD methods. Table 5-1 lists the results of a GTD analysis of a half-wavelength horizontal dipole located λ/4 over a limited square ground plane. An inﬁnite ground plane and dipole combination has an inﬁnite front-toback (F/B) ratio with the ﬁelds vanishing in the ground-plane direction. By using the methods of Section 3-3, we calculate a 120◦ half-power beamwidth for the two-element half-wavelength spaced array of the dipole and its image. The F/B ratio increases as the reﬂector (ground plane) size increases. Unfortunately, F/B is only the ratio of two pattern angles. We could tune the size of the ground plane to produce a high F/B ratio for a nonsquare ground plane, but it holds for only a small range of angles. Figure 5-8 illustrates the general increase in F/B as the size of the ground plane increases. We expect zero ﬁelds at θ = 90◦ on an inﬁnite ground plane, and Table 5-1 shows a decrease of the ﬁelds with an increase of the ground plane. The half-power beamwidth cycles about 120◦ as the ground plane increases in size. Phase center is the apparent radiation center placed at the focus of a paraboloidal reﬂector when used as a feed. The phase center of the equivalent two-element array is located on the ground plane. As we decrease the ground plane, the effect of the image decreases and causes the phase center to move toward the dipole. In the limit of no ground plane, the phase center is on the dipole. Table 5-1 shows the small gain changes that occur as the relative phase of the ground-plane scattered ﬁelds and the dipole direct ﬁelds add in the far ﬁeld. The small ground plane at λ/2 square fails to signiﬁcantly limit radiation and gain drops. Peak gain occurs when the ground plane is 1.2λ square, but this result would not necessarily hold for a circular ground plane. In most applications the dipole cannot be mounted directly above the ground-plane center, but we can add a small ground plane to control

228

DIPOLES, SLOTS, AND LOOPS

H-Plane, 1l dia.

E-Plane, 10l dia. E-Plane, 1l dia.

H-Plane, 1l dia.

FIGURE 5-9 V-dipole spaced 0.35λ over and tilted 35◦ toward 1λ- and 10λ-diameter ground planes.

the pattern and then place the combination on a pedestal over the larger ground plane. Most cases should be analyzed or measured in the ﬁnal conﬁguration. The dipole E-plane pattern null can be reduced by tilting the two poles down toward the ground plane. Figure 5-9 illustrates the calculated pattern of a tilted element dipole above a ﬁnite disk ground plane. The feed point of the dipole has been raised to 0.35λ to allow for the 35◦ tilt to the poles. Tilt and ground-plane height give additional parameters to control the pattern of the dipole mounted over a ﬁnite ground plane. For example, a horizontal dipole located λ/2 over an inﬁnite ground plane forms an odd-mode (0◦ , 180◦ ) two-element array using the dipole and its image. The simple ray-tracing argument given in Section 3-1 predicts a pattern null at zenith. But when placed over a ﬁnite ground plane, the fainter image fails to produce a complete null. We sometimes mount a dipole spaced away from a metal cylinder that provides a ground plane to restrict radiation. The curved ground plane allows greater radiation around the cylinder when rays spread as they scatter from it. Figure 5-10 shows the horizontal plane pattern for a vertical dipole mounted near a 1λ-diameter cylinder for spacing of 0.25λ, 0.4λ, 0.5λ, and 0.75λ. When we space a dipole λ/2 above a large ﬂat ground plane, the pattern has a null normal to the plane. The cylinder is unable to generate a full image of the dipole to produce this null, but the pattern does dip 11.2 dB from the peak. A dipole spaced 3λ/4 over a ground plane produces a threelobed pattern that we can see in Figure 5-10 except that the cylinder can produce only 8-dB dips. If we mount the dipole over a 2λ-diameter cylinder, the pattern is similar to Figure 5-10 except that F/B increases and the nulls have greater depths. Table 5-2 summarizes pattern results for vertical dipoles mounted over small cylinders.

DIPOLE MOUNTED OVER FINITE GROUND PLANES

229

0.4l l/2 3l/4

l/4 Spacing

FIGURE 5-10 Horizontal plane pattern for a vertical dipole mounted near a 1λ-diameter cylinder at 0.25λ, 0.4λ, 0.5λ, and 0.75λ distances.

TABLE 5-2

Dipole Mounted Over a Cylinder Aligned with a Cylinder Gain (dB)

Height (λ) Over Cylinder

Cylinder Diameter (λ)

At 0

0.25

0.25 0.50 1.0 2.0 0.25 0.50 1.0 2.0 0.25 0.50 1.0 2.0 0.25 0.50 1.0 2.0

3.5 6.1 6.7 7.3 3.2 3.6 2.2 2.2 0.5 −2.9 −5.9 −8.7 3.3 5.0 5.1 6.4

0.4

0.5

0.75

◦

At 180◦

Gain Peak

Peak Angle

−2.1 −2.7 −6.1 −10.7 0.3 −1.3 −5.3 −9.7 0.9 −1.8 −4.2 −8.5 −0.2 −0.9 −3.2 −6.8

3.6 6.1 6.7 7.3 4.9 6.0 5.1 6.0 5.2 4.8 5.2 5.9 3.4 4.7 4.6 5.2

0 0 0 0 64 62 60 54 80 80 76 70 102 102 98 90

230

DIPOLES, SLOTS, AND LOOPS

l/4

0.4l l/2 3l/4 Spacing

FIGURE 5-11 Horizontal dipole mounted over a vertical 1λ diameter pole at 0.25λ, 0.4λ, 0.5λ, and 0.75λ distances.

TABLE 5-3 Dipole Mounted Over a Cylinder Perpendicular to a Cylinder Gain (dB) Height (λ) Cylinder Gain Peak Peak Angle Over Cylinder Diameter (λ) At 0◦ At 180◦ Perpendicular to Plane Perpendicular to Plane 0.25

0.4

0.5

0.75

0.25 0.50 1.0 2.0 0.25 0.50 1.0 2.0 0.25 0.50 1.0 2.0 0.25 0.50 1.0 2.0

3.8 6.6 7.1 7.0 3.0 2.9 0.6 1.2 0.0 −4.2 −5.6 −6.7 3.5 5.3 5.7 6.4

−2.6 −4.4 −8.2 −8.6 −0.2 −2.5 −7.1 −7.3 0.5 −2.8 −5.7 −6.3 −0.4 −1.6 −4.3 −5.7

3.8 6.6 7.1 7.4 4.9 6.5 6.4 7.3 5.1 5.2 6.6 7.3 2.7 4.3 5.1 5.8

0 0 0 30 46 48 50 54 54 54 56 60 66 66 66 70

CROSSED DIPOLES FOR CIRCULAR POLARIZATION

231

To complete the analysis, the dipole was rotated so that its axis is perpendicular to the pole. Figure 5-11 illustrates the patterns calculated for a horizontally polarized dipole mounted above a vertical pole. We expect a pattern null at 90◦ in this horizontal plane due to the dipole polarization null, but the dipole induces curved currents on the cylinder that radiate and ﬁll in these nulls. The null due to the dipole does narrow the pattern in the horizontal plane compared to Figure 5-10, and in many cases peak radiation occurs in the vertical plane. Table 5-3 lists the characteristics of the horizontal dipole mounted over a vertical pole for various dipole spacing above the pole and its diameter.

5-6 CROSSED DIPOLES FOR CIRCULAR POLARIZATION We produce a circularly polarized antenna by placing two dipoles along the x- and y-axes over a ground plane and feeding them with equal amplitudes and quadrature phase (0◦ and −90◦ for RHC). Without the ground plane the combination radiates LHC in the −z direction. The ground plane changes the sense of circular polarization of the wave radiated in the −z-direction and it adds with the direct radiated wave. The dipoles are fed from either dual folded baluns that produce two separate inputs or by a split coax balun connecting both dipoles in shunt. The shunt connection requires differing lengths for the dipoles to produce the 90◦ phase difference that we call the turnstile conﬁguration. The dual-feed antenna uses either a quadrature hybrid equal-amplitude power divider to feed the two ports or an equal phase and amplitude power divider with an extra line length on one of the two ports. The hybrid power divider feed produces an antenna with a wide impedance and axial ratio bandwidth. The hybrid power divider has two inputs that provide ports for both RHC and LHC polarizations. The signals reﬂected from the two equal-length dipoles when fed from one port of the hybrid reﬂect into the second port due to the phasing in the hybrid coupler. When measuring at one port of the hybrid, the impedance bandwidth is quite broad because the reﬂected power is dissipated in the load on the other port. This dissipated power lowers the efﬁciency of the antenna, a hidden loss unless you measure the coupling between the inputs of the hybrid. The second conﬁguration, using the extra line length, produces an antenna with a narrowed axial ratio bandwidth and a wider impedance bandwidth compared to a single dipole. The extra 180◦ round-trip total signal path in one arm causes the equal reﬂections to cancel. Figure 5-12 gives the circularly polarized pattern from a pair of crossed dipoles over a ground plane with a perfect feed. The E-plane dipole null limits the angular range of good circular polarization. We improve the circular polarization by raising the dipoles a little and tilting them down to widen the E-plane beamwidth. Figure 5-12 shows the pattern for the tilted dipole pair and illustrates the improved cross polarization and the wider beamwidth. The placement on a ﬁnite ground plane complicates this result somewhat and will require extra design effort. Turnstile feeding exploits the impedance properties of the dipole to shift the relative phase between two different dipoles when shunt connected to the same port. When we shorten a dipole below resonance, its impedance is capacitive and its current has positive phase relative to the resonant-length dipole, while the lengthened dipole has an inductive reactance and a negatively phased current. We determine the lengths of

232

DIPOLES, SLOTS, AND LOOPS

FIGURE 5-12 Crossed dipoles fed for circular polarization: (a) λ/4 height and 0◦ tilt; (b) 0.3λ height and 30◦ tilt.

the two dipoles by a perturbation technique using the Q of the resonant circuit of the dipole. Q is related to the VSWR bandwidth: BW =

VSWR − 1 √ Q VSWR

Q=

VSWR − 1 √ BW VSWR

(5-10)

We derive the lengths of the two dipoles in terms of the resonant (zero reactance)-length dipole, L0 : L0 Lx = √ 1 + 1/Q

Ly = L0 1 +

1 Q

RHC polarization

(5-11)

A dipole of 0.014λ diameter located 0.3λ above a ground plane and tilted down 30◦ has a resonant length of 0.449λ. The 2 : 1 VSWR bandwidth for 70 is 18.3% or a Q of 3.863 by using Eq. (5-10). When we insert this Q in Eq. (5-11), we calculate the two lengths for a turnstile design: Lx = 0.400λ and Ly = 0.504λ for RHC polarization. The +x and +y poles are fed from the same port. Figure 5-13 plots the Smith chart of this design. The trace on a Smith chart rotates clockwise for increasing frequency. The cusp in the trace is the frequency with the best axial ratio, which did not occur at the frequency of best match. Nevertheless, the 2 : 1 VSWR bandwidth of the antenna has increased to 41.5% because the combined reactance of the two dipoles cancels over a large frequency range. At center frequency the pattern is similar to Figure 5-12 except that the patterns in the two planes have slightly different beamwidths due to the dipole lengths. When the frequency shifts off center, the axial ratio degrades. The axial ratio bandwidth is far less than the impedance bandwidth, and the design gives a 16.4% 6-dB axial ratio bandwidth. An axial ratio of 6 dB produces 0.5-dB polarization loss similar to the 0.5 dB reﬂected power loss of 2 : 1 VSWR. This illustrates the importance of considering not only the impedance bandwidth but also the pattern characteristics over the frequency band. We can increase the beamwidth of the turnstile dipole located over a ground plane by adding a notched cone under it. Figure 5-14 illustrates the arrangement of the slightly

CROSSED DIPOLES FOR CIRCULAR POLARIZATION

233

Finish

Best Axial Ratio

Start

FIGURE 5-13 Smith chart response of a turnstile dipole pair Lx = 0.400λ and Ly = 0.504λ mounted 0.30λ over a ground plane with 30◦ tilt.

FIGURE 5-14 Turnstile dipole mounted over a notched cone on a ﬁnite circular ground plane with radial line chokes to reduce the backlobe.

less than λ/4-long notches in a 45◦ cone with the turnstile dipoles located about λ/4 above the ground plane. A split-tube coaxial balun feeds the two dipoles sized as a turnstile with dipoles of longer and shorter length. The upper feed jumper excites RHC radiation. The dipoles excite magnetic currents in the slots that radiate a broad pattern to ﬁll in the E-plane nulls of the dipoles. On an inﬁnite ground plane the horizontal

234

DIPOLES, SLOTS, AND LOOPS

FIGURE 5-15 Pattern of a turnstile dipole mounted over a notched cone with a 0.75λ ground plane and two radial chokes.

polarization must vanish along the ground plane, and the RHC and LHC components would be equal at 90◦ similar to the pattern shown in Figure 5-12. By using a ﬁnitesize ground plane, the horizontal component does not vanish, and a wide beamwidth is obtained with circular polarization at 90◦ as shown in Figure 5-15, which uses a 0.75λdiameter ground plane and 0.5λ-base-diameter cone. To reduce the backlobe below the ground plane, two short-circuited radial transmission line chokes were placed around the edge to form a soft surface. We size the inner radius so that the transmission line produces an open-circuit impedance at the outer rim that reduces the edge diffraction and the backlobe [6, p. 88]. From a PO perspective the radial line choke is a slot that supports a magnetic current loop. This example illustrates that slots or notches can be used to shape the patterns of small antennas.

5-7 SUPER TURNSTILE OR BATWING ANTENNA [7] The super turnstile or batwing antenna was developed for TV transmitter antennas. The antenna combines a slot with a dipole batwing to produce an antenna with a wide impedance bandwidth. Figure 5-16 shows the normal conﬁguration, with four wings placed around a central support metal mast. Each wing connects to the mast at the top and bottom with a metal-to-metal connection. The inner vertical rod and the support

SUPER TURNSTILE OR BATWING ANTENNA

235

FIGURE 5-16 Super turnstile or batwing antenna using an open rod construction.

V-Pol. H-Pol.

FIGURE 5-17 Elevation pattern of a single bay of a super turnstile antenna showing horizontal and vertical polarization components.

236

DIPOLES, SLOTS, AND LOOPS

Reflection, dB

mast form a two-line slot fed by a jumper located at the center of each wing. To produce an omnidirectional pattern about the mast, a feed power divider located inside the mast phases the inputs for circular polarization (0◦ , 90◦ , 180◦ , 270◦ ). The antenna radiates horizontal polarization in the horizontal plane but radiates cross-polarization that increases with elevation (depression) angle as shown in Figure 5-17. A four-wing antenna produces a horizontal plane pattern ripple of about 1.5 dB. Adding more wing antennas around a larger central mast reduces the ripple. The extraordinary characteristic of the antenna is its impedance bandwidth. Figure 5-18 gives the return loss frequency response for a wire frame antenna. The 1.1 : 1 VSWR bandwidth is about 35%; if adjusted to 1.25 : 1 VSWR, the antenna has a 51% bandwidth. You make small adjustments to the spacing between the mast and the inner rod to tune the VSWR. Table 5-4 lists the parameters of batwing antennas with both wire frame and solid panel wings. The solid panels lower the input impedance to 75 from the 100 of the wire frame antenna. Using an antenna with only two wings

Adjusted for 1.25:1 VSWR

Adjusted for 1.1:1 VSWR

Normalized Frequency

FIGURE 5-18 Super turnstile wire frame antenna return-loss response adjusted for 1.1 : 1 and 1.25 : 1 VSWR. TABLE 5-4 Dimensions of a Super Turnstile Antenna in Wavelengths for Four Wings Center-Fed for Circular Polarization Parameter

Wire Frame

Impedance () Height Wing upper Wing middle Gap Rod diameter Mast diameter

100 0.637 0.2254 0.0830 0.0169 0.0508 0.0847

Solid Wing 75 0.637 0.229 0.0847 0.0216 0.0847

CORNER REFLECTOR

237

changes the input impedance from the value for an antenna with four wings because the close coupling between the wings alters the impedance. It depends on the feeding mode. This holds for any antenna with close coupling: for example, a spiral antenna. You must feed it in the operating mode to measure the correct input impedance. The transmitter antenna will consist of a number of these antennas stacked vertically to produce a narrow pattern directed at the horizon. 5-8 CORNER REFLECTOR [8, p. 328] The usual corner reﬂector (Figure 5-19) has a dipole located between two ﬂat plates that limit directions of radiation. The angle between the reﬂectors can be any value, but 90◦ seems to be the most effective. On paper, decreased angles give better results, but only marginally. We could consider the ﬂat plate as a limiting case. The tangential electric ﬁelds must vanish at the surface of the ﬂat plates. We discover a greater restriction, since the ﬁelds can only decrease gradually in the limited space between the ground planes and the dipole. Most of the power is concentrated in lower-order spherical modes. In the limit of zero vertex distance, the single mode possible restricts the beamwidth to 45◦ in the H -plane. We analyze the 90◦ corner reﬂector as an array by using the three images of the dipole in the ground planes (Figure 5-19) plus the real dipole. The array factor of the array of dipole and images is 2 j kd cos θ e + e−j kd cos θ − (ej kd sin θ sin φ + e−j kd sin θ sin φ )

Length

d d d Height

Images

FIGURE 5-19 A 90◦ corner reﬂector.

238

DIPOLES, SLOTS, AND LOOPS

In the H -plane, φ = 90◦ and we evaluate terms to get 4[cos(kd cos θ ) − cos(kd sin θ )]2 where d is the distance from the vertex to the dipole and θ is the H -plane pattern angle from the axis. We must multiply this by the pattern of the dipole to obtain the radiation intensity. We consider only the H -plane, where the maximum radiation intensity is found from Eq. (5-9): U = 4[cos(kd cos θ ) − cos(kd sin θ )]2

η|I0 |2 (2π)2

1 − cos

kL 2

2 (5-12)

where η is the impedance of free space, I0 the dipole current, and L the dipole length. The radiated power of the single dipole is √ (5-13) Pin = |I0 |2 [R11 + R12 (2d) − 2R12 ( 2 d)] where R11 is the self-resistance of the dipole and R12 (x) is the mutual resistance function between the dipole and its images. The directivity is found from directivity(θ ) =

4πU (θ ) Pin

(5-14)

We combine Eqs. (5-12) and (5-13) into Eq. (5-14) to compute directivity of the 90◦ corner reﬂector with inﬁnite sides: directivity(θ ) =

4η[1 − cos(kL/2)]2 [cos(kd cos θ ) − cos(kd sin θ )]2 √ R11 + R12 (2d) − 2R12 ( 2 d)

(5-15)

Table 5-5 gives the directivity, beamwidth, and impedance of a 90◦ corner reﬂector fed from a dipole 0.42λ long and 0.02λ in diameter. We must shorten the dipole further than a free-space dipole length at resonance to compensate for the mutual coupling between dipoles. Directivity increases as the vertex distance decreases, but the effects of superdirectivity cause the efﬁciency and gain to fall as the vertex is approached. The antenna has a 50- input impedance for d = 0.37λ. This point shifts when we increase the dipole’s diameter to increase its bandwidth. Kraus gives the following guidelines for the size of the sides. Each plate should be at least twice the length of the dipole-to-vertex distance, and the plate height (the dipole direction) should be at least 0.6λ. To evaluate those guidelines, a GTD analysis was performed on various combinations (Table 5-6) with d = 0.37λ. The H -plane beamwidth decreases with an increase in plate length. After about 1.5λ sides, the H -plane beamwidth ﬂuctuates about 45◦ as the sides increase. Even with 5λ sides the beamwidth is below 45◦ . The E-plane beamwidth ﬂuctuates with the plate height. The directivity was estimated from the beamwidths. In one case—1.5λ sides and 1.5λ high—the estimated directivity exceeds the directivity of the inﬁnite-side case. The edge diffractions add to the reﬂected and direct radiation of the rest of the antenna. Refer to Section 2-4.2 for an example using PO to analyze a corner reﬂector. Similar to inﬁnite plate analysis, the reaction of the image dipoles in the ﬁnite plates can be used to ﬁnd the input impedance and gain of the antenna. When we analyze the corner

CORNER REFLECTOR

239

TABLE 5-5 Characteristics of a 90◦ Corner Reﬂector with Inﬁnite Sides and 0.42λ Dipole Vertex Directivity Beamwidth Input Impedance Distance (λ) (dB) (deg) () 0.30 0.32 0.34 0.36 0.37 0.38 0.40 0.42 0.44 0.46 0.48 0.50 0.52 0.54 0.56 0.58 0.60

12.0 12.0 11.9 11.9 11.9 11.8 11.8 11.7 11.7 11.6 11.5 11.4 11.4 11.3 11.2 11.1 10.9

44.7 44.6 44.5 44.3 44.2 44.1 43.9 43.6 43.3 42.9 42.4 41.8 41.1 40.2 39.2 38.1 36.8

29.1 − j 1.1 34.9 + j 0.4 40.9 + j 1.1 47.0 + j 0.8 50.0 + j 0.3 53.0 − j 0.5 58.8 − j 2.8 64.1 − j 6.0 68.8 − j 10.0 72.7 − j 14.9 75.7 − j 20.3 77.7 − j 26.2 78.6 − j 32.2 78.4 − j 38.4 77.0 − j 44.3 74.6 − j 49.8 71.3 − j 54.8

TABLE 5-6 Results of a GTD Analysis of a 90◦ Corner Reﬂector with Finite Sides and Vertex Distance 0.37λ Beamwidth

Side Length (λ)

Plate Height (λ)

E-Plane

H -Plane

F/B (dB)

Estimated Directivity (dB)

0.75 1.00 1.50 0.75 1.00 1.50 0.75 1.00 1.50 5.00

0.75 0.75 0.75 1.00 1.00 1.00 1.50 1.50 1.50 5.00

70.4 73.6 72.6 60.2 61.0 58.5 53.4 51.6 48.2 68.8

97.4 72.4 50.8 91.6 62.8 46.0 81.6 60.0 42.6 43.4

18.4 17.3 18.2 23.4 22.7 23.8 34.0 39.0 46.3 63.5

7.7 8.8 10.0 8.5 10.1 11.4 9.3 11.0 12.6 10.8

reﬂector using GTD, the method does not determine input impedance and gain must be estimated from the patterns. We can use the method of moments to analyze the corner reﬂector. One preferred construction method is to use rods for the reﬂector so that the antenna has minimum wind loading. Figure 5-20 illustrates a corner reﬂector made with only six rods on each side. Figure 5-21 gives the pattern of this antenna from a moment method calculation. This small antenna produces excellent results. We can use the angle of the sides as a design parameter. A geometric optics analysis that uses images restricts the angle, but nothing stops the antenna from working for

240

DIPOLES, SLOTS, AND LOOPS

FIGURE 5-20 Corner reﬂector constructed from 0.6λ-long rods spaced 1/6λ with a dipole spaced 0.37λ from the vertex.

E-Plane H-Plane

FIGURE 5-21 Pattern of a corner reﬂector made from 0.6λ-long rods spaced 1/6λ with dipole 0.37λ from vertex.

241

CORNER REFLECTOR

TABLE 5-7 Corner Reﬂector with Varying Angle H -Plane 0.9λ-Wide Plates Connected to a 0.2λ Central Plate, 1λ E -Plane Width, Dipole 0.3λ Above the Central Plate Beamwidth

Side Angle

E-Plane

60 55 50 45 40 35 30 25 20

58.1 57.1 56.3 55.8 55.4 55.4 55.6 56.2 57.4

Beamwidth

H -Plane

Gain (dB)

F/B (dB)

Side Angle

E-Plane

H -Plane

Gain (dB)

F/B (dB)

59.4 56.0 52.3 49.1 46.8 45.5 45.6 47.9 53.7

9.1 9.9 10.5 10.8 11.1 11.2 11.0 10.7 10.2

21.7 22.6 23.0 23.4 22.8 24.3 24.9 25.5 20.3

15 10 5 0 −5 −10 −15 −20 −25

59.2 61.6 64.7 67.8 70.2 71.5 71.8 71.7 71.8

65.9 83.3 99.2 108.6 117.0 125.8 135 143.8 152.2

9.5 8.7 7.8 7.5 7.2 6.9 6.6 6.3 5.4

19.3 26.3 25.1 23.6 22.0 16.5 19.0 17.8 16.1

Dipole 0.9l Side Plate Hinge 0.2l Center Plate

FIGURE 5-22

Corner reﬂector with variable-angle side plates and a center ﬂat plate.

arbitrary side angles. It is convenient to have a small plate between the tilted sides for the mounting brackets, and these side plates could be mounted on hinges and rotated to vary the H -plane beamwidth. Table 5-7 lists the parameters of a corner reﬂector 1λ along the E-plane, a central plate 0.2λ wide in the H -plane, and sides 0.9λ long where the side angle is varied. The dipole is located 0.3λ above the central plate. We measure the side plate angle from the plane containing the small central ground plane; zero corresponds to a ﬂat plane ground plane and 45◦ the usual corner reﬂector. Negative side-plate angle means that the side plates are tilted behind the central plate away from the dipole. Figure 5-22 illustrates the H -plane cross section of this corner reﬂector with 30◦ side plates. We should not design corner reﬂectors with large sides since the gain is limited. The gain of paraboloid reﬂectors of the same size soon exceeds that of a corner reﬂector. A 2λ-diameter paraboloid reﬂector at 50% efﬁciency has a gain of 13 dB, and its gain exceeds that of a corner reﬂector. Any corner reﬂector with a vertex angle given by 180◦ /N , where N is an integer, can be analyzed by the method of images. Corner reﬂectors with N greater than 2 have only marginally higher gains. The 90◦ corner reﬂector gives the best result for a given amount of material. Elkamchouchi [9] adds a cylindrical surface between the plates centered on the vertex. This surface adds another set of images within the cylinder. The images increase the gain by about 2 dB and decrease the frequency dependence of the impedance.

242

DIPOLES, SLOTS, AND LOOPS

10l dia. Disk

2l dia.

1l dia.

FIGURE 5-23 A λ/4 monopole located on 1λ-, 2λ-, and 10λ-diameter disk ground planes.

5-9 MONOPOLE A monopole consists of a single conductor fed out of a ground plane from the center conductor of a coax. When we include its image (Figure 5-6), the monopole equates to a dipole for analysis. The ﬁelds vanish below the ground plane and restricting the ﬁelds to the upper hemisphere doubles the gain over a dipole, since only half the input power of the dipole is needed to produce the same ﬁeld strength. The input impedance decreases to half that of the equivalent dipole. We can form the image of the voltage source feeding the monopole in the ground plane. The voltage across the input of the equivalent dipole is twice that of the monopole to produce the same current. Therefore, the impedance of the monopole is half the impedance of the dipole. The large value of edge diffraction greatly limits the F/B ratio of a monopole when it is placed on a ﬁnite ground plane. Figure 5-23 shows the pattern of a monopole when placed on 1λ-, 2λ-, and 10λ-diameter circular ground planes. The back radiation can be reduced by placing the monopole over a ground plane with circular corrugations that forms a soft surface at the edge when the corrugations are slightly deeper than λ/4 [10]. When the corrugations are less than λ/4, the ground plane can support surface waves. 5-10

SLEEVE ANTENNA [8, p. 422; 11, Chap. 5; 12; 13, p. 278]

A sleeve around the monopole (Figure 5-24) moves the virtual antenna feed up the monopole. The bandwidth increases because the current at the feed point remains nearly

SLEEVE ANTENNA

Virtual Feed Point

l/4

243

l/2

I Sleeve

I

I

Image Feed Point

Images

Coax

FIGURE 5-24 Sleeve monopole and current distributions.

constant over a wide band. Currents at the input for the case when the monopole is a quarter-wavelength long and when it is a half-wavelength long are about the same (Figure 5-24). The input resistance remains constant as the frequency changes. The sleeve shields possible radiation from the internal currents while the currents on the outside of the sleeve radiate. The pattern changes little from that of an unshielded monopole. The internal structure is available as a series-matching stub and a transformer to broadband the antenna. Design consists of adjusting the parts until a suitable compromise input impedance match is achieved over the band. Dipole sleeve antennas (Figure 5-25) require symmetrical sleeves on the arms to maintain the symmetry of the currents. It is equivalent to feeding the antenna in two places. The balun is made an integral part of the base. In both antennas, strips or rods can replace the total coaxial sleeve [14]. The currents on the rods cancel the radiation from the currents on the internal feeder. Figure 5-26 illustrates an open-sleeve dipole using two rods designed to be mounted over a ground plane. The antenna is fed from a folded balun that consists of a grounded vertical coax with one pole connected to the outer shield and a matching tube connected to the second pole. The center conductor jumps across the gap to the second pole. Following are the design dimensions in wavelengths normalized to the lower-frequency band edge: Dipole length Sleeve length Dipole-to-sleeve spacing Input taper

0.385 0.2164 0.0381 0.056

Dipole diameter Sleeve diameter Dipole height above ground

0.0214 0.0214 0.1644

Figure 5-27 plots the return-loss response of the antenna for various conﬁgurations and models of the antenna. The dipole without the sleeves has its best return loss over a narrow band centered at a normalized frequency of 1.05. The sleeves have little effect on this response at the low-frequency end. Adding sleeves produces a second resonance, which combines with the lower one to produce a broad bandwidth. An initial method of moments analysis used constant-diameter rods for the antenna, and Figure 5-27 shows the poor impedance match response of the antenna. A key element of the experimental

244

DIPOLES, SLOTS, AND LOOPS

Sleeve Dipole with Folded Balun

Sleeve Dipole

Type II Sleeve Balun

FIGURE 5-25 Sleeve dipoles.

FIGURE 5-26 Open-sleeve dipole with conical input taper.

(a) Single Dipole

Reflection, dB

(b) With Sleeves, No Taper (d) Large Ground (e) 1l Ground Plane

(c) Free Space with Sleeves

Normalized Frequency

FIGURE 5-27 Return-loss response of an open sleeve dipole: (a) dipole without sleeves; (b) open sleeve antenna; (c) open sleeve antenna with tapered input; (d) open sleeve antenna with tapered input located λ/4 over ground plane; (e) open sleeve antenna over a 1λ-diameter ground plane.

CAVITY-MOUNTED DIPOLE ANTENNA

245

antenna is the tapered input. Adding this feature to the model produced the improved broad response of the experimental antenna. The constant-diameter model response shows a notable capacitive term on a Smith chart, and the tapered input produced the necessary inductance to reduce this effect. If the antenna is located in free space, the impedance response improves as shown. Figure 5-27 points out the importance of analyzing an antenna in its operating environment. The dot-dashed curve illustrates the response when the antenna was mounted over a one-wavelength-square ground plane. The ﬁnite ground plane produces a small but noticeable change in the input impedance. The effects of small changes in the analytical model warn us that we cannot expect antennas to match their models exactly and that small mechanical details can be used to improve performance. An open-sleeve antenna can be made using a wire cage. Since the diameters of the dipole and sleeve rods are large, the weight can be reduced by using a circular array of wires for each conductor. The effective diameter of the cage, deff , is given as deff = d

nd0 d

1/n or

1 d0 = d n

deff d

n

(5-16)

The diameter of the individual wires is d0 , the cage diameter is d, and n is the number of wires.

5-11 CAVITY-MOUNTED DIPOLE ANTENNA A dipole can be placed in a cup, and the assembly can be ﬂush-mounted in a ground plane. The antenna shown in Figure 5-28 has disk sleeves located above and below the dipoles to stretch the bandwidth over a 1.8 : 1 range [15]. Following are the dimensions normalized to the dipole length: D = 2.57 L T = 0.68 L

H = 0.070 L G = 0.40 L

S = 0.505 L

The operating range is 0.416λ ≤ L ≤ 0.74λ. The antenna cavity ranged from 0.28λ to 0.50λ deep and can no longer be considered thin. The cup antenna has a nearly constant gain (±0.5 dB) of 10.5 dB over the band. Mounting the antenna in a cavity opens up new possibilities, because extra parameters are added to the design. At the low-frequency end, the cavity diameter is 1.07λ, which grows to 1.90λ at the high end. We can use a dipole in a cup as a reﬂector feed. Excellent pattern and impedance response is obtained with the dipole mounted in a truncated cone cup with a 0.88λ aperture diameter, a 0.57λ-diameter base, and a 0.44λ depth [16, pp. 106–108]. The dipole is foreshortened to 0.418λ for an element diameter of 0.013λ and mounted 0.217λ above the base to achieve a 21% 2 : 1 VSWR bandwidth for a single element. When we use a cross-polarized pair fed from a hybrid coupler to radiate CP, the impedance match at the input port improves. The signals reﬂected from the two dipoles add in phase at the isolated port and cancel at the input port. The load dissipates the reﬂected power, and the antenna through the hybrid presents an excellent impedance match.

246

DIPOLES, SLOTS, AND LOOPS

FIGURE 5-28 Cavity-mounted sleeve dipole antenna.

RHC Pol.

LHC Pol.

FIGURE 5-29 Circular polarization response of a crossed dipole mounted 0.217λ above the bottom of a truncated cone 0.44λ deep with a 0.88λ aperture and a 0.57λ base.

FOLDED DIPOLE

247

TABLE 5-8 Illumination Losses When Pattern of Figure 5-29 Feeds a Paraboloidal Reﬂector Loss (dB) f/D

Average

Maximum

0.36 0.38 0.40 0.42 0.44 0.46 0.48 0.50 0.52

1.69 1.60 1.54 1.50 1.49 1.50 1.52 1.55 1.60

1.74 1.66 1.65 1.65 1.68 1.72 1.77 1.83 1.91

Figure 5-29 plots its pattern when excited for CP. The cross-polarization is about 30 dB below the peak co-polarization response over the entire 10-dB beamwidth cone. It has the following illumination losses when the antenna is used as a paraboloidal reﬂector feed (see Section 8-2); for f/D = 0.44 and averaged over the 21% bandwidth: spillover loss = 0.72 dB

amplitude taper loss = 0.65 dB

cross-polarization loss = 0.12 dB Table 5-8 demonstrates the broad optimum reﬂector f/D for a phase center 0.02λ inside the aperture plane, where we position it at the reﬂector focus.

5-12 FOLDED DIPOLE A half-wavelength folded dipole increases the input impedance of a normal dipole fourfold while radiating the pattern of a single dipole. With the two elements closely coupled, we analyze the antenna using even and odd modes (Figure 5-30). The even mode divides the antenna into separate dipoles because the magnetic wall halfway between them is a virtual open circuit. The input current to the even mode becomes Ie =

V 2(Z11 + Z12 ) l/2

+

V/2

−

Magnetic

+ Wall

+

− V/2

=

V/2

−

Electric

+ Wall

−

+ V/2

FIGURE 5-30 Folded dipole analysis modes.

V

−

248

DIPOLES, SLOTS, AND LOOPS

where Z11 is the self-impedance of one of the dipoles and Z12 is the mutual impedance between the closely coupled dipoles. The odd mode reduces the antenna to the series connection of two nonradiating λ/4 stubs: Io =

V j Z0 tan(kL/2)

where Z0 is the characteristic impedance between the two rods. The input current is the sum of the even- and odd-mode currents. Near L = λ/2, the odd-mode current is quite small because its input impedance is an open circuit, and the input impedance is then determined by the even mode only: Zin =

V = 2(Z11 + Z12 ) Ie

For closely coupled lines, Z11 = Z12 and the input impedance becomes Zin = 4Z11 , where Z11 is the self-impedance of the dipole. Higher input impedance levels can be obtained by adding more elements. A second method of altering the step ratio from 4 is to use unequal feed and shorted element diameters [17,18]. Given a driven element radius a1 , parasitic element radius a2 , and center-to-center spacing b, Hansen [18] gives a convenient formula for the step-up ratio (1 + γ 2 ): γ =

5-13

cosh−1 [(v 2 − u2 + 1)/2v] cosh−1 [(v 2 + u2 − 1)/2uv]

where u = a2 /a1 and v = b/a1

(5-17)

SHUNT FEEDING [19, p. 118]

Shunt feeding grows out of the folded dipole. The T-match (Figure 5-31) starts as a folded dipole when the taps are at the ends. As the taps move toward the center, the impedance of the dipole dominates at ﬁrst, since the admittance of the shunt stub in the odd mode is small and the input impedance is capacitive. At some point, as the taps move toward the center, the inductive admittance of the stub will cancel the capacitive admittance of the dipole and produce antiresonance with its high input resistance. The location and magnitude of this peak resistance depends on the diameters of the rods in the T-match section and the diameter of the radiator. The input resistance decreases as we continue to move the tap point toward the center after the feed location passes the antiresonance point. The input impedance is inductive and match is achieved by using symmetrical series capacitors. The T-match is fed from a balanced line. The center short on the dipole allows the direct connection of the dipole to ground. Direct connection of broadcast towers (monopoles) to ground gives some lightning protection because the transmitter is capacitively connected to the tower. Shunt feeding with a T-match enables solid conductors, such as the skin of an aircraft, to be excited as a dipole. Horizontal shunt-fed dipoles can be connected directly to vertical towers with a metal-to-metal connection to increase the strength of the antenna to withstand adverse weather conditions. A gamma match (Figure 5-31) can be fed from an unbalanced coax line. The shield of the coax connects to the shorted center of the dipole while the center conductor

DISCONE ANTENNA

Tee

Match

249

Gamma Match

Balanced Line Coax

FIGURE 5-31 Shunt-fed dipoles.

taps into one side of the solid rod. Moving the tap away from the center increases the input resistance. The inductive reactance is series-tuned with a capacitor. Both of these connections reduce the bandwidth of the antenna as the input impedance is raised because the combination of the series capacitor and the shunt inductive stub increases the stored energy and Q of the antenna. 5-14 DISCONE ANTENNA The discone antenna (Figure 5-32) is a modiﬁcation of the dipole where the upper pole becomes a disk and the lower pole turns into a cone. We feed the antenna by locating a coax in the center of the cone and by connecting its outer shield to the lower cone at its top while we extend the coax center conductor and connect it to the disk. We obtain an antenna with a wide impedance bandwidth and a dipolelike pattern. As frequency increases the pattern peak moves toward the cone and gives a downward-pointing pattern. Figure 5-33 shows the pattern of a discone antenna at the design frequency and at two, three, and four times this frequency. The antenna produces less useful patterns as frequency increases. The antenna that gives the patterns in Figure 5-33 has a VSWR less than 3 : 1 from 1 to 10 times the design frequency. The cone upper diameter determines the high-frequency end of good impedance match. Typical slant length dimensions versus cone angle are as follows [20, pp. 128–130]: Total Cone Angle

25

35

60

70

90

Slant Length (λ)

0.318

0.290

0.285

0.305

0.335

The upper disk diameter equals 0.7 times the lower cone diameter. The spacing between the top of the cone and the upper disk equals 0.3 times the diameter of the upper cone. The diameter of the upper cone determines the upper frequency limit, but practice shows that the antenna patterns are good only over a 4 : 1 to 4.5 : 1 frequency range. The impedance bandwidth is much wider than the pattern bandwidth. To reduce weight and wind loading, the cone and disk can be made from rods, with a typical implementation having at least eight.

250

DIPOLES, SLOTS, AND LOOPS

FIGURE 5-32 Discone antenna with coaxial feed with a center conductor connected to the upper disk and a shield connected to the lower cone.

Freq. = 2

Freq. = 4 Freq. = 3

Freq. = 1

FIGURE 5-33 Elevation pattern of a 60◦ discone antenna at normalized frequencies = 1, 2, 3, and 4.

BALUNS

251

5-15 BALUNS [21; 22, pp. 167–180] A balun properly connects a balanced transmission line to an unbalanced transmission line. Simple arguments about impedances to the balanced and unbalanced modes of the three-wire transmission lines explain its operation. Considering one of the lines of a transmission line as ground misleads us. A ground plane under the transmission-line feeder becomes the third conductor of a three-wire line. Currents ﬂowing in the ground plane can unbalance the currents in the feeder. A balanced three-wire transmissionline mode carries equal and opposite currents in the feeder lines. The capacitances per unit length of the two lines to ground are the same. Coax is an example of an unbalanced line structure (Figure 5-34). The inner conductor has no direct capacitance to ground. The two-wire line shown in Figure 5-34 is a balanced line having equal capacitances to ground, but we must judge a balanced line by the currents, not just the physical structure. Before we analyze baluns, we must consider the fundamental modes of a threewire transmission line. Figure 5-35 shows circuit representations of the modes without showing the ground conductor. Equal loads terminate ports 3 and 4. The even mode applies equal voltages on ports 1 and 2 and forms a magnetic wall between the conductors where the magnetic ﬁeld vanishes to produce a virtual open circuit. The unbalanced mode—equal current directions—is associated with the even mode. Equal and opposite voltages on ports 1 and 2 form the odd mode and set up an electric wall between the conductors. The electric wall is a virtual short circuit. The odd mode excites equal and opposite currents—balanced mode—on the two lines. When the loads on ports 3 and 4 are unequal, the modes separate according to the voltages, even and odd, or the currents, unbalanced and balanced. Dipoles present loads between the lines and not to ground. Balanced line

Unbalanced line

Ground

FIGURE 5-34

+V

−V

Ground

Physically balanced and unbalanced transmission lines.

l 1

l Electrical Wall

2 l

4 3

Balanced Mode (Odd Mode)

V

V

1

Magnetic Wall

2 l

4 3

Unbalanced Mode (Even Mode)

FIGURE 5-35 Balanced and unbalanced modes on a three-wire transmission line.

252

DIPOLES, SLOTS, AND LOOPS

Unbalanced mode circuits radiate. Only closely spaced equal and opposite currents, the balanced mode, cancel the far-ﬁeld radiation from the currents on the feed lines. The radiating feeder line adds radiation components to the antenna. These components can radiate unwanted polarizations and redirect the beam peak of the antenna (squint). In reception, the unwanted currents excited on the feeder by passing electromagnetic waves reach the receiver terminals without a balun to block them. We analyze baluns by using either transmitting or receiving antennas, depending on convenience, because reciprocity applies to baluns as well as antennas. We detect balance problems from pattern squint and cross polarization. An impedance-measuring setup can detect some balance problems. Radiating unbalanced currents cause changes in the impedance. The radiation shows when the impedance changes as ﬁngers are run over the coax line from the equipment. If we feed a dipole from a coax without a balun, the current on the outer conductor splits between the dipole conductor and the outside of the conductor. Patterns and impedance measurements detect this current. Unbalanced currents on the arms of the dipole and feeder currents cause pattern squint, but the cross-polarization radiated is usually a greater concern. 5-15.1 Folded Balun A folded balun (Figure 5-36) allows the direct connection of a coax line to the dipole. A dummy coax outer conductor is connected to the pole fed from the center conductor. It runs alongside the feeder coax for λ/4 and connects to ground. The other pole connects directly to the shield of the feeder coax. The outer conductor of the coax and the extra line are two lines in a three-wire line with ground. We analyze the structure by using balanced (odd) and unbalanced (even) modes. Unbalanced-mode excitation

Balanced Output to Dipole Arms

l/4

Unbalanced Coax

FIGURE 5-36 Folded balun.

BALUNS

Ys

Ya

253

Antenna Ys = −jY0 cot kl

Coax

Quarter Wavelength Line with a Short Circuit Termination

FIGURE 5-37

Folded balun equivalent circuit (balanced mode).

at the dipole forms a magnetic wall through the ground connection between the two coax shields. The circuit reduces to a single line with an open circuit at the ground connection. The open circuit transforms through the quarter-wavelength line to a short circuit at the dipole. Any unbalanced currents induced on the dipole or the coax outer conductor are shorted at the input. Balanced-mode excitation at the dipole forms an electric wall through the ground connection. The balanced-mode circuit of the two coax shields is a λ/4 short-circuited stub connected in shunt with the dipole (Figure 5-37). We analyze the frequency response from Figure 5-37. The bandwidth of the balun, although narrow, exceeds the bandwidth of the dipole. The Roberts balun [23] design adds an open-circuited stub λ/4 long inside the dummy coax of the folded balun. Instead of connecting the center conductor of feeding coax to the outer shield, we connect it to the open-circuited stub. The equivalent circuit for the balanced mode includes the short-circuited stub of the folded balun plus the open-circuited stub. The two reactances shift in opposite directions as frequency changes and produce a dual resonance we see as a loop on the Smith chart plot of impedance. The frequency bandwidth increases to almost 3 : 1, a more suitable choice for wide-bandwidth antennas. 5-15.2 Sleeve or Bazooka Baluns An outer jacket shields the outer conductor of the coax feeder in a sleeve balun (Figure 5-38). The sleeve and outer conductor of the coax form a series stub between the coax feeder and ground when the cup is short circuited to the coax outer conductor. The λ/4 stub presents a high impedance to the unbalanced currents at the top of the cup (Figure 5-39). A second sleeve below the ﬁrst one and directed away from the dipole further prevents currents excited on the coax from reaching the input. When the frequency shifts, the connection to ground through the sleeve unbalances the transmission line. This balun is inherently narrowband. Adding a stub to the center conductor (Figure 5-40) increases the bandwidth because the stubs track each other when the frequency changes. Figure 5-39 demonstrates the circuit diagrams of the two types of sleeve baluns. The type II sleeve balun has matching series stubs on the outputs. The lines remain balanced at all frequencies, but the stubs limit the bandwidth of efﬁcient operation. Marchand [21] adds an open-circuited λ/4 stub inside the matching type II extra shorted stub of the sleeve balun and connects it to the coax center conductor in the same manner as the Roberts balun. The Roberts balun is a folded balun version of the Marchand compensated sleeve balun. The coaxial dipole is a variation of the sleeve or bazooka balun. We rotate the right pole in Figure 5-38 until it is vertical and remove the left pole. We turn over the

254

DIPOLES, SLOTS, AND LOOPS

l/4

FIGURE 5-38 Sleeve or bazooka balun. Coax Inner Conductor Unbalanced

Balanced line

Inside of Outer Conductor

Outside of Outer Conductor

Zs

Zs

l/4

l/4

Sleeve

Type II Only Zg

FIGURE 5-39

Schematic of types I and II sleeve or bazooka baluns. Balanced Mode

Unbalanced Mode

l/4

l/4

FIGURE 5-40 Type II sleeve balun.

sleeve and connect the short-circuit end to the outer conductor of the coax. The sleeve becomes the second pole of the dipole. The short-circuited stub at the bottom of the dipole between the outer conductor of the coax and sleeve transforms to an open-circuit impedance at the end of the lower pole. This prevents current ﬂow farther down the coax. Some references call this a sleeve dipole, which should not be confused with the

BALUNS

255

sleeve dipole used to increase the impedance bandwidth. The coaxial dipole has the inherently narrow bandwidth of the bazooka balun, but is a convenient construction. 5-15.3 Split Coax Balun [24, p. 245] A split coax balun allows the connection of both arms of a dipole to the outer shield of the coax that maintains symmetry to the dipole arms. Its rigidity helps to overcome vibration problems. Slots cut in the outer shield (Figure 5-41) enable the coax line to support two modes and make it equivalent to a three-wire line. A shorting pin excites the TE11 mode in the slotted coax (Figure 5-42) to feed the dipole in the balanced mode. Analysis of a split coax balun is similar to that of a folded balun. The ends of the slots are equivalent to the ground connection of the two coax shields of the folded balun. A virtual open circuit forms at the ends of the slots in the unbalanced (even) mode. It transforms to a short circuit at the dipole and shorts the unbalanced mode at the input. The virtual short circuit at the end of the slots in the balanced mode transforms to an open circuit at the input. Figure 5-37 gives its circuit diagram. Symmetry improves the performance of a split coax balun over a folded balun. The shorting pin is used only to excite the TE11 mode to feed the dipole arms. The extra wire length of the center conductor jumper of the folded balun introduces phase shift to the second arm and squints the beam. For that reason, the split coax balun is a better high-frequency balun. The phase shift problem of the jumper also occurs with the “inﬁnite” balun of the log-periodic antenna. Dipole Arm Shorting Pin

l/4

Slot Dipole Arm

FIGURE 5-41

TEM

Split coax balun. (From [24], Fig. 8-5, 1948 McGraw-Hill.)

TE11

Pin E=0

FIGURE 5-42 Coaxial transmission-line modes in a split coax balun. (From [24], Fig. 8-6, 1948 McGraw-Hill.)

256

DIPOLES, SLOTS, AND LOOPS

5-15.4 Half-Wavelength Balun A half-wavelength balun (Figure 5-43) works by cancellation of the unbalanced-mode currents at the input to the coax. The impedance transforms by a factor of 4 from unbalanced- to balanced-mode ports. In the unbalanced (even) mode, equal voltages are applied to the two output ports. When the voltage wave on the upper line propagates through λ/2, its phase changes by 180◦ . This signal cancels the signal connected directly to the coax center conductor. A load across a balanced-mode transmission line has a virtual short circuit halfway through it. The load on each balanced-mode line is 2Z0 , where Z0 is the coax characteristic impedance. The load on the end of the λ/2-long line is transformed by the transmission line to the identical impedance when it circles the entire Smith chart. The two loads, each 2Z0 , are connected in shunt at the coax input and combine to Z0 . A balanced-mode impedance of 4Z0 transforms to Z0 at the coax input. The λ/2long cable can be rolled up for low frequencies. The balun transforms 300- input impedances of folded dipoles to 75 by using RG-59 cable (75 ). 5-15.5 Candelabra Balun A candelabra balun (Figure 5-44) transforms the unbalanced-mode impedance fourfold to the balanced-mode port. The coax cables on the balanced-mode side connect in series, whereas those on the unbalanced-mode side connect in parallel. We can divide the balanced-mode impedance in two and connect each half to a 2Z0 impedance transmission line. These lines then connect in shunt at the unbalanced-mode port. The unbalanced-mode currents short out at the input to the 2Z0 coax lines in the same manner as does the folded balun. More lines can be stacked in series and higher-impedance transformations obtained, but construction becomes more difﬁcult. 5-15.6 Ferrite Core Baluns Ferrite cores can be used to increase the load impedance to unbalanced-mode currents and reduce them. At low frequencies (l/2

lg/2

TEM Input

Ground Planes

FIGURE 5-55 Stripline series slot.

SHALLOW-CAVITY CROSSED-SLOT ANTENNA

269

the cavity. The cavity reactance slope limits the bandwidth of the stripline-fed slot to a few percent. Increasing the impedance of the waveguide cavity transmission line reduces the reactance slope contributed by the cavity. We increase the bandwidth by using greater distances between the ground planes and thereby increase the waveguide transmission-line impedance. In general, greater volumes for an antenna increase the impedance bandwidth. Rotating the slot relative to the stripline feeding line reduces its load on the transmission line. The waveguide top wall series slot relation [Eq. (5-34)] applies in this case. The slot maintains its polarization while the nonradiating stripline center conductor approaches the slot at an angle. Rotated slots in waveguide must be paired symmetrically to reduce cross-polarization. A longitudinal array [37] can be made by placing all the slots on the centerline of a boxed stripline. Either edge plating or a series of plated-through holes forms a waveguide structure that supports only the TE10 mode. Slots placed on the centerline (as in Figure 5-59, slot c) fail to interrupt the waveguide mode currents. The stripline meanders below and varies the excitation by changing the angle between the slot and the stripline center conductor. The slight loading of each slot excites very little of the parallel-plate mode that causes unwanted slot coupling. Both traveling-wave and resonant linear arrays are possible. See Section 5-26 for a discussion of slot arrays.

5-22 SHALLOW-CAVITY CROSSED-SLOT ANTENNA We can feed the slot in Figure 5-55 by exciting the cavity in an odd mode from two points on opposite sides of the slot. To be able to excite both polarizations, we divide the slot in two and rotate the two parts in opposite directions by 45◦ to form a cross. We use a square cavity to maintain symmetry and replace the shorting pins with solid walls (Figure 5-56c). Since we feed across the diagonal between the crossed slots, we excite both slots. The sum of the ﬁelds radiated from the two slots is polarized in the direction of the diagonal. We increase the radiation conduction by lengthening the crossed slots to the maximum, which lowers the Q (increased bandwidth). The cavity compensates for the slot susceptance to obtain resonance. A crossed-slot antenna was built [38] with the following dimensions: Cavity edge Cavity depth Slot length

0.65λ 0.08λ 0.915λ

The measured 2 : 1 VSWR bandwidth was 20.8%. √ The bandwidth exceeded that of a microstrip patch of the same thickness by about 2. Lindberg [39] found that the resonant length of the slot depends on the cavity depth and requires some experimental adjustment. King and Wong [41] added ridges (Figure 5-56b) to increase the bandwidth. Antennas with ridges need a larger cavity width and a longer slot than the unridged design. The ridges can be stepped as shown to increase the bandwidth. Adding ridges gives us extra parameters to adjust for best input match performance. The following design with uniform ridges produces a 58.7% 2.5 : 1 VSWR bandwidth with a double resonance curve.

270

DIPOLES, SLOTS, AND LOOPS Feed Probes (4)

16

16 (a) b d

5.5

h1 h2

(b)

Feed Probes (4) (0.35 in from Corner) 180° Hybrid l1

L

l1

l1

l1

l2

l2 180° Hybrid 90° Hybrid (c)

FIGURE 5-56 Shallow-cavity crossed-slot antenna: (a) cavity with ridge; (b) cavity with ridge; (c) typical slot conﬁguration. All dimensions are in inches. (From [38], Fig. 2, 1975 IEEE.)

Cavity edge Slot length Ridge width Cavity thickness

0.924λ 1.3λ 0.087λ 0.115λ

Slot width, W2 Ridge height Feed width, W1

0.058λ 0.076λ 0.144λ

Both the ridge and slot shapes can be varied to improve the performance. As fed in Figure 5-56c, the antenna radiates circular polarization on a boresight. Near the horizon (90◦ from the boresight), the polarization reduces to linear as we enter the null of one of the slots.

5-23

WAVEGUIDE-FED SLOTS [24, p. 291; 40, p. 95]

Waveguide is an ideal transmission line for feeding slots. Although its impedance cannot be deﬁned uniquely, all possible candidates—voltage and current, power and

RECTANGULAR-WAVEGUIDE WALL SLOTS

271

current, or power and voltage—yield high values that match the high values of impedance of half-wavelength slots. Waveguide provides a rigid structure with shielded ﬁelds. The slots couple to the internal ﬁelds and allow the easy construction of linear arrays fed from traveling waves or standing waves in the waveguide. By controlling the position of the slots in the walls, the amplitude of the slot excitation can be controlled. The waveguide ﬁelds excite a slot when the slot interrupts the waveguide wall currents. When excited, the slot loads the waveguide transmission line. We make the following assumptions about the wall slots. 1. The slot width is narrow. When a slot grows in width, we must either consider it to be an aperture in the wall or assume that it is excited by interrupting currents in two coordinate directions. 2. The slot is a resonant length and its length is near λ/2. The waveguide environment, the wall thickness, and the position in the wall all affect the resonant length. In most cases, experiments must determine the resonant length. 3. The electric ﬁeld is directed across the narrow width of the slot and varies sinusoidally along its length and is independent of the excitation ﬁelds. This reiterates assumptions 1 and 2. An aperture radiates the polarization of the incident ﬁelds, but resonant-length slots can be excited only with a sinusoidal voltage standing wave. The slot direction determines polarization. 4. The waveguide walls are perfectly conducting and inﬁnitely thin. Even though the walls have thickness, the difference has a small effect on the general form of the slot excitation formulas. As in the case of the resonant length, experiments determine a few values from which the rest must be interpolated, or the values provide the constants for more elaborate models. 5-24 RECTANGULAR-WAVEGUIDE WALL SLOTS The lowest-order mode (TE10 ) in a rectangular waveguide has the following ﬁelds [41, p. 69]: Ey = E0 sin(kc x)e−j kg z kg E 0 sin(kc x)e−j kg z ωµ kc E 0 cos(kc x)e−j kg z Hz = − j ωµ

Hx = −

(5-24)

where kc = π/a, kg2 = kc2 − k 2 , and a is the guide width with cutoff wavelength λc = 2a. We can separate TE10 -mode rectangular waveguide ﬁelds into two plane waves that propagate at an angle to the axis and reﬂect from the two narrow walls. We denote as ξ the angle of the waves measured from the centerline of the waveguide or with respect to the wall. We relate the waveguide propagation to this angle: ξ = sin−1 (λ/λc )

(5-25)

At high frequencies, ξ → 0 and the waves travel straight through the guide as though the walls are not there. As the wavelength approaches cutoff, ξ → 90◦ and the waves

272

DIPOLES, SLOTS, AND LOOPS

reﬂect back and forth between the sidewalls instead of propagating down the guide. This angle factors into the expressions for slot loading to the waveguide transmission line and can be related to propagation: λ λ = cos ξ 1 − (λ/λc )2 c phase velocity Vph = and group velocity = c cos ξ cos ξ λ = cos ξ relative propagation constant P = λg guide wavelength, λg =

For analysis we divide the ﬁelds bouncing down the waveguide into z-directed ﬁelds of the axial wave moving down the guide and x-directed ﬁelds of the transverse wave, a standing wave between the two narrow walls. A standing wave causes a 90◦ separation of the voltage and currents in a transmission line as shown in Figure 5-1. The phase of the currents excited in the waveguide walls due to the ﬁelds will be 90◦ relative to the electric ﬁeld. The wall currents Js are determined by J = n × H, where n is the unit normal to the wall. When we apply this boundary condition to the walls, we obtain the following wall currents: Sidewalls: Jy = −j

E0 kc −j kg z e ωµ

Bottom wall (y = 0 ): Js =

E0 −j kg z e [kg sin(kc x)ˆz + j kc cos(kc x)ˆx] ωµ

Js =

−E0 −j kg z e [kg sin(kc x)ˆz + j kc cos(kc x)ˆx] ωµ

(5-26)

Top wall (y = b):

Equation (5-26) shows that transverse wave currents are 90◦ out of phase with respect to the electric ﬁeld E0 . The current alternates between the two types of current as the wave propagates down the waveguide. In the case of a standing wave along the z-axis caused by a short circuit, the axial wave currents are 90◦ out of phase with the electric ﬁeld across the waveguide (Figure 5-1). The peak amplitude of the transverse wave currents occurs at the same point as the electric ﬁeld in a standing wave along the z-axis, since both are 90◦ out of phase with the axial wave currents. The sidewalls Jy have only transverse wave currents. The top and bottom broad walls have both xdirected transverse wave and z-directed axial wave currents. Figure 5-57a shows the direction and amplitude distribution of these transverse waves. Slots interrupting these

RECTANGULAR-WAVEGUIDE WALL SLOTS IJI

273

IJI

J J

E

J

E J

Traveling Wave Currents

Transverse Standing Wave Currents (a)

(b)

FIGURE 5-57 TE10 -mode rectangular waveguide wall currents: (a) transverse wave currents; (b) axial wave currents.

Axial Wave Current

Transverse Wave Current

Slot

Slot Short

FIGURE 5-58 Short-circuited waveguide axial and transverse wave currents and the location of longitudinal wall slots.

currents are shunt loads to the waveguide. In an axial wave along the z-axis, these transverse waves propagate in the z-axis direction. Equation (5-26) shows that the transverse wave currents are 90◦ phase with respect to the axial wave currents. Figure 5-58 shows the two types of currents along the zaxis when the guide has a short circuit at its end. When measuring slots that interrupt transverse wave currents, we need to place the waveguide short at λg /4 or 3λg /4 away from the slot. This locates the peak of the transverse wave currents ﬂowing around the waveguide walls at the slot shown in Figure 5-58 because the axial wave currents are at a minimum. The second consideration is the shunt load on the waveguide. The λg /4 section of waveguide transforms a short circuit on the end of the waveguide (to the axial wave currents) to an open circuit at the slot. From a voltage point of view the susceptance of the shorted stub is at a minimum. We place the short circuit at

274

DIPOLES, SLOTS, AND LOOPS

λg /2 from the last slot for a series loading slot that interrupts the axial wave currents. This locates the current maximum at the slot and causes maximum interaction with the waveguide ﬁelds. Figure 5-58 illustrates the placement of the next slot λg /2 down the guide at the next current maximum. Figure 5-57a indicates the transverse wave current ﬂow and we see that the currents ﬂow toward the centerline, producing currents 180◦ out of phase on the two sides of the centerline. The two slots in Figure 5-58 are excited by oppositely directed currents that add 180◦ phase shift between the slots. This phase shift compensates for internal standing-wave current phasing of 180◦ due to the λg /2 spacing. Longitudinal top and bottom wall slots cut x-directed transverse shunt currents. The central slot c, located at a current null, fails to be excited. We use this nonradiating slot to insert a traveling probe to measure VSWR. When moved off center, slots d and e cut x-directed currents and are excited. The shunt conductance has the relation g = g1 sin2

πx a

(5-27)

where x is the distance from the guide centerline. Shunt currents on either side of the centerline of the top or bottom wall (Figure 5-57a) have different directions. Besides any traveling-wave phase, slots d and e (Figure 5-59) are 180◦ out of phase. Top-wall longitudinal slots generate no cross-polarization, since all maintain the same orientation. We relate the peak conductance g1 to the direction of the waves in the guide [41]: g1 = 2.09

a cos2 [(π/2) cos ξ ] b cos ξ

(5-28)

Equation (5-28) indicates that the conductance increases for a given spacing off the centerline as the frequency approaches cutoff and ξ → π/2. We cannot use Eqs. (5-27) and (5-28) for design because they do not include the wall thickness and we need to determine the exact length for resonance. The resonant length depends on the spacing from the centerline. Fortunately, the coupling between longitudinal slots is small enough that measurements can be made on single slots. Elliott suggests a measurement plan for longitudinal slots [3]. We build a series of slotted waveguides each containing a single slot at different distances from the centerline. Seven cases are sufﬁcient to generate a curve for design. We need

d

x′ q b

c

q

e g

x′

h

f

a

Shunt Loads

Series Loads

FIGURE 5-59 TE10 -mode rectangular waveguide wall slots.

RECTANGULAR-WAVEGUIDE WALL SLOTS

275

to locate a sliding short circuit farther down the waveguide and adjust it until the standing-wave current peaks at the slot to produce maximum radiation and conductance. With a network analyzer we measure the conductance normalized with respect to the waveguide impedance. Initially, we machine the slots too short, measure the results, and then machine longer slots using the same guides and remeasure until they pass through resonance. Since the manufacturing cost of test slots is high and they require careful measurements, analytical methods of determining slot parameters become attractive. FEM programs can model the details of the slot, the waveguide, and the wall thickness. A number of runs similar to those of the measurements allows design curves to be created. Sidewall slots (Figure 5-59) interrupt shunt transverse waves. Slot a fails to cut surface currents and is not excited. By tilting slot b, currents are cut. The sidewall slot conductance is given for θ < 30◦ by g = g0 sin2 θ

(5-29)

where g0 is the peak conductance. Note that the sidewall slots must cut into the top and bottom walls to achieve a resonant length. The peak conductance can be related to the direction of the waves in the waveguide [Eq. (5-25)] [1, p. 82]: g0 = 2.09

a sin4 ξ b cos ξ

(5-30)

Equation (5-30) shows the relationship of the slot load conductance versus the frequency. As frequency increases, ξ decreases and the conductance falls off as the fourth power of the sine of the angle. The complete theory of Stevenson gives the conductance for an arbitrary tilt [42]: a sin4 ξ g = 2.09 b cos ξ

sin θ cos[(π/2) cos ξ sin θ ] 1 − cos2 ξ sin2 θ

2 (5-31)

Tilting the slots to interrupt currents introduces cross-polarization components in the array pattern. We alternate the direction of tilt to reduce cross-polarization. Two things prevent the total cancellation of cross-polarization. First, the amplitude taper of the array changes the amplitude from element to element and the ﬁelds do not cancel. Alternating the tilt of the slots symmetrically about the centerline in an array with an even number of elements prevents cross-polarization on the boresight. Off the boresight, the array effect of the spaced elements introduces a cross-polarization pattern, since cross-polarization is not canceled at each element. Although Eqs. (5-29) and (5-30) give the slot conductance, they cannot be used for design. They assume an inﬁnitely thin wall and ignore the high level of radiation along the waveguide wall. These slots readily couple to neighboring slots. The effective conductance needs to include the mutual conductance. For these slots we build a series of slotted waveguides containing a group of slots all tilted to the same angle and cut so that they are a resonant length. This means that we will ﬁrst need to build the slots about 5% shorter than resonance length, make measurements, and then machine the slots longer and repeat the measurements to ﬁnd the resonant length. We space the slots at the same distance as will be used in the ﬁnal design and either place a

276

DIPOLES, SLOTS, AND LOOPS

short-circuit beyond the last slot to produce a maximum current at all slots or load the waveguide to form a nonresonant array. We measure the load of the group of slots on the waveguide transmission line using a network analyzer and divide the conductance by the number of slots to get an incremental conductance. This conductance is larger than the one measured on a single slot. We ﬁt the group of measurements to a curve that replaces Eq. (5-29) for design. Axial z-directed waves (Figure 5-57b) peak in the center of the broad walls and taper to zero at the edges. They remain zero on the sidewalls. When centered, transverse slots f and g (Figure 5-59) interrupt the maximum current. When moved off center, g, their series loading to the waveguide drops: R = R0 cos2

πx a

(5-32)

The maximum resistance is related to the direction of the waves in the waveguide: R0 = 2.09

a sin2 ξ 2 π cos sin ξ b cos3 ξ 2

(5-33)

An evaluation of Eq. (5-33) shows that the resistance increases as frequency approaches cutoff for a given location of the slot, a result similar to that for other slot conﬁgurations. The mutual coupling between these series slots is high. We perform incremental resistance experiments similar to the procedure used for sidewall slots to discover the true values of resistance versus offset. Rotating the broadwall transverse slot, h, reduces the z-axis directed current interrupted. When the slot is centered, equal and opposite shunt currents are cut by the slot and the slot fails to present a load to shunt currents: R = R0 cos2 θ

(5-34)

We can excite slots a and c by probe coupling into the waveguide. A probe placed next to the slot and extending into the guide feeds the slot. The longer the probe, the more it disturbs the waveguide ﬁelds to excite the slot. Probes placed on opposite sides of the slots induce ﬁelds 180◦ out of phase with respect to each other.

5-25

CIRCULAR-WAVEGUIDE SLOTS

Figure 5-60 shows the transverse wave and axial wave currents of the circular waveguide TE11 dominant mode. Slots may be placed successfully only at the current maximums without affecting the polarization of the internal wave. A longitudinal slot placed halfway between the current maximums, 45◦ , interrupts only shunt transverse waves. Since any polarization is possible in the circular waveguide, analytically we divide the incident wave into two waves. One is polarized in the direction of the slot; the other is polarized perpendicular to the slot axis. The wave polarized perpendicular to the slot location has its current maximum at the slot and it removes power from the wave. The other wave produces a current null on the slot. When we combine the two ﬁelds after the slot, the unloaded wave is larger and the combined wave rotates its polarization toward the slot. Circumferential slots interrupting axial wave currents

CIRCULAR-WAVEGUIDE SLOTS

J

J

IJI

E J

E

277

IJI

J

Transverse Standing Wave Currents

Traveling Wave Currents

FIGURE 5-60 TE11 -mode circular waveguide wall currents.

also cause polarization rotation of the wave when not centered 90◦ from the electric ﬁeld direction. Slots placed at the maximum of transverse currents cut them when rotated about the axis of the waveguide. Like rectangular-waveguide sidewall slots, the slots oriented perpendicular to the guide axis, circumferential, do not load the waveguide. Rotating the slot increases the shunt load on the waveguide. Slots placed at the maximum of the axial wave cut z-directed currents. Field probes can monitor the internal ﬁelds of the waveguide through a longitudinal slot without causing radiation from the slot. When the slot is rotated away from the axis direction, it interrupts series axial wave currents, loads the waveguide, and radiates. Coaxial TEM-mode transmission line and TM01 -mode circular waveguide have the same outer wall currents (Figure 5-61). Slots can be excited and load the waveguide only by interrupting these axial wave currents. In Figure 5-61, slot a fails to cut currents and is not excited. VSWR measuring probes use this slot. Slots b and c interrupt the currents and series-load the guide. Slot c, whose total length is resonant, is excited by the small portion in the center cutting z-directed currents. We can probe feed slot a, but the probe shunt loads the waveguide or TEM coax that would be series loads on the waveguide if they directly interrupted the axial wave currents.

IJI

Traveling Wave Currents

a

b

c

Wall Slots

FIGURE 5-61 Coax or TM01 -mode circular waveguide wall currents and slots.

278

5-26

DIPOLES, SLOTS, AND LOOPS

WAVEGUIDE SLOT ARRAYS [4, p. 402]

Waveguide slot arrays can produce low sidelobe antennas for pencil beams with good aperture efﬁciency. Array fabrication requires close manufacturing tolerances to achieve the desired amplitude distribution because random errors in manufacture produce unwanted sidelobes and raise the general sidelobe level. Producing these arrays is an art requiring careful analysis of all slot interactions, slot dimensioning determined from models and measurements, and precision machining and assembly. An array consists of a set of waveguides loaded with slots and joined with a corporate feed into the total array. The corporate feed can also be a slotted array feeding the individual waveguides that contain the radiating slots. Aperture size and distribution determine the beamwidth and sidelobes in the various planes. We divide arrays of slots into two groups: nonresonant, excited by traveling waves, and resonant, excited by standing waves. Waves either travel along the guide into a terminating load or reﬂect from a short and set up standing waves along the z-axis (Figure 5-58). Traveling-wave currents excite the slots as they pass, and slots may be placed anywhere relative to the load. The distance between slots and the propagation constant determine the relative phases. Standing waves set up a ﬁxed sinusoidal current pattern along the waveguide axis at a given frequency. The standing-wave phase is either 0◦ or 180◦ . Slots placed in the current nulls of standing waves interrupt no currents and fail to be excited by the waveguide. We can vary the amplitude by the z-axis placement of the slots. The termination determines the array type. Do not confuse transverse waves that produce shunt currents and z-axis standing waves caused by a short-circuit termination. Both traveling and standing waves on the z-axis have shunt currents. Standing waves (resonant array) produce beams normal to the array axis. A resonant array maintains its beam direction when frequency changes, but the standing-wave pattern shifts and changes the excitation of the slots (Figure 5-62). The amplitudes of the slots farthest from the short circuit change the most, since the standing waves have shifted farther. The length of the resonant array determines its bandwidth. The pattern shape changes because distribution and input impedance change as the loads change when the standing-wave currents shift. Nonresonant array (traveling-wave) beam directions are functions of the propagation constant of the wave exciting the slots. Changing the frequency shifts the beam direction. If the load on the end reﬂects a wave, another beam forms from the reﬂected traveling wave. The second beam appears at the same angle to the axis of the waveguide

Frequency Shift by 10%

Waveguide Axis

Slot

Half Guide Wavelength Slot Slot

Short Slot

FIGURE 5-62 Standing-wave currents in resonant array relative to slots and after 10% frequency shift.

279

WAVEGUIDE SLOT ARRAYS

as the ﬁrst but measured from the −z-axis. The ﬁrst-pass radiated power and return loss of the load determine the level of this second beam relative to the ﬁrst. Both resonant and nonresonant waveguide slot arrays use resonant-length slots. We space the slots λg /2 apart in the resonant array, as shown in Figure 5-58. We place the slots at alternating positions about the centerline of the broadwall or at alternating tilt angles in the sidewall to give the additional 180◦ phase shift to produce a broadside beam. The admittances of the slots of the resonant array add at the input because the λg /2 spacing produces a complete rotation around the Smith chart. In nonresonant arrays a traveling wave is used to excite the slots. We space the slots at other than λg /2 distances and terminate the waveguide with a load. We assume a matched system throughout the antenna in a ﬁrst-order analysis suitable for most designs. The beam of most nonresonant slot arrays is designed to backﬁre at an angle to broadside. 5-26.1 Nonresonant Array [43] In a nonresonant waveguide, slot array resonant-length slots are used in a travelingwave antenna terminated at the end with a load. The antenna radiates at an angle to the normal of the waveguide face determined by wave velocity and slot spacing. We vary the slot loading along the waveguide so that each slot radiates the proper amount of the remaining power. A termination absorbs the power remaining after the last slot. With a mismatched termination the reﬂected power radiates a second lower-amplitude beam as the wave travels to the source. We design with either shunt- or series-loading slots. A shunt slot radiates the power |V |2 gi /2, where gi is the normalized slot conductance. Similarly, a series slot radiates the power |I |2 ri /2, where ri is the normalized slot resistance. We normalize the conductance or resistance to a per unit length function: g(z) or r(z). The attenuation equation (4-78) becomes 1 dP = −g(z) or P (z) dz

− r(z)

(5-35)

Equation (5-35) modiﬁes the normalized attenuation equation (4-79) [24, p. 291]: g(z)L =

|A(z)|2

L

[1/(1 − R)] 0

z

|A(z)|2 dz −

(5-36) |A(z)|2 dz

0

where the aperture runs ±L/2 and R is the ratio of the input power absorbed by the termination. A(z) is the normalized aperture distribution on the interval ± 21 . We change to r(z)L in Eq. (5-36) for series-loading slots. Equation (5-36) assumes light loading by the slots so that the waveguide transmission line is matched at all points. This approximation improves as the length increases. Equation (5-36) is the same as Eq. (4-79) except for a constant. We divide the values in Table 4-28 or Figure 4-26 by 4.34 to calculate normalized conductance (resistance) of shunt (series) slots times the array length. Each slot provides the loading over the spacing between slots: gi =

d/2

−d/2

where d is the spacing of the slot at zi .

g(z) dz g(zi ) d

280

DIPOLES, SLOTS, AND LOOPS

We space the slots at other than λg /2. At λg /2 spacings, all reﬂections from the mismatches (slots) add in phase at the input. The small mismatches from each slot add with various phase angles for element spacing different from λg /2 and cancel each other to some extent to give a good input match over a reasonable bandwidth. When we increase the array length, we can no longer ignore the waveguide losses. The slot conductances become very small and radiate power on the same order as the losses. We modify Eq. (5-36) to include the losses as in Eq. (4-79), and the slot conductance increases to compensate for the ohmic losses in the walls. A small slot conductance is difﬁcult to achieve with longitudinal broadwall slots because one edge of the slot must be over the centerline of the waveguide wall and the results become unpredictable. The achievable conductances limit possible distributions in a slotted waveguide array. Mutual coupling between slots changes the distribution and we must modify the slot offsets to account for mutual coupling using Eq. (3-23). If we specify the radiating power of each slot in a discrete sequence Pi , we modify Eq. (5-36). The integrals become summations, since |A(z)|2 = δ(z − id)Pi where d is the slot spacing, δ(x) the Dirac delta (impulse) function, and Pi the power coefﬁcient of the ith slot. The power radiated is N

z 0

|A(z)|2 dz

0

i=1

The integral reduces to

L

Pi = Pin (1 − R) =

|A(z)|2 dz is the power radiated by the preceding slots. Equation (5-36) gi = ri =

1−

Pi i−1 n=1

Pn

(5-37)

Dissipating more power in the termination decreases each Pi and the required conductance (resistance) range of the slots. We alternate the locations of longitudinal slots about the centerline of the broadwall to add 180◦ phase shift between elements. Similarly, sidewall slot directions are alternated along the array. The additional phase shifts cause backﬁre of the beam in most cases. The element spacing, as well as the traveling-wave phase velocity, determines the beam direction. The phasing equation in the array factor for beam peak becomes kd cos θ + 2nπ = P kd − π, where θ is measured from the array axis, P is the relative propagation constant (P < 1), and n is an arbitrary integer. We solve for the beam peak direction and the necessary spacing to get a particular beam direction: (n + 12 )λ −1 θ = cos P− (5-38) d n + 12 d = λ P − cos θmax We usually work with n = 0 because using n > 0 produces multiple beams.

(5-39)

WAVEGUIDE SLOT ARRAYS

281

Example Compute slot spacing to produce a beam at θ = 135◦ in a waveguide of width 0.65λ. Calculate the relative propagation constant from the general equation for a waveguide. 2 λ P = 1− λc For λc = 2a,

P =

1−

1 1.3

2 = 0.640 =

λ λg

From Eq. (5-39), using n = 0, we determine spacing in free space: d/λ = 0.371. The waveguide spacing is given by d d = P = 0.371(0.640) = 0.237 λg λ If we use n = 1, then d/λ = 1.11, which radiates an additional beam at θ = 79◦ for n = 0 [Eq. (5-38)]. Beams enter visible space at cos θ = −1 (180◦ ) and move toward end ﬁre (θ = 0) as the spacing increases. We calculate the region of single-beam operation from Eq. (539). The minimum d/λ occurs when θ = 180◦ for n = 0, and the maximum occurs when θ = 180◦ for n = 1: d 1.5 0.5 ≤ ≤ (5-40) 1+P λ 1+P We substitute the upper bound into Eq. (5-38) and use n = 1 to derive the minimum angle of single-beam operation: 1+P θmin = cos−1 P − 3

(5-41)

Example Determine the minimum scan angle (toward end ﬁre) for P = 0.6, 0.7, 0.8, and 0.9 that has a single beam. We substitute these values into Eq. (5-41) to ﬁnd: P θmin

0.6

0.7

0.8

0.9

86.2◦ 82.3◦ 78.5◦ 74.5◦

If we scan to θ = 90◦ , the spacing becomes λg /2 and the mismatches from each slot add to the input and produce a resonant array. The array with a forward ﬁring beam has a slot spacing greater than λg /2. Given a waveguide with P = 0.8, we use Eq. (5-39) to calculate spacing to give beams at 80◦ and 100◦ : 0.5 d = = 0.798 and λ 0.8 − cos 80◦ d d = P = 0.639 and λg λ

0.5 d = = 0.514 λ 0.8 − cos 100◦ d d = P = 0.411 λg λ

282

DIPOLES, SLOTS, AND LOOPS

A nonresonant array has a backﬁre beam that scans toward broadside as frequency (and P ) increases. Hansen [44, p. 90] gives the slope of the beam shift with frequency change: d sin θ 1 f = − sin θ (5-42) df P where f is the frequency. 5-26.2 Resonant Array We space the slots at λg /2 and terminate the waveguide end with a short circuit either λg /4 or 3λg /4 from the last one for shunt loading slots in a resonant array. The beam radiates broadside to the array. The 2 : 1 VSWR bandwidth of the array is approximately 50%/N for N elements in the array. The antenna is narrowband. The admittances of all elements add at the input. To have a matched input, N g = 1, where gi is the i=1 i normalized slot conductance. If we deﬁne Pi as the normalized power radiated by the ith slot, then N gi = Pi where Pi = 1 i=1

5-26.3 Improved Design Methods The methods given above ignore the interaction of slots and their effect on the transmission line. We can describe the array as a loaded transmission line and consider the interactions of the slots by accounting for the transmission-line mismatches [45, pp. 9–11]. We ignore the mutual coupling for longitudinal broadwall slots because it is small, but sidewall slots have high mutual coupling and require an adjustment of the effective slot impedance. We use an incremental admittance, found from the measured change in admittance, when one slot is added to the array or total conductance of the array divided by the number. This accounts somewhat for the mutual coupling. Elliott and Kurtz [46] relate the self-admittance of a longitudinal broad-wall slot, measured or calculated, to the mutual admittance of the array of slots found from equivalent dipoles. They use Babinet’s principle and the mutual impedance of equivalent dipoles. The method requires solution of a set of 2N equations in the location and length of the slots to give the desired excitation while accounting for mutual coupling. Their formulation ignores slot interaction in the waveguide beyond the ﬁrst-order mode. Elliott [47] extends this method to the analysis and design of nonresonant arrays. Of course, when we design a planar array, the slots between waveguide sticks couple readily and we need to account for the mutual coupling between them. The voltage excitation needs to be adjusted to account for this coupling or the desired distribution will not be achieved. Dielectric loaded waveguide arrays require additional analysis because the approximation of a piecewise sinusoidal distribution, such as dipole current, fails to model the slot distribution adequately. Elliott [48] uses a slot distribution E(x) = cos

πx 2b

where b is the length. Mutual impedances between dipoles that have the wrong distribution are not used; instead, the active admittances are found from forward and

REFERENCES

283

back scattering between the slots directly. The method still requires the solution of 2N equations for the slot lengths and locations. REFERENCES 1. R. F. Harrington, Time-Harmonic Electromagnetic Fields, McGraw-HiIl, New York, 1961. 2. C. A. Balanis, Antenna Theory, Analysis and Design, 2nd ed., Wiley, New York, 1997. 3. H. G. Booker, Slot aerials and their relation to complementary wire aerials, Proceedings of IEE, vol. 92, pt. IIIA, 1946, pp. 620–626. 4. R. S. Elliott, Antenna Theory and Design, Prentice-Hall, Englewood Cliffs, NJ, 1981. 5. V. H. Rumsey, Frequency Independent Antennas, Academic Press, New York, 1966. 6. W. H. Watson, Wave Guide Transmission and Antenna Systems, Oxford University Press, London, 1947. 7. R. W. Masters, Super-turnstile antenna, Broadcast News, vol. 42, January 1946. 8. J. D. Kraus, Antennas, McGraw-Hill, New York, 1950. 9. H. M. Elkamchouchi, Cylindrical and three-dimensional corner reﬂector antennas, IEEE Transactions on Antennas and Propagation, vol. AP-31, no. 3, May 1983, pp. 451–455. 10. S. Maci et al., Diffraction at artiﬁcially soft and hard surfaces by using incremental diffraction coefﬁcients, IEEE AP-S Symposium, 1994, pp. 1464–1467. 11. E. L. Bock, J. A. Nelson, and A. Dome, Very High Frequency Techniques, McGraw-Hill, New York, 1947. Chapter 5. 12. A. J. Poggio and P. E. Mayes, Pattern bandwidth optimization of the sleeve monopole antenna, IEEE Transactions on Antennas and Propagation, vol. AP-14, no. 5, September 1966, pp. 623–645. 13. W. L. Stutzman and G. A. Thiele, Antenna Theory and Design, Wiley, New York, 1981. 14. H. E. King and J. L. Wong, An experimental study of a balun-fed open-sleeve dipole in front of a metallic reﬂector, IEEE Transactions on Antennas and Propagation, vol. AP-19, no. 2, March 1972, pp. 201–204. 15. J. L. Wong and H. E. King, A cavity-backed dipole antenna with wide bandwidth characteristics, IEEE Transactions on Antennas and Propagation, vol. AP-21, no. 5, September 1973, pp. 725–727. 16. A. Kumar and H. D. Hristov, Microwave Cavity Antennas, Artech House, Boston, 1989. 17. Y. Mushiake, An exact step-up impedance ratio chart of folded antenna, IRE Transactions on Antennas and Propagation, vol. AP-2, 1954, p. 163. 18. R. C. Hansen, Folded and T-match dipole transformation ratio, IEEE Transactions on Antennas and Propagation, vol. AP-30, no. 1, January 1982, pp. 161–162. 19. The ARRL Antenna Book, American Radio Relay League, Inc., Newington, CT, 1974. 20. R. A. Burberry, VHF and UHF Antennas, Peter Peregrinus, London, 1992. 21. N. Marchand, Transmission-line conversion transformers, Electronics, December 1941, pp. 142–145. 22. W. L. Weeks, Antenna Engineering, McGraw-Hill, New York, 1968. 23. W. K. Roberts, A new wide-band balun, Proceedings of IRE, vol. 45, December 1957, pp. 1628–1631. 24. S. Silver, ed., Microwave Antenna Theory and Design, McGraw-Hill, New York, 1949. 25. J. W. Duncan and V. P. Minerva, 100 : 1 Bandwidth balun transformer, Proceedings of IRE, vol. 48, February 1960, pp. 156–164. 26. B. A. Munk, Baluns, Chapter 23 in J. D. Kraus and R. J. Marhefka, Antennas, McGrawHill, New York, 2002.

284

DIPOLES, SLOTS, AND LOOPS

27. P. K. Park and C. T. Tai, Receiving antennas, Chapter 6 in Y. T. Lo and S. W. Lee, eds., Antenna Handbook, Van Nostrand Reinhold, New York, 1993. 28. A. Alford and A. G. Kandoian, Ultrahigh frequency loop antennas, AIEE Transactions, vol. 59, 1940, pp. 843–848. 29. A. J. Fenn, Arrays of horizontally polarized loop-fed slotted cylinder antennas, IEEE Transactions on Antennas and Propagation, vol. AP-33, no. 4, April 1985, pp. 375–382. 30. T. Tsukiji and S. Tou, On polygonal loop antennas, IEEE Transactions on Antennas and Propagation, vol. AP-28, no. 4, July 1980, p. 571. 31. W. C. Wilkinson et al., Two communication antennas for the Viking lander spacecraft, 1974 IEEE Antennas and Propagation Symposium Digest, vol. 12, June 1974, pp. 214–216. 32. C. C. Kilgus, Multielement fractional turn helices, IEEE Transactions on Antennas and Propagation, vol. AP-16, no. 4, July 1968, pp. 499–500. 33. C. C. Kilgus, Resonant quadriﬁlar helix, IEEE Transactions on Antennas and Propagation, vol. AP-17, no. 3, May 1969, pp. 349–351. 34. A. A. Oliner, Equivalent circuits for discontinuities in balanced strip transmission line, IRE Transactions on Microwave Theory and Techniques, vol. MTT-3, March 1955, pp. 134–143. 35. J. S. Rao and B. N. Ras, Impedance of off-centered stripline fed series slot, IEEE Transactions on Antennas and Propagation, vol. AP-26, no. 6, November 1978, pp. 893, 894. 36. J. Van Bladel, Small hole in waveguide wall, Proceedings of IEE, vol. 118, January 1971, pp. 43–50. 37. P. K. Park and R. S. Elliott, Design of collinear longitudinal slot arrays fed by boxed stripline, IEEE Transactions on Antennas and Propagation, vol. AP-29, no. 1, January 1981, pp. 135–140. 38. H. E. King and J. L. Wong, A shallow ridged-cavity cross-slot antenna for the 240- to 400MHz frequency range, IEEE Transactions on Antennas and Propagation, vol. AP-23, no. 5, September 1975, pp. 687–689. 39. C. A. Lindberg, A shallow-cavity UHF crossed-slot antenna, IEEE Transactions on Antennas and Propagation, vol. AP-17, no. 5, September 1969, pp. 558–563. 40. R. C. Hansen, ed., Microwave Scanning Antennas, Vol. II, Academic Press, New York, 1966. 41. R. F. Harrington, Time-Harmonic Electromagnetic Fields, McGraw-Hill, New York, 1961, p. 69. 42. A. F. Stevenson, Theory of slots in rectangular waveguides, Journal of Applied Physics, vol. 19, January 1948, pp. 24–38. 43. A. Dion, Nonresonant slotted arrays, IRE Transactions on Antennas and Propagation, vol. AP-7, October 1959, pp. 360–365. 44. R. C. Hansen, in A. W. Rudge et al., eds., The Handbook of Antenna Design, Vol. II, Peter Peregrinus, London, 1983. 45. M. J. Ehrlich, in H. Jasik, ed., Antenna Engineering Handbook, McGraw-Hill, New York, 1961. 46. R. S. Elliott and L. A. Kurtz, The design of small slot arrays, IEEE Transactions on Antennas and Propagation, vol. AP-26, no. 2, March 1978, pp. 214–219. 47. R. S. Elliott, On the design of traveling-wave-fed longitudinal shunt slot arrays, IEEE Transactions on Antennas and Propagation, vol. AP-27, no. 5, September 1979, pp. 717–720. 48. R. S. Elliott, An improved design procedure for small arrays of shunt slots, IEEE Transactions on Antennas and Propagation, vol. AP-31, no. 1, January 1983, pp. 48–53.

6 MICROSTRIP ANTENNAS

Microstrip antennas are planar resonant cavities that leak from their edges and radiate. We can utilize printed circuit techniques to etch the antennas on soft substrates to produce low-cost and repeatable antennas in a low proﬁle. The antennas fabricated on compliant substrates withstand tremendous shock and vibration environments. Manufacturers for mobile communication base stations often fabricate these antennas directly in sheet metal and mount them on dielectric posts or foam in a variety of ways to eliminate the cost of substrates and etching. This also eliminates the problem of radiation from surface waves excited in a thick dielectric substrate used to increase bandwidth. As electronic devices continue to shrink in size, the antenna designer is pushed to reduce the antenna size as well. Cavity antennas use valuable internal volume, but we have the conﬂict that restricting the volume limits impedance bandwidth. Bandwidths widen with increased circuit losses (material losses) or by efﬁcient use of the restricted volume. Bounds on bandwidth can be found by enclosing the antenna in a sphere and expanding the ﬁelds into TE and TM spherical modes [1,2]. Each mode radiates, but it requires more and more stored energy as the mode number increases. Decreasing the volume increases the Q value of each mode and a sum, weighted by the energy in each mode, determines the overall Q value. Antennas that use the spherical volume efﬁciently and reduce power in the higher-order modes have the greatest bandwidths. A single lowest-order mode puts an upper bound on bandwidth, given the size of the enclosing sphere. Greater volumes have potential for greater bandwidth provided that the energy in higher-order spherical modes is restricted. Increasing material losses or adding small resistors increases bandwidth beyond the single-mode bound [2]. We discover that increasing the volume of ﬂush antennas increases the impedance bandwidth provided that the radiation mode on the structure can be maintained. Thicker substrates develop greater bandwidths, but they increase the possibility of higher-ordermode excitation and surface-wave losses. Losses limit the lower bound of bandwidth Modern Antenna Design, Second Edition, By Thomas A. Milligan Copyright 2005 John Wiley & Sons, Inc.

285

286

MICROSTRIP ANTENNAS

as we reduce the thickness because efﬁciency degrades to a point where the bandwidth remains constant. Microstrip consists of a metal strip on a dielectric substrate covered by a ground plane on the other side. Unlike stripline, the single ground plane shields the circuit on only one side, but normal packaged microstrip—within a receiver, for example—has a second shielding ground plane to reduce circuit interactions. The dielectric substrate retains most of the power because the shielding ground plane is spaced a few substrate thicknesses away. Removing the shield in antenna applications allows radiation from resonant cavities. We also discover feeding circuits etched on the substrate radiate to some extent, but their radiation is comparatively small. Arrays of antennas can be photoetched on the substrate, along with their feeding networks, and microstrip provides easy connections to active devices and allows placement of preamps or distributed transmitters next to the antenna elements. Diode phase-shifter circuits etched in the microstrip form single-board phased arrays. Microstrip circuits make a wide variety of antennas possible through the use of the simple photoetching techniques. The vast literature on microstrip antennas concentrates on the microwave circuit analysis of the internal parts of the antenna used to control the internal modes. Designers have increased the bandwidth of the antenna by coupling to multiple resonators, such as vertically stacked or coplanar coupled patches or by using internal slots and apertures. These multiple resonators increase the impedance bandwidth, and in the best cases the antenna continues to radiate the same pattern. As antenna designers we need to concentrate ﬁrst on obtaining the desired pattern while working to increase the impedance bandwidth. Simple microstrip antennas have much larger pattern bandwidths than impedance bandwidths, but as more resonators are added to increase the impedance bandwidth, spreading in the horizontal plane alters the radiated pattern and we must return to concentrate on the pattern. Microstrip patch antennas consist of metal patches large with respect to normal transmission-line widths. A patch radiates from fringing ﬁelds around its edges. Impedance match occurs when a patch resonates as a resonant cavity. When matched, the antenna achieves peak efﬁciency. A normal transmission line radiates little power because the fringing ﬁelds are matched by nearby counteracting ﬁelds. Power radiates from open circuits and from discontinuities such as corners, but the amount depends on the radiation conductance load to the line relative to the patches. Without proper matching, little power radiates. The edges of a patch appear as slots whose excitations depend on the internal ﬁelds of the cavity. A general analysis of an arbitrarily shaped patch considers the patch to be a resonant cavity with metal (electric) walls of the patch and the ground plane and magnetic or impedance walls around the edges. The radiating edges and fringing ﬁelds present loads along the edges. In one analysis [3] the patch effective size is increased to account for the capacitive susceptance of fringing ﬁelds, and the radiation admittance is ignored to calculate resonant frequency. The far ﬁeld is integrated to compute radiated power and the equivalent radiation conductance. The second method [4] is to retain the patch size but satisfy boundary conditions into a loaded wall whose load is determined by radiation and fringing ﬁelds. Assuming a constant electric ﬁeld from the ground plane to the substrate allows solutions in terms of modes TM to the substrate thickness. Boundary conditions determine possible modes and correspond to the dual TE modes of waveguides having electric walls. Patches in the shape of standard coordinate system

MICROSTRIP ANTENNA PATTERNS

287

axes, such as rectangular and circular, give solutions in terms of tabulated functions. Numerical techniques used for arbitrarily shaped waveguides can be applied to patches with nonstandard shapes. We consider only rectangular and circular patches. 6-1 MICROSTRIP ANTENNA PATTERNS We start our discussion of patches with their pattern characteristics. It is difﬁcult to separate a discussion of pattern from the internal construction consideration, but we will only brieﬂy discuss the internal structures that affect the pattern. The small size of microstrip antennas limits control of the pattern and we must use arrays of patches to control its pattern seriously. Rectangular and circular are the most common shapes for microstrip antennas and they radiate similar broad patterns. When we load the cavity to shrink its size, it radiates wider beamwidth patterns that lower directivity (gain). Antennas that couple to coplanar patches to increase the impedance bandwidth will radiate narrower beams, but the basic patch has a wide beamwidth. If we couple to multiple coplanar patches, we can expect the pattern to narrow or vary its shape as the mixture of modes on the various patches changes over the frequency range of operation. Patches consist of metal plates suspended over large ground planes. We excite the cavity in a variety of ways that we discuss later. Electric currents ﬂow on the plate and on the ground plane around the antenna, and these radiate. If we use vertical probes to excite the antenna from coaxial lines, the currents ﬂowing on these radiate and add to the pattern. We can reduce the antenna size by adding vertical shorting plates (quarter-wave patches) or shorting pins near the feed pins (compact patches), and these also radiate from the current ﬂow on them. Remember that the patch radiates from real electric currents, although the distribution is complicated. We simplify the problem of computing patch radiation by using magnetic currents along the edges. Figure 6-1 illustrates the fringing electric ﬁelds around the edges of square and circular patch antennas excited in the lowest-order cavity modes. The arrow sizes indicate the magnitude of the ﬁelds. The square patch has nearly uniform ﬁelds along two edges we call the width, and a sinusoidal variation along the other two edges, called the resonant length. The ﬁelds vanish along a virtual electrically short-circuited plane halfway across the patches. On either side of the short-circuit plane, the ﬁelds are directed in opposite directions. Looking from above the ﬁelds along the width, both edges are in the same direction. The circular patch fringing ﬁelds distribution varies as cos φ, where the angle φ along the rim is measured from the peak electric ﬁeld. Magnetic currents found from the fringing electric ﬁelds can replace the electric currents located on the patch and the surrounding ground plane for pattern analysis. Figure 6-2 shows the distribution of magnetic currents around the edges, with the size of the arrowhead indicating magnitude. Our use of magnetic currents around the patch perimeter reduces the pattern calculation to equivalent slots. A two-element array consisting of slots with equivalent uniform magnetic currents produces the E-plane radiation of a rectangular patch. To √ ﬁrst order, the slots are spaced λ/2 εr and we can determine the pattern from the equivalent two-element array. The magnetic currents along the resonant length sides individually cancel because the current changes direction halfway across the edge. The currents also cancel from side to side. These cancellations eliminate pattern contributions to the E- and H -plane patterns. The slot length determines the H -plane pattern.

288

MICROSTRIP ANTENNAS

(a)

(b)

FIGURE 6-1 Fringing electric ﬁelds around microstrip patches: (a) square; (b) circular. (From L. Diaz and T. A. Milligan, Antenna Engineering Using Physical Optics, Figs. 3.12 and 3.19, 1996 Artech House, Inc.).

The H -plane of the slot has the same pattern as the E-plane of a dipole and produces a null along its axis. Figure 6-3 illustrates the pattern of a patch on an inﬁnite ground plane using a free-space substrate. The two-element slot array in the E-plane generates a null along the ground plane because the elements are spaced λ/2. The H -plane dashed curve shows the null along the ground plane due to the polarization of the slots. The light curves give the Huygens source polarization (Section 1-11.2) patterns in the diagonal planes. The antenna radiates cross-polarization (dashed curve) in this plane from the combination of separated magnetic currents along the resonant-length sides and from the unbalance in the beamwidths in the principal planes. When we design a microstrip patch on a dielectric substrate, the size reduction brings the two slots closer together and widens the E-plane beamwidth and eliminates its null along the ground plane. Figure 6-4 illustrates the pattern of a patch designed for a substrate with εr = 2.2. The H -plane pattern retains its null along the ground plane due to the slot pattern. The cross-polarization of the Huygens source in the diagonal plane increases because of the increased difference between the beamwidths of the principal plane patterns. Table 6-1 gives the directivity of a square and circular patch on an inﬁnite ground plane found by integrating the pattern. The range of directivity

MICROSTRIP ANTENNA PATTERNS

289

(a)

(b)

FIGURE 6-2 Equivalent magnetic currents on the edges of microstrip patches: (a) square; (b) circular.

of a patch is limited. Increasing the width of a rectangular patch increases directivity by shrinking the H -plane beamwidth. We gain some control of the pattern by placing the patch on a ﬁnite ground plane. Figure 6-5 shows the pattern of a square patch on a 2.21 dielectric constant substrate when located on circular disks 5λ, 2λ, and 1λ in diameter. On a 5λ ground plane, edge diffraction adds ripple to the pattern. As the ground plane increases, the angular separation between the ripples decreases, due to the increased array size of the radiation from the two edges. The H -plane pattern widens signiﬁcantly for 1λ- and 2λ-diameter ground planes, as the limited ground plane can no longer support the currents that make the patch edge radiate like a slot. Although the principal-plane beamwidths are more nearly equal for the patch on the 2λ-diameter disk, the cross-polarization in the diagonal plane increases relative to the pattern on the inﬁnite ground plane. The 1λ ground plane increases the gain of the patch by about 1 dB relative to the patch on an

290

MICROSTRIP ANTENNAS

E-Plane

X-Pol Diagonal Plane Diagonal Plane H-Plane

FIGURE 6-3 Patterns of microstrip patch on a free-space substrate mounted on an inﬁnite ground plane.

H-Plane X-Pol

E-Plane

Diagonal

FIGURE 6-4 Patterns of microstrip patch on a dielectric substrate εr = 2.2 over an inﬁnite ground plane. TABLE 6-1 Estimated Directivity of Square and Circular Microstrip Patches on a Large Ground Plane Dielectric Constant

Square Patch (dB)

Circular Patch (dB)

1.0 2.0 3.0 4.0 6.0 8.0 10.0 16.0

8.4 7.7 7.2 7.0 6.7 6.5 6.4 6.3

9.8 7.6 6.7 6.2 5.8 5.5 5.4 5.1

MICROSTRIP ANTENNA PATTERNS

291

X-Pol., Diag.

H-Plane

E-Plane Diag.

(a)

H-Plane

X-Pol., Diag, E-Plane

Diag.

( b)

FIGURE 6-5 Patterns of microstrip patches with dielectric substrate εr = 2.2 mounted over ﬁnite circular ground planes: (a) 5λ diameter; (b) 2λ diameter. (continued )

292

MICROSTRIP ANTENNAS

H-Plane

X-Pol., Diag. E-Plane

(c)

FIGURE 6-5 (continued ) (c) 1λ diameter.

RHC Principal Planes Diag. LHC

Diag. Planes

FIGURE 6-6 Circularly polarized patch mounted on a 1λ-diameter ground plane.

MICROSTRIP PATCH BANDWIDTH AND SURFACE-WAVE EFFICIENCY

293

inﬁnite ground plane. In this case the edge diffractions add constructively to narrow the beamwidths. We can take advantage of the nearly equal E- and H -plane patterns in the forward hemisphere to produce a pattern with excellent circular polarization over the entire hemisphere when we feed the patch for circular polarization. Figure 6-6 gives the circular polarization pattern when the patch is fed in two spots with equal signals phased 90◦ apart. The cross-polarization is 13 dB below the co-polarization at θ = 90◦ in the principal planes and −7 dB in the diagonal plane. We retain these excellent polarization characteristics over a large ground plane if we place the ﬁnite ground plane on a 1λ or greater pedestal above the ground plane. 6-2 MICROSTRIP PATCH BANDWIDTH AND SURFACE-WAVE EFFICIENCY Microstrip patches radiate from the currents induced on the patch or equivalently, the magnetic currents around the periphery of the patch and from surface waves induced in the dielectric slab. The surface waves radiate when they reach the edges of the substrate and their radiation contributes to the normal patch radiation. The fringing ﬁelds from the patch to the ground plane readily excite the lowest-order surface-wave TM0 mode that has no low frequency cutoff. Any thickness dielectric slab supports this mode. We can control the surface-wave radiation by limiting the substrate area or by adding etched photonic bandgap patterns to the open areas of the substrate, but generally, surface waves are undesirable. As the substrate thickness or dielectric constant increases, the ratio of the power in surface waves increases. When we calculate the microstrip patch antenna impedance bandwidth, we must include the directly radiated power and the surface-wave power. For most cases we consider surface-wave radiation as reducing radiation efﬁciency, but for a single patch on a substrate with limited area, its radiation can add constructively. We eliminate surface waves by using metal plate patches without dielectric substrates or low-density foam supports of the patch. Surface waves are bound to the dielectric similar to any transmission line except that the ﬁeld decays exponentially in the direction normal to the surface. Because the surface wave is excited along the ﬁnite edges of the patch, it spreads in the horizontal√plane. The radiation spreads like a two-dimensional wave and the ﬁelds decay as 1/ r, where r is the horizontal distance from the edge. This is a far-ﬁeld approximation, and close to the edge it is a near-ﬁeld problem. Unfortunately, these surface waves increase the coupling between patches fabricated on the same substrate. Simple formulas have been derived for the impedance bandwidth of rectangular patches that include the surface-wave loss [5]. Since substrates can be both electric and magnetic, we deﬁne the index of refraction of a patch substrate that includes both √ parameters: n = εr µr . The idea is that the ratio of space-wave radiation to surfacewave radiation can be found for any small antenna mounted on the substrate and we can then apply it to a patch. By integrating the power density in the radiation from a horizontal Hertzian (incremental) dipole spaced the substrate thickness over a ground plane, we obtain the space-wave radiated power in closed form given the substrate thickness h and the free-space propagation constant k: PRh k 2 (kh)2 · 20 µ2r C1 C1 = 1 −

1 0.4 + 4 n2 n

(6-1)

294

MICROSTRIP ANTENNAS

We express the current on the patch as an integral of Hertzian dipoles. The surfacewave power generated in the substrate by the Hertzian dipole can be simpliﬁed when the substrate is thin: 1 3 h 2 3 3 PSW = k (kh) · 15πµr 1 − 2 (6-2) n We deﬁne the surface-wave radiation efﬁciency by the ratio of radiated power to total power: Ph 4C1 (6-3) ηSW = h r h = 2 3 4C + 3πkhµ PR + PSW 1 r (1 − 1/n ) We relate the power radiated by a patch to a Hertzian dipole by integrating the surface current on the patch consisting of a distribution of small dipoles to calculate the total space-wave power of the patch: PR = PRh m2eq = PRh

2 JS dx dy

(6-4)

S

For a rectangular patch the ratio of PR to PRh m2eq , p, can be approximated by a simple formula given the resonant length L, the width W , and the propagation constant k: p =1−

0.16605(kW )2 0.02283(kW )4 0.09142(kL)2 + − 20 560 10

(6-5)

The 2 : 1 VSWR of the rectangular patch is related to the quality factor Q that includes the space- and surface-wave radiations: BW = √

16C1 p 1 h W = √ 2Q 3 2 ηSW εr λ0 L 1

(6-6)

Figure 6-7 plots the 2 : 1 VSWR bandwidth given by Eq. (6-6) for common substrates versus the free-space thickness in wavelengths and includes the radiation due to surface waves. The surface-wave radiation found using Eq. (6-3) becomes a signiﬁcant part of the total radiation as the substrate thickness increases or the dielectric constant increases, as shown in Figure 6-8 of the surface-wave loss. For a single resonator circuit model for a patch, Eq. (6-6) computes bandwidth from the Q and the allowable input VSWR: BW =

VSWR − 1 √ Q VSWR

or

Q=

VSWR − 1 √ BW VSWR

(6-7)

We determine bandwidth at different VSWR levels by manipulating Eq. (6-7): √ BW2 (VSWR2 − 1) VSWR1 √ = (6-8) BW1 (VSWR1 − 1) VSWR2 Quality factor Q is another way of expressing efﬁciency. The Q used in Eq. (6-6) is the combination of the space-wave radiation QR and the surface-wave radiation QSW : 1 1 PR + PSW 1 1 = + = = Q Qrad QR QSW ωWT

2:1 VSWR Bandwidth, %

MICROSTRIP PATCH BANDWIDTH AND SURFACE-WAVE EFFICIENCY

295

4.5 2.2

6.0

2.94 9.8 1.0

Substrate Thickness, l (Free Space)

Surface Wave Loss, dB

FIGURE 6-7 2:1 VSWR bandwidth of square microstrip patches versus substrate thickness in free-space wavelengths, including surface-wave radiation.

2.2 2.94 4.5 9.8

6.0

Substrate Thickness, l (Free Space)

FIGURE 6-8 Surface-wave loss of microstrip patches versus substrate thickness for common substrate dielectric constants.

WT is the energy stored in the patch and the surface wave and ω = 2πf , the radian frequency. Equation (6-3) can be expressed in terms of Q: ηSW =

Q Qrad = QR QR

296

MICROSTRIP ANTENNAS

The surface wave is not a dissipation loss, but potentially an uncontrolled radiation. Dielectric and conductor losses increase the impedance bandwidth of the patch, but reduce its gain. We express these losses as Q to compute patch efﬁciency. Given the dielectric loss tangent, tan δ, and the patch conductivity σ , we have two more Q terms that reduce the overall Q of the patch in terms of impedance bandwidth: Qd =

1 tan δ

and Qc = h πf µ0 σ

(6-9)

The total quality factor QT is found from the sum of the inverses: 1 1 1 1 1 = + + + QT QR QSW Qd Qc

(6-10)

If we attempt to fabricate a patch on a thin substrate, Qd and Qc become commensurate with the radiation Qrad and efﬁciency suffers. The impedance bandwidth increases due to the dissipation in the microstrip patch. Figure 6-7 does not include these losses. Dielectric Slab Surface Wave We consider the dielectric slab surface wave because it can be excited not only by a microstrip patch but by any wave that passes across it. The slab binds a portion of the wave and releases it when it diffracts at its edges. The surface-wave device slows the wave velocity of this wave relative to the spacewave signal, and when it radiates from the edges it no longer adds in phase with the space wave. The surface-wave ﬁelds decrease exponentially in the direction normal to the surface, and the exponential rate increases as the binding increases and the wave propagates more slowly. A dielectric slab on a ground plane will support TM modes when thin and TE modes when thick. The TM mode is polarized normal to the slab surface, whereas the TE mode is polarized parallel to the slab surface. A TM mode requires an inductive surface such as a corrugated ground plane to bind the wave. While corrugations prevent propagation between the slots, the wave propagates in the dielectric slab by bouncing between the two interfaces at an angle with respect to the surfaces. The second surface can be either free space or a conductor. To solve for the ﬁelds, we equate not only the wave impedance at the boundary but the propagation constants in the two regions as well. We deduce the grounded dielectric slab solution from a slab twice as thick in free space that has an odd-mode electric ﬁeld excitation on the slab sides. The center becomes a virtual short circuit for the odd-mode excitation. We divide the space around the slab into three regions: 1 above the slab, 2 in the slab, and 3 below the slab and then derive the ﬁelds from potential functions [6, p. 129]: −2πbx ψ1 = A1 exp exp(−j kz z) λ 2πpx x sin λ ψ2 = A2 (6-11) exp(−j kz z) cos 2πpx x λ 2πbx ψ3 = ±A1 exp exp(−j kz z) λ

MICROSTRIP PATCH BANDWIDTH AND SURFACE-WAVE EFFICIENCY

297

where the sign of ψ3 depends on satisfying continuous tangential ﬁelds across the lower slab boundary. The center of the coordinate normal to the slab (x) is the slab center. Equating the propagation constants and x-directed wave impedances produces transcendental equations in the transverse propagation constant in the slab px : πpx a tan 2 2 ωa πpx a πpx a λ (6-12) (ε1 µ1 − ε0 µ0 ) − = ±B0 2 λ λ cot πpx a λ where B0 = µ0 /µ1 for TE waves. B0 = ε0 /ε1 for TM waves, ω is the radian frequency (2πf ), a is the slab thickness, and ε1 and µ1 are the permittivity and permeability of the slab. We solve for px [Eq. (6-12)] numerically or graphically and use πp 2 ωa 2 πb x (ε1 µ1 − ε0 µ0 ) − (6-13) a= λ 2 λ to determine attenuation constant b and the relative propagation constant P of the slab surface wave: kz = P k = k 1 + b2 or P = 1 + b2 (6-14) For the TM0 mode we can use an approximate expression for P instead of solving Eq. (6-12) numerically when the slab is thin [7]: P2 = 1 +

(εr µr − 1)2 (2 ka)2 (εr µr )2

(6-15)

Equation (6-12) has an inﬁnite number of solutions, corresponding to the multiple values of the tangent and cotangent functions. Order 0 corresponds to the tangent function from 0 to 90◦ ; order 1 corresponds to the cotangent function from 90 to 180◦ ; and so on. Even-mode orders use the tangent function, and odd-mode orders use the cotangent function. We deﬁne the cutoff frequency as the point where α = 0, the transition between attached and detached waves: 2a ε1 µ1 λc = −1 (6-16) n ε0 µ 0 The cutoff frequency for the zeroth-order mode is zero. Only the TM0 mode has odd symmetry, required for the grounded slab. The grounded slab supports even-order TM modes and odd-order TE modes. Equation (6-12) coupled to Eq. (6-13) has been solved numerically to generate Tables 6-2 and 6-3. Table 6-4 lists the thicknesses of a slab in free space supporting the TM0 mode for a given P . The grounded slab is one-half the thickness of the values in Table 6-2. Similarly, Table 6-3 lists the thicknesses for the TE1 mode. Equation (6-16) can be solved for the minimum thickness to support the TE1 mode. Below that thickness the waves do not bind to the surface. Besides microstrip patches, we feed these surfaces from either a small horn or a parallel-plate transmission line. We match the feed polarization to the mode on the slab, but the slab binds only part of the power. The rest radiates directly from the feed or reﬂects to the feed input. We can feed an ungrounded slab by centering it in a waveguide. The TE10 waveguide mode excites the TE0 slab mode when the mode velocity determining thickness is in the H -plane. Like the grounded slab with the TM0

298

MICROSTRIP ANTENNAS

TABLE 6-2 Thickness (λ0 ) of a Dielectric Slab Supporting a TM0 Modea Dielectric Constant P 1.001 1.002 1.005 1.01 1.02 1.04 1.06 1.08 1.10 1.12 1.14 1.16 1.18 1.20 1.25 1.30 1.35 1.40 a

2.21

2.94

4.50

6.00

9.80

0.02699 0.03672 0.05792 0.08162 0.1147 0.1607 0.1956 0.2253 0.2520 0.2770 0.3012 0.3251 0.3493 0.3741 0.4426 0.5282 0.6492

0.02152 0.03041 0.04784 0.06713 0.09355 0.1289 0.1545 0.1752 0.1930 0.2088 0.2233 0.2369 0.2499 0.2625 0.2934 0.3250 0.3593 0.3986

0.01831 0.02574 0.04032 0.05623 0.07746 0.1046 0.1231 0.1374 0.1491 0.1590 0.1677 0.1756 0.1827 0.1894 0.2045 0.2182 0.2314 0.2444

0.01839 0.02420 0.03744 0.05195 0.07094 0.09444 0.1099 0.1215 0.1307 0.1384 0.1450 0.1508 0.1560 0.1607 0.1712 0.1803 0.1887 0.1966

0.03446 0.04710 0.06316 0.08180 0.09331 0.1015 0.1078 0.1129 0.1171 0.1208 0.1240 0.1269 0.1329 0.1380 0.1424 0.1463

Use half-thickness for a slab on a ground plane.

TABLE 6-3 Thickness (λ0 ) of a Dielectric Slab Supporting a TE1 Modea Dielectric Constant P 1.001 1.002 1.005 1.01 1.02 1.04 1.06 1.08 1.10 1.15 1.20 1.25 1.30 a

2.21

2.94

4.50

6.00

9.80

0.4469 0.4720 0.4829 0.4961 0.5164 0.5494 0.5790 0.6078 0.6368 0.7140 0.8046 0.9182 1.0712

0.3689 0.3709 0.3765 0.3843 0.3962 0.4150 0.4313 0.4465 0.4613 0.4982 0.5372 0.5802 0.6291

0.2743 0.2774 0.2770 0.2810 0.2873 0.2968 0.3049 0.3122 0.3191 0.3356 0.3518 0.3683 0.3856

0.2260 0.2272 0.2302 0.2330 0.2373 0.2438 0.2492 0.2540 0.2585 0.2690 0.2790 0.2890 0.2992

0.1701 0.1705 0.1717 0.1736 0.1761 0.1797 0.1825 0.1851 0.1874 0.1928 0.1978 0.2026 0.2073

Use half-thickness for a slab on a ground plane.

mode, the TE0 mode has no cutoff frequency for a free-space slab. Table 6-4 lists the slab thicknesses for a given relative propagation constant for the TE0 mode. The surface-wave power was found in terms of the relative propagation constant P [7]:

PSW = n2

15πk 2 n2 µ3r (P 2 − 1) √ n4 (P 2 − 1) P2 − 1 1 + kh 1 + + 2 √ n − P2 n2 − P 2 P2 − 1

(6-17)

RECTANGULAR MICROSTRIP PATCH ANTENNA

TABLE 6-4

299

Thickness (λ0 ) of a Dielectric Slab Supporting a TE0 Mode Dielectric Constant

P 1.001 1.002 1.005 1.01 1.02 1.04 1.06 1.08 1.10 1.15 1.20 1.25

2.21

2.94

4.50

6.00

9.80

0.01274 0.01684 0.02649 0.03772 0.05409 0.07872 0.09935 0.1184 0.1368 0.1833 0.2348 0.2968

0.00994 0.01174 0.01666 0.02341 0.03345 0.04823 0.06027 0.07104 0.08113 0.1051 0.1290 0.1542

0.00661 0.00920 0.01289 0.01846 0.02640 0.03275 0.03832 0.04343 0.05507 0.06595 0.07661

0.00410 0.00638 0.00905 0.01286 0.01839 0.02276 0.02656 0.03002 0.03779 0.04489 0.05168

0.00250 0.00364 0.00514 0.00729 0.01040 0.01284 0.01494 0.01684 0.02106 0.02483 0.02835

We combine Eq. (6-1) for the space-wave power with Eq. (6-17) for the surface-wave power to calculate efﬁciency in the same manner as Eq. (6-3). The results are similar. 6-3 RECTANGULAR MICROSTRIP PATCH ANTENNA Although design equations will be given below for single-layer rectangular and circular patches, serious design work should use one of the excellent available commercial design codes [8]. Their use reduces the need to modify the ﬁnal dimensions using a knife to remove metal or metal tape to increase the patches. Antennas can be built with tuning tabs, but the labor to trim these increases cost. Tuning tabs are unsuitable for arrays when the input port to individual antennas cannot be accessed. As we add layers to increase bandwidth, a cut-and-try method becomes extremely difﬁcult, and numerical methods are a necessity. Rectangular patch antennas can be designed by using a transmission-line model [9] suitable for moderate bandwidth antennas. Patches with bandwidths of less than 1% or greater than 4% require a cavity analysis for accurate results, but the transmissionline model covers most designs. The lowest-order mode, TM10 , resonates when the effective length across the patch is a half-wavelength. Figure 6-9 demonstrates the patch fed below from a coax along the resonant length. Radiation occurs from the fringing ﬁelds. These ﬁelds extend the effective open circuit (magnetic wall) beyond the edge. The extension is given by [10] εeff + 0.300 W/H + 0.262 = 0.412 H εeff − 0.258 W/H + 0.813

(6-18)

where H is the substrate thickness, W the patch nonresonant width, and εeff the effective dielectric constant of a microstrip transmission line the same width as the patch. A suitable approximation for εeff is given by [11] εeff

εr + 1 εr − 1 10H −1/2 + = 1+ 2 2 W

(6-19)

300

MICROSTRIP ANTENNAS

l 2√ eff ∋

Feed E

X

E

W

A

Polarization

A

Patch Shorting Pin

L ∆

Center Pin Soldered to Path E

AA Ground Soldered Coax

FIGURE 6-9 Coax-fed microstrip patch antenna.

where εr is the substrate dielectric constant. The transmission-line model represents the patch as a low-impedance microstrip line whose width determines the impedance and effective dielectric constant. A combination of parallel-plate radiation conductance and capacitive susceptance loads both radiating edges of the patch. Harrington [6, p. 183] gives the radiation conductance for a parallel-plate radiator as πW (kH )2 G= 1− ηλ0 24

(6-20)

where λ0 is the free-space wavelength. The capacitive susceptance relates to the effective strip extension: W B = 0.01668 (6-21) εeff H λ Example Design a square microstrip patch antenna at 3 GHz on a 1.6-mm substrate with a dielectric constant of 2.55 (woven Teﬂon ﬁberglass). The patch will be approximately a half-wavelength long in the dielectric.

RECTANGULAR MICROSTRIP PATCH ANTENNA

301

Assume at ﬁrst that the width is λ/2. W =

2f

c √

εr

= 31.3 mm

By Eq. (6-19), εeff = 2.405. On substituting that value into Eq. (6-18), we obtain the effective cutback on each edge; = 0.81 mm. The resonant length is L=

2f

c √

εeff

− 2 = 30.62 mm

When we use this length as the width (square patch) to calculate the effective dielectric constant, we obtain 2.403, very close to the initial value. We can iterate it once more and obtain 30.64 mm for the resonant length. The input conductance of the patch fed on the edge will be twice the conductance of one of the edge slots [Eq. (6-20)]: 30.64 mm [2π(1.6)/100]2 G= 1− = 2.55 mS 120(100 mm) 24 R=

1 = 196 2G

A microstrip feeding line can be attached to the center of one of the radiating edges but 50- transmission lines become inconveniently wide on low-dielectric-constant substrates. More convenient, 100- narrower lines have about the same low loss and are generally used in feed networks. To transform the 196- input resistance of the example above to 100 , we use a 140- quarter-wavelength transformer. The bandwidth of the transformer far exceeds that of the antenna. In the example above, we have a square patch. Why doesn’t the antenna radiate from the other two edges? We can equally well say that the patch is a transmission line in the other direction. The equal distances from the feed point to the nonradiating edges produce equal ﬁelds from the patch to ground. Equal ﬁelds on the edges set up a magnetic wall (virtual open circuit) through the centered feed line and create a poor impedance match to the feed. We expand the radiating edge ﬁelds in an odd mode, since the power traveling across the patch loses 180◦ of phase. The odd mode places a virtual short circuit halfway through the patch. A shorting pin through the center (Figure 6-9) has no effect on radiation or impedance, but it allows a low-frequency grounding of the antenna. The patch can be fed by a coax line from underneath (Figure 6-9). The impedance varies from zero in the center to the edge resistance approximately as Ri = Re sin2

πx L

0≤x≤

L 2

(6-22)

where Ri is the input resistance, Re the input resistance at the edge, and x the distance from the patch center. The feed location does not signiﬁcantly affect the resonant frequency. By using Eq. (6-22), we locate the feed point given the desired input impedance: L −1 Ri x = sin (6-23) π Re

302

MICROSTRIP ANTENNAS

Compute the 50- feed point in the example above: 30.64 −1 sin x= π

50 = 5.16 mm 196

The feed pin currents add to the pattern by radiating a monopole pattern. Figure 6-10 shows this radiation for a patch using a free-space substrate where the E-plane radiating edges are spaced λ/2. The pattern of Figure 6-10 has a null along the ground plane in the E-plane, but the monopole radiation increases the radiation along the ground plane. On one side the radiation adds and on the other it subtracts from the E-plane pattern to form a null tilted above the ground plane. The H -plane pattern now contains crosspolarization. We can reduce the monopole radiation by feeding the patch at a second port located an equal distance from the center on the opposite side. This requires an external feed network that divides the power equally between the two ports with a 180◦ phase difference. The problem with this feed arrangement is that signiﬁcant power is coupled between the two feeds in the equivalent microwave circuit of the patch. The estimated value of −6 dB coupling between the ports causes a portion of the input power to be dissipated in the second port. At this level the patch efﬁciency drops 1.25 dB. We can reduce the monopole radiation by coupling to a second short-circuited probe to the patch instead of directly feeding it. The gap between the second probe and the patch is adjusted until the antenna radiates minimum cross-polarization in the H -plane. This uses the microstrip patch as the feed network, and the second probe has no resistive load to dissipate power. The feed probe across the microstrip patch substrate is a series inductor at the input. Higher-order modes excited in the patch by this feeding method add to the inductive component of the antenna. Below resonance, the antenna is inductive and has nearzero resistance. As the frequency increases, the inductance and resistance grow as the parallel resonance is approached. Above the resonant frequency, the antenna is capacitive as the impedance sweeps clockwise around the Smith chart (Figure 6-11) and ﬁnally back to a slight inductive component near a short circuit. Increasing the

E-Plane H-Plane, X-Pol H-Plane

FIGURE 6-10 Pattern of coax-fed, microstrip patch including feed pin radiation for free-space substrate.

RECTANGULAR MICROSTRIP PATCH ANTENNA

303

OverCoupled UnderCoupled Critically Coupled

FIGURE 6-11 Smith chart frequency response of under-, critically, and overcoupled patches as the feed point moves toward one radiating edge of a rectangular patch.

input resistance by changing the feed point causes the resonant frequency response circle to grow on the Smith chart and cross the resistance line at a higher level. We call the left-hand curve the undercoupled case because the sweep of the curve fails to enclose the center of the chart. The center curve is critically coupled and the right curve is the overcoupled case. This general impedance response also holds for circular patches. We use these terms for all resonance curves when they sweep around or toward the Smith chart center from any peripheral point. Figure 6-12 plots the Smith chart for a design with a patch on a 0.05λ-thick substrate with dielectric constant 1.1 that includes the inductance of the feed pin. The response locus lies above the real axis and is always inductive. We can tune this impedance locus by adding a series capacitor at the input with a reactance −j 50 at the center frequency. The series capacitor moves the locus down until it sweeps around the center of the chart in an overcoupled response. Figure 6-13 shows implementation of the capacitor as a disk on the end of the feed pin. The pin passes through a hole in the patch so that the only connection is through the capacitor disk. The disk can be placed below the patch on a separate substrate in a multiple-layer construction. Other conﬁgurations use an annular ring capacitor etching in the patch at the feed point for small capacitors. Adding to this a series inductor and adjusting the series capacitor improves the impedance match over a larger frequency range, as shown in Figure 6-14, where the locus encircles the origin [12]. The patch with the single added

304

MICROSTRIP ANTENNAS

W/Pin Inductance

Added Series Capacitor

FIGURE 6-12 Impedance improvement by adding a series capacitor to a patch on a thick substrate. Patch

Disk Capacitor Feed Pin

Coax Input

FIGURE 6-13 Cross section of a probe-fed patch with an added series capacitor.

series capacitor has a 9.1% 10-dB return-loss bandwidth while adjusting the series capacitor, and adding a series inductor increases the impedance bandwidth to 15.4%. Matching networks have limited ability to add resonances to broadband the impedance match, but construction becomes difﬁcult. Later, we will obtain extra resonances by adding antenna elements. We can feed patches from the edge by using an inset microstrip line as shown in Figure 6-15, where the gap on either side of the microstrip line equals its width. A FDTD analysis shows that the inset disturbs the transmission line or cavity model

305

RECTANGULAR MICROSTRIP PATCH ANTENNA

w/Pin Inductance

Series Capacitor Series Inductor Matching Network

FIGURE 6-14 Impedance response of a patch with a two-element matching network.

Substrate

Patch Feed Inset

FIGURE 6-15

Inset-fed square patch.

and increases the impedance variation with distance compared to a coaxial probe feed given a patch resonant length L and feed position x from the center [13]: Ri = Re sin4

πx L

0≤x≤

L 2

(6-24)

306

MICROSTRIP ANTENNAS

Equation (6-24) is an approximate solution because at x = 0, the resistance remains ﬁnite. We locate the feed from the equation using a radian angle measure: L −1 Ri 1/4 x = sin (6-25) π Re Compute the 50- feed point in the example above: 30.64 −1 50 0.25 x= = 7.71 mm sin π 196 The inset distance (7.3 mm) is less than the distance of the probe (9.8 mm) from the edge. Aperture Feed [14,15] A microstrip patch is a planar resonant cavity with opencircuited sidewalls that leak power in radiation. We can also think of the rectangular patch operating in the lowest-order mode as a low-impedance transmission line with end susceptance and radiation conductance. Both models predict a resonant structure with signiﬁcant Q. Resonant cavities are readily excited by coupling to a transmission line through an aperture or by direct feeding from a transmission line. The Q of the resonant cavity limits the excitation ﬁelds to one of the modes. We can expand the excitation in the cavity modes, but the lowest-order mode is usually the most signiﬁcant and contains most of the stored energy. We generally consider the voltage distribution in a patch with its null plane located halfway across the patch through the center. Whether we consider it as a cavity or a transmission line the standing-wave voltage has a standing-wave current associated with it. This current is out of phase with the voltage and its peak occurs along the virtual short circuit through the centerline. Along the resonant length the current has a cosinusoidal distribution that vanishes at the radiating edges in a single half-cycle for the lowest-order mode. The current has a uniform distribution along the patch width. We produce maximum coupling to a patch through a slot by distorting the currents in the ground plane of the patch where they are maximum in the center of the patch. To ﬁrst order the currents ﬂow along the resonant length. This means that we align the slot perpendicular to the current ﬂow for maximum excitation in the same manner as slots in waveguides (Section 5-24). To excite the slot we pass a microstrip transmission line across it perpendicularly. This leads to a three-layer structure. The patch is located on the top layer. Its ground plane contains a coupling aperture usually placed under the center of the patch for maximum coupling. The third layer contains a microstrip transmission using the same ground plane as the patch and located under the center of the slot to maximize coupling. Figure 6-16a shows an exploded view of the patch, ground with its aperture, and the microstrip transmission line ﬂipped over relative to the patch. Figure 6-16b gives the general parameters associated with the slot aperture. Although xos and yos are usually zero to maximize coupling, the patch current distribution tells us how the coupling varies with slot location. Because the current in the ground plane is uniform across the patch width W , coupling is independent of xos until the slot starts to overlap the edge of the patch. The cosinusoidal distribution current distribution along the resonant length direction L means that the coupling falls off slowly as yos is moved off zero. The sign of yos does not matter because the distribution is an even function. The slow variation of current near the patch center means that the slot location has a loose tolerance.

RECTANGULAR MICROSTRIP PATCH ANTENNA

b ε r

307

db

Ground Plane with Aperture

a ε r

da

(a ) Lp

Wap

Wf

y Xos Wp

Ls

Lap

yos

x ( b)

FIGURE 6-16 Aperture feed of square patch. (From [15], Fig. 1, 1986 IEEE.)

The microstrip transmission line excites the slot (aperture) from a standing-wave with its maximum current located at the slot. We maximize the standing-wave current by either using a shorting via from the microstrip line to the ground plane or by using a quarter-wave open-circuited transmission line stub of length Ls . Ls will be less than a quarter-wave in the effective dielectric constant of the microstrip line because the open-circuit end has fringing capacitance and its capacitance must overcome the higherorder modes of the microstrip patch, which load the input inductively. The reactance of the stub, a series load to the input, is given by the equation ZS = −j Z0 cot(keff LS ) where Z0 is the characteristic impedance of the microstrip feed line, keff the effective propagation constant in the microstrip substrate, and Ls the stub length ≈0.22λeff . We increase the coupling to the patch resonant cavity by increasing the aperture size. Figure 6-17 shows the Smith chart variation with aperture size as the coupling

308

MICROSTRIP ANTENNAS

e

rtur

g asin

Ape

e

Incr

FIGURE 6-17 Effect of aperture size on coupling to a patch where larger openings move the response to the right.

varies left to right as undercoupled, critically coupled, and overcoupled. When we increase the bandwidth, we lower the Q, and the coupling aperture size must increase. Waterhouse [8] suggests starting with a slot about one-half the patch width and using a commercial code to analyze the response while adjusting dimensions before fabrication. We control the rotational position on the Smith chart by varying the open-circuited stub length. Shorter lengths, below λ/4, increase the capacitive reactance and the coupling loop will rotate around a constant-resistance circle with its diameter determined by the aperture size, as shown in Figure 6-18. Figure 6-19 gives aperture shapes in order of increasing coupling. The longer slot of (b) compared to slot (a) increases coupling. Widening the aperture as in (c) increases coupling relative to (a). The H-shaped slot has a more uniform distribution along the horizontal slot and increased coupling. The bowtie and hourglass apertures increase coupling from a consideration of increased path length around the opening. The smooth curve of the hourglass reduces current discontinuities at the edges and increases coupling [16, pp. 158–159]. Aperture feeding eliminates the vertical pin structure in the microstrip patch and eases construction but at the cost of a multiple-layer etching. The elimination of the vertical pin removes the added monopole pattern, which increases cross-polarization. When the patch is edge fed, whether directly or inset, the substrate for good patch radiation does not match the one needed for good microstrip lines. With an aperturefed patch, each structure can use its optimum substrate, because they are independent and connected only through the aperture. As we try to feed broadband patches, the

RECTANGULAR MICROSTRIP PATCH ANTENNA

309

FIGURE 6-18 Effect of varying length of an open-circuited stub in an aperture-fed patch when critically coupled.

0.55

0.15 1.1

1.4

(a)

(b)

1.1 (c) 0.55

0.15 1.1 (d )

1.1 (e)

1.1 ( f)

FIGURE 6-19 Aperture shapes to increase coupling and bandwidth. (From [16], Fig. 4-29, 2003 Artech House, Inc.)

Q decreases and the aperture size grows. This slot, although below a resonant size, increases its radiation and decreases the front-to-back ratio because it radiates equally on both sides. One solution is to enclose the microstrip line in a box to prevent slot radiation on the back side. If we use a high-dielectric-constant substrate for the microstrip line, the coupling through the aperture remains high, but the second ground plane of the microstrip will reduce the coupling. The slot aperture adds a pole to the patch circuit that can be used to broadband the impedance response. To use this pole effectively, we must increase the aperture size until it becomes a signiﬁcant radiator.

310

MICROSTRIP ANTENNAS

6-4 QUARTER-WAVE PATCH ANTENNA When operation is in the lowest mode, a virtual short circuit forms through a plane centered between the two radiating edges. We can make an antenna by using half the patch and supplying the short circuit (Figure 6-20). The E-plane pattern broadens to that of a single slot. The resonant length is about a quarter-wavelength in the dielectric of the substrate. We use the effective dielectric constant εeff of a microstrip line of patch width W and given by Eq. (6-18) to determine the resonant length L of the quarter-wave patch: L λ − (6-26) = √ 2 4 εeff We can implement the short circuit with a series of pins or etched vias between the ground plane and the patch. These add an inductive component to the transmission-line model of the antenna. The effective shorting plane occurs further along the transmission line. The equivalent extra length l is found from the parallel-plate circuit model of a row of evenly spaced pins [17, p. 104]. Given the pin center spacing S, their radius √ r, and the wavelength in the dielectric λd = λ0 / εr , we compute the patch-length reduction from the equation 2 S S S 2πr 2 l = ln (6-27) + 0.601 − 2π 2πr S λd We have only the conductance and susceptance of a single edge that doubles the resonant resistance at the edge as compared with the full patch. It becomes difﬁcult to feed the antenna from microstrip because this raises the quarter-wavelength transformer impedance and requires narrower lines. We can increase the edge width to reduce the edge input resistance, but the antenna is usually fed from underneath. Equation (6-23) gives the approximate feed location measured from the short circuit. The resonant frequency shifts slightly as the feed point moves. During tuning for impedance match, the length of the cavity will have to be adjusted to maintain the desired resonant frequency. Quarter-wave and full-patch antennas have the same Q. A half-patch antenna has half Input Coax l 4√ eff ∋

∆

Feed Pin

L/2

E E E

Substrate

FIGURE 6-20 Quarter-wave patch.

Shorting wall

QUARTER-WAVE PATCH ANTENNA

311

the radiation conductance but only half the stored energy of a full-patch antenna. Its bandwidth is approximately the same as that of the full patch. Example Design a half-patch antenna at 5 GHz on a 0.8-mm-thick substrate (εr = 2.21) with a radiating width of 0.75λ. The edge width is 0.75(300 mm)/5 = 45 mm. By using Eq. (6-19), we compute the effective dielectric constant in the cavity: εeff = 2.16. Equation (6-18) gives us the cutback for fringing ﬁelds: = 0.42 mm. The resonant length becomes L λ − = 10.20 − 0.42 = 9.78 mm = √ 2 4 εeff The radiation conductance from the single edge is [Eq. (6-20)] G=

45 = 6.25 mS 120(60)

or

R = 160

The 50- feed point is found from Eq. (6-23): 19.56 −1 x= sin π

50 = 3.69 mm 160

where x is the distance from the short. The short circuit of this antenna is quite critical. The low impedance of the microstrip cavity raises the currents in the short circuit. Without a good low-impedance short, the antenna will detune and have spurious radiation. If the antenna is made from a machined cavity, careful attention must be paid to the junction between the top plate and the cavity to assure good electrical contact. Figure 6-21 shows the calculated pattern of a quarter-wave patch on a free-space substrate 0.04λ thick on an inﬁnite ground plane. The antenna radiates primarily from the single edge located opposite the short-circuited edge. A vertical probe feeds the

E-Plane

H-Plane X-Pol.

H-Plane

FIGURE 6-21 Pattern of a quarter-wave patch on a free-space substrate.

312

MICROSTRIP ANTENNAS

E-Plane H-Plane X-Pol.

H-Plane

(a)

H-Plane X-Pol E-Plane

H-Plane

(b)

FIGURE 6-22 Pattern of a quarter-wave patch mounted on (a) 2λ- and (b) 10λ-diameter ground planes.

CIRCULAR MICROSTRIP PATCH

313

antenna directly. The E-plane has a broad, nearly constant pattern. Radiation from current on the probe and shorting pins adds to positive angle radiation of the equivalent single slot while subtracting in the opposite direction. The H -plane pattern (dashed curve) retains its null along the ground plane. The light-line curve gives the cross-polarization in the H -plane. The feed probe and shorting pin currents produce a pattern similar to that from a monopole. The model that uses equivalent magnetic currents fails to predict the high radiation from these currents. When the quarter-wave patch is mounted on a ﬁnite ground plane, it exhibits behavior similar to that of a monopole. Figure 6-22a,b plots the pattern when it is mounted on 2λ- and 10λ-diameter ground planes. These show a monopole-type pattern, where radiation spreads readily behind the ground plane. Currents ﬂowing in the feed pin and shorting wall distort the E-plane and cause asymmetry. The magnetic currents ﬂowing along the side slots no longer cancel as in the square patch and increase cross-polarization. If we close off the nonradiating edges with metal walls, the walls convert the parallel-plate line into a waveguide and we use the waveguide propagation constant to calculate the quarter-wavelength cavity depth. The slot ﬁelds vanish on the ends and establish a sinusoidal slot distribution. We can offset the feed toward both the back wall and the sidewall to reduce the input impedance. The peak voltage (minimum current and peak resistance) occurs at the slot center. Figure 6-23 illustrates the pattern of the waveguide quarter-wave patch on an inﬁnite ground plane. The sidewalls reduce the monopole radiation, and the H -plane cross-polarization is reduced compared to a quarter-wave patch. When mounted on a 2λ-diameter disk, centered on the feed pin, the pattern (Figure 6-24) exhibits lower-level radiation in the backlobe because the monopole pattern has been reduced. The high radiation level at the disk edges still causes considerable edge diffraction in the E-plane. 6-5 CIRCULAR MICROSTRIP PATCH In some applications, a circular patch ﬁts in the available space better than a rectangular one. In a triangularly spaced array, they maintain a more uniform element

E-Plane H-Plane X-Pol

H-Plane

FIGURE 6-23 Pattern of a quarter-wave waveguide patch.

314

MICROSTRIP ANTENNAS

H-Plane E-Plane

H-Plane X-Pol

FIGURE 6-24 Quarter-wave waveguide patch mounted on a 2λ-diameter disk.

environment. No suitable transmission-line model presents itself, and the cavity model must determine the resonant frequency and bandwidth. The cutoff frequencies of TE modes of circular waveguides give the resonant frequencies of circular patch antennas. The patch with its magnetic walls and TM modes is the dual of the waveguide. The resonant frequencies are given by fnp =

Xnp c √ 2πaeff εr

(6-28)

where Xnp are the zeros of the derivative of the Bessel function Jn (x) of order n, as is true of TE-mode circular waveguides. The term aeff is an effective radius of the patch [18]: 2H πa ln + 1.7726 (6-29) aeff = a 1 + πaεr 2H where a is the physical radius and H is the substrate thickness. Using the effective radius gives the resonant frequency within 2.5%. We combine Eqs. (6-28) and (6-29) to determine radius to give a particular resonant frequency: Xnp c (6-30) aeff = √ 2πfnp er

CIRCULAR MICROSTRIP PATCH

315

Since a and aeff are nearly the same, we can iterate Eq. (6-29) to compute a, the physical radius [19, p. 119]: a=

aeff 1 + 2H /πaεr ln (πa/2H ) + 1.7726

(6-31)

We start by using aeff for a in Eq. (6-31), which converges rapidly. The lowest-order mode, TM11 , uses X11 (1.84118) and produces a linearly polarized ﬁeld similar to a square patch. The TM01 mode (X01 = 3.83171) produces a monopole-type pattern from a uniform edge fringing ﬁeld. Example Design a circular microstrip patch antenna (TM11 mode) at 3 GHz on a 1.6-mm substrate that has a dielectric constant of 2.55 (woven Teﬂon ﬁberglass). We calculate the effective radius from Eq. (6-30): aeff =

1.84118(300 × 109 mm/s) = 18.35 mm √ 2π(3 × 109 Hz) 2.55

The physical radius will be slightly less. By using aeff in the denominator of Eq. (631), we obtain a physical radius: a = 17.48 mm. We can then substitute this back into Eq. (6-31) and obtain a = 17.45 mm. Equation (6-31) converges in two iterations to a reasonable tolerance, since another iteration gives the same value. Actually, a single iteration gives the value within 0.2% on a formula accurate to only 2.5%. The ﬁelds of the TM11 mode produce a virtual short circuit at the center of the patch. We can reinforce the short circuit with a pin soldered between the patch and ground. The radial line along which the feed is placed determines the direction of the linear polarization. The nonuniform radiation along its edge gives a larger edge impedance than the square patch. Experience shows that the 50- feed point is located from the center at about one-third the radius. Experiments, actual or numerical, will be required to locate the proper point. Use a network analyzer with a Smith chart display to measure the input impedance. If the resonance circle swings around the origin, the impedance is too high (overcoupled). Move the feed toward the center. A scalar return-loss display cannot give you the direction of movement required. Like the rectangular patch, mismatching the impedance at center frequency to about 65 increases the bandwidth slightly. Derneryd [20] gives an approximate expression for the radial impedance variation: Rin = Re

J12 (kε ρ) J12 (kε a)

(6-32)

where Re is the edge resistance, ρ the radial distance, and J1 the Bessel function of the √ ﬁrst kind. kε is the propagation constant in the substrate dielectric constant: kε = k εr . Figure 6-25 gives the 2 : 1 VSWR bandwidth of a circular patch on various substrates as a function of the substrate thickness. It has a slightly smaller bandwidth than that of a square patch because it has a smaller volume. The curves on Figure 6-25 include surface-wave radiation (or losses).

316

MICROSTRIP ANTENNAS

2:1 VSWR Bandwidth, %

2.94

6.0

2.2

9.8

4.5 1.0

Substrate Thickness, l (Free Space)

FIGURE 6-25 2 : 1 VSWR bandwidth of circular microstrip patches versus substrate thickness in free-space wavelengths, including surface-wave radiation.

6-6 CIRCULARLY POLARIZED PATCH ANTENNAS Figure 6-26 show methods of achieving circular polarization with square patches fed with two inputs. The patches are fed by equal signals 90◦ out of phase. The branchline hybrid (Figure 6-26a) consists of four transmission lines connected in a square. The hybrid shown (100- system) produces equal outputs 90◦ out of phase at center frequency. The two inputs produce patterns with opposite senses of circular polarization. Both the VSWR and axial ratio bandwidths far exceed the singly fed patch bandwidth. Reﬂections due to the patch mismatch are routed to the opposite input. Patch input reﬂections, undetected at the input, reduce the efﬁciency of the antenna by the same amount as the singly fed patch mismatches. The antenna can be fed from below in two places by using a coupled line hybrid, but it suffers from the same efﬁciency problem. The cross-fed antenna (Figure 6-26b) splits the signal to feed both edges. A quarterwavelength-longer line provides the extra 90◦ phase shift to give circular polarization. Shifting the impedance from one input through a quarter-wavelength line before adding the two in shunt cancels some of the reﬂection from the second line and increases the impedance bandwidth. The impedance bandwidth approximately doubles compared to the singly fed patch. The 6-dB axial ratio bandwidth roughly equals the singly fed square-patch bandwidth. The polarization loss (0.5 dB) of a 6-dB axial ratio equals the 2 : 1 VSWR mismatch loss. The antennas in Figure 6-27 use asymmetries to perturb the resonance frequencies of two possible modes and achieve circular polarization [21]. The approximately square patches have been divided into two groups: type A, fed along the centerline, and type B, fed along the diagonal. All these antennas radiate RHC. We can understand the operation of these patches from an analysis of the turnstile dipole antenna (Figure 6-28). The orthogonal dipoles could be of equal length and fed from a 90◦ hybrid to achieve

CIRCULARLY POLARIZED PATCH ANTENNAS

l/4

317

100 Ω

RHC 100 Ω

l/4

LHC 70.7 Ω

(a)

−90°

LHC

∆ ∆ + l/4

0° Cross-fed patch

l/4

(b)

FIGURE 6-26 Dual-fed circularly polarized patch antennas: (a) branchline hybrid fed; (b) cross-fed patch.

FIGURE 6-27 Classes of perturbed microstrip patches to generate circular polarization from a single feed. (From R. Garg et al., Microstrip Patch Handbook, Fig. 8-15, 1999 Artech House, Inc.)

circular polarization (like the patch in Figure 6-26a). Instead, the lengths are changed to shift the phase of each dipole by 45◦ at resonance. If we lengthen the dipole beyond resonance, the input impedance becomes inductive. The current becomes I=

V V (R2 − j X2 ) = R2 + j X2 R22 + X22

318

MICROSTRIP ANTENNAS

FIGURE 6-28 Turnstile dipole antenna.

The radiated ﬁeld phase decreases relative to the resonant-length dipole. Shortening the dipole from resonance increases the far-ﬁeld phase. We adjust the lengths until the phase difference of the radiated ﬁelds is 90◦ and the susceptances from the two dipoles cancel at center frequency. The combination of the two modes produces a Smith chart response with a small loop or kink (see Figure 5-13). The best circular polarization occurs at the frequency of the kink, and the response degrades below and above this frequency. The axial ratio bandwidth is far less than the impedance bandwidth, because the combination of the two modes causes a cancellation of transmission-line reﬂections from the two modes and increases the impedance bandwidth. The phase required for good circular polarization changes rapidly. We denote the total change in area S to achieve two resonances for a normal patch area of S and it is proportional to the Q. A type A patch, fed along the square patch axis, requires less area change than a type B patch, fed along the diagonal: type A:

1 S = S 2Q

type B:

S 1 = S Q

(6-33a,b)

We achieve the same effect with a patch by perturbing the lengths of a square patch and feeding both polarizations. An input along the diagonal (type B) feeds all edges in two separate resonances. The ratio of the edge lengths is found in terms of Q by a perturbation technique [4]. We rearrange Eq. (6-33b) to derive the ratio of these lengths: b 1 =1+ (6-34) a Q We calculate resonant frequencies for the two lengths from Eq. (6-34): f0 1 f2 = f0 1 + f1 = √ Q 1 + 1/Q

(6-35)

COMPACT PATCHES

319

Q is related to the VSWR bandwidth by Eq. (6-7). The 3-dB axial ratio bandwidth of the antenna is limited to 35%/Q or 35% of the frequency difference between f1 and f2 . Example Compute resonant lengths for a corner-fed patch on a 1.6-mm substrate with εr = 2.55 at 3 GHz. We have λ = 100 mm and thickness/λ = 0.016. From Figure 6-7 we read the 2 : 1 VSWR bandwidth: 1.61%. From Eq. (6-7) we calculate Q: Q=

1

√ = 43.9 0.0161 2

We use Eq. (6-35) to determine the resonant frequencies: 3 = 2.966 GHz 1 + 1/43.9 f2 = 3 1 + 1/43.9 = 3.034 GHz

f1 = √

By using the techniques of Section 6-3, we calculate the resonant lengths: a = 30.27 mm, b = 31.01 mm. All perturbations by small areas in a circular patch can only be type A feeding. The perturbation equations are related to the circular patch separation constant X11 (1.84118): 1 S 2 S = type B: = (6-36) type A: S X11 Q S X11 Q A circular patch perturbed into an elliptical patch radiates circular polarization when fed on a 45◦ diagonal from the major or minor axis and produces type B feeding. The ratio of major to minor axes is related to Q [4]: b 1.0887 =1+ a Q with resonant frequencies f0 f1 = √ 1 + 1.0887/Q

and f2 = f0 1 +

1.0887 Q

(6-37)

We compute Q by using Eq. (6-7) and read the bandwidths from Figure 6-25 for circular patches. Use the techniques of Section 6-5 to calculate the physical radius of the major and minor axes from the frequencies [Eq. (6-35)]. 6-7 COMPACT PATCHES The desire to produce small patches for cellular telephone handset use has lead to the development of compact designs. The ideal antenna is one whose location the user is unaware of and which is as small as possible. Because most signals arrive at the user

320

MICROSTRIP ANTENNAS

after many bounces and edge diffractions, polarization is arbitrary. We do not need to control the radiation pattern or its polarization carefully and it opens up a range of possibilities. Shorting pins placed close to the feed pin reduce the patch size to about 1 λ on a side, but its polarization is poorly controlled. If we can force the current to take 8 a longer path along the resonant-length path, we can shrink the overall size. We etch notches in the patch to make the current wander or use various spiral-wound networks on a ﬂat substrate. Three-dimensional solutions consist of folding a patch by using the vertical direction or some sort of winding around a cylinder. Many variations on these ideas appear in the literature and in collections of these ideas [16,22,23]. Adding a shorting pin closely spaced to the vertical feed pin (Figure 6-29) greatly reduces the resonant frequency of a given-size patch and produces a compact patch [24]. The idea is to make the current ﬂow over a longer path from the feed point to the radiation site; in other words, the transmission line has been folded to make the path longer in the resonant cavity. We use this concept for all compact patches. In this conﬁguration the resonant wavelength is found from the patch perimeter. Given the width W and the length L of the patch on a dielectric substrate εr , the resonant wavelength is given by √ λ0 = 4 εr (L + W ) (6-38) which reduces to a square patch λ/8 on a side. This patch has one-half the length and one-fourth the area of a quarter-wave path, with its short circuit along an entire √ edge. The circular shorting pin compact patch resonant diameter equals 0.14λ0 / εr . Making a patch this small produces highly inductive input impedance, which we can see by looking at the Smith chart of a coaxial probe-fed patch (Figure 6-12). The curve sweeps clockwise as the frequency increases. At low frequency (or small size) the patch is highly inductive. Figure 6-12 shows that using thicker substrates to increase bandwidth makes the patch impedance even more inductive. The shorting pin next to the feed pin forms a transmission line with it and adds a capacitive component to the input impedance that counteracts the patch and feed pin inductance. As the shorting pin is moved farther away from the feed pin, the capacitance decreases and the shorting pin becomes an inductive component, as it is in the quarter-wave patch. y

y

(xps, yps) R

x

x

W

(xp, yp)

(xp, yp)

(xps, yps) L

(a) z q Shorting Post εr

d

(b)

FIGURE 6-29 Compact patch with a shorting pin near the feed. (From [24], Fig. 1, 1998 IEEE.)

COMPACT PATCHES

321

The recommended position of the shorting pin is 80 to 90% of the distance from the center to the outer edge and a diameter of 0.008λ. You will need to iterate the position of the feed probe, usually one-half the diameter of the shorting pin, to achieve an impedance match. Table 6-5 lists the bandwidth achieved versus substrate thickness on foam, εr = 1.07 [25, p. 207]. Figure 6-30 gives the calculated pattern of a shorting pin compact patch on 0.034λ free-space substrate. The broad E-plane pattern has 10-dB dip on the broadside matched by the H -plane Eφ component. The large current in the shorting pin produces a significant monopole pattern seen in the Eθ radiation in the H -plane. This small antenna is a combination of a top-loaded monopole and a patch. Thinner antennas have a lower pattern dip broadside to the substrate because the monopole is shorter. The planar inverted F antenna (PIFA) is similar electrically to the shorting pin compact patch. We move the shorting pin to one corner and often make it a small shorting plate. We locate the feed pin close to the small shorting plate to again form a transmission line whose capacitance with the feed pin counteracts the inductive component of the small patch. We use Eq. (6-38) to determine its resonant wavelength. If we rotate the coordinates so that the shorting plate and diagonal lie on the x-axis, we obtain the pattern response of Figure 6-30. Since there is practically no difference between the two antennas, Table 6-5 gives the bandwidth of the PIFA versus thickness [26]. TABLE 6-5 Bandwidth of a Single Shorting Pin Compact Patch Thickness (λ0 ) 0.01 0.02 0.03 0.04 0.05 0.06

Bandwidth, 2 : 1 VSWR (%)

Feed-to-Pin Center Distance (λ0 )

1.6 2.2 2.7 3.4 4.3 5.7

0.0071 0.0076 0.0081 0.0085 0.0101 0.0135

E-Plane

H-Plane X-Pol

H-Plane

FIGURE 6-30 Pattern of a compact patch.

322

MICROSTRIP ANTENNAS

(b)

(a)

FIGURE 6-31 Reduced-size microstrip patches using meandered current paths. (From [22], Fig. 1-3, 2002 John Wiley & Sons, Inc.)

Modest size reduction can be obtained by making the currents ﬂow along a longer path along the resonant length. Figure 6-31 shows two planar antennas where slits cut from the width sides and disrupting the resonant-length path cause wandering of the current. The bowtie patch also makes the current path longer. These antennas radiate normal patch patterns with broader beamwidths in the E-plane because the notches bring the radiating edges closer together. The antennas in Figure 6-32 shrink the resonant length by folding the antennas vertically. The total length along the path is approximately λ/2, but the radiating edges are closer together. A large number of variations using slots have been investigated and offer interesting approaches to both shrink the patch size and produce dual-frequency antennas by using both the patch mode and slot radiation [22]. Ground Plane

Bent Edge

Air-Substrate Thickness ( a) Ground Plane

Folded Edge (b) Ground Plane

Double-Folded Edge (c)

FIGURE 6-32 Folding microstrip patches to reduce size. (From [22], Fig. 1-4, 2002 John Wiley & Sons, Inc.)

DIRECTLY FED STACKED PATCHES

323

6-8 DIRECTLY FED STACKED PATCHES Figure 6-7 illustrates the limited impedance bandwidth achievable from a singleresonator microstrip patch. When we increase the substrate thickness to widen the bandwidth, the antenna excites more surface waves (Figure 6-8) difﬁcult to control, and we accept them as losses. In Section 6-3 we discussed the use of external circuit elements to improve the impedance response. These have limited usefulness, although the simple series capacitor input to overcome the inductance of a long feed probe and the inductive nature of the higher-order modes is easily implemented. These external elements add poles to the resonant circuit to increase bandwidth. We can increase the number of poles by adding antenna elements instead. One solution is to couple to additional patches located around the fed patch on the same substrate surface. This increases the antenna size and reduces pattern beamwidths. This solution is difﬁcult to use in an array because the large spacing between elements produces grating lobes. Stacking patches vertically above the driven patch and coupling to them electromagnetically produces the best solution in terms of pattern response. The disadvantage of this approach is the additional fabrication cost. Our discussion of aperture-coupled patches in Section 6-3 points out that large apertures also add resonant poles that can increase the bandwidth. These added resonant elements complicate the design and call for the application of analytical tools instead of a cut-and-try approach. Although either patch in a two-element stacked patch design can be fed, feeding the lower element produces a design with minimum feed pin inductance. Aperture coupling through the ground plane feeds the lower patch directly as well. If we use an edge feed, we want the input transmission line to be as narrow as possible to reduce radiation by feeding the lower patch. Initially, we consider the probe-fed stacked patch [27]. A coaxial probe feeds directly a lower substrate of thickness d1 and dielectric constant εr1 through a hole in the ground plane. Figure 6-12 shows that the feed probe adds inductance for a thick substrate and the resonant loop is located on the upper inductive portion of the Smith chart. When we couple the lower patch to an upper patch with thickness d2 and dielectric constant εr2 , its circuit response becomes more inductive. We need to start with the impedance locus of the lower patch to be capacitive without the upper patch. This can be achieved by using an overcoupled feed. Figure 6-11 illustrates the overcoupled patch whose impedance locus sweeps around the origin of the Smith chart. The inductance of the feed probe rotates these curves clockwise around the center of the chart and the overcoupled response has signiﬁcant capacitive reactance when it sweeps around the origin. If we matched the lower patch critically, upward movement of the locus due to the coupled patch would reduce the impedance bandwidth. Figure 6-33 illustrates these design steps. Figure 6-33b shows that increasing the lower patch thickness leads to a longer feed probe that sweeps across the center from a more inductive portion of the Smith chart. Adding the second patch fails to increase the bandwidth relative to the thinner optimum lower patch. The thickness of the upper patch substrate d2 controls the tightness of the resonant loop. A greater thickness d2 produces a tighter loop in the Smith chart response that leads to a lower VSWR over a narrower bandwidth. Remember that we cannot use Eq. (6-7) to determine the bandwidth for different VSWR levels because we now have multiple resonators. If we use a foam upper substrate, the dielectric constant and thickness of the lower substrate determines the surface-wave efﬁciency. Waterhouse [27] used a dielectric

324

MICROSTRIP ANTENNAS Single layer + Stacked

+

+ ++ +

++

+ + + +

+

+++ ++ ++ + + ++ + + ++ + +

(a)

Single layer + Stacked

+ + + + + + + + + ++

+ ++ + ++ ++ ++ ++ +++ + + + + + + + + + + + +++ + + ++ + + ++ + + + +

(b)

FIGURE 6-33 Effect of coupling to a second patch: (a) overcoupled single lower patch response forms resonant loop with the second coupled patch; (b) increasing lower patch thickness causes rotation on a Smith chart and lower bandwidth. (From [27], Fig. 3, 1999 IEEE.)

constant of 2.2 for the lower substrate with a thickness of 0.04λ0 and a foam upper substrate 0.06λ0 to achieve optimum bandwidth with acceptable surface-wave losses. The lower patch was overcoupled so that it swept through the 250- resistance point at resonance. Since the impedance locus sweeps clockwise on the Smith chart as frequency increases, this resonant point should be slightly below the lower end of the desired frequency band. We adjust the second substrate thickness to move the resonant loop on the Smith chart in the vertical direction. As we increase the size of the upper patch, the loop moves around an arc in the clockwise direction, which we use to center the impedance response on the Smith chart for optimum bandwidth. This method produces impedance bandwidths of around 25%. The pattern bandwidth exceeds this bandwidth and we expect little change in pattern over this frequency range. Another successful stacked patch fed from a coaxial probe is the hi-lo conﬁguration, in which a high dielectric substrate (εr1 = 10.4) is used for the lower substrate and a foam (εr2 = 1.07) for the upper substrate [25, pp. 178–182]. The upper patch captures the surface wave of the lower patch and greatly improves the overall efﬁciency by radiating this power in a space wave. Although the two patches have different sizes, the coupling remains sufﬁcient to produce a broadband antenna with impedance bandwidths approaching 30%. In this design the lower patch is designed for the high dielectric of the lower substrate with little consideration for the upper patch except for making it a little overcoupled. The upper patch can be designed using the substrate thickness and dielectric constant assuming that the high dielectric substrate acts as the ground plane. When we mount the upper patch over the smaller lower patch, small adjustments must be made to the dimensions to achieve a 50- impedance match. The example given used a lower substrate thickness of 0.032λ0 with εr1 = 10.4, and by Figure 6-8 would have −1.3-dB surface-wave loss. Locating the second patch on a 0.067λ0 -thick foam substrate directly over the ﬁrst patch reduced the surface-wave loss to better than −0.7 dB over the entire band.

APERTURE-COUPLED STACKED PATCHES

325

6-9 APERTURE-COUPLED STACKED PATCHES The discussion on aperture feeding of a patch in Section 6-3 stated that we can utilize the aperture as another resonator to broadband the antenna. Figure 6-34 shows the stacked patch antenna fed from an aperture. In this implementation we make the coupling slot long enough to be one of the resonators, which increases the number of resonators to three: the aperture, the lower patch, and the upper patch. We must use element spacing to control coupling because frequencies control resonator sizes. By careful control of parameters two loops will form in the Smith chart response of impedance and be made to wrap tightly around the center of the chart [28] as shown in Figure 6-35b. We form these loops by coupling resonators. Undercoupling produces small tight loops; overcoupling produces large loops. Figure 6-35 illustrates the effect of aperture size. The left Smith chart shows undercoupling between the aperture and the lower patch by the small left loop. We increase the coupling by increasing the aperture slot length (Figure 6-35b) or by increasing the

Patch 2(PL2, PW2, Layer N2)

Layer N(εrN, dN, tan dN)

Layer N2(εrN2, dN2, tan dN2) Patch 1(PL1, PW1, Layer N1)

Layer N1(εrN1, dN1, tan dN1) Layer 1(εr1, d1, tan d1)

Ground Plane Aperture(SL, SW) Feedline(Wf, doff, Lstub)

Feed Substrate (εrf, df, tan df)

FIGURE 6-34 Construction of a resonant aperture coupled dual patch in exploded view. (From [28], Fig. 1, 1998 IEEE.)

(a)

(b)

(c)

FIGURE 6-35 Effect on increasing slot length, SL, of apertures in stacked dual patches: (a) SL = 8 mm; (b) SL = 10 mm; (c) SL = 12 mm. (From [28], Fig. 4, 1998 IEEE.)

326

MICROSTRIP ANTENNAS

lower patch size or reducing the lower patch thickness. The best results have been obtained by having the lower-frequency (left) loop determined by the lower patch and aperture. Overcoupling the aperture to the lower patch produces the impedance locus of Figure 6-35c. We control the upper loop size by varying the upper patch size, the relative size between the two patches, and the upper substrate thickness. The lower patch size is a critical parameter because it affects the coupling and size of both loops while shifting their center frequencies. For ﬁxed sizes of the other two resonators, decreasing the lower patch size decreases the coupling to the aperture while increasing the coupling to the upper patch. Increasing aperture size increases coupling to the aperture and decreases coupling to the upper patch. By remembering that overcoupling produces larger Smith chart loops, we determine in which direction to change parameters by observing changes in analytical results on the Smith chart to produce optimum designs. Because the slot aperture is one of the three resonators, we cannot vary its length to determine coupling to the lower patch. The overcoupled large slot produces high resistance at the microstrip input. We can lower this impedance by offset feeding the slot or by using a wide transmission line. A single offset line will unbalance the ﬁelds in the slot and lead to unbalanced excitation of the patches. This unbalanced excitation on the patches increases cross-polarization. The dual balanced offset feeding shown in Figure 6-36, where we join the two lines in a reactive power divider, both lowers the resistance and balances the patch excitation. A design using rectangular patches for a single linear polarization achieved a 67% 2 : 1 VSWR bandwidth [28]. The only signiﬁcant problem with the design is the poor front-to-back ratio, which is reduced to 6 dB at the upper frequencies as the aperture radiation increases. Placing a reﬂector patch below the microstrip feed line, it can be sized to reduce the F/B ratio by forming a Yagi–Uda antenna with the stacked patch [29]. Figure 6-37 illustrates an exploded view of a dual polarized aperture stacked patch. The potential bandwidth shrinks because we lose width as a parameter with square patches to optimize impedance. The key element of this design is the feed crossed slot [30]. The crossed-slot feeding aperture is located on a ground plane shared by microstrip networks located below and above the aperture. Each network consists of a reactive power divider to raise the impedance of the feed lines and allow offset feeding of the slot for each polarization. The balanced feed reduces cross-polarization and cross coupling between the two ports that would occur in both the crossed slot

2doff

Low Impedance Line Lstub

50Ω Line

Lstub

100Ω Line

50Ω Line (a)

(b)

FIGURE 6-36 Impedance matching for resonant aperture dual stacked patches: (a) wide transmission line; (b) dual offset feed. (From [28], Fig. 3, 1998 IEEE.)

PATCH ANTENNA FEED NETWORKS

Patch 2 (W2, L2)

327

εr8, h8

W

εr7, h7

L

εr6, h6 Patch 1 (W1, L1)

εr5, h5

Feed 2 (W2F, L2STUB)

εr4, h4

Slot 1 (W1SL, L1SL) Slot 2 (W2SL, L2SL)

L1STUB

εr3, h3 εr2, h2

Feed 1 (W1F, L1STUB) εr1, h1 Reflector (W1REF, L1REF) Reflector (W2REF, L2REF)

FIGURE 6-37 Exploded view of construction of a dual polarized aperture-fed stacked patch utilizing a crossed strip reﬂector. (From [8], Fig. 3.6.22, 2003 Kluwer Academic Publishers.)

and the patch elements. This shows that the slot that couples to a patch resonator can be fed by a microstrip line located either below or above the slot. The ground plane between the two networks for each polarization eliminates direct coupling between the microstrip networks and symmetrical feeding reduces coupling in the slot. Because we use long slots to feed the lower microstrip patch in an overcoupled excitation, direct coupling of the upper microstrip to the lower patch is minimal in comparison. We use thin substrates of moderate dielectric constant (εr = 2.2) to support the etched patches and foam layers between to separate the patches to increase bandwidth and control coupling. Figure 6-37 shows a crossed dipole used as a reﬂector element below the microstrip feed lines to reﬂect direct radiation from the crossed slot that reduces the F/B ratio. 6-10 PATCH ANTENNA FEED NETWORKS Patch antenna arrays may be fed from below (Figure 6-9) by using a stripline distribution network. The connections between the boards greatly complicate the assembly. A connection made vertically from the center strip of a stripline unbalances the ﬁelds and induces parallel-plate modes. Shorting pins between the ground planes suppress this mode. It is far easier to etch the feed network on the microstrip and use either edge feeds or aperture feeds with the network located below the patch layer. Feed networks radiate very little in comparison with the patches when etched on the same substrate because radiation from fringing ﬁelds on the two sides of the microstrip lines cancel each other except at discontinuities (corners and steps). Consider the equally fed array (Figure 6-38). Equal amplitude and phase feeding generates virtual magnetic walls between the patches as shown. We can join the edges

328

MICROSTRIP ANTENNAS Magnetic Walls

l/4 l/4

Transformer 50 Ω

100 Ω 70.7 Ω 100 Ω

FIGURE 6-38 Equally fed microstrip patch array.

between the patches without effect, since the midpoint remains a virtual open circuit and the separate patches join into a continuous strip. The feeds must be spaced close enough together to prevent grating lobes and to provide uniform amplitude along the edges. These antennas can be wrapped around missiles to provide omnidirectional coverage about the roll axis. To eliminate pattern ripple, feeds must be spaced about every 0.75λ in a circular array. The resistance at each feed at resonance will be the combination of the radiation conductances from the portions of the edges between the magnetic walls. Figure 6-38 illustrates an equally fed four-element array. Starting from the patch, a quarter-wavelength transformer reduces the roughly 200- impedance to 100 . Two 100- lines join in shunt to 50 at their juncture. A 70.7- quarter-wavelength line transforms the 50 back to 100 . We continue this sequence for any 2N -array for reactive power dividers at each junction. Equal path lengths from the input excite them with equal phases. Arrays with the number of elements different from 2N -are possible, but they require more difﬁcult feed networks. A 100- system was picked because 50- lines on low-dielectric-constant substrates are quite wide. The reactive power divider (Figure 6-38) has more bandwidth than the patch while it is matched at the input but not at its outputs. The network can be analyzed by using even and odd modes and shows that the output return loss is 6 dB, and it provides only 6 dB of isolation between outputs. The power reﬂected from a damaged antenna distributes to the other elements of the array and produces an effect greater than that of just a missing element. Making power dividers with isolation resistors reduces this problem, but we cannot justify the added difﬁculty of mounting resistors when both good etchings and low probability of damage make them unnecessary. We must be wary of coupling between different parts of the feed network. We want to pack the feed network into the smallest area, but coupled signals between the lines produce unexpected anomalies. Distinguishing direct radiation from the feed and coupling redistribution is difﬁcult. Although couplings are predictable, they appear as random errors when we cannot perform a full analysis. Unfortunately, the coupling between microstrip lines falls off quite slowly. Table 6-6 lists the coupling and peak errors for 100- lines; those of 50- lines are very similar. We read the amplitude and phase errors from Scales 1-8 and 1-9.

SERIES-FED ARRAY

329

TABLE 6-6 Peak Feed Errors Due to Microstrip Coupling for 100- Lines (εr = 2.4) Spacing/Substrate Thickness 1.0 2.0 3.0 4.0 5.0

Coupling (dB)

Amplitude Error (dB)

Phase Error (deg)

16 23 28 32 35

1.5 0.7 0.4 0.22 0.12

9.0 4.0 2.3 1.5 1.0

6-11 SERIES-FED ARRAY If we reduce the width of the patch, the radiation conductance is insufﬁcient to match the input. We can use the microstrip patch as a transmission line and connect a line opposite the feed to lead to other patches (Figure 6-39). If we space the patches by halfwavelengths, the impedances of the patches will add in phase at the input, because it rotates once around the Smith chart in λ/2. The characteristic impedance of the connecting lines has no effect at center frequency. The junction of the transmissionline feeder and the patch introduces excess phase shift. In arrays of a few elements, the extra phase shift can be ignored, but arrays with a large number of elements, or when we design for critical amplitude taper, must account for δ. Of course, traveling-wave or resonant arrays can be designed. The frequency dispersion of the traveling-wave array can be used to frequency-scan the beam. Various experimental methods have been devised to measure the parameters of the series array. Metzler [31] performed experiments on uniform-width element arrays to determine the radiation conductance and excess phase shift. Measuring the transmission loss through the array as a network with input and output connectors determines the radiation conductance of the patches. An empirical equation was obtained:

W G = 0.0162 λ0

1.757 0.033 ≤

W ≤ 0.254 λ

(6-39)

where G is the total radiation conductance of each patch, with half from each edge. Measurement of the beam direction of the uniform traveling-wave array determines the excess phase shift in each patch. Jones et al. [32] model the patch (Figure 6-39) with extensions due to the fringing ﬁelds as a transmission line: L + 2 long. The other excess phase shift, due to the step, is modeled as extensions to the input lines (δ). Jones et al. perform measurements on single elements to establish these lengths. is found from the resonant frequency of the √ patch: L + 2 = λ/2 εeff , where εeff is given by Eq. (6-19). When the transmissionline phase is measured through the patch at resonance, the excess phase beyond π is equated to a phase shift length in the narrow feeder lines: 2δ =

λN φexcess 2π

where λN is the wavelength in the narrow line.

330

MICROSTRIP ANTENNAS

(a)

(b)

(c)

FIGURE 6-39 Series-fed patch and its equivalent circuit. (From [32], Fig. 2, 1982 IEEE.)

When designing the array, we vary the widths of the patches to achieve the desired √ amplitude taper. The voltage distribution at each patch is given by V g, where g is the patch conductance. Standing-wave (resonant) arrays require that the sum of the conductances be equal to the input conductance desired. We have some latitude when we feed the array through a quarter-wavelength transformer. The nonresonant array requires a matched load on the end to prevent standing waves. We must pick the ratio of the power dissipated to the radiated power that gives us an extra parameter with which to optimize the design. We control the beam direction by spacing the elements to achieve the phase shift required. 6-12

MICROSTRIP DIPOLE [33]

As the width W of a patch narrows, the input impedance increases. When the width approaches that of a microstrip feed line, either the patch fails to be a resonator or the

331

MICROSTRIP DIPOLE

feed line becomes very narrow in trying to transform the impedance. The microstrip dipole solves these problems by having a coupled line feeder. The dipole is a halfwavelength strip whose width equals that of a microstrip feed line. A line etched on a substrate below feeds the dipole by coupling into the strip (Figure 6-40a). The equivalent circuit (Figure 6-40b) transforms the high impedance of the dipole through the unequal coupled lines. By varying the coupling, we can change the input impedance at resonance. Best results occur for quarter-wavelength overlap where the equivalent stubs (Figure 6-40b) do not contribute reactance. We vary the coupling by changing the thickness of the substrate between the strips or by offsetting the lower strip. The dipole radiates as a narrow patch and not as a dipole. No pattern nulls appear along the axis of the strip, but they occur more strongly in the direction of the equivalent magnetic currents of the edges. The H -plane pattern becomes quite broad for the narrow strip width. The feed distribution circuit is etched on the substrate below the dipoles. With the feed circuit on a separate level, we have greater freedom in the feed network design to excite desired distributions. Also, because the dipoles are small, we can use density tapering of the dipoles to that end. Proper design requires measurement [34] Overlap Region Microstrip Feed Line

L/2

L

W

Ground Plane

Substrates (a)

a

Zoo = vLa =

Cb vF

b

Zoo = vLb =

q a

Cb vF

q

q

a

Zoe, Zoo 1

a

=

1

a

Zoe − Zoo = vLab = Cab vF 2

2

2 b

b

Zoe, Zoo

(b)

F = CaCb + CaCab + CbCab

FIGURE 6-40 (a) Microstrip dipole; (b) equivalent circuit. [(b) From G. L. Matthaei et al., Microwave Filters, Impedance Matching Networks, and Coupling Structures, 1980 Artech House, Inc.]

332

MICROSTRIP ANTENNAS

to obtain the desired effect, since mutual coupling will change the distribution by changing the active impedance of each dipole. The feed network must compensate for the coupling. 6-13

MICROSTRIP FRANKLIN ARRAY [35]

An electrically long line with a standing wave on it fails to radiate on broadside because the many cycles cancel each other. We obtain a pattern with many nulls and lobes. By folding the lines with out-of-phase standing-wave currents close together, we can prevent their radiation. The other portions are free to radiate (Figure 6-41a). The Franklin array consists of straight sections λ/2 long connected by λ/4 shorted stubs. The standing-wave currents on the straight portions add in phase. We can construct a microstrip version (Figure 6-41b). Half-wavelength lines act as radiators (patches). We connect them with half-wavelength lines folded into stubs so that the counteracting standing-wave currents do not radiate. The straight lines are narrow patches. The total radiation conductance of each strip is 1 W 2 G= (6-40) 45 λ for narrow strip widths W , where λ is the free-space wavelength. Using lines for the stubs whose impedance is twice the radiating strip impedance reduces unwanted internal reﬂections. The two stubs add in shunt. Since the antenna is quite narrowband and the length of the lines between patches is a half-wavelength long, the impedance of these connecting arms has a secondary effect. Example Design an eight-wavelength array at 10 GHz. There are 16 patches in the array. The radiation conductances add for elements spaced at λ/2 intervals. For a 100- input, each patch supplies a conductance 0.01/16. We solve Eq. (6-40) for the width: 0.01(45) W =λ = 0.168λ 16 l/2

l/2

l/2

l/2

l l

l/4

l (a)

L

l/2

W

Connecting line

Patch

(b)

FIGURE 6-41 (a) Dipole and (b) microstrip Franklin arrays.

MICROSTRIP ANTENNA MECHANICAL PROPERTIES

333

If we use Eq. (6-39) from the series patch, we obtain W = 0.157λ, within the range of the empirical formula. For 10 GHz, W = 4.71 mm. On an 0.8-mm substrate (εr = 2.21), W/H = 5.89 and the impedance of the strip radiator Z0 = 44.01 . We need to ﬁnd the effective dielectric constant of the strip to determine the patch length and impedance. From Eq. (6-19), εr = 1.97. We calculate the cutback from each end by using Eq. (6-18); = 0.40 mm. Each radiating strip is 300 × 109 L= − 2(0.40) = 9.88 mm √ 1010 (2) 1.97 √ The radiating-strip impedance is Z0 / εeff = 31.3 . We need 62.6- connecting lines in the stubs to achieve the broadest bandwidth. With so few radiators, we could use 100- connecting arms with little change in bandwidth and have more reasonable connecting arm widths: 0.71 mm. The example shows that the microstrip Franklin array works best for high frequencies or long arrays. The elements are narrow, and the interconnecting arms are thin.

6-14 MICROSTRIP ANTENNA MECHANICAL PROPERTIES A microstrip patch antenna has very desirable mechanical properties. It can withstand tremendous shock and vibration. Because the antenna is on a solid substrate, the patch cannot ﬂex, and small changes in the substrate thickness have only a minor effect on the resonant frequency. The commonly used soft substrate (Teﬂon and ﬁberglass) has a good damped resilience. Microstrip patch antennas have been used to telemeter data from artillery shells and high-velocity rockets, which have high shock and vibration levels. The repeatability of the dimensions of the patches depends only on the etcher’s art. Complicated shapes and feed networks are produced as cheaply as simple ones. The antennas can withstand exposure to high temperatures when covered by a radome made of the same soft dielectric as the substrate. The cover protects the metal patches but has only a minor effect on the resonant frequency [36]. High temperatures on the surface of the radome or ablation fail to change the resonance signiﬁcantly because the radome itself has only a minor effect. Variation in the dielectric constant of the substrate from lot to lot causes problems with repeatability. The narrowband antennas require measurement of the dielectric constant of each lot, and sometimes of each sheet, to get the center frequency desired. A series of etching masks can be made to cover the expected range. The antennas can be tuned with inductive shorting pins or capacitive screws, but tuning is prohibitive when the number of elements in an array is large. Careful quality control of the dielectric constant is the answer. Close monitoring of the etching process may also be needed to prevent excessive undercutting. Temperature variations can be a problem with thin substrates when the bandwidth is narrow. The patch and substrate size grow when the temperature rises, but they are overshadowed by the change in dielectric constant of soft substrates. Instead of decreasing the resonant frequency because of the increased patch size, a lowered dielectric constant raises the center frequency. Whenever we need more bandwidth than a microstrip patch can provide, we must turn to cavity antennas. We increase the antenna volume by penetrating the vehicle for the cavity, but we gain a design parameter.

334

MICROSTRIP ANTENNAS

REFERENCES 1. R. F. Harrington, Effects of antenna size on gain, bandwidth, and efﬁciency, Journal of Research, NBS, D, Radio Propagation, vol. 64D, January–February 1960, pp. 1–12. 2. R. C. Hansen, Fundamental limitations in antennas, Proceedings of IEEE, vol. 69, no. 2, February 1981, pp. 169–173. 3. Y. T. Lo et al., Study of microstrip antenna elements, arrays, feeds, losses, and applications, Final Report, RADC-TR-81-98, Rome Air Development Center, Rome, NY, June 1981. 4. K. R. Carver and E. L. Coffey, Theoretical investigation of the microstrip antenna, Technical Report PT-00929, Physical Science Laboratory, New Mexico State University, Las Cruces, NM, January 1979. 5. D. R. Jackson and N. G. Alexopoulas, Simple approximate formulas for input resistance, bandwidth, and efﬁciency of a resonant rectangular patch, IEEE Transactions on Antennas and Propagation, vol. 39, no. 3, March 1991, pp. 407–410. 6. R. F. Harrington, Time-Harmonic Electromagnetic Fields, McGraw-Hill, New York, 1961. 7. D. M. Pozar, Rigorous closed-form expressions for the surface wave loss of printed antennas, Electronics Letters, vol. 26, no. 13, June 21, 1990, pp. 954–956. 8. R. B. Waterhouse, ed., Microstrip Patch Antennas, A Designer’s Guide, Kluwer Academic, Boston, 2003. 9. R. E. Munson, Conformal microstrip antennas and microstrip phased arrays, IEEE Transactions on Antennas and Propagation, vol. AP-22, no. 1, January 1974, pp. 74–78. 10. E. O. Hammerstad, Equations for microstrip circuit design, Proceedings of the 5th European Micro-strip Conference, Hamburg, Germany, September 1975, pp. 268–272. 11. M. V. Schneider, Microstrip lines for microwave integrated circuits, Bell System Technical Journal, vol. 48, May–June 1969, pp. 1421–1444. 12. D. A. Paschen, Practical examples of integral broadband matching of microstrip antenna elements, Proceedings of the 1986 Antenna Applications Symposium, Monticello, IL. 13. T. Samaras, A. Kouloglou, and J. N. Sahalos, A note on impedance variation of a rectangular microstrip patch antenna with feed position, IEEE Antennas and Propagation Magazine, vol. 46, no. 2, April 2004. 14. D. M. Pozar, Microstrip antenna aperture-coupled to a microstrip-line, Electronics Letters, vol. 21, no. 2, January 1985, pp. 49–50. 15. P. L. Sullivan and D. H. Schaubert, Analysis of an aperture coupled microstrip antenna, IEEE Transactions on Antennas and Propagation, vol. AP-34, no. 8, August 1986, pp. 977–984. 16. G. Kumar and K. P. Ray, Broadband Microstrip Antennas, Artech House, Boston, 2003. 17. J. R. James, P. S. Hall, and C. Wood, Microstrip Antennas: Theory and Design, Peter Peregrinus, London, 1981. 18. L. C. Shen et al., Resonant frequency of a circular disk, printed circuit antenna, IEEE Transactions on Antennas and Propagation, vol. AP-25, no. 4, July 1977, pp. 595–596. 19. I. J. Bahl and P. Bhartia, Microstrip Antennas, Artech House, Dedham, MA, 1980. 20. A. G. Derneryd, Analysis of the microstrip disk antenna element, IEEE Transactions on Antennas and Propagation, vol. AP-27, no. 5, September 1979, pp. 660–664. 21. J. L. Kerr, Microstrip antenna developments, Proceedings of the Workshop on Printed Circuit Antennas, New Mexico State University, Las Cruces, NM, October 1979, pp. 3.1–3.20. 22. K.-L. Wong, Compact and Broadband Microstrip Antennas, Wiley, New York, 2002. 23. K.-L. Wong, Planar Antennas for Wireless Communications, Wiley, New York, 2003.

REFERENCES

335

24. R. B. Waterhouse, S. D. Targonski, and D. M. Kokotoff, Design and performance of small printed antennas, IEEE Transactions on Antennas and Propagation, vol. AP-46, no. 11, November 1998, pp. 1629–1633. 25. R. B. Waterhouse, ed., Microstrip Patch Antennas: A Designer’s Guide, Kluwer Academic, Boston, 2003. 26. T. Taga and K. Tsunekawa, Performance analysis of a built-in planar inverted-F antenna for 800 MHz band portable radio units, IEEE Transactions on Selected Areas in Communication, vol. SAC-5, no. 5, June 1987, pp. 921–929. 27. R. B. Waterhouse, Design of probe-fed stacked patches, IEEE Transactions on Antennas and Propagation, vol. AP-47, no. 12, December 1999, pp. 1780–1784. 28. S. D. Targonski, R. B. Waterhouse, and D. M. Pozar, Design of wide-band aperture-stacked patch microstrip antennas, IEEE Transactions on Antennas and Propagation, vol. AP-46, no. 9, September 1998, pp. 1245–1251. 29. S. D. Targonski and R. B. Waterhouse, Reﬂector elements for aperture and aperture coupled microstrip antennas, IEEE Antennas and Propagation Symposium Digest, Montreal, Quebec, Canada, July 1997, pp. 1840–1843. 30. J. R. Sanford and A. Tengs, A two substrate dual polarized aperture coupled patch, IEEE Antennas and Propagation Symposium Digest, 1996, pp. 1544–1547. 31. T. Metzler, Microstrip series arrays, Proceedings of the Workshop on Printed Circuit Antennas, New Mexico State University, Las Cruces, NM, October 1979, pp. 20.1–20.16. 32. B. B. Jones, F. V. M. Chow, and A. W. Seeto, The synthesis of shaped patterns with seriesfed microstrip patch arrays, IEEE Transactions on Antennas and Propagation, vol. AP-30, no. 6, November 1982, pp. 1206–1212. 33. D. A. Huebner, An electrically small dipole planar array, Proceedings of the Workshop on Printed Circuit Antennas, New Mexico State University, Las Cruces, NM, October 1979, pp. 17.1–17.16. 34. R. S. Elliott and G. J. Stern, The design of microstrip dipole arrays including mutual coupling, parts I and II, IEEE Transactions on Antennas and Propagation, vol. AP-29, no. 5, September 1981, pp. 757–765. 35. K. Solbach, Microstrip-Franklin antenna, IEEE Transactions on Antennas and Propagation, vol. AP-30, no. 4, July 1982, pp. 773–775. 36. I. J. Bahl et al., Design of microstrip antennas covered with a dielectric layer, IEEE Transactions on Antennas and Propagation, vol. AP-30, no. 2, March 1982, pp. 314–318.

7 HORN ANTENNAS

Horn antennas have a long history, traced in part in the collection of papers by Love [1] together with papers on every other horn topic. Horns have a wide variety of uses, from small-aperture antennas to feed reﬂectors to large-aperture antennas used by themselves as medium-gain antennas. Horns can be excited in any polarization or combination of polarizations. The purity of polarization possible and the unidirectional pattern make horns good laboratory standards and ideal reﬂector feeds. Horns also closely follow the characteristics predicted by simple theories. Horns are analyzed using a variety of techniques. Barrow and Chu [2] analyzed a sectoral horn, ﬂaring in only one plane, by solving the boundary value problem in the wedge. They expanded the ﬁelds in terms of Hankel functions in cylindrical coordinates. The ﬁelds form an equiphase surface over a cylindrical cap to which the Kirchhoff–Huygens equivalent current method [Eq. (2-23)] can be applied to compute the pattern. Similarly, Schorr and Beck [3] use spherical Hankel and Legendre functions to analyze conical horns. The integration surface consists of a spherical cap. Schelkunoff and Friis [4] use the mouth of the horn as the aperture and approximate the phase distribution as quadratic. Both aperture theories have the same valid pattern range. The method predicts patterns accurately in the area in front of the aperture. The error increases as the plane of the aperture is approached. The predicted pattern remains continuous and gives no indication of its increasing error. GTD methods [5] predict the pattern both in back and in front of the aperture while providing estimates of the error in the predictions. Most of the details needed for design can be obtained from the aperture theory. Only GTD predicts sidelobes accurately, since no assumption of zero ﬁelds outside the horn aperture is made. Figure 7-1 shows the general horn geometry. The input waveguide can be either rectangular or circular (elliptical). W is the width of a rectangular aperture, and a is the radius of a circular aperture. The distance from the junction of the projected sides Modern Antenna Design, Second Edition, By Thomas A. Milligan Copyright 2005 John Wiley & Sons, Inc.

336

RECTANGULAR HORN (PYRAMIDAL)

FIGURE 7-1

337

General geometry of a horn.

to the aperture is the slant radius R. The distance along the centerline from the aperture to the waveguide is the axial length. We derive the aperture ﬁeld amplitude from the input waveguide mode while the phase distribution is approximately quadratic across the aperture. We assume that the aperture ﬁelds radiate in spherical waves from the projected juncture of the sides, and the extra distance along the sides compared with the distance to the center of the aperture is given by =R−

R2 − a2

= R 1 −

1−

a R2 2

a2 ≈R 1− 1− 2R 2

=

a2 W2 = 2R 8R

We divide by wavelength to obtain the dimensionless constant S of the quadratic phase distribution: W2 a2 S= = = (7-1) λ 8λR 2λR Since the semiﬂare angle θ0 of most practical horns is small, we use the quadratic phase error approximation.

7-1 RECTANGULAR HORN (PYRAMIDAL) The rectangular horn ﬂares out of a rectangular or square waveguide with ﬂat metal walls. Figure 7-2 shows the horn geometry. The slant radiuses along the sides will be unequal, in general. The input waveguide dimensions are width a and height b.

338

HORN ANTENNAS

FIGURE 7-2 Rectangular horn geometry.

The aperture has width W in the H -plane and height H in the E-plane. Each aperture coordinate has its own quadratic phase distribution constant: Se =

H2 8λRe

Sh =

W2 8λRh

(7-2)

The TE10 mode of the lowest-order waveguide mode has the ﬁeld distribution Ey = E0 cos

πx a

Combining these ideas, the aperture electric ﬁeld is approximated by 2 πx 2y 2 2x exp −j 2π Se + Sh Ey = E0 cos W H W

(7-3)

The ratio of the electric and magnetic ﬁelds approaches the impedance of free space for large apertures. In this case we use the Huygens source approximation and need only the electric ﬁeld with Eq. (2-24) to ﬁnd the pattern. Small-aperture horns require Eq. (2-23) with an arbitrary ratio of the magnetic and electric ﬁelds. We compute the E-plane pattern by using a uniform aperture distribution and the H plane pattern from a cosine distribution. Both have a quadratic phase error. Figures 7-3 and 7-4 plot the E- and H -plane universal patterns in U -space of the Taylor distribution with S as a parameter. We can use them to determine the pattern of a general rectangular horn. Example Compute the pattern level at θ = 15◦ in the E- and H -planes of a horn with the following measured dimensions: Aperture: W (H -plane) = 28.9 cm, H (E-plane) = 21.3 cm Input waveguide: width a = 3.50 cm, height b = 1.75 cm

RECTANGULAR HORN (PYRAMIDAL)

339

0.9

0.8

S = 0.6

Relative Field Strength

0.7 S= 0.6

H2 8lRe

0.4 0.5

0.5 0.3 0.4 0.2

0.3

0.1 0.2 0 0.1 0 1

2 H/l sin q

3

4

FIGURE 7-3 E-plane universal pattern of a rectangular, TE10 mode.

The slant distance from the aperture to the waveguide along the center of each plate of the ﬂare was measured: Dh = 44.8 cm and De = 44.1 cm. We calculate the slant radius from similar triangles: Rh W = Dh W −a

H Re = De H −b

(7-4)

Slant radius: Rh = 50.97 cm, Re = 48.05 cm The frequency is 8 GHz (λ = 3.75 cm). Using Eq. (7-2), we compute Sh = 0.55 and Se = 0.31. We use Figures 7-3 and 7-4 to determine the universal pattern ﬁeld intensity (voltage): W H sin θ = 2.0 sin θ = 1.47 λ λ The ﬁelds from the ﬁgures are 0.27 (H -plane) and 0.36 (E-plane). We must include the obliquity factor of the Huygens source element pattern: (1 + cos θ )/2 to obtain the proper pattern level. At θ = 15◦ , the obliquity factor is 0.983. We calculate the pattern level in decibels from 20 times the logarithm of the product of the ﬁeld intensity from

340

HORN ANTENNAS

0.9 S=1 0.8 S=

Relative Field Strength

0.7

W2 8lRh

0.8

0.6 0.6 0.5

0.4

0.3 0.4 0.2 0.2 0.1 0 0 1

FIGURE 7-4

2 W/l sin q

3

4

H -plane universal pattern of a rectangular, TE10 mode.

the ﬁgures and the obliquity factor: H -plane : −11.5 dB

E-plane : −9 dB

We can calculate gain of this horn by using aperture efﬁciencies: H -plane (cosine) (Table 4-1) : 0.91 dB

E-plane (uniform) : 0 dB

These values hold for all rectangular horns excited by the TE10 mode. The quadratic phase distributions give us the phase error loss. From Table 4-42 we interpolate these losses: Sh = 0.55 cosine distribution

PEL = 2.09 dB

Se = 0.31 uniform distribution

PEL = 1.50 dB

The directivity is given by directivity = 10 log

4πW H − ATLh − ATLe − PELh − PELe = 22.9 dB λ2

The aperture efﬁciency is 35.5%, since ATL + PELh + PELe = 4.5 dB.

(7-5)

RECTANGULAR HORN (PYRAMIDAL)

341

We usually equate gain to directivity, since the wall losses are very small. Of course, for millimeter-wave horns we must include wall losses. Although we can use Table 4-42 along with the ﬁxed-amplitude taper loss of 0.91 dB to determine the aperture efﬁciency of a rectangular horn, Schelkunoff and Friis [4] give the following closed-form equation for the directivity: directivity =

8Rh Re {[C(u) − C(v)]2 + [S(u) − S(v)]2 }[C 2 (z) + S 2 (z)] WH

where 1 u= √ 2

√

λRh W +√ W λRh

and

x

C(x) = 0

1 v=√ 2

πt 2 dt cos 2

√

λRh W −√ W λRh

x

S(x) =

sin 0

z= √

H 2λRe (7-6)

πt 2 dt 2

are the Fresnel integrals. Closed-form expressions for these integrals are available [6]. 7-1.1 Beamwidth We can use Figures 7-3 and 7-4 to compute beamwidths. The 3-dB point corresponds to 0.707 and the 10-dB point to 0.316 on the graphs. Table 7-1 lists the 3- and 10dB points (values of U ) for different quadratic phase constants S in the H -plane. Similarly, Table 7-2 lists the E-plane points. The tables are more convenient than the graphs. Because we remove the obliquity factor to get a universal pattern, we must modify the beamwidths found by using the tables. We ﬁnd the beamwidth, but then we must add the obliquity factor to the beamwidth level. The beamwidth is found for

TABLE 7-1

Rectangular-Horn H -Plane Beamwidth Points, TE10 Mode (W/λ) sin θ

(W/λ) sin θ

S

3 dB

10 dB

S

3 dB

10 dB

0.00 0.04 0.08 0.12 0.16 0.20 0.24 0.28 0.32 0.36 0.40 0.44 0.48

0.5945 0.5952 0.5976 0.6010 0.6073 0.6150 0.6248 0.6372 0.6526 0.6716 0.6951 0.7243 0.7609

1.0194 1.0220 1.0301 1.0442 1.0652 1.0949 1.1358 1.1921 1.2700 1.3742 1.4959 1.6123 1.7143

0.52 0.56 0.60 0.64 0.68 0.72 0.76 0.80 0.84 0.88 0.92 0.96 1.00

0.8070 0.8656 0.9401 1.0317 1.1365 1.2445 1.3473 1.4425 1.5320 1.6191 1.7071 1.7991 1.8970

1.8062 1.8947 1.9861 2.0872 2.2047 2.3418 2.4876 2.6246 2.7476 2.8618 2.9744 3.0924 3.2208

342

HORN ANTENNAS

TABLE 7-2 Rectangular-Horn E -Plane Beamwidth Points, TE10 Mode (H /λ) sin θ

(H /λ) sin θ

S

3 dB

10 dB

S

3 dB

10 dB

0.00 0.04 0.08 0.12 0.16 0.20

0.4430 0.4435 0.4452 0.4482 0.4527 0.4590

0.7380 0.7405 0.7484 0.7631 0.7879 0.8326

0.24 0.28 0.32 0.36 0.40 0.44

0.4676 0.4793 0.4956 0.5193 0.5565 0.6281

1.4592 1.5416 1.6034 1.6605 1.7214 1.8004

a lower pattern level than speciﬁed. Since the beamwidth levels are close, we use a relation by Kelleher [7, p. 12–5] with good results: BW2 level2 (dB) (7-7) = BW1 level1 (dB) Example Compute 3- and 10-dB beamwidths for the horn in the preceding example. We have Sh = 0.55 and Se = 0.31. From Tables 7-1 and 7-2,

θh3

W sin θh3 λ W sin θh10 λ 28.9 W = λ 3.75 = 6.34 θh10

= 0.851 = 1.8726 = 7.707 = 14.06

H sin θe3 = 0.4915 λ H sin θe10 = 1.588 λ H 21.3 = = 5.68 λ 3.75 θe3 = 4.96 θe10 = 16.24

We consider the obliquity factor, (1 + cos θ )/2, at these angles, and apply Eq. (7-7) to reduce the beamwidths found. ◦

BWh3 = 12.68 at 3.03 dB ◦

BWe3 = 9.92 at 3.02 dB

◦

BWh3 x = 12.62 at 3.01 dB ◦

BWe3 = 9.89 at 3.01 dB

◦

BWh10 = 27.94 at 10 dB

◦

BW10 = 32.2 at 10 dB

BWh10 = 28.12 at 10.13 dB BWe10 = 32.48 at 10.18 dB

◦

◦

Including the obliquity factor has a very small effect on the results, but the effect increases with larger beamwidths (smaller apertures). Aperture theory fails for small horns because the beam is determined more by edge diffraction than the aperture ﬁelds. Empirical data were collected and reduced to simple formulas for small rectangular horns based on aperture size only [8, p. 46–22]: ◦

BW10e = 88

λ H

◦

◦

and BW10h = 31 + 79

λ W

RECTANGULAR HORN (PYRAMIDAL)

343

7-1.2 Optimum Rectangular Horn A rectangular horn has extra parameters, which we can use to design various optimum horns. Given a desired gain, we can design any number of horns with the same gain. Any optimum design depends on the requirements. Without any particular requirements, we will pick an antenna with equal E- and H -plane 3-dB beamwidths [9], but even this does not determine the design totally. If we pick a constant slant radius and vary the aperture width, the gain increases with increasing aperture width, but the quadratic phase error loss increases faster and produces a maximum point. The maximum occurs in the two planes at approximately constant phase deviations independent of the slant radius: Sh = 0.40

Se = 0.26

(7-8)

At these points we read the 3-dB points from Tables 7-1 and 7-2: H sin θ = 0.4735 λ

W sin θ = 0.6951 λ

On dividing these equations to eliminate the constant sin θ in the two planes, we derive the ratio of width to height to give a constant 3-dB beamwidth in the two planes for this optimum point: H = 0.68 (7-9) W The ratio depends on the beamwidth level. For 10-dB beamwidths, H = 1.00 W

(7-10)

These values of S determine the efﬁciency of the optimum horn. We read the PEL of the quadratic phase distribution from Table 4-42 by using a cosine distribution for the H -plane and a uniform distribution for the E-plane. The H -plane distribution has an ATL of 0.91 dB. PELh = 1.14 dB

PELe = 1.05 dB

or an efﬁciency of 49%: gain =

ATL = 0.91 dB

4πH W 0.49 λ2

We solve for H and W for a given gain, since we know the ratio between them [Eq. (7-9)]: W = λ

gain 4π(0.68)(0.49)

W = 0.489 gain λ

H = λ

gain(0.68) 4π(0.49)

H = 0.332 gain λ

(7-11)

344

HORN ANTENNAS

We combine Eqs. (7-8) and (7-11) to calculate slant radiuses: Rh = 0.0746 · gain λ

Re = 0.0531 · gain λ

(7-12a, b)

If, given gain, we use Eqs. (7-11) and (7-12) to design a horn, the dimensions will not be practical with an arbitrary input waveguide. The axial lengths from the waveguide to the aperture must be equal in the E- and H -planes, so the horn will meet the waveguide in a single plane. Given waveguide dimensions a and b, the axial lengths are 2 W −a W H −a H2 Rh2 − Re2 − Le = (7-13a, b) Lh = W 4 H 4 We have a choice between retaining the E- or H -plane slant radius given by Eq. (7-12) and forcing the other radius to give the same axial length. The primary factor affecting gain is the aperture dimensions, which we retain from Eq. (7-11). The slant radius is secondary. We retain the H -plane radius calculated from Eq. (7-12) and modify the E-plane radius. Modifying the H -plane radius would give us a second optimum design: H (H − b)2 (7-14) L2 + Re = H −b 4 To obtain the proper gain, we must iterate, since we cannot use both Eq. (7-12). Design the horn by using Eqs. (7-11), (7-12a), (7-13a), and (7-14). Calculate the gain from the dimensions and obtain a new design gain from Gd,new =

Grequired Gd,old Gactual

(7-15)

where Grequired is the required gain, Gactual is the actual gain, and Gd,old is the old design gain. Example Design a horn fed from WR-90 waveguide to have 22 dB of gain at 10 GHz. The waveguide dimensions are 2.286 cm × 1.016 cm (0.9 in. × 0.4 in.) and Greq = Gd = 1022/10 = 158.5. On substituting in Eq. (7-11), we calculate aperture dimensions: W = 18.47 cm and H = 12.54 cm. From Eqs. (7-12a) and (7-13a), Rh = 35.47 cm and L = 30.01 cm. To get the same axial length in the E-plane [Eq. (7-14)], Re = 33.25 cm. We now calculate gain and compare it with the gain required. The amplitude taper loss and phase error loss in the H -plane remain constant, since Sh is ﬁxed. ATL = 0.91 dB

PELh = 1.14 dB at Sh = 0.40

Calculate Se : H2 = 0.197 PELe = 0.60 dB (Table 4-42) 8λRe 4πH W Gactual (dB) = 10 log − ATL − PELh − PELe = 22.45 dB λ2 = 175.8 Se =

RECTANGULAR HORN (PYRAMIDAL)

345

We must pick a new design gain [Eq. (7-15)]: Gd,new =

158.5(158.5) = 142.9 175.8

A second iteration with this gain gives the following dimensions: W = 17.54 cm

H = 11.91 cm

L = 26.75 cm

Rh = 31.98 cm

Re = 29.84 cm Se = 0.198 PELe = 0.60 4πH W − 0.91 − 1.14 − 0.60 = 22.00 dB gain = 10 log λ2 We obtain the gain desired, but the 3-dB beamwidths are only approximately equal: H -plane: 13.66◦ , E-plane: 13.28◦ . Scales 7-1 to 7-3 provide the dimensions of the optimum rectangular horn for a given gain. During their generation, a waveguide with a 2 : 1 aspect was used, but they are close to the proper values for nearby aspects. They design horns to within 0.1 dB. These scales produce short rapidly ﬂaring horns for low-gain antennas. In these cases it is better to deviate from the optimum design that gives the lightest horn for a given gain and design a horn with a small value of S. Scales 7-4 to 7-6 give designs for S = 0.1 to be used for low-gain designs. Aperture Width, l

Optimum Rectangular Horn Gain, dB

SCALE 7-1 Aperture width of an optimum pyramidal horn for a 2 : 1-aspect waveguide.

Aperture Height, l

Optimum Rectangular Horn Gain, dB

SCALE 7-2 Aperture height of an optimum pyramidal horn for a 2 : 1-aspect waveguide. Aperture Length, l

Optimum Rectangular Horn Gain, dB

SCALE 7-3 Axial length of an optimum pyramidal horn for a 2 : 1-aspect waveguide.

346

HORN ANTENNAS Aperture Width, l

S = 0.1 Rectangular Horn Gain, dB

SCALE 7-4 Aperture width of a pyramidal horn with S = 0.1 connected to a 2 : 1 waveguide. Aperture Height, l

S = 0.1 Rectangular Horn Gain, dB

SCALE 7-5 Aperture height of a pyramidal horn with S = 0.1 connected to a 2 : 1 waveguide. Aperture Length, l

S = 0.1 Rectangular Horn Gain, dB

SCALE 7-6 Axial length of a pyramidal horn with S = 0.1 connected to a 2 : 1 waveguide.

7-1.3 Designing to Given Beamwidths The beamwidths in the two planes of a rectangular horn can be designed independently. The axial lengths in the two planes must be equal if the design is to be realizable, but the aperture width and height can be adjusted to give the desired beamwidths. We pick S in one plane and then vary S in the other plane to produce the required beamwidth and the same axial length as in the ﬁrst plane. Since the ﬁrst S is arbitrary, the design is not unique, but in only a limited range of S will designs be realizable. Example Design a rectangular horn for the following 10-dB beamwidths: 30◦ H plane and 70◦ E-plane at 7 GHz using a 3.5-cm × 1.75-cm waveguide. Since the H -plane has the narrower beamwidth and therefore the wider aperture, we use it to determine length. Pick Sh = 0.20 (an arbitrary choice). The obliquity factor at 15◦ is 0.15 dB. When using Table 7-1 we must design for wider than a 30◦ beamwidth to compensate for the obliquity factor: 10.15 ◦ ◦ BWd = (30 ) = 30.22 10 The horn width to provide that beamwidth is 1.0949 W = = 4.200 λ sin(BWd /2)

RECTANGULAR HORN (PYRAMIDAL)

W = 18.00 cm at 7 GHz

347

Rh = 47.25 [Eq. (7.2)]

Use Eq. (7-13a) to determine the axial length: Lh = L = 37.36 cm. Because the Eplane beamwidth is wider than the H -plane beamwidth, the E-plane aperture will be smaller. We try a smaller value of Se than Sh for our initial guess: Se = 0.04. The obliquity factor at 35◦ adds 0.82 dB to the pattern loss and requires a larger design beamwidth. 10.82 ◦ ◦ BWd = 70 = 72.82 10 H 0.7405 = = 1.248 (Table 7-2) H = 5.246 λ sin(BWd /2) We use Eqs. (7-2) and (7-13b) to calculate slant radius and axial length: Re = 20.84 cm and Le = 13.90 cm. The axial lengths in the two planes do not match, so we pick a smaller Se because the E-plane is shorter than the H -plane beamwidth. At Se = 0.02, H = 5.337 cm, Re = 41.54 cm, and Le = 27.86 cm. Le has doubled when Se changes from 0.04 to 0.02, but H changes by only 0.01 cm. We change our method and pick H = 5.33 cm and force Re to give the same axial length as the H -plane: Re = 55.69 cm [from Eq. (7-14)] or Se = 0.0149. 7-1.4 Phase Center We deﬁne the phase center as the point from which it appears that an antenna radiates spherical waves. Measurements show that the phase center is seldom a unique point in a plane, but depends on the pattern angle. The E- and H -plane phase centers will also be unequal in general. Usually, they are extremes, and the axial position varies elliptically between the planes. Even with the phase-center location fuzzy, it is a useful point. We place the phase center of a feed at the focus of a parabolic reﬂector to minimize the reﬂector aperture phase error loss. An aperture with a quadratic phase distribution appears to be radiating from a point behind the aperture. Without quadratic phase error (S = 0), the phase center is located at the aperture plane. Increasing S moves the phase center toward the apex of the horn. Muehldorf [10] has calculated the phase-center location as a function of S, and Table 7-3 summarizes his results. The phase center located inside the aperture is given as a ratio of the slant length. TABLE 7-3 Phase-Center Axial Location of a Rectangular Horn (TE10 Mode) Behind the Aperture as a Ratio of the Slant Radius S 0.00 0.04 0.08 0.12 0.16 0.20 0.24

H -Plane Lph /Rh

E-plane Lph /Re

0.0 0.0054 0.022 0.048 0.086 0.134 0.191

0.0 0.011 0.045 0.102 0.182 0.286 0.416

S 0.28 0.32 0.36 0.40 0.44 0.48 0.52

H -plane Lph /Rh

E-plane Lph /Re

0.258 0.334 0.418 0.508 0.605 0.705 0.808

0.572 0.755

348

HORN ANTENNAS

Example above.

Calculate the phase-center location for the beamwidth design example Sh = 0.20

Rh = 47.25 cm

Se = 0.015

Re = 55.69 cm

From Table 7-3, we interpolate H -plane phase center = 0.134(47.25 cm) = 6.33 cm E-plane phase center = 0.004(55.69 cm) = 0.22 cm The difference in the phase-center location in the two planes is 1.43λ. As in the example, antennas with widely differing beamwidths will have widely separated phase centers.

7-2 CIRCULAR-APERTURE HORN With a circular-aperture horn, we lose independent control of the beamwidths in the principal planes. The circular waveguide can support any orientation of the electric ﬁeld and thereby allows any polarization in the horn. We use the same aperture ﬁeld method as with the rectangular horn and the waveguide mode determines the aperture amplitude. The cone of the horn projects to a point in the feed waveguide where we assume a point source radiating to the aperture. The aperture phase is approximately quadratic. The waveguide ﬁelds are given by x11 ρ E0 Eρ = J1 cos φc ρ a E0 x11 x11 ρ sin φc J1 E φc = − (7-16) a a where J1 is the Bessel function, ρ the radial component in the waveguide, a the radius, and φc the cylindrical coordinate. x11 (1.841) is the ﬁrst zero of J1 (x). Equation (7-16) has its maximum electric ﬁeld directed along the φc = 0 plane. We add the quadratic phase factor to Eq. (7-16) and calculate the Fourier transform on the circular aperture to determine the far ﬁeld. The direction of the electric ﬁeld changes from point to point in the aperture. For a given direction (θ , φc ) we must project the ﬁelds in the aperture onto the θˆ and φˆ directions before integrating over the aperture:

2π a J1 (x11 ρ/a) θˆ žρˆ θˆ žφˆ c x11 x11 ρ sin φc Eθ = E0 cos φc − J ρ cos θ a 1 a cos θ 0 0

ρ 2 (7-17) × ρ exp j kρ sin θ cos(φ − φc ) − 2πS dρ dφc a

CIRCULAR-APERTURE HORN

x11 ρ sin φc φˆ · φˆ c Eφ = E0 a 0 0 ρ 2 × ρ exp j kρ sin θ cos(φ − φc ) − 2πS (7-18) dρ dφc a

2π

a

349

x J1 (x11 ρ/a) cos φc φˆ · ρˆ − 11 J1 ρ a

θˆ · ρˆ = cos θ (cos φ cos φc + sin φ sin φc ) θˆ · φˆ c = cos θ (sin φ cos φc − cos φ sin φc ) φˆ · ρˆ = cos φ sin φc − sin φ cos φc φˆ · φˆ c = cos φ cos φc + sin φ sin φc By a suitable change of variables in the integrals, universal radiation patterns can be generated for the E-and H -planes (Figures 7-5 and 7-6). The equality of S in the two planes ties the curves together. The axis is the k-space variable. We can calculate a few pattern points for a given horn with those curves if we remember to add the obliquity factor to the values taken from the curves.

0.9

S = 0.7

0.8 S=

a2 2lR

0.7 Relative Field Intensity

0.6 0.6 0.5 0.5 0.4 0.4

0.3

0.3 0.2

0.2

0.1 0.1 0 0 1

2

3

4

5

6 7 2pa sin q l

8

9

10

11

12

FIGURE 7-5 E-plane universal pattern of a circular, TE11 mode. (From T. Milligan, Universal patterns ease circular horn design, Microwaves, vol. 20, no. 3, March 1981, p. 84.)

350

HORN ANTENNAS

0.9

0.8 S=

a2 2lR

Relative Field Intensity

0.7 0.6

S = 0.7

0.5 0.6 0.4 0.5 0.3 0.4 0.2 0.2

0.1

0

0 1

2

3

4

5

6 7 2pa sin q l

8

9

10

11

12

FIGURE 7-6 H -plane universal pattern of a circular, TE11 mode. (From T. Milligan, Universal patterns ease circular horn design, Microwaves, vol. 20, no. 3, March 1981, p. 84.)

Example A horn has an aperture radius of 12 cm and a slant radius of 50 cm. Compute the pattern level at θ = 20◦ at 5 GHz. From Figures 7-5 and 7-6 we interpolate the pattern voltage level: H -plane level = 0.18 (−14.7 dB)

E-plane level = 0.22

(−13.1 dB)

The obliquity factor is 20 log[(1 + cos 20◦ )/2] = −0.3 dB, and the plane level at 20◦ becomes H -plane level = −15 dB E-plane level = −13.4 dB 7-2.1 Beamwidth Table 7-4 lists the 3- and 10-dB points from Figures 7-5 and 7-6. We can use them to compute beamwidths from dimensions. Example Calculate 10-dB beamwidths of the horn in the example above. S = 0.24, a = 12 cm, and λ = 6 cm.

351

CIRCULAR-APERTURE HORN

TABLE 7-4

Circular-Horn Beamwidths, TE11 Mode (2πa/λ) sin θ 3 dB

10 dB

S

E-Plane

H -Plane

E-Plane

H -Plane

ATL + PEL (dB)

0.00 0.04 0.08 0.12 0.16 0.20 0.24 0.28 0.32 0.36 0.40 0.44 0.48 0.52 0.56 0.60 0.64 0.68 0.72

1.6163 1.6175 1.6212 1.6273 1.6364 1.6486 1.6647 1.6855 1.7123 1.7471 1.7930 1.8552 1.9441 2.0823 2.3435 3.4329 4.3656 4.8119 5.1826

2.0376 2.0380 2.0391 2.0410 2.0438 2.0477 2.0527 2.0592 2.0676 2.0783 2.0920 2.1100 2.1335 2.1652 2.2089 2.2712 2.3652 2.5195 2.8181

2.7314 2.7368 2.7536 2.7835 2.8296 2.8982 3.0024 3.1757 3.5720 4.6423 5.0492 5.3139 5.5375 5.7558 6.0012 6.3500 7.6968 8.4389 8.8519

3.5189 3.5211 3.5278 3.5393 3.5563 3.5799 3.6115 3.6536 3.7099 3.7863 3.8933 4.0504 4.2967 4.6962 5.2173 5.6872 6.0863 6.4622 6.8672

0.77 0.80 0.86 0.96 1.11 1.30 1.54 1.82 2.15 2.53 2.96 3.45 3.99 4.59 5.28 5.98 6.79 7.66 8.62

From Table 7-4 we read the k-space values at 10 dB: 2πa sin θh = 3.6115 λ 2πa sin θe = 3.0024 E-plane k-space = λ 3.6115 3.0024 ◦ ◦ = 33.40 = 27.65 BWh = 2 sin−1 BWe = 2 sin−1 4π 4π H -plane k-space =

We must add the obliquity factor to the 10-dB universal pattern level:

BWh10 =

1 + cos 16.7◦ : 2 1 + cos 13.8◦ : 2 10 ◦ ◦ 33.40 = 33.10 10.18

0.18 dB

H -plane

0.13 dB

E-plane

BWe10 =

10 ◦ ◦ 27.65 = 27.48 10.13

We can also use Table 7-4 to design a horn to a given beamwidth, but we can design to only one plane. Any number of horns can be designed to a given beamwidth, since S is an independent parameter. Table 7-4 lists the combined amplitude taper loss and phase error loss as a function of S for the circular horn. With this table we can easily estimate the gain of a given horn or design a horn to a given gain.

352

HORN ANTENNAS

Example Compute the gain of a horn with a 12-cm aperture radius and 50-cm slant radius at 5 GHz. From the examples above we read S = 0.24 and λ = 6 cm: πD − GF where GF = ATL + PEL (dB) λ 24π = 20 log − 1.54 = 20.4 dB 6

gain = 20 log

(7-19)

Example Design a circular horn at 8 GHz with a gain of 22 dB. The quadratic phase distribution constant S is arbitrary. Pick S = 0.20. Rearrange Eq. (7-19) to ﬁnd the diameter: λ · 10(gain+GF)/20 π 3.75 · 10(22+1.30)/20 = 17.45 cm = π D2 = 50.77 cm R= 8λS

D=

(7-20)

We can determine an optimum circular horn in the sense of minimizing the slant radius at a given gain. When we plot gain as a function of aperture radius for a ﬁxed slant radius, we discover a broad region in which the gain peaks. By plotting a series of these lines with a voltage gain ordinate, we see that a single line corresponding to S = 0.39 can be drawn through the peaks. This is our optimum with GF = 2.85 dB (ATL + PEL). Example Design an optimum horn at 8 GHz with gain of 22 dB. From Eq. (7-20), D=

3.75 · 10(22+2.85)/20 = 20.86 cm π

R=

20.862 D2 = = 37.2 cm 8λS 8(3.75)(0.39)

The optimum is quite broad. A horn designed with S = 0.38 has a 0.07-cm-longer slant radius and a 0.25-cm-smaller aperture diameter. 7-2.2 Phase Center The phase-center location behind the aperture plane is a function of S. Table 7-5 lists the phase-center location as a ratio of the slant radius. As S increases, the phase center migrates toward the horn apex and the difference between the phase-center locations in the E- and H -planes increases. Example Use Table 7-5 to compute phase-center locations in the E- and H -planes for the circular horns of the preceding two examples. R = 50.77 cm S = 0.20 H -plane phase center = 0.117(50.77) = 5.94 cm E-plane phase center = 0.305(50.77) = 15.48 cm

CIRCULAR (CONICAL) CORRUGATED HORN

353

TABLE 7-5 Phase-Center Axial Location of a Circular-Waveguide Horn TE11 Mode Behind the Aperture as a Ratio of the Slant Radius S

H -Plane Lph /Rh

E-Plane Lph /Re

0.00 0.04 0.08 0.12 0.16 0.20 0.24

0.0 0.0046 0.018 0.042 0.075 0.117 0.171

0.0 0.012 0.048 0.109 0.194 0.305 0.416

S

H -Plane Lph /Rh

E-Plane Lph /Re

0.28 0.32 0.36 0.40 0.44 0.48

0.235 0.310 0.397 0.496 0.604 0.715

0.603 0.782 0.801 0.809 0.836 0.872

The phase centers differ by 2.5 wavelengths at 8 GHz. The optimum horn has the dimensions R = 37.2 cm S = 0.39 H -plane phase center = 0.471(37.2) = 17.5 cm E-plane phase center = 0.807(37.2) = 30.0 cm The phase centers of the optimum horn differ by 3.3 wavelengths. The difference increases with increasing S.

7-3 CIRCULAR (CONICAL) CORRUGATED HORN Normal smooth-wall horns present problems that can be eliminated by corrugating the walls. Many applications require dual linear or circular polarizations. The horn aperture must be square or circular and the beamwidths in the two planes are unequal. When the smooth-wall horn feeds a reﬂector, we have astigmatism (unequal phase centers in orthogonal planes). A horn has higher sidelobes in the E-plane than in the H -plane. Finally, the diffraction off E-plane walls causes backlobes. The aperture theory fails to predict them, but measurement or a GTD analysis shows them. Corrugating the walls can prevent all these problems. Figure 7-7 shows the cross sections of two types of corrugated horns. The smallﬂare horn (a) is nominally the corrugated horn, and the wide-ﬂare horn (b) is the scalar horn of Simmons and Kay [11]. Many papers on corrugated horns appear in Section VI of Love’s [1] collection of papers. Thomas [12] provides a good design summary in a topic with many papers. The corrugations that extend circumferentially should be cut normal to the slant radius as in (b), but they may be cut normal to the axis (a) for small ﬂare angles. Horns can be built either way, but when cut normal to the axis, the depth is different on the two sides. The corrugated wall presents the same boundary conditions to the electric and magnetic ﬁelds when it is capacitive (slots λ/4 to λ/2 deep). When excited in the transition between the smooth-wall waveguide and the corrugated-wall cone, the TE11 and TM11 waveguide modes, have equal phase velocities. The combination of these

354

HORN ANTENNAS

t w

p E11

g a1 q0 d1 R

a

∆ Rq0 d

w t

(a)

a0 q0 0

g

p

Rq0

∆ d1

R

d (b)

FIGURE 7-7 (a) Corrugated horn; (b) scalar horn. (From [12], 1978 IEEE.)

modes forms the hybrid mode HE11 when the mode phases are equal. When the modes are out of phase by 180◦ , the hybrid mode is denoted EH11 . The ratio of the modes is called γ , and γ = 1 for the balanced HE11 mode. γ = 0 corresponds to having only the TM11 mode and γ = ∞ to having only the TE11 mode. γ = 1 occurs when the corrugation depth is λ/4, but the horn parameters vary slowly with changing γ [13]. We consider only γ = 1. Changing γ has its biggest effect on the cross-polarization [12,14]. When γ = 1, the amplitude of the aperture ﬁelds is given by [15] x ρ 01 cos φc Eρ = J0 a (7-21) x ρ 01 sin φc Eφ = −J0 a where x01 = 2.405 is the ﬁrst zero of J0 (x), the Bessel function. The ﬁelds vanish at the walls and prevent edge diffractions that produce sidelobes and backscatter. The lower-order slope diffractions still produce small sidelobes and backlobes, but we get H -plane-type lobes in all planes. In amplitude the aperture ﬁelds are symmetrical about the axis and all patterns through the cone axis are identical. A Huygens source analysis of the aperture ﬁelds with a quadratic phase distribution produces Figure 7-8, valid when the 10-dB beamwidth is less than 74◦ [12]. For greater beamwidths the ﬂange changes the beamwidths of the small-aperture horn in the two planes and we should use the scalar horn. Table 7-6 lists the 3-, 10-, and 20-dB points

CIRCULAR (CONICAL) CORRUGATED HORN

355

S=1 0.9

0.9

0.8 0.8

S=

a2 2lR

0.7 Relative Field Intensity

0.7 0.6

0.5 0.6 0.4 0.5 0.3 0.4

0.2

0.3 0.1 0.2 0 2

4

0 6 2pa sin q l

8

10

12

FIGURE 7-8 Universal pattern of a circular corrugated horn: HE11 mode. TABLE 7-6 Mode

Resonant Circular Corrugated Horn Beamwidth Points (2πa/λ) sin θ , HE11

S

3 dB

10 dB

20 dB

ATL + PEL (dB)

S

3 dB

10 dB

20 dB

ATL + PEL (dB)

0.00 0.04 0.08 0.12 0.16 0.20 0.24 0.28 0.32 0.36 0.40 0.44 0.48

2.0779 2.0791 2.0827 2.0887 2.0974 2.1088 2.1234 2.1415 2.1637 2.1906 2.2231 2.2624 2.3103

3.5978 3.6020 3.6150 3.6371 3.6692 3.7129 3.7699 3.8433 3.9372 4.0572 4.2112 4.4090 4.6578

4.6711 4.6878 4.7405 4.8387 5.0061 5.3052 5.8451 6.3379 6.6613 6.9179 7.1534 7.3939 7.6633

1.60 1.62 1.66 1.73 1.83 1.96 2.12 2.30 2.52 2.76 3.04 3.34 3.68

0.52 0.56 0.60 0.64 0.68 0.72 0.76 0.80 0.84 0.88 0.92 0.96 1.00

2.3688 2.4411 2.5317 2.6469 2.7966 2.9946 3.2597 3.6061 4.0189 4.4475 4.8536 5.2331 5.5984

4.9532 5.2720 5.5878 5.8913 6.1877 6.4896 6.8134 7.1788 7.6042 8.0852 8.5773 9.0395 9.4701

7.9936 8.4261 8.9472 9.4352 9.8514 10.2337 10.6250 11.0735 11.6356 12.2658 12.8236 13.3059 13.7706

4.04 4.44 4.86 5.31 5.79 6.30 6.83 7.39 7.96 8.54 9.13 9.72 10.29

356

HORN ANTENNAS

from Figure 7-8. We use the table to ﬁnd beamwidths of given horns or design to given beamwidths. Table 7-6 also lists the sum of ATL and PEL(GF) for various S. We estimate gain or design to a given gain with this listing. Example Calculate 10-dB beamwidth and gain of a corrugated conical horn with an aperture radius of 12 cm and a slant radius of 50 cm at 5 GHz: S=

a2 122 = = 0.24 2λR 2(6)(50)

From Table 7-7 we read the k-space value at 10 dB for S = 0.24: 2πa sin θ = 3.7699 λ

or

◦

θ = 17.46

We include the obliquity factor, since the pattern loss will be greater than 10 dB at θ = 17.46◦ : 10 ◦ ◦ ◦ (1 + cos 17.46 )/2 : −0.20 dB BW10 = 34.92 = 34.57 10.20 A smooth-wall horn with the same dimensions has a similar H -plane beamwidth (33.4◦ ). We calculate gain from the distribution losses and aperture area: GF = ATL + PEL = 2.12 dB

gain = 20 log

πD − GF = 19.86 dB λ

The smooth-wall horn has about 0.5 dB more gain (20.4 dB). Example Design a circular corrugated-wall horn at 8 GHz with a gain of 22 dB. We use Eq. (7-20) with the GF from Table 7-6. Choose S = 0.20 (arbitrary): 3.75 cm · 10(22+1.96)/20 = 18.83 cm π D2 = 59.10 cm R= 8λS

D=

TABLE 7-7 Phase-Center Axial Location of a Circular Corrugated Horn (HE11 Mode) Behind the Aperture as a Ratio of the Slant Length S

Lp /R

S

Lp /R

0.00 0.04 0.08 0.12 0.16 0.20 0.24 0.28 0.32

0.0 0.005 0.020 0.045 0.080 0.124 0.178 0.240 0.310

0.36 0.40 0.44 0.48 0.52 0.56 0.60 0.64 0.68

0.386 0.464 0.542 0.614 0.673 0.718 0.753 0.783 0.811

CIRCULAR (CONICAL) CORRUGATED HORN

357

The phase center, like that of other horns, starts at the aperture for S = 0 (R = ∞) and moves toward the apex as S increases. Table 7-7 lists the phase-center location as a ratio of the slant radius. Because the aperture distribution is the same along all radial lines of the aperture, the phase center is the same in all planes. We measure some variation, since the balance between modes will not be perfect, and some higher-order modes will be generated. The phase center moves least over a frequency band for long horns (small S). A wide-ﬂare-angle horn has its phase center at the apex and is better described as a scalar horn. 7-3.1 Scalar Horn A scalar horn has a wide half-ﬂare angle, θ0 . Its beamwidth depends on the half-ﬂare angle. Since the phase distribution in the aperture is large, there is an optimum diameter for a given ﬂare angle. Table 7-8 lists the optimum diameter versus the ﬂare angle. The beamwidth is approximately a linear function of the half-ﬂare angle, θ0 , for the optimum horn: BW3 dB = 0.74θ0 BW10 dB = 1.51θ0

(7-22)

BW20 dB = 2.32θ0 A scalar horn has a wider bandwidth as a reﬂector feed than that of a small-ﬂare-angle corrugated horn, because the phase center is ﬁxed at the horn apex. 7-3.2 Corrugation Design The corrugations present a capacitive reactance to the passing wave. When a corrugated surface is inductive, it will support surface waves. The depth of corrugations must be between λ/4 and λ/2. Less than λ/4 or greater than λ/2, it is inductive. Between 3λ/2 and λ it will be capacitive again, but this second passband is seldom used. A quarterwavelength corrugation depth balances the two modes and gives the best results. The corrugations need be only λ/4 at the aperture. Before the aperture we ﬁnd it better to deepen the slots. Quarter-wavelength-deep corrugations mismatch the horn in the transition region, where the TM11 mode is generated from the TE11 mode and depths approaching λ/2 have the least effect on match. TABLE 7-8 Optimum Diameter of a Scalar Horn Half-Flare Angle, θ0 (deg)

Aperture Diameter (λ)

Half-Flare Angle, θ0 (deg)

Aperture Diameter (λ)

15 20 25 30 35 40

10.5 8.0 6.4 5.2 4.5 3.9

45 50 55 60 65 70

3.5 3.0 2.8 2.6 2.4 2.3

Source: [16, p. 429].

358

HORN ANTENNAS

To design for a particular band, limited to about 1.5 : 1 for a good match, we design with tapered corrugation depths. We make the corrugations λ/4 deep at the aperture at the low-frequency end. The high-frequency band edge determines the corrugation depth at just short of λ/2 in the throat. The horn needs at least four corrugations per wavelength along the slant radius. The high-frequency end determines the corrugation spacing. The ﬁrst few corrugations can be used to match the horn to the waveguide, and we can improve the match by shaping the corrugation widths [14]. The slots should be as wide as practical spacers will allow. Mechanical considerations, such as shock and vibration, will determine the limits on the thinness of spacers, but corrugations greatly increase the strength of the bell. The circular geometry of the horn changes the corrugation depth necessary for the balanced HE11 mode from λ/4. An empirical formula for the depth is given by [17] 1 λ d = exp ka > 2 (7-23) 4 2.5 ka We increase the depth slightly at the horn aperture. 7-3.3 Choke Horns We can place corrugations in the ﬂanges of small-aperture horns and design widebeamwidth antennas with good pattern symmetry and low cross-polarization. The choke horn (Figure 7-9) is the limit of a scalar horn opened to θ0 = 90◦ . The corrugations consist of concentric rings about the aperture and are generally a quarter-wavelength deep. The best location for the corrugated rings may not be in the same plane as the aperture but instead somewhat behind as reported for a feed for f/D = 0.3 reﬂector [18;19, pp. 200–209]. The design for the reﬂector feed [18] used four corrugations. James [20] and Kumer [21] show that using only one corrugation is quite effective. More corrugations improve the design but only add marginally. Small apertures need the corrugations to reduce the cross-polarization that peaks in the diagonal planes between the E- and H -planes. We usually assume a Huygens source in the aperture plane of the horn. This approximation collapses as we shrink the aperture to achieve wide beamwidths. In the limit

FIGURE 7-9

Choke horn.

CORRUGATED GROUND PLANE

359

we have only a slot analyzed from magnetic currents replacing the electric ﬁeld in the aperture. The magnetic ﬁeld is ignored in the slot while a Huygens source assumption is that the ratio of the electric to magnetic ﬁeld is the same as the impedance of free space. Waveguides have high wave impedances, which implies small magnetic ﬁelds. To calculate the far ﬁeld, Eq. (2-23) must be used with the actual ratio of ﬁelds in the aperture instead of Eq. (2-24), with its Huygens source approximation. Restricting the aperture dimensions to achieve wide beamwidths will limit the bandwidth as well as the cross-polarization isolation, because reducing volume raises Q. 7-3.4 Rectangular Corrugated Horns We can design rectangular horns with corrugated walls, but we only need to cut corrugations in the E-plane walls to produce a cosine distribution in the E-plane. Only for dual polarization do we need corrugations in the H -plane walls. We analyze the horn as having an H -plane distribution (cosine) in both planes and use the results of Section 7-1. The larger aperture dimension in the diagonal plane decreases the beamwidth slightly, but the rectangular horn still provides an acceptable design. Both planes have a linear amplitude taper loss of 0.91 dB. We use the cosine column of Table 4-42 for the quadratic phase error loss. We design beamwidths using Table 7-1. Equalizing the distributions in both planes of square horns results in equal phase centers, given by Table 7-3 (H -plane). Example Compute the gain of a square corrugated horn with an aperture width of 24 cm and a slant radius of 50 cm at 5 GHz. From Eq. (7-1), W2 242 S= = = 0.24 8λR 8(6)(50) We use the cosine column of Table 4-42 for the phase error loss: PELx = PELy = 0.42 (cosine). The amplitude taper loss is the same in both planes: 0.91 dB. 4πW 2 − ATLx − ATLy − PELx − PELy λ2 = 23.03 − 0.91 − 0.91 − 0.42 − 0.42 = 20.4 dB

Gain = 10 log

A circular corrugated horn with a diameter equal to the width and having the same slant radius has a gain of 19.9 dB or 0.5 dB less. The larger aperture area increases the gain over the circular horn. 7-4 CORRUGATED GROUND PLANE The corrugated surface (Figure 7-10) supports surface waves (TM) when the slot depth is less than λ/4 (inductive). As with the corrugated horn, we assume many slots per wavelength along the direction of propagation. The ﬁelds attenuate exponentially above the ends of the corrugations in a surface wave. We derive the ﬁelds from a potential function −2πbx ψ = A1 exp (7-24) exp(−j kz z) λ

360

HORN ANTENNAS x g

t d z

FIGURE 7-10 Corrugated ground plane.

above the corrugations, where A1 is an amplitude constant, x the distance out of the corrugations, and α(= 2πb/λ) the attenuation constant of the ﬁelds above the corrugations. We expand the ﬁelds and take the ratio of the z-directed electric ﬁeld to the y-directed magnetic ﬁeld to ﬁnd the wave impedance into the corrugated surface: 1 2πb 2 A1 2πb (k 2 − kz2 )ψ = − exp − exp(−j kz z) j ωε0 λ j ωε0 λ 2πb ∂ψ 2πb = A1 exp − Hy = − exp(−j kz z) ∂x λ λ √ j (2πb/λ) µ0 Ez j (2πb/λ) j (kb) = = √ = Z−x = η = j bη (7-25) √ Hy ωε0 ω ε0 µ 0 ε0 k Ez =

where η is the impedance of free space and b is related to α [Eq. (7-24)]. The structure must present this impedance to the wave. The corrugated surface is a parallel-plate transmission line to Ez , and it presents a per unit length impedance of Zc = j η tan kd

(7-26)

where d is the corrugation depth. We equate Eqs. (7-25) and (7-26) to determine the constant b: b = tan kd (7-27) We use Eq. (7-27) in Eq. (10-16) for the relative propagation constant: P = 1 + b2 = 1 + tan2 kd

(7-28)

We include the effect of the corrugation thickness by averaging between the parallelplate impedance and the zero impedance along the corrugation edges. Equation (7-28) becomes 2 g P = 1+ tan2 kd (7-29) g+t where g is the gap distance and t is the corrugation thickness. P increases without bound as the depth d approaches λ/4. The ﬁelds become tightly bound to the surface and attenuate rapidly to zero above the corrugations—the normal electric ﬁeld vanishes as in a corrugated horn E-plane wall. We design the depth of the corrugations by using d=

λ g+t tan−1 √ 2π g P2 − 1

(7-30)

CORRUGATED GROUND PLANE

361

When the corrugation depth approaches λ/4, the surface impedance [Eq. (7-26)] approaches inﬁnity and the tangential magnetic ﬁeld vanishes on the surface to create an artiﬁcial PMC (Section 2-3) for waves polarized along the z-axis. The reﬂection coefﬁcient is +1 instead of −1 for the PEC surface. Waves polarized along the y-axis encounter closely spaced corrugations that approximately produce a PEC surface with the usual metal wall reﬂection coefﬁcient of −1. Whereas a PEC reﬂects an incident circularly polarized wave with opposite sense of circular polarization, the artiﬁcial PMC (soft) surface [22, pp. 276–280] reﬂects the wave with the same sense of polarization. We can use these surfaces to shape the pattern of a wide-beamwidth circularly polarized antenna to narrow the beamwidth without generating the opposite sense polarization, which would be generated by metal walls. A ground plane covered with circular coaxial corrugations λ/4 deep reduces the edge diffraction that produces a large backlobe for a monopole antenna mounted in the center (Figure 5-23). The artiﬁcial soft wall causes the reduction of circumferential magnetic ﬁelds and the associated GTD diffraction (Section 2-7.11) [23]. This increases the forward gain by reducing the backlobe. It is not necessary to corrugate the entire top surface. Figure 7-11 illustrates a surface with only two coaxial corrugations around the outer rim. These reduce the backlobe for a dipole mounted over the ground plane without generating signiﬁcant cross-polarization from a pair of orthogonal dipoles fed for a circular polarization. Corrugating the entire surface would cause radiation of crosspolarization because the region below the dipole pair radiates oppositely sensed circular polarization. The corrugated surface reﬂects the same sense of circular polarization as incident. The PEC surface reverses the sense of circular polarization of the reﬂected wave and both waves add. The corrugations only reduce the backlobe. The choke horn uses the same type of corrugations to reduce the backlobe radiated from the small-diameter horn aperture. The corrugations can be placed radially below the ground plane by using shortcircuited radial transmission lines (Figure 7-12) and also reduce the backlobe. We

FIGURE 7-11

Ground plane with two coaxial corrugations to reduce edge diffraction.

FIGURE 7-12 Ground plane with short-circuited radial transmission-line corrugations.

362

HORN ANTENNAS

TABLE 7-9 Radial Transmission Outer Choke Depth at Resonance Outer Radius (λ)

Depth (λ)

Outer Radius (λ)

Depth (λ)

0.25 0.30 0.35 0.40 0.50 0.60

0.188 0.199 0.208 0.213 0.222 0.227

0.70 0.80 1.0 2.0 4.0

0.230 0.233 0.236 0.243 0.247

compute the reactance at the outer radius from the following equation, which uses Bessel and Neumann functions: X=j

ηb N0 (kr)J0 (kro ) − J0 (kr)N0 (kro ) 2πr J1 (kr)N0 (kro ) − N1 (kr)J0 (kro )

The short-circuited wall is located at radius ro and the spacing between the plates is b. Reactance X grows rapidly as r approaches resonance. For a large outer radius the difference r − ro approaches λ/4 but is less for a small radius. Table 7-9 gives the difference in radius versus the outer radius for resonance. Corrugations on the upper surface are more effective than radial corrugations, but the radial line chokes ﬁt easily behind a small ground plane. In both cases the corrugations enhance radiation behind the ground plane at frequencies below resonance λ/4 depths, because a surface wave is generated along the corrugations. Corrugated surfaces are useful structures because they can be used to enhance or reduce radiation, depending on their depth.

7-5 GAUSSIAN BEAM Corrugated horns and simple reﬂector feeds can be approximated with Gaussian beams. An inﬁnite circularly symmetrical Gaussian aperture distribution located in the x –y plane radiates a Gaussian beam along the z-axis. The radial exponent of the Gaussian distribution determines the spread of the wave as it propagates along the z-axis. We use the distribution to calculate the radiation pattern and then add the Huygens source (Section 2-2.2) for polarization. The analysis is divided into far- and near-ﬁeld approximations. The near-ﬁeld approximation consists of a paraxial wave. The Gaussian beam satisﬁes Maxwell’s equations by using the free-space Helmholtz equation and produces correct patterns when applied with physical optics (PO). The free-space Green’s function satisﬁes the Helmholtz equation: e−j kR /R. We derive the Gaussian beam from a point source placed at a complex position along the z-axis: z0 = −j b. A source at this position produces a Gaussian distribution in the z = 0 plane. 2 −ρ exp with ρ 2 = x 2 + y 2 W02

GAUSSIAN BEAM

363

W0 is the beam waist radius where amplitude has dropped by 1/e. We relate the waist radius W0 to the position b by [24, pp. 80–90] W02 =

2b k

where k =

2π λ

As the wave propagates along the z-axis, its amplitude retains the Gaussian distribution in the radial direction ρ but the waist spreads: z 2

W 2 (z) = W02 1 + b The waist surface is a hyperboloid with a ring focus at radius b located at z = 0. The wave amplitude reduces by the ratio of the waists and combines with the radial Gaussian distribution:

W0 ρ2 exp − 2 W (z) W (z) The phase of the paraxial (near-ﬁeld) wave has two terms. The ﬁrst is the normal z-directed wave phase exp (−j kz) and the second is a quadratic phase term that arises from the complex location of the point source at z = −j b. The quadratic phase term slant radius depends on the location along the z-axis: 2 b Rc (z) = z 1 + z The paraxial Gaussian beam has an additional slippage phase term ζ(z) = tan−1 (z/b). The paraxial Gaussian beam phase term is the sum

ρ2 + j ζ(z) exp −j kz − j k 2Rc (z) The constant phase (eikonal) surfaces between the hyperboloid amplitude surfaces are ellipsoids with a ring focus at radius b located at z = 0. At z = 0 the eikonal surface is planar. We combine the amplitude and phase terms for the complete paraxial Gaussian beam equation:

ρ2 θ W0 ρ2 − j E0 cos2 exp − 2 + j ζ(z) exp −j kz − j k 2 W (z) W (z) 2Rc (z) × (θˆ cos φ − φˆ sin φ)

(7-31)

The Huygens source polarization for an x-directed wave [Eq. (1-38)] and the obliquity factor [Eq. (2-14)] have been added to Eq. (7-31). We determine the constant E0 by equating the radiation between this paraxial beam and the far-ﬁeld expression for a Gaussian beam with a given input power. The recommended distance to equate the two representations is z = 200W02 /λ. We calculate the far-ﬁeld Gaussian beam by substituting the point source position into e−j kR /R and approximating R with a far-ﬁeld expression [25, pp. 96–106]: R=

x 2 + y 2 + z2 − b2 + j 2bz = r 2 − b2 + j 2br cos θ

(7-32)

364

HORN ANTENNAS

In the far ﬁeld we can ignore b2 and expand Eq. (7-32) in a Taylor series and retain the ﬁrst two terms, which reduces e−j kR /R to ekb cos θ e−j kr /r. We combine this term with the Huygens source radiation to produce the far-ﬁeld Gaussian beam equation for an x-directed linear polarization in the aperture normalized at θ = 0 to directivity: E(r, θ, φ) =

e−j kr P0 · directivity · η θ cos2 ekb(cos θ−1) (θˆ cos φ − φˆ sin φ) 4π 2 r

(7-33)

The directivity is found by integrating the pattern of Eq. (7-33): directivity =

4(2 kb)2 2(2 kb) − 2 + 1/(2 kb) − e−2(2 kb) /(2 kb)

(7-34)

Scale 7-7 gives the relationship between gain and the 10-dB beamwidth for a Gaussian beam. Given the beamwidth (BW) at a given level L(dB), we solve Eq. (7-33) for the complex-plane point source position b: b=

2 log[cos(BW/4)] + |L/20| k[1 − cos(BW/2)] log e

(7-35)

Scale 7-8 relates Gaussian beam half-depth of focus, b, to its 10-dB beamwidth, and Scale 7-9 gives the minimum waist diameter. We simplify the expression for the Gaussian beam for small angles by expanding cos θ in a Taylor series cos θ ≈ 1 − θ 2 /2, which reduces Eq. (7-33): e−j kr θ 2 E(r, θ, φ) = E0 cos2 e−(θ/θ0 ) (θˆ cos φ − φˆ sin φ) 2 r

(7-36)

10-dB Beamwidth (degrees)

Gaussian Beam Gain, dB

SCALE 7-7

Gaussian beam gain compared to a 10-dB beamwidth.

Half Depth of Focus b, l

Gaussian Beam, 10-dB Beamwidth (degrees)

SCALE 7-8

Gaussian beam half-depth of focus, b, compared to a 10-dB beamwidth.

RIDGED WAVEGUIDE HORNS

365

Minimum Waist Diameter 2W, l

Gaussian Beam, 10-dB Beamwidth (degrees)

SCALE 7-9

Gaussian beam minimum waist diameter compared to a 10-dB beamwidth.

The angle θ0 is the beam divergence [24, pp. 80–90], given by 2 θ0 = = kW0

2 kb

Equation (7-36) cannot be used beyond θ0 because it is based on a small-angle approximation. We can use a Gaussian beam to approximate the pattern of a corrugated horn [26, pp. 170–176]. The minimum waist is located behind the horn aperture Lp , the phase-center distance given the aperture radius a and the slant radius R: Lp =

R 1 + [2R/k(0.644a)2 ]2

(7-37)

Lp is the location of z = 0 of the Gaussian beam. The minimum waist radius W0 is given by W02 k 0.644a (7-38) b = W0 = 1 + [k(0.644a)2 /2R]2 2 For a 22 dB-gain corrugated horn, Eq. (7-38) produces a Gaussian beam with the same gain as the horn for S = 0.134. For different values of S, Eq. (7-38) gives only approximate Gaussian beams to match the gain of corrugated horns. The Gaussian beam has a 10-dB beamwidth of 27.5◦ and the corrugated horn has a beamwidth of 27.2◦ . The phase center of the Gaussian beam given by Eq. (7-37) is 2.44λ behind the aperture and the actual horn phase center is at 0.89λ. The Gaussian beam approximation ﬁnds the near-ﬁeld pattern of the corrugated horn because it includes the ﬁnite waist size instead of assuming a point source at the phase center of the horn. A PO analysis using the equivalent currents in the aperture [27, pp. 141–156] also ﬁnds the near-ﬁeld pattern but requires greater calculation effort.

7-6 RIDGED WAVEGUIDE HORNS Inserting ridges in the E-plane of a waveguide lowers the cutoff frequency compared to a waveguide of the same width. The ridges raise the cutoff frequencies of the next two higher modes and can produce a waveguide that operates over a 10 : 1 frequency range or more. If we use this as the input waveguide to a horn and taper the ridges until they do not block the horn aperture, the horn radiates a pattern similar to the smooth-wall horn. Near the aperture the horn can support many higher-order waveguide modes

366

HORN ANTENNAS

as frequency increases. The horn generates some higher-mode content to the ﬁelds which distorts the pattern over narrow frequency ranges, but for many applications such distortions are acceptable. Initial designs [28] used dual ridges for a single linear polarization, while later designs increased the number of ridges to four (quad-ridge) to allow for dual linear (or circular) polarization. Design concentrates on the input waveguide dimensions. We apply transverse resonance to the waveguide to calculate its cutoff frequencies. A rectangular waveguide with the electric ﬁeld parallel to the narrow wall can be considered as a parallel-plate transmission line with the wave traveling between the two narrow wall shorts at cutoff (see Section 5-24). The parallel-plate transmission-line impedance is ηb for a height of b (meters). The lowest-order mode cutoff frequency for a normal rectangular waveguide occurs when the width a = λ/2. The transverse resonance method considers half the width as a transmission line and cutoff occurs when the impedance at the centerline is an open circuit (odd-order mode) or a short circuit (even-order mode) (i.e., a/2 = N λ/4) for mode TEN0 . Of course, we ignore the impedance of the parallel-plate line because it is uniform. Figure 7-13a shows the cross section of a dual-ridged waveguide. The diagram illustrates feeding the waveguide with a coaxial line running through the center of one ridge. The center conductor extends across the gap to feed the second ridge. The center pin does not need to touch the second ridge but can be coupled capacitively. The transverse resonance circuit of a dual-ridged horn used to determine cutoff frequencies consists of two transmission-line segments with a shunt capacitor due to the step. The capacitance depends on the ratio of the heights α = b2 /b1 , where b2 < b1 [29]: C=

ε0 π

4α 1 + α2 α2 + 1 cosh−1 − 2 ln α 1 − α2 1 − α2

(7-39)

For the dual-ridged waveguide we analytically place a ground plane halfway across the waveguide E-plane and divide the waveguide into two half-height waveguides. Later we will consider the impedance, and the total impedance of the guide is these two

(a)

(b)

FIGURE 7-13 Coaxial feeds of ridged waveguides: (a) dual ridge; (b) quad ridge.

RIDGED WAVEGUIDE HORNS

367

half-height guides in series. Given the waveguide width a1 and height 2b1 , and the ridge width a2 with gap 2b2 , we solve for the cutoff frequency using a transcendental equation in admittance at the transition point between the two half-height waveguides. Odd-order TE modes have a virtual open circuit in the center of the ridge and a short circuit at the wall. Cutoff occurs for kc = 2π/λc = 2πfc /c for c equal to the speed of light [30]: tan(kc a2 /2) cot[kc (a1 − a2 )/2] + kc cC − =0 (7-40) ηb2 ηb1 We solve Eq. (7-40) numerically for kc for odd-order modes. The even modes have a virtual short circuit in the center, which leads to a similar equation for the cutoff number kc : cot[kc (a1 − a2 )/2] cot(kc a2 /2) + kc cC − =0 (7-41) − ηb2 ηb1 We use Eq. (7-40) to calculate the cutoff wavelengths of modes TE10 and TE30 and Eq. (7-41) to compute the cutoff wavelength of mode TE20 for given dimensions. We design the waveguide to have a suitable low-frequency cutoff with an impedance equal to the input coax, whose outer conductor is connected to one ridge, with the center conductor jumping the gap to feed the other. The impedance at an inﬁnite frequency is given by the equations Y∞

1 = kηb2

Z∞ =

ka2 sin ka2 b2 cos2 (ka2 /2) ka1 sin ka1 + + − 4 4 b1 sin2 (ka1 /2) 4 4

π b2 ka2 2b2 ln csc cos2 + λ 2 b1 2

1 Y∞

(7-42)

The impedance at a ﬁnite frequency increases: Z∞ Z0 = 1 − (fc /f )2

(7-43)

An approximate value for the gap can be found from the impedance of microstrip. The inﬁnite impedance equals slightly less than twice the impedance of microstrip the same width as the ridge with one-half the gap. The extra fringing capacitance between the sides of the ridges lowers the impedance compared to microstrip. You can use a microstrip line design program to ﬁnd an approximate gap and a few evaluations of Eq. (7-42) to determine the correct gap. Design for Z∞ because the impedance approaches Z∞ rapidly as frequency increases by Eq. (7-43) and ridged horns operate over a large bandwidth. Quad-ridged waveguide, illustrated in Figure 7-13b, requires modiﬁcations at the input to a horn. To achieve Z∞ = 50 , the gap must be reduced and the ridges made with a rooftop shape so that they ﬁt within each other. The capacitance between the ridges for one polarization is a series combination of the two capacitors to the ridges for the second polarization. Similar to the dual-ridged waveguide, we divide the

368

HORN ANTENNAS

waveguide along the centerline through the second set of ridges and analyze a singleridged waveguide. Given a square waveguide with width w, ridge width s, and gap g between the ridges of different polarizations, the equivalent single-ridged waveguide has parameters given by the expressions √ √ a1 = w + s( 2 − 1) a2 = s 2 b1 = w − s/2

b2 = g

(7-44)

For the quad-ridged waveguide the inﬁnite impedance equals slightly less than four times the impedance of microstrip the same width as the equivalent ridge a2 with onehalf the gap. We use the parameters of Eq. (7-44) in Eqs. (7-39) through (7-43) to ﬁnd the parameters of quad-ridged waveguide. Figure 7-13b demonstrates that the feed pin of one coax passes over the other to reduce coupling between them. The difference in distance to the waveguide shorting wall for the two coaxial lines produces different impedances for the two inputs. We can use the expressions above for circular waveguides. We design with a width equal to the diameter. The inﬁnite impedance is lower by the factor π/4. The cutoff frequency is about 1.25 times the cutoff frequency of the equivalent square waveguide [31]. Figure 7-14 gives a cross-sectional drawing of a ridged horn and demonstrates the key elements of design. A coax is fed through the center of one ridge and the center conductor jumps the gap and feeds the second ridge. We locate the coax close to the end of the ridge truncated before it reaches the waveguide back wall short circuit, leaving a small gap. Without ridges the waveguide is cutoff at the low-frequency end of the horn operation. Operating the waveguide below its cutoff frequency does not

FIGURE 7-14

Dual-ridged waveguide horn cross section.

RIDGED WAVEGUIDE HORNS

369

prevent the wave from reaching the back wall because the distance is short. The original horns [28] used waveguides in cutoff at the feed point at the lowest frequencies. By tapering the sidewalls the waveguide operates above cutoff in a short distance from the feed and the waves propagated to that region. Cutoff only means that a wave will not propagate in a long waveguide, but it attenuates as it moves along the waveguide. Figure 7-14 shows optional shorting pins between the back wall and the ridges. These prevent an additional resonance in impedance that may arise at a frequency when the height of a single ridge approaches λ/2. Not all designs need these pins. We space the ridges to form a transmission line matched to the feed coax at the input. A uniform section of ridged waveguide extends to the throat of the horn. The horn shown in Figure 7-14 uses an exponentially tapered ridge that has an additional linear taper with slope 0.02 [28] empirically found to improve the impedance match. It would seem that designing a classical tapered impedance transformer would give a better impedance match, but the simple exponential physical taper produces an excellent impedance match. The gain of the horn falls short of the equivalent open horn because multiple modes are excited and beamwidth broadens. In a dual-ridged horn the power concentrates between the ridges in the E-plane, and we can replace the H -plane sidewalls with a few rods. We space the rods close enough to block radiation at the lower frequencies and allow high-frequency radiation through the spaces. Since the ﬁelds are concentrated between the ridges at high frequencies, the side H -plane walls have little effect on the pattern. A quad-ridged horn requires solid walls. A circular quad-ridged horn was measured as a possible feed for a Cassegrain reﬂector from 6 to 18 GHz. The horn has a 13.2-cm aperture diameter and a 37.6-cm slant radius and operates from 2 to 18 GHz. Figure 7-15 plots the measured E- and H -plane 10-dB beamwidths along with the beamwidths of both smooth and corrugated wall horns of the same size. Neither smooth wall nor corrugated wall horns could be designed to operate over this wide bandwidth; they are shown only for comparison. The

FIGURE 7-15 Measured 10-dB beamwidths of a circular quad-ridged horn compared to the calculated beamwidths of smooth- and corrugated-wall horns.

370

HORN ANTENNAS

FIGURE 7-16 Measured directivity of a circular quad-ridged horn compared to those of smooth- and corrugated-wall horns.

quad-ridged horn has wider beamwidths in both planes compared to the other horns. This reduces the gain shown in Figure 7-16. Similar to the corrugated horn, the quadridged horn operates with multiple modes. We can determine the circular waveguide modes radiated by using physical optics analysis on the measured pattern. We radiate a plane wave into a circular aperture plane equal to the physical horn aperture and placed at the average phase center. Each plane wave, weighted by the pattern level and sin θ , excites Huygens source currents on the patches that cover the aperture by using Eq. (2-33). We normalize the currents to 1 watt and project the currents for each mode of a circular waveguide horn onto the incident wave currents by integrating over the aperture to determine their excitation levels bm : bm =

Ja ·J∗m dS

(7-45)

S

We use the aperture currents Ja and mode currents Jm in Eq. (7-45), where we take the complex conjugate of the vector for projection in the same manner as polarization calculations (section 1-11). We operate on the electric currents only because the magnetic currents are proportional to the electric currents for Huygens sources. Figure 7-17 plots the levels of the TE11 , TM11 , and diagonally oriented TE21 modes. TE11 and TM11 modes are also excited in a corrugated horn, but the level of the TM11 mode is approximately −5 dB relative to the TE11 mode. Further measurements of the horn show that it has approximately equal power in the TE11 and TM11 modes, all the way down to 2.7 GHz. Below that frequency the horn aperture will not support the TM11 mode and the pattern reverts to the TE11 mode only, which narrows the beamwidth. Analysis shows that the diagonally oriented TE21 mode peaks at an angle halfway around from the ridges and increases cross-polarization in the diagonal plane. The unmatched beamwidths in the E- and H -planes also increases the Huygens source

RIDGED WAVEGUIDE HORNS

371

TM11 TE11

TE12

FIGURE 7-17 Modal decomposition into circular waveguide modes of the measured pattern of a circular quad-ridged horn.

cross-polarization (section 1-11.2) in the diagonal plane. Square quad-ridged horns have similar modes. Measurements on a 63.5-cm-square aperture horn with a 140-cm slant length produced nearly equal levels of TE10 and TM12 modes, similar to the TE11 and TM11 circular modes in ﬁeld distribution. The TE10 and TM12 modes have approximately the same phase. The horn radiated the TE12 mode at the higher-frequency end of the band, which caused pattern distortion over a narrow frequency range. Both the TM12 and TE12 modes are excited by the electric ﬁeld between the ridges. The interplay of these three modes causes rapid changes in the beam shape as frequency changes. The horn exhibits these changes at the high end of the frequency band when all three modes exist with nearly equal power. Measurements on a dual-ridged horn produced patterns that reduced to the same three dominant modes as the square quad-ridged horn radiated and produced similar results. We fail to obtain a close match with the measured pattern of the quad-ridged horn by using the aperture currents beyond the 10-dB beamwidth for an aperture small in wavelengths. If we include currents excited along the outside of the horn bell in the physical optics analysis, we better match the measured pattern. This illustrates that the pattern of a horn is determined not only by aperture ﬁelds but also by the currents that ﬂow down the bell. Figure 7-18 shows the measured E- and H -plane patterns and the cross-polarization in the diagonal plane. The three-dimensional measured pattern plot in Figure 7-19 at 6 GHz shows the four cross-polarization lobes in the diagonal planes. The average pattern beamwidth matches a reﬂector with f/D = 1 and has an average illumination loss of 3 dB, with the value ranging from 2.5 to 4 dB. The average taper loss is 1.07 dB and the average spillover loss 1.08 dB. The cross-polarization exhibited in Figure 7-19 contributes an average 0.7 dB of loss. The phase-center location measurements show that the horn has up to 2λ astigmatism, which contributes 0.4 dB of loss when used as a reﬂector feed.

372

HORN ANTENNAS

FIGURE 7-18 Measured pattern of a circular quad-ridged horn.

Huygens Polarization, Co-Polarization

Huygens Polarization, Cross-Polarization

FIGURE 7-19 Spherical radiation pattern of a circular quad-ridged horn showing four-way symmetry of cross-polarization in diagonal planes.

7-7 BOX HORN [32, pp. 377–380] With a box horn (Figure 7-20), multiple waveguide modes are used to decrease the H -plane amplitude taper loss and axial length of the horn. We add the TE30 mode to

BOX HORN

373

FIGURE 7-20 Box horn.

the TE10 mode to reduce the cosine distribution taper of the H -plane. By phasing the modes 180◦ out of phase in the center of the aperture, the cos 3πx distribution subtracts from the TE10 -mode distribution in the center and adds in the region near the edges. A step in the width of a waveguide generates TEN0 modes when fed by the TE10 mode. Any modes not cut off by the waveguide will propagate to the aperture. If we maintain symmetry about the axis of the waveguide, only odd-order modes (TE30 , TE50 , etc.) will be generated. The width W of the waveguide (Figure 7-20) beyond the step determines the possible propagating modes: λc = 2W/N , where N is the mode number. If we limit the modes to the TE10 and TE30 modes in the aperture, the cutoff wavelength of the TE50 mode determines the maximum width: Wmax = 2.5λ. The TE30 -mode cutoff wavelength establishes the minimum width: Wmin = 1.5λ. Within this range, short horns with good aperture efﬁciency can be designed. We can ﬂare the E-plane to increase its aperture (Figure 7-20), but the limited axial length of the horn bounds the possible ﬂare without an excessive phase error loss. The H -plane can also be ﬂared, but ﬂaring it complicates the design for the proper length L. The step generates smaller amplitudes of higher-order modes with each increase in N . Small amounts of higher-order modes (TE50 , TE70 , etc.) will decrease the efﬁciency only marginally, since the mode amplitudes are small. The step generates modes in phase with the input TE10 mode, since the higher-order modes must peak in the center and subtract from the TE10 ﬁelds on the back wall of the larger waveguide section. The aperture distribution is a sum of TE10 and TE30 modes: Ey (x) = a1 cos

πx 3πx exp(−j k10 L) + a3 cos exp(−j k30 L) W W

(7-46)

where k10 and k30 are the propagation constants of the two modes. The amplitude distribution in the H -plane will be more nearly uniform if the phase between the modes is 180◦ . The modes travel from the step with different phase velocities, depending on their cutoff frequencies. We adjust the length L to give a 180◦ phase difference between the modes: (k10 − k30 )L = π where k10 = k 1 − (λ/2 W )2 and k30 = k 1 − (3λ/2 W )2 . We solve for the length: λ/2 L= 2 1 − (λ/2 W ) − 1 − (3λ/2 W )2

(7-47)

374

HORN ANTENNAS

TABLE 7-10 TE30 /TE10 (Voltages)

Box Horn Characteristics Ratio of Input Waveguide to Aperture

Linear ATLw (dB)

3 dB

10 dB

1.000 0.940 0.888 0.841 0.798 0.758 0.719 0.682 0.645 0.609 0.573 0.537 0.500 0.462 0.424

0.91 0.78 0.67 0.58 0.52 0.48 0.46 0.46 0.47 0.50 0.54 0.60 0.66 0.74 0.82

0.594 0.575 0.558 0.544 0.530 0.518 0.507 0.496 0.487 0.479 0.471 0.463 0.456 0.450 0.444

1.019 0.981 0.947 0.917 0.890 0.866 0.844 0.824 0.806 0.790 0.775 0.761 0.749 0.737 0.726

0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70

(W /λ) sin θ

The ratio of the modes generated by the step can be found from mode matching on the input waveguide aperture of width a:

a/2

cos(πx/a) cos(N πx/W )dx aN −a/2 = a/2 a1 cos(πx/a) cos(πx/W )dx

(7-48)

−a/2

where aN is the TEN0 mode amplitude. Table 7-10 lists the step dimensions needed to design to a given ratio of modes. The amplitude taper loss is a minimum at a3 /a1 = 0.32. The possible 3-dB beamwidths with a single mode, TE10 , range from 20 to 44◦ . Example Design a box horn with an H -plane 10-dB beamwidth of 50◦ . We pick a3 /a1 = 0.35. From Table 7-10, (W/λ) sin θ = 0.824. The obliquity factor at 25◦ adds 0.42 dB of loss. We must design with a wider 10-dB beamwidth. This is within the permissible range for only two modes in the aperture. We calculate the length to phase the modes by 180◦ by using Eq. (7-47): L = 1.451λ. The horn is shorter than the aperture width.

7-8 T-BAR-FED SLOT ANTENNA A T-bar-fed slot antenna (Figure 7-21) looks more like an open-circuited waveguide to coax transition than a slot. Like a slot, its pattern is very broad. The antenna has been designed experimentally [33, pp. 184–190] and those dimensions provide a good starting point. Table 7-11 lists two designs [33] referred to Figure 7-21. The aperture admittance is a combination of the radiation admittance and a capacitive

T-BAR-FED SLOT ANTENNA

375

b

a w x

x

Outline of Cavity G

b

Back Wall

b/2

F E

D H

Aperture a Front View

W Side View

FIGURE 7-21 T-bar-fed cavity slot antenna. (From [34], 1975 IEEE.)

TABLE 7-11 Dimensions for Two Antenna Designs Dimensions

Antenna 1

Antenna 2

b/a W/a x/a D/a I/a E/a F/a

0.323 0.323 0.118 0.118 0.059 0.118 0.057

0.226 0.295 0.113 0.090 0.045 0.090 0.045

susceptance. Behind the feed point, the length of short-circuited waveguide adds an inductive susceptance that grows as frequency decreases. The horizontal bar produces a capacitive susceptance at the input to counteract the back-wall susceptance. These susceptances track with frequency changes; each one decreases to maintain the sum near resonance. Later experimental work [34] revealed further properties of the antenna. Measurements on antenna 1 show that the lower-end 2 : 1 VSWR band edge occurs when a = 0.57λ and the upper end when a = 0.9λ. The bandwidth is about 1.6 : 1.

376

HORN ANTENNAS

Antenna 2 was reported [33] as having less bandwidth than antenna 1. When the round rod was replaced with a ﬂat strip, whose width across the guide was the same as the diameter of the rod, almost identical results were obtained. We have a choice. The ﬂat strip is an easier construction, but the round rod gives better mechanical support in all axes to withstand shock and vibration. The ﬂat strip adds to the freedom of design. The bandwidth potential increases when H is decreased while I is held constant. Newman and Thiele [34] found that when H was decreased, the nominal impedance level was raised. When the input impedance is plotted on the Smith chart, the locus is centered about a higher resistance. By adding a broadband impedance transformer on the input, we can achieve the higher bandwidth potential. Newman and Thiele achieved a nearly 2 : 1 VSWR bandwidth from a = 0.52λ to a = 1.12λ, or 2.3 : 1 bandwidth.

7-9 MULTIMODE CIRCULAR HORN [35] A step in the diameter of a circular waveguide generates a TM11 mode to satisfy the boundary conditions. The ﬁelds of the TM11 mode can be phased to cancel the ﬁelds from the TE11 mode at the edges of the aperture in the E-plane. The tapering of the ﬁelds in the aperture reduces the E-plane sidelobes while broadening the beamwidth. Equalizing the ﬁeld distributions in the two planes helps to bring the E- and H -plane phase centers closer together. The modes generated by the step are more complicated than those for the box horn. Symmetry eliminates generation of the unwanted modes: TM01 , TE21 , and TE01 . The step transition shifts the phase of the TM11 mode relative to the TE11 mode [36]. Since the waveguide modes have different phase velocities, they can be phased to produce the desired ﬁeld at the aperture. Although calculated information [36] is helpful, the designs must be completed empirically. The required phasing to achieve ﬁeld cancellation limits the bandwidth, but for narrowband applications a stepped horn is cheaper than a corrugated horn. Satoh [37] loads the ﬂare of a conical horn with a conical dielectric step to generate the TM11 mode. Symmetry prevents the excitation of unwanted modes. He places the step at a diameter where the TM11 mode can propagate. By using two steps, the bandwidth can be increased because the lengths can be adjusted to give perfect mode cancellation at two frequencies. We can replace the dielectric cone by metal steps each of which generates the TM11 mode and thereby achieve good results, in theory, at multiple frequencies.

7-10

BICONICAL HORN [4]

A biconical horn consists of two cones with a common vertex. The angle of the generating lines of the cones is measured from a common axis. The cones of the usual antenna have angles that sum to 180◦ . Spherical modes describe the ﬁelds between the cones, but we can use approximations with good results. The lowest-order mode is TEM between the cones and is easily excited by a coax line. The outer conductor connects to one cone, and the second cone feeds out of the center conductor. The electric ﬁeld of the TEM mode is polarized in the direction of the axis. The ﬁrst higher-order mode

BICONICAL HORN

377

has a circulating electric ﬁeld with the magnetic ﬁeld in the direction of the axis. It can be excited either from a TE01 -mode circular waveguide or by an array of slots on a cylinder. The distance between cones must be at least λ/2 at the point of excitation of the TE01 biconical mode. We approximate the distribution of the zeroth-order mode, TEM, as uniform along the axis. The ﬁrst-order mode, TE01 , distribution is approximately cosine along the axis. We calculate gain by using aperture distribution losses. We describe the horn with a slant radius along the generating line and a height between the ends of the cones. The expansion in spherical modes requires integration over a spherical cap at the aperture if a constant phase surface is used. We obtain good results by using a cylindrical aperture and a quadratic phase distribution. The antenna has circular symmetry about the z-axis that bounds the directivity to 2L/λ. We use linear distribution efﬁciencies to compute directivity (gain): 2L gain = 10 log (7-49) − ATLx − PELx λ The TEM mode has a uniform distribution, so we use the “uniform” column of Table 4-42 to calculate phase error loss. The uniform distribution has no amplitude taper loss. The cosine distribution of the ﬁrst-order mode requires an ATL = 0.91 dB and use of the cosine distribution quadratic phase error loss of Table 4-42. Given the height between the ends of the cones, H , and the slant radius R, we determine the quadratic phase distribution constant from S=

H2 8λR

(7-50)

Example Compute the gain of a biconical horn with a slant radius of 10λ and cone angles of 75◦ and 105◦ . H = 2R cos 75◦ = 5.176λ and S = 0.33. From Table 4-42, we read PELTEM = −1.76 dB

PELTE−01 = −0.79 dB

Vertical mode, TEM: gain = 10 log[2(5.176)] + PELTEM : 10.15 dB − 1.76 dB = 8.4 dB Horizontal mode, TE01 : gain = 10 log[2(5.176)] + PELTE−01 + ATLcosine : 10.15 dB − 0.79 dB − 0.91 dB = 8.45 dB We can calculate beamwidths by using the results of the rectangular horn, where we measure the angles from θ = 90◦ for the complementary-angled biconical horn. Example Compute the 3-dB beamwidths of the horn above. S = 0.33 and H = 5.176λ. Use Table 7-2 with the TEM mode and α as the angle from θ = 90◦ : H sin α = 0.5015 λ

◦

α = 5.56

378

HORN ANTENNAS

The obliquity factor is insigniﬁcant. ◦

HPBW = 11.1

TEM mode

Use Table 7-1 with the TE01 mode. H sin α = 0.6574 λ

◦

HPBW = 14.6

TE01 mode

The two modes have about the same gain, but the TE01 mode has a greater beamwidth. When we refer to Figures 7-3 and 7-4, we see that the TEM-mode horn has about 7-dB sidelobes and the TE01 -mode horn has practically no sidelobes. The sidelobes reduce the gain of the TEM mode with its narrower beamwidth.

REFERENCES 1. A. W. Love, ed., Electromagnetic Horn Antennas, IEEE Press, New York, 1976. 2. W. L. Barrow and L. J. Chu, Theory of electromagnetic horn, Proceedings of IRE, vol. 27, January 1939, pp. 51–64. 3. M. C. Schorr and E. J. Beck, Electromagnetic ﬁeld of the conical horn, Journal of Applied Physics, vol. 21, August 1950, pp. 795–801. 4. S. A. Schelkunoff and H. Friis, Antenna Theory and Practice, Wiley, New York, 1952. 5. P. M. Russo et al., A method of computing E-plane patterns of horn antennas, IEEE Transactions on Antennas and Propagation, vol. AP-13, no. 2, March 1965, pp. 219–224. 6. J. Boersma, Computation of Fresnel integrals, Mathematics of Computation, vol. 14, 1960, p. 380. 7. K. S. Kelleher, in H. Jasik, ed., Antenna Engineering Handbook, McGraw-Hill, New York, 1961. 8. D. G. Bodnar, Materials and design data, Chapter 46 in R. C. Johnson, ed., Antenna Engineering Handbook, 3rd ed., McGraw-Hill, New York, 1993. 9. E. H. Braun, Gain of electromagnetic horns, Proceedings of IRE, vol. 41, January 1953, pp. 109–115. 10. E. I. Muehldorf, The phase center of horn antennas, IEEE Transactions on Antennas and Propagation, vol. AP-18, no. 6, November 1970, pp. 753–760. 11. A. J. Simmons and A. F. Kay, The scalar feed: a high performance feed for large paraboloid reﬂectors, Design and Construction of Large Steerable Aerials, IEE Conference Publication 21, 1966, pp. 213–217. 12. B. M. Thomas, Design of corrugated conical horns, IEEE Transactions on Antennas and Propagation, vol. AP-26, no. 2, March 1978, pp. 367–372. 13. T. S. Chu and W. E. Legg, Gain of corrugated conical horn, IEEE Transactions on Antennas and Propagation, vol. AP-30, no. 4, July 1982, pp. 698–703. 14. G. L. James, TE11 to HE11 mode converters for small angle corrugated horns, IEEE Transactions on Antennas and Propagation, vol. AP-30, no. 6, November 1982, pp. 1057–1062. 15. P. J. B. Clarricoats and P. K. Saha, Propagation and radiation of corrugated feeds, Proceedings of IEE, vol. 118, September 1971, pp. 1167–1176. 16. A. W. Rudge et al., eds., The Handbook of Antenna Design, Vol. 1, Peter Peregrinus, London, 1982.

REFERENCES

379

17. G. L. James and B. M. Thomas, TE11 -to-HE11 corrugated cylindrical waveguide mode converters using ring-loaded slots, IEEE Transactions on Microwave Theory and Techniques, vol. MTT-30, no. 3, March 1982, pp. 278–285. 18. R. Wohlleben, H. Mattes, and O. Lochner, Simple small primary feed for large opening angles and high aperture efﬁciency, Electronics Letters, vol. 8, September 21, 1972, pp. 474–476. 19. A. D. Olver et al., Microwave Horns and Feeds, IEEE Press, New York, 1994. 20. G. L. James, Radiation properties of 90◦ conical horns, Electronics Letters, vol. 13, no. 10, May 12, 1977, pp. 293–294. 21. A. Kumer, Reduce cross-polarization in reﬂector-type antennas, Microwaves, March 1978, pp. 48–51. 22. P.-S. Kildal, Foundations of Antennas, Studentlitteratur, Lund, Sweden, 2000. 23. S. Maci et al., Diffraction at artiﬁcially soft and hard edges by using incremental theory of diffraction, IEEE Antennas and Propagation Symposium, 1994, pp. 1464–1467. 24. B. A. Saleh and M. C. Teich, Fundamentals of Photonics, Wiley, New York, 1991. 25. K. Pontoppidan, ed., Technical Description of Grasp 8, Ticra, Copenhagen, 2000 (self published and available at www.ticra.com). 26. P. F. Goldsmith, Quasioptical Systems, IEEE Press, New York, 1998. 27. L. Diaz and T. A. Milligan, Antenna Engineering Using Physical Optics, Artech House, Boston, 1996. 28. J. L. Kerr, Short axial length broadband horns, IEEE Transactions on Antennas and Propagation, vol. AP-21, no. 5, September 1973, pp. 710–714. 29. J. R. Whinnery and H. W. Jamieson, Equivalent circuits for discontinuities in transmission lines, Proceedings of IRE, vol. 32, no. 2, February 1944, pp. 98–114. 30. S. B. Cohn, Properties of ridge waveguide, Proceedings of IRE, vol. 35, no. 8, August 1947, pp. 783–788. 31. M. H. Chen, G. N. Tsandoulas, and F. W. Willwerth, Modal characteristics of quadrupleridged circular and square waveguide, IEEE Transactions on Microwave Theory and Techniques, vol. MTT-21, August 1974, pp. 801–804. 32. S. Silver, ed., Microwave Antenna Theory and Design, McGraw-Hill, New York, 1949. 33. A. Dome and D. Lazarno, Radio Research Laboratory Staff, Very High Frequency Techniques, McGraw-Hill, New York, 1947, pp. 184–190. 34. E. H. Newman and G. A. Thiele, Some important parameters in the design of T-bar fed slot antennas, IEEE Transactions on Antennas and Propagation, vol. AP-23, no. 1, January 1975, pp. 97–100. 35. P. D. Potter, A new horn antenna with suppressed sidelobes and equal beamwidths, Microwaves, vol. 6, June 1963, pp. 71–78. 36. W. J. English, The circular waveguide step-discontinuity mode transducer, IEEE Transactions on Microwave Theory and Techniques, vol. MTT-21, no. 10, October 1973, pp. 633–636. 37. T. Satoh, Dielectric loaded horn antenna, IEEE Transactions on Antennas and Propagation, vol. AP-20, no. 2, March 1972, pp. 199–201.

8 REFLECTOR ANTENNAS

The importance of reﬂector antennas cannot be overstated. Large-aperture antennas can be built only with reﬂectors or arrays and reﬂectors are far simpler than arrays. The arrays give us more degrees of freedom than is necessary in many applications. With plenty of room and slow scan rates, a reﬂector becomes a better design than an array. Of course, there can be many valid reasons for using an array in an application, but a reﬂector should always be considered. An array needs an elaborate feed network, whereas a reﬂector uses a simple feed and free space as its feed network. Most reﬂector designs require extensive calculations together with full characterization of the feed antenna. Many types of analysis have been developed. As with horn antennas, Love [1] has collected signiﬁcant papers on reﬂector antennas. In his classic book, Silver [2] provides the foundation for an analysis based on aperture theory and physical optics (induced currents on the reﬂector). Aperture theory or physical optics reduced to aperture theory is still used for most designs. Rusch and Potter fully develop aperture and physical optics theories for the design and analysis of both prime focus and dual-reﬂector (Cassegrain) antennas [3]. Other methods have been developed either to increase the range of valid patterns or to decrease the pattern calculation time so that optimization techniques can be applied. Wood [4] collects ideas for designing by using spherical wave expansions that allow for an overall system optimization using only a few terms. GTD methods [5,6] ﬁnd increasing applications as an analysis technique suitable for a full pattern analysis except at boresight. Improved methods of calculating the secondary pattern have been developed using aperture ﬁelds, such as FFT methods [7] and Jacobi–Bessel series [8]. Many of these techniques and hardware implementations of reﬂectors are summarized in a handbook [9, Chaps. 2 and 3]. Although all these methods are available, aperture theory and physical optics remain the main techniques of reﬂector design and analysis. Modern Antenna Design, Second Edition, By Thomas A. Milligan Copyright 2005 John Wiley & Sons, Inc.

380

PARABOLOIDAL REFLECTOR GEOMETRY

381

8-1 PARABOLOIDAL REFLECTOR GEOMETRY Figure 8-1 shows the geometry of the parabolic reﬂector. We form the reﬂector by rotating the ﬁgure about its axis or by moving the ﬁgure along an axis out of the paper to form a cylindrical reﬂector. Because the cylindrical reﬂector requires a line source, it is less important than the circularly symmetrical reﬂector fed from a single point source. A paraboloidal reﬂector transforms a spherical wave radiated by the feed located at its focus into a plane wave. Although the feed wave spreads from the focus, which reduces its amplitude, geometric optics predicts a plane wave reﬂection that remains constant. The reﬂected wave does not remain a plane wave but spreads because the ﬁelds must be continuous across the reﬂection boundary of the beam plane wave column because ﬁelds can be discontinuous only across physical boundaries. Nevertheless, we will use the aperture theory on the projected diameter to predict its performance. Since the reﬂected rays are parallel, we can place the aperture plane anywhere along the axis, but somewhat close in front of the reﬂector. The equations for the reﬂector surface are f cos2 (ψ/2) polar coordinates

ρ=

r 2 = 4f (f + z) rectangular coordinates

(8-1)

where f is the focal length, D the diameter, ρ the distance from the focus to the reﬂector, and ψ the feed angle from the negative z-axis. The reﬂector depth from the rim to the center is z0 = D 2 /16f .

Parabolic Reflector

^ n r

r y

Focus z

D F y0

FIGURE 8-1

Geometry of a parabolic reﬂector.

382

REFLECTOR ANTENNAS Feed Subtended Angle (degrees)

Parabola f/D

SCALE 8-1

Parabola f/D compared to a feed total subtended angle.

We eliminate the dimensions of the reﬂector by using the ratio f /D. The half subtended angle of the reﬂector, ψ0 , relates to f/D by ψ0 = 2 tan−1

1 4f/D

(8-2)

Scale 8-1 computes the total feed subtended angle from reﬂector f/D. When we place the aperture plane at the focus, the ray path distance becomes ρ + ρ cos ψ = 2ρ cos2

ψ = 2f 2

all ray path lengths are equal, and the aperture plane is a constant-phase surface (eikonal). The normal unit vector at a point on the reﬂector (r, z) is found from the feed angle: nˆ = − sin

ψ ψ rˆ + cos zˆ 2 2

At this point we need the radius of curvatures in the principal planes to apply Eq. (2-77) reﬂection from a curved surface: R1 in the r –z plane and R2 in the φ –z plane: R1 =

2f cos3 (ψ/2)

and R2 =

2f cos(ψ/2)

The spherical wave spreads from the feed as 1/ρ. At the surface of the reﬂector the wave curvature changes to a plane wave and propagates to the aperture plane at a constant amplitude. The spherical wave spreading multiplies the feed distribution by [Eq. (8-1)] cos2 (ψ/2) in the aperture. Then added edge taper = cos2

ψ0 2

voltage

(8-3)

Deeper reﬂectors (smaller f/D) have greater edge tapers than shallow reﬂectors (larger f/D). Scale 8-2 provides a quick calculation of the added edge taper due to spherical wave spreading. Example Calculate the edge taper of a paraboloidal reﬂector for f/D = 0.5 and an isotropic feed. From Eq. (8-2), ψ0 = 2 tan−1 12 = 53.13◦ . The edge taper is [Eq. (8-3)] edge taper = 20 log cos2

53.13◦ = −1.94 dB 2

PARABOLOIDAL REFLECTOR APERTURE DISTRIBUTION LOSSES

383

Added Edge Taper, dB

Parabola f/D

SCALE 8-2 Added edge taper due to a spherical wave from feed.

If the feed has its 10-dB pattern point directed toward the reﬂector edge, the aperture edge taper is 11.9 dB.

8-2 PARABOLOIDAL REFLECTOR APERTURE DISTRIBUTION LOSSES We manipulate Eq. (4-2) for ATL to eliminate the dimensions and relate the integrals to the feed pattern: 0

ATL =

2π

b 2π a

πa 2

a

Ea (r , φ ) r dr dφ

b

0

2

Ea (r , φ )2 r dr dφ

(8-4)

where a is the aperture radius, b the central blockage radius, and Ea (r , φ ) the aperture ﬁeld. We make the following substitutions into Eq. (8-4): r = ρ sin ψ = 2 sin

ψ ψ ψ f cos = 2f tan 2 2 2 cos (ψ/2) 2

ψ = ρdψ dr = f sec 2

(8-5)

2

The aperture ﬁeld is related to the feed pattern by Ea (r , φ ) =

E(ψ , φ ) ρ

These substitutions eliminate dimensions in Eq. (8-4):

2π 0

ATL = π[tan2 (ψ

0 /2)

−

ψ0 ψb

2 |E(ψ, φ)| tan(ψ/2) dψ dφ

tan2 (ψ

2π

b /2)] 0

ψ0 ψb

(8-6) |E(ψ, φ)| sin ψ dψ dφ 2

where ψb = 2 tan−1 [b/(2f )]. When we substitute the relations in Eq. (8-5) into Eq. (4-9) to eliminate dimensions in the integrals, we obtain an expression with only the feed

384

pattern:

REFLECTOR ANTENNAS

2π ψ0 2 E(ψ, φ) tan(ψ/2) dψ dφ ψ 0 PEL = b 2 2π ψ0 |E(ψ, φ)| tan(ψ/2) dψ dφ

(8-7)

ψb

0

PEL is the efﬁciency at the boresight. We modify Eq. (8-7) when we scan the beam to give off-boresight values as in Eq. (4-3). The amplitude taper efﬁciency (ATL) of Eq. (8-6) and the phase error efﬁciency (PEL) of Eq. (8-7) do not account for the total directivity loss of the aperture. The reﬂector does not intercept all the power radiated by the source and some of it spills over the edge. Spillover adds little to the pattern except as sidelobes, since usual feeds have small backlobes. We consider this spilled-over power as a loss (SPL):

2π

2π

0

SPL =

ψ0 ψb π

0

|E(ψ, φ)|2 sin ψ dψ dφ (8-8) |E(ψ, φ)| sin ψ dψ dφ 2

0

This expression for spillover includes the scattered portion of the central blockage efﬁciency, but not the loss of potential aperture. We include the remainder in the directivity calculation. We have ignored the cross-polarized power radiated by the source. We deﬁne crosspolarization efﬁciency (XOL) as XOL =

2π

0

π

2π 0

π

|EC (ψ, φ)|2 sin ψ dψ dφ

0

(8-9)

(|EC (ψ, φ)| + |EX (ψ, φ)| ) sin ψ dψ dφ 2

2

0

where Ec is the co-polarized ﬁeld and Ex is the cross-polarized ﬁeld. These polarizations correspond to Ludwig’s [10] third deﬁnition of cross-polarization. A Huygens source produces straight reﬂector surface currents when projected to the aperture plane. Including the cross-polarization efﬁciency gives us the true average radiation intensity as in Eq. (1-17). If we express the efﬁciencies as ratios, the directivity will be found from directivity =

π 2 λ

(Dr2 − Db2 )SPL · ATL · PEL · XOL (ratio)

(8-10)

where Dr is the reﬂector diameter and Db is the diameter of the central blockage. Equation (8-10) includes the nonscattered blockage loss of potential aperture. Equation (8-10) can be expressed in terms of decibel ratios: π 2 2 2 directivity = 10 log (Dr − Db ) + SPL(dB) + ATL(dB) λ + PEL(dB) + XOL(dB)

(8-11)

APPROXIMATE SPILLOVER AND AMPLITUDE TAPER TRADE-OFFS

385

Of course, all the decibel ratios of the efﬁciencies will be negative and subtract from the directivity calculated from the area. When measuring an actual feed, we can ignore the cross-polarized power. We measure the efﬁciency as the difference between directivity and gain. Actual directivity includes the co-polarizations and cross-polarizations in the average radiation intensity. If we ignore the cross-polarization, the measured efﬁciency decreases by the crosspolarization loss because the measured and true directivity differ by that loss. We must measure the cross-polarization pattern distribution of the feed if we want to calculate the cross-polarized secondary (reﬂector) pattern. When the cross-polarization pattern is not required, we save time without loss of accuracy by measuring only the co-polarized feed pattern. Equations (8-8) and (8-9) are by no means unique. We could include the crosspolarized power in the spillover calculation [Eq. (8-8)] and limit the integration limits in Eq. (8-9) to the reﬂector. A set of efﬁciency relations is correct when the equations account for all the power radiated by the feed. When we use calculated feed patterns, we must determine cross-polarization efﬁciency, since we can only estimate the efﬁciency due to material losses. The cross-polarization efﬁciency cannot be included as it is in measurements, and the division of cross-polarized power between Eqs. (8-8) and (8-9) is arbitrary.

8-3 APPROXIMATE SPILLOVER AND AMPLITUDE TAPER TRADE-OFFS We use the approximate pattern cos2N (ψ/2) for a feed pattern to establish trends. Of course, if the actual feed pattern distribution is available, we should use Eqs. (86) to (8-9). We obtain closed-form expressions when we substitute this pattern into Eqs. (8-6) and (8-8). Ignoring any central blockage, we get spillover efﬁciency = 1 − u2(N+1) amplitude taper efﬁciency =

4(N + 1)(1 − uN )2 ψ0 cot2 N 2 [1 − u2(N+1) ] 2

(8-12) (8-13)

where u = cos(ψ0 /2). We combine Eqs. (8-12) and (8-13) and plot their combination to ﬁnd the beamwidth for minimum loss. In Figure 8-2 the loss versus the 10-dB beamwidth for various f/D values is plotted. At narrow beamwidths little feed power spills over the reﬂector edge, but the reﬂector is underilluminated. Increasing the beamwidth improves the illumination but increases the spillover. The efﬁciency peaks when the feed 10-dB beamwidth is approximately the subtended angle of the reﬂector. Figure 8-2 shows a broad peak for any given f/D. Small changes in the beamwidth near the peak have no practical effect on the reﬂector’s gain. Scale 8-3 relates the average illumination loss reduction given the feed pattern level in the direction of the reﬂector rim for typical antennas. Example Estimate the amplitude taper loss for a reﬂector with f/D = 0.5 whose feed has a 10-dB edge taper. Compare the loss with that of the circular Gaussian and the Hansen single-parameter distributions: ψ0 = 2 tan−1 12 = 53.13◦ . The 10-dB beamwidth of the feed is then 106.26◦ . We modify Eq. (1-20) to compute the exponent N of the cos2N (ψ/2) feed

386

REFLECTOR ANTENNAS

Spillover + Amplitude Taper Loss, dB

0

F/D 2 1.5

1 0.8

0.6

1

0.5

0.4 0.3

2

3 0

50°

100°

150°

200°

Feed 10-dB Beamwidth

FIGURE 8-2 Sum of spillover and amplitude taper losses versus feed 10-dB beamwidth. Feed Level in Direction of Rim, dB

Mean Illumination Loss Change f /D : 0.3 - 0.6, dB

SCALE 8-3 Mean illumination loss change of a reﬂector given the feed pattern level in the rim direction.

pattern approximation, N=

log 0.1 = 10.32 2 log cos(106.26◦ /4)

From Eq. (8-13), u = cos(53.13◦ /2) = 0.894: 4(1 − 0.89410.32 )2 (11.32) ◦ cot2 (53.13 /2) = 0.864 10.322 (1 − 0.89422.64 ) = 10 log 0.864 = −0.63 dB

ATL(dB) =

The extra distance from the feed to the reﬂector edge compared with the center distance adds 1.94 dB and increases the aperture amplitude taper to 11.94 dB. We interpolate Table 4-29 for the circular Gaussian distribution and Table 4-30 for the Hansen singleparameter distribution to ﬁnd the following data: Gaussian ATL(dB) = −0.62 dB Sidelobe level = 26.3 dB Beamwidth factor = 1.142

Hansen ATL(dB) = −0.57 dB Sidelobe level = 24.7 dB Beamwidth factor = 1.136

PHASE ERROR LOSSES AND AXIAL DEFOCUSING

387

We multiply Eq. (4-83) by the beamwidth factor to estimate the reﬂector beamwidth: ◦

HPBW = 67.3

λ D

◦

and HPBW = 67

λ D

These compare well with the approximation, HPBW = 70◦ λ/D for a parabolic reﬂector. An integration of the aperture distribution for the far-ﬁeld pattern gives the following results: ◦

HPBW = 67.46

λ D

sidelobe level = 27 dB

8-4 PHASE ERROR LOSSES AND AXIAL DEFOCUSING All rays starting at the reﬂector focus travel the same distance through reﬂection to the aperture plane. The aperture plane is any convenient plane in front of the dish whose normal is the axis of the reﬂector. If we could build a feed with a unique phase center and place it at the focus of a perfect paraboloidal reﬂector, we would eliminate phase error loss in the aperture plane because it would have a constant phase. The feed, the positioning of the feed, and the reﬂector surface all contribute to the phase error loss. We discussed techniques for obtaining unique phase centers in the various planes for horns. Unlike smooth-wall horns, corrugated horns can have equal phase centers in all planes through the axis, but even their position will wander with changes in frequency. We measure the feed pattern distribution (amplitude and phase) to predict the contribution of the feed to the overall efﬁciency. From those measurements we deﬁne the practical phase center as the point on the feed leading to the minimum phase error loss when placed at the focus. The random and systematic phase error contributions can be measured directly on the feed and calculated numerically using Eq. (8-7). The feed phase center cannot always be placed at the focus. The phase-center location wanders with changes in frequency, and in any wideband application we expect axial defocusing. For example, the location of the phase center of a log-periodic antenna moves toward the apex as frequency increases. Figure 8-3 is a plot of the phase error loss due to axial defocusing. Each feed has its 10-dB beamwidth equal to the reﬂector subtended angle. Axial defocusing affects deep dishes (lower f/D) more than shallow dishes. We can estimate the axial defocusing phase error loss by approximating the distribution with a quadratic aperture phase distribution. Given z as the axial defocusing, the maximum phase deviation in cycles is z 1 −1 S= 1 − cos 2 tan λ 4f/D

(8-14)

We combine this with the quadratic phase error loss of the circular Gaussian distribution to estimate the loss. With z = λ we obtain a scaling factor for S (Scale 8-4) given z from Eq. (8-14). The scaling factor decreases with increasing f/D. Example Estimate the phase error loss for z = 2λ when f/D = 0.6 and the feed 10-dB beamwidth equals the reﬂector subtended angle.

388

REFLECTOR ANTENNAS

2 1 1.6 1.4

Phase Error Loss, dB

2

1.2

3

4

5

6

0.3

0.6 1

2

0.8

F/D = 1

3 4 Axial Defocusing, l

5

6

FIGURE 8-3 Paraboloidal reﬂector phase error loss due to axial defocusing of the feed.

Axial Defocusing Quadratic Phase Factor, S

Parabola f /D

SCALE 8-4

Quadratic phase factor S for axial defocusing of a paraboloidal reﬂector.

From Scale 8-4, S = 0.30(2) = 0.60. We use Eq. (8-3) to compute the edge taper: 1 ◦ = 45.2 2.4 ψ0 = −1.4 dB edge taper = 20 log cos2 2 ψ0 = 2 tan−1

An equivalent truncated Gaussian aperture distribution tapers to 10 dB + 1.4 dB = 11.4 dB

ρ=

11.4 = 1.31 8.69

We use Eq. (4-118) to calculate phase error efﬁciency of the truncated Gaussian distribution: PEL = 0.305 or PEL (dB) = −5.2 dB. This matches the value from Figure 8-3 found by integration of the actual distribution. The optimum feed beamwidth produces an average aperture edge taper of 11.8 dB. Scale 8-5 evaluates Eq. (4-118) for this taper. We detect axial defocusing by looking at the patterns of the reﬂector. Axial defocusing ﬁlls-in nulls between sidelobes. We adjust the feed location to maximize the null

ASTIGMATISM

389

Truncated (-11.8 dB) Circular Gaussian Distribution PEL, dB

Quadratic Phase Factor, S

SCALE 8-5 given S.

Truncated circular Gaussian distribution (−11.8 dB taper) phase error loss

depth, but antenna range errors and receiver sensitivity limit our ability to eliminate this defocusing. 8-5 ASTIGMATISM [11] Both the feed and the reﬂector can have astigmatism: unequal phase centers in different planes. We measure the feed by itself to discover its astigmatism. When the feed is mounted in the reﬂector, we detect astigmatism by the depth of nulls in the various pattern planes. A series of measurements can separate the feed and reﬂector astigmatism, but the feed must be able to move along the reﬂector axis and to rotate by 90◦ during the measurements. Move the feed along the axis to ﬁnd the locations that give maximum nulls. The extrema of the reﬂector focuses may not occur in the E- and H -planes and will require a search in the other planes. At this point we cannot separate the feed astigmatism from the reﬂector astigmatism. We rotate the feed and repeat the measurements. The feed phase center locations shift, and the reﬂector focuses remain ﬁxed. Simple manipulation of the data from the two measurements separates the two sources of astigmatism. The reﬂector can be shimmed to remove its astigmatism, or the feed phase centers can be matched to the reﬂector focuses. Figure 8-4 shows the magnitude

1.5 1

Phase Error Loss, dB

1 2 0.8 0.6

3 0.5 4 F/D = 0.3 0.4 5

6

FIGURE 8-4

1

4 5 2 3 Distance Between Phase Centers, l

6

Paraboloidal reﬂector phase error loss due to feed astigmatism.

390

REFLECTOR ANTENNAS

of phase error losses due to feed astigmatism. Astigmatism loss is not as severe as axial defocusing because in two planes the feed phase center is at the reﬂector focus. As is true of axial defocusing loss, deep dishes are affected more than shallow reﬂectors.

8-6 FEED SCANNING Moving the phase center of the feed off axis laterally scans the reﬂector beam to a limited extent without severe pattern problems. Figure 8-5 shows the k-space pattern effects of feed scanning. The sidelobes show the effects of coma (cubic phase errors) where the sidelobes on the boresight side grow and the sidelobes on the other side decrease. We call these coma lobes, although no new lobes are generated. In fact, we see one lobe disappearing as a vestigial lobe with increased scan (Figure 8-5). Suppose that the feed is offset from the axis by a distance d. We measure the offset angle ψS from the axis to a line from the feed to the reﬂector vertex: d = f tan ψS . We ignore the slight amplitude distribution change due to the small lateral offset. Referred to the focus, the movement produces a phase factor in the feed pattern: −kd sin ψ cos φc when the feed is moved along the negative x-axis. Equation (8-7) predicts only the boresight phase error loss. Like Eq. (4-3), we must calculate the phase error efﬁciency at any angle to determine the loss at the pattern peak:

PEL(θ, φ) =

2π 0

ψ0 0

2 E(ψ, φc ) tan(ψ/2)ej k2f tan(ψ/2) sin θ cos(φ−φc ) dψ dφc

2

(8-15)

|E(ψ, φc )| tan(ψ/2) dψ dφc When we include the offset along φ = 0, the phase factor becomes ψ exp j kf cos φc 2 tan sin θ − tan ψS sin ψ 2

0 2pa tan y = 8 s l

Pattern Level, dB

−10

ys Feed

−20

a

−30 −40 −50 −60

−20

−15

−10

−5

5 0 2pa sin q l

10

15

20

FIGURE 8-5 Feed-scanned paraboloidal reﬂector f/D = 0.5 and feed beamwidth = 60◦ .

FEED SCANNING

391

For large reﬂectors we make the approximations ψS ≈ tan ψS and θ ≈ sin θ . The pattern scale and the offset phase factor become kaθ and kaψS . A ﬂat plate would reﬂect the ray at an equal angle on the other side of the axis for an offset feed, but a curved reﬂector modiﬁes that result slightly. The offset factor in Figure 8-5 is 8, and the beam peak is at 7. We call the ratio of the beam maximum to offset angle the beam deviation factor (BDF) [12]: BDF =

θm 7 = ψS 8

θm = BDF · ψS

The BDF varies from less than 1 for a concave reﬂector to greater than 1 for a convex reﬂector. BDF equals 1 for a ﬂat reﬂector. Table 8-1 lists the BDF values for various f/D and Scale 8-6 gives the relationship. The BDF approaches 1 as f/D approaches inﬁnity (ﬂat plate). The approximate expression for BDF is BDF =

(4f/D)2 + 0.36 (4f/D)2 + 1

(8-16)

Feed scanning increases the phase error loss. When normalized to beamwidths of scan, a single loss curve can be drawn for each f/D (Figure 8-6). Scanning also raises the sidelobes. Table 8-2 gives the approximate level of the peak coma lobe for a given scan loss. It is almost independent of f/D.

TABLE 8-1 Feed-Scanned Paraboloidal Reﬂector Beam Deviation Factor f/D

BDF

f/D

BDF

0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75

0.724 0.778 0.818 0.850 0.874 0.893 0.908 0.921 0.930 0.938

0.80 0.85 0.90 1.00 1.10 1.20 1.40 1.60 1.80 2.00

0.945 0.951 0.957 0.965 0.970 0.975 0.981 0.986 0.989 0.991

Beam Deviation Factor

Parabola f/D

SCALE 8-6 Feed-scanned reﬂector beam deviation factor given f/D.

392

REFLECTOR ANTENNAS 0

Scanning Loss, dB

1.5 1.2 1.0 1

2 F/D 0.3

0.4

0.5

0.6

0.7

3 2

4

6

8 10 12 14 Beamwidths of Scan

16

20

FIGURE 8-6 Feed-scanning loss of a paraboloidal reﬂector.

TABLE 8-2 Sidelobe Level of a Feed-Scanned Paraboloidal Reﬂector Scanning Loss (dB)

Sidelobe Level (dB)

Scanning Loss (dB)

Sidelobe Level (dB)

0.50 0.75 1.00 1.25 1.50

14.1 12.9 11.9 11.2 10.6

1.75 2.0 2.5 3.0

10.1 9.7 9.0 8.5

Example A reﬂector with a 50λ diameter is feed-scanned to 6◦ . Compute the offset distance and scanning loss when f/D = 0.6. Use the approximation HPBW = 70◦ λ/D = 1.4◦ . The reﬂector is scanned 6/1.4 = 4.3 beamwidths: scanning loss (Figure 8-6) = 0.4 dB sidelobe level (Table 8-2) = 14.6 dB The angle between the axis and the feed point to vertex must be greater than the scan angle, since the reﬂector is concave: ψS =

θS 6◦ ◦ = = 6.61 BDF 0.908

(Table 8-1)

The offset distance is f tan 6.61◦ = 0.6(50λ) tan 6.61◦ = 3.48λ.

RANDOM PHASE ERRORS

393

The scalar analysis of this section gives only approximate results. Large feed scanning produces higher-order aberrations other than coma [13–15]. The optimum gain point moves off the focal plane but fails to follow the curve predicted from optics for reﬂectors extremely large in wavelengths [14]. The reﬂector f /D and illumination taper determine the maximum gain contour for feed-scanning a reﬂector. Vector analysis improves the match between calculated and measured results [15].

8-7 RANDOM PHASE ERRORS Reﬂector anomalies reduce the gain predicted from the feed analysis. We must specify reasonable manufacturing tolerances for the frequency of operation. It would appear that gain can be increased without bound by increasing the reﬂector diameter, but the tolerance problems of large reﬂectors limit the maximum gain. We consider only surface anomalies so small that on average the reﬂector retains its basic shape. The surface imperfections change the optical path length from the feed to the reﬂector aperture plane by δ(r, φ), which gives us 2π a 2 j δ(r,φ) E(r, φ)e r dr dφ 0 0 PEL = 2 2π a |E(r, φ)|r dr dφ 0

(8-17)

0

Cheng [16] bounds the phase error loss by using a limit on the integrals. Given a peak phase error of m (radians), the change in gain is bounded: 2 G m2 ≥ 1− G0 2

(8-18)

This gain loss estimate is too conservative, but it is useful as an upper bound. Ruze [17] improved the random surface error loss estimate by using a Gaussian distributed error correlated over regions. Dents or segments making up the reﬂector are correlated with the errors over a nearby region. The error at a point depends on the location of nearby points in the correlation region. The phase error efﬁciency becomes an inﬁnite series: 2

PEL = exp(−δ ) +

1 η

2C D

2

2

exp(−δ )

2 ∞ (δ )n n · n! n=1

(8-19)

where C is the correlation distance, D the diameter, and η the aperture efﬁciency 2 (ATL). δ is the mean-square phase deviation, given by 2

δ =

2π 0

a

0 2π

0

|E(r, φ)|δ 2 (r, φ)r dr dφ a |E(r, φ)|r dr dφ 0

(8-20)

394

REFLECTOR ANTENNAS

If we include the correlation distance, PEL decreases. The inﬁnite series [Eq. (8-19)] converges rapidly. When the correlation distance is small compared with the diameter, the phase error efﬁciency becomes −4πε0 2 2 PEL = exp = exp(−δ ) (8-21) λ where ε0 is the effective reﬂector tolerance. We use 4π instead of 2π because the wave travels to and from the reﬂector and the phase distance is twice the reﬂector tolerance. From Eq. (8-20) we derive the effective RMS tolerance: 2π a |E(r, φ)|ε2 (r, φ)r dr dφ 0 0 2 εo = (8-22) 2π a |E(r, φ)|r dr dφ 0

0

Ruze gives the distance ε in terms of the z-axis deviation z and the surface normal ε=

z 1 + (r/2f )2

n ε=

1 + (r/2f )2

We evaluate the constants in Eq. (8-21) and convert to a decibel ratio: ε 2 0 PEL(dB) = −685.8 λ

(8-23)

(8-24)

Example Compute the required reﬂector tolerance at 30 GHz to limit the RMS surface tolerance phase error loss to 1 dB. Using Eq. (8-24), we get ε0 1 = = 0.038 λ 685.8 At 30 GHz, λ = 1 cm and ε0 = 0.38 mm. We can also use Eq. (8-18), which gives the upper bound on surface error loss:

4πε0 G m= = 0.466 at 1 dB = 2 1− λ G0 ε0 = 0.037λ or ε0 = 0.37 mm at 30 GHz. Both methods give about the same answer in this case. Zarghamee [18] extended tolerance theory to include the effects of the surface error distribution. Some antennas have better support and construction in some areas and are more accurate in those areas. This improves the reﬂector performance. Zarghamee deﬁned a second variation of surface deviations by 2π a |E(r, φ)|[ε2 (r, φ) − ε02 ]r dr dφ 0 0 4 η0 = 2π a |E(r, φ)|r dr dφ 0

0

RANDOM PHASE ERRORS

395

The phase error efﬁciency becomes

−4πε0 PEL = exp λ

2 exp

πη 4 0

λ

The correlation of random errors increases the probable sidelobe level. The sidelobe level increases with the size of the correlation interval and decreases for larger aperture diameters. Increasing the amplitude taper of the distribution makes the aperture pattern more susceptible to random-error sidelobes, since increasing the taper is somewhat equivalent to decreasing the aperture diameter. Blockage and feed diffraction also limit the achievable sidelobe level in a reﬂector. A simple feed cannot carefully control the aperture distribution necessary for low sidelobes. Hansen [19, p. 74] discusses sidelobe limitations caused by random phase error in some detail. Paraboloidal reﬂectors can be made in an umbrella shape where the ribs are parabolic and wire mesh is stretched between them [20]. The gore shape causes phase error loss and their periodicity produces extra sidelobes. Given the number of gores NG and the focal length of the ribs fr , the surface is given by f (ψ) = fr

cos2 (π/NG ) cos2 ψ

where ψ is measured from the centerline between the ribs. We calculate the average focal length by integrating across the gore half-angle π/NG and dividing by π/NG : fav = fr

sin(2π/NG ) 2π/NG

(8-25)

We use Eq. (8-25) to calculate the rib focal length given the average focal length of the reﬂector. The peak sidelobe due to the periodic gores occurs at an angle θp found from the number of gores and the diameter D: λ (8-26) θp = sin−1 1.2NG πD Given the average f /D of the reﬂector, we determine the peak-to-peak phase deviation across the gore by the approximate equation =

800 − 500(f/D − 0.4) D λ NG2

(8-27)

Scale 8-7 lists the phase error loss for a feed edge taper of 10 dB. Increasing the feed taper decreases the phase error loss due to gore construction. When we use a 20-dB feed taper, the values given by Scale 8-7 reduce by 0.16 dB for 0.5 dB of loss, 0.31 dB for 1 dB, and 0.45 dB for 1.5 dB. The gain losses due to underillumination by the 20-dB edge taper feed exceed these values. Example Given a reﬂector with D/λ = 35 with a limit of 0.5 dB loss due to gore construction for f/D = 0.34, we discover that the allowable peak-to-peak phase error from Scale 8-7 is 124◦ . Using Eq. (8-27), we solve for the number of ribs: NG2 =

830 35 = 234 124

or

NG = 16

396

REFLECTOR ANTENNAS Phase Error Loss, dB

Peak-to-Peak Phase Error due to Gores (degrees)

SCALE 8-7 Phase error loss due to gore construction of a paraboloidal reﬂector.

We use Eq. (8-26) to compute the angle of the peak gore sidelobe, θp = 10.0◦ . Equation (8-27) shows that the phase deviation is proportional to frequency. If the frequency increases by 1.5 times, then by Eq. (8-27), 124◦ increases to 186◦ and we read 1.1 dB of loss from Scale 8-7 while the peak gore sidelobe becomes θp = 6.7◦ . 8-8 FOCAL PLANE FIELDS We improve the efﬁciency and pattern response of a reﬂector if we match the feed ﬁelds to the focal plane ﬁelds. GO assumes a point focus, but an actual focus is extended. We determine the reﬂector and feed efﬁciency from the ﬁeld match over the focal plane. When the reﬂector f /D value is large, we use the diffraction pattern of a circular aperture, the Airy function: J1 (krψ0 ) E= (8-28) krψ0 where ψ0 is the half subtended angle of the reﬂector (radians), r the radial coordinate, k the propagation constant, and J1 the Bessel function. In a more exact method the currents induced on the reﬂector (2n × H) and the magnetic vector potential are used to calculate the focal plane ﬁelds. As f /D decreases, the currents on the reﬂector interact and modify their distribution, but it is a secondary effect [21]. Iterative physical optics analysis (section 2-4) can ﬁnd these current modiﬁcations. We calculate the reﬂector efﬁciency from the ﬁeld match of the focal plane ﬁelds (E1 , H1 ) and the ﬁelds of the feed (E2 , H2 ) using Robieux’s theorem [4]:

η=

2 (E1 × H2 − E2 × H1 ) · dS S

4P1 P2

(8-29)

where P1 and P2 are the input powers to produce the ﬁelds and η is the efﬁciency. Equation (8-29) is the magnitude squared of Eq. (2-35), the reactance equation equivalence applied to Eq. (1-55) for the coupling between two antennas S21 . The ﬁnite size of the feed causes spillover. The extent of amplitude and phase mismatch between the two ﬁelds determines the efﬁciency. By illuminating the reﬂector with a crosspolarized wave, we compute the cross-polarization radiation level through its ﬁeld match [Eq. (8-29)]. We maximize efﬁciency [Eq. (8-29)] by conjugate-matching the focal plane ﬁelds with the feed ﬁelds. Corrugated horns can be designed by expanding the focal plane

FEED MISMATCH DUE TO THE REFLECTOR

397

ﬁelds in axial hybrid modes of the horn and mode matching [22,23]. Wood [4] expands the reﬂector and feed ﬁelds in spherical harmonics and matches them at a boundary. Both sets of ﬁelds can be approximated very well by just a few terms, and this method can handle dual-reﬂector and offset reﬂector systems as well as axisymmetric prime focus reﬂectors. We can feed the reﬂector with an array to match the focal plane ﬁelds [24–26]. The array samples the focal plane ﬁeld and conjugate-matches it so that the powers sum in phase. The array can form multiple beams and also correct reﬂector aberrations [24]. By using the multiple feeds of the array, coma can be reduced for scanned beams and efﬁciency improved. However, quantization of the array element locations and excitations, amplitude, and phase reduces efﬁciency and raises the sidelobe level [27]. We apply Eq. (1-55) for the coupling between two antennas to determine the feeding coefﬁcients of an array feed for a dish. Assume an incident ﬁeld distribution on the reﬂector that includes the incident wave direction and the desired aperture distribution for the reﬂector. Using physical optics, we calculate the currents induced on the reﬂector surface. If the reﬂector has signiﬁcant curvature so that the patches face each other, iterative PO can be used to account for their interaction. We calculate the ﬁelds radiated by each feed on the reﬂector surface and apply Eq. (1-55) to calculate coupling. This method applies the feed pattern to the calculation instead of the point matching used in a focal plane solution. Similar to scanning of an array, we use conjugative matching for the feed array elements to produce the beam desired. This method can determine array feed element amplitude and phase for any composite reﬂector aperture distribution that includes aperture distribution to control sidelobes or include multiple beams. The method reduces coma to the minimum possible with a given array. Analysis ﬁnds the array distribution desired, but we do not achieve this distribution merely by designing the feed network to produce these amplitudes and phases because the feed elements have signiﬁcant mutual coupling. We need to include the effect of the paraboloidal reﬂector when computing mutual coupling because the ﬁeld radiated by one feed induces currents on the reﬂector that couple to other feed elements. Below we show that the effect of the reﬂector diminishes as the reﬂector diameter increases. If the mutual coupling is signiﬁcant whether direct or due to the reﬂector, we need to apply the corrections given in Section 3-11 to adjust the feeding coefﬁcients of the array. 8-9 FEED MISMATCH DUE TO THE REFLECTOR The feed receives some of its transmitted power because it reﬂects from the parabola and returns as a mismatch at the feed terminals. We calculate the reﬂected ﬁeld at the feed by using surface currents and the magnetic vector potential. The only signiﬁcant contribution comes from areas near where the normal of the reﬂector points at the feed. Around every other point, the phase of the reﬂection varies rapidly and cancels and we need to consider only points of stationary phase. We calculate the reﬂection from each point of stationary phase from [2] Gf (ρ0 ) ρ1 ρ2

= −j e−j 2kρ0 (8-30) 4kρ0 (ρ1 + ρ0 )(ρ2 + ρ0 )

398

REFLECTOR ANTENNAS

where is the reﬂection coefﬁcient, ρ0 the distance to the stationary phase point, Gf (ρ0 ) the feed gain in the direction of ρ0 , and ρ1 and ρ2 the radiuses of curvature of the reﬂector at ρ0 . The vertex is the only point of stationary phase on a paraboloidal reﬂector: ρ1 = ρ2 = −2f and ρ0 = f . Equation (8-30) reduces to

= −j

Gf (0) −j 2kf e 2kf

(8-31)

Example Suppose that we have a reﬂector with f/D = 0.40. Compute reﬂector mismatch for a source with its 10-dB beamwidth equal to the reﬂector subtended angle. Half subtended angle [Eq. (8-2)] ψ0 = 2 tan(1/1.6) = 64◦ . By using the feed approximation cos2N (θ /2), we have N=

log 0.1 = 6.98 2 log cos(64◦ /2)

The feed gain at the boresight is N + 1 [Eq. (1-20c)]: [Eq. (8-31)]

| | =

λ 8λ = 1.59 4πf D

Increasing the reﬂector diameter in wavelengths decreases the reaction of the reﬂector on the feed. For example, given a 3-m reﬂector at 4 GHz, we calculate reﬂector reﬂection coefﬁcient as 0.04, or VSWR = 1.08. We can express the reﬂector reﬂection of a paraboloidal reﬂector as | | = V

λ D

(8-32)

and calculate Scale 8-8 of V versus f /D for feeds with 10-dB beamwidths equal to the reﬂector subtended angle. Higher reﬂector f /D values produce larger feed reﬂections, since the feed gain increases faster than the reduced area of the reﬂector seen from the feed. Narrowband corrections to these reﬂections can be designed by using a vertex plate (Silver [2]) or by designing sets of concentric ring ridges in the reﬂector (Wood [4]). The rings can match the feed at more than one frequency. By any of these methods, the free-space mismatch of the feed could be corrected for, but, of course, the feed itself can be mismatched to compensate for the reﬂector reaction.

Feed Reflection Coefficient Factor, V

Parabola f/D

SCALE 8-8 Feed reﬂection scale factor V given f/D.

OFFSET-FED REFLECTOR

399

Front/Back Increase K Factor, dB

Parabola f/D

SCALE 8-9 Paraboloidal reﬂector front-to-back ratio increase K given f/D.

8-10 FRONT-TO-BACK RATIO Figure 2-9 illustrates the pattern response of a paraboloidal reﬂector and shows that the pattern behind the reﬂector peaks along the axis. The diffractions from all points along the rim add in-phase along the axis and produce a pattern peak. We can reduce this rim diffraction by using a rolled, serrated, or castellated edge to reduce diffraction. An absorber-lined cylindrical shroud extending out to enclose the feed will greatly reduce back radiation, including spillover, and allows the close spacing of terrestrial microwave antennas with reduced crosstalk. For a normal truncated circular reﬂector rim, the following equation estimates the front-to-back ratio given the reﬂector gain G, the feed taper T , and feed gain Gf [28]: F/B = G + T + K − Gf

dB

(8-33)

The constant K, given by Scale 8-9, is related to f /D: K = 10 log 1 +

1 (4f/D)2

(8-34)

Example Estimate F/B for a reﬂector with f/D = 0.34 and 40 dB of gain. We read the feed subtended angle from Scale 8-1 to be 143◦ . A 10-dB edge taper feed has a gain of about 8.1, found from Scale 1-2. Using Eq. (8-33), we estimate F/B = 40 + 10 + 1.9 − 8.1 = 43.8 dB.

8-11 OFFSET-FED REFLECTOR Moving the feed out of the aperture eliminates some of the problems with axisymmetrical reﬂectors. Blockage losses and diffraction-caused sidelobes and cross-polarization disappear. We can increase the size of the feed structure and include more if not all of the receiver with the feed. For example, the reﬂector may be deployed from a satellite, with the feed mounted on the main satellite body. Figure 8-7 shows the offset-fed reﬂector geometry. We form the reﬂector out of a piece of a larger paraboloid. Every piece of the paraboloidal reﬂector converts spherical waves from the focus into a plane wave moving parallel with its axis. We point the feed toward the center of the reﬂector to reduce the spillover, but we still locate the feed phase center at the focus of the reﬂector. The aperture plane projects to a circle, although the rim shape is an ellipse. ψ0 is the angle from the axis of the parabola to

400

REFLECTOR ANTENNAS

ru

yc

D

ye

yf

D′

rL

ye

yo

H

yu

ƒ

FIGURE 8-7 Parameters of an offset-fed parabolic reﬂector.

the center of the cone of the reﬂector, and the reﬂector subtends an angle 2ψe about this centerline. Given the aperture plane diameter D and the height H of the center, we ﬁnd the lower rim offset D = H − D/2. From these parameters we determine the angle of the center of the rim cone from the z-axis: ψ0 = tan−1

16fH 2f (D + 2D ) −1 = tan 16f 2 + D 2 − 4H 2 4f 2 − D (D + D )

(8-35)

The half cone angle deﬁnes the rim: ψe = tan−1

8fD 2f D = tan−1 16f 2 + 4H 2 − D 2 4f 2 + D (D + D )

(8-36)

We direct the feed an angle ψf from the z-axis to the center of the projected diameter different from the angle ψ0 of the rim cone axis: ψf = 2 tan−1

H 2D + D = 2 tan−1 2f 4f

(8-37)

The rim lies in a plane at an angle ψc with respect to the z-axis: ψc = tan−1

2f 4f = tan−1 H 2D + D

(8-38)

401

OFFSET-FED REFLECTOR

The rim is an ellipse in this plane with major and minor axes given by ae =

D 2 sin ψc

and be =

D 2

(8-39)

The offset angle modiﬁes the f /D of the reﬂector: cos ψe + cos ψ0 f = D 4 sin ψe

(8-40)

We calculate the rim offset from the cone angles: D = 2f tan

ψ0 − ψe 2

(8-41)

Manufacturing an offset reﬂector requires speciﬁcation of the reﬂector when laid on its rim in the x –y plane so that the mold can be machined. We center the major axis of the reﬂector elliptical rim L = 2ae along the x-axis and the minor axis D along the y-axis. In this position the reﬂector depth d(x, y) is found from the expression [29] √ 2f L3 xD 2 L2 − D 2 D 2 (L2 − D 2 ) D 2 2 d(x, y) = 1+ + −y D(L2 − D 2 ) f L3 4f 2 L4 4 √ xD 2 L2 − D 2 −1− 2f L3

(8-42)

The deepest point of the reﬂector dmax occurs along the x-axis at xb : √ D 2 L2 − D 2 xb = − 16f L

where dmax =

D3 16f L

(8-43)

After measuring D, L, and dmax , we determine the offset focal length from the equation f =

D3 16Ldmax

(8-44)

We calculate the center height of the offset from H = 2f

L2 −1 D2

(8-45)

We calculate the reﬂector half cone angle ψe and the cone axis angle from the z-axis and ψ0 from the focal length f , ellipse major diameter L, and minor diameter D:

ψe ψ0

= tan−1

D D L L ∓ tan−1 −1+ −1− D2 4f D2 4f 2

2

(8-46)

402

REFLECTOR ANTENNAS

To align the reﬂector, we use the angle of the reﬂector rim major axis ψc = sin−1 (D/L) with respect to the z-axis and the radial distances from the lower and upper edges of the reﬂector in the offset plane, since the center offset H is not a distinguishable point:

ρU ρL

=

2 L D2 fL2 ± D + − 1 D2 16f D2

(8-47)

We analyze the offset reﬂector with the same tools as those used with the axisymmetric reﬂector: aperture ﬁeld, physical optics, and GTD. The asymmetry of the reﬂector to feed geometry introduces anomalies. Huygens sources no longer eliminate crosspolarization, because the source must be tilted. Symmetry prevents cross-polarization in the plane containing the x-axis (Figure 8-8), but cross-polarization for linear polarization increases in the plane containing the y-axis (symmetry plane) as f /D decreases

Diameter

2ye y0

x

Focal le

ngth

y

Focus

(a)

x

Diameter y 2ye y0 Focus Focal length (b)

FIGURE 8-8 Offset-fed paraboloidal reﬂector geometry: (a) perspective; (b) orthographic representation.

OFFSET-FED REFLECTOR

403

Amplitude, dB

E-Plane

Symmetrical Plane Cross-Polarization H-Plane

Pattern Angle (degrees)

FIGURE 8-9 Pattern of an offset-fed reﬂector with linearly polarized feed.

LHC

Amplitude, dB

RHC

Pattern Angle (degrees)

FIGURE 8-10 Pattern of an offset-fed reﬂector with circularly polarized feed.

(Figure 8-9). The Condon lobes move off the diagonal planes and into the plane containing the y-axis. The asymmetry along the x-axis tapers the amplitude distribution from a symmetrical feed, since the spherical wave travels farther to the outer edge of the reﬂector than to the lower edge. The offset-fed reﬂector geometry squints circularly polarized pattern peaks in the symmetrical (y-axis) plane without generating cross-polarization (Figure 8-10). An approximate formula for the squint is [9] ψs = sin−1

λ sin ψ0 4πf

(8-48)

404

REFLECTOR ANTENNAS

where ψs is the squint angle. Opposite senses of circular polarization squint in opposite directions and cause a problem with dual circularly polarized feed systems. In all cases increasing the f /D or the effective f /D through a subreﬂector reduces these problems. The subreﬂector should be kept out of the aperture of the main reﬂector. We can feed-scan the offset-fed reﬂector by moving the feed laterally along a line that lies perpendicular to the boresight of the feed (the line deﬁned by ψ0 ). We must modify the beam deviation factor (BDF): BDFoffset

fed

= BDFcenter

fed

(f/D)offset (f/D)center fed

(8-49)

Example Given an offset-fed reﬂector with ψ0 = 45◦ and ψe = 40◦ , compute the beam deviation factor. From Eq. (8-49), f cos 40◦ + 1 = = 0.687 D center fed 4 sin 40◦ cos 40◦ + cos 45◦ f = = 0.573 D offset fed 4 sin 40◦ From Table 8-1 we interpolate BDFcenter fed = 0.928, and we substitute the values into Eq. (8-49) to calculate BDFoffset fed = 0.774. We must laterally offset the feed farther than with a center-fed reﬂector to achieve the same feed scanning. Periscope Conﬁguration The periscope consists of an offset paraboloidal reﬂector with ψ0 = 90◦ with a long focal length fed by a paraboloidal reﬂector located at the focus. This eliminates the need to run a transmission line up a tower. Periscope antennas can be made using a ﬂat-plate reﬂector, but the long focal length means that the parabolic splash plate antenna has only a small deviation from ﬂat. The ﬂat plate is limited to a gain of only 6 dB more than the feed reﬂector for optimum conditions with a large plate. The gain of the offset paraboloidal reﬂector is determined by the diameter of the splash plate, not the feed reﬂector. Because the splash plate is in the near ﬁeld of the feeding reﬂector, gain is reduced by phase error, whereas spillover and amplitude taper losses also contribute to gain loss. Design starts with determining the splash reﬂector center height H required to clear obstacles along the transmission path. We calculate the splash reﬂector aperture diameter from the required gain and beamwidth. The periscope conﬁguration contributes to gain loss, but with proper selection of the feed paraboloidal reﬂector these losses are minor and can be compensated for by using a larger splash reﬂector. Having the splash reﬂector directly overhead corresponds to a parent reﬂector design with f/Dp = 0.25 and f = H /2. An analysis using a radial parabolic aperture distribution in the feed reﬂector determined that the optimum feed reﬂector diameter is found from the ratio of height to projected splash reﬂector aperture diameter Ds [30]: Df =

2λH αλH = Ds Ds

or

α=

Ds Df Ds Df F = λH Hc

(8-50)

Whereas α = 2 is the optimum dimensions at a particular frequency, we account for shift from the optimum with this factor. The parameter α is the frequency response

REFLECTIONS FROM CONIC SECTIONS

405

factor for frequency F and speed of light c. The illumination efﬁciency is the product of the feed reﬂector illumination efﬁciency and the periscope efﬁciency factor ηp : 4 1 − (1 − K)J0 (m) − K(2/m)J1 (m) ηp = (8-51) m2 (1 − K/2)2 J0 and J1 are Bessel functions, m = απ/2, and K = 1 − 10−[ET(dB)/20] for the feed reﬂector edge illumination taper ET(dB). Table 8-3 lists the added illumination loss of a periscope, given geometry using α in Eq. (8-50) for a 12-dB edge taper in a feeding reﬂector. Example A periscope antenna system placed a 3-m projected aperture splash reﬂector 30 m above the feed reﬂector to operate at 12 GHz (λ = 0.05 m). Using Eq. (8-50), we calculate the feed reﬂector diameter to be 0.5 m for α = 2. If we assume that the feed reﬂector has an efﬁciency of 60% (−2.22 dB), the efﬁciency of the splash reﬂector will be −2.6 dB: gain (dB) = 20 log

3π πDs − 2.6 = 20 log − 2.6 = 42.9 λ 0.05

The focal length f of the splash reﬂector is H /2 = 15 m. Since the angle of the splash reﬂector rim is 45◦ , L = Ds / sin(45◦ ) = 4.24 m. We determine the maximum depth of the reﬂector by using Eq. (8-43) to be 2.65 cm located 2.65 cm off center. The splash reﬂector increased the gain relative to the feed reﬂector by 7.4 dB. 8-12 REFLECTIONS FROM CONIC SECTIONS We use reﬂectors made from conic sections other than the parabola as subreﬂectors. The ellipse and hyperbola rotated about their axes to form solid ﬁgures that reﬂect incident spherical waves into spherical waves with different caustics (focal points). Reﬂectors formed by moving the ﬁgure along a line change the caustics of cylindrical waves. We consider only spherical waves, but we need only convert to cylindrical waves for cylindrical reﬂectors. All conic-section reﬂectors convert spherical waves from one focus into spherical waves directed toward the other focus. The ellipse has its two focuses located within the TABLE 8-3 Added Illumination Loss of a Periscope, Given Geometry Using α in Eq. (8-50) for 12-dB Edge Taper in a Feeding Reﬂector (dB) α 1.0 1.2 1.4 1.6 1.8 2.0

ηp

α

ηp

3.17 2.06 1.28 0.77 0.48 0.38

2.2 2.4 2.6 2.8 3.0 3.2 3.4

0.45 0.68 1.04 1.52 2.11 2.79 3.53

406

REFLECTOR ANTENNAS

ﬁgure. As we let one focus approach inﬁnity, the ellipse transforms into a parabola. If the focus is pushed through inﬁnity to the negative axis, the ﬁgure becomes a hyperbola located between the two focuses. Figure 8-11 shows the ray tracing for axisymmetrical conic section reﬂectors. A spherical source at one focus is reﬂected to the second focus by the reﬂector, although it is virtual (not actually reached) in some cases. We describe all conic sections with the same polar equation: ρ=

eP 1 − e cos θ

(8-52)

where P is the distance between the origin, the focus, to a line called the directrix (Figure 8-12). The eccentricity e is the ratio of the distance from the origin to a point on the curve to the distance from the same point to the directrix: r1 = er2 . In an ellipse, e < 1; in a parabola, e = 1; and in a hyperbola, e > 1. The distance between the focuses is 2P e2 (8-53) 2c = 1 − e2

2c

F1

F2

2b

Ellipse 2a

Parabola Focus

Hyperbola

Focus

Hyperbola

Focus Focus

Focus

FIGURE 8-11 Reﬂections from conic-section reﬂectors.

REFLECTIONS FROM CONIC SECTIONS

407

R2 Directrix

R1

q Focus

P

R= e=

eP 1 − e cos q R1 R2

FIGURE 8-12 Conic-section geometry.

A hyperbola with its axis containing the two focuses along the z-axis, located at ±c, intersects the z-axis at ±a and satisﬁes the equation r2 z2 − =1 a2 b2

where b2 = c2 − a 2

and e =

c >1 a

(8-54)

When we take the portion of the hyperbola along the +z-axis that intersects the axis at +a, we deﬁne the angles from the two focuses from the line between them because we place the feed at the left focus and locate a parabola focus at the right hyperbola focus. The left angle θ is the feed angle, and the right angle ψ is the parabola angle in a dual-reﬂector antenna. Given a point on the hyperbola, the distance from the left focus is ρ1 and the distance from the right focus is ρ2 : ρ1 =

b2 a(e2 − 1) = e cos θ − 1 e cos θ − 1

and ρ2 =

b2 a(e2 − 1) = e cos ψ + 1 e cos ψ + 1

(8-55)

We determine the radial position off the axis from either polar equation: r = ρ1 sin θ = ρ2 sin ψ

(8-56)

The two angles are related by the eccentricity e: (e + 1) tan

θ ψ = (e − 1) tan 2 2

(8-57)

At a given point on the hyperbola, the angle of the normal u relative to the radial line ρ1 is half the sum of the two angles, u = (θ + ψ)/2. We need the radius of curvatures in the principal planes to apply Eq. (2-77) for reﬂection from a curved surface: R1 in

408

REFLECTOR ANTENNAS

the r –z plane and R2 in the φ –z plane: R1 =

b2 a cos2 u

and R2 =

b2 a cos u

(8-58)

When we offset-feed a hyperboloid, a cone determines the rim and it lies in a planar ellipse. Similar to the offset paraboloid, we center the cone at an angle θ0 from the axis and deﬁne the cone by the half feed edge angle θe . We compute the distances from the focus to the upper and lower rim along the major axis of the rim ellipse: ρL =

b2 e cos(θ0 − θe ) − 1

and ρU =

b2 e cos(θ0 + θe ) − 1

We determine the major axis diameter (2ae ) of the elliptical rim from the triangle with sides ρL and ρU and angle 2θe between them: 2ae = ρL2 + ρU2 − 2ρL ρU cos 2θe (8-59) The minor axis diameter (2be ) is given by the equation

2be = (2ae )2 − (ρL − ρU )2

(8-60)

An ellipsoid can also be used as a subreﬂector in a dual-reﬂector antenna. Its equations are similar to the hyperboloid: r2 z2 + = 1 and b2 = a 2 − c2 a2 b2

with

e=

c 0

where Jn is the Bessel function and Hn(2) is the outward-traveling Hankel function, with Jn the derivative of Jn , and so on. Table 8-8 lists these factors for both polarizations versus strut radius. Using only physical optics we can determine accurately the blockage effects of struts that are at least 3λ in diameter. PO excites currents only on the visible half of the struts. For smaller-diameter struts, currents creep to the far side and alter the results. No matter how thin the struts, currents will be excited on them and affect the pattern. The physical optics analysis of a dual reﬂector includes currents excited on the struts a number of times. Assume that the antenna is transmitting. The feed illuminates the subreﬂector and the struts. The current excited on the struts also radiates a ﬁeld that illuminates the subreﬂector. If the strut blocks the path between the feed and the subreﬂector, PO analysis uses strut current to calculate its blockage. At this point we use the currents TABLE 8-8

IFRE and IFRH for a Circular Strut

a sin θ0

Re(IFRE )

Im(IFRE )

0.005 0.010 0.020 0.050 0.10 0.20 0.50 1.00 2.00

−5.148 −3.645 −2.712 −1.982 −1.641 −1.414 −1.215 −1.145 −1.092

14.088 6.786 3.964 2.003 1.225 0.758 0.381 0.255 0.160

Re(IFRH )

Im(IFRH )

−0.0001 −0.0005 −0.004 −0.054 −0.292 −0.552 −0.781 −0.858 −0.914

−0.0198 −0.050 −0.103 −0.272 −0.448 −0.374 −0.258 −0.188 −0.126

FEED AND SUBREFLECTOR SUPPORT STRUT RADIATION

419

on the subreﬂector to compute additional currents excited on the struts. These add to the ﬁrst set of strut currents. Radiation from the subreﬂector, all currents on the struts, and stray feed illumination add to illuminate the main reﬂector. The radiation from the main reﬂector current excites additional current on the subreﬂector and a third set of currents on the struts. We apply the far-ﬁeld Green’s function on the sum of all currents to calculate the pattern. PO currents are modiﬁed on thin struts to account for creeping-wave currents. We multiply the PO strut currents by the induced current ratio (ICR) to obtain equivalent currents suitable for predictions. The factor ICR includes a strut current distribution and a complex value: Js = 2n × Hinc · ICR(a, θ0 , φ ) (8-76) ICR depends on the strut radius a, the incidence angle θ0 with respect to the strut axis, and the angle around the strut φ from the direction of the plane wave and the incident wave polarization. We ﬁnd the incident magnetic ﬁeld Hinc at the point where the plane wave touches the strut and use the current excited at this point to calculate the current in a ring around the strut. Remember that the primary effect of the strut is to block the radiation from the main reﬂector that approximates a plane wave in the near ﬁeld where the struts are located. We solve for ICR by considering two-dimensional scattering of the strut cross section by a plane wave. By applying moment methods to a two-dimensional scattering problem, we can solve for the current distribution on any strut cross section, but here we consider only circular struts that have a closed-form solution [40, pp. 209–219]. To simplify the problem, consider a strut lying along the z-axis. For actual analysis you will need to rotate the strut into place and rotate the incident wave into the strut coordinate system to use ICR to calculate the current distribution. In two-dimensional space the incident wave is either TM or TE with respect to the z-axis. The TM wave has its electric ﬁeld in the plane containing the strut axis. A TE wave has its magnetic ﬁeld in this plane. The TM case produces the following equation from the scattering of a plane wave: ICRE zˆ = ICRTM (a, θ0 , φ )ˆz =

∞ zˆ e−j ka cos φ j m εm cos mφ πka sin θ0 m=0 Hm(2) (ka sin θ0 )

(8-77)

The phase factor e−j ka cos φ shifts the reference plane from the strut center to the attachment point. Equation (8-77) expands the current in a cos mφ Fourier series around the strut. ICRE has a complex value because Hm(2) has a complex value. As the strut radius a increases, ICR approaches 1 at the location φ = 0. For practical purposes ICR is 1 for a/λ > 1.5. Equation (8-77) gives the current distribution on the circular strut relative to the current excited at the initial contact point of the incident plane wave. Table 8-9 lists ICRE evaluated when φ = 0. The constant term grows rapidly as a → 0, with its imaginary part growing faster than the real part because it approximates the vector potential of a ﬁlamentary current element with its −j factor between the current and the ﬁeld [Eq. (2-1)]. Small struts have nearly constant current around their periphery. Equation (8-77) requires more and more terms as the strut diameter increases, and ﬁnally, a simple PO solution produces the same results. Table 8-9 lists the ICR factors versus the strut radius.

420

REFLECTOR ANTENNAS

TABLE 8-9 ICRE and ICRH PO Current Multipliers When φ = 0 for a Circular Strut a sin θ0 Re(ICRE ) Im(ICRE ) Re(ICRH ) Im(ICRH ) 0.002 0.004 0.01 0.02 0.05 0.10 0.50 1.00 1.50

−7.932 −4.530 −2.228 −1.353 −0.751 −0.479 −0.140 −0.076 −0.052

3.660 2.735 2.018 1.687 1.376 1.200 1.030 1.010 1.005

0.500 0.500 0.499 0.500 0.546 0.748 0.948 0.982 0.991

0.003 0.013 0.033 0.071 0.195 0.275 0.109 0.068 0.049

A TE incident wave produces similar results for scattering from a circular cylinder but has co- and cross-polarization terms: ICRH = ICRTE (a, θ0 , φ )φˆ =

∞ ˆ −j ka cos φ j φe j m εm cos mφ πka sin θ0 m=0 Hm(2) (ka sin θ0 )

(8-78)

Equation (8-78) has the same form as Eq. (8-77) and is expanded in the even function cos mφ as Eq. (8-77) with coefﬁcients using the derivative of the Hankel function. When the incident wave approaches the strut at an angle θ0 other than 90◦ , the strut scatters cross-polarization for a TE incident wave: JCRH = JCRTE zˆ =

j cos θ0 zˆ e−j ka cos φ (ka sin θ0 )2

∞

mj m ej mφ

m=−∞

Hm(2) (ka sin θ0 )

(8-79)

JFRH is an odd function around the perimeter of the strut with a zeroth term of zero. We can expand Eq. (8-79) in terms of sin mφ :

−2j cos θ0 zˆ e−j ka cos φ (I1 sin φ + j 2I2 sin 2φ − 3I3 sin 3φ − · · ·) JCRH = (ka sin θ0 )2 1 Im = (2) (8-80) Hm (ka sin θ0 ) Consider a plane wave incident on a straight strut at an angle θ0 to its axis. As the wave sweeps across the strut it excites current whose phase velocity is c/cos θ0 with respect to the strut axis. This is the same situation as a waveguide with two waves traveling back and forth between the sidewalls that produces a central phase velocity greater than c (Section 5-24). A thin strut with its constant current distribution around the circumference radiates a cone-shaped pattern peaked at an angle determined by the current phase velocity. The current is a fast or leaky wave radiator that radiates in a cone at an angle θ0 from the axis, while the length of the strut in wavelengths determines the narrowness of the radiation beamwidth. As the diameter of the strut increases, the peripheral current distribution alters the radiation level around the cone, but the peak radiation occurs along the cone determined by the incident angle.

DISPLACED-AXIS DUAL REFLECTOR

421

We use incident plane waves to derive strut blockage and scattering and to modify the formulation for spherical wave incidence. First divide the struts into coin sections. For a wave incident from a given point, we trace a ray from the point to the strut axis through a given coin section. We determine the incident magnetic ﬁeld and calculate the surface current density at the point of intersection of this ray and the strut. The intersection point is φ = 0. We apply ICRE , ICRH , and JCRH to calculate the currents around the coin cross section. This near-ﬁeld case does not radiate a strut cone pattern because it is not a plane wave incident on the strut. As the strut diameter grows, this method leads directly to a PO formulation for strut scattering. 8-15 GAIN/NOISE TEMPERATURE OF A DUAL REFLECTOR Collins [41] has developed a procedure for calculating the noise temperature of Cassegrain antennas pointed near the horizon. First, the diffraction pattern of the feed and subreﬂector combination is calculated. Some of the diffraction is added to the main-reﬂector spillover. At low elevation angles the antenna points about one-half of the spillover on the ground. It is a major noise temperature contribution, 12 (1 − SPL)TG , where TG is the ground temperature and SPL is the spillover efﬁciency (ratio). The scattered portion of the blockage produces wide-angle sidelobes, half of which see the ground. The gain is reduced by the spillover loss, and a uniform distribution for the blockage is assumed (ATL = 1): 1 Sb (SPL)TG 2 Sa where Sb is the blocked area and Sa is the total potential aperture. The main beam points toward the sky and collects noise, SPLηb ηm TS , where ηb is the blockage efﬁciency, ηm the ratio of the power in the main beam and the ﬁrst few sidelobes (ηm ≈ 0.99), and Ts the sky temperature. We include a group of minor contributors: 1 (1 − SPL)Ts 2

1 Sb (SPL)Ts 2 Sa

1 SPLηb (1 − ηm )(TG + Ts ) 2

Equation (1-56) can be used when the temperature distribution is known, but the procedure of Collins gives good, although slightly conservative results. Refer to Section 1-15 to calculate the gain noise temperature of the receiving system. 8-16 DISPLACED-AXIS DUAL REFLECTOR A displaced-axis dual reﬂector uses a paraboloidal main reﬂector with a ring focus that transforms the vertex into a ring. GO rays reﬂected from the main paraboloid miss the subreﬂector and reduce the blockage loss to a nonexcitation area instead of scattered blockage. This reﬂector achieves high aperture efﬁciency by using a subreﬂector that directs the higher feed radiation at the boresight to the outer rim of the parabola, where the differential area is the largest. For the moment, consider Figure 8-14 of the Gregorian dual reﬂector in two dimensions. The parabola and ellipse retain their reﬂecting properties because we can extend them out of the page into cylindrical reﬂectors and use a linear array as a feed. Mentally, remove the lower half of the parabola and the upper half of the ellipse. Rays

422

REFLECTOR ANTENNAS

from the left ellipse focus (feed) reﬂect from the remaining half of the lower ellipse to the upper half of the parabola, which transforms them into plane waves. If we ﬁx the right ellipse focus at the focus of the parabola, we can rotate the ellipse axis about the right focus without changing the ray tracing from a feed at the left focus. We use a slightly different portion of the ellipse determined by the rays traced to the edges of the remaining half parabola. Place a horizontal axis at the lower edge of the ellipse and rotate the ellipse axis until the feed focus is on this axis. Rotate the two-dimensional ﬁgure about this horizontal axis to form a three-dimensional reﬂector, and it becomes a displaced-axis dual reﬂector. Both the focus and vertex of the main reﬂector have become rings. The subreﬂector has a matching ring focus at the same diameter as that of the main vertex and a point focus at the feed (Figure 8-16). Now rays from the upper portion of the subreﬂector reﬂect to the upper portion of the parabola. Rays from the center of the subreﬂector terminate on the outer edge of the main reﬂector, while outer subreﬂector edge rays reﬂect to the ring vertex of the main reﬂector [42]. The reﬂector geometry has been found in closed form [43]. Given the main reﬂector diameter D, focal length f , diameter of the subreﬂector Ds , and feed half-subtended angle θ0 , the distance along the reﬂector axis from the vertex to the feed Lm is Lm =

fD Ds cos θ0 + 1 − D − Ds 4 sin θ0

(8-81)

We tilt the axis of the ellipse φ to collapse the ring focus to a point at the feed: tan φ =

2 (cos θ0 + 1)/ sin θ0 − 4f/(D − Ds )

Ring Focus

f

2c Feed

q f

Subreflector

Ring Vertex Ellipse

Parabola

FIGURE 8-16

Displaced-axis reﬂector antenna.

(8-82)

DISPLACED-AXIS DUAL REFLECTOR

423

The parameters of the ellipse are given by the equations Ds c= 4 sin φ

Ds and a = 8

4f cos θ0 + 1 + sin θ0 D − Ds

(8-83)

The half-subtended angle of the main reﬂector ψ0 is found from the normal parabola with the subreﬂector removed and the ring focus collapsed to a point [Eq. (8-2)]: ψ0 = 2 tan−1

D − Ds 4f

We compute the distance between the feed and the subreﬂector along the axis Ls from the geometry [45]: Ds Ls = 2c cos φ + 2 tan ψ0 We determine aperture power distribution A(r ) by tracing rays from the feed to the aperture radius r of the main reﬂector and by equating power in differential areas: P (θ ) sin θ dθ = A(r ) dr

(8-84)

Figure 8-17 gives the aperture distribution for a displaced-axis reﬂector designed for a main reﬂector f/D = 0.27 and an effective feff /D = 1.2 from the feed (θ0 = 23.54◦ ) for various feed edge tapers. The plot shows that increasing the feed edge taper increases the aperture power at larger radiuses but reduces the center amplitude. The center 20% of the diameter of the aperture is not excited, but this corresponds to only 4% lost area, or −0.18 dB. When calculating the amplitude taper loss [Eq. (4-8)], we

Feed Edge Taper

Normalized Aperture Voltage

10-dB 12-dB 14-dB 16 dB 18-dB

Normalized Aperture Radius

FIGURE 8-17 Aperture distribution in a displaced-axis reﬂector given feed edge taper for a particular antenna. (From [43], Fig. 3, 1997 IEEE.)

424

REFLECTOR ANTENNAS

TABLE 8-10 Illumination Losses of a Displaced-Axis Dual Reﬂector, f /D = 0.27, feff /D = 1.2 for Ds = 0.2D and 0.1D Versus a Feed Edge Taper Ds /D (%) 20

10

Edge Taper (dB)

SPL (dB)

ATL (dB)

Total (dB)

Sidelobe

SPL (dB)

ATL (dB)

Total (dB)

Sidelobe

10 11 12 13 14 15 16 17 18 19 20

0.434 0.341 0.268 0.212 0.167 0.132 0.105 0.083 0.066 0.052 0.041

0.476 0.455 0.442 0.434 0.432 0.436 0.444 0.456 0.472 0.491 0.513

0.910 0.796 0.710 0.646 0.600 0.568 0.548 0.539 0.537 0.543 0.555

15.2 14.9 14.5 14.2 14.0 13.7 13.5 13.3 13.1 12.9 12.7

0.434 0.341 0.268 0.212 0.167 0.132 0.105 0.083 0.066 0.052 0.041

0.411 0.377 0.350 0.330 0.316 0.308 0.304 0.306 0.311 0.320 0.333

0.845 0.718 0.619 0.542 0.483 0.440 0.409 0.388 0.377 0.372 0.374

18.3 17.8 17.3 16.8 16.4 16.0 15.6 15.3 14.9 14.6 14.3

use the full radius a = D/2, which accounts for the lost aperture center area. Table 8-10 lists the illumination losses of this reﬂector. The antenna has 88.2% aperture efﬁciency, including blockage loss for Ds /D = 0.2 and 91.8% for Ds /D = 0.1. These numbers do not include diffraction loss due to the subreﬂector and main reﬂector size in wavelengths or strut blockage. Similar to a Cassegrain reﬂector, increasing the subreﬂector diameter beyond the GO design by 1 to 2λ decreases the diffraction loss. Figure 8-17 shows the complete taper of the aperture to zero voltage at the edge. We can increase the aperture efﬁciency slightly by designing the antenna with an effective main reﬂector diameter slightly larger than the real diameter and produce a ﬁnite aperture edge taper at the cost of increased spillover past the main reﬂector. Table 8-11 lists the illumination losses for designs the same as Table 8-10 except that the effective main reﬂector is 2% greater. Four versions of displaced-axis reﬂectors have been derived from Gregorian and Cassegrain antennas [44]. One other case, the double-offset Cassegrain, crosses the feed illumination so that boresight feed amplitude reﬂects to the outer rim of the main reﬂector. This antenna, similar to the case covered above, has a high aperture efﬁciency, whereas the other two cases have modest aperture efﬁciencies. Equations to specify all four antennas are available [45]. The normal displaced-axis dual reﬂector has less sensitivity to feed axial defocusing than does a normal Cassegrain or Gregorian antenna, but it is more sensitive to lateral offset of the feed [46].

8-17

OFFSET-FED DUAL REFLECTOR

When we offset-feed a dual reﬂector, we can eliminate subreﬂector central blockage of the Cassegrain or Gregorian reﬂectors. This design adds parameters to give more convenient packaging that ﬁts in the available space, such as on a spacecraft. More important, by rotating the subreﬂector axis relative to the main reﬂector axis, we can

425

OFFSET-FED DUAL REFLECTOR

TABLE 8-11 Illumination Losses of a Displaced-Axis Dual Reﬂector, f /D = 0.27, feff /D = 1.2 for Ds = 0.2D, and 0.1D Versus a Feed Edge Taper (Effective Main Diameter = 102% Actual) Ds /D (%) 20

10

Edge Taper (dB)

SPL (dB)

ATL (dB)

Total (dB)

Aperture Taper (dB)

SPL (dB)

ATL (dB)

Total (dB)

Aperture Taper (dB)

10 11 12 13 14 15 16 17 18 19 20

0.455 0.352 0.280 0.225 0.181 0.147 0.120 0.099 0.083 0.070 0.060

0.371 0.352 0.340 0.335 0.334 0.339 0.349 0.363 0.380 0.401 0.425

0.816 0.705 0.621 0.559 0.515 0.486 0.469 0.462 0.463 0.471 0.485

10.6 10.4 10.1 9.9 9.6 9.5 9.3 9.1 9.0 8.8 8.7

0.443 0.350 0.278 0.222 0.178 0.144 0.117 0.096 0.079 0.066 0.056

0.304 0.271 0.246 0.227 0.215 0.208 0.206 0.209 0.216 0.227 0.241

0.746 0.621 0.524 0.449 0.393 0.352 0.323 0.305 0.295 0.293 0.297

13.5 12.5 11.5 10.6 10.4 10.2 10.0 9.8 9.7 9.5 9.4

Main Reflector

Subreflector

yf

qe qe a b

2c

2ye yu

Focus

Focus

Feed

Feed ( a)

(b)

FIGURE 8-18 Dual offset-fed Cassegrain reﬂector, including Mizugutch feed axis tilt: (a) feed and subreﬂector geometry; (b) dual reﬂector.

greatly reduce cross-polarization or beam squint of dual circularly polarized feeds in the offset reﬂector. Figure 8-18 illustrates the geometry of an offset-fed Cassegrain reﬂector, and Figure 8-19 shows the offset-fed Gregorian geometry. Refer to Figure 8-7 for the parameters of the offset main reﬂector. We point the feed at the subreﬂector center to reduce spillover and to equalize the amplitude distribution in the aperture of the

426

REFLECTOR ANTENNAS

Main Reflector Focus b a

Feed

yu qe

yf

Subreflector

qe

Focus Feed

(a)

(b)

FIGURE 8-19 Dual offset-fed Gregorian reﬂector, including Mizugutch feed axis tilt: (a) feed and subreﬂector geometry; (b) dual reﬂector.

main reﬂector. Similar to a displaced-axis dual reﬂector, we tilt the axis between the focuses of the subreﬂector relative to the main reﬂector axis β (α in Ticra [48, App. B]). The small amount of tilt to the subreﬂector axis converts the equivalent parabola of the dual reﬂector to an axisymmetric geometry [47]. It takes ﬁve parameters to specify the antenna if the Mizugutch angle requirement is applied to the feed tilt angle α (ψ0 Ticra) relative to the subreﬂector axis determined from the magniﬁcation M given the subreﬂector eccentricity e [Eq. (8-64)]: M tan

β α = tan 2 2

or

M tan

α ψ0 = tan 2 2

† for M =

e+1 e−1

(8-85)

The equation for M in Eq. (8-85) is used for both Cassegrain and Gregorian reﬂectors. We start the design with the diameter of the main reﬂector D because it determines the gain and beamwidth. Ticra [48, App. B] uses main reﬂector focal length f , the half distance between focuses of subreﬂector c, subreﬂector eccentricity e, and axis tilt β. Granet [49,50] supplies equations to calculate the reﬂector dimensions for 17 different sets of ﬁve input parameters. These sets of equations allow the direct application of various mechanical constraints to the design or electrical constraints, such as subreﬂector size to limit diffraction loss. All of Granet’s sets apply the Mizugutch relationship, because this small change should be applied to all designs. By tracing rays through the reﬂectors, the center offset H is found: H = −2f

tan(β/2) − M tan(α/2) 1 + M tan(β/2) tan(α/2)

(8-86)

Given H , D, and f , we compute main reﬂector parameters from Eqs. (8-35) to (8-47). We compute the half subtended angle of feed θe by tracing the ray to the upper rim of

HORN REFLECTOR AND DRAGONIAN DUAL REFLECTOR

427

the main reﬂector with feed angles ψU and α: ψU = −2 tan−1

2H + D 4f

1 ψU − β and θe = 2 tan−1 tan − α M 2

(8-87)

The feed subtended angle of the subreﬂector is 2θe . We calculate the rim ellipse of the subreﬂector determined by the cone axis with angle α and cone angle θe using Eq. (8-59) and (8-60).

8-18 HORN REFLECTOR AND DRAGONIAN DUAL REFLECTOR The horn reﬂector shown in Figure 8-20 consists of a pyramidal or conical input section excited with a rectangular or circular waveguide mode, respectively, that feeds an offset paraboloidal reﬂector. The beam exits horizontally. The horn reﬂector geometry is an offset reﬂector with offset angle ψ0 = 90◦ and center offset H = 2f the same as the periscope conﬁguration. Figure 8-21 gives the pattern of a 3-m-diameter reﬂector operating at 6 GHz (diameter = 60λ) with f = 3.215 m (ψe = 15◦ ). The antenna radiates cross-polarization −23 dB relative to the beam peak in the horizontal plane. It cannot be used for two channels with different polarizations, because similar to all offset-fed reﬂectors, circularly polarized beams squint right and left in this horizontal plane. The antenna radiates a signiﬁcant sidelobe 90◦ from the boresight in the horizontal plane that can be controlled using serrated-edge blinders [51]. A Dragonian dual reﬂector uses a hyperbola subreﬂector that curves toward the main reﬂector in a Cassegrain system. This produces a dual reﬂector with magniﬁcation

b d bc

FIGURE 8-20 Horn reﬂector with serrated side blinders. (From [51], Fig. 3, 1973 IEEE.)

428

REFLECTOR ANTENNAS

Amplitude, dB

Symmetrical Plane Cross-Polarization

Pattern Angle (degrees)

FIGURE 8-21 Pattern of a 3-m horn reﬂector at 6 GHz.

M < 1 and we need a feed antenna with a wider beamwidth than would be required for efﬁcient feed of the main reﬂector. We use long focal lengths for the main reﬂector, which ﬂattens its curve, so that the feed beamwidth can remain small. Jones and Kelleher [52] applied this Cassegrain arrangement to a horn reﬂector and located the feed horn in the middle of the paraboloidal main reﬂector. Dragone [53] derived the generalized Mizugutch criterion for multiple reﬂectors and showed that it could be applied to this Cassegrain system to eliminate cross-polarization. Figure 8-22 shows a Dragonian dual reﬂector fed by a corrugated horn feed located beyond the rim of the paraboloidal reﬂector to replace the 3-m horn reﬂector. Pattern analysis produces the same curves as in Figure 8-21 except that cross-polarization in the horizontal plane is eliminated. The design given in Figure 8-22 has f = 9.8 m, D = 3 m, θe = 20◦ , and tilts the subreﬂector axis by −73◦ to place the main reﬂector and subreﬂector in different quadrants. The Mizugutch criterion between the subreﬂector axis tilt and the feed tilt locates the feed axis at −97.5◦ relative to the main reﬂector axis and −24.5◦ relative to the subreﬂector axis. The parameters of the reﬂector were adjusted so that the plane wave radiated from the main reﬂector misses both the feed corrugated horn and the subreﬂector. All dimensions can be found using available equation sets [54] that require various sets of ﬁve inputs to totally specify the dual reﬂector. By using these equations, we discover that the hyperboloidal subreﬂector rim is an ellipse with 2.05- and 2.60-m diameters. To use the hyperbola close to the feed, specify a negative eccentricity and the equations curve the hyperbola toward the main reﬂector. Similarly, the equation for magniﬁcation produces a value of less than 1 for a negative eccentricity. Given the e = −1.832 for the reﬂector of Figure 8-22, M=

e+1 −1.832 + 1 = = 0.2938 e−1 −1.832 − 1

SPHERICAL REFLECTOR

429

Feed

Main Reflector Beam

Subreflector

FIGURE 8-22 Dragonian dual-reﬂector geometry.

We compute feed tilt relative to the subreﬂector axis to satisfy the Mizugutch criterion: β −73 ◦ α = 2 tan−1 M tan = 2 tan−1 0.2938 tan = −24.5 2 2 8-19 SPHERICAL REFLECTOR When we feed-scan a paraboloidal reﬂector, the pattern sidelobes develop coma and the beam shape generally degrades. Feed scanning is limited. In a spherical reﬂector a feed moved in an arc from the center of the sphere and sees the same reﬂector geometry if we discount the edge effects. Greater scanning is possible, but the spherical reﬂector fails to focus an incident plane wave to a point and requires more elaborate feeds. We can design many types of feeds for the spherical reﬂector. The reﬂector can be fed from a point source for large f /D by assuming that it is a distorted parabola [55,56]. It can be fed with a line source to follow the axis ﬁelds. Corrector subreﬂectors can be designed to correct the spherical aberrations [58]. Like the parabolic reﬂector, we can design arrays [24] to compensate for spherical aberrations and give multiple beams. Figure 8-23 shows the geometry and ray tracing of a spherical reﬂector illuminated by a plane wave. All rays intersect a radial line of the sphere (the axis) in the direction of the incident wave because the reﬂector has circular symmetry about all axes. The diagram traces rays hitting the outer portion of the reﬂector as passing through the axis closer to the vertex than do the rays reﬂected from areas closer to the axis. The reﬂector has a line focus. A distorted paraboloidal reﬂector with a line focus exhibits spherical aberration because the focal length depends on the radial distance from the axis of the reﬂection point. The spherical reﬂector has a cusplike caustic where GO

430

REFLECTOR ANTENNAS

Paraxial Focus

y y

R/2

z

H

R/2 z

y R/2

Center

Caustic

FIGURE 8-23 Ray tracing in a spherical reﬂector.

predicts inﬁnite ﬁelds. The second side of Figure 8-23 traces a single ray. We can easily solve the isosceles triangle for the results: R/2 z=

1 − H 2 /R 2 R2 H 2 = R2 1 − 2 4z

(8-88) (8-89)

where z is the location of the focus for a given ray. As H approaches zero, with rays near the axis, the reﬂected ray passes through the paraxial focus (z = R/2). We use Eq. (8-89) to ﬁnd the power distribution on the axis by using the conservation of power. The power in a differential area of the plane wave reﬂects into a differential length on the axis: dA = 2πH dH . We differentiate Eq. (8-89) implicitly: 2H dH =

R2 dz 2z3

The power distribution along the axis is Pz =

P0 R 3 8z3

(8-90)

where P0 is the power at the paraxial focus. The peak power occurs at the paraxial focus and drops by one-eighth (−9 dB) at the vertex. We determine the required length of the line source feed from the rotation angle ψ of the illuminated portion of the reﬂector: R(1/ cos ψ − 1) feed length = (8-91) 2

SPHERICAL REFLECTOR

431

Example If the half-rotation angle of the illuminated region is 30◦ , the feed length is 0.0774R from Eq. (8-91). The amplitude decreases by [Eq. (8-90)]

R3 Pz = = 0.65 (−1.9 dB) P0 8(R/2 + 0.0774R)3

The rays intersecting the axis are not at constant phase. The path length from the aperture plane through the reﬂector origin is path length =

R2 +z 2z

(8-92)

We can approximate Eq. (8-92) by a linear function if the feed length is short. Example The feed length is 0.0774R long; calculate the phase change required along the feed. The feed starts at the paraxial focus (z = R/2): path length =

R2 R + = 1.5R R 2

At z = R/2 + 0.0774R, the path length = 1.443R. The phase change is (2π/λ)R(0.0566). If we plot the phase change over the region of the feed, we can approximate the phase change by a linear function very accurately. The spherical reﬂector can be fed from a point source when the f /D is large [55]. The center of the reﬂector approximates a parabola. The optimum focal point is f =

!

1 R + R 2 − (D/2)2 4

(8-93)

The maximum phase path length error is [56] L 1 D 1 = λ 2048 λ (f/D)3

(8-94)

The approximate gain loss is L 2 G = 3.5092 G λ or

L PEL(dB) = 10 log 1 − 3.5092 λ

2 (8-95)

A path length deviation of 0.25λ reduces the gain by 1.08 dB. Example Determine the f /D value of a spherical reﬂector to limit L to 1/16λ for a reﬂector diameter of 50λ.

432

REFLECTOR ANTENNAS

By rearranging Eq. (8-94), we ﬁnd that f 16(50) = = 0.73 D 2048 8-20

SHAPED REFLECTORS

Shaped reﬂectors spread cylindrical or spherical waves into a desired pattern that depends on geometric optics. Shaped reﬂectors do not radiate patterns exactly as prescribed by GO. In all cases we must apply techniques such as aperture diffraction, induced currents, or geometric theory of diffraction (GTD) to compute the actual pattern. We consider only the ﬁrst-order GO for design, although analysis requires more elaborate techniques. We use two principles to design shaped reﬂectors. The ﬁrst is GO reﬂection expressed as a differential equation. The second is the conservation of power in ray tubes, which can be expressed either in terms of differential areas or integrals over sections of the feed and reﬂection patterns. We deﬁne two angles for the GO reﬂection equation. The feed points toward the reﬂector, and we measure its pattern angle ψ from an axis pointed toward the reﬂector. The reﬂector reradiates the incident feed pattern in a far-ﬁeld pattern whose angle θ is measured from the axis pointing away from the reﬂector. The differential equation of reﬂection is [2] θ +ψ dρ tan = (8-96) 2 ρdψ where ρ is the distance of the reﬂector from the feed. The edges of the reﬂector are deﬁned by angles ψ1 and ψ2 measured from the feed axis and reﬂect in directions θ1 and θ2 . We integrate this differential equation for a solution: ρ(ψ) = ln ρ0 (ψ1 )

ψ

tan ψ1

θ (ψ) + ψ dψ 2

(8-97)

where ψ1 is some initial angle of the feed and ρ(ψ1 ) is the initial radius vector locating the reﬂector at ψ1 . GO, the zero-wavelength approximation, is consistent at any size. All parabolic reﬂectors collimate spherical waves radiated from the focus regardless of size. Only by considering diffraction or currents induced on the reﬂector can we compute gain and beamwidth of the antenna. Example

From Eq. (8-97), determine the reﬂector surface to give θ (ψ) = 0. ln

ρ(ψ) = ρ0 (ψ1 )

ψ

tan ψ1

ψ ψ1 ψ dψ = −2 ln cos − ln cos 2 2 2

By the properties of the ln function, this becomes ln

ρ(ψ) cos2 (ψ1 /2) = ln ρ0 (ψ1 ) cos2 (ψ/2)

433

SHAPED REFLECTORS

By taking the exponential of each side, we get the polar equation of the reﬂector: ρ(ψ) = ρ0 (ψ1 )

cos2 (ψ1 /2) cos2 (ψ/2)

We let ψ1 = 0 and set ρ(ψ) = f , and we obtain the equation for the parabola [Eq. (8-1)]. The differential equation of reﬂection tells us only the shape of the reﬂector locally to produce a reﬂection in a direction θ for an incident angle ψ. We must still ﬁnd the power density in various directions. The ratio of differential areas gives us these power densities. Given the pattern of the feed Gf (ψ, φ) and the pattern of the reﬂection P (θ, φ), KP(θ, φ) dA(θ, φ) = Gf (ψ, φ) dAf (ψ, φ) (8-98) where dA(θ, φ) is the differential area of the reﬂection pattern, dAf (ψ, φ) the differential area of the feed pattern, and K a constant found by equating the total incident and reﬂected powers. Equation (8-98) is based on the assumption that reﬂections are 1 : 1 with the feed pattern. 8-20.1 Cylindrical Reﬂector Synthesis We feed cylindrical reﬂectors with line sources. The reﬂector determines the beam shape in one plane and the line-source distribution in the other. The problem reduces to designing a single two-dimensional curve moved along an axis to deﬁne the reﬂector. The power radiated by the feed is given by Gf (ψ) dψ. The reﬂector directs this power at an angle θ whose power density is P (θ ) dθ . These are proportional [Eq. (8-98)]: KP (θ ) dθ = Gf (ψ) dψ

(8-99)

At the limits of the reﬂector, feed angles ψ1 and ψ2 reﬂect to angles θ1 and θ2 . We calculate the constant K by equating the power in each pattern:

ψ2

ψ1

K=

Gf (ψ) dψ

θ2 θ1

(8-100) P (θ ) dθ

We integrate Eq. (8-99) to derive a formal solution usually arrived at numerically. By combining Eqs. (8-99) and (8-100), we eliminate K:

θ

θ1 θ2 θ1

ψ

=

ψ1 ψ2

P (θ ) dθ P (θ ) dθ

ψ1

Gf (ψ) dψ (8-101) Gf (ψ) dψ

We use Eq. (8-101) with a known feed pattern Gf (ψ) and a desired pattern function P (θ ) to establish the relation θ (ψ). We insert θ (ψ) into the differential equation for

434

REFLECTOR ANTENNAS

the reﬂection [Eq. (8-97)] and determine radial distance as a function of ψ: ρ(ψ) = ρ0 (ψ1 ) exp

ψ

tan ψ1

θ (ψ) + ψ dψ 2

(8-102)

With Eq. (8-102) we calculate the reﬂector coordinates to within a scale factor ρ0 (ψ1 ). 8-20.2 Circularly Symmetrical Reﬂector Synthesis The synthesis of circularly symmetrical reﬂectors is readily reduced to a twodimensional problem. We must assume that the feed pattern is also axisymmetrical. We describe the feed pattern by Gf (ψ) and the reﬂector pattern by P (θ ). The differential areas are sin ψ dψ and sin θ dθ . Equation (8-98) becomes KP (θ ) sin θ dθ = Gf (ψ) sin ψ dψ

(8-103)

We integrate Eq. (8-103) to ﬁnd the function θ (ψ):

θ

θ1 θ2 θ1

P (θ ) sin θ dθ = P (θ ) sin θ dθ

ψ

ψ1 ψ2 ψ1

Gf (ψ) sin ψ dψ (8-104) Gf (ψ) sin ψ dψ

We use Eq. (8-102) from the reﬂection differential equation with θ (ψ) to determine the polar equation of the reﬂector. The design of a shaped reﬂector can best be explained with an example. The cylindrical reﬂector synthesis follows parallel steps. Example Design a reﬂector to transform the feed pattern of Figure 8-24a into the pattern of Figure 8-24b. The required pattern drops by about 9 dB from 50◦ to 0◦ . We will use the feed pattern from 4◦ to 54◦ and design a reﬂector with a ring caustic. The feed pattern at 4◦ is reﬂected to 50◦ , and the feed pattern at 54◦ is reﬂected to 0◦ . The geometric optics rays cross somewhere in front of the reﬂector. We have the following limits: Feed: ψ1 = 4◦ , ψ2 = 54◦ Reﬂection: θ1 = 50◦ , θ2 = 0 We insert the patterns of Figure 8-24 into both sides of Eq. (8-104) and evaluate the ratio of the integrals. Table 8-12 gives the results of these integrals for θ and ψ as normalized in Eq. (8-104). Given ψ, we ﬁnd θ by equating integrals in the table. For example, follow the line from ψ = 28◦ (feed) to its integral value, match the values of the integrals, and determine the reﬂection angle θ = 42◦ . We trace a number of these through interpolation to generate Table 8-13 of reﬂection angles θ (ψ) for given feed angles. We use Table 8-13 of θ (ψ) in the integral of Eq. (8-102) to calculate the normalized polar equation of the reﬂector listed in Table 8-14. Figure 8-25 shows the reﬂector shape as well as the geometry of axisymmetric-shaped reﬂectors. There is a hole in the center because the procedure fails to specify that region. Note in Table 8-12 how

SHAPED REFLECTORS

435

( a)

(b)

FIGURE 8-24 Axisymmetrical reﬂector pattern transformation: (a) feed pattern; (b) reﬂector pattern.

436

REFLECTOR ANTENNAS

TABLE 8-13 Shaped Reﬂector Synthesis Reﬂection Angles for Given Feed Angles, θ (ψ) Feed Angle, ψ (deg)

Reﬂection Angle, θ (deg)

Feed Angle, ψ (deg)

Reﬂection Angle, θ (deg)

4 6 8 10 12 14 16 18 20 22 24 26 28

50.0 50.0 49.9 49.8 49.6 49.3 48.8 48.2 47.4 46.4 45.2 43.7 42.1

30 32 34 36 38 40 42 44 46 48 50 52 54

40.2 38.2 36.0 33.6 31.1 28.4 25.6 22.6 19.5 16.2 12.6 8.4 0.0

TABLE 8-14 Shaped Reﬂector Synthesis Normalized Polar Equation of a Reﬂector, ρ(ψ) Feed Angle, ψ (deg)

Normalized Radius, ρ

Feed Angle, ψ (deg)

Normalized Radius, ρ

Feed Angle, ψ (deg)

Normalized Radius, ρ

4 6 8 10 12 14 16 18 20

1.000 1.018 1.038 1.058 1.080 1.103 1.128 1.153 1.180

22 24 26 28 30 32 34 36 38

1.208 1.238 1.268 1.299 1.332 1.365 1.398 1.433 1.468

40 42 44 46 48 50 52 54

1.503 1.539 1.575 1.611 1.648 1.684 1.719 1.753

much of the inner portion of the reﬂector must reﬂect rays near 50◦ to achieve the high gain required. We could repeat the example and design without a ring caustic where rays from the feed near 54◦ reﬂect to 50◦ and would produce a ﬂatter reﬂector. Because diffraction effects spread the pattern, we could approximate the pattern of Figure 8-24b by designing a reﬂector to point the beam in a cone about the axis. If we take Eq. (8-97) and let θ (ψ) = θ0 , a constant, we get the surface ρ(ψ) = ρ0

cos2 [(ψ0 + θ0 )/2] cos2 [(ψ + θ0 )/2]

(8-105)

Example Estimate the directivity of a 40λ-diameter reﬂector shaped by Eq. (8-105) to scan in a cone to θ0 = 50◦ .

SHAPED REFLECTORS

437

q y

r(y)

y Feed

y1 q1

Reflector

y2 q2

FIGURE 8-25 Circularly symmetrical reﬂector designed with a caustic reﬂector.

Only half of the diameter is used for each side. The effective scanned aperture width becomes (40λ/2) cos 50◦ = 12.8λ. If we assume a uniform-amplitude aperture distribution, we obtain an upper bound. From Eq. (4-83), HPBW = 59◦ /12.8 = 4.6◦ . We use Eq. (1-24) to estimate the directivity: directivity =

2 = 32.5 (15 dB) cos(50 − 2.3 ) − cos(50◦ + 2.3◦ ) ◦

◦

The boresight gain of the aperture with a uniform distribution is 42 dB. Spreading the reﬂection into a cone greatly reduces gain. The shaped reﬂector above will have even less directivity because it has a greater edge taper. 8-20.3 Doubly Curved Reﬂector for Shaped Beams It is a common radar requirement to have a narrow beam in one plane and a shaped beam in the other. Such beams can be obtained from shaped cylindrical reﬂectors, but it is simpler to replace the line source with a single feed. We only specify the pattern in the principal planes denoted: θV , the shaped pattern coordinate, and θH , the pencil

438

REFLECTOR ANTENNAS

beam [2,59]. Similarly, we specify the feed antenna pattern in terms of ψV and ψH . For a given feed angle ψV , the reﬂected wave angle is θV . The only θH value allowed is zero. The reﬂector collimates the wave in the horizontal plane. This collimation requires a symmetrical reﬂector made from parabolic curves in the horizontal plane. We design the vertical curve only through the center of the reﬂector. For a given feed angle ψV (Figure 8-26a), all incoming rays at an angle θV must be reﬂected into the feed. The incoming rays form the x –z plane in Figure 8-26, and the reﬂector collimates these to the feed by a parabola in the plane. We call this parabola a rib of the reﬂector. Figure 8-26b shows the plane and two rays reﬂecting into the feed from a wave arriving at an angle θV in the x –z plane. For the beam to focus, the optical path lengths must be equal: BP + P O = AN + N O

(8-106)

Equation (8-106) establishes the curve of the rib in the x –z plane as a parabola with focal length θV (ψV ) + ψV f = ρc (ψV ) cos2 (8-107) 2 with the focus located on the z -axis. Using the parabolic ribs reduces the problem to the design of the central curve ρc (ψV ).

y

Central Curve z qV

yV z

Boresight

(a) y

Rib

Central Curve

Incoming Wave, x − z′ Plane z′

qV

P

qV r(yV)

N yV

Focus O

x

z

(b)

FIGURE 8-26

Doubly curved shaped reﬂector.

SHAPED REFLECTORS

439

The reﬂected and feed power densities modify Eq. (8-98) to KP (θV )dθV ρc (ψV ) dψH = Gf (ψV ) dψV dψH

(8-108)

We integrate Eq. (8-108) and normalize to the total power:

θV

θ1 θ2 θ1

P (θV ) dθV = P (θV ) dθV

ψV

ψ1 ψ2 ψ1

[Gf (ψV )/ρc (ψV )] dψV (8-109) [Gf (ψV )/ρc (ψV )] dψV

Equation (8-109) is similar to Eqs. (8-101) and (8-104) except that the feed pattern integral value depends on the radial distance to the central rib. We must know ρc (ψV ) before we can determine θV (ψV ), which will be required to compute ρc (ψV ) from the reﬂection differential equation [Eq. (8-102)]. The solution can be found only by an iterative process. We must assume a ρc (ψV ), solve for θV (ψV ), and use the result to compute a new ρc (ψV ). After a few iterations, the values of ρc (ψV ) converge. We use the normalized ρc with the foregoing ratio of integrals. We start with a parabola: cos2 (ψ1 /2) ρc (ψV ) = ρc (ψ1 ) cos2 (ψV /2) The surface generated by following the method may not be deﬁned uniquely. We pick a constant width for the reﬂector in the horizontal plane. We deﬁne the surface with a continuous series of parabolas each in a x –z plane determined by the reﬂection angle θV , which changes direction along the central rib. We must plot the curve of the vertical coordinate of the edge versus ψV to see if it is monotonic. If there are loops in the curve, the surface deﬁned is not unique. Given the width x, we calculate the vertical coordinate of the edge by the following development. The location of the rib on the central curve is given by ρc (ψV ) sin ψV . The rib is a parabola in the x –z plane with its focus given by Eq. (8-107). The z coordinate at the edge is z = x 2 /4f (ψV ). We determine the vertical dimension by projecting this point onto the y-axis: y = ρc (ψV ) sin ψV . Elliott [60, p. 500] points out that by following this method, one does not get the proper slope for reﬂection at all points, but we will get the desired pattern when we design for only small deviations from a pencil beam. The surface can be designed with or without a caustic depending on the reﬂection angles at the edges. Reﬂectors designed with caustic edge reﬂections have a better chance of being unique [60]. Carberry [61] presents a method of analysis that involves physical optics. When we apply these methods, we must subdivide the reﬂector into many patches because the phases of the currents change rapidly with position on the reﬂector, and the analysis must be repeated with ﬁner and ﬁner patches until the result converges. 8-20.4 Dual Shaped Reﬂectors We can design a dual-reﬂector antenna to produce an arbitrary phase and amplitude in the aperture plane by shaping both reﬂectors. By using both the conservation of power

440

REFLECTOR ANTENNAS

and the differential equations of reﬂection on the two surfaces, Galindo [62] derived a pair of differential equations in terms of the aperture radius. Runge–Kutta or any other suitable numerical method can be used to solve the simultaneous differential equations instead of an integration of the power equation. Williams [63] ﬁnds a solution to Cassegrain antennas within the restriction of equal amplitude and phase in the aperture plane by integration of the power equation. Collins [64] considers using a parabolic reﬂector for the main reﬂector, since the difference between the shaped main reﬂector and a parabola is small. He accepts a quadratic phase error in the aperture. Existing large reﬂectors can be retroﬁtted with a shaped subreﬂector to improve performance. For the method to work, an axisymmetric feed such as a corrugated horn is required. Galindo-Israel and Mittra [65] use a pair of reﬂectors offset from each other to transform a spherical wave from a feed antenna into a second spherical wave with a modiﬁed pattern amplitude. This combination of a feed with two reﬂectors can illuminate either prime focus paraboloidal reﬂectors or Cassegrain systems without modiﬁcation of the existing reﬂector surfaces. For example, a sec4 (θ/2) pattern can be realized from an ordinary pattern source to increase the aperture efﬁciency of the overall reﬂector system. The reﬂectors maintain equal GO path lengths for all rays, but they only approximate the desired pattern amplitude from the virtual focus. The procedure can be used to determine the contours of the reﬂectors along radial lines through numerical solution of differential equations. The equations develop from simplifying assumptions that depend on the extra degree of freedom introduced by the second reﬂector. In most cases, solution of the equations produces usable designs, although the method is not exact. Lee et al. [66] developed a method to shape offset-fed dual reﬂectors that reduces to the solution of a differential equation similar to that of the single-reﬂector design given above. The reﬂection properties of the subreﬂector determine the main reﬂector amplitude distribution to ﬁrst order. This method does not produce exact results but is close enough for engineering purposes. We start with a desired aperture power distribution P (r, φc ) and a known feed power pattern Gf (θ , φc ) given in the radial direction φc . Most cases use distributions independent of φc , but the design is performed along these planes. For a circularly symmetric design we only need to solve the differential equation along one plane, but the general case requires solutions along enough planes to allow splines along the coordinate φc to ﬁnd every point on both reﬂectors. A differential expression relates the feed power to the aperture power: Gf (θ , φc ) sin θ dθ = P (r, φc )r dr This leads to a ratio of integrals:

θ

−θe θe −θ e

Gf (θ ) sin θ dθ

Gf (θ ) sin θ dθ

"R R

P (r )r dr

R1

P (r )r dr

= " R12

(8-110)

Equation (8-110) covers the general case where the offset subreﬂector directs power from a lower angle −θe to an offset radius R1 that changes for each plane φc . For a circularly symmetric design, −θe = 0 and R1 = 0. Although many designs attempt to generate a uniform aperture distribution for the main reﬂector, we can substitute

SHAPED REFLECTORS

441

any distribution, such as a circular Taylor distribution to control the sidelobes into Eq. (8-110). Given the aperture distribution and the feed pattern, we calculate a table similar to Table 8-12 for each plane φc that gives the feed angle as a function of aperture radius. We interpolate on this table to determine every value. We start at the center of the subreﬂector described in spherical coordinates (ρ0 , 0, 0) relative to the axis of the subreﬂector centered at the feed focus. The subreﬂector axis may be tilted relative to the main reﬂector axis. The rectangular coordinates of the subreﬂector are (ρ sin θ cos φc , ρ sin θ sin φc , ρ cos θ ). The incident wave reﬂects to a point on the main reﬂector: (H ± R cos φc , ±R sin φc , z) using + Cassegrain, −Gregorian. We calculate the unit vector between the subreﬂector point and the main reﬂector. The normal vector on the subreﬂector is expressed as a differential: 1 1 ∂ρ 1 ∂ρ n= aθ − aφ (8-111) aρ − ρ ∂θ ρ sin θ ∂φc c

where =

1+

1 ∂ρ ρ ∂θ

2 +

1 ∂ρ ρ sin θ ∂φc

2

We apply both equations of Snell’s law [Eq. (2-67)] to the subreﬂector reﬂection and gather terms to form a pair of differential equations: QV ∂ρ = 2 ∂θ Q + U2

and

∂ρ U V sin θ = 2 ∂φc Q + U2

(8-112)

The terms of Eq. (8-112) are given by the expressions a cos θ cos φc + b sin θ sin φc − c sin θ ρ b cos φc − a sin φc U= ρ

Q=

V = L + a sin θ cos φc + b sin θ sin φc + c sin θ

(8-113)

a = H ± R cos φc − ρ sin θ cos φc b = ±R sin φc − ρ sin θ sin φc c = z − ρ cos θ where the vector (a, b, c) is from the subreﬂector to the main reﬂector and L = √ a 2 + b2 + c2 . We choose z = 0 as the aperture and equate path lengths along every ray. This gives an equation for the z-position of the main reﬂector: OL = ρ0 + L0 − z0 = ρ + L − z We solve for z: z=

a 2 + b2

1 + (ρ cos θ − ρ + OL) 2(ρ cos θ − ρ + OL) 2

(8-114)

442

REFLECTOR ANTENNAS

To solve for the reﬂector surfaces, we choose a starting point, usually the center of the reﬂector as the ﬁrst ray from the feed to the subreﬂector, and calculate the initial distance L0 between the subreﬂector and the main reﬂector to ﬁnd the path length. We select a polar plane φc and solve the left differential equation [Eq. (8-112)] for both surfaces using a Runge–Kutta numerical solution. We repeat this in a sufﬁcient number of planes φc to specify the surface totally. If the antenna is circularly symmetric, we solve the equation only once. For an offset dual reﬂector we can improve the cross-polarization by computing an equivalent subreﬂector using least squares and use its eccentricity to calculate the Mizugutch subreﬂector axis rotation.

8-21 OPTIMIZATION SYNTHESIS OF SHAPED AND MULTIPLE-BEAM REFLECTORS Silver [2] discusses using a linear array feed to shape the beam of a paraboloidal reﬂector. The method is quite empirical and involves the addition of a number of offset beams. A similar technique is used in three-dimensional radar, but the feeds are kept separate so that multiple beams can scan a larger area in a given time. An array feed provides the best solution to beam shaping in many cases. The number of elements in the array limits the number of variables to a ﬁnite set to which optimization techniques can be applied. A second method uses optimization to shape the reﬂector and possible subreﬂectors. This requires distortion functions on the reﬂectors. We start with conic-section reﬂectors and add distortions. These distortions can be global Zernike functions deﬁned over the total surface, or they could be localized functions such as B-splines [67]. A B-spline uses a grid of points on the reﬂector, but the spline coefﬁcients apply only over a limited area. In both cases we obtain a set of coefﬁcients used in the optimization algorithms. We have the choice of combining these coefﬁcients, or we can iterate between different sets of coefﬁcients. Optimization is an art. Because the reﬂector is an aperture antenna, we pick a set of directions in (u, v) = (sin θ cos φ, sin θ sin φ) space to evaluate the pattern. The number of points should exceed the number of coefﬁcients and be spaced close enough to fully describe the main-beam pattern: 0.5λ 0.25λ u and v ∼ to (8-115) D D We calculate the pattern power Pm (u, v) at these points and compare them to the desired pattern Pmd (u,v) using a suitable cost function. We weight each pattern direction ωm and use a summation cost [68] with a gradient minimization technique: F (x) =

M

|ωm (Pm (um , vm )) − Pmd (um , vm )|2

(8-116)

m=1

A second choice is a min–max optimization [69]. This algorithm minimizes the maximum error: max[ωm (Pm (um , vm )) − Pmd (um , vm )] (8-117)

REFERENCES

443

If we optimize the reﬂector shape, we express the distortion as B-splines speciﬁed at evenly spaced points across the aperture with the number determined by the maximum pattern angle θmax and the reﬂector diameter D [69]: Nx = Ny =

πD sin θmax +2 λ

(8-118)

Given a Zernike polynomial expansion with maximum azimuthal mode expansion Mmax and maximum polar mode index Nmax , we have similar mode number requirements: Mmax = Nmax =

πD sin θmax +2 λ

(8-119)

Shaping starts with a paraboloid main reﬂector whose beamwidth may be so narrow that a portion of the speciﬁed u–v space area may lie in the sidelobe region. In this case the optimization may become trapped because it cannot satisfy this area when changes effecting the main beam region positively affect the sidelobe region negatively. We must distort the main reﬂector before starting the optimization [69]. First surround the u–v space area of speciﬁed points with an ellipse centered at (u0 , v0 ) with major radius ω1 and minor radius ω2 tilted an angle α. Given a paraboloid with diameter D, focal length f , and center offset (x0 , y0 ), we deﬁne rotated coordinates on the aperture. x = (x − x0 ) cos α + (y − y0 ) sin α y = −(x − x0 ) sin α + (y − y0 ) cos α Using these coordinates, we alter the z-axis position of the reﬂector:

1 x2 + y2 + z = − 2 8f 2

ω1 x 2 + ω2 y 2 + u0 (x − x0 ) + v0 (y − y0 ) D

(8-120)

We have a choice with ω1 and ω2 because they can be both either positive or negative. Positive values ﬂatten the reﬂector while negative values cause a caustic reﬂection to broaden the beam. REFERENCES 1. A. W. Love, Reﬂector Antennas, IEEE Press, New York, 1978. 2. S. Silver, ed., Microwave Antenna Theory and Design, McGraw-Hill, New York, 1949. 3. W. V. T. Rusch and P. D. Potter, Analysis of Reﬂector Antennas, Academic Press, New York, 1970. 4. P. J. Wood, Reﬂector Antenna Analysis and Design, Peter Peregrinus, London, 1980. 5. C. A. Mentzer and L. Peters, A GTD analysis of the far-out sidelobes of Cassegrain antennas, IEEE Transactions on Antennas and Propagation, vol. AP-23, no. 5, September 1975, pp. 702–709. 6. S. W. Lee et al., Diffraction by an arbitrary subreﬂector: GTD solution, IEEE Transactions on Antennas and Propagation, vol. AP-27, no. 3, May 1979, pp. 305–316. 7. A. D. Craig and P. D. Simms, Fast integration techniques for reﬂector antenna pattern analysis, Electronics Letters, vol. 18, no. 2, January 21, 1982, pp. 60–62.

444

REFLECTOR ANTENNAS

8. V. Galindo-Israel and R. Mittra, A new series representation for the radiation integral with application to reﬂector antennas, IEEE Transactions on Antennas and Propagation, vol. AP25, no. 5, September 1977, pp. 631–641. 9. A. W. Rudge et al., eds., The Handbook of Antenna Design, Vol. 1, Peter Peregrinus, London, 1982. 10. A. C. Ludwig, The deﬁnition of cross polarization, IEEE Transactions on Antennas and Propagation, vol. AP-21, no. 1, January 1973, pp. 116–119. 11. J. R. Cogdell and J. H. Davis, Astigmatism in reﬂector antennas, IEEE Transactions on Antennas and Propagation, vol. AP-21, no. 4, July 1973, pp. 565–567. 12. Y. T. Lo, On the beam deviation factor of a parabolic reﬂector, IEEE Transactions on Antennas and Propagation, vol. AP-8, no. 3, May 1960, pp. 347–349. 13. J. Ruze, Lateral-feed displacement in a paraboloid, IEEE Transactions on Antennas and Propagation, vol. AP-13, no. 5, September 1965, pp. 660–665. 14. W. V. T. Rusch and A. C. Ludwig, Determination of the maximum scan-gain contours of a beam scanned paraboloid and their relation to the Petzval surface, IEEE Transactions on Antennas and Propagation, vol. AP-21, no. 2, March 1973, pp. 141–147. 15. W. A. Imbriale, P. G. Ingerson, and W. C. Wong, Large lateral feed displacements in a parabolic reﬂector, IEEE Transactions on Antennas and Propagation, vol. AP-22, no. 6, November 1974, pp. 742–745. 16. D. K. Cheng, Effect of arbitrary phase errors on the gain and beamwidth characteristics of radiation pattern, IEEE Transactions on Antennas and Propagation, vol. AP-3, no. 4, July 1955, pp. 145–147. 17. J. Ruze, Antenna tolerance theory: a review, Proceedings of IRE, vol. 54, no. 4, April 1966, pp. 633–640. 18. M. S. Zarghamee, On antenna tolerance theory, IEEE Transactions on Antennas and Propagation, vol. AP-15, no. 6, November 1967, pp. 777–781. 19. R. C. Hansen, ed., Microwave Scanning Antennas, Academic Press, New York, 1964. 20. W. V. T. Rusch and R. D. Wanselow, Boresight gain loss and gore related sidelobes of an umbrella reﬂector, IEEE Transactions on Antennas and Propagation, vol. AP-30, no. 1, January 1982, pp. 153–157. 21. W. H. Watson, The ﬁeld distribution in the focal plane of a paraboloidal reﬂector, IEEE Transactions on Antennas and Propagation, vol. AP-12, no. 5, September 1964, pp. 561–569. 22. T. B. Vu, Optimization of efﬁciency of reﬂector antennas: approximate method, Proceedings of IEE, vol. 117, January 1970, pp. 30–34. 23. B. M. Thomas, Theoretical performance of prime focus paraboloids using cylindrical hybrid modes, Proceedings of IEE, vol. 118, November 1971, pp. 1539–1549. 24. N. Amitay and H. Zucker, Compensation of spherical reﬂector aberrations by planar array feeds, IEEE Transactions on Antennas and Propagation, vol. AP-20, no. 1, January 1972, pp. 49–56. 25. V. Galindo-Israel, S. W. Lee, and R. Mittra, Synthesis of laterally displaced cluster feed for a reﬂector antenna with application to multiple beams and contoured patterns, IEEE Transactions on Antennas and Propagation, vol. AP-26, no. 2, March 1978, pp. 220–228. 26. B. Popovich et al., Synthesis of an aberration corrected feed array for spherical reﬂector antennas, IEEE/APS Symposium Digest, May 1983. 27. V. Mrstik, Effect of phase and amplitude quantization errors on hybrid phased-array reﬂector antennas, IEEE Transactions on Antennas and Propagation, vol. AP-30, no. 6, November 1982, pp. 1233–1236. 28. C. M. Knop, On the front to back ratio of a parabolic dish antenna, IEEE Transactions on Antennas and Propagation, vol. AP-24, no. 1, January 1976, pp. 109–111.

REFERENCES

445

29. M. Uhm, A. Shishlov, and K. Park, Offset-paraboloid geometry: relations for practical use, IEEE Antennas and Propagation Magazine, vol. 38, no. 3, June 1996, pp. 77–79. 30. R. F. H. Yang, Illuminating curved passive reﬂector with defocused parabolic antenna, 1958 IRE Wescon Convention Record, August 1958, pp. 260–265. 31. C. Granet, Designing axially symmetric Cassegrain and Gregorian dual-reﬂector antennas from combinations of prescribed geometric parameters, IEEE Antennas and Propagation Magazine, vol. 40, no. 2, April 1998, pp. 76–82. 32. C. Granet, Designing axially symmetric Cassegrain and Gregorian dual-reﬂector antennas from combinations of prescribed geometric parameters, part 2: minimum blockage condition while taking into account the phase-center of the feed, IEEE Antennas and Propagation Magazine, vol. 40, no. 3, June 1998, pp. 82–85. 33. W. V. T. Rusch, Phase error and associated cross polarization effects in Cassegrainian-fed microwave antennas, IEEE Transactions on Antennas and Propagation, vol. AP-14, no. 3, May 1966, pp. 266–275. 34. P.-S. Kildal, The effects of subreﬂector diffraction on the aperture efﬁciency of a conventional Cassegrain antenna: an analytical approach, IEEE Transactions on Antennas and Propagation, vol. AP-31, no. 6, November 1983, pp. 903–909. 35. A. M. Isber, Obtaining beam-pointing accuracy with Cassegrain antennas, Microwaves, August 1967, pp. 40–44. 36. W. V. T. Rusch and R. Wohlleben, Surface tolerance loss for dual-reﬂector antennas, IEEE Transactions on Antennas and Propagation, vol. AP-30, no. 4, July 1982, pp. 784–785. 37. A. F. Kay, Electrical design of metal space frame radomes, IEEE Transactions on Antennas and Propagation, vol. AP-13, no. 2, March 1965, pp. 188–202. 38. P.-S. Kildal, E. Olsen, and J. A. Aas, Losses, sidelobes, and cross polarization caused by feed-support struts in reﬂector antennas: design curves, IEEE Transactions on Antennas and Propagation, vol. AP-36, no. 2, February 1988, pp. 182–190. 39. W. V. T. Rusch et al., Forward scattering from square cylinders in the resonance region with application to aperture blockage, IEEE Transactions on Antennas and Propagation, vol. AP-24, no. 2, March 1976, pp. 182–189. 40. G. T. Ruck, ed., Radar Cross Section Handbook, Vol. 1, Plenum Press, New York, 1970. 41. G. W. Collins, Noise temperature calculations from feed system characteristics, Microwave Journal, vol. 12, December 1969, pp. 67–69. 42. B. E. Kinber, On two-reﬂector antennas, Radioengineering and Electronics, vol. 7, no. 6, 1962, pp. 973–980. 43. A. P. Popov and T. A. Milligan, Amplitude aperture-distribution control in displaced-axis two reﬂector antennas, IEEE Antennas and Propagation Magazine, vol. 39, no. 6, December 1997, pp. 58–63. 44. S. P. Morgan, Some examples of generalized Cassegrainian and Gregorian antennas, IEEE Transactions on Antennas and Propagation, vol. AP-12, no. 6, November 1964, pp. 685–691. 45. C. Granet, A simple procedure for the design of classical displaced-axis dual-reﬂector antennas using a set of geometric parameters, IEEE Antennas and Propagation Magazine, vol. 41, no. 6, December 1999, pp. 64–72. 46. T. A. Milligan, The effects of feed movement on the displaced-axis dual reﬂector, IEEE Antennas and Propagation Magazine, vol. 40, no. 3, June 1998, pp. 86–87. 47. Y. Mizugutch, M. Akagawa, and H. Yokoi, Offset dual reﬂector antenna, IEEE Symposium on Antennas and Propagation Digest, 1976, pp. 2–5. 48. P. H. Nielson and S. B. Sørensen, Grasp8 Software Users Manual, Ticra, Copenhagen, 2001.

446

REFLECTOR ANTENNAS

49. C. Granet, Designing classical offset Cassegrain or Gregorian dual-reﬂector antennas from combinations of prescribed geometric parameters, IEEE Antennas and Propagation Magazine, vol. 44, no. 3, June 2002, pp. 114–123. 50. C. Granet, Designing classical offset Cassegrain or Gregorian dual-reﬂector antennas from combinations of prescribed geometric parameters, part 2: feed-horn blockage conditions, IEEE Antennas and Propagation Magazine, vol. 45, no. 6 December 2003, pp. 86–89. 51. D. T. Thomas, Design of multiple-edge blinders for large horn reﬂector antennas, IEEE Transactions on Antennas and Propagation, vol. AP-21, no. 2, March 1973, pp. 153–158. 52. S. R. Jones and K. S. Kelleher, A new low noise, high gain antenna, IEEE International Convention Record, March 1963, pp. 11–17. 53. C. Dragone, Offset multireﬂector antennas with perfect pattern symmetry and polarization discrimination, Bell System Technical Journal, vol. 57, no. 7, September 1978, pp. 2663–2684. 54. C. Granet, Designing classical dragonian offset dual-reﬂector antennas from combinations of prescribed geometric parameters, IEEE Antennas and Propagation Magazine, vol. 43, no. 6, December 2001, pp. 100–107. 55. T. Li, A study of spherical reﬂectors as wide-angle scanning antennas, IEEE Transactions on Antennas and Propagation, vol. AP-7, no. 4, July 1959, p. 223–226. 56. R. Woo, A multiple-beam spherical reﬂector antenna, JPL Quarterly Technical Review, vol. 1, no. 3, October 1971, pp. 88–96. 57. A. W. Love, Spherical reﬂecting antennas with corrected line sources, IEEE Transactions on Antennas and Propagation, vol. AP-10, no. 5, September 1962, pp. 529–537. 58. F. S. Bolt and E. L. Bouche, A Gregorian corrector for spherical reﬂectors, IEEE Transactions on Antennas and Propagation, vol. AP-12, no. 1, January 1964, pp. 44–47. 59. A. S. Dunbar, Calculation of doubly curved reﬂectors for shaped beams, Proceedings of IRE, vol. 36, no. 10, October 1948, pp. 1289–1296. 60. R. S. Elliott, Antenna Theory and Design, Prentice-Hall, Englewood Cliffs, NJ, 1981. 61. T. F. Carberry, Analysis theory for the shaped beam doubly curved reﬂector antenna, IEEE Transactions on Antennas and Propagation, vol. AP-17, no. 2, March 1969, pp. 131–138. 62. V. Galindo, Design of dual reﬂector antennas with arbitrary phase and amplitude distributions, IEEE Transactions on Antennas and Propagation, vol. AP-12, no. 4, July 1964. pp. 403–408. 63. W. F. Williams, High efﬁciency antenna reﬂector, Microwave Journal, vol. 8, July 1965, pp. 79–82. 64. C. Collins, Shaping of subreﬂectors in Cassegrainian antennas, IEEE Transactions on Antennas and Propagation, vol. AP-21, no. 3, May 1973, pp. 309–313. 65. V. Galindo-Israel and R. Mittra, Synthesis of offset dual shaped subreﬂector antennas for control of Cassegrain aperture distributions, IEEE Transactions on Antennas and Propagation, vol. AP-32, no. 1, January 1984, pp. 86–92. 66. J. J. Lee, L. I. Parad, and R. S. Chu, A shaped offset-fed dual-reﬂector antenna, IEEE Transactions on Antennas and Propagation, vol. AP-27, no. 2, March 1979, pp. 165–171. 67. M. E. Mortenson Geometric Modeling, Wiley, New York, 1985. 68. C. C. Han and Y. Hwang, Satellite antennas, Chapter 21 in Y. T. Lo and S. W. Lee, eds., Antenna Handbook, Van Nostrand Reinhold, New York, 1993. 69. H.-H. Viskum, S. B. Sørensen, and M. Lumholt, User’s Manual for POS4, Ticra, Copenhagen, 2003.

9 LENS ANTENNAS

In lenses, as in parabolic reﬂectors, we utilize free space as a feed network to excite a large aperture. Because we locate the feed behind the aperture, the conﬁguration eliminates aperture blockage and allows direct connection of the feed to the transmitter or receiver. When frequencies are above microwaves, this feeding method removes lossy transmission lines that increase system noise. Lenses have only half the tolerance requirements of reﬂectors because the wave passes by the anomaly only once. In a reﬂector the wave path deviates by twice the distance as the wave travels to and from the reﬂector. At low microwave frequencies the lens is prohibitively heavy, but zoning and the use of artiﬁcial dielectrics reduce this problem. Both zoning and artiﬁcial dielectrics present mechanical stability problems and narrow the bandwidth. We organize the design of lenses by the available degrees of freedom. A single lens with a uniform dielectric has two surfaces and is equivalent to a dual reﬂector because each surface is a degree of freedom. We start our discussion with single-surface lenses where we eliminate one degree of freedom by making the second surface match either the incoming or outgoing wave. Shaping both surfaces lets us correct one lens anomaly. We can remove either coma to improve the feed scanning or design to convert a given feed pattern to a desired aperture distribution. Bootlace lenses have three possible degrees of freedom. They consist of back-to-back arrays with cables connecting the sides. Normally, we use the degrees of freedom of the bootlace lens to increase the number of focal points. We give up degrees of freedom in many designs to simplify the mechanical layout. Finally, we discuss the use of a variable index of refraction in the Luneburg lens. We base the design of lenses on geometric optics. Like parabolic reﬂectors, lenses have no inherent frequency bandwidth limitation. We are limited by the feeds and mechanical problems of large sizes. Because we borrow from optics, lenses have great high-frequency potential. Modern Antenna Design, Second Edition, By Thomas A. Milligan Copyright 2005 John Wiley & Sons, Inc.

447

448

LENS ANTENNAS

9-1 SINGLE REFRACTING SURFACE LENSES Single-surface lenses convert the wave type, such as spherical to planar, at one surface through refraction. The constant-phase surface (eikonal) of the wave type determines the shape of the second surface of the lens. The common lens converts an incident spherical or cylindrical wave to a plane wave. Conversion to a cylindrical wave requires a line source feed and cylindrical surfaces on the lens. Spherical waves use point feeds and axisymmetrical surfaces. Like the reﬂector, which also converts spherical waves from the feed to plane waves by geometric optics (GO), diffraction from an aperture determines the far-ﬁeld pattern. Consider the second or nonrefracting surface. If the surface toward the feed converts the wave type, the wave exits the second surface as a plane wave and it is a plane. Similarly, when the surface away from the feed converts the exiting wave to a plane wave, the inner surface toward the feed follows the incident wave eikonal, a cylindrical or spherical surface. Figure 9-1 shows the two types of single refracting surface lenses. We can determine the refracting surface shape by either of two different approaches. Snell’s law can be applied to the refracting surface, and the surface slope can be determined for each feed angle. Equivalently, we can apply Fermat’s principle to equalize the optical path length from the feed through the lens to an aperture plane. The designs are easily found [1,2]. For Figure 9-1a, ρ(ψ) =

(n − 1)f n cos ψ − 1

(9-1)

where n is the index of refraction, given by n=

√ εr µ r

(9-2)

and εr and µr are the relative permittivity and permeability of the lens medium. When n > 1, Eq. (9-1) describes a hyperbola with the feed at one focus. The distance from the feed to the hyperbola along the axis is f . The asymptotes of the hyperbola limit the collimated portion of the feed radiation: ψa = cos−1

1 n

(9-3)

We must limit the lens edge to angles less than ψa because the asymptotes imply an inﬁnite aperture size. Similar to the paraboloidal reﬂector, we have feed spillover, considered to be lost in sidelobes. But, for example, placing the lens at the aperture of a horn eliminates spillover. We calculate the lens diameter from D = 2ρ sin ψe =

2(n − 1)f sin ψe n cos ψe − 1

(9-4)

where ψe is the edge angle, subject to the restriction of Eq. (9-3). The surfaces of the lens in Figure 9-lb have the polar equations ρ1 = constant

ρ2 (ψ) =

(n − 1)f n − cos ψ

(9-5)

SINGLE REFRACTING SURFACE LENSES

449

r(y)

Feed

Focus Hyperbolical

Plane

(a)

r2(y) Feed y r1

Focus

Focus Spherical (Circular)

Elliptical

(b)

FIGURE 9-1 Single-surface lenses.

The inner surface must either be a circular cylinder (cylindrical lens) or be spherical (axisymmetrical lens). The outer surface ρ2 (ψ) is elliptical for n > 1. The junction of the circle and ellipse determines the feed angle limitation: cos ψe = n −

(n − 1)f ρ1

(9-6)

We can, of course, truncate the lens before the two curves meet. Equation (9-6) gives the limitation on ρ1 at the lens edge as well: ρ1 ≤

(n − 1)f n − cos ψe

(9-7)

Example Compute f and ρ1 at the edge for an elliptical lens (Figure 9-lb) with D = 10λ, n = 1.6 (polystyrene), and ψe = 50◦ . Solve Eq. (9-5) for f /D: f n − cos ψe = D 2 sin ψe (n − 1) D ρ1 ≤ 2 sin ψe

f = 10.41λ

(9-8) (9-9)

450

LENS ANTENNAS

If ρ1 remained constant to the center of the lens, it would be 3.88λ thick. In narrowband applications we can remove the multiple wavelength thicknesses by zoning and reduce weight and dielectric material loss. Lenses change the amplitude distribution of the feed in the aperture plane. We relate the feed pattern to the aperture distribution through the conservation of power in differential areas. For an axisymmetrical lens: F (ψ, φ) sin ψ dψ dφ feed power

=

A(r, φ)r dr dφ aperture power

(9-10)

where ψ is the feed angle and r, ρ sin ψ, is the aperture radial distance; F (ψ, φ) is the feed power pattern and A(r, φ) is the aperture power distribution: A(r, φ) sin ψ dψ = F (ψ, φ) r dr

(9-11)

For a cylindrical lens, we also equate differential area multiplied by the feed or aperture power: F (ψ, y) dψ dy = A(r, y) dr dy dψ A(r, y) = F (ψ, φ) dr

(9-12)

We substitute Eqs. (9-11) and (9-12) into Eq. (9-1) for the hyperbolical lens to calculate the aperture distribution relative to the feed power pattern. Axisymmetrical A(r, φ) (n cos ψ − 1)3 = 2 F (ψ, φ) f (n − 1)2 (n − cos ψ)

Cylindrical (n cos ψ − 1)2 A(r, y) = F (ψ, y) f (n − 1)(n − cos ψ)

(9-13a, b)

The ﬁeld distribution is the square root of Eq. (9-13). We substitute Eqs. (9-11) and (9-12) into Eq. (9-5) for the elliptical lens: Axisymmetrical (n − cos ψ)3 A(r, φ) = 2 F (ψ, φ) f (n − 1)(n cos ψ − 1)

Cylindrical (n − cos ψ)2 A(r, y) = F (ψ, φ) f (n − 1)(n cos ψ − 1) (9-14a, b) The hyperbolical and elliptical lenses concentrate the aperture power in different ways. The hyperbolical lens reduces the feed power directed toward the edges and produces an additional aperture taper. On the other hand, an elliptical single-surface lens increases the power toward the edges as compared with the center. Example For axisymmetrical lenses with ψe = 50◦ and n = 1.6, compute the edge taper due to the lenses.

451

ZONED LENSES

We divide Eq. (9-13b) with ψe = ψ by the same equation with ψ = 0 to determine the ratio of power at the edge to that at the center of the aperture (assuming an isotropic feed). We do the same calculation with Eq. (9-14a):

Ae = Ac

(n cos ψe − 1)3 (n − 1)2 (n − cos ψe ) (n − cos ψe )3 (n − 1)2 (n cos ψe − 1)

hyperbolical lens

(9-15)

elliptical lens

(9-16)

By substituting ψe and n, we compute the edge taper. hyperbolical lens: 0.038 (−14.2 dB)

elliptical lens : 7.14 (8.5 dB)

The increased taper of the hyperbolical lens reduces sidelobes, and the elliptical lens increases aperture efﬁciency by compensating for some of the feed antenna pattern taper to make the aperture distribution more uniform.

9-2 ZONED LENSES Lenses designed by the methods of Section 9-1 have bandwidth limitations determined only by the invariability of the dielectric constant. Zoning removes multiples-ofwavelength path lengths in the lens to reduce weight, to reduce the lens-induced amplitude taper, or to thin the lens. The act of changing dimensions by wavelengths implies narrowing the frequency bandwidth. We step the lens in either the nonrefracting or refracting surface. Stepping the nonrefracting surface (Figure 9-2a, b) has the least effect. The edges of the steps, parallel with the waves, will diffract waves and cause some change in the aperture ﬁelds, but GO predicts no effect. Stepping the refracting surface introduces losses either as misdirected feed power (Figure 9-2c, d) or as unexcited aperture (Figure 92e, f). But stepping the refracting surface reduces the lens-induced aperture taper by changing the focal lengths in various zones. Figure 9-2 shows the limits in the two types of refracting surface steps, since we could compromise between the directions and have both feed spillover and unexcited aperture. We can easily calculate the step dimensions in Figure 9.2a and b. We equate the path lengths inside and outside the dielectric along the step with a difference of or some integer multiple of λ: n = +λ inside outside The step becomes =

λ n−1

(9-17)

In Figure 9-2c–f we determine the change in focal lengths instead of the step dimensions. Zoning affects the optical path lengths in the center of the lens. We calculate the edge focal length from the unzoned case. The focal length increases by Eq. (9-17)

452

LENS ANTENNAS

l/(n − 1)

(a)

l/(n − 1)

( b)

Feed Dead Zones

(c)

(d )

Aperture Dead Zones

(e)

(f )

FIGURE 9-2 Zoning of single-surface lenses.

for each inner step of hyperbolical lenses and decreases by the same amount for each inner step of elliptical lenses. We derive the lens-induced taper relative to the center by using ratios of Eq. (9-13) or (9-14). For axisymmetrical lenses, fc2 (n cos ψ − 1)3 (9-18) hyperbolical A(r, φ) f 2 (n − 1)2 (n − cos ψ) = Ac f 2 (n − cos ψ)3 2 c 2 (9-19) elliptical f (n − 1) (n cos ψ − 1) where fc is the focal length in the center, f the focal length in the feed direction ψ, and Ac the central amplitude. Example Design an axisymmetrical hyperbolical lens (n = 1.6) by using the three types of zoning shown in Figure 9-2, using an aperture diameter of 30λ and a maximum feed angle of 35◦ with a 70◦ 10-dB beamwidth feed.

ZONED LENSES

453

The minimum allowable thickness is 0.5λ with a 0.3λ edge thickness. By working through the geometry, we obtain dimensions for the following cases: Figure 9-2a, nonrefracting surface zoning, we compute steps = λ/(n − 1) = 1.67λ. 1

2

3

4

12.418

10.009

7.403

4.256

Step Aperture Radius (λ)

We estimate the feed spillover loss from Eq. (8-12) for the 10-dB feed edge taper, 0.41 dB. Equation (9-15) gives the edge taper (9.72 dB), since the refracting surface is not zoned. Combined with the feed edge taper of 10 dB, we have a 19.72-dB amplitude taper in the aperture plane. We use Eq. (4-8) to calculate the amplitude taper loss (1.8 dB) for this axisymmetric distribution. Figure 9-2c, zoning in the refracting surface of the hyperbolical lens (parallel with outgoing waves), we compute the dimensions starting with the edge focal length found from a rearrangement of Eq. (9-4) (Table 9-1). The changing focal lengths in the zones alter the aperture amplitude distribution. The edge taper becomes [Eq. (918)] 6.24 dB and reduces the amplitude taper loss [Eq. (4-8)] to 1.19 dB. The portion of the feed directed to the dead zones refracts out of the aperture and forms sidelobes that reduce the aperture efﬁciency. We consider this as a second spillover loss (0.81 dB). Figure 9-2e, zoning in the refracting surface of the hyperbolical lens (parallel with the feed wave rays), again we start with the edge focal length and increase it by λ/(n − 1) at each step and determine the feed angles where the change in the focal length will give the minimum allowable thickness. The dimensions given in Table 9-2 were obtained. Since the focal lengths are the same as for Fig. 9-2c, the lens-induced edge taper is 6.24 dB. The dead zones in the aperture distribution increase the amplitude taper loss to 3.10 dB. These ring dead zones can be considered as radiating and adding to the fully excited aperture pattern. They radiate patterns with closely spaced lobes that raise the near sidelobes of the total antenna. The three designs are compared in Table 9-3. Example Similar to the example above, we can design zoned elliptical axisymmetrical lenses that have the same problems of feed angle or aperture dead zones. The edge taper of the elliptical lens counteracts some of the feed taper and reduces amplitude taper loss. The losses calculated for those designs are given in Table 9-4.

TABLE 9-1 Zoned Hyperbolical Lens, Figure 9-2c

Zone

Focal Length (λ)

Aperture Radius of Step (λ)

Thickness (λ)

Feed Dead Zone Angles (deg)

1 2 3 4 5

20.21 18.54 16.87 15.21 13.54

0 5.12 8.42 10.84 12.89

1.52 2.09 1.98 1.90 1.83

13.57–14.02 21.64–23.10 27.06–28.68 31.28–32.95

454

LENS ANTENNAS

TABLE 9-2

Zoned Hyperbolical Lens, Figure 9-2e

Zone

Focal Length (λ)

Feed Angle (deg)

Thickness (λ)

Aperture Dead Zones (λ)

1 2 3 4 5

20.21 18.54 16.87 15.21 13.54

0 13.57 21.64 27.06 31.28

1.51 2.25 2.41 2.60 2.83

0 4.70–5.12 7.66–8.42 9.77–10.84 11.48–12.89

TABLE 9-3 Aperture Illumination Losses of Three Hyperbolical Lenses, Figure 9-2 Design (Figure 9-2)

SPL (dB)

ATL (dB)

Sum (dB)

(a) (c) (e)

0.41 1.22 0.41

1.80 1.19 3.10

2.21 2.41 3.51

TABLE 9-4 Aperture Illumination Losses of Three Elliptical Lenses, Figure 9-2 Design (Figure 9-2)

SPL (dB)

ATL (dB)

Sum (dB)

(b) (d) (f)

0.41 1.43 0.41

0.06 0.14 1.43

0.47 1.57 1.84

Zoning reduces the frequency bandwidth. At the center frequency the optical path length difference between the central ray and edge ray is K − 1 for K zones. A common maximum allowed deviation of feed to aperture path is 0.125λ, which leads to a bandwidth of 25% B (9-20) K −1 For the ﬁve zone lenses of the examples above, Eq. (9-20) gives a 6% bandwidth. We determine the loss at band edge by tracing rays from the feeds-to-aperture plane and calculating the phase error loss using Eq. (4-9). The loss at band edge is about 0.3 dB for all the designs. Bandwidth is greatly underestimated by Eq. (9-20) if a greater phase error loss is allowed. The 1-dB phase error loss bandwidth is 45%/(K − 1). 9-3 GENERAL TWO-SURFACE LENSES Optical lens designs use either ﬂat or spherical surfaces, an approximation useful for long focal lengths. We design lenses exactly. In Section 9-1 we discussed lenses where the rays refracted at only one surface. The shape curve of these two lenses could be

GENERAL TWO-SURFACE LENSES

455

derived easily. In this section we pick the curve for one lens surface and use a numerical search for the second surface that generates a table of (r, z) values. We generate a cubic spline from the table and use it to calculate the surface normal vector and the radius of curvature useful in manufacture. The initial numerical search generates ray paths through the lens that produce a table of feed angle versus aperture radius. We apply Eq. (9-11) to compute the aperture distribution after we generate the cubic spline whose evaluation includes the derivative. If we specify the surface on the feed side, we start at the feed and trace rays to the inner surface. We know the inner surface normal vector because it is a speciﬁed surface. We compute the direction of the ray inside the lens given the incident medium index of refraction ni , no inside the lens, the surface unit normal n directed into the lens, and the incident ray unit vector Si . First determine if the ray will exit the incident ray medium, because if ni > no , it can act as a prism and have total reﬂection: Ra = n2o − n2i [1 − (Si · n)2 ] If Ra < 1, the ray is totally reﬂected. For Ra > 1, we determine the direction of the output ray unit vector So from the following operations [3, p. 355]: γ =

Ra − ni (n · Si )

ﬁnd the vector t = ni Si + γ n

So =

t |t|

(9-21)

For our case the normal vector, incident, and refracted rays are in two-dimensional space (z, r), because the lens has axisymmetry. To start, specify the z-axis distance from the feed to the outer rim of the lens, the initial radius, and the rim thickness along the initial internal ray direction. The feed-side position of the lens is (z1 , r1 ). Apply Eq. (9-21) and ﬁnd the outer lens point given the rim thickness tr , by tracing along the vector t to point (z2 , r2 ). Since the lens collimates the beam, we trace the ray to a plane z = z3 whose normal is the z-axis. Figure 9-3 illustrates the rim and internal ray paths for a lens with a spherical inner surface. We calculate the electrical path length from the feed located at z = zf to the output plane by including the index

(z1, r1)

(z2, r2)

Focus

Spherical Surface

FIGURE 9-3

Ray tracing in a single-surface lens with a speciﬁed spherical inner surface.

456

LENS ANTENNAS

of refraction n of the lens for the ray path length (PL) through the lens, and the input and output ray lengths: PL =

r12 + (z1 − zf )2 + n (r2 − r1 )2 + (z2 − z1 )2 + (z3 − z2 )

(9-22)

The design consists of stepping in r1 , tracing the ray to the inner lens surface, computing the direction of the ray internal by Eq. (9-21), and determining thickness tr when Eq. (9-22) minus the initial path length is zero. By Fermat’s principle of equal path length, the outer surface refraction direction adds to the inner surface refraction to produce parallel output rays. This procedure generates a table of (z2 , r2 ) pairs that we convert to a cubic spline. By using the cubic spline, we produce an evenly spaced table of (z2 , r2 ) for machining, and if necessary, a table of radius of curvature from the second derivative to assist machining operations. The next step is to calculate a cubic spline between the aperture radius r2 and the feed angle ψ because its output includes dψ/dr2 . By rearranging Eq. (9-11), we compute the aperture power distribution given the feed power pattern: A(r2 , φ) =

F (ψ, φ) sin ψ dψ r2 dr2

(9-23)

The design steps for a lens with the outer surface speciﬁed are similar except that we trace rays from the output plane backward to the feed point. Again, we start at the lens rim, use Eq. (9-21) to calculate the internal ray direction by using the −z-directed ray and the known surface normal to travel along this ray by the rim thickness to the inner surface. Equation (9-22) gives the electrical path length for this lens, as well. We repeat the root-ﬁnding procedure used above to determine a series of points (z1 , r1 ) along the surface by equating all electrical path lengths. We generate the same series of cubic splines to obtain machining dimensions and differential dψ/dr2 used in Eq. (9-23) for the aperture power distribution. Figure 9-4 shows the aperture distributions for various lenses with focus points located 1.5 times the radius below the lens. The curves include lenses of Section 9-1. The focal spot is not a singularity as drawn on the ﬁgures using geometric optics, but spreads due to the ﬁnite wavelength. We use Gaussian beams to evaluate the size of focal spots. For a lens with a collimated output, we assume a Gaussian beam on the output with minimum waist diameter 2W0 = D, the lens diameter, radiating into free space, and a matching Gaussian beam on the feed side which tapers to the focal plane. The lens transforms one Gaussian beam into another. The focal length f = zf and we determine the diameter of the focal spot 2W0 and the half depth of focus b from the lens F -number F# = f /D [4, pp. 91–95]: 2W0 =

4λ 4λ f F# = π π D

b=

8λ 2 F π #

(9-24)

A small modiﬁcation to Eq. (9-22) allows the design of lenses with a second focus at a ﬁnite position z = z3 on the axis: PL =

r12 + (z1 − zf )2 + n (r2 − r1 )2 + (z2 − z1 )2 + r22 + (z3 − z2 )2

(9-25)

457

Aperture Distribution, dB

GENERAL TWO-SURFACE LENSES

Normalized Radius

FIGURE 9-4 Aperture distribution for various single-surface lenses with f/D = 1.5.

Focus

Focus

FIGURE 9-5 Ray tracing in lens designed for two ﬁnite focuses.

We follow the same steps as in the design above except that we need to generate a table of feed angles given the output ray angle with respect to the z-axis. Figure 9-5 illustrates a typical design and shows the ray tracing. The lens designs noted above narrow the beamwidth. We design lenses to spread the beam by using a virtual focus located behind the output surface of the lens, as shown in Figure 9-6. Because the rays trace backward to the virtual focus, we change the sign of the last term in Eq. (9-25): 2 2 2 2 PL = r1 + (z1 − zf ) + n (r2 − r1 ) + (z2 − z1 ) − r22 + (z3 − z2 )2

(9-26)

The iterative design procedure should be modiﬁed to start in the center of the lens because the concave lens has a signiﬁcant outer thickness. We compute focal length f of the lens by the distance of the two focuses from the lens: 1 1 1 + = f zf z3

(9-27)

458

LENS ANTENNAS

Virtual Focus

Focus

Focus

(a)

Virtual Focus

(b)

FIGURE 9-6 Lenses designed with virtual focuses to widen beam: (a) spherical outer surfaces; (b) spherical inner surfaces.

To calculate the pattern we trace rays through the lens either to a planar surface output side of the lens or the actual surface. By using a cubic spline between the feed angle and the aperture position we calculate the amplitude due to spreading [Eq. (9-23)]. We replace the ﬁelds with currents and use physical optics to calculate the pattern. A second simpler approach uses a Gaussian beam approximation for lenses that accounts for electrical size. Both the input and output Gaussian beam have the same waist at the lens plane. Each Gaussian beam decreases in a hyperbola to the minimum waist at the location of the phase center or focus. The output does not pass through a narrow caustic, as shown in Figure 9-5, but reaches a ﬁnite-diameter waist related to the feed beamwidth and the lens diameter. At this point we design the lenses scaled to wavelengths. Given the feed beamwidth, we calculate the half depth of focus b by using Eq. (7-35). The lens magniﬁes the output beam waist compared to the input beam waist by M. First, we calculate an input magniﬁcation factor Mr : f (9-28) Mr = zf − f With b given in wavelengths, we compute the ratio of b to the shift of the feed relative to the focus: b r= (9-29) zf − f We combine Eqs. (9-28) and (9-29) to calculate the lens magniﬁcation M: M=√

Mr 1 + r2

(9-30)

SINGLE-SURFACE OR CONTACT LENSES

459

The output Gaussian beam propagates from the second focus with new parameters: b , the half depth of focus, waist W0 , and divergence angle θ0 [4]: b = M 2 b

W0 = MW0

θ0 =

θ0 M

(9-31)

Example The lens in Figure 9-5 was designed for a diameter of 20λ, zf = 15λ, and z3 = 25λ. We calculate the focal length from Eq. (9-27) as f = 9.375λ. A feed with a 68◦ 10-dB beamwidth is approximated by a Gaussian beam with b = 0.99λ (Scale 7-8) and 2W0 = 1.12λ (Scale 7-9). We use b to ﬁnd the magniﬁcation M by using Eqs. (9-28) to (9-30), M = 1.64. The Gaussian beam at the second focus has a minimum waist 2W0 = 1.64(1.12) = 1.84λ. Using Scales 7-7 to 7-9, we read the Gaussian beam output values: 10-dB beamwidth = 42◦ (gain = 18.4 dB). Scale 7-7 gives the gain of the feed as 14.3 dB. The half depth of focus increased from 0.99λ to 3.39λ. Example When we repeat the lens calculations for the lens of Figure 9-6a for a diameter = 20λ, zf = 15λ, and z3 = −5λ, we calculate f = −7.5. Starting with a feed 10-dB beamwidth = 80◦ , we use Scale 7-8 to ﬁnd b = 0.7λ and waist diameter 2W0 = 0.94λ. The negative focal length produces a magniﬁcation below 1: M = 0.333. This decreases b to 0.0775λ and 2W0 = 0.314. The output Gaussian beam gain drops to 6.6 dB from a feed gain of 12.9 dB.

9-4 SINGLE-SURFACE OR CONTACT LENSES We can alter the pattern of antennas with planar surfaces such as spirals and microstrip patches by placing a lens directly on the planar surface. The lens can be spaced a small distance away to avoid potential damage and have little pattern impact. The lens modiﬁes the original pattern of the antenna by using the refraction at the single output surface. Because the lens contacts the antenna, the pattern inside the lens is the same as radiated into free space except for the dielectric loading on the antenna. This loading shifts the operating frequency and in the case of a spiral improves its efﬁciency and widens its beamwidth (see Section 11-5.1). We design these antennas by ﬁrst generating a mapping between the feed angle ψ and the output angle θ . This could be as simple as θ (ψ) = constant or a function to generate a shaped beam, in the same manner as a shaped reﬂector (see Section 8-20). The relationship θ (ψ) enables calculation of the surface normal at every point along the outer surface. The surface normal is found from the gradient of the radial vector, and equating the two values produces a differential equation between the radius r and the feed angle ψ. Given the feed angle and the output angle at a point on the lens, we calculate the surface normal from n = nSi − So , where n is the index of refraction, Si the incident unit vector, and So the exiting ray unit vector. We normalize n to a vector v. The gradient of r(ψ) gives a second expression for the normal vector: ∇r(ψ) = ar + aψ

1 ∂r(ψ) r(ψ) ∂ψ

(9-32)

The coefﬁcient of aψ in Eq. (9-32) is the tangent of the angle α between the normal vector and the radial vector (incident ray). The tangent can be found from the unit

460

LENS ANTENNAS

vector v and the incident ray unit vector Si : tan α =

1 ∂r(ψ) (Si × v) = Si · v r(ψ) ∂ψ

(9-33)

The cross product between the incident ray and the unit vector v only has a z-axis component, because the incident ray and the surface normal lie in the x –y plane. We design the lens surface by solving the differential equation (9-33). A numerical technique such as the Runge–Kutta method easily solves the equation when we start at one feed angle and an arbitrary lens radius and step through feed angles. The method determines only the shape of the lens to an arbitrary size that we scale to the diameter desired. Figure 11-9a illustrates a lens designed to redirect all feed rays to θ = 0. Figure 9-7 illustrates the shape of a contact designed to redirect all rays between feed angles 0 through 60◦ to output rays at 30◦ . For a 3λ lens diameter and a feed 12-dB beamwidth of 120◦ , the lens spreads the beam to form a ﬂat-topped output beam shown in Figure 9-8. Contact lenses can greatly modify radiation from a feed with electrically small lenses. Exit Rays

Incident Rays

Feed

FIGURE 9-7 Contact lens designed to direct the beam in a cone at 30◦ .

FIGURE 9-8 Pattern of contact 3λ-diameter lens for a feed with a 12-dB beamwidth of 120◦ .

METAL PLATE LENSES

461

9-5 METAL PLATE LENSES The phase velocity of a wave in waveguide exceeds that of a wave in free space and produces a medium with an effective refractive index of less than 1. We can make a microwave lens by spacing parallel metal plates and feeding the lens with a wave polarized in the direction of the plates. For plates spaced a distance a, the index of refraction is 2 λ (9-34) n= 1− 2a where λ is the wavelength in the medium between the plates. The index of refraction is frequency dependent. The lens can be made polarization independent by forming an egg crate of orthogonal plates. We divide an arbitrary polarization into orthogonal polarizations normal to each set of plates. If we substitute n from Eq. (9-34) into Eq. (9-1), we obtain the equation of a front single-surface lens: (1 − n)f (9-35) ρ(ψ) = 1 − n cos ψ Equation (9-35) is an ellipse with f as the distance from the far focus of the ellipse to the center of the lens front surface. This surface refracts waves parallel with the axis and determines the second surface: a plane. The parallel plates constrain the waves parallel with the axis and prevent the design of an outer single-surface lens. The cutoff wavelength 2a and the possibility of higher-order modes restrain the range of n. The second-order-mode cutoff occurs when λ = a, and it limits n to 0.866 [Eq. (9-34)]. At cutoff, λ = 2a and n equals zero. Reasonable values lie between 0.3 and 0.7. The variation of n versus frequency limits bandwidth. When the phase variation in the aperture is limited to λ/8, the bandwidth is approximately [1]: bandwidth(%) =

λ 25n 1 − n (1 − n)t

(9-36)

where n is the center-frequency index of refraction and t is the maximum thickness. An acceptable bandwidth is underestimated by Eq. (9-36) through restriction of the band edge phase error. The elliptical surface increases the aperture distribution toward the edges. When we substitute Eq. (9-35) into Eqs. (9-11) and (9-12), we obtain the aperture amplitude distribution relative to the feed pattern: (1 − n cos ψ)3 A(r, φ) = 2 F (ψ, φ) f (1 − n)2 (cos ψ − n)

axisymmetrical

A(r, y) (1 − n cos ψ)2 = F (ψ, y) f (1 − n)(cos ψ − n)

cylindrical

(9-37a, b)

Example Design an axisymmetric parallel-plate lens with a diameter of 30λ, maximum feed angle of 35◦ , n = 0.625, and minimum thickness of λ. ◦

ρ(35 ) =

30λ = 26.15λ 2 sin 35◦

462

LENS ANTENNAS

Rearrange Eq. (9-35) to calculate the focal length at the edge: f =

(1 − n cos 35◦ )ρ(35◦ ) = 34.03λ 1−n

Equation (9-34) gives us the plate separation with a slight rearrangement: 1 = 0.64λ a= √ 2 1 − n2 The amplitude variation from the center to the edge caused by the ellipse is given by Eq. (9-37a): A(ψe ) (1 − n cos ψe )3 = = 4.26 (6.3 dB) A(0) (1 − n)2 (cos ψe − n) A feed with its 10-dB beamwidth equal to the subtended angle of the lens at the feed produces −3.7 dB edge taper in the aperture. The edge thickness is given by ◦

◦

t = f + 1 − ρ(35 ) cos(35 ) = 13.61λ Equation (9-36) predicts a bandwidth of 1.9%. A detailed analysis using ray tracing through the lens and varying n with changes in frequency predicts a 0.2-dB loss at this band edge. The 1-dB bandwidth is about 4.5%. Zoning a parallel-plate lens increases its bandwidth by limiting the maximum thickness, since the variation of the optical path length due to the varying n exceeds that due to the zoning. Figure 9-9 gives the central curve of the three possible types of zoning. The lens in Figure 9-9a only suffers loss due to diffractions from edges. The other two zoned lenses (Figure 9-9b, c) have dead zones in the aperture. These dead zones produce additional amplitude taper loss and high close-in sidelobes. The feed-side zoning has different focal lengths in each zone. The outer zone remains the same as the unzoned lens. The focal lengths of the inner zones are reduced by λ/(1 − n) at each step and the stepping reduces the amplitude taper of the lens by varying f : f 2 (1 − n cos ψ)3 A(ψ) = 2 c (9-38) A(0) f (1 − n)2 (cos ψ − n)

r(y) y Feed Elliptical

(a)

(b)

(c)

FIGURE 9-9 Central section of zoned parallel metal plate lenses.

SURFACE MISMATCH AND DIELECTRIC LOSSES

463

where fc is the central focal length and f is the focal length of the ellipse at ψ. The zoned lens bandwidth is approximately [1] BW =

25% K − 1 + [(1 + n)(1 − n)t/n]

(9-39)

where K is the number of zones and t is the maximum thickness. Equation (9-39) also is an underestimation of an acceptable loss-level bandwidth. Example The lens of the example above was zoned as in Figure 9-9a and c with ﬁve zones. The maximum thickness is 3.4λ. Equation (9-39) predicts a bandwidth of 3.4%. By tracing rays through the lens and applying Eq. (4-9) to calculate phase error loss, we predict a 10% bandwidth for a 1-dB loss. Zoning the nonrefracting surface has no effect on the aperture distribution except for edge diffractions that modify the ﬁelds slightly. Zoning the refracting surfaces causes aperture dead zones and reduces the lens-induced amplitude taper. The focal length of the ellipse at the edge remains at 34.03λ. The focal length of the center ellipse is reduced by 4λ/(1 − n) from the edge ellipse to 23.36λ. The edge taper [Eq. (9.38)] becomes 2.01 (3 dB). The aperture dead zones increase the loss by 2 dB. The bandwidth of a parallel-plate lens can be increased by a method of compounding lenses into a doublet [5]. We can make a lens by using a uniform waveguide length between the input and output surfaces and placing a phase shifter in each line to compensate for the optical path-length differences and produce an eikonal at the aperture plane. If we combine a refracting surface waveguide plate lens with a differential phase shift lens, we can match the aperture phase at two frequencies. This matching broadbands the antenna like an optical achromatic doublet. 9-6 SURFACE MISMATCH AND DIELECTRIC LOSSES The reﬂection coefﬁcient of a wave normally incident on a dielectric is =

1−n 1+n

(9-40)

valid for both dielectric and metal lenses. The actual reﬂection coefﬁcient of any ray depends on the angle of incidence and the polarization. Both surfaces of the lens have reﬂections, and these interact to produce the actual reﬂection. With plane surfaces, such as those assumed for radomes, we can analyze the combination of reﬂections using transmission-line mismatch equivalence. Since the reﬂections from second surfaces may not return to the same point as the incident waves and may have their caustic distances changed by the curved surfaces, transmission-line models of the two surfaces have limited use for lenses. Equation (9-40) gives us a reasonable approximation to expected mismatch, since one surface of the single refracting surface lens will be normal to the wave incident from the feed and reﬂect into the feed. The second surface fails to focus power back to the feed and has a minor effect. Example A lens with n = 1.6 or n = 0.625 has a reﬂection coefﬁcient magnitude of 0.23 [Eq. (9-40)] at one surface that focuses into the feed. This gives a feed mismatch if VSWR =

1 + 0.23 = 1.6 1 − 0.23

464

LENS ANTENNAS

the same as n (n > 1) or 1/n (n < 1). The mismatch loss becomes 1 − ||2 = 0.95 (0.2 dB). The wave may be matched to the surface by a quarter-wavelength transformer with an index of refraction n1/2 , but adding matching transformers narrows the bandwidth. The surface, which reﬂects power back to the feed, should be matched ﬁrst, since the second surface has a minor effect on feed mismatch. Also, the primary reﬂecting surface has normally incident waves and does not suffer from the need to vary the thickness to match waves off normal incidence. Transformers to match waves off normal incidence are polarization-sensitive. Simple methods can be used to reduce lens-caused feed mismatch [2]. The lens can be tilted to cause the reﬂection to miss the feed. Offsetting half the lens by λ/4 causes cancellation of the reﬂection from the two halves. Tilting does not reduce the mismatch loss but does produce backlobe power in the pattern. Similarly, the reﬂected power from the hyperbolical surface forms backlobes. These reﬂections reduce the antenna efﬁciency below that predicted by aperture theory alone. Cohn and Morita [2,6] developed methods of matching the surface of the lenses by removing some of the dielectric for a quarter wavelength. The surfaces are either corrugated, have arrays of holes, or have arrays of rods. With this method, the lens can be made from a single dielectric slab. The design depends on the angle of incidence and the polarization of the waves. The lens dissipates power by the attenuation constant of the material: 27.3n tan δ α= dB/length (9-41) λ where tan δ is the loss tangent of the dielectric. Waveguide losses reduce the power transmitted through metal plate lenses. Zoning eliminates material and its associated loss to improve efﬁciency, but for most materials, this effect is small. Artiﬁcial dielectrics [2] reduce excessive weight and material losses of lenses. We make them by embedding metal particles or plated microspheres in foam with a dielectric constant near 1. The metal parts may be strips or disks made from metal foil. Similarly, solid metal parts can be hollow. Since the effective dielectric constant depends on the size of the metal particles in wavelengths, lenses made from artiﬁcial dielectrics will be narrowband if the particles are large, but the use of plated microspheres dispersed in the foam reduces this problem.

9-7 FEED SCANNING OF A HYPERBOLOIDAL LENS [7] The hyperboloidal lens has no cross-polarization when fed from an electric dipole source. Kreutel [7] analyzed the effects of off-axis dipole sources on the pattern of the hyperboloidal lens. The coma increases more rapidly for the lens than for a paraboloidal reﬂector for the same scanning. Like the paraboloidal reﬂector, the hyperboloidal lens beam scans less than the deviation angle of the feed relative to the vertex and axis and has a beam deviation factor (Table 9-5). The scanning loss (Table 9-6) decreases with increasing n. The peak coma lobe (Table 9-7) limits the possible scan before unusable patterns are obtained. The paraboloidal reﬂector can be scanned further (Table 8-2) for the same coma.

DUAL-SURFACE LENSES

TABLE 9-5 f/D 0.8 1.0 1.2

Beam Deviation Factor for a Feed-Scanned Hyperboloidal Lens √ √ √ n= 2 n=2 f/D n= 2 n=2 f/D n= 2

465

0.75 0.80 0.83

0.84 0.87 0.89

1.4 1.6 1.8

0.86 0.89 0.92

0.91 0.92 0.94

2.0 2.5 3.0

0.93 0.95 0.97

n=2 0.95 0.96 0.98

TABLE 9-6 Scanning Loss for a Hyperboloidal Lens (dB) √ n=2 n= 2 Beamwidth of Scan f/D = 1 f/D = 2 f/D = 1 f/D = 2 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5

0.03 0.06 0.12 0.23 0.36 0.51 0.69 0.90 1.09

0.00 0.01 0.03 0.05 0.09 0.13 0.19 0.24 0.31 0.38 0.44

0.01 0.04 0.07 0.12 0.20 0.28 0.37 0.49 0.60 0.75

0.00 0.01 0.02 0.04 0.06 0.08 0.11 0.14 0.18 0.22 0.25

TABLE 9-7 Coma Sidelobe Level for a Scanned Hyperboloidal Lens (dB) √ n=2 n= 2 Beamwidth of Scan f/D = 1 f/D = 2 f/D = 1 0 1 2 3 4 5 6

20.9 17.6 15.2 13.1 11.3 9.8 8.8

18.5 17.5 16.5 15.5 14.8 14.0 13.3

19.5 17.6 15.7 14.2 12.9 11.5 10.4

9-8 DUAL-SURFACE LENSES The second surface of the lens offers an additional degree of freedom that can be used to control the pattern characteristics. Ruze [8] developed methods to reduce coma for feed-scanned cylindrically shaped metal plate lenses, which constrain the wave parallel with the axis. Both surfaces are used to satisfy focusing requirements. We will develop a method for axisymmetrical dielectric lenses to eliminate coma for small feed

466

LENS ANTENNAS

displacements. In a second design we can also use the second surface shape to control the amplitude distribution in the aperture plane. 9-8.1 Coma-Free Axisymmetric Dielectric Lens [9] The design of the coma-free axisymmetric antenna reduces to the numerical solution of a differential equation with side conditions to produce a collimated beam and satisfy the Abbe sine condition [10, p. 157]. Successful designs require a number of iterations, since the ultimate lens shape depends heavily on the initial conditions. Solution of the differential equation will sometimes diverge into unrealizable designs or fail to continue to satisfy the side conditions. A lens satisfying the Abbe sine condition is free of coma aberrations for small deviations of the feed from the axis. The deviations produce higher-order aberrations that eventually distort the beam with continued scanning, but coma is removed. For a lens focused at inﬁnity, the Abbe sine condition requires that the surface which refracts the waves parallel with the axis must be spherical with its center on the effective focus of the lens. The dielectric lens refracts waves parallel with the axis on the outer surface (away from the feed). Given the aperture radial component r, r = fe sin ψ

(9-42)

where fe is the effective focus and ψ is the feed angle. A waveguide plate lens satisﬁes the Abbe sine condition by having a spherical inner surface [6], since the waves are parallel with the axis in the lens because the waveguide plates constrain the wave to be parallel with the axis. The second surface must produce the conditions for a uniform phase in the aperture plane. The waveguide plate lens only has to equalize path lengths. In a dielectric lens, the inner surface must refract the waves in the proper direction to satisfy the Abbe sine condition, and it must be so placed as to equalize the path lengths from the feed to the aperture plane. The locations of both surfaces along the axis are varied to equalize the path lengths. Figure 9-10 shows the coordinates of the coma-free dielectric lens. The polar equation ρ (ψ) describes the inner surface and ψ is the angle of the refracted wave with the axis. The distance from the feed to the center of the lens inner surface is f , and T is the thickness. The coordinates (r, z) describe the outer lens surface, where z is the axis dimension and r is the aperture radial component. Snell’s law reduces to a differential equation at the inner surface: dρ n sin(ψ − ψ )ρ = dψ n cos(ψ − ψ ) − 1 where

(9-43)

r − ρ sin ψ z − ρ cos ψ

(9-44)

(fe − ρ) sin ψ z − ρ cos ψ

(9-45)

tan ψ = By use of Eq. (9-42), this reduces to tan ψ =

DUAL-SURFACE LENSES

467

(r,z) y′ r(y) D

y f

T

Feed

FIGURE 9-10 Coma-corrected dual-surface axisymmetric lens; n = 1.6, D = 35, f = 45, T = 6.5, fe = 49.

The requirement for equal optical path lengths to the aperture determines a quadratic equation in z: Az2 + Bz + C = 0 (9-46) where A = n2 − 1 B = 2(ρ − K) − 2n2 ρ cos ψ C = n2 ρ 2 cos2 ψ + n2 (fe − ρ)2 sin2 ψ − (ρ − K)2 K = T (n − 1) √ −B + B 2 − 4AC z= 2A Design consists of the numerical solution of the differential equation (9-43) subject to the conditions of Eqs. (9-45) and (9-46). Realizable solutions depend on the initial conditions. Most failures to produce a usable design occur in Eq. (9-46), which satisﬁes the requirement of equal aperture phase. Example Figure 9-10 shows a scale drawing of a realizable design for n = 1.6, focal distance f = 45, diameter D = 35, center thickness T = 6.5, and effective focal length fe = 49. Table 9-8 lists a few points of the solution obtained by a Runge–Kutta numerical method for the differential equation (9-43). The example above contains only relative dimensions. The solution is size and frequency independent, since it is obtained by geometric optics. We can zone the lens for a given frequency by cutting along ray paths. Each step is λ/(n − 1). Table 9-8, as in the example above, determines the ray paths through the lens. Zoning will produce either feed or aperture dead zones. The reduction of weight must be balanced with

468

LENS ANTENNAS

TABLE 9-8 Design of Figure 9-11 for a Coma-Free Lens Feed Angle, ψ (deg)

Inner Surface, ρ(ψ)

Horizontal Distance, z

Radius, r

Thickness Along Ray, T

0 5 10 15 20 20.92

45.00 45.18 45.70 46.59 47.63 47.61

51.50 51.19 50.27 48.71 46.28 45.40

0 4.27 8.51 12.68 16.76 17.50

6.50 6.20 5.29 3.77 1.59 1.06

the loss in efﬁciency to achieve some compromise. Because we used the degrees of freedom of the second surface to satisfy the Abbe sine condition, we lose control of the aperture distribution through the lens surfaces. Most practical designs produce an amplitude taper near that of the feed antenna. We must achieve low sidelobes, if required, through a tapered illumination from the feed. The feed pattern plays no part in the design and gives us degrees of freedom for amplitude taper. An antenna designed and built with a diameter of 32 wavelengths [9] showed no coma in a scanning of ±2 beamwidths. Like the paraboloidal reﬂectors, increasing the focal length for a given diameter allows greater scanning without signiﬁcant coma. 9-8.2 Speciﬁed Aperture Distribution Axisymmetric Dielectric Lens [11] We use the desired aperture amplitude distribution to specify the relation between the aperture radius r, and the feed angle ψ. Earlier, the Abbe sine condition established this relation. Given a feed power pattern F (ψ) and a required aperture distribution A(r), we relate the two through differential areas: F (ψ) sin ψ dψ = A(r)r dr where an axisymmetrical pattern is assumed. We derive the relation between ψ and r through normalized integrals as in Section 8-20:

ψ

0 ψm 0

F (ψ) sin ψ dψ = F (ψ) sin ψ dψ

r

0rm

A(r)r dr (9-47) A(r)r dr

0

In any particular design we generate a table, such as Table 8-12, of the feed angle versus its normalized feed pattern integral and the aperture radius versus its normalized aperture distribution integral. For a given feed angle ψ we equate the normalized integrals to compute the corresponding aperture radius. We generate a table of aperture radius versus feed angle, such as Table 8-13, using interpolation techniques. The design is very dependent on feed pattern, because changing the feed pattern alters the table. Low sidelobe aperture distributions require close tolerances and a good speciﬁcation of the feed pattern. Once we have the relation between ψ and r, the design follows steps similar to those taken in designing the lens for the Abbe sine condition. We solve the differential

DUAL-SURFACE LENSES

469

equation (9-43) numerically. We specify the aperture radius by the table generated for the transformation of the feed pattern into the aperture distribution. The requirement for equal path length through the lens determines the axis location z of the outer surface. Ax 2 + Bx + C = 0

x =z−f

A=n −1 2

B = 2[n2 (f − ρ cos ψ)] + ρ − K − f

(9-48)

C = [n(ρ cos ψ − f )]2 + (r − ρ sin ψ)2 − (f + K − ρ)2 √ −B + B 2 − 4AC K = T (n − 1) z=f + 4A where T is the central thickness and f is the axis focal length. A successful design requires a number of iterations, starting with different initial conditions. The differential equation solution will diverge to shapes unable to satisfy the equal-path-length side requirement with poor initial conditions. Each design has a narrow range of satisfactory initial conditions. In most cases, increasing the thickness increases the chance for a successful design. Example A lens was designed to transform the feed pattern of a conical corrugated horn into a circular Taylor distribution with 40-dB sidelobes (n = 8). The initial conditions were n = 1.6, diameter D = 32, focal distance f = 35, central thickness T = 9, and maximum feed angle ψm = 20◦ . A table of the feed angle and its normalized power pattern integral, along with the aperture radius and its normalized power distribution integral, was generated. A table of feed angle and the corresponding aperture radius follows from equating normalized integrals. This table is independent of the lens thickness but not the feed pattern. A Runge–Kutta numerical method is used to solve the differential equation (9-43) subject to the conditions imposed by the aperture radius table and Eq. (9-48) for equal optical path lengths. Figure 9-11 shows a design for a 36◦ 10-dB beamwidth feed horn (12-dB feed edge taper). The horn dimensions are aperture radius = 1.90λ and slant radius = 9λ for a maximum quadratic phase deviation S = 0.2. A few points of the design are listed in Table 9-9. Other antennas designed with small changes in the feed pattern beamwidth show signiﬁcant changes in the lens shape near the edges for a constant center thickness. Axisymmetric dielectric lenses can be designed to be independent of frequency because only relative sizes are speciﬁed. The lenses tend to be thick to allow room to satisfy the requirement for an equal optical path length. Zoning lens reduces weight while decreasing bandwidth. Low-sidelobe designs are inherently narrowband, since small changes in the beamwidth of the feed alter the aperture distribution and sidelobe levels. Zoning may not reduce the bandwidth significantly. An antenna designed and tested using the technique above revealed a number of requirements on the design [12]. The sidelobe levels exceeded the design speciﬁcation for three main reasons. First, the feed pattern was speciﬁed as sin(πU )/πU , an oversimpliﬁcation of the actual feed pattern. Realistic feed patterns must be used because small changes in the feed pattern require new designs. Second, the surfaces must be matched with quarter-wavelength sections to prevent reﬂections, unaccounted

470

LENS ANTENNAS

(r,z) y′ r

r(y) Feed

D

y ym

T

f

FIGURE 9-11 Dual-surface axisymmetric lens for circular Taylor aperture distribution (40-dB, nˆ = 8). Lens: n = 1.6, D = 32, f = 35, T = 9, ψm = 20◦ . Feed: conical corrugated horn, 36◦ 10-dB beamwidth. S = 0.2.

TABLE 9-9 Design of Figure 9-12 for a Speciﬁed Aperture Distribution Axisymmetric Dielectric Lens Feed Angle, ψ (deg)

Inner Surface, ρ(ψ)

Horizontal Distance, z

Radius, r

Thickness Along Ray, T

0 5 10 15 20

35.00 35.31 36.14 37.69 38.71

44.00 43.95 43.67 43.25 40.58

0 3.27 6.59 10.11 16.00

9.00 8.78 8.08 6.86 5.03

for in the design, that change the aperture distribution. Third, diffractions from the edges affect the distribution. Increasing the aperture diameter or using a low edge illumination feed reduces these effects. As designed, the lens exhibits severe coma when scanned by feed lateral offset. Since most lenses are quite thick, zoning can be used to approximate the Abbe sine condition on the inner surface. The lens refracts most of the rays parallel with the axis by the inner surface. If the zones approximate a spherical surface on the average, coma is reduced for the scanned beams. These coma-corrected lenses are useful for multibeam applications when each beam is fed from an offset feed.

9-9 BOOTLACE LENS The bootlace lens consists of a set of receiving antennas on a surface connected by cables to a set of transmitting antennas on the second surface (Figure 9-12). The cables constrain the path through the lens. We have three degrees of freedom with this lens: (1) input surface, (2) output surface, and (3) cable length. We can change the lens characteristics dynamically by placing phase shifters and/or attenuators in the lines

BOOTLACE LENS

471

y

Radiators Radiators F1 F2 x F2 F1

Focal Arc

C1

C2 Flexible Cables

FIGURE 9-12 Bootlace lens. (From [13], Fig. 3, 1965 IEEE.)

between the input and output radiators and scan one or more beams. The input and output surfaces are arrays, and we can generate multiple beams by placing more than one feed on a focal arc determined by the lens geometry. The simplest bootlace lens consists of a spherical input surface connected to a plane output surface by equal-length cables. This lens converts spherical waves radiated by the feed into plane waves at the output surface. The lens uses true time delay, which removes bandwidth limitations. Most bootlace lenses are line sources or twodimensional lenses fed by line sources. The general lens can have four focal points [13] placed symmetrically about the axis of the symmetric structure. The focal arc is chosen on a curve through the focal points to minimize defocusing when feeds are placed off the focal points. A feed at each point along the focal arc produces an output beam in a different direction. Because each feed uses the full aperture, it achieves the full array gain less the loss of projection of the aperture length in the beam direction. The number of focal points is reduced to three when the lens is further restrained. The Ruze design for a metal plate lens [8] has three focal points, since the waveguides between the surfaces travel in straight lines. There is one central axis focal point and two symmetrically placed focal points. The Rotman [14] lens loses one possible focal point because the output surface is limited to a straight line. A parallel-plate structure in the Rotman lens leads from the possible feed locations to the feed-side surface, which normally is excited by probes in the parallel-plate guide. The lens becomes a feed network that produces multiple beams whose directions depend on the location of the feed on the focal arc. Although perfect focusing is achieved at only three points, the phase error loss associated with points between them is small. Because the Rotman lens feed network is a true time-delay array feed network, we can achieve bandwidths greater than an octave from it. Rao [15] extends the design of bootlace lenses to three dimensions and shows that the number of focal points cannot be extended beyond four. Because the lens is not

472

LENS ANTENNAS

axisymmetric, it has different scanning capabilities in orthogonal planes. Rao designs lenses with two, three, and four focal points on a focal line. Decreasing the number of focal points in one plane increases the scanning capability in the orthogonal plane for a given phase error level.

9-10

LUNEBURG LENS [16, p. 545]

A Luneburg lens, a spherically symmetric lens with a variable index of refraction, radiates a beam in any direction for a feed located opposite the beam. We place the feed phase center either on the surface of the lens or a short distance away. We form multiple beams by feeding the lens at a number of places. Our only restriction is the blockage due to other feeds or support structures. We can rapidly scan a beam by moving a lightweight feed around the sphere or by switching between multiple feeds. When we place the feed on the outer surface of the sphere, the required index of refraction is

r 2 n= 2− (9-49) a where a is the lens outer radius and r is the inner radius. The dielectric constant n2 must vary between 2 in the center and 1 on the outer surface. Few feeds have their phase centers on a surface that can be mounted against a sphere. We can move the feed away from the surface by changing the variation of the index of refraction from that given by Eq. (9-49), but the required center index of refraction decreases as we move the feed away from the lens surface. We calculate the variation of the index of refraction from an integral equation, and the curves follow the general shape of Eq. (9-49). For the feed-to-sphere radius of 1.1, the proper center dielectric constant is 1.83 and it varies smoothly to 1 at the lens surface. Similarly, the center dielectric constant starts at 1.68 for the feed-to-sphere radius of 1.2. The lens changes the amplitude distribution in the aperture compared with the feed. Given the ratio of feed radius to lens radius ri , the aperture plane power distribution becomes F (ψ) A(r) = 2 (9-50) r1 cos ψ where ψ is the feed angle, F (ψ) the feed power pattern, and A(r) the aperture power distribution. Equation (9-50) shows that the lens refracts power toward the edge of the aperture. Lenses have been made by using a series of concentric spherical shells each with a constant dielectric. A minimum of 10 shells is needed for an adequate approximation of the required variation of the dielectric constant. REFERENCES 1. J. R. Risser, Chapter 11 in S. Silver, ed., Microwave Antenna Theory and Design, McGrawHill, New York, 1948. 2. S. B. Cohn, Chapter 14 in H. Jasik, ed., Antenna Engineering Handbook, McGraw-Hill, New York, 1961. 3. S. Cornbleet, Microwave Optics, Academic Press, London, 1976.

REFERENCES

473

4. B. E. A. Saleh and M. C. Teich, Fundamentals of Photonics, Wiley, New York, 1991. 5. A. R. Dion, A broadband compound waveguide lens, IEEE Transactions on Antennas and Propagation, vol. AP-26, no. 5, September 1978, pp. 751–755. 6. T. Morita and S. B. Cohn, Microwave lens matching by simulated quarter-wave transformers, IEEE Transactions on Antennas and Propagation, vol. AP-4, no. 1, January 1956, pp. 33–39. 7. R. W. Kreutel, The hyperboloidal lens with laterally displaced dipole feed, IEEE Transactions on Antennas and Propagation, vol. AP-28, no. 4, July 1980, pp. 443–450. 8. J. Ruze, Wide-angle metal-plate optics, Proceedings of IRE, vol. 38, no. 1, January 1950, pp. 53–59. 9. J. J. Lee, Numerical methods make lens antennas practical, Microwaves, vol. 21, no. 9, September 1982, pp. 81–84. 10. R. Kingslake, Lens Design Fundamentals, Academic Press, New York, 1978. 11. J. J. Lee, Dielectric lens shaping and coma-correcting zoning, part I: analysis, IEEE Transactions on Antennas and Propagation, vol. AP-31, no. 1, January 1983, pp. 211–216. 12. J. J. Lee and R. L. Carlise, A coma-corrected multibeam shaped lens antenna, part II: experiments, IEEE Transactions on Antennas and Propagation vol. AP-31, no. 1, January 1983, pp. 216–220. 13. M. L. Kales and R. M. Brown, Design considerations for two dimensional symmetric bootlace lenses, IEEE Transactions on Antennas and Propagation, vol. AP-13, no. 4, July 1965, pp. 521–528. 14. W. Rotman and R. F. Turner, Wide-angle microwave lens for line source applications, IEEE Transactions on Antennas and Propagation, vol. AP-11, no. 6, November 1963, pp. 623–632. 15. J. B. L. Rao, Multifocal three-dimensional bootlace lenses, IEEE Transactions on Antennas and Propagation, vol. AP-30, no. 6, November 1982, pp. 1050–1056. 16. R. S. Elliott, Antenna Theory and Design, Prentice-Hall, Englewood Cliffs, NJ, 1981.

10 TRAVELING-WAVE ANTENNAS

Traveling-wave antennas consist of transmission-line structures that radiate. We develop a uniﬁed theory for end-ﬁre line antennas because length and propagation constant along the structure determine most of their properties. To ﬁrst order, length determines gain and bandwidth. The size and shape of the structure produce secondary effects such as polarization nulls and narrower beamwidths. Most of these structures are slow wave-transmission structures that bind waves to it and radiate at discontinuities. We use surface-wave structures to radiate end-ﬁre beams and leaky wave structures to radiate beams at an angle to the axis of the line source. In both cases there are planar conﬁgurations that have their uses, but in this chapter we concentrate on long, thin geometries. We combine leaky wave line-source radiators, such as slotted rectangular waveguides, into planar arrays, but the line source remains the building block. We make traveling-wave antennas from structures that guide waves. Surface-wave structures bind the power to the transmission line and radiate from discontinuities such as bends or dimensional changes. In some cases we analyze the surface wave as radiating throughout its extent on the transmission line. Both methods provide insight. Leaky wave antennas carry waves internally, such as a waveguide, and radiate at openings that allow power to escape. The radiation mechanism differs in the two cases, but we use similar mathematics to describe both types. We may have trouble distinguishing the radiation mode because the structures may be similar because with small changes in structure, some antennas can radiate in either mode. We separate traveling-wave antennas from other antennas by the presence of a wave traveling along the structure, with most of its power propagating in a single direction. We divide antennas by their structure: line and planar. We usually analyze planar structures as being inﬁnite in the direction normal to the wave propagation. Similarly, we usually ignore the diameter of line sources in a ﬁrst-order analysis. The diameter is important for determining the mode structure, but to ﬁrst order we calculate patterns Modern Antenna Design, Second Edition, By Thomas A. Milligan Copyright 2005 John Wiley & Sons, Inc.

474

GENERAL TRAVELING WAVES

475

based on a thin line source since length and propagation constant determine the pattern and bandwidth. The width of a planar structure determines the pattern beamwidth in that plane. Increasing the diameters of the rods of line sources will decrease the pattern beamwidth and increase gain, but the effect is secondary. Only when we include the diameter can we make the transition to aperture-type structures considered to be radiating from the end. In this chapter we must consider unusual transmission-line structures. Properly designed dimensions provide the proper phase velocity to establish a single end-ﬁre beam for slow-wave antennas or to point the beam of a leaky wave antenna. We calculate some of the dimensions by analysis (an ever-expanding list), but we can also measure the velocities and leakage and proceed to design. 10-1 GENERAL TRAVELING WAVES A wave traveling in a single direction has a ﬁeld representation: E = E0 (z)e−kP z

(10-1)

where z is the direction of propagation, k the free-space propagation constant (wave number) 2π/λ, and P the relative propagation constant. E0 (z) describes the amplitude variation: P > 1 surface waves P < 1 leaky waves

(10-2)

For a planar structure in the y –z plane, we consider separable distributions: E = E0 (z)E1 (y)e−j kP z We compute the pattern from L f = 0

a −a

E0 (z)E1 (y)e−j kP z ej kz z ej ky y dz dy

(10-3)

where kz = k cos θ . Similarly, for circular distributions we have E = E0 (z)E1 (φ)e−j kP z and

L

f =

E0 (z)e

−j kP z j kz z

e

2π

dz

0

E1 (φc )aej ka sin θ cos(φ−φc ) dφc

(10-4)

0

where a is the radius. The second integral includes vector dot products to project the ring aperture ﬁelds on to the far-ﬁeld polarizations (see Section 7-2). We consider only the term along the z-axis, and we can consider the effect of the other coordinate separately. The pattern response is L f = E0 (z)e−j kz (P −cos θ) dz (10-5) 0

476

TRAVELING-WAVE ANTENNAS

We use the results of Chapter 4 with these separable distributions. Maximum gain comes from a uniform distribution reduced by the amplitude taper efﬁciency for tapered distributions. In Eq. (10-3) the y-axis distribution and size determine the gain factor as a product for aperture area. Equation (10-4) has a separable φ distribution that separates directivity into a product. We ignore these factors for now and concentrate on the z-axis pattern and associated directivity. Linear-rod antennas have increased directivity because of dipole φ distributions and their increased radius. A traveling wave with a uniform distribution has pattern response sin(ψ/2) ψ/2

where ψ = kL(P − cos θ )

(10-6)

for θ measured from the z-axis. The y or φ distribution determines the pattern in the other coordinate. For P > 1, a slow wave, the beam peak approaches θ = 0 when P → 1. The length bounds the range P for an end-ﬁre pattern peak. Leaky waves, P < 1, have a pattern peak when P = cos θ , or θmax = cos−1 P

(10-7)

The pattern peak approaches end ﬁre (θ = 0) as P → 1. By increasing P beyond 1, the directivity increases and reaches maximum value for a given P , depending on the length [1] 0.465 P =1+ (10-8) L Equation (10-8) is the Hansen and Woodyard criterion for increased directivity of a long end-ﬁre structure commonly approximated by [2] P =1+

1 2L

(10-9)

The phase increase of 180◦ [Eq. (10-9)] along the length gives the maximum directivity for a long structure with a uniform distribution. The amplitude distribution for most surface-wave devices (P > 1) peaks near the input and the taper reduces the gain by the amplitude taper efﬁciency [Eq. (4-8)]. We reduce the relative propagation constant from that given by Eq. (10-8) depending on the length [3]: P =1+

1 RL

(10-10)

where R = 6 at L = λ, diminishing to 3 from L = 3λ to L = 8λ and tapering to 2 [Eq. (10-9)] at L = 20λ. Zucker [4] uses R = 6 for the amplitude, which peaks by 3 dB at the input for all lengths. Equations (10-8) and (10-10) give designs with only small differences in gain. The value of P controls one edge of the visible region. Setting P = 0 centers the visible region about ψ = 0. End ﬁre occurs at P = 1. As we increase P beyond 1, the beam peak of the distribution in ψ space moves into invisible space and the sidelobe level increases. A progression of the distribution sidelobes becomes beam peaks as P increases. Since the amplitude difference between sidelobes decreases as the sidelobe

477

GENERAL TRAVELING WAVES

number increases, the sidelobe level of the pattern increases as the pattern degrades because P exceeds the value given by Eq. (10-9). Figure 10-1 shows the effects on directivity of varying P on an axisymmetrical traveling-wave antenna with a uniform amplitude distribution. For broadside radiation (P = 0) and for P near end ﬁre, the directivity is a constant value with scan: directivity =

2L λ

(10-11)

When the broadside conical beam is scanned until the cone joins into a single end-ﬁre beam, the directivity increases. For end ﬁre, P = 1: directivity =

4L λ

end ﬁre

(10-12)

SL λ

(10-13)

The directivity peaks for P given by Eq. (10-8): directivity =

10l

18

8l

Directivity, dB

16

6l 5l

14 4l 3l

12 2l

10

8

6 0.8

0.9

1.0

1.1

1.2

1.3

1.4

Relative Propagation Constant, P

FIGURE 10-1 Directivity of an axisymmetrical uniform-distribution traveling wave.

478

TRAVELING-WAVE ANTENNAS

20 HE11 Mode

18

Directivity, dB

16

Hansen-Woodyard Criterion

Helical Antenna

P=1

14 12 10 8 6 4 1

2

3

4

5

7 6 Length, l

8

9

10

11

12

FIGURE 10-2 Directivity of an end-ﬁre traveling-wave antenna.

L/λ

2

4

6

10

20

S

7.92

7.58

7.45

7.33

7.25

Figure 10-2 plots the maximum directivity of end-ﬁre structures versus length. For the case P = 1, Eq. (10-12) gives the directivity on the curve. The Hansen and Woodyard criterion increases the directivity as shown in Figure 10-2 for an inﬁnitesimal-diameter structure. The distribution on the ﬁnite diameter of the helical wire antenna in the axial mode increases the directivity over that for the Hansen and Woodyard increased directivity criterion. The hybrid mode with its linear polarization has a dipole null normal to the traveling-wave axis whose elemental pattern also increases directivity. Figure 10-3 is a plot of the corresponding beamwidths of those structures. Figure 10-2 sets an upper bound to the possible directivity of a small-diameter end-ﬁre travelingwave structure of given length. 10-1.1 Slow Wave A slow wave exists on an open transmission-line structure that binds the wave by slowing a passing wave and bending it in the direction of the structure. In the same manner, a lens bends waves toward regions of higher index of refraction (increased slowing). We designate x as the direction normal to a planar structure and the radial coordinate ρ as the direction normal to the cylindrical slow-wave structure. The relation between propagation constants in various directions is found in any electromagnetics text [5]: kz2 + kx2 = k 2 or kz2 + kρ2 = k 2 (10-14) Since x (or ρ) is unbounded, the waves must attenuate exponentially from the surface: α = j kx

or

α = j kρ

(10-15)

GENERAL TRAVELING WAVES

479

90

Beamwidth (degrees)

80 70 60 Hansen-Woodyard Criterion

50

P=1

40

HE11 Helical

30 20 10 1

2

3

4

5

7 6 Length, l

8

9

10

11

12

FIGURE 10-3 Beamwidth of a traveling-wave end-ﬁre antenna.

The z-directed propagation constant becomes kz2 = k 2 + α 2 = P 2 k 2

where P =

α2 1+ 2 = k

1+

λα 2π

2 (10-16)

P , the relative propagation constant, becomes a measure of the wave binding to the surface. We rearrange Eq. (10-16): α=

2π 2 2π 2 P − 1 (Np/λ) = 8.63 P − 1 (dB/λ) λ λ

As P increases, the wave is more tightly bound to the surface. Figure 10-4 is a plot of the distances normal to the surface of constant-ﬁeld contours versus P . The ﬁelds attenuate rapidly normal to the surface. For P → 1, the slow-wave structure only diffracts passing plane waves without capturing the power. This is the sense of a cutoff frequency for the structure. Most surface-wave antennas consist of three regions. The feed region launches the wave on the structure with P between 1.2 and 1.3 [4]. The structure tapers in a short section until P suitable for the length is reached. We design for a given phase shift along the entire length. For example, a long antenna would be designed so that the wave on the structure has an excess phase shift of 180◦ [Eq. (10-9)] over the traveling wave in free space. Near the end we sometimes taper the structure to reduce the end reﬂection given approximately by [4] P 2 − 1 (power). This end taper can be quite short and achieve good results.

480

TRAVELING-WAVE ANTENNAS 4

Distance, l

3

2 60 dB 50 dB 40 dB

1

30 dB 20 dB 6 dB 3 dB

1

10 dB

1.1

1.2

1.3

1.4

1.5

Relative Propagation Constant, P

FIGURE 10-4

Constant ﬁeld contours off the surface of a surface-wave structure.

10-1.2 Fast Waves (Leaky Wave Structure) Only closed structures such as waveguides support fast waves. An open structure requires a negative α [Eq. (10-16)] for fast waves, which implies an exponentially increasing wave away from the structure. The structure soon radiates all its power and no longer guides t

THOMAS A. MILLIGAN

IEEE PRESS

A JOHN WILEY & SONS, INC., PUBLICATION

MODERN ANTENNA DESIGN

MODERN ANTENNA DESIGN Second Edition

THOMAS A. MILLIGAN

IEEE PRESS

A JOHN WILEY & SONS, INC., PUBLICATION

Copyright 2005 by John Wiley & Sons, Inc. All rights reserved. Published by John Wiley & Sons, Inc., Hoboken, New Jersey. Published simultaneously in Canada. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning, or otherwise, except as permitted under Section 107 or 108 of the 1976 United States Copyright Act, without either the prior written permission of the Publisher, or authorization through payment of the appropriate per-copy fee to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, 978-750-8400, fax 978-646-8600, or on the web at www.copyright.com. Requests to the Publisher for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, (201) 748-6011, fax (201) 748-6008. Limit of Liability/Disclaimer of Warranty: While the publisher and author have used their best efforts in preparing this book, they make no representations or warranties with respect to the accuracy or completeness of the contents of this book and speciﬁcally disclaim any implied warranties of merchantability or ﬁtness for a particular purpose. No warranty may be created or extended by sales representatives or written sales materials. The advice and strategies contained herein may not be suitable for your situation. You should consult with a professional where appropriate. Neither the publisher nor author shall be liable for any loss of proﬁt or any other commercial damages, including but not limited to special, incidental, consequential, or other damages. For general information on our other products and services please contact our Customer Care Department within the U.S. at 877-762-2974, outside the U.S. at 317-572-3993 or fax 317-572-4002. Wiley also publishes its books in a variety of electronic formats. Some content that appears in print, however, may not be available in electronic format. Library of Congress Cataloging-in-Publication Data: Milligan, Thomas A. Modern antenna design / by Thomas A. Milligan.—2nd ed. p. cm. Includes bibliographical references and index. ISBN-13 978-0-471-45776-3 (cloth) ISBN-10 0-471-45776-0 (cloth) 1. Antennas (Electronics)–Design and construction. I. Title. TK7871.6.M54 2005 621.382 4—dc22 Printed in the United States of America. 10 9 8 7 6 5 4 3 2 1

2004059098

To Mary, Jane, and Margaret

CONTENTS

Preface 1 Properties of Antennas

xv 1

1-1 1-2 1-3 1-4 1-5 1-6 1-7 1-8

Antenna Radiation, 2 Gain, 3 Effective Area, 6 Path Loss, 6 Radar Range Equation and Cross Section, 7 Why Use an Antenna? 9 Directivity, 10 Directivity Estimates, 11 1-8.1 Pencil Beam, 11 1-8.2 Butterﬂy or Omnidirectional Pattern, 13 1-9 Beam Efﬁciency, 16 1-10 Input-Impedance Mismatch Loss, 17 1-11 Polarization, 18 1-11.1 Circular Polarization Components, 19 1-11.2 Huygens Source Polarization, 21 1-11.3 Relations Between Bases, 22 1-11.4 Antenna Polarization Response, 23 1-11.5 Phase Response of Rotating Antennas, 25 1-11.6 Partial Gain, 26 1-11.7 Measurement of Circular Polarization Using Amplitude Only, 26 1-12 Vector Effective Height, 27 1-13 Antenna Factor, 29 1-14 Mutual Coupling Between Antennas, 29 1.15 Antenna Noise Temperature, 30 vii

viii

CONTENTS

1-16 Communication Link Budget and Radar Range, 35 1-17 Multipath, 36 1-18 Propagation Over Soil, 37 1-19 Multipath Fading, 39 References, 40 2 Radiation Structures and Numerical Methods 2-1 Auxiliary Vector Potentials, 43 2-1.1 Radiation from Electric Currents, 44 2-1.2 Radiation from Magnetic Currents, 49 2-2 Apertures: Huygens Source Approximation, 51 2-2.1 Near- and Far-Field Regions, 55 2-2.2 Huygens Source, 57 2-3 Boundary Conditions, 57 2-4 Physical Optics, 59 2-4.1 Radiated Fields Given Currents, 59 2-4.2 Applying Physical Optics, 60 2-4.3 Equivalent Currents, 65 2-4.4 Reactance Theorem and Mutual Coupling, 66 2-5 Method of Moments, 67 2-5.1 Use of the Reactance Theorem for the Method of Moments, 68 2-5.2 General Moments Method Approach, 69 2-5.3 Thin-Wire Moment Method Codes, 71 2-5.4 Surface and Volume Moment Method Codes, 71 2-5.5 Examples of Moment Method Models, 72 2-6 Finite-Difference Time-Domain Method, 76 2-6.1 Implementation, 76 2-6.2 Central Difference Derivative, 77 2-6.3 Finite-Difference Maxwell’s Equations, 77 2-6.4 Time Step for Stability, 79 2-6.5 Numerical Dispersion and Stability, 80 2-6.6 Computer Storage and Execution Times, 80 2-6.7 Excitation, 81 2-6.8 Waveguide Horn Example, 83 2-7 Ray Optics and the Geometric Theory of Diffraction, 84 2-7.1 Fermat’s Principle, 85 2-7.2 H -Plane Pattern of a Dipole Located Over a Finite Strip, 85 2-7.3 E-Plane Pattern of a Rectangular Horn, 87 2-7.4 H -Plane Pattern of a Rectangular Horn, 89 2-7.5 Amplitude Variations Along a Ray, 90 2-7.6 Extra Phase Shift Through Caustics, 93 2-7.7 Snell’s Laws and Reﬂection, 93 2-7.8 Polarization Effects in Reﬂections, 94 2-7.9 Reﬂection from a Curved Surface, 94 2-7.10 Ray Tracing, 96

42

CONTENTS

2-7.11 2-7.12 2-7.13 2-7.14 2-7.15 References, 100

ix

Edge Diffraction, 96 Slope Diffraction, 98 Corner Diffraction, 99 Equivalent Currents, 99 Diffraction from Curved Surfaces, 99

3 Arrays

102

3-1 Two-Element Array, 104 3-2 Linear Array of N Elements, 109 3-3 Hansen and Woodyard End-Fire Array, 114 3-4 Phased Arrays, 115 3-5 Grating Lobes, 117 3-6 Multiple Beams, 118 3-7 Planar Array, 120 3-8 Grating Lobes in Planar Arrays, 125 3-9 Mutual Impedance, 127 3-10 Scan Blindness and Array Element Pattern, 127 3-11 Compensating Array Feeding for Mutual Coupling, 128 3-12 Array Gain, 129 3-13 Arrays Using Arbitrarily Oriented Elements, 133 References, 135 4 Aperture Distributions and Array Synthesis 4-1 Amplitude Taper and Phase Error Efﬁciencies, 137 4-1.1 Separable Rectangular Aperture Distributions, 139 4-1.2 Circularly Symmetrical Distributions, 140 4-2 Simple Linear Distributions, 140 4-3 Taylor One-Parameter Linear Distribution, 144 4-4 Taylor n Line Distribution, 147 4-5 Taylor Line Distribution with Edge Nulls, 152 4-6 Elliott’s Method for Modiﬁed Taylor Distribution and Arbitrary Sidelobes, 155 4-7 Bayliss Line-Source Distribution, 158 4-8 Woodward Line-Source Synthesis, 162 4-9 Schelkunoff’s Unit-Circle Method, 164 4-10 Dolph–Chebyshev Linear Array, 170 4-11 Villeneuve Array Synthesis, 172 4-12 Zero Sampling of Continuous Distributions, 173 4-13 Fourier Series Shaped-Beam Array Synthesis, 175 4-14 Orchard Method of Array Synthesis, 178 4-15 Series-Fed Array and Traveling-Wave Feed Synthesis, 188 4-16 Circular Apertures, 191 4-17 Circular Gaussian Distribution, 194 4-18 Hansen Single-Parameter Circular Distribution, 195 4-19 Taylor Circular-Aperture Distribution, 196 4-20 Bayliss Circular-Aperture Distribution, 200

136

x

CONTENTS

4-21 4-22 4-23 4-24 4-25

Planar Arrays, 202 Convolution Technique for Planar Arrays, 203 Aperture Blockage, 208 Quadratic Phase Error, 211 Beam Efﬁciency of Circular Apertures with Axisymmetric Distribution, 214 References, 215 5 Dipoles, Slots, and Loops 5-1 5-2 5-3 5-4 5-5 5-6 5-7 5-8 5-9 5-10 5-11 5-12 5-13 5-14 5-15

5-16 5-17 5-18 5-19 5-20 5-21 5-22 5-23 5-24 5-25 5-26

Standing-Wave Currents, 218 Radiation Resistance (Conductance), 220 Babinet–Booker Principle, 222 Dipoles Located Over a Ground Plane, 223 Dipole Mounted Over Finite Ground Planes, 225 Crossed Dipoles for Circular Polarization, 231 Super Turnstile or Batwing Antenna, 234 Corner Reﬂector, 237 Monopole, 242 Sleeve Antenna, 242 Cavity-Mounted Dipole Antenna, 245 Folded Dipole, 247 Shunt Feeding, 248 Discone Antenna, 249 Baluns, 251 5-15.1 Folded Balun, 252 5-15.2 Sleeve or Bazooka Baluns, 253 5-15.3 Split Coax Balun, 255 5-15.4 Half-Wavelength Balun, 256 5-15.5 Candelabra Balun, 256 5-15.6 Ferrite Core Baluns, 256 5-15.7 Ferrite Candelabra Balun, 258 5-15.8 Transformer Balun, 258 5-15.9 Split Tapered Coax Balun, 259 5-15.10 Natural Balun, 260 Small Loop, 260 Alford Loop, 261 Resonant Loop, 263 Quadriﬁlar Helix, 264 Cavity-Backed Slots, 266 Stripline Series Slots, 266 Shallow-Cavity Crossed-Slot Antenna, 269 Waveguide-Fed Slots, 270 Rectangular-Waveguide Wall Slots, 271 Circular-Waveguide Slots, 276 Waveguide Slot Arrays, 278 5-26.1 Nonresonant Array, 279 5-26.2 Resonant Array, 282

217

CONTENTS

xi

5-26.3 Improved Design Methods, 282 References, 283 6 Microstrip Antennas

285

6-1 Microstrip Antenna Patterns, 287 6-2 Microstrip Patch Bandwidth and Surface-Wave Efﬁciency, 293 6-3 Rectangular Microstrip Patch Antenna, 299 6-4 Quarter-Wave Patch Antenna, 310 6-5 Circular Microstrip Patch, 313 6-6 Circularly Polarized Patch Antennas, 316 6-7 Compact Patches, 319 6-8 Directly Fed Stacked Patches, 323 6-9 Aperture-Coupled Stacked Patches, 325 6-10 Patch Antenna Feed Networks, 327 6-11 Series-Fed Array, 329 6-12 Microstrip Dipole, 330 6-13 Microstrip Franklin Array, 332 6-14 Microstrip Antenna Mechanical Properties, 333 References, 334 7 Horn Antennas

336

7-1 Rectangular Horn (Pyramidal), 337 7-1.1 Beamwidth, 341 7-1.2 Optimum Rectangular Horn, 343 7-1.3 Designing to Given Beamwidths, 346 7-1.4 Phase Center, 347 7-2 Circular-Aperture Horn, 348 7-2.1 Beamwidth, 350 7-2.2 Phase Center, 352 7-3 Circular (Conical) Corrugated Horn, 353 7-3.1 Scalar Horn, 357 7-3.2 Corrugation Design, 357 7-3.3 Choke Horns, 358 7-3.4 Rectangular Corrugated Horns, 359 7-4 Corrugated Ground Plane, 359 7-5 Gaussian Beam, 362 7-6 Ridged Waveguide Horns, 365 7-7 Box Horn, 372 7-8 T-Bar-Fed Slot Antenna, 374 7-9 Multimode Circular Horn, 376 7-10 Biconical Horn, 376 References, 378 8 Reﬂector Antennas 8-1 Paraboloidal Reﬂector Geometry, 381 8-2 Paraboloidal Reﬂector Aperture Distribution Losses, 383

380

xii

CONTENTS

8-3 8-4 8-5 8-6 8-7 8-8 8-9 8-10 8-11 8-12 8-13

Approximate Spillover and Amplitude Taper Trade-offs, 385 Phase Error Losses and Axial Defocusing, 387 Astigmatism, 389 Feed Scanning, 390 Random Phase Errors, 393 Focal Plane Fields, 396 Feed Mismatch Due to the Reﬂector, 397 Front-to-Back Ratio, 399 Offset-Fed Reﬂector, 399 Reﬂections from Conic Sections, 405 Dual-Reﬂector Antennas, 408 8-13.1 Feed Blockage, 410 8-13.2 Diffraction Loss, 413 8-13.3 Cassegrain Tolerances, 414 8-14 Feed and Subreﬂector Support Strut Radiation, 416 8-15 Gain/Noise Temperature of a Dual Reﬂector, 421 8-16 Displaced-Axis Dual Reﬂector, 421 8-17 Offset-Fed Dual Reﬂector, 424 8-18 Horn Reﬂector and Dragonian Dual Reﬂector, 427 8-19 Spherical Reﬂector, 429 8-20 Shaped Reﬂectors, 432 8-20.1 Cylindrical Reﬂector Synthesis, 433 8-20.2 Circularly Symmetrical Reﬂector Synthesis, 434 8-20.3 Doubly Curved Reﬂector for Shaped Beams, 437 8-20.4 Dual Shaped Reﬂectors, 439 8-21 Optimization Synthesis of Shaped and Multiple-Beam Reﬂectors, 442 References, 443 9 Lens Antennas

447

9-1 9-2 9-3 9-4 9-5 9-6 9-7 9-8

Single Refracting Surface Lenses, 448 Zoned Lenses, 451 General Two-Surface Lenses, 454 Single-Surface or Contact Lenses, 459 Metal Plate Lenses, 461 Surface Mismatch and Dielectric Losses, 463 Feed Scanning of a Hyperboloidal Lens, 464 Dual-Surface Lenses, 465 9-8.1 Coma-Free Axisymmetric Dielectric Lens, 466 9-8.2 Speciﬁed Aperture Distribution Axisymmetric Dielectric Lens, 468 9-9 Bootlace Lens, 470 9-10 Luneburg Lens, 472 References, 472 10

Traveling-Wave Antennas 10-1 General Traveling Waves, 475

474

CONTENTS

xiii

10-1.1 Slow Wave, 478 10-1.2 Fast Waves (Leaky Wave Structure), 480 10-2 Long Wire Antennas, 481 10-2.1 Beverage Antenna, 481 10-2.2 V Antenna, 482 10-2.3 Rhombic Antenna, 483 10-3 Yagi–Uda Antennas, 485 10-3.1 Multiple-Feed Yagi–Uda Antennas, 492 10-3.2 Resonant Loop Yagi–Uda Antennas, 495 10-4 Corrugated Rod (Cigar) Antenna, 497 10-5 Dielectric Rod (Polyrod) Antenna, 499 10-6 Helical Wire Antenna, 502 10-6.1 Helical Modes, 503 10-6.2 Axial Mode, 504 10-6.3 Feed of a Helical Antenna, 506 10-6.4 Long Helical Antenna, 507 10-6.5 Short Helical Antenna, 508 10-7 Short Backﬁre Antenna, 509 10-8 Tapered Slot Antennas, 512 10-9 Leaky Wave Structures, 516 References, 518 11 Frequency-Independent Antennas Spiral 11-1 11-2 11-3 11-4 11-5

Antennas, 522 Modal Expansion of Antenna Patterns, 524 Archimedean Spiral, 526 Equiangular Spiral, 527 Pattern Analysis of Spiral Antennas, 530 Spiral Construction and Feeding, 535 11-5.1 Spiral Construction, 535 11-5.2 Balun Feed, 536 11-5.3 Inﬁnite Balun, 538 11-5.4 Beamformer and Coaxial Line Feed, 538 11-6 Spiral and Beamformer Measurements, 538 11-7 Feed Network and Antenna Interaction, 540 11-8 Modulated Arm Width Spiral, 541 11-9 Conical Log Spiral Antenna, 543 11-10 Mode 2 Conical Log Spiral Antenna, 549 11-11 Feeding Conical Log Spirals, 550 Log-Periodic Antennas, 550 11-12 Log-Periodic Dipole Antenna, 551 11-12.1 Feeding a Log-Periodic Dipole Antenna, 556 11-12.2 Phase Center, 558 11-12.3 Elevation Angle, 559 11-12.4 Arrays of Log-Periodic Dipole Antennas, 560 11-13 Other Log-Periodic Types, 561 11-14 Log-Periodic Antenna Feeding Paraboloidal Reﬂector, 563

521

xiv

CONTENTS

11-15 V Log-Periodic Array, 567 11-16 Cavity-Backed Planar Log-Periodic Antennas, 569 References, 571 12

Phased Arrays

573

12-1 12-2 12-3 12-4

Fixed Phase Shifters (Phasers), 574 Quantization Lobes, 578 Array Errors, 580 Nonuniform and Random Element Existence Arrays, 582 12-4.1 Linear Space Tapered Array, 582 12-4.2 Circular Space Tapered Array, 584 12-4.3 Statistically Thinned Array, 587 12-5 Array Element Pattern, 588 12-6 Feed Networks, 590 12-6.1 Corporate Feed, 590 12-6.2 Series Feed, 592 12-6.3 Variable Power Divider and Phase Shifter, 592 12-6.4 Butler Matrix, 594 12-6.5 Space Feeding, 596 12-6.6 Tapered Feed Network with Uniform-Amplitude Subarrays, 597 12-7 Pattern Null Formation in Arbitrary Array, 599 12-8 Phased Array Application to Communication Systems, 601 12-9 Near-Field Measurements on Phased Arrays, 602 References, 604 Index

607

PREFACE

I wrote this book from my perspective as a designer in industry, primarily for other designers and users of antennas. On occasion I have prepared and taught antenna courses, for which I developed a systematic approach to the subject. For the last decade I have edited the “Antenna Designer’s Notebook” column in the IEEE antenna magazine. This expanded edition adds a combination of my own design notebook and the many other ideas provided to me by others, leading to this collection of ideas that I think designers should know. The book contains a systematic approach to the subject. Every author would like to be read from front to back, but my own career assignments would have caused to me to jump around in this book. Nevertheless, Chapter 1 covers those topics that every user and designer should know. Because I deal with complete antenna design, which includes mounting the antenna, included are the effects of nearby structures and how they can be used to enhance the response. We all study ideal antennas ﬂoating in free space to help us understand the basics, but the real world is a little different. Instead of drawing single line graphs of common relationships between two parameters, I generated scales for calculations that I perform over and over. I did not supply a set of computer programs because I seldom use collections supplied by others, and younger engineers ﬁnd my programs quaint, as each generation learns a different computer language. You’ll learn by writing your own. IEEE Antennas and Propagation Society’s digital archive of all material published from 1952 to 2000 has changed our approach to research. I have not included extensive bibliographies, because I believe that it is no longer necessary. The search engine of the archive can supply an exhaustive list. I referred only to papers that I found particularly helpful. Complete sets of the transactions are available in libraries, whereas the wealth of information in the archive from conferences was not. I have started mining this information, which contains many useful design ideas, and have incorporated some of them in this book. In this ﬁeld, 40-year-old publications are still useful and we should not reinvent methods. Many clever ideas from industry are usually published xv

xvi

PREFACE

only once, if at all, and personally, I’ll be returning to this material again and again, because all books have limited space. As with the ﬁrst edition, I enjoyed writing this book because I wanted to express my point of view of a rewarding ﬁeld. Although the amount of information available is overwhelming and the mathematics describing it can cloud the ideas, I hope my explanations help you develop new products or use old ones. I would like to thank all the authors who taught me this subject by sharing their ideas, especially those working in industry. On a personal note I thank the designers at Lockheed Martin, who encouraged me and reviewed material while I wrote: in particular, Jeannette McDonnell, Thomas Cencich, Donald Huebner, and Julie Huffman. THOMAS A. MILLIGAN

1 PROPERTIES OF ANTENNAS

One approach to an antenna book starts with a discussion of how antennas radiate. Beginning with Maxwell’s equations, we derive electromagnetic waves. After that lengthy discussion, which contains a lot of mathematics, we discuss how these waves excite currents on conductors. The second half of the story is that currents radiate and produce electromagnetic waves. You may already have studied that subject, or if you wish to further your background, consult books on electromagnetics. The study of electromagnetics gives insight into the mathematics describing antenna radiation and provides the rigor to prevent mistakes. We skip the discussion of those equations and move directly to practical aspects. It is important to realize that antennas radiate from currents. Design consists of controlling currents to produce the desired radiation distribution, called its pattern. In many situations the problem is how to prevent radiation from currents, such as in circuits. Whenever a current becomes separated in distance from its return current, it radiates. Simply stated, we design to keep the two currents close together, to reduce radiation. Some discussions will ignore the current distribution and instead, consider derived quantities, such as ﬁelds in an aperture or magnetic currents in a slot or around the edges of a microstrip patch. You will discover that we use any concept that provides insight or simpliﬁes the mathematics. An antenna converts bound circuit ﬁelds into propagating electromagnetic waves and, by reciprocity, collects power from passing electromagnetic waves. Maxwell’s equations predict that any time-varying electric or magnetic ﬁeld produces the opposite ﬁeld and forms an electromagnetic wave. The wave has its two ﬁelds oriented orthogonally, and it propagates in the direction normal to the plane deﬁned by the perpendicular electric and magnetic ﬁelds. The electric ﬁeld, the magnetic ﬁeld, and the direction of propagation form a right-handed coordinate system. The propagating wave ﬁeld intensity decreases by 1/R away from the source, whereas a static ﬁeld Modern Antenna Design, Second Edition, By Thomas A. Milligan Copyright 2005 John Wiley & Sons, Inc.

1

2

PROPERTIES OF ANTENNAS

drops off by 1/R 2 . Any circuit with time-varying ﬁelds has the capability of radiating to some extent. We consider only time-harmonic ﬁelds and use phasor notation with time dependence ej ωt . An outward-propagating wave is given by e−j (kR−ωt) , where k, the wave number, is given by 2π/λ. λ is the wavelength of the wave given by c/f , where c is the velocity of light (3 × 108 m/s in free space) and f is the frequency. Increasing the distance from the source decreases the phase of the wave. Consider a two-wire transmission line with ﬁelds bound to it. The currents on a single wire will radiate, but as long as the ground return path is near, its radiation will nearly cancel the other line’s radiation because the two are 180◦ out of phase and the waves travel about the same distance. As the lines become farther and farther apart, in terms of wavelengths, the ﬁelds produced by the two currents will no longer cancel in all directions. In some directions the phase delay is different for radiation from the current on each line, and power escapes from the line. We keep circuits from radiating by providing close ground returns. Hence, high-speed logic requires ground planes to reduce radiation and its unwanted crosstalk.

1-1 ANTENNA RADIATION Antennas radiate spherical waves that propagate in the radial direction for a coordinate system centered on the antenna. At large distances, spherical waves can be approximated by plane waves. Plane waves are useful because they simplify the problem. They are not physical, however, because they require inﬁnite power. The Poynting vector describes both the direction of propagation and the power density of the electromagnetic wave. It is found from the vector cross product of the electric and magnetic ﬁelds and is denoted S: S = E × H∗

W/m2

Root mean square (RMS) values are used to express the magnitude of the ﬁelds. H∗ is the complex conjugate of the magnetic ﬁeld phasor. The magnetic ﬁeld is proportional to the electric ﬁeld in the far ﬁeld. The constant of proportion is η, the impedance of free space (η = 376.73 ): |S| = S =

|E|2 η

W/m2

(1-1)

Because the Poynting vector is the vector product of the two ﬁelds, it is orthogonal to both ﬁelds and the triplet deﬁnes a right-handed coordinate system: (E, H, S). Consider a pair of concentric spheres centered on the antenna. The ﬁelds around the antenna decrease as 1/R, 1/R 2 , 1/R 3 , and so on. Constant-order terms would require that the power radiated grow with distance and power would not be conserved. For ﬁeld terms proportional to 1/R 2 , 1/R 3 , and higher, the power density decreases with distance faster than the area increases. The energy on the inner sphere is larger than that on the outer sphere. The energies are not radiated but are instead concentrated around the antenna; they are near-ﬁeld terms. Only the 1/R 2 term of the Poynting vector (1/R ﬁeld terms) represents radiated power because the sphere area grows as R 2 and

GAIN

3

gives a constant product. All the radiated power ﬂowing through the inner sphere will propagate to the outer sphere. The sign of the input reactance depends on the near-ﬁeld predominance of ﬁeld type: electric (capacitive) or magnetic (inductive). At resonance (zero reactance) the stored energies due to the near ﬁelds are equal. Increasing the stored ﬁelds increases the circuit Q and narrows the impedance bandwidth. Far from the antenna we consider only the radiated ﬁelds and power density. The power ﬂow is the same through concentric spheres: 4πR12 S1,avg = 4πR22 S2,avg The average power density is proportional to 1/R 2 . Consider differential areas on the two spheres at the same coordinate angles. The antenna radiates only in the radial direction; therefore, no power may travel in the θ or φ direction. Power travels in ﬂux tubes between areas, and it follows that not only the average Poynting vector but also every part of the power density is proportional to 1/R 2 : S1 R12 sin θ dθ dφ = S2 R22 sin θ dθ dφ Since in a radiated wave S is proportional to 1/R 2 , E is proportional to 1/R. It is convenient to deﬁne radiation intensity to remove the 1/R 2 dependence: U (θ, φ) = S(R, θ, φ)R 2

W/solid angle

Radiation intensity depends only on the direction of radiation and remains the same at all distances. A probe antenna measures the relative radiation intensity (pattern) by moving in a circle (constant R) around the antenna. Often, of course, the antenna rotates and the probe is stationary. Some patterns have established names. Patterns along constant angles of the spherical coordinates are called either conical (constant θ ) or great circle (constant φ). The great circle cuts when φ = 0◦ or φ = 90◦ are the principal plane patterns. Other named cuts are also used, but their names depend on the particular measurement positioner, and it is necessary to annotate these patterns carefully to avoid confusion between people measuring patterns on different positioners. Patterns are measured by using three scales: (1) linear (power), (2) square root (ﬁeld intensity), and (3) decibels (dB). The dB scale is used the most because it reveals more of the low-level responses (sidelobes). Figure 1-1 demonstrates many characteristics of patterns. The half-power beamwidth is sometimes called just the beamwidth. The tenth-power and null beamwidths are used in some applications. This pattern comes from a parabolic reﬂector whose feed is moved off the axis. The vestigial lobe occurs when the ﬁrst sidelobe becomes joined to the main beam and forms a shoulder. For a feed located on the axis of the parabola, the ﬁrst sidelobes are equal.

1-2 GAIN Gain is a measure of the ability of the antenna to direct the input power into radiation in a particular direction and is measured at the peak radiation intensity. Consider the

4

PROPERTIES OF ANTENNAS

Angle 3 dB

Half-power beamwidth

Pattern level

10 dB

Tenth-power beamwidth

20 dB

Sidelobes

Vestigial lobe 30 dB

Null beamwidth

40 dB

FIGURE 1-1

Antenna pattern characteristics.

power density radiated by an isotropic antenna with input power P0 at a distance R: S = P0 /4πR 2 . An isotropic antenna radiates equally in all directions, and its radiated power density S is found by dividing the radiated power by the area of the sphere 4πR 2 . The isotropic radiator is considered to be 100% efﬁcient. The gain of an actual antenna increases the power density in the direction of the peak radiation: P0 G |E|2 1 P0 Gη S= = (1-2) or |E| = = Sη 4πR 2 η R 4π Gain is achieved by directing the radiation away from other parts of the radiation sphere. In general, gain is deﬁned as the gain-biased pattern of the antenna: P0 G(θ, φ) 4πR 2 P0 G(θ, φ) U (θ, φ) = 4π S(θ, φ) =

power density radiation intensity

(1-3)

The surface integral of the radiation intensity over the radiation sphere divided by the input power P0 is a measure of the relative power radiated by the antenna, or the antenna efﬁciency: 2π π Pr G(θ, φ) sin θ dθ dφ = ηe = efﬁciency P0 4π 0 0

GAIN

5

where Pr is the radiated power. Material losses in the antenna or reﬂected power due to poor impedance match reduce the radiated power. In this book, integrals in the equation above and those that follow express concepts more than operations we perform during design. Only for theoretical simpliﬁcations of the real world can we ﬁnd closed-form solutions that would call for actual integration. We solve most integrals by using numerical methods that involve breaking the integrand into small segments and performing a weighted sum. However, it is helpful that integrals using measured values reduce the random errors by averaging, which improves the result. In a system the transmitter output impedance or the receiver input impedance may not match the antenna input impedance. Peak gain occurs for a receiver impedance conjugate matched to the antenna, which means that the resistive parts are the same and the reactive parts are the same magnitude but have opposite signs. Precision gain measurements require a tuner between the antenna and receiver to conjugate-match the two. Alternatively, the mismatch loss must be removed by calculation after the measurement. Either the effect of mismatches is considered separately for a given system, or the antennas are measured into the system impedance and mismatch loss is considered to be part of the efﬁciency. Example Compute the peak power density at 10 km of an antenna with an input power of 3 W and a gain of 15 dB. First convert dB gain to a ratio: G = 1015/10 = 31.62. The power spreads over the sphere area with radius 10 km or an area of 4π(104 )2 m2 . The power density is S=

(3 W)(31.62) = 75.5 nW/m2 4π × 108 m2

We calculate the electric ﬁeld intensity using Eq. (1-2): |E| = Sη = (75.5 × 10−9 )(376.7) = 5333 µV/m Although gain is usually relative to an isotropic antenna, some antenna gains are referred to a λ/2 dipole with an isotropic gain of 2.14 dB. If we approximate the antenna as a point source, we compute the electric ﬁeld radiated by using Eq. (1-2): e−j kR P0 G(θ, φ)η E(θ, φ) = (1-4) R 4π This requires only that the antenna be small compared to the radial distance R. Equation (1-4) ignores the direction of the electric ﬁeld, which we deﬁne as polarization. The units of the electric ﬁeld are volts/meter. We determine the far-ﬁeld pattern by multiplying Eq. (1-4) by R and removing the phase term e−j kR since phase has meaning only when referred to another point in the far ﬁeld. The far-ﬁeld electric ﬁeld Eff unit is volts: 2 P0 G(θ, φ)η 4π 1 Eff (θ, φ) (1-5) Eff (θ, φ) = or G(θ, φ) = 4π P0 η During analysis, we often normalize input power to √ 1 W and can compute gain easily from the electric ﬁeld by multiplying by a constant 4π/η = 0.1826374.

6

PROPERTIES OF ANTENNAS

1-3 EFFECTIVE AREA Antennas capture power from passing waves and deliver some of it to the terminals. Given the power density of the incident wave and the effective area of the antenna, the power delivered to the terminals is the product Pd = SAeff

(1-6)

For an aperture antenna such as a horn, parabolic reﬂector, or ﬂat-plate array, effective area is physical area multiplied by aperture efﬁciency. In general, losses due to material, distribution, and mismatch reduce the ratio of the effective area to the physical area. Typical estimated aperture efﬁciency for a parabolic reﬂector is 55%. Even antennas with inﬁnitesimal physical areas, such as dipoles, have effective areas because they remove power from passing waves. 1-4 PATH LOSS [1, p. 183] We combine the gain of the transmitting antenna with the effective area of the receiving antenna to determine delivered power and path loss. The power density at the receiving antenna is given by Eq. (1-3), and the received power is given by Eq. (1-6). By combining the two, we obtain the path loss: Pd A2 G1 (θ, φ) = Pt 4πR 2 Antenna 1 transmits, and antenna 2 receives. If the materials in the antennas are linear and isotropic, the transmitting and receiving patterns are identical (reciprocal) [2, p. 116]. When we consider antenna 2 as the transmitting antenna and antenna 1 as the receiving antenna, the path loss is Pd A1 G2 (θ, φ) = Pt 4πR 2 Since the responses are reciprocal, the path losses are equal and we can gather and eliminate terms: G1 G2 = = constant A1 A2 Because the antennas were arbitrary, this quotient must equal a constant. This constant was found by considering the radiation between two large apertures [3]: 4π G = 2 A λ

(1-7)

We substitute this equation into path loss to express it in terms of the gains or effective areas: Pd A1 A2 λ 2 = G1 G2 = 2 2 (1-8) Pt 4πR λR We make quick evaluations of path loss for various units of distance R and for frequency f in megahertz using the formula path loss(dB) = KU + 20 log(f R) − G1 (dB) − G2 (dB)

(1-9)

RADAR RANGE EQUATION AND CROSS SECTION

7

where KU depends on the length units: KU

Unit

km 32.45 nm 37.80 miles 36.58 m −27.55 ft −37.87 Example Compute the gain of a 3-m-diameter parabolic reﬂector at 4 GHz assuming 55% aperture efﬁciency. Gain is related to effective area by Eq. (1-7): G=

4πA λ2

We calculate the area of a circular aperture by A = π(D/2)2 . By combining these equations, we have πD 2 πDf 2 G= ηa = ηa (1-10) λ c where D is the diameter and ηa is the aperture efﬁciency. On substituting the values above, we obtain the gain:

3π(4 × 109 ) G= 0.3 × 109

2 (0.55) = 8685

(39.4 dB)

Example Calculate the path loss of a 50-km communication link at 2.2 GHz using a transmitter antenna with a gain of 25 dB and a receiver antenna with a gain of 20 dB. Path loss = 32.45 + 20 log[2200(50)] − 25 − 20 = 88.3 dB What happens to transmission between two apertures as the frequency is increased? If we assume that the effective area remains constant, as in a parabolic reﬂector, the transmission increases as the square of frequency: Pd A1 A2 1 A1 A2 f 2 = = = Bf 2 Pt R 2 λ2 R2 c where B is a constant for a ﬁxed range. The receiving aperture captures the same power regardless of frequency, but the gain of the transmitting antenna increases as the square of frequency. Hence, the received power also increases as frequency squared. Only for antennas, whose gain is a ﬁxed value when frequency changes, does the path loss increase as the square of frequency. 1-5 RADAR RANGE EQUATION AND CROSS SECTION Radar operates using a double path loss. The radar transmitting antenna radiates a ﬁeld that illuminates a target. These incident ﬁelds excite surface currents that also radiate

8

PROPERTIES OF ANTENNAS

to produce a second ﬁeld. These ﬁelds propagate to the receiving antenna, where they are collected. Most radars use the same antenna both to transmit the ﬁeld and to collect the signal returned, called a monostatic system, whereas we use separate antennas for bistatic radar. The receiving system cannot be detected in a bistatic system because it does not transmit and has greater survivability in a military application. We determine the power density illuminating the target at a range RT by using Eq. (1-2): PT GT (θ, φ) Sinc = (1-11) 4πRT2 The target’s radar cross section (RCS), the scattering area of the object, is expressed in square meters or dBm2 : 10 log(square meters). The RCS depends on both the incident and reﬂected wave directions. We multiply the power collected by the target with its receiving pattern by the gain of the effective antenna due to the currents induced: RCS = σ =

powerreﬂected Ps (θr , φr , θi , φi ) = power density incident PT GT /4πRT2

(1-12)

In a communication system we call Ps the equivalent isotropic radiated power (EIRP), which equals the product of the input power and the antenna gain. The target becomes the transmitting source and we apply Eq. (1-2) to ﬁnd the power density at the receiving antenna at a range RR from the target. Finally, the receiving antenna collects the power density with an effective area AR . We combine these ideas to obtain the power delivered to the receiver: AR PT GT σ (θr , φr , θi , φi ) Prec = SR AR = (4πRT2 )(4πRR2 ) We apply Eq. (1-7) to eliminate the effective area of the receiving antenna and gather terms to determine the bistatic radar range equation: Prec GT GR λ2 σ (θr , φr , θi , φi ) = PT (4π)3 RT2 RR2

(1-13)

We reduce Eq. (1-13) and collect terms for monostatic radar, where the same antenna is used for both transmitting and receiving: Prec G2 λ2 σ = PT (4π)3 R 4 Radar received power is proportional to 1/R 4 and to G2 . We ﬁnd the approximate RCS of a ﬂat plate by considering the plate as an antenna with an effective area. Equation (1-11) gives the power density incident on the plate that collects this power over an area AR : PC =

PT GT (θ, φ) AR 4πRT2

The power scattered by the plate is the power collected, PC , times the gain of the plate as an antenna, GP : Ps = PC GP =

PT GT (θi , φi ) AR GP (θr , φr ) 4πRT2

WHY USE AN ANTENNA?

9

This scattered power is the effective radiated power in a particular direction, which in an antenna is the product of the input power and the gain in a particular direction. We calculate the plate gain by using the effective area and ﬁnd the scattered power in terms of area: PT GT 4πA2R Ps = 4πRT2 λ2 We determine the RCS σ by Eq. (1-12), the scattered power divided by the incident power density: σ =

Ps 4πA2R GR (θi , φi )GR (θr , φr )λ2 = = λ2 4π PT GT /4πRT2

(1-14)

The right expression of Eq. (1-14) divides the gain into two pieces for bistatic scattering, where the scattered direction is different from the incident direction. Monostatic scattering uses the same incident and reﬂected directions. We can substitute any object for the ﬂat plate and use the idea of an effective area and its associated antenna gain. An antenna is an object with a unique RCS characteristic because part of the power received will be delivered to the antenna terminals. If we provide a good impedance match to this signal, it will not reradiate and the RCS is reduced. When we illuminate an antenna from an arbitrary direction, some of the incident power density will be scattered by the structure and not delivered to the antenna terminals. This leads to the division of antenna RCS into the antenna mode of reradiated signals caused by terminal mismatch and the structural mode, the ﬁelds reﬂected off the structure for incident power density not delivered to the terminals. 1-6 WHY USE AN ANTENNA? We use antennas to transfer signals when no other way is possible, such as communication with a missile or over rugged mountain terrain. Cables are expensive and take a long time to install. Are there times when we would use antennas over level ground? The large path losses of antenna systems lead us to believe that cable runs are better. Example Suppose that we must choose between using a low-loss waveguide run and a pair of antennas at 3 GHz. Each antenna has 10 dB of gain. The low-loss waveguide has only 19.7 dB/km loss. Table 1-1 compares losses over various distances. The waveguide link starts out with lower loss, but the antenna system soon overtakes it. When the path length doubles, the cable link loss also doubles in decibels, but an antenna link TABLE 1-1 Losses Over Distance Distance (km)

Waveguide Loss (dB)

Antenna Path Loss (dB)

2 4 6 10

39.4 78.8 118.2 197

88 94 97.6 102

10

PROPERTIES OF ANTENNAS

increases by only 6 dB. As the distance is increased, radiating between two antennas eventually has lower losses than in any cable. Example A 200-m outside antenna range was set up to operate at 2 GHz using a 2-mdiameter reﬂector as a source. The receiver requires a sample of the transmitter signal to phase-lock the local oscillator and signal at a 45-MHz difference. It was proposed to run an RG/U 115 cable through the power and control cable conduit, since the run was short. The cable loss was 36 dB per 100 m, giving a total cable loss of 72 dB. A 10-dB coupler was used on the transmitter to pick off the reference signal, so the total loss was 82 dB. Since the source transmitted 100 mW (20 dBm), the signal was −62 dBm at the receiver, sufﬁcient for phase lock. A second proposed method was to place a standard-gain horn (15 dB of gain) within the beam of the source on a small stand out of the way of the measurement and next to the receiver. If we assume that the source antenna had only 30% aperture efﬁciency, we compute gain from Eq. (1-10) (λ = 0.15 m): G=

2π 0.15

2 (0.3) = 526

(27.2 dB)

The path loss is found from Eq. (1-9) for a range of 0.2 km: 32.45 + 20 log[2000(0.2)] − 27.2 − 15 = 42.3 dB The power delivered out of the horn is 20 dBm − 42.3 dB = −22.3 dBm. A 20-dB attenuator must be put on the horn to prevent saturation of the receiver (−30 dBm). Even with a short run, it is sometimes better to transmit the signal between two antennas instead of using cables. 1-7 DIRECTIVITY Directivity is a measure of the concentration of radiation in the direction of the maximum: maximum radiation intensity Umax directivity = = (1-15) average radiation intensity U0 Directivity and gain differ only by the efﬁciency, but directivity is easily estimated from patterns. Gain—directivity times efﬁciency—must be measured. The average radiation intensity can be found from a surface integral over the radiation sphere of the radiation intensity divided by 4π, the area of the sphere in steradians: 2π π 1 average radiation intensity = U (θ, φ) sin θ dθ dφ = U0 (1-16) 4π 0 0 This is the radiated power divided by the area of a unit sphere. The radiation intensity U (θ, φ) separates into a sum of co- and cross-polarization components: U0 =

1 4π

2π 0

π 0

[UC (θ, φ) + U× (θ, φ)] sin θ dθ dφ

(1-17)

DIRECTIVITY ESTIMATES

11

Both co- and cross-polarization directivities can be deﬁned: directivityC =

UC,max U0

directivity× =

U×,max U0

(1-18)

Directivity can also be deﬁned for an arbitrary direction D(θ, φ) as radiation intensity divided by the average radiation intensity, but when the coordinate angles are not speciﬁed, we calculate directivity at Umax . 1-8 DIRECTIVITY ESTIMATES Because a ratio of radiation intensities is used to calculate directivity, the pattern may be referred to any convenient level. The most accurate estimate is based on measurements at equal angle increments over the entire radiation sphere. The average may be found from coarse measurements by using numerical integration, but the directivity measured is affected directly by whether the maximum is found. The directivity of antennas with well-behaved patterns can be estimated from one or two patterns. Either the integral over the pattern is approximated or the pattern is approximated with a function whose integral is found exactly. 1-8.1 Pencil Beam By estimating the integral, Kraus [4] devised a method for pencil beam patterns with its peak at θ = 0◦ . Given the half-power beamwidths of the principal plane patterns, the integral is approximately the product of the beamwidths. This idea comes from circuit theory, where the integral of a time pulse is approximately the pulse width (3-dB points) times the pulse peak: U0 = θ1 θ2 /4π, where θ1 and θ2 are the 3-dB beamwidths, in radians, of the principal plane patterns: directivity =

4π 41,253 (rad) = (deg) θ1 θ2 θ1 θ2

(1-19)

Example Estimate the directivity of an antenna with E- and H -plane (principal plane) pattern beamwidths of 24◦ and 36◦ . Directivity =

41,253 = 47.75 (16.8 dB) 24(36)

An analytical function, cos2N (θ/2), approximates a broad pattern centered on θ = 0◦ with a null at θ = 180◦ : U (θ ) = cos2N (θ/2)

or

E = cosN (θ/2)

The directivity of this pattern can be computed exactly. The characteristics of the approximation are related to the beamwidth at a speciﬁed level, Lvl(dB): beamwidth [Lvl(dB)] = 4 cos−1 (10−Lvl(dB)/20N ) N=

−Lvl(dB) 20 log[cos(beamwidthLvl(dB) /4)]

directivity = N + 1 (ratio)

(1-20a) (1-20b) (1-20c)

12

PROPERTIES OF ANTENNAS Directivity, dB

3-dB Beamwidth

SCALE 1-1 3-dB beamwidth and directivity relationship for cos2N (θ /2) pattern.

Directivity, dB

10-dB Beamwidth

SCALE 1-2 10-dB beamwidth and directivity relationship for cos2N (θ /2) pattern.

Scales 1-1 and 1-2, which give the relationship between beamwidth and directivity using Eq. (1-20), are useful for quick conversion between the two properties. You can use the two scales to estimate the 10-dB beamwidth given the 3-dB beamwidth. For example, an antenna with a 90◦ 3-dB beamwidth has a directivity of about 7.3 dB. You read from the lower scale that an antenna with 7.3-dB directivity has a 159.5◦ 10-dB beamwidth. Another simple way to determine the beamwidths at different pattern levels is the square-root factor approximation: BW[Lvl 2(dB)] Lvl 2(dB) = BW[Lvl 1(dB)] Lvl 1(dB) By this factor, beamwidth10 dB = 1.826 beamwidth3 dB ; an antenna with a 90◦ 3-dB beamwidth has a (1.826)90◦ = 164.3◦ 10-dB beamwidth. This pattern approximation requires equal principal plane beamwidths, but we use an elliptical approximation with unequal beamwidths: U (θ, φ) = cos2Ne (θ/2) cos2 φ + cos2Nh (θ/2) sin2 φ

(1-21)

where Ne and Nh are found from the principal plane beamwidths. We combine the directivities calculated in the principal planes by the simple formula directivity (ratio) =

2 · directivitye · directivityh directivitye + directivityh

(1-22)

Example Estimate the directivity of an antenna with E- and H -plane pattern beamwidths of 98◦ and 140◦ . From the scale we read a directivity of 6.6 dB in the E-plane and 4.37 dB in the H -plane. We convert these to ratios and apply Eq. (1-22): directivity (ratio) =

2(4.57)(2.74) = 3.426 or 4.57 + 2.74

10 log(3.426) = 5.35 dB

DIRECTIVITY ESTIMATES

13

Many analyses of paraboloidal reﬂectors use a feed pattern approximation limited to the front hemisphere with a zero pattern in the back hemisphere: U (θ ) = cos2N θ

or

E = cosN θ

◦

for θ ≤ π/2(90 )

The directivity of this pattern can be found exactly, and the characteristics of the approximation are beamwidth [Lvl(dB)] = 2 cos−1 (10−Lvl(dB)/20N ) −Lvl(dB) N= 20 log[cos(beamwidthLvl(dB) /2)] directivity = 2(2N + 1) (ratio)

(1-23a) (1-23b) (1-23c)

We use the elliptical model [Eq. (1-21)] with this approximate pattern and use Eq. (1-22) to estimate the directivity when the E- and H -plane beamwidths are different. 1-8.2 Butterﬂy or Omnidirectional Pattern Many antennas have nulls at θ = 0◦ with rotational symmetry about the z-axis (Figure 1-2). Neither of the directivity estimates above can be used with these patterns because they require the beam peak to be at θ = 0◦ . We generate this type of antenna pattern by using mode 2 log-periodic conical spirals, shaped reﬂectors, some higherorder-mode waveguide horns, biconical horns, and traveling-wave antennas. A formula similar to Kraus’s can be found if we assume that all the power is between the 3-dB beamwidth angles θ1 and θ2 : 1 θ2 cos θ1 − cos θ2 sin θ dθ = U0 = 2 θ1 2 Rotational symmetry eliminates integration over φ: directivity =

2 Umax = U0 cos θ1 − cos θ2

(1-24)

FIGURE 1-2 Omnidirectional antenna pattern with sidelobes scanned above the horizon.

14

PROPERTIES OF ANTENNAS Directivity, dB of Omnipattern at q = 90°

3-dB Beamwidth

SCALE 1-3 Relationship between 3-dB beamwidth of omnidirectional pattern and directivity.

Example A pattern with rotational symmetry has half-power points at 35◦ and 75◦ . Estimate the directivity. Directivity =

2 = 3.57 (5.5 dB) cos 35◦ − cos 75◦

If the pattern also has symmetry about the θ = 90◦ plane, the integral for the average radiation intensity has limits from 0 to π/2. Equation (1-24) reduces to directivity = 1/ cos θ1 . Example A rotationally symmetric pattern with a maximum at 90◦ has a 45◦ beamwidth. Estimate the directivity. θ1 = 90◦ − 45◦ /2 = 67.5◦ , so directivity =

1 = 2.61 cos 67.5◦

(4.2 dB)

The pattern can be approximated by the function U (θ ) = B sin2M (θ/2) cos2N (θ/2) but the directivity estimates found by integrating this function show only minor improvements over Eq. (1-24). Nevertheless, we can use the expression for analytical patterns. Given beam edges θL and θU at a level Lvl(dB), we ﬁnd the exponential factors. √ ln[cos(θU /2)] − ln[cos(θL /2)] AA = and T M2 = tan−1 AA ln[sin(θL /2)] − ln[sin(θU /2)] −|Lvl(dB)|/8.68589 N= AA{ln[sin(θL /2)] − ln(sin T M2 )} + ln(cos(θL /2)) − ln(cos T M2 ) M = AA(N ) A second pattern model of an omnidirectional pattern based on the pattern function with minor sidelobes and a beam peak at θ0 measured from the symmetry axis is sin[b(θ0 − θ )] b(θ0 − θ ) We estimate the directivity from the half-power beamwidth (HPBW) and the beam peak θ0 [5]: directivity(dB) = 10 log

101 (HPBW − 0.0027HPBW2 ) sin θ0

(1-25)

DIRECTIVITY ESTIMATES

15

Scan Factor, dB

Beam Direction, q

SCALE 1-4 pattern.

Additional directivity of omnidirectional pattern when scanned into conical

Scale 1-3 evaluates this formula for a beam at θ0 = 90◦ given HPBW, and Scale 1-4 gives the additional gain when the beam peak scans toward the axis. The directivity of butterﬂy patterns with unequal beamwidths in the principal planes cannot be estimated directly from the foregoing formulas. Similarly, some pencil beam patterns have large sidelobes which decrease the directivity and cannot be estimated accurately from Eq. (1-19). Both problems are solved by considering the directivity as an estimate of the average radiation intensity. Example A butterﬂy pattern peak is at 50◦ in both principal planes, but the beamwidths are 20◦ and 50◦ . Estimate the directivity. The 3-dB pattern points are given by: Cut 1 (40◦ and 60◦ ): U01 =

cos 40◦ − cos 60◦ = 0.133 2

U02 =

cos 25◦ − cos 75◦ = 0.324 2

Cut 2 (25◦ and 75◦ ):

Average the two pattern integral estimates: 0.133 + 0.324 = 0.228 2 1 Umax = = 4.38 (6.4 dB) directivity = U0 0.228 U0 =

Suppose that the beams are at different levels on the same pattern. For example, the lobe on the right of the ﬁrst pattern is the peak and the left lobe is reduced by 3 dB. The peaks of the second pattern are reduced by 1 dB. We can average on one pattern alone. Each lobe contributes Umax (cos θ1 − cos θ2 )/4 to the integral. The integral of the ﬁrst pattern is approximated by 0.266 + 0.266 × 10−3/10 = 0.100 4 The integral of the second pattern is reduced 1 dB from the peak. The average radiation intensity is found by averaging the two pattern averages: 0.100 + 0.324 × 10−0.1 = 0.178 2 1 = 5.602 (7.5 dB) directivity = U0 U0 =

16

PROPERTIES OF ANTENNAS

Pencil beam patterns with large sidelobes can be averaged in a similar manner: Up = 1/directivity. By using Eq. (1-19) and assuming equal beamwidths, we have Up = HPBW2 /41, 253, where Up is the portion of the integral due to the pencil beam and HPBW is the beamwidth in degrees. Example Consider a pencil beam antenna with pattern beamwidths of 50◦ and 70◦ in the principal planes. The second pattern has a sidelobe at θ = 60◦ down 5 dB from the peak and a 30◦ beamwidth below the 5 dB. What is the effect of the sidelobe on the directivity estimate? Without the sidelobe the directivity estimate is directivity =

41253 = 11.79 50(70)

(10.7 dB)

Consider each pattern separately: UP1 =

502 = 0.0606 41,253

UP2 =

702 = 0.1188 41,253

The sidelobe adds to the second integral: UPS2 =

(cos 45◦ − cos 75◦ )10−5/10 = 0.0354 4

Averaging the integrals of the parts gives us 0.1074: directivity =

1 = 9.31 (9.7 dB) U0

If there had been a sidelobe on each side, each would have added to the integral. Estimating integrals in this manner has limited value. Remember that these are only approximations. More accurate results can be obtained by digitizing the pattern and performing numerical integration on each pattern by using Eq. (1-16) or (1-17). 1-9 BEAM EFFICIENCY Radiometer system designs [6, p. 31–6] specify the antenna in terms of beam efﬁciency. For a pencil beam antenna with the boresight at θ = 0, the beam efﬁciency is the ratio (or percent) of the pattern power within a speciﬁed cone centered on the boresight to the total radiated power. In terms of the radiation intensity U ,

θ1

beam efﬁciency =

0

0

π

2π

U (θ, φ) sin θ dθ dφ (1-26)

0 2π

U (θ, φ) sin θ dθ dφ

0

where U includes both polarizations if necessary. Extended noise sources, such as radiometry targets, radiate noise into sidelobes of the antenna. Beam efﬁciency measures the probability of the detected target being located within the main beam (θ ≤ θ1 ).

INPUT-IMPEDANCE MISMATCH LOSS

17

Sometimes we can calculate directivity more easily than the pattern everywhere required by the denominator of Eq. (1-26): for example, a paraboloidal reﬂector. We use Eqs. (1-15) and (1-16) to calculate the denominator integral: π 2π 4πUmax U (θ, φ) sin θ dθ dφ = directivity 0 0 This reduces Eq. (1-26) to

θ1

2π

directivity beam efﬁciency =

0

U (θ, φ) sin θ dθ dφ

0

4πUmax

(1-27)

Equation (1-27) greatly reduces the pattern calculation requirements to compute beam efﬁciency when the directivity can be found without pattern evaluation over the entire radiation sphere. 1-10 INPUT-IMPEDANCE MISMATCH LOSS When we fail to match the impedance of an antenna to its input transmission line leading from the transmitter or to the receiver, the system degrades due to reﬂected power. The input impedance is measured with respect to some transmission line or source characteristic impedance. When the two are not the same, a voltage wave is reﬂected, ρV , where ρ is the voltage reﬂection coefﬁcient: ρ=

ZA − Z0 ZA + Z0

(1-28)

ZA is the antenna impedance and Z0 is the measurement characteristic impedance. On a transmission line the two traveling waves, incident and reﬂected, produce a standing wave: Vmax = (1 + |ρ|)Vi Vmin = (1 − |ρ|)Vi 1 + |ρ| Vmax = VSWR = Vmin 1 − |ρ|

(1-29) (1-30)

VSWR is the voltage standing-wave ratio. We use the magnitude of ρ, a complex phasor, since all the terms in Eq. (1-28) are complex numbers. The reﬂected power is given by Vi2 |ρ|2 /Z0 . The incident power is Vi2 /Z0 . The ratio of the reﬂected power to the incident power is |ρ|2 . It is the returned power ratio. Scale 1-5 gives the conversion between return loss and VSWR: return loss = −20 log |ρ| VSWR

Return Loss, dB

SCALE 1-5

Relationship between return loss and VSWR.

(1-31)

18

PROPERTIES OF ANTENNAS Reflected Power Loss, dB

Return Loss, dB

SCALE 1-6 Reﬂected power loss due to antenna impedance mismatch.

The power delivered to the antenna is the difference between the incident and the reﬂected power. Normalized, it is expressed as 1 − |ρ|2 or reﬂected power loss(dB) = 10 log(1 − |ρ|2 )

(1-32)

The source impedance to achieve maximum power transfer is the complex conjugate of the antenna impedance [7, p. 94]. Scale 1-6 computes the power loss due to antenna impedance mismatch. If we open-circuit the antenna terminals, the reﬂected voltage equals the incident voltage. The standing wave doubles the voltage along the transmission line compared to the voltage present when the antenna is loaded with a matched load. We consider the effective height of an antenna, the ratio of the open-circuit voltage to the input ﬁeld strength. The open-circuit voltage is twice that which appears across a matched load for a given received power. We can either think of this as a transmission line with a mismatch that doubled the incident voltage or as a Th´evenin equivalent circuit with an open-circuit voltage source that splits equally between the internal resistor and the load when it is matched to the internal resistor. Path loss analysis predicts the power delivered to a matched load. The mathematical Th´evenin equivalent circuit containing the internal resistor does not say that half the power received by the antenna is either absorbed or reradiated; it only predicts the circuit characteristics of the antenna load under all conditions. Possible impedance mismatch of the antenna requires that we derate the feed cables. The analysis above shows that the maximum voltage that occurs on the cable is twice that present when the cable impedance is matched to the antenna. We compute the maximum voltage given the VSWR using Eq. (1-29) for the maximum voltage: Vmax =

1-11

2Vi 2 VSWR(Vi ) = VSWR + 1 1 + 1/VSWR

(1-33)

POLARIZATION

The polarization of a wave is the direction of the electric ﬁeld. We handle all polarization problems by using vector operations on a two-dimensional space using the far-ﬁeld radial vector as the normal to the plane. This method is systematic and reduces chance of error. The spherical wave in the far ﬁeld has only θ and φ components of the electric ˆ Eθ and Eφ are phasor components in the direction of the unit ﬁeld: E = Eθ θˆ + Eφ φ. ˆ We can also express the direction of the electric ﬁeld in terms of vectors θˆ and φ. a plane wave propagating along the z-axis: E = Ex xˆ + Ey yˆ . The direction of propagation conﬁnes the electric ﬁeld to a plane. Polarization is concerned with methods

POLARIZATION

19

f (y)

wt

E

t q(x)

FIGURE 1-3

Polarization ellipse.

of describing this two-dimensional space. Both of the above are linear polarization expansions. We can rewrite them as ˆ E = Eθ (θˆ + ρˆL φ) E = Ex (ˆx + ρˆL yˆ )

Eφ Eθ Ey ρˆL = Ex ρˆL =

(1-34)

where ρˆL is the linear polarization ratio, a complex constant. If time is inserted into the expansions, and the tip of the electric ﬁeld traced in space over time, it appears as an ellipse with the electric ﬁeld rotating either clockwise (CW) or counter clockwise (CCW) (Figure 1-3). τ is the tilt of the polarization ellipse measured from the x-axis (φ = 0) and the angle of maximum response. The ratio of the maximum to minimum linearly polarized responses on the ellipse is the axial ratio. If ρˆL = e±j π/2 , the ellipse expands to a circle and gives the special case of circular polarization. The electric ﬁeld is constant in magnitude but rotates either CW (left hand) or CCW (right hand) at the rate ωt for propagation perpendicular to the page. 1-11.1 Circular Polarization Components The two circular polarizations also span the two-dimensional space of polarization. The right- and left-handed orthogonal unit vectors deﬁned in terms of linear components are ˆ = √1 (θˆ − j φ) ˆ R 2 1 ˆ Lˆ = √ (θˆ + j φ) 2

or or

ˆ = √1 (ˆx − j yˆ ) R 2 1 Lˆ = √ (ˆx + j yˆ ) 2

(1-35a) (1-35b)

The electric ﬁeld in the polarization plane can be expressed in terms of these new unit vectors: ˆ + ER R ˆ E = EL L

20

PROPERTIES OF ANTENNAS

When projecting a vector onto one of these unit vectors, it is necessary to use the complex conjugate in the scalar (dot) product: EL = E · Lˆ

∗

ˆ ER = E · R

∗

ˆ onto itself, we obtain When we project R ˆ ·R ˆ ∗ = 1 (θˆ − j φ) ˆ · (θˆ + j φ) ˆ = 1 (1 − j · j ) = 1 R 2 2 Similarly,

ˆ ∗ = 1 (θˆ + j φ) ˆ · (θˆ + j φ) ˆ = 1 (1 + j · j ) = 0 Lˆ · R 2 2

The right- and left-handed circular (RHC and LHC) components are orthonormal. A circular polarization ratio can be deﬁned from the equation ρˆc =

ˆ E = EL (Lˆ + ρˆc R)

ER = ρc ej δc EL

Let us look at a predominately left-handed circularly polarized wave when time and space combine to a phase of zero for EL . We draw the polarization as two circles (Figure 1-4). The circles rotate at the rate ωt in opposite directions (Figure 1-5), with the center of the right-handed circular polarization circle moving on the end of the vector of the left-handed circular polarization circle. We calculate the phase of the circular polarization ratio ρˆc from the complex ratio of the right- and left-handed circular components. Maximum and minimum electric ﬁelds occur when the circles

f (y)

ER

L +

E

R

E

EL

E

L −

E

R

ER

EL EL

q (x)

E

ER dc t

FIGURE 1-4 Polarization ellipse LHC and RHC components. (After J. S. Hollis, T. J. Lyons, and L. Clayton, Microwave Antenna Measurements, Scientiﬁc Atlanta, 1969, pp. 3–6. Adapted by permission.)

POLARIZATION

21

RHC f

LHC f Wave Propagating Out of Paper

E

wt q

E

wt

q

r^L = e −jp/2

r^L = e jp/2

FIGURE 1-5 Circular polarization components. (After J. S. Hollis, T. J. Lyons, and L. Clayton, Microwave Antenna Measurements, Scientiﬁc Atlanta, 1969, pp. 3–5. Adapted by permission.)

Axial Ratio, dB

Circular Cross-polarization, dB

SCALE 1-7 Circular cross-polarization/axial ratio.

alternately add and subtract as shown in Figure 1-4. Scale 1-7 shows the relationship between circular cross-polarization and axial ratio: √ √ Emax = (|EL | + |ER |) / 2 Emin = (|EL | − |ER |) / 2 1 + |ρˆc | |EL | + |ER | Emax = LHC = |EL | − |ER | 1 − |ρˆc | axial ratio = Emin (1-36) |ρˆc | + 1 |ER | + |EL | Emax = RHC = Emin |ER | − |EL | |ρˆc | − 1 |ρˆc | < 1 LHC 0 ≤ 1 < 1 RHC ρˆc axial ratio(dB) = 20 log

Emax Emin

The tilt angle of the polarization ellipse τ is one-half δc , the phase of ρˆc . Imagine time moving forward in Figure 1-5. When the LHC vector has rotated δc /2 CW, the RHC vector has rotated δc /2 CCW and the two align for a maximum. 1-11.2 Huygens Source Polarization When we project the currents induced on a paraboloidal reﬂector to an aperture plane, Huygens source radiation induces aligned currents that radiate zero cross-polarization

22

PROPERTIES OF ANTENNAS

in the principal planes. We separate feed antenna radiation into orthogonal Huygens sources for this case. To calculate the far-ﬁeld pattern of a paraboloid reﬂector, we can skip the step involving currents and integrate over the Huygens source ﬁelds in the aperture plane directly. We transform the measured ﬁelds of the feed into orthogonal Huygens sources by

Ec Eθf cos φ − sin φ = (1-37) sin φ cos φ Ex Eφf where Ec is the φ = 0 direction of polarization in the feed pattern and Ex is the φ = 90◦ polarization. This division corresponds to Ludwig’s third deﬁnition of crosspolarization [8]. The following matrix converts the Huygens source polarizations to the normal far-ﬁeld components of spherical coordinates:

Ec cos φ sin φ Eθ = (1-38) − sin φ cos φ Eφ Ex 1-11.3 Relations Between Bases In problems with antennas at arbitrary orientations, circularly polarized components have an advantage over linear components. When the coordinate system is rotated, both the amplitude and phase change for ρˆL , the linear polarization ratio, whereas the circular polarization ratio ρˆc magnitude is constant under rotations and only the phase changes. In other words, the ratio of the diameters of the circles (Figure 1-4) is constant. The circular components can be found from linear polarization components by projection. ˆ ∗ = √1 (Eθ θˆ + Eφ φ) ˆ ·R ˆ · (θˆ + j φ) ˆ ER = (Eθ θˆ + Eφ φ) 2 1 ER = √ (Eθ + j Eφ ) 2

(1-39)

Similarly, 1 EL = √ (Eθ − j Eφ ) 2 The linear polarizations can be found in terms of the circular components in the same manner: 1 j Eθ = √ (EL + ER ) Eφ = √ (EL − ER ) 2 2 These relations enable the conversion between polarizations. Good circularly polarized antennas over a wide bandwidth are difﬁcult to build, but good linearly polarized antennas are obtained easily. After we measure the phase and amplitude of Eθ and Eφ component phasors, we compute the circular components from Eq. (1-39), the axial ratio by using Eq. (1-36), and the polarization ellipse tilt τ from one-half the phase of ER /EL . We employ a leveled phase-locked source to record two patterns with orthogonal linear sources (or the same linear source is rotated between patterns). Afterward, we use the equations given above to convert polarization

23

POLARIZATION

to any desired polarization components. We calculate the maximum and minimum linear components by projecting the linear components into the rotated coordinate system of the polarization ellipse: Emax = Eθ cos τ + Eφ sin τ Emin = −Eθ sin τ + Eφ cos τ 1-11.4 Antenna Polarization Response The path loss formulas assume that the two antennas have matched polarizations. Polarization mismatch adds an extra loss. We determine polarization efﬁciency by applying the scalar (dot) product between normalized polarization vectors. An antenna transmitting in the z-direction has the linear components Ea = E1 (ˆx + ρˆL1 yˆ ) The incident wave on the antenna is given by Ei = E2 (ˆx + ρˆL2 yˆ ) where the wave is expressed in the coordinates of the source antenna. The z-axis of the source is in the direction opposite that of the antenna. It is necessary to rotate the coordinates of the receiving antenna wave. Rotating about the x-axis is equivalent to changing the sign of the tilt angle or taking the complex conjugate of Ea . The measurement antenna projects the incident wave polarization onto the antenna polarization. The antenna measures the incident ﬁeld, but we need to normalize the antenna polarization to a unit vector to calculate polarization efﬁciency: E2 · E∗1 =

∗ E2 E1∗ (1 + ρˆL2 ρˆL1 ) 2 1 + |ρˆL1 |

We normalize both the incident wave and antenna responses to determine loss due to polarization mismatch: xˆ + ρˆL2 yˆ Ei = ∗ |Ei | 1 + ρˆL2 ρˆL2

∗ xˆ + ρˆL1 E∗a yˆ = ∗ |Ea | 1 + ρˆL1 ρˆL1

The normalized voltage response is ∗ ρˆL2 Ei · E∗a 1 + ρˆL1 = ∗ ∗ |Ei ||Ea | 1 + ρˆL1 ρˆL1 1 + ρˆL2 ρˆL2

(1-40)

When we express it as a power response, we obtain the polarization efﬁciency : =

1 + |ρˆL1 |2 |ρˆL2 |2 + 2|ρˆL1 ||ρˆL2 | cos(δ1 − δ2 ) |Ei · E∗a |2 = |Ei |2 |Ea |2 (1 + |ρˆL1 |2 )(1 + |ρˆL2 |2 )

(1-41)

This is the loss due to polarization mismatch. Given that δ1 and δ2 are the phases of the polarization ratios of the antenna and the incident wave. As expressed in terms of linear polarization ratios, the formula is awkward because when the antenna is rotated to determine the peak response, both the amplitudes and phases change. A formula

24

PROPERTIES OF ANTENNAS

using circular polarization ratios would be more useful, because only phase changes under rotation. Two arbitrary polarizations are orthogonal ( = 0) only if |ρˆ1 | =

1 |ρˆ2 |

◦

and δ1 − δ2 = ±180

(1-42)

This can be expressed as vectors by using unit vectors: a1 · a∗2 = 0; a1 and a2 are the orthonormal generalized basis vectors for polarization. We can deﬁne polarization in terms of this basis with a polarization ratio ρ. By paralleling the analysis above for linear polarizations, we obtain the polarization efﬁciency for an arbitrary orthonormal polarization basis: =

1 + |ρˆ1 |2 |ρˆ2 |2 + 2|ρˆ1 ||ρˆ2 | cos(δ1 − δ2 ) (1 + |ρˆ1 |2 )(1 + |ρˆ2 |2 )

(1-43)

It has the same form as Eq. (1-41) derived for linear polarizations. We can use Eq. (1-43) with circular polarizations whose polarization ratio ρc magnitudes are constant with rotations of the antenna. The maximum and minimum polarization efﬁciencies occur when δ1 − δ2 equals 0◦ and 180◦ , respectively. The polarization efﬁciency becomes (1 ± |ρˆ1 ||ρˆ2 |)2 (1-44) max / min = (1 + |ρˆ1 |2 )(1 + |ρˆ2 |2 ) In all other vector pair bases for polarization, the magnitude of the polarization ratio ρ changes under rotations. Figure 1-6 expresses Eq. (1-44) as a nomograph. If we have ﬁxed installations, we can rotate one antenna until the maximum response is obtained and realize minimum polarization loss. In transmission between mobile antennas such as those mounted on missiles or satellites, the orientation cannot be controlled and the maximum polarization loss must be used in the link analysis. Circularly polarized antennas are used in these cases. Example A satellite telemetry antenna is RHC with an axial ratio of 7 dB. The ground station is RHC with a 1.5-dB axial ratio. Determine the polarization loss. Because the orientation of the satellite is unknown, we must use the maximum polarization loss. To ﬁnd it, use the RHC ends of the scales in Figure 1-6. Draw a line from 7 on the leftmost scale to 1.5 on the center scale. Read the loss on the scale between: 0.9 dB. The measured cross-polarization response of a linearly polarized antenna is the reciprocal of the axial ratio, the same absolute magnitude in decibels. Example Suppose that the linear cross-polarization responses of two antennas in a stationary link are given as 10 and 20 dB. Compute the minimum polarization loss. We rotate one of the antennas until the maximum response is found. The speciﬁcation of cross-polarization response does not state whether an antenna is predominately leftor right-handed circularly polarized. It must be one or the other. Suppose that the 20dB cross-polarization antenna is LHC. If the other antenna also is LHC, we use a line drawn from the lower portion of the center scale in Figure 1-6 to the rightmost LHC scale and read 0.2 dB of loss on the scale between the two. The second possibility is

POLARIZATION

25

FIGURE 1-6 Maximum and minimum polarization loss. (After A. C. Ludwig, A simple graph for determining polarization loss, Microwave Journal, vol. 19, no. 9, September 1976, p. 63.)

that the antenna could be predominately RHC. On drawing a line to the RHC (lower) scale, we read 0.7 dB on the center scale. When polarization is expressed in terms of linearly polarized components, it is ambiguous to give only magnitudes and no information of the circular polarization sense. 1-11.5 Phase Response of Rotating Antennas The polarization sense of an antenna can be determined from the phase slope of a rotating antenna. Before starting the phase measurement, determine that the setup is proper. Some older phase–amplitude receivers are ambiguous, depending on whether the local oscillator frequency was above or below the signal frequency. We use the convention that increased distance between antennas gives decreased phase. Move the antenna away from the source and observe decreasing phase or correct the setup. A rotating linearly polarized source ﬁeld is given by Es = E2 (cos α xˆ + sin α yˆ ) where α is clockwise rotation viewed from the direction of propagation (forward). A horizontally polarized linear antenna has the response Ea = E1 xˆ . It responds to the

26

PROPERTIES OF ANTENNAS

rotating linear source ﬁeld, E1 E2 cos α. The phase is constant under rotation until the null is passed and it ﬂips 180◦ through the null. An RHC polarized antenna has the response E1 (ˆx − j yˆ ). It responds to the rotating linear source ﬁeld, E1 E2 (cos α − j sin α) = E1 E2 e−j α The magnitude remains constant, but the phase decreases with rotation. Phase increases when the antenna is LHC. By observing the phase slope, the sense of the predominant polarization can be determined: RHC = negative phase slope; LHC = positive phase slope. It is easily remembered by considering the basis vectors of circular polarization: ˆ = √1 (ˆx − j yˆ ) R 2 In rotation from the x-axis to the y-axis, the phase decreases 90◦ . 1-11.6 Partial Gain If we measure the antenna gain to one polarization (e.g., RHC) and operate it in a link with an antenna also measured to one polarization, Eq. (1-44) fails to predict the response. Polarization efﬁciency assumes that the antenna gain was measured using a source ﬁeld with matched polarization. Gains referred to a single polarization are partial gains. If we align the two polarization ellipses of the two antennas, the response increases. Similarly, when the ellipses are crossed, the link suffers polarization loss. To obtain the full gain, we add the factor 10 log(1 + |ρ|2 )

(1-45)

to the partial gain, an expression valid using ρ for either circular or linear polarization. In terms of axial ratio A for circular polarization, the conversion is 2(1 + A2 ) 20 log 1+A When using measured partial gains for both antennas, the range of polarization efﬁciency is given by polarization efﬁciency = 20 log(1 ± ρ1 ρ2 )

(1-46)

We can convert Eq. (1-46) to expressions that use the axial ratio of the two antennas: 2(A1 A2 + 1) (A1 + 1)(A2 + 1) 2(A1 + A2 ) minimum polarization efﬁciency = 20 log (A1 + 1)(A2 + 1)

maximum polarization efﬁciency = 20 log

1-11.7 Measurement of Circular Polarization Using Amplitude Only The analyses given above assume that you can measure both amplitude and phase response of antennas, whereas in some cases only amplitude can be measured. If

VECTOR EFFECTIVE HEIGHT

27

you do not know the sense of circular polarization, it will be necessary to build two antennas that are identical except for their circular polarization sense. For example, you can build two identically sized counter-wound helical wire antennas. You determine polarization sense by using both sources and comparing measured levels. Once you establish the polarization sense, mount a linearly polarized measurement antenna with low cross-polarization. For a given pointing direction of the antenna under test, rotate the source antenna and record the maximum and minimum levels. The ratio of the maximum to the minimum is the axial ratio. To measure gain, rotate the measurement linearly polarized antenna to determine the peak response. Replace the antenna under test with a linearly polarized gain standard (horn) and perform a gain comparison measurement. Given the antenna axial ratio A, you adjust the linearly polarized gain by the correction factor: A+1 gain correction factor(dB) = 20 log √ 2A

(1-47)

We obtain the RHC and LHC response from 1 ER = √ (Emax + Emin ) 2

1 and EL = √ (Emax − Emin ) 2

assuming that the antenna is predominately RHC. 1-12 VECTOR EFFECTIVE HEIGHT The vector effective height relates the open-circuit voltage response of an antenna to the incident electric ﬁeld. Although we normally think of applying effective height to a line antenna, such as a transmitting tower, the concept can be applied to any antenna. For a transmitting tower, effective height is the physical height multiplied by the ratio of the average current to the peak current: VOC = Ei · h∗

(1-48)

The vector includes the polarization properties of the antenna. Remember from our discussion of antenna impedance mismatch that the open-circuit voltage√VOC is twice that across a matched load ZL for a given received power: VOC = 2 Prec ZL . The received power is the product of the incident power density S and the effective area of the antenna, Aeff . Gathering terms, we determine the open-circuit voltage from the incident ﬁeld strength E and a polarization efﬁciency : ZL Aeff VOC = 2E η We calculate polarization efﬁciency by using the scalar product between the normalized incident electric ﬁeld and the normalized vector effective height: =

|Ei ·h∗ |2 |Ei |2 |h|2

(1-49)

28

PROPERTIES OF ANTENNAS

Equation (1-49) is equivalent to Eq. (1-41) because both involve the scalar product between the incident wave and the receiving polarization, but the expressions have different normalizations. You can substitute vector effective height of the transmitting antenna for the incident wave in Eq. (1-49) and calculate polarization efﬁciency between two antennas. When an antenna rotates, we rotate h. We could describe polarization calculations in terms of vector effective height, which would parallel and repeat the discussion given in Section 1-11. We relate the magnitude of the effective height h to the effective area Aeff and the load impedance ZL : ZL Aeff (1-50) h=2 η The mutual impedance in the far ﬁeld between two antennas can be found from the vector effective heights of both antennas [9, p. 6–9]. Given the input current I1 to the ﬁrst antenna, we ﬁnd the open-circuit voltage of the second antenna: Z12 =

(V2 )OC j kηe−j kr h1 ·h∗2 = I1 4πr

(1-51)

When we substitute Eq. (1-50) into Eq. (1-51) and gather terms, we obtain a general expression for the normalized mutual impedance of an arbitrary pair of antennas given the gain of each in the direction of the other antenna as a function of spacing r: √ Z12 j G1 G2 −j kr h1 ·h∗2 = e (1-52) √ kr |h1 ||h2 | ZL1 ZL2 The magnitude of mutual impedance increases when the gain increases or the distance decreases. Of course, Eq. (1-52) is based on a far-ﬁeld equation and gives only an approximate answer, but it produces good results for dipoles spaced as close as 1λ. Figure 1-7 gives a plot of Eq. (1-52) for isotropic gain antennas with matched polarizations which shows the 1/R amplitude decrease with distance and that resistance and

Normalized Mutual Impedance

Reactance

Resistance

Spacing, l

FIGURE 1-7 Normalized mutual impedance (admittance) from the vector effective length for two antennas with 0 dB gain along the line between them.

MUTUAL COUPLING BETWEEN ANTENNAS

29

reactance curves are shifted out of phase. The cosine and sine factors of the complex exponential produce this effect. We multiply these curves by the product of the antenna gains, but the increased gain from larger antennas means that it is a greater distance to the far ﬁeld. When we bring two antennas close together, the currents on each antenna radiate and excite additional currents on the other that modify the result given by Eq. (1-52). But as we increase the distance, these induced current effects fade. Equivalent height analysis can be repeated using magnetic currents (e.g., used with microstrip patches), and Eqs. (1-51) and (1-52) become mutual admittance. Figure 1-7 is also valid for these antennas when we substitute normalized mutual admittance for normalized mutual impedance. For antennas with pattern nulls directed toward each other, the mutual impedance decreases at the rate 1/R 2 , due to the polarization of current direction h. 1-13 ANTENNA FACTOR The EMC community uses an antenna connected to a receiver such as a spectrum analyzer, a network analyzer, or an RF voltmeter to measure ﬁeld strength E. Most of the time these devices have a load resistor ZL that matches the antenna impedance. The incident ﬁeld strength Ei equals antenna factor AF times the received voltage Vrec . We relate this to the antenna effective height: AF =

Ei 2 = Vrec h

(1-53)

AF has units meter−1 but is often given as dB(m−1 ). Sometimes, antenna factor is referred to the open-circuit voltage and it would be one-half the value given by Eq. (1-53). We assume that the antenna is aligned with the electric ﬁeld; in other words, the antenna polarization is the electric ﬁeld component measured: 1 η 4π AF = = ZL Aeff λ ZL G This measurement may be corrupted by a poor impedance match to the receiver and any cable loss between the antenna and receiver that reduces the voltage and reduces the calculated ﬁeld strength. 1-14 MUTUAL COUPLING BETWEEN ANTENNAS The simplest approach for coupling between antennas is to start with a far-ﬁeld approximation. We can modify Eq. (1-8) for path loss and add the phase term for the ﬁnite distance to determine the S-parameter coupling: S21 =

e−j kr E1 ·E∗2 G1 G2 2kr |E1 ||E2 |

(1-54)

Equation (1-54) includes the polarization efﬁciency when the transmitted polarization does not match the receiving antenna polarization. We have an additional phase term

30

PROPERTIES OF ANTENNAS

because the signal travels from the radiation phase center along equivalent transmission lines to the terminals of each antenna. Equations (1-52) and (1-54) have the same accuracy except that Eq. (1-54) eliminates the need to solve the two-port circuit matrix equation for transmission loss. These formulas assume that antenna size is insigniﬁcant compared to the distance between the antennas, and each produces approximately uniform amplitude and phase ﬁelds over the second element. We can improve on Eq. (1-54) when we use the current distribution on one of the two antennas and calculate the near-ﬁeld ﬁelds radiated by the second antenna at the location of these currents. Since currents vary across the receiving antenna, we use vector current densities to include direction: Jr electric and Mr magnetic. Although magnetic current densities are ﬁctitious, they simplify the representation of some antennas. We compute coupling from reactance, an integral across these currents [see Eq. (2-34)]: j S21 = √ (Et ·Jr − Ht · Mr ) dV (1-55) 2 Pr Pt The input power to the transmitting antenna Pt produces ﬁelds Et and Ht . The power Pr into the receiving antenna excites the currents. The scalar product between the incident ﬁelds and the currents includes polarization efﬁciency. If we know the currents on the transmitting antennas, we calculate the near-ﬁeld pattern response from them at the location of the receiving antenna. Similar to many integrals, Eq. (1-55) is notional because we perform the integral operations only where currents exist. The currents could be on wire segments or surfaces. A practical implementation of Eq. (1-55) divides the currents into patches or line segments and performs the scalar products between the currents and ﬁelds on each patch and sums the result. A second form of the reactance [see Eq. (2-35)] involves an integral over a surface surrounding the receiving antenna. In this case each antenna radiates its ﬁeld to this surface, which requires near-ﬁeld pattern calculations for both. Equation (1-55) requires adding the phase length between the input ports and the currents, similar to using Eq. (1-54). When we use Eq. (1-55), we assume that radiation between the two antennas excites insigniﬁcant additional currents on each other. We improve the answer by using a few iterations of physical optics, which ﬁnds induced currents from incident ﬁelds (Chapter 2). We improve on Eq. (1-55) by performing a moment method calculation between the two antennas. This involves subdividing each antenna into small elements excited with simple assumed current densities. Notice the similarity between Eqs. (1-52) and (1-54) and realize that Eq. (1-55) is a near-ﬁeld version of Eq. (1-54). We use reactance to compute the mutual impedance Z21 between the small elements as well as their selfimpedance. For the moment method we calculate a mutual impedance matrix with a row and column for each small current element. We formulate a matrix equation using the mutual impedance matrix and an excitation vector to reduce coupling to a circuit problem. This method includes the additional currents excited on each antenna due to the radiation of the other.

1.15 ANTENNA NOISE TEMPERATURE [10] To a communication or radar system, an antenna contributes noise from two sources. The antenna receives noise power because it looks out on the sky and ground. The

ANTENNA NOISE TEMPERATURE

31

FIGURE 1-8 Antenna sky temperature. Noise temperature of an idealized antenna (lossless, no Earth-directed sidelobes) located at the Earth’s surface, as a function of frequency, for a number of beam elevation angles. Solid curves are for geometric-mean galactic temperature, sun noise 10 times quiet level, sun in unity-gain sidelobe, cool temperate-zone troposphere, 2.7 K cosmic blackbody radiation, zero ground noise. The upper dashed curve is for maximum galactic noise (center of galaxy, narrow-beam antenna). Sun noise 100 times quiet level, zero elevation, other factors the same as solid curves. The lower dashed curve is for minimum galactic noise, zero sun noise, 90◦ elevation angle. (The bump in the curves at about 500 MHz is due to the sun-noise characteristic. The curves for low elevation angles lie below those for high angles at frequencies below 400 MHz because of reduction of galactic noise by atmospheric absorption. The maxima at 22.2 and 60 GHz are due to the water-vapor and oxygen absorption resonance.) (From L. V. Blake, A guide to basic pulse-radar maximum-range calculation, Naval Research Laboratory Report 5868, December 1962.)

ground generates noise because it is about 290 K and a portion of the antenna pattern falls on it. Similarly, the sky adds noise dependent on the elevation angle and the operating frequency. Figure 1-8 gives the sky temperature versus frequency and elevation angle. The frequency range of lowest noise occurs in the middle of microwave frequencies of 1 to 12 GHz. The graphs show a large variation between the dashed curves, which occurs because of antenna direction and the pointing relative to the galactic center. In the middle of microwaves the sky noise temperatures are around 50 K, whereas near zenith the temperature is under 10 K. Near the horizon it rises because of the noise from oxygen and water vapor. The exact value must be determined for each application. As frequency decreases below 400 MHz, the sky temperature rises rapidly and becomes independent of antenna pointing. The curve continues the rapid rise at the same slope for lower frequencies. Low-frequency sky temperatures are often given as decibels relative to 290 K.

32

PROPERTIES OF ANTENNAS

An antenna receives this blackbody noise from the environment, but the value that affects the communication system depends both on the pattern shape and the direction of the main beam. We determine the antenna noise temperature by integrating the pattern times the environmental noise temperature distribution: Ta =

1 4π

2π 0

π

G(θ, φ)Ts (θ, φ) sin θ dθ dφ

(1-56)

0

where G(θ, φ) is the antenna gain pattern and Ts (θ, φ) is the angle-dependent blackbody radiation of the environment. Changing the antenna pointing changes Ta . Equation (1-56) is a weighted average of the environment noise temperature, usually referred to as the sky temperature. The second source of noise in the antenna is that of components that have both dissipative losses and reﬂection losses that generate noise. A receiving system needs to maximize the signal-to-noise ratio for given resources. System considerations, such as bit error rate, establish the required S/N ratio. We determine the noise power from the product N = k0 B n T e

(1-57)

where k0 is Boltzmann’s constant (1.38 × 10−23 W/K · Hz = −228.6 dB) and Bn is the receiver bandwidth (Hz). Te is the effective noise temperature (K). When referring noise temperature to other parts of the network, we increase or decrease it by the gain or loss, since it represents power and not a true temperature. Antenna gain is a measure of the signal level, since we can increase gain independent of the noise temperature, although the gain pattern is a factor by Eq. (1-56). The antenna conductor losses have an equivalent noise temperature: Te = (L − 1)Tp

(1-58)

where Tp is the antenna physical temperature and L is the loss (a ratio > 1). From a systems point of view, we include the transmission line run to the ﬁrst ampliﬁer or mixer of the receiver. We do not include the current distribution losses (aperture efﬁciencies) that reduce gain in Eq. (1-58) because they are a loss of potential antenna gain and not noise-generating losses (random electrons). The antenna–receiver chain includes mismatch losses, but these do not generate random electrons, only reﬂected waves, and have a noise temperature of zero. We include them in a cascaded devices noise analysis as an element with loss only. Noise characteristics of some receiver components are speciﬁed as the noise ﬁgure FN (ratio), and cascaded devices’ noise analysis can be analyzed using the noise ﬁgure, but we will use noise temperature. Convert the noise ﬁgures to noise temperature using TE = (FN − 1)T0

(1-59)

T0 is the standard reference temperature 290 K. We calculate noise temperature for the entire receiver chain of devices at a particular point normally at the input to the ﬁrst device. To calculate the S/N ratio we use the transmitter power, path loss (including antenna gain and polarization efﬁciency), and the gains (losses) of any devices for signal to the location in the receiver chain where

ANTENNA NOISE TEMPERATURE

33

noise temperature is being calculated. We characterize a given antenna by the ratio G/T , a measure independent of transmitter power and path loss, but including the receiver noise characteristics. Using the input of the ﬁrst device as the noise reference point, we calculate the input noise temperature from component noise temperatures and gains: T2 T3 T4 T = T1 + + + + ··· (1-60) G1 G1 G2 G1 G2 G3 Equation (1-60) merely states that noise temperatures are powers that decrease when we pass backward through a device with gain G. Each noise term is referred to the input of the device, and we pass backward to all previous devices and reduce noise temperature by 1/G. If we decided to locate the noise reference point at the input to the second device, the noise initially referred to the chain input would increase by the gain of the ﬁrst device. The system noise temperature becomes T(2) : T(2) = T1 G1 + T2 +

T3 T4 + + ··· G2 G2 G3

The signal also passes through the ﬁrst device and the new gain at the input to the second device becomes GG1 . The gain and the noise temperature change by the same factor G1 and produce a constant ratio. By extending these operations to any location in the receiver chain, we show that G/T is constant through the receiver device chain. It is easiest to illustrate G/T noise calculations with an example. A ground station has a 5-m-diameter paraboloid reﬂector with 60% aperture efﬁciency with the system operating at 2.2 GHz (λ = 0.136 m). We compute antenna directivity using the physical area and aperture efﬁciency:

π · Dia directivity = 0.60 λ

2

5π = 0.60 0.136

2 = 7972

(39 dB)

The reﬂector feed loss is 0.2 dB and it has a VSWR of 1.5 : 1. The cable between the feed and the ﬁrst ampliﬁer (LNA) has a 0.5-dB loss. These are elements under control of the antenna designer. We calculate the noise temperature of these by using Eq. (1-58) when we use a physical antenna temperature of 37.7◦ C (100◦ F) (310.8 K). Feed loss: T1 = (100.2/10 − 1)310.8 = 14.65 K Feed mismatch: T2 = 0 K Cable: T3 = (100.5/10 − 1)310.8 = 37.92 K The gains of these devices are G1 = 10−0.2/10 = 0.955 (feed loss), G2 = 10−0.18/10 = 0.959 (reﬂected power loss for 1.5 : 1 VSWR), and G3 = 10−0.5/10 = 0.891 (cable loss). The antenna sees the environment that generates noise due to blackbody radiation from the sky and ground. A typical value for the antenna pointed at 5◦ elevation is 50 K. This is not a physical temperature but represents an equivalent received power. Remember that the 60% aperture efﬁciency has no noise or loss contribution, because it only represents the loss of potential gain, since no random electrons are generated. We must consider the rest of the receiver chain when calculating the total input noise temperature. For this example we assume that the LNA has a noise ﬁgure of

34

PROPERTIES OF ANTENNAS

2 dB with 20 dB gain. The ﬁnal portion of the receiver includes the mixer and IF of the receiver, which we assume has a 10-dB noise ﬁgure. We use Eq. (1-59) to convert noise ﬁgure to noise temperature. LNA noise temperature T4 = 290(102/10 − 1) = 162.62 K Receiver noise temperature T5 = 290(1010/10 − 1) = 2610 K The 20-dB (100) LNA gain greatly reduces the effect of the 2610-K receiver. We calculate the contribution of each device to the input noise temperature by applying Eq. (1-60) to each device. We pass the noise temperature of the receiver through the four devices, and its temperature is reduced by the gain of each device: Te5 =

T5 2610 = 31.98 K = G1 G2 G3 G4 0.955(0.959)(0.891)(100)

The gain of the LNA greatly reduced the effective noise of the receiver at the antenna input. This operation shows that cascading noise temperature involves passing each device’s noise temperature through the gains of all preceding devices to the input and reducing it by the product of their gains. Similarly, we perform this operation on all the other noise temperatures. T4 169.62 = 207.86 K = G1 G2 G3 0.955(0.959)(0.891) T3 37.92 = 41.40 K Te3 = = G1 G2 0.955(0.959) T2 0 =0 = Te2 = G1 0.955 Te1 = T1 = 14.65 Te4 =

These operations illustrate that the cascaded devices’ noise temperature equation (1-60) is easily derived by considering the passage of noise temperature (power) through devices with gain to a common point where we can add the contributions. The sky temperature is not an input noise temperature but the noise power delivered at the ﬁctitious point called the antenna directivity, where gain = directivity. Since noise temperature represents power, we convert it to decibels and subtract it from directivity to compute G/T : G/T (dB) = 39 − 10 log(345.9) = 13.6 dB This G/T is a measure of the antenna and receiver combined performance when the antenna is pointed to 5◦ elevation. Changing the pointing direction affects only the sky temperature added directly to the ﬁnal result. We use G/T in the link budget of the communication system. We can supply a single value for the antenna gain and noise temperature at the output port connected to the receiver. Recognize that the ﬁrst three noise temperatures and the sky temperature are associated with the antenna. We moved the noise reference of each device to the input by dividing by the gain of the preceding devices. To move

COMMUNICATION LINK BUDGET AND RADAR RANGE

35

to the output of the antenna, we increase the noise temperature and the antenna gain by the product of the gain for the devices: T = (Tsky + Te1 + Te2 + Te3 )G1 G2 G3 = (50 + 14.65 + 0 + 37.92)10−0.88/10 = 83.7 K gain(dB) = directivity(dB) − 0.88 dB = 39 − 0.88 = 38.12 dB This reduces the antenna to a single component similar to the directivity and sky temperature that started our analysis. 1-16 COMMUNICATION LINK BUDGET AND RADAR RANGE We illustrate communication system design and path loss by considering a sample link budget example. The 5-m-diameter reﬂector is pointing at a satellite in an orbit 370 km above the Earth with a telemetry antenna radiating 10 W at 2.2 GHz. Since the antenna pattern has to cover the visible Earth, its performance is compromised. Considering the orbit geometry and antenna pointing is beyond the scope of this discussion. The range from a satellite at 370 km to a ground station pointing at 5◦ is 1720 km. The satellite antenna pointing angle from nadir is 70.3◦ , and a typical antenna for this application would have gain = −2 dBiC (RHC gain relative to an isotropic antenna) and an axial ratio of 6 dB. Assume that the ground station antenna has a 2-dB axial ratio. We apply the nomograph of Figure 1-6 to read the maximum polarization loss of 0.85 dB since we cannot control the orientation of the polarization ellipses. The link budget needs to show margin in the system, so we take worst-case numbers. When we apply Eq. (1-9) for path loss, we leave out the antenna gains and add them as separate terms in the link budget (Table 1-2): free-space path loss = 32.45 + 20 log[2200(1720)] = 164 dB The link budget shows a 4.4-dB margin, which says that the communication link will be closed. This link budget is only one possible accounting scheme of the system parameters. Everyone who writes out a link budget will separate the parameters differently. This budget shows typical elements. Radar systems have similar link budgets or detection budgets that consider S/N : S/N =

Prec PT GT (directivity)λ2 σ (EIRP)λ2 (G/T )σ = = KT B (4π)3 R 4 KT B (4π)3 R 4 KB

The radar has a required S/N value to enable it to process the information required, which leads to the maximum range equation:

(EIRP)λ2 (G/T )σ R= (4π)3 (S/N )req KB

1/4 (1-61)

Equation (1-61) clearly shows the role of the transmitter, EIRP; the receiver and antenna noise; G/T ; and the requirement for signal quality, S/Nreq , on the radar range for a given target size σ . Equation (1-61) applies to CW radar, whereas most radars use pulses. We increase radar performance by adding many pulses. We ignore the aspects of pulse train encoding

36

PROPERTIES OF ANTENNAS

TABLE 1-2 Link Budget Frequency Transmit power Transmit antenna gain EIRP (effective isotropic radiated power) Free-space path loss Polarization loss Atmospheric loss Rain loss Pointing loss Receive antenna directivity G/T Boltzmann’s constant Carrier/noise (C/N) ratio (ignores bandwidth) Bit rate: 8 Mb/s Eb /N0 (energy per bit/noise density) Implementation loss Eb /N0 required Margin

2.2 GHz 10 dBW −2 dB 8 dBW 164 dB 0.85 0.30 0.00 0.00 39 dB 13.6 dB 228.6 dB 85 dB 69 dB 16 dB 2 dB 9.6 dB 4.4 dB

Information only 10 log(10) Transmit power dBW + antenna gain dB Isotropic antenna path loss Maximum for uncontrolled orientation 5◦ elevation at 2.2 GHz Little loss at this frequency Location in receiver chain for G/T calculation From preceding section EIRP + G/T − path loss − polarization loss −atmospheric loss − rain loss + 228.6 10 log(bit rate) bandwidth Eb /N0 = C/N − 10 log(bit rate) Groups extra system losses For bit error rate (BER) = 10−5 in QPSK Eb /N0 − required Eb /N0 − implementation loss

that allow coherent addition. Radar range is determined by the total energy contained in the pulses summed. We replace EIRP with GT (energy) since PT × time = energy. It is the total energy that illuminates the target that determines the maximum detection range. Using antennas in radar leads to speaking of the radiated energy correct for pulsed systems, but when we do not integrate pulse shape times time, the antenna radiates power. To be correct we should call radiation that we integrate over angular space to ﬁnd power, “power density.” To say “energy radiated in the sidelobes” is poor physics unless it is a radar system, because it is power. 1-17

MULTIPATH

Multipath means that the ﬁeld intensity at a particular point is the sum of a number of waves that arrive from different directions or from different sources. It arises from signal transmission paths such as edge reﬂections from the mounting structure around an antenna and general reﬂections from objects near the antenna. Nearby reﬂections only seem to modify the antenna pattern, while reﬂections from additional objects cause rapid ripple with changing pattern angle. In Section 3-1 we discuss how to use the ripple angular rate and pattern distribution to locate its source. Multipath causes degraded system performance or measurement errors. Of course, multipath can improve performance as well. In fact, we add nearby objects, such as ground planes, to improve antenna performance. We specify pattern response in terms of the power response, but we add ﬁelds. An extra signal −20 dB relative to the main signal is 0.01 in power but 0.1 in ﬁeld strength

PROPAGATION OVER SOIL

37

Peak-to-Peak Ripple, dB

Interference Signal Level, dB

SCALE 1-8 Signal peak-to-peak amplitude ripple due to multipath signal. Maximum Phase Error (degrees)

Interference Signal Level, dB

SCALE 1-9 Peak phase error due to multipath signal.

(voltage). Since the extra signal can have any phase relative to the main signal, it can add or subtract. Given an extra signal MP(dB), the pattern ripple is ripple(dB) = 20 log

1 + 10MP/20 1 − 10MP/20

(1-62)

where MP(dB) has a negative sign. Scale 1-8 gives the relationship between peakto-peak amplitude ripple and the level of the multipath signal. Equation (1-62) is numerically the same as the relationship between return loss and 20 log(VSWR). The multipath signal can change the phase when summed with the main signal over a range given by maximum phase error = ± tan−1 (10MP/20 ) (1-63) Scale 1-9 calculates the peak phase error due to a multipath signal. 1-18 PROPAGATION OVER SOIL When we position antennas over soil and propagate the signal any signiﬁcant distance, it will reﬂect from soil or water and produce a large multipath signal. Soil is a conductive dielectric that reﬂects horizontally and vertically polarized signals differently. Typical ground constants are listed in Table 1-3. Given the grazing angle ψ measured between the reﬂected ray and ground, the voltage reﬂection coefﬁcients are sin ψ − εr − j x − cos2 ψ (εr − j x) sin ψ − εr − j x − cos2 ψ ρh = and ρv = sin ψ + εr − j x − cos2 ψ (εr − j x) sin ψ + εr − j x − cos2 ψ (1-64) where x = σ/ωε0 = 17, 975σ /frequency(MHz). Figure 1-9 gives the reﬂection coefﬁcient for the two polarizations versus grazing angle. Horizontal polarization reﬂects from soil about the same as a metal surface. Vertical polarization reﬂection produces a more interesting curve. The graph shows that the reﬂection is low over a region of grazing angles. The minimum reﬂection direction is called the Brewster angle. At this angle the reﬂected wave is absorbed

38

PROPERTIES OF ANTENNAS

TABLE 1-3

Typical Ground Constants

Surface Dry ground Average ground Wet ground Fresh water Seawater

Dielectric Constant

Conductivity (S)

4–7 15 25–30 81 81

0.001 0.005 0.020 0.010 5.0

Reflection, dB

Horizontal Polarization

Vertical Polarization

Grazing Angle

FIGURE 1-9

Average soil reﬂection for horizontal and vertical polarization.

into the soil. At high grazing angles ρh has a phase near 180◦ and ρv a phase of 0◦ . When the grazing angle decreases and becomes less than the Brewster angle, the vertical polarization reﬂection changes from 0◦ to 180◦ . Remember that for most general response nulls, the signal phase changes by 180◦ when passing through the transition. As the grazing angle approaches zero both reﬂection coefﬁcients approach −1 and multipath is independent of polarization. The electric ﬁeld at the receiving antenna is the sum of the direct wave plus the reﬂected wave, which traveled along a longer path: E = Ed [1 − exp(−j φ)] = Ed (1 − cos φ + j sin φ) We compute the magnitude |E| = |Ed | 1 + cos2 φ + sin2 φ − 2 cos φ = 2|Ed | sin(φ/2) for the small phase difference between the two equal-amplitude signals. The received power Prec is proportional to E 2 . The path loss for this multipath link is modiﬁed from the free-space equation: λ 2 hT hR 2 2 2πhT hR Prec = 4PT → PT GT GR GT GR sin (1-65) 4πd λd d2

MULTIPATH FADING

39

Equation (1-65) states that the power received is proportional to 1/d 4 and increases by h2 for either antenna. We can approximate the propagation over soil by a region for closely spaced antennas when the results consist of the free-space transmission with 1/d 2 average transmission with signiﬁcant variation due to multipath and a second region proportional to 1/d 4 with small multipath variations. The breakpoint between the two models occurs at a distance d = 4hT hR /λ. Experiments at mobile telephone frequencies showed that Eq. (1-65) overestimates the received power when the receiving antenna height is less than 30 m and a more correct model modiﬁes the exponent of hR [11, p. 38]: h2T hCR (1-66) d4 Below 10 m, C = 1 and the exponent varies linearly between 10 and 30 m: C = hR /20 + 12 . On a narrow-beam terrestrial propagation path, scattering from an object along a path an odd multiple of λ/2 produces a signal that reduces the main path signal. Given an obstacle at a distance h radial from the direct ray path and located dT from the transmitter and a distance dR from the receiving antenna, we determine the differential path length as λ nλdT dR h2 dT + dR =n (1-67) or clearance height h = = 2 dT dR 2 dT + dR Prec = PT GT GR

We call these Fresnel clearance zones of order n. The direct path should clear obstacles by at least one clearance zone distance h to prevent the scattered signal from having a negative impact on the communication link. The ﬁrst Fresnel zone touches ground when dT = 4hT hR /λ is the breakpoint distance between 1/d 2 and 1/d 4 propagation models. 1-19 MULTIPATH FADING Most mobile communication occurs when there is no direct path between the base station antennas and the mobile user. The signal reﬂects off many objects along the path between the two. This propagation follows a Rayleigh probability distribution about the mean signal level: r r2 R2 pr (r) = 2 exp − 2 prob[r < R] = PR (R) = 1 − exp − 2 α 2α 2α √ R is the signal level,√ α the value of the peak in the distribution, with mean = α π/2 and median RM = α 2 ln(2) = 1.1774α. The median signal level is found by ﬁtting measured data for various localities (town, small town, open country, etc.) into a prediction model. The signal will have large signal fades where the level drops rapidly. The Rayleigh model can be solved for the average distance between fades given the level. As a designer it is important to realize the magnitude of the problem [12, pp. 125–130]: 2(R/RM ) average distance between fades = λ √ 2π ln(2)(R/RM ) 2

(1-68)

40

PROPERTIES OF ANTENNAS Distance between Fades, l

Rayleigh Multipath Signal Fade, dB

SCALE 1-10

Average distance between fades and depth of fade for a Rayleigh multipath.

Length of Fade Region, l

Rayleigh Multipath Signal Fade, dB

SCALE 1-11

Average fade length and depth of fade for a Rayleigh multipath.

R is the fade level (ratio) and RM is the median signal level found from a propagation model. Scale 1-10 shows the relationship between the average distance between fades and the depth of fade for Rayleigh multipath. A mobile channel operating at 1.85 GHz (λ = 16.2 cm) has a 15-dB fade every 2.75λ which equals 44.5 cm, while 10-dB fades occur every 1.62λ = 26.25 cm. The communication system must overcome these fades. Fortunately, the deep fades occur over a short distance: 2(R/RM ) − 1 average length of fade = λ √ 2π ln(2)(R/RM ) 2

The signal fades and then recovers quickly for a moving user. Scale 1-11 shows the average fade length along a path given the depth of fade. For the 1.85-GHz channel the 15-dB fade occurs only over 0.06(16.2) = 0.97 cm, and the 10-dB fade length is 0.109(16.2) = 1.76 cm. The solution to mobile communication multipath fading is found either in increasing the link margin with higher gain base station antennas or the application of diversity techniques. We use multiple paths between the user and the base station so that while one path experiences a fade, the other one does not. Diversity has no effect on the median signal level, but it reduces the effects of the nulls due to the Rayleigh distribution propagation. REFERENCES 1. S. A. Schelkunoff and H. Friis, Antenna Theory and Practice, Wiley, New York, 1952. 2. R. F. Harrington, Time-Harmonic Electromagnetic Fields, McGraw-Hill, New York, 1961. 3. H. Friis, A note on a simple transmission formula, Proceedings of IRE, vol. 34, May 1946, pp. 254–256.

REFERENCES

41

4. J. D. Kraus, Antennas, McGraw-Hill, New York, 1950. 5. N. McDonald, Omnidirectional pattern directivity in the presence of minor lobes: revisited, IEEE Antennas and Propagation Magazine, vol. 41, no. 2, April 1999, pp. 63–65. 6. W. F. Croswell and M. C. Bailey, in R. C. Johnson and H. Jasik, eds., Antenna Engineering Handbook, McGraw-Hill, New York, 1984. 7. C. G. Montgomery, R. H. Dicke, and E. M. Purcell, Principles of Microwave Circuits, McGraw-Hill, New York, 1948. 8. A. C. Ludwig, The deﬁnition of cross polarization, IEEE Transactions on Antennas and Propagation, vol. AP-21, no. 1, January 1973, pp. 116–119. 9. K. P. Park and C. T. Tai, Receiving antennas, Chapter 6 in Y. T. Lo and S. W. Lee, eds., Antenna Handbook, Van Nostrand Reinhold, New York, 1993. 10. L. V. Blake, Prediction of radar range, Chapter 2 in Radar Handbook, M. Skolnik, ed., McGraw-Hill, New York, 1970. 11. K. Fujimoto and J. R. James, Mobile Antenna Systems Handbook, 2nd ed., Artech House, Boston, 2001. 12. J. D. Parsons, The Mobile Radio Propagation Channel, Wiley, New York, 1992.

2 RADIATION STRUCTURES AND NUMERICAL METHODS

Antenna analysis, an important part of design, requires a compromise between extensive calculations and the fabrication and measurement of prototypes, which depends on your working environment. You should minimize cost, which means reducing the time from the start of a design to completion of a working model. In some cases you should not rush to build a prototype. For example, when designing large and expensive antennas, such as paraboloidal reﬂectors, the high fabrication cost justiﬁes the time required for analysis. Management will not let you proceed before knowing the design will work. You should develop a cost model for each design in which analysis is one factor. Analyses allow optimization of a design. You can design a number of antennas and adjust the dimensions until you ﬁnd the best one. Again, you should be considering the costs of your time. At some point the incremental improvements are not worth the extra time for further analyses. In any case, when you build the prototype, you can expect differences. You soon determine that you can achieve only limited knowledge about a design because fabrication and measurement errors mask the true response of the antenna. You are doing engineering, not a science project. Textbooks contain many analyses of ideal antennas, and this book is no exception. You need to consider the application and the ﬁnal antenna environment. The mounting structure has little effect on the pattern of a large antenna with narrow beamwidth because little radiation strikes it. The overall radiation characteristics of narrow- or wide-beam antennas depend signiﬁcantly on the shape of the vehicle and how the antenna is mounted. In later chapters we discuss how to use antenna mounting to improve performance, so you can take advantage of it. The size of the mounting structure limits the type of analysis used. In this chapter we discuss physical optics (PO) and geometric optics (GO) [geometric theory of diffraction (GTD)] for large structures. In physical optics we compute the current induced on the vehicle due to antenna radiation and include their radiation in the Modern Antenna Design, Second Edition, By Thomas A. Milligan Copyright 2005 John Wiley & Sons, Inc.

42

AUXILIARY VECTOR POTENTIALS

43

overall pattern. But the PO analysis cost rises rapidly as the number of small current patches increases for larger structures. PO analysis works well with large antennas, such as paraboloidal reﬂectors, that produce focused beams. Geometric optics uses ray optics techniques whose computation cost is independent of the size of the vehicle and whose accuracy improves as structure size increases. GO provides insight because we can visualize the combination of direct, reﬂected, and diffracted (GTD) rays to calculate the pattern, but it requires the solution of difﬁcult geometry problems. Smaller structures allow the use of multiple methods. For example, the moment method divides the surroundings into small patches and uses an expansion of the current in predetermined basis functions. This method uses integral equations of the boundary conditions to calculate a matrix equation involving coefﬁcients of the current expansion. Numerical methods invert the matrix to solve for the coefﬁcients, but it is a costly numerical operation and limits the size of the problem that can be handled to a few wavelengths. The ﬁnite-difference time-domain (FDTD) technique computes the ﬁelds on the structure in the time domain. This method handles moderate-sized structures and readily includes complex material properties such as biological features. FDTD divides the region into cubic cells and when excited by pulse feeding functions, it produces wide frequency bandwidth responses. Finite-element methods (FEMs) also divide the problem into cubic cells, but the analysis is performed in the frequency domain. FEM analysis must be repeated at every frequency of interest. FDTD and FEMs require a program to divide the structure into a mesh before starting the solution. Both methods calculate currents on a boundary surface by using the equivalence theorem with the incident ﬁelds and then calculate the far-ﬁeld radiation pattern from these boundary currents. Most methods start by assuming a current distribution on the antenna or, equivalently, a distribution of ﬁelds on an aperture. The ﬁelds on the aperture can be reduced to a current distribution. The moment method uses a summation of assumed basis function currents and solves for the coefﬁcients of the expansion, but it, too, starts with assumed currents over small regions. You will discover that the radiation pattern can be found with greater accuracy than the input impedance. For antennas constructed from wires, the moment method computes the input current for a given excitation voltage and we calculate impedance from the ratio. Interaction of an antenna with the currents induced on a structure has little effect on impedance for narrow-beam antennas. Even for wide-beam antennas, such as dipoles, the structure effect on impedance can be found by using source mutual coupling with its images. In the end, antenna impedance should be measured when mounted in the ﬁnal conﬁguration. An antenna has both a radiation pattern bandwidth and an impedance bandwidth, but you must give the pattern primary consideration. Too many designs concentrate on the wideband impedance characteristics of an antenna when, in fact, the antenna pattern has changed over the frequency range of the impedance bandwidth. Your primary task should be to design for the radiation pattern desired. In Chapter 1 we detailed the system aspects of impedance mismatch (Section 1-10), and you may determine the overall system impact of small impedance mismatch. 2-1 AUXILIARY VECTOR POTENTIALS We do not use vector potentials in design. It seems as though they would be useful, but only a few simple antennas ﬁt their direct use. You cannot measure them

44

RADIATION STRUCTURES AND NUMERICAL METHODS

because they are not physical entities, so they seem artiﬁcial. Physical optics (PO) calculates the radiation directly from currents using dyadic Green’s functions but uses long expressions. Nevertheless, many analysis techniques ﬁnd them more efﬁcient than PO expressions and you should be aware of them. We illustrate their use with a couple of simple antennas. We use vector potentials to introduce a few antenna concepts. In the ﬁrst example we apply the magnetic vector potential to calculate the radiation from a short-length current element (dipole) and show how to obtain the pattern. Integration of the radiation pattern power density (Section 1-2) determines the total power radiated. Because we know the input current and the total radiated power, the ratio of the power to the input current squared gives the radiation resistance. We combine the low radiation resistance with the material resistance to compute the antenna efﬁciency. Electric vector potentials used with ﬁctitious magnetic currents illustrate analysis by duality. We apply this to the analysis of a small loop and show that it has the same pattern as that of a small dipole. 2-1.1 Radiation from Electric Currents Normal electron currents radiate when time varying. The simplest example is a ﬁlamentary current on wire, but we include surface and volumetric current densities as well. We analyze them by using the magnetic vector potential. Far-ﬁeld electric ﬁelds are proportional to the magnetic vector potential A: E = −j ωA

(2-1)

We determine the magnetic ﬁeld from |E| = η|H|

(2-2)

and realizing the cross product of the electric ﬁeld with the magnetic ﬁeld points in the direction of power ﬂow, the Poynting vector. Since the electric ﬁeld direction deﬁnes polarization, we usually ignore the magnetic ﬁeld. We derive the magnetic vector potential from a retarded volume integral over the current density J: A=µ

J(r )e−j k|r−r | dV 4π|r − r |

(2-3)

where r is the ﬁeld measurement point radius vector, r the source-point radius vector, µ the permeability (4π × 10−7 A/m in free space), and k, the wave number, is 2π/λ. As written, Eq. (2-3) calculates the potential A everywhere: near and far ﬁeld. The vector potential can be written in terms of a free-space Green’s function: e−j kR where R = |r − r | 4πR A=µ g(R)J(r ) dV

g(R) =

(2-4)

Radiation Approximation When we are interested only in the far-ﬁeld response of an antenna, we can simplify the integral [Eq. (2-3)]. An antenna must be large in terms

AUXILIARY VECTOR POTENTIALS

45

of wavelengths before it can radiate efﬁciently with gain, but at great distances it still appears as a point source. Consider the radiation from two different parts of an antenna. Far away from the antenna, the ratio of the two distances to the different parts will be nearly 1. The phase shift from each part will go through many cycles before reaching the observation point, and when adding the response from each part, we need only the difference in phase shift. In the radiation approximation we pick a reference point on the antenna and use the distance from that point to the far-ﬁeld observation point for amplitudes, 1/R, for all parts of the antenna. The direction of radiation deﬁnes a plane through the reference point. This plane is deﬁned by the radius normal vector, given in rectangular coordinates by rˆ = sin θ cos φ xˆ + sin θ sin φ yˆ + cos θ zˆ We compute the phase difference to the far-ﬁeld point by dropping a normal to the reference plane from each point on the antenna. This distance multiplied by k, the propagation constant, is the phase difference. Given a point on the antenna r , the phase difference is kr · rˆ . When we substitute these ideas into Eq. (2-3), the equation becomes e−j kr µ J ej kr ·ˆr dV A= (2-5) 4πr In rectangular coordinates kr · rˆ becomes k(x sin θ cos φ + y sin θ sin φ + z cos θ ) We can combine k and rˆ to form a k-space vector: k = kˆr = k sin θ cos φ xˆ + k sin θ sin φ yˆ + k cos θ zˆ and the phase constant becomes k · r . Currents in ﬁlaments (wires) simplify Eq. (2-5) to a single line integral. Magnetic vector potentials and electric ﬁelds are in the same directions as the wires that limit the directions of current. For example, ﬁlamentary current along the z-axis produces z-directed electric ﬁelds. Spherical waves (far ﬁeld) have only θˆ and φˆ components found from the projection of Ez onto those axes. Filamentary currents on the z-axis produce only z-directed electric ﬁelds with a null from θˆ · zˆ = − sin θ at θ = 0. In turn, x- or y-directed currents produce electric ﬁelds depending on the scalar products (projections) of the xˆ and yˆ unit vectors onto the θˆ and φˆ vectors in the far ﬁeld: θˆ · xˆ = cos θ cos φ θˆ · yˆ = cos θ sin φ

φˆ · xˆ = − sin φ φˆ · yˆ = cos φ

By examining antenna structure you can discover some of its characteristics without calculations. Without knowing the exact pattern, we estimate the polarization of the waves by examining the directions of the wires that limit the current density. Consider various axes or planes of symmetry on an antenna: for example, a center-fed wire along the z-axis. If we rotate it about the z-axis, the problem remains the same, which means that all conical polar patterns (constant θ ) must be circles; in other words, all great

46

RADIATION STRUCTURES AND NUMERICAL METHODS

circle patterns must be the same. An antenna with the same structure above and below the x –y plane radiates the same pattern above and below the x –y plane. Always look for axes and planes of symmetry to simplify the problem. We can extend the magnetic vector potential [Eq. (2-1)] to determine near ﬁelds: E = −j ωA + H=

∇(∇ · A) j ωεµ

1 ∇ ×A µ

(2-6)

The electric ﬁeld separates into far- and near-ﬁeld terms, but the equation for the magnetic ﬁeld, the deﬁning equation of the potential, does not separate. If we substitute the free-space Green’s function from Eq. (2-4) into Eq. (2-6), expand, and gather terms, we can determine the ﬁelds directly from the electric currents and eliminate the use of a vector potential. 1 j j ηk 2 − + 3 3 J(r ) − E(r) = 4π kR k 2 R 2 k R V

3 j 3j + 2 2− 3 3 e−j kR dV kR k R k R k2 1 j ˆ H(r) = + 2 J(r ) × R e−j kR dV 4π kR k R ˆ R ˆ +[J(r ) · R]

V

ˆ = r−r = r−r R |r − r | R

since

(2-7) (2-8)

R = |r − r |

Terms with 1/R dependence are the far-ﬁeld terms. The radiative near-ﬁeld terms have 1/R 2 dependence and near-ﬁeld terms have 1/R 3 dependence. The impedance of free space, η, is 376.7 . We can rearrange Eqs. (2-7) and (2-8) so that they become the integral of the dot product of the current density J with dyadic Green’s functions [1]. It is only a notation difference that leads to a logic expression. Except for a few examples given below, we leave the use of these expressions to numerical methods when designing antennas. Example Use the magnetic vector potential to derive the far ﬁeld of a short-length current element. Assume a constant current on the wire. The current density is I lδ(r ), where δ(r ) is the Dirac delta distribution and l is the length over which the far-ﬁeld phase is constant. The integral in Eq. (2-4) easily reduces to µI le−j kr 4πr The current element is so short that the phase distances from all parts of the element are considered to be equal; e−j kr is the retarded potential phase term. The electric ﬁeld is found from Az using Eq. (2-1): I l −j kr Ez = −j ωµ e 4πr I l −j kr Eθ = Ez zˆ · θˆ = j ωµ sin θ e 4πr Az =

AUXILIARY VECTOR POTENTIALS

47

√ √ √ µ µ, and divide and multiply by ε: √ j I l2πf µε µ −j kr sin θ Eθ = e 4πr ε

Evaluate ω as 2πf , split µ in

The following terms can be recognized as l f = c λ

1 c= √ µε

η=

µ ε

The far-ﬁeld electric ﬁeld becomes j I lη −j kr sin θ e 2λr The magnetic ﬁeld is found from the electric ﬁeld using Eq. (2-2): Eθ =

Hφ =

j I l −j kr Eθ = e sin θ η 2λr

The term j can be evaluated as ej π/2 , a phase shift term. The power density Sr is Sr = Eθ Hφ∗ =

|I |2 l 2 η sin2 θ 4λ2 r 2

The normalized power pattern is equal to sin2 θ . Figure 2-1 gives the polar pattern of this antenna as a dashed plot. The dashed circle is the −3-dB pattern level. We measure the angular separation between the 3-dB points to determine the beamwidth (half-power beamwidth). For comparison, Figure 2-1 shows the pattern of a half-wavelength-long dipole as a solid curve. At a length about 5% shorter than a half wavelength, the reactive component of the impedance vanishes. The ﬁgure illustrates that a short dipole has about the same pattern as a long-resonant-length (reactance equals zero) dipole. We determine directivity (Section 1-7) by calculating the average radiation intensity, often normalized to the peak of the power pattern: π/2 2 Uavg = sin2 θ sin θ dθ = 3 0 Umax = 1 Umax directivity = = 1.5 (1.76 dB) Uavg The resonant-length dipole (≈ λ/2) has a directivity of 2.15 dB or only 0.39 dB more than that of the very short dipole. The total power radiated by the antenna is found by integrating the Poynting vector magnitude over a sphere: 2π π Pr = Sr r 2 sin θ dθ dφ

0

0 π

|I |l = 2λ 0 0 2 2π |I |l η = 3 2λ 2π

2 η sin θ dθ dφ

48

RADIATION STRUCTURES AND NUMERICAL METHODS

l/2 Dipole

Short Dipole

FIGURE 2-1 Pattern of a short current element and small loop (dashed curve) compared to a λ/2 -long dipole (solid curve) located along the 0 to 180◦ axis.

We represent the radiated power as a radiation resistance at the input of the antenna: 2 l Pr 2π η = RR = |I |2 3 λ While a short dipole with a length λ/20 has a radiation resistance of about 2 , a resonant-length dipole has about a 50- radiation resistance and is more efﬁcient because the relative material resistance is low. The input resistance of the antenna is the sum of the radiation resistance and the resistance due to material losses: Pin = (RR + RL )|I |2 The gain of an antenna is the ratio of the peak radiation intensity to the input power averaged over the radiation sphere: gain =

Sr,peak r 2 Umax = Pin Pin 4π 4π

By using the idea of radiation resistance, we rewrite this as gain =

4πUmax (RR + RL )|I |2

AUXILIARY VECTOR POTENTIALS

49

Efﬁciency is the ratio of radiated power to input power: ηe =

Pr Rr |I |2 Rr = = Pin (Rr + RL )|I |2 Rr + RL

Instead of integrating the pattern to calculate the total power radiated, we sometimes compute the input power of the antenna from currents induced on the antenna elements by given voltage sources on various terminals of the antenna in analysis: Pin = Re(V1 I1∗ ) + Re(V2 I2∗ ) + · · · + Re(VN IN∗ ) The gain can be found from gain =

Sr (θ, φ)r 2 U (θ, φ) = Pin Pin 4π 4π

This method is considerably easier than integrating the radiation intensity to compute directivity. By integrating the pattern, we found only the input resistance of the short antenna, not the reactive component. A short antenna has a large capacitive reactance term that limits the impedance bandwidth when combined with a match network. The short antenna has a large pattern bandwidth but a narrow impedance bandwidth. Of course, an active network could be designed to impedance-match the antenna at any frequency, but the instantaneous bandwidth is narrow. The moment method of analysis gives us the currents for given input voltages and calculates the complete input impedance. 2-1.2 Radiation from Magnetic Currents Magnetic currents are ﬁctitious, but they enable slot radiation to be solved by the same methods as electric currents on dipoles by using duality. Slot radiation could be calculated from the surface currents around it, but it is easier to use magnetic currents to replace the electric ﬁeld in the slot. Magnetic currents along the long axis of slots in ground planes replace the electric ﬁelds across the slots by application of the equivalence theorem. Similarly, current loops can be replaced by magnetic dipole elements to calculate radiation. We use the electric vector potential F with magnetic currents. The far-ﬁeld magnetic ﬁeld is proportional to the electric vector potential: H = −j ωF

(2-9)

We determine the magnitude of the electric ﬁeld from Eq. (2-2); it is perpendicular to H. The electric vector potential is found from a retarded volume integral over the magnetic current density M. Applying the radiation approximation, it is e−j kr F= (2-10) ε M ej k·r dV 4πr where ε is the permittivity (8.854 × 10−12 F/m in free space). Equation (2-9) is the dual of Eq. (2-1), and Eq. (2-10) is the dual of Eq. (2-5). The dual of Eq. (2-3) is valid in both the near- and far-ﬁeld regions.

50

RADIATION STRUCTURES AND NUMERICAL METHODS

The magnetic currents in a slot are perpendicular to the slot electric ﬁelds: M = E × n, ˆ where nˆ is the normal to the plane with the slot. The ﬁlamentary currents of thin slots reduce Eq. (2-10) to a line integral, and magnetic current direction limits the direction of the electric vector potential and the magnetic ﬁeld. Since the electric ﬁeld (far ﬁeld) is orthogonal to the magnetic ﬁeld, the electric ﬁeld is in the same direction as the ﬁeld across the slots. We use the direction of the electric ﬁeld across the slots to estimate the polarization of the far ﬁeld. As with ﬁlamentary electric currents, the far ﬁeld is zero along the axis of the magnetic current. The electric vector potential can also be used to derive the near ﬁeld: H = −j ωF +

∇(∇ · F) j ωµε

1 E = − ∇ ×F ε The magnetic ﬁeld separates into near- and far-ﬁeld terms in the electric vector potential; the electric ﬁeld does not. We can determine the radiated ﬁelds directly in terms of the magnetic currents and avoid using the vector potential: E(r) = −

k2 4π

ˆ M(r ) × R

V

1 j + 2 e−j kR dV kR k R

1 j j − + 3 3 M(r ) − kR k 2 R 2 k R V j 3j 3 ˆ R ˆ e−j kR dV +[M(r ) · R] + 2 2− 3 3 kR k R k R

k H(r) = 4πη 2

(2-11)

(2-12)

Equations (2-11) and (2-12) can be rearranged to ﬁnd the dyadic Green’s functions for magnetic currents. These differ from the dyadic Green’s functions for electric currents by only constants. Example Derive the ﬁelds radiated from a small constant-current loop. We could use the magnetic vector potential and calculate over the currents in the wire but must account for changing current direction around the loop. Place the loop in the x –y plane. The electric ﬁeld radiated by the loop is in the φ direction because the currents in the loop can only be in the φˆ direction. When solving the integral for the magnetic vector potential, note that the direction of the current on the loop, φˆ at a ˆ unit vector. The integral general point is not in the same direction as the ﬁeld point, φ, must be solved with a constant vector direction, one component at a time. Although the magnetic vector potential can be computed, it is easier to replace the current loop with an incremental magnetic current element. The equivalent magnetic current element is Im l = j ωµI A where A is the area of the loop. The magnetic current density is M = Im lδ(r )ˆz = j ωµI Aδ(r )ˆz

APERTURES: HUYGENS SOURCE APPROXIMATION

51

The electric vector potential is found using Eq. (2-10): Fz =

j ωµεI A −j kr e 4πr

The magnetic ﬁeld is found from this electric vector potential using Eq. (2-9): Hz = −j ωFz =

ω2 µεI A −j kr e 4πr

We calculate Hθ by projection: ω2 µεI A −j kr e sin θ Hθ = Hz zˆ · θˆ = − 4πr Eφ and Hθ are related in the far ﬁeld because the wave propagates in the r direction: Eφ = −ηHθ =

ω2 µεI Aη −j kr e sin θ 4πr

The small current loop and small current element have the same pattern shape, sin θ , but opposite polarizations. The directivity is 1.5 (1.76 dB). Figure 2-1 uses a dashed curve to plot the response of the small loop, while the solid curve gives the pattern of a half-wavelength slot that radiates on both sides of the ground sheet. 2-2 APERTURES: HUYGENS SOURCE APPROXIMATION Many antennas, such as horns or paraboloid reﬂectors, can be analyzed simply as apertures. We replace the incident ﬁelds in the aperture with a combination of equivalent electric and magnetic currents. We calculate radiation as a superposition of each source by using the vector potentials. Most of the time, we assume that the incident ﬁeld is a propagating free-space wave whose electric and magnetic ﬁelds are proportional to one another. This gives us the Huygens source approximation and allows the use of integrals over the electric ﬁeld in the aperture. Each point in the aperture is considered to be a source of radiation. The far ﬁeld is given by a Fourier transform of the aperture ﬁeld: f(kx , ky ) = Eej k·r ds (2-13) S

This uses the vector propagation constant k = kx xˆ + ky yˆ + kz zˆ kx = k sin θ cos φ

ky = k sin θ sin φ

kz = k cos θ

where f(kx , ky ) is the pattern in k-space. We multiply the Fourier transform far ﬁeld by the pattern of the Huygens source: j e−j kr (1 + cos θ ) 2λr

(2-14)

When apertures are large, we can ignore this pattern factor. In Eq. (2-13), f(kx , ky ) is a vector in the same direction as the electric ﬁeld in the aperture. Each component

52

RADIATION STRUCTURES AND NUMERICAL METHODS

is transformed separately. The far-ﬁeld components Eθ and Eφ are found by projection (scalar products) from f(kx , ky ) times the pattern factor of the Huygens source [Eq. (2-14)]. If we have a rectangular aperture in which the electric ﬁeld is expressed as a product of functions of x and y only, the integral reduces to the product of two single integrals along each coordinate. The Fourier transform relationships provide insight into pattern shape along the two axes. Large apertures radiate patterns with small beamwidths. An antenna with long and short axes has a narrow-beamwidth pattern in the plane containing the long dimension and a wide beamwidth in the plane containing the short dimension. This is the same as the time and frequency dual normally associated with the Fourier transform. We draw on our familiarity with signal processing to help us visualize the relationship between aperture distributions and patterns. Large apertures give small beamwidths, just as long time pulses relate to low-frequency bandwidths in normal time–frequency transforms. The sidelobes of the pattern correspond to the frequency harmonics of an equivalent time waveform under the Fourier transform and rapid transitions in the time response lead to high levels of harmonics in the frequency domain. Rapid amplitude transitions in the aperture plane produce high sidelobes (harmonics) in the far-ﬁeld response (Fourier transform). Step transitions on the aperture edges produce high sidelobes, while tapering the edge reduces sidelobes and we relate the sidelobe envelope of peaks to the derivative of the distributions at the edges. To produce equal-level sidelobes, we need Dirac delta functions in the aperture that transform to a constant level in the pattern domain. Another example is periodic aperture errors that produce single high sidelobes. When we discuss aperture distribution synthesis, we see that the aperture extent in wavelengths limits our ability to control the pattern. A uniform amplitude and phase aperture distribution produces the maximum aperture efﬁciency and gain that we determine from the following argument. An aperture collects power from a passing electromagnetic wave and maximum collectible power occurs at its peak amplitude response. If the amplitude response somewhere else in the aperture is reduced from the maximum, that portion will collect less power. The amplitude response can be reduced only by adding loss or reﬂecting power in reradiation. The antenna with the highest aperture efﬁciency reﬂects the least amount of power when illuminated by a plane wave. Similarly, if the phase shift from the collecting aperture to the antenna connector is different for different parts of the aperture, the voltages from the various parts will not add in phase. Gain is directly proportional to aperture efﬁciency [Eq. (1-10)]. Therefore, a uniform amplitude and phase aperture distribution has maximum gain. All this assumes that the input match on various aperture distribution antennas is the same. For example, consider the pattern of a uniform aperture distribution in a rectangular aperture a × b. We use the Fourier transform and ignore the polarization of the electric ﬁeld in the aperture. (This assumes that the ﬁeld has a constant polarization or direction.) b/2 a/2 ej k·r dx dy f (kx , ky ) = E0 = E0

−b/2 b/2 −b/2

−a/2 a/2 −a/2

ej kx x ej ky y dx dy

APERTURES: HUYGENS SOURCE APPROXIMATION

53

We separate the integral into a product of two integrals each with the form a/2 ej kx a/2 − e−j kx a/2 a sin(kx a/2) ej kx x dx = = j k kx a/2 x −a/2 On combining the two similar integrals, we have f (kx , ky ) =

ab sin(kx a/2) sin(ky b/2) kx a/2 ky b/2

where kx = k sin θ cos φ, ky = k sin θ sin φ, and kz = k cos θ and k = 2π/λ. The pattern in both planes is given by a k-space function, sin u/u. Figure 2-2 plots this pattern function as a solid curve using kx -space [(ka/2) sin θ ] as the abscissa to produce a universal curve independent of aperture size a. The half-power points occur when sin u 1 =√ u 2

or

u = 1.39156

When we substitute for u, we have in the principal planes πa sin θ = 1.39156 λ By solving for θ , we compute the half-power beamwidth (HPBW): HPBW = 2 sin−1

0.4429λ a

By using the approximation u = sin u for small angles, the half-power beamwidth can be estimated as ◦λ HPBW = 50.76 a

Amplitude, dB

Cosine Uniform

ka/2 sin q

FIGURE 2-2 Universal k-space pattern for the radiation from uniform (solid curve) and cosine (dashed curve) aperture distributions.

54

RADIATION STRUCTURES AND NUMERICAL METHODS

Note that we have ignored the (1 + cos θ )/2 pattern of the Huygens source, which reduces the beamwidth for radiation from small apertures. We discover on Figure 2-2 the pattern nulls occur at integer multiples of π and the ﬁrst sidelobe amplitude is 13.2 dB below the peak. The gain of a uniform amplitude and phase aperture distribution is given by Eq. (1-7), where A is the area of the aperture. Chapter 4 develops amplitude taper efﬁciency for nonuniform aperture amplitude distributions to calculate the gain reduction. Phase error efﬁciency gives the gain reduction due to phase anomalies. Each of these is found from the distribution of aperture ﬁelds. Figure 2-2 plots the pattern of a half cosine aperture distribution as a dashed curve. The distribution peaks in the center and tapers linearly to zero at the edges. Tapering the aperture distribution widens the beamwidth and reduces both gain and sidelobe levels. The pattern beamwidth is 1.342 times wider than the uniform distribution beamwidth. A cosine distribution produces a −0.91-dB amplitude taper loss, and the distribution edge taper causes the sidelobes to fall off at a faster rate. Example Compute the length of the aperture with a uniform distribution that will give a 10◦ beamwidth. 50.76◦ a = 5 wavelengths λ 10◦ We can calculate radiated power by integrating the Poynting vector magnitude over the radiation sphere, but there is an easier way. We assumed that the aperture ﬁelds are free-space waves. The total power radiated is in the aperture, Pr |E|2 Pr = ds Pavg = Uavg = η 4π aperture

where η is the impedance of free space. The radiated electric ﬁeld is e−j kr (1 + cos θ ) f(kx , ky ) 2λr The Poynting vector magnitude is E=j

Sr =

(1 + cos θ )2 |E|2 = |f(kx , ky )|2 η 4λ2 r 2

(2-15)

By combining Eqs. (2-14) and (2-15), we determine directivity: Sr r 2 U (θ, φ) = Uavg Pr /4π 2 j k·r Ee ds π(1 + cos θ ) = λ2 |E|2 ds

directivity(θ, φ) =

(2-16)

By considering electric and magnetic ﬁelds separately in the aperture, we eliminate the requirement that the ratio electric and magnetic ﬁelds are the same as free space used in the Huygens source approximation. Given the ﬁelds in an aperture, we can equate

APERTURES: HUYGENS SOURCE APPROXIMATION

55

them to magnetic and electric currents: Ms = E a × n

Js = n × Ha

(2-17)

where Ea and Ha are the aperture ﬁelds and n is the outward normal. The equivalence theorem [2, p. 113] results in exact solutions by using the total aperture ﬁeld, incident and reﬂected. When using the equivalence theorem, we replace the total ﬁelds present with equivalent currents. The induction theorem equates currents only to the incident ﬁelds on the aperture, which ignores wave reﬂection and results in approximate solutions: Ms e−j k|r−r | Js e−j k|r−r | ds ds A=µ (2-18) F=ε 4π|r − r | 4π|r − r | s

s

We derive the radiated ﬁelds from each distribution of currents by using vector potentials where r is the ﬁeld point and r is the source point in the aperture. These expressions are valid in the near and far ﬁelds. By integrating over only a ﬁnite aperture, we assume zero ﬁelds outside the aperture, while rigorous expressions require integrals over closed boundaries. A planar aperture must extend to inﬁnity, but the ﬁelds outside the aperture are nearly zero and contribute little. 2-2.1 Near- and Far-Field Regions The radiative near- and far-ﬁeld regions are characterized by the approximations made to the integrals [Eq. (2-18)]. The radiative near-ﬁeld region lies between the near ﬁeld, with no approximations, and the far-ﬁeld region. In both approximations the ﬁeld (observation) distance r is substituted for |r − r | in the amplitude term. The vector potentials reduce to µ ε −j k|r−r | Ms e ds A= Js e−j k|r−r | ds (2-19) F= 4πr 4πr s

s

We handle the phase term differently in the two regions. First, we expand the phase term in a Taylor series, 1 |r − r | = r 2 + r 2 − 2r · r = r − rˆ · r + [r 2 − (ˆr · r )2 ] · · · 2r where rˆ is the unit vector in the ﬁeld point direction. We retain the ﬁrst two terms for the far-ﬁeld approximation and the vector potentials become e−j kr ε F= Ms ej k·r ds , etc. (2-20) 4πr s

where we have combined k, the propagation constant, with the unit vector rˆ : k = kˆr = k(sin θ cos φ xˆ + sin θ sin φ yˆ + cos θ zˆ ) The magnetic vector potential integral parallels Eq. (2-20) as in Eq. (2-19). In the radiative near-ﬁeld zone approximation the terms in r 2 are retained and we obtain the following integral for the electric vector potential: kr 2 (k · r )2 e−j kr ε Ms exp j (k · r ) + − (2-21) F= ds 4πr 2rk 2r

56

RADIATION STRUCTURES AND NUMERICAL METHODS

No clear boundary between the three regions exists because the ﬁelds are continuous. Common boundaries are r far ﬁeld L λ where L is the maximum dimension on the aperture. Example Determine the maximum difference between the radiative near- and far-ﬁeld approximations at a point normal to the maximum aperture dimension when r = L2 /λ and r = 2L2 /λ. Normal to the maximum dimension, rˆ · r = 0. The phase difference is

2 kr max 2r

= where rmax

phase difference φ =

L 2

2πL2 8λr

φ = π/4 at r = L2 /λ

and φ = π/8 at r = 2L2 /λ

The usual minimum distance used for antenna patterns is 2L2 /λ, where L is the maximum dimension of the antenna. At that distance, the phase error across the aperture from a point source antenna is π/8. The distance is not sufﬁcient for low-sidelobe antennas [3] because quadratic phase error raises the measured sidelobes. We can use vector potentials in the aperture after determining equivalent currents, but we will ﬁnd it more convenient to use the ﬁelds directly. Deﬁne the following integrals: j k·r Ea e ds g= Ma ej k·r ds (2-22) f= s

s

using the far-ﬁeld approximation. Near-ﬁeld integrals require additional phase terms. Given an aperture, we calculate the vector potentials in terms of Ea and Ha through the currents by using either the equivalence or inductance theorems, and we use the integrals of Eq. (2-22) in the vector potentials. We combine the ﬁelds in the far ﬁeld due to each partial source: E = −j ωA − j ηωF × rˆ For an aperture in the x –y plane, we carry out these steps by using the inductance theorem and obtain the following far ﬁelds from the incident aperture ﬁelds j ke−j kr [fx cos φ + fy sin φ + η cos θ (−gx sin φ + gy cos φ)] 4πr −j ke−j kr Eφ = [(fx sin φ − fy cos φ) cos θ + η(gx cos φ + gy sin φ)] 4πr Eθ =

(2-23)

BOUNDARY CONDITIONS

57

where f and g have been expanded in terms of their x- and y-components and η is the impedance of free space. 2-2.2 Huygens Source The Huygens source approximation is based on the assumption that the magnetic and electric ﬁelds are related as in a plane wave in the aperture: ηgy = fx

and

− ηgx = fy

ηHy = Ex

and

− ηHx = Ey

since

With this approximation, the far ﬁeld [Eq. (2-23)] becomes j ke−j kr (1 + cos θ )(fx cos φ + fy sin φ) 4πr −j ke−j kr Eφ = (1 + cos θ )(fx sin φ − fy cos φ) 4πr Eθ =

(2-24)

The two-dimensional vector Fourier transform f = (fx , fy ) of the aperture electric ﬁeld in the x –y plane determines the far-ﬁeld components. We derive the radiated components by projecting (vector scalar product) this ﬁeld onto the vectors θˆ / cos θ ˆ The transform f expands the ﬁeld in k-space [usually, (kx , ky )]. This normalizes and φ. the pattern and removes the direct dependence on aperture length. We separate out all but f when we consider aperture distributions. We drop the terms for the radiation from a point source and the pattern of a Huygens point source [Eq. (2-14)] and limit our discussions to Huygens sources and far ﬁelds. General aperture ﬁelds require Eq. (2-23), and for any region other than the far ﬁeld, additional phase terms are needed in the transforms [Eq. 2-21)]. 2-3 BOUNDARY CONDITIONS Material boundaries cause discontinuities in the electric and magnetic ﬁelds. The effects can be found by considering either vanishing small pillboxes or loops that span the boundary between the two regions. By using the integral form of Maxwell’s equations on these differential structures, the integrals reduce to simple algebraic expressions. These arguments can be found in most electromagnetic texts and we give only the results. Conversely, we will discover that artiﬁcial boundaries such as shadow and reﬂection boundaries used in geometric optics (ray optics) cannot cause a discontinuity in the ﬁelds because they are not material boundaries. The idea that the ﬁelds remain continuous across the boundary leads to the necessity of adding terms to extend ray optics methods. We discuss these ideas when considering the uniform theory of diffraction (UTD) method used with ray optics. Suppose that we have a locally plane boundary in space described by a point and a unit normal vector nˆ that points from region 1 to region 2. We compute the tangential ﬁelds from the vector (cross) product of the ﬁelds and the normal vector. The ﬁelds

58

RADIATION STRUCTURES AND NUMERICAL METHODS

can be discontinuous at the interface between the two regions if surface magnetic MS or electric current JS densities exist on the surface. nˆ × (E2 − E1 ) = −MS

nˆ × (H2 − H1 ) = JS

(2-25a,b)

The normal components of the ﬁelds change due to the differing dielectric and magnetic properties of the materials and the charges induced on the surface: nˆ · (ε2 E2 − ε1 E1 ) = ρS

nˆ · (µ2 H2 − µ1 H1 ) = τS

(2-26)

with ρS and τS given as electric and magnetic surface charge densities, respectively. Perfect dielectric and magnetic materials can have no currents, which reduces Eq. (2-25) to nˆ × (E2 − E1 ) = 0 nˆ × (H2 − H1 ) = 0 (2-27) Equation (2-27) means that the tangential ﬁelds are continuous across the boundary. These boundary conditions are used in the method of moment analyses to determine currents. The method applies the boundary condition in integral equations to determine the coefﬁcients of the expansion of currents in the sum of basis functions. The currents described as these sums do not satisfy the boundary conditions at all points but do when integrated over a region. This method leads to approximations that will converge as more terms are included in the expansions. When doing analysis we ﬁnd two types of surfaces convenient. We use these surfaces to reduce analysis effort by using planes of symmetry. The ﬁrst one is the perfect electric conductor (PEC). A PEC surface causes the ﬁelds to vanish inside and to have electric currents induced on it: nˆ × E2 = 0

nˆ × H2 = JS

(PEC)

(2-28a,b)

A PEC surface is also called an electric wall. The second surface is the perfect magnetic conductor (PMC) and is a hypothetical surface. Whereas good conductors approximate PEC, there are no PMC materials. The PMC has no internal ﬁelds like the PEC and forces the tangential magnetic ﬁeld to be zero: nˆ × E2 = −MS

nˆ × H2 = 0

(PMC)

(2-29)

A PMC surface supports the hypothetical magnetic current density MS . We ﬁnd that the magnetic wall (PMC) concept simpliﬁes analysis.

I

I

I

M V

I

I

I Electric Currents

FIGURE 2-3

M

M

+ − + −

V Voltage Source

M

M

M

Magnetic Currents

Ground-plane images.

PHYSICAL OPTICS

59

We use images of currents to include material boundaries in analysis. Figure 2-3 illustrates ground-plane images. When we analyze radiation from currents in the presence of a boundary, we include the actual antenna and its image to compute the ﬁelds. The ﬁgure shows an inﬁnite ground plane, but a ﬁnite ground-plane image can be used in the angular region where a reﬂected wave occurs in the ﬁnite plane. We consider this idea further when discussing geometric optics. We can use images in dielectric boundaries provided that we calculate the polarization sensitive reﬂection coefﬁcients to adjust the magnitude and phase of the image. 2-4 PHYSICAL OPTICS Physical optics uses things that can be measured. We can measure both currents and ﬁelds, but auxiliary vector potentials have no physical reality, only mathematical artifacts that simplify Maxwell’s equations. Nevertheless, the auxiliary vector potentials provide simple models for problems that enable simple mental pictures, as shown earlier, but we cannot easily formulate them into a systematic analysis tool for antenna problems. The physical optics analysis method combines the use of Green’s functions to calculate ﬁelds radiated by a given distribution of currents and then uses boundary conditions to determine the currents induced on objects due to incident ﬁelds. We compute the effects of a mounting structure by inducing currents on it and adding their radiation to the antenna pattern. The method assumes that radiation from the induced currents on the structure does not change the initial currents. We start analyses from either currents or incident ﬁelds and work from those. The resonant structure of many antennas determines the approximate current distribution that we normalize to the radiated power. We calculate the ﬁelds from these currents. Physical optics can use an iterative technique to calculate incremental currents induced on the original radiators and improve the solution, but we usually just sum the radiation from the original currents to the radiation from the induced currents. The second starting point for physical optics can be incident ﬁelds. These could be plane waves or could be ﬁelds found from the measured radiation patterns of antennas: for example, the pattern of a reﬂector feed. We add the radiation from the induced currents to the incident waves. 2-4.1 Radiated Fields Given Currents The radiated ﬁelds can be found from distribution of the electric and magnetic currents by the use of dyadic Green’s functions that contain source and ﬁeld coordinates. We sometimes refer to the Green’s functions as vector propagators or transfer functions between currents and ﬁelds. We calculate the ﬁelds from integrals over the source points of the dot (scalar) product between the dyadic and current densities. The dyadic Green’s function contains both near- and far-ﬁeld terms and requires slightly different expressions for the electric and magnetic ﬁelds. The general propagator from electric and magnetic currents has separate terms for electric and magnetic currents, which when used with surface patch currents can be reduced to short subroutines or procedures easily programmed [1]: (2-30) E(r) = GEJ (r, r ) · J(r ) dV + GEM (r, r ) · M(r ) dV H(r) = GHJ (r, r ) · J(r ) dV + GHM (r, r ) · M(r ) dV (2-31)

60

RADIATION STRUCTURES AND NUMERICAL METHODS

These expressions integrate over the currents located at source points r for a dyadic Green’s function that changes at each ﬁeld point r and source point r . Although these Green’s functions are valid at all ﬁeld points in space both near and far ﬁeld, they are singular at a source point. Only retaining terms with 1/R dependence for the far ﬁeld greatly simpliﬁes the expressions. When ﬁelds are incident on a perfect electric conductor (PEC), the combination of incident and reﬂected tangential magnetic ﬁelds induces an electric current density on the surface. The ﬁelds inside the conductor are zero. We assume locally plane surfaces on patches and compute currents that satisfy the boundary condition. Given the local unit normal nˆ to the surface, the induced current density is given by JS = nˆ × (Hincident + Hreﬂected ) Hincident = Hreﬂected JS = 2nˆ × Hincident

(2-32)

The reﬂected magnetic ﬁeld equals the incident magnetic ﬁeld because the ﬁeld reﬂects from the conductive surface. The sum of the tangential electric ﬁelds must be zero. Because the reﬂected wave changes direction, the vector (cross) product of the electric and magnetic ﬁelds must change direction. The reﬂected tangential electric ﬁeld changes direction by 180◦ , so the tangential magnetic ﬁeld must not change direction because the Poynting vector changed its direction. Equation (2-32) is the magnetic ﬁeld equation applied on a PEC. Equation (2-25b) is the general magnetic ﬁeld equation at a boundary. Physical optics starts with a given current distribution that radiates, or the measured pattern of an antenna. When an object is placed in the radiated ﬁeld, the method calculates induced current on the object to satisfy the internal ﬁeld condition. For example, PEC or PMC have zero ﬁelds inside. When we use simple functions such as constant-current surface patches, the sum of the radiation from the incident wave and the scattered ﬁelds from induced surface currents produces only approximately zero ﬁelds inside. As the patch size decreases, the method converges to the correct solution. To obtain the radiated ﬁeld everywhere, we sum the incident wave and scattered waves. The ﬁelds radiated by the induced currents produce the shadow caused by the object. With geometric optics techniques such as UTD, the object blocks the incident wave and we determine the ﬁelds in the shadow regions from separate diffraction waves. In physical optics the incident wave continues as though the object were not present. Only geometric optics techniques use blockage. We can calculate the ﬁelds radiated from antennas in free space or measure them in an anechoic chamber that simulates free space, but we mount the antenna on ﬁnite ground planes, handsets, vehicles, over soil, and so on, when we use them. Physical optics is one method of accounting for the scattering. We show in later chapters that the mounting conﬁguration can enhance the patterns. 2-4.2 Applying Physical Optics In this book we do not discuss how to develop numerical techniques, but it is important to understand how to apply methods. Whether you develop your own codes or use commercial codes, certain rules should be applied. Consider Eq. (2-32). The normal to

PHYSICAL OPTICS

61

the surface points in the direction of the incident wave: outward. If the normal pointed inward, the sign of induced electric current density would change. Most codes have made the assumption that the normal points outward, but some codes may check on the direction of the normal relative to the incident wave and make the necessary sign change. We must keep track of the direction of the normal, and it may be necessary to rotate the normal depending on the expected direction of the incident wave. If an object can have radiation from both sides, it may be necessary to use two objects in the analysis. Many codes store each object as a separate entity in a disk ﬁle. In some cases we need to store an object multiple times. Take, for example, a Cassegrain dual reﬂector. The feed antenna illuminates the subreﬂector and induces currents on it. These currents radiate and excite currents on the main reﬂector. When the main reﬂector–induced currents radiate, the subreﬂector intercepts or blocks part of the ﬁelds. We account for this blockage by using a second subreﬂector object on which the code calculates a new set of induced currents by using the main reﬂector currents as the source. We could add these currents to the existing disk ﬁle object or merely keep the second object. We want to keep the second object separate so that we can calculate additional currents induced on the main reﬂector using these currents as sources. These currents will be reduced from the initial set, but they are an important contribution to the ﬁelds radiated behind the reﬂector. This example illustrates iterative PO. When objects face each other signiﬁcantly, iterative PO is necessary to calculate correct patterns. The method converges rapidly in most cases. Figure 2-4 illustrates the geometry of a corner reﬂector. A half-wavelength-long dipole is placed between two metal plates usually bent to form a 90◦ angle. We can use other angular orientations between the plates, but this is the usual design. The ﬁgure does not show the feed line to the dipole, which usually starts at the juncture of the two plates and runs up to the dipole. This feed line contains the balun discussed in Section 5-15. Although the ﬁgure shows the plates as solid, many implementations use metal rods to reduce weight and wind loading. The analysis starts with assumed currents on the dipole. We divide the plates analytically into small rectangular patches, which can be small (≈ λ/8 to λ/4) on a side since it takes only a few to cover the plates. You should repeat the analysis with

FIGURE 2-4

Corner reﬂector with a dipole located between two ﬂat plates.

62

RADIATION STRUCTURES AND NUMERICAL METHODS

different-sized current patches to determine if the analysis has converged. In a similar manner, we break down the current on the dipole into short linear segments, each with constant amplitude. By using a near-ﬁeld version of Eq. (2-31), we calculate the magnetic ﬁeld incident on each patch on the plates. This ﬁeld induces electric currents on the plates calculated from Eq. (2-32). Remember that we combine radiation from the source dipole with that radiated from the induced currents to reduce the radiation behind the antenna. The currents were induced to satisfy the boundary condition of the plate, but only with both radiations present. Figure 2-5a illustrates this process of inducing currents. Figure 2-6 shows the antenna pattern calculated using these currents. The E-plane pattern drawn as a solid line produces a null at 90◦ because the dipole pattern has this null. The plates cause the narrowing of the beam in the H -plane. The plates reduced the back radiation to −22 dB relative to the forward radiation, called the front-to-back ratio (F/B). The gain has increased from the 2.1 dB expected from a dipole to 9.3 dB. An equivalent geometric optics analysis uses two images in the plates, as shown in Figure 2-5b, for the analysis. If you look at Figure 2-4 or 2-5, you should notice that the two plates face each other. Currents on one plate will radiate toward the other plate and induce another set of currents on it. We could ignore these induced currents if the radiation was insigniﬁcant, but to produce correct patterns we must include them. The solution to this problem calls for an iterative technique where we calculate the radiation from the currents on the ﬁrst plate and induce incremental currents on the second plate. These incremental currents produce further radiation that induces additional currents on the other plate. The method converges rapidly. Figure 2-7 gives the antenna pattern after the iterations have been completed and we include radiation from all currents. The actual F/B ratio of the antenna is 29 dB, and the additional currents increased the gain by 0.7 dB to 10 dB. Adding the two plates in the original analysis increased the gain by 7.2 dB, whereas the iterative technique had a much smaller effect. Figure 2-8 illustrates the iterative technique and shows that the equivalent geometric optics analysis adds a third image to represent the reﬂection between the plates. Remember when you mount the antenna in an application, the structure will change the realized pattern, but the high F/B ratio reduces this effect. The mounting structure used when measuring the antenna changes the pattern as well, which limits our knowledge of the real pattern.

Dipole Currents

Image (a)

Image (b )

FIGURE 2-5 Cross-sectional view of a corner reﬂector: (a) magnetic ﬁeld radiated from a dipole induces currents on plates; (b) plate currents replaced with image dipoles.

PHYSICAL OPTICS

E-plane

63

H-Plane

FIGURE 2-6 Pattern calculated from a combination of dipole and plate currents in a corner reﬂector with 1 × 0.9λ plates without induced current iteration.

Physical optics can determine the impedance effects of the limited images in the ground planes, such as the corner reﬂector. The local nature of impedance effects allows the use of images to calculate the mutual impedance effects of ground planes. We use impedance calculations not only to determine the bounds of ground-plane effects on input impedance, but to calculate the total power radiated by the antenna. The images (excited currents on ground planes) radiate but do not receive input power. A ground plane at least λ/2 on a side located about λ/4 away from the antenna produces nearly the same impedance effects as an inﬁnite ground plane, but the ground plane alters the radiation pattern greatly because it restricts possible radiation directions. It has commonly been thought that physical optics could compute the ﬁeld only in the main beam pattern direction of a paraboloidal reﬂector. The method can determine this pattern region accurately by using only a few patches, each one being many wavelengths on a side. As the processing power of computers increases, the patch size can be shrunk until PO can calculate the pattern in every direction, including behind the reﬂector. It is important to remember to include the feed pattern behind the reﬂector even though its radiation is obviously blocked by the main reﬂector. Physical optics uses induced currents to cancel the ﬁelds inside objects when the incident ﬁelds and the radiation from the induced currents are added. We can calculate the pattern behind a reﬂector using UTD (GTD), the uniform (geometric) theory of diffraction. This

64

RADIATION STRUCTURES AND NUMERICAL METHODS

E-Plane

H-plane

FIGURE 2-7 Pattern of corner reﬂector with 1 × 0.9λ plates with induced current iteration equivalent to multiple reﬂectors between the plates.

Dipole

Image

Image

( a)

Image (b)

FIGURE 2-8 (a) Wall currents on plates radiate magnetic ﬁelds that induce additional currents on facing plates; (b) added induced currents equivalent to additional image dipole.

geometric optics-based method blocks the radiation from the feed and uses diffractions from the rim edge to calculate the pattern behind the reﬂector. We discuss UTD in Section 2-7. A comparison of UTD and physical optics calculations [1,4] of the pattern behind shows that the two methods match. The dashed curve of Figure 2-9 plots the results of the PO analysis of a 20λ-diameter centrally fed paraboloidal reﬂector. The feed antenna radiation tapers to −12 dB at the

65

Gain, dB

PHYSICAL OPTICS

Pattern Angle

FIGURE 2-9 Physical optics analysis of a 20λ-aperture-diameter paraboloidal reﬂector (dashed curve) compared to analysis that includes PTD (solid curve).

reﬂector rim. Figure 2-9 shows the feed power spillover peaking at angles off the boresight near 100◦ . PO analysis computes the currents on a patch by assuming that it is embedded in an inﬁnite plate. The reﬂector rim violates this assumption and we need extra terms to calculate the pattern behind the reﬂector accurately. Adding PTD (the physical theory of diffraction) to PO improves the match between the two methods behind the reﬂector as shown by the solid curve on Figure 2-9. PTD handles caustic regions of PO in a manner similar to the equivalent current method based on diffraction coefﬁcients of UTD with geometric optics for shadow and reﬂection boundaries. For this example, the additional PTD currents add with the same phase because of the symmetry of the reﬂector geometry and produce the maximum effect. The PTD currents on the rim of an offset reﬂector will not add and produce a peak effect behind the reﬂector but will produce a more defuse effect. We only need PTD over a limited pattern angular range to reduce error, and the cost of implementing the ﬁx may exceed the necessity of knowing the pattern in these regions. Similarly, UTD needs the addition of edge currents for accurate calculation of the radiation near 180◦ , behind the reﬂector. Although any model for the feed pattern can be used with PO, results matching UTD exactly occur only over all regions of the back radiation when the feed satisﬁes Maxwell’s equations in the near and far ﬁelds [4, p. 212]. One such feed is the Gaussian beam approximation. Again, like PTD ﬁxes, the small errors when using other feed antenna approximations occur only at limited pattern regions that may be unimportant. 2-4.3 Equivalent Currents We can relate the concept of equivalent currents to physical optics. In this case we generate an artiﬁcial surface that covers a source of radiation. The incident ﬁelds generate surface electric and magnetic current whose radiation cancels the internal

66

RADIATION STRUCTURES AND NUMERICAL METHODS

ﬁelds and generates the external pattern. We use these at the apertures of antennas such as horns. By using the dyadic Green’s functions we can calculate the near-ﬁeld patterns and the coupling between antennas when the assumption is made that the presence of a second antenna does not alter the aperture ﬁelds. Given the outward normal n, ˆ we calculate the equivalent currents by nˆ × Eincident = −MS

nˆ × Hincident = JS

(2-33)

We must use both electric and magnetic current densities on the surface to replace the internal ﬁelds. If the ratio of the electric ﬁeld to the magnetic ﬁeld equals the impedance of free space (376.7 ), the combination of the two currents produces the radiation of the Huygens aperture source when used with the dyadic Green’s function. We use equivalent currents for a variety of analyses over ﬂat apertures such as horns and paraboloidal reﬂectors, but they can also be used with curved structures or apertures. We can, for example, use equivalent currents for calculation of the effects of radomes. Locally, we assume that the incident waves are plane waves and use boundary conditions to calculate reﬂected and transmitted waves. It is necessary to separate the incident wave into parallel and perpendicular polarizations, the ray-ﬁx representation discussed in Section 2-7.8. These polarizations have differing reﬂection and transmission coefﬁcients. We generate one surface on the inside of the radome and another on the outside. We use locally free-space waves for the reﬂected and transmitted waves lying outside the radome. Both these waves can be replaced with equivalent currents. The equivalent currents produce null ﬁelds inside the radome when combined with the incident wave radiation [4, p. 155]. Including these equivalent currents in a PO analysis, we add the effect of the radome. Equivalent currents can also be used with lenses. We use the incident waves combined with the idea of locally plane waves to calculate reﬂected and transmitted waves at each surface and replace them with equivalent currents. We include the dielectric constant of the lens in the dyadic Green’s functions for the internal radiation of the lens to calculate the ﬁelds at the second surface. We apply locally plane waves at the second surface to determine the transmitted and reﬂected rays and then replace them with equivalent currents. Because the lens has internal reﬂections, we need to apply an iterative PO analysis to calculate the multiple reﬂections between the two surfaces. The method converges rapidly because the internal reﬂections are small. 2-4.4 Reactance Theorem and Mutual Coupling In Section 1-14 we discussed how the coupling between two antennas can be found from reactance. Given a transmitting antenna that generates a ﬁeld at the receiving antenna, the reactance is described by an integral equation [5]: reactance =

(Et · Jr − Ht · Mr ) dV = t, r

(2-34)

The volume integral is over the receiving antenna currents, but it is often reduced to a surface or line integral. A second form of Eq. (2-34) uses the ﬁelds radiated by

METHOD OF MOMENTS

67

both antennas. Given a surface that surrounds the receiving antenna, the integral for reactance is taken over this surface: (Er × Ht − Et × Hr ) · ds = t, r

reactance =

(2-35)

Sr

The differential normal ds is pointed away from the receiving antenna. When we represent the two antennas and the transmission between them as an impedance matrix, it implies that we know the input currents to both antennas. By expressing the coupling as an impedance matrix, we compute mutual impedance from the reactance integral: −1 t, r (2-36) Z12 = I1 I2 Antennas that we describe by input currents only have electric current densities excited on their surfaces. The mutual impedance formula using reactance reduces to Z12 =

−1 I1 I2

Et · Jr dV

(2-37)

Vr

The volume integral reduces to a line integral in most cases. Antennas with given input voltages such as slots can be described using magnetic currents and we use a mutual admittance matrix for the antenna pair: Y12 =

1 −1 · reactance = Vt Vr Vt Vr

Ht · Mr dV

(2-38)

Vr

By using reciprocity antennas made of linear, isotropic materials, we have equal crossmatrix terms: Z12 = Z21 and Y12 = Y21 (2-39) We calculate self-impedance terms by integrating over the surface of the antenna: for example, the radius of a dipole with the source of the ﬁeld located at the center of wires or slots. 2-5 METHOD OF MOMENTS The method of moments (MOM) [6] expands the currents on an antenna (or scattering object) in a linear sum of simple basis functions. The approximate solution is a ﬁnite series of these basis functions: N fa = ai fi (2-40) i=1

We compute the coefﬁcients by solving integral equations to satisfy boundary conditions on the surface of the antenna (or object). The integral equation can be expressed in the form Lfa = g, where L is a linear operator, usually a scalar product using an integral, fa the unknown currents given by Eq. (2-40), and g the known excitation

68

RADIATION STRUCTURES AND NUMERICAL METHODS

or source function. We substitute the summation of Eq. (2-40) into the linear operator equation and use the scalar product integral to calculate the terms in a matrix equation. The solution of the matrix equation determines the coefﬁcients of current expansion. The MOM produces ﬁlled matrices that require time-consuming numerical methods for inversion. The art of the MOM is in choosing basis functions and deriving efﬁcient expressions for evaluating the ﬁelds using the basis function currents. Common basis functions are simple staircase pulses, overlapping triangles, trigonometric functions, or polynomials. The method does not satisfy boundary conditions at every point, only over an integral average on the boundaries. By increasing the number of basis functions, the method will converge to the correct solution. We need to judge how many terms are required for an adequate engineering evaluation. Spending excessive time on the solution cannot be justiﬁed if it greatly exceeds our ability to measure antenna performance accurately using real hardware. 2-5.1 Use of the Reactance Theorem for the Method of Moments We can use the reactance theorem to generate a moment method solution to the currents on a thin-wire antenna. Thin-wire solutions assume that there are no circumferential currents and reduces the problem to ﬁlamentary currents. An electric ﬁeld integral equation (EFIE) satisﬁes the boundary condition of Eq. (2-25a), a zero tangential ﬁeld at the surface of the wires, but it does not seem explicit in the derivation. The reactance theorem produces an impedance matrix whose inversion yields the coefﬁcients of the current expansion [7]. Similar to many other methods, the Green’s function has been solved explicitly to reduce run time. This method [7] uses overlapping sinusoidal currents on V-dipoles as basis function currents and uses the Green’s function to calculate the radiation from one V-dipole at the location of a second V-dipole. Both the radiating and receiving dipoles use the same expansion function. Galerkin’s method uses the same weighting (or testing) function as the basis function and yields the most stable solutions. The reactance equation (2-37) calculates the mutual impedance between the two dipoles when each has unity current. We compute self-impedance by spacing a second V-dipole one radius away and by using the reactance theorem to calculate mutual impedance, a technique equivalent to the induced EMF method. The scalar (dot) product between the incident vector electric ﬁeld and the current density along the dipole reduces the vectors to scalars that can be integrated. The current density acts as the testing or weighting function for the method of moments. Performing the integration means that the current density only satisﬁes the zero tangential electric ﬁeld boundary condition in an average sense. If series impedances are placed in the V-dipole, their impedance is added to the diagonal elements of the mutual impedance matrix. To excite the structure, we place a delta voltage source in series with the Vdipole terminals. The solution for the currents can be found by inverting the matrix equation and using the voltage excitation vector starting with the matrix equation [Zmn ][Im ] = [Vn ]

(2-41)

After computing the matrix inverse and specifying the input voltage vector, the complex current values are found on the structure: [In ] = [Zmn ]−1 [Vm ]

(2-42)

METHOD OF MOMENTS

69

Given the input voltage and the solution for the currents, the input impedance can be calculated. Similarly, the far- and near-ﬁeld patterns can be calculated by using Eqs. (2-30) and (2-31) of the dyadic Green’s function. The code must satisfy Kirchhoff’s current law at the junction between groups of Vdipoles, which adds a constraint to the currents. Because an overlapping sinusoidal basis function closely follows the actual currents normally excited on dipoles, the segments can be on the order of a quarter-wavelength long or more and yield acceptable results. Basis functions that closely follow expected current distributions are sometimes called entire domain functions. These reduce the size of the matrix to be inverted but require more complicated calculations for matrix terms and radiation. Although the concept of a V-dipole was expanded to a V rectangular plate [8], the method is only a subset of general integral equation solutions. This approach generates a simple impedance matrix formulation easily understood from an engineering point of view. 2-5.2 General Moments Method Approach The method of moments can solve other types of electromagnetic problems: for example, electrostatic problems involving charges and dielectrics [9]. These solutions can determine the characteristic impedance of transmission lines useful in the design of antenna feeders. All moment method solutions are found from the solution of integral equations over boundary conditions. The boundary conditions can be either the tangential electric ﬁeld (EFIE) or magnetic ﬁeld (MFIE) conditions given by Eq. (2-25a,b) or a combination applied using an integral scalar product. We need a combination for closed bodies near an internal resonance frequency (resonant cavity) because the solutions exhibit resonances that make the solution invalid over a narrow frequency range. The method of moments can be applied to dielectric bodies when we use the constitutive relations of Eqs. (2-25) and (2-26), where the formulations for dielectric bodies use either volume or surface integrals [9]. Consider the use of the electric ﬁeld integral equation (EFIE) with metal surfaces. We expand the currents on the objects using basis functions Bm (r ) with coefﬁcients Im : J(r ) =

Im Bm (r )

(2-43)

The basis functions can be applied over a limited range of the structure in piecewise linear functions, which can be staircase pulses, overlapping triangular functions, or sinusoidal basis functions, whereas multiple functions can be applied over the whole or part of the structure for entire domain basis functions. For example, these could be a sum of sinusoidal functions which form a Fourier series representation. On a PEC surface the tangential electric ﬁeld vanishes [Eq. (2-28a)]. At ﬁeld point r along the surface S, nˆ × [Eincident (r) + Escattered (r)] = 0 M Im Bm (r ) · G(r, r ) ds Escattered = m=1

(2-44)

s

We can only satisfy Eq. (2-44) using a ﬁnite sum in the average sense of an integral. Since the integral and summation operate on a linear function, we can interchange them.

70

RADIATION STRUCTURES AND NUMERICAL METHODS

We introduce weighting (or testing) vector functions tangent to the surface Wn (r) and take the scalar (dot) product of this vector with the sum of electric ﬁelds. This limits the result to the tangential component of the electric ﬁeld: [Eincident (r) · Wn (r) + Escattered (r) · Wn (r)] ds = 0

(2-45)

s

We identify the weighted integral of the incident ﬁeld with the source and weighted integral of the ﬁeld radiated by the basis functions (scattered ﬁeld) as the impedance matrix terms. The integrals over the boundaries are one form of scalar product represented by · notation. Using unity current on each basis function, we calculate the matrix terms by using the scalar product:

Bm · G(r, r ) ds , Wn (r) =

Zmn = s

Vn = −Eincident (r), Wn (r) = −

Bm · G(r, r ) · Wn (r) ds ds (2-46)

s

s

Eincident (r) · Wn (r) ds

(2-47)

s

The combination of Eqs. (2-46) and (2-47) when integrated over each portion of the source gives a matrix equation: [Zmn ][Im ] = [Vn ]

(2-48)

The weighting functions could be as simple as pulse functions, overlapping triangular functions on lines or surfaces (rooftop), piecewise sinusoidal functions, or others. The type of basis functions determines the convergence more than the weighting (testing) functions, which only determine the averaging. Realize that the moment method converges to the exact solution when we increase the number of basis functions, but it is a matter of engineering judgment to determine how many terms give acceptable answers. Equation (2-47) deﬁnes the source voltage occurring over a segment when the formulation uses a piecewise function expansion. The incident voltage is the weighted integral of the incident electric ﬁeld. For example, the NEC formulation applies an excitation voltage across one segment. The reaction integral formulation of Section 25.1 applies a voltage source at the end of a segment. The modeling of sources is an important part of the art in the method of moments. The expansion of Eq. (2-44) is only one possible moment method solution. We could use the boundary condition on the magnetic ﬁeld, a combination of the electric and magnetic ﬁeld conditions on a PEC. If the surface has ﬁnite conductivity, the boundary conditions are modiﬁed. The moment method is a general method that computes approximate solutions to the currents. Unlike physical optics, the currents do not have to be assumed beforehand but are found as a ﬁnite series approximation. Antenna designers discover that adequate codes are available for most problems. Moment method solutions are typically limited to objects only one or two wavelengths in size, although any method can be stretched. Analysis of large structures becomes intractable because of the large amount of computer memory required and the length

METHOD OF MOMENTS

71

of time needed to calculate the solution. Coarse models may not give totally accurate results but can be useful in determining trends. Given these ideas, remember that physical models can be built that solve the electromagnetic problem instantaneously. We found that it takes considerable time to learn any code, and a new code has to offer considerable advantages or solve problems that the present one cannot solve before we invest our time. 2-5.3 Thin-Wire Moment Method Codes Thin-wire codes that assume only ﬁlamentary currents are readily available. We have experience with NEC, the Richmond code (ASAP), and AWAS [10], a commercial code. All have advantages, but they take time to learn. A commercial code with a graphical interface makes the input and output easier: for example, for NEC. These pay for themselves quickly by saving time. NEC can include plates, but since it uses a MFIE (magnetic ﬁeld integral equation) for them, it is limited to closed bodies. When accuracy becomes important, it is necessary to decrease the segment length and increase their number. These codes use matrix inversion with calculation time proportional to N 3 and a matrix ﬁll time proportional to N 2 . Run time increases enormously as the number of segments increases. The commercial code AWAS determines the segmentation, while the user of NEC must specify it. The rule is to use at least 10 segments per wavelength, but initial analysis can tolerate the errors due to using fewer segments. The segments should be longer than the diameter, and care must be taken that the segments do not overlap because the radius of the wires is too large. Solid objects, such as plates, can be modeled as wire frames, with the rule that the perimeter of the wire equal the spacing between the wires [11]. This rule can be violated, but a test of the convergence should be made. When we model slots in a solid object, we cannot apply the perimeter equalspacing rule because the slot will disappear. These codes compute the radiation pattern more accurately than the input impedance due to simplistic source models, and we may have to build the antenna to determine the true input impedance. Of course, an antenna with a good input impedance response that does not have the required pattern is useless. We can reduce NEC run time if the antenna has symmetry with multiple inputs. The code reduces input by allowing the user to specify symmetry. For example, a multiarm spiral analysis requires only the input of one arm. The various mode voltages are entered after the basic structure impedance matrix has been solved. If an object has M-way symmetry, the matrix ﬁll time is reduced by M 2 and the solution time by M 3 . The various voltage modes can be applied afterward. If we add another wire segment after specifying symmetry, the symmetry is destroyed and the program uses the full matrix. The only advantage we gain is in specifying the model because the program solves the full matrix instead of the reduced matrix. 2-5.4 Surface and Volume Moment Method Codes Antennas made of plates or containing ﬁnite plate ground planes can be solved by using wire meshing of a thin-wire code. The method of moments code has been extended to plates [12,13] using a rooftop basis function on both rectangular and triangular patches. The number of basis functions (i.e., matrix size) grows rapidly. One solution is to use entire domain basis functions. These require more complicated integrals, but

72

RADIATION STRUCTURES AND NUMERICAL METHODS

they reduce the matrix size. Dielectric portions of the problem lead to either volumetric integrals or various forms of surface integrals that use equivalent currents to replace the internal ﬁelds [9,14]. These problems lead to a variety of boundary conditions solved using a ﬁnite series of basis function and integral equations to satisfy those boundary conditions approximately. MOM analysis of antennas mounted on dielectric substrates requires special techniques. Commercial codes determine the currents ﬂowing on these antennas while accounting for the dielectric. Often, Green’s functions are found numerically, which increases the execution time. Since the currents are located on the surface and the integrals of the boundary conditions are over the same surface, the singularity of the Green’s function causes a numerical problem. For example, the free-space Green’s function has the term 1/|r − r |, which becomes inﬁnite on the surface. Spectral domain methods remove the singularity by using a sum of current sheets on the surface as an entire domain basis function. A uniform plane wave propagating at an angle to the surface excites the current sheet. The actual current ﬂowing on the metal portions is expanded as a sum of these current sheets [15, p. 208ff; 16]. The uniform current sheets are expanded in a spatial Fourier transform as well as the Green’s function, and the MOM problem is solved. The Fourier-transformed Green’s function no longer has the singularity. When the metallization can be expressed as an inﬁnite periodic structure, the current is expanded as a Fourier series. The inﬁnite periodic structure is used with frequency-selective surfaces and inﬁnite arrays. In this case the ﬁelds and currents are expanded in Floquet modes (harmonics). 2-5.5 Examples of Moment Method Models Figure 2-10 demonstrates the use of a wire mesh to replace a solid plate. We located a resonant (≈ λ/2) dipole λ/4 distance over a λ-wide ground plane in the H -plane and offset 3/8λ from one edge. This is repeated in Figure 2-20 using GTD analysis. The rods only run parallel to the dipole because cross wires do not have currents induced on them in the ideal world of analysis. The circumference of the rods equals the spacing between the rods and forms an equivalent solid plate. An actual antenna could use smaller-diameter rods and work as effectively as the solid plate and would reduce weight and wind loading. NEC analysis produces the same pattern as the GTD analysis of Section 2-7.2, except that the E-plane size of the rods alters the backlobe predicted by GTD to some extent, because that analysis assumes inﬁnite-length rods.

FIGURE 2-10 Use of a wire mesh to replace a solid plate for dipole over a ground plane in a MOM calculation.

METHOD OF MOMENTS

73

Over most of the pattern angles the two analyses produce identical results. The NEC analysis accounts for the mutual impedance between the dipole and its image in the ﬁnite ground plane. For impedance calculations a small ground plane gives almost the same reaction to the antenna as an inﬁnite ground plane. Figure 2-11 shows a wire frame model of a cell phone. The model contains more wires than necessary for λ/10 spacing, but more wires improve the geometry match. When using crossed wires that shield both polarizations, we reduce the wire circumference in half since the wires approach the squares from four sides. The small wire antenna must be connected to the wire grid of the model to generate proper currents on the box. Either we restrict possible locations of the antenna or we must distort the wire grid locally. You should write an automatic grid generator if you use this analysis often. Consider that you need to specify whether an edge wire should be generated when two plates share the same edge. The hand holding the cell phone and the head nearby have signiﬁcant effect on the antenna performance. The model given in Figure 2-11 has limited use. We need either a moment method analysis, such as WIPL-D, which includes volume dielectric structures, or FDTD, which can include complex material structures to model the head and produce good results. Figure 2-12 illustrates a wire frame model of an airplane used for low-frequency analysis. Antennas mounted on free-ﬂying models such as airplanes or spacecraft will excite the structure. Electrically, small antennas can excite the entire vehicle as an antenna. For example, a small antenna mounted on a large ground plane that would produce vertical polarization can excite the wings or fuselage and the entire system will

FIGURE 2-11 Wire frame MOM model of a cellular telephone handset with an antenna connected to the mesh.

74

RADIATION STRUCTURES AND NUMERICAL METHODS

FIGURE 2-12 Wire frame MOM model of an airplane.

radiate horizontal polarization. Models similar to Figure 2-12 can eliminate surprises. The model restricts antenna mounting locations to the wire positions and may require local distortions of the grid. Moment methods can include solid plates. Figure 2-13 shows an open waveguide horn analysis that uses a combination of plates and a single-feed wire monopole [12]. Locating the monopole or a small dipole inside the waveguide produces excitation of the waveguide mode that feeds the horn. Even though the model does not necessarily produce accurate impedance information, the model accurately calculates the pattern generated by the currents excited in the walls. We can either use an aperture method

FIGURE 2-13 MOM model of a pyramidal horn using ﬂat plates fed by a small dipole. (From [14, p. 229].)

METHOD OF MOMENTS

75

PEC (a)

PMC

(b)

FIGURE 2-14 Use of electric and magnetic walls to reduce the model size in MOM analysis of a pyramidal horn: (a) PEC wall divides the horn; (b) PMC wall divides the horn.

for the horn that replaces the aperture ﬁelds or use the currents excited in the walls to calculate the pattern. Either method works for the front lobe. The moment method calculation requires signiﬁcantly greater calculation time but produces results that better match measurements in all directions. Figure 2-14 demonstrates how to reduce calculation time by using planes of symmetry in a moment method analysis. In this case the small dipole feed is separated by two equally fed closely spaced dipoles. The right–left symmetry of the antenna allows reduction of the model by half. A vertical PMC wall divides the antenna into two parts, with only one remaining in the analysis. A horizontal PEC conductor divides the remaining model in half because halfway between the dipole feed is a virtual short circuit. Figure 2-14 contains only one-fourth the size of the original problem. Since matrix inversion requires N 3 calculations for an N × N matrix, dividing the analytical model down to one-fourth size reduces this calculation by a 64 : 1 factor. This also reduces the matrix element (ﬁll time) calculations by 16 : 1. Reducing the model by using symmetry planes enables the solution of larger problems and reduces calculation time. Analyses in later chapters use the moment method to predict antenna performance. Wire frame and plate analyses determine vehicle and mounting structure pattern effects. The moment method produces excellent analyses because it determines the approximate

76

RADIATION STRUCTURES AND NUMERICAL METHODS

current distribution as a sum of simple basis functions and we need not start with an assumed current distribution on the antenna.

2-6 FINITE-DIFFERENCE TIME-DOMAIN METHOD The ﬁnite-difference time-domain (FDTD) method solves the coupled Maxwell’s curl equations directly in the time domain by using ﬁnite time steps over small cells in space. The method reduces the differential equations to difference equations that can be solved by sets of simple equations. The method alternates between the electric and magnetic ﬁelds solved at locations a half-step apart because central differences are used to approximate derivatives. A 1966 paper by Yee [17] described the basic method that many authors have improved upon, but the original method remains the approach of choice. FDTD can solve many types of electromagnetic problems, of which antenna analyses are only one type. Computer memory and speed limit the size of problems that can be solved, but larger and larger problems can be solved as the cost of computing keeps reducing. Besides antenna problems, the method is applied to microwave circuits, biological interaction with electromagnetic waves, optics, and radar cross-section problems. The number of uses expands every day. The method allows each cell to be made of different materials, leading to the solution of volumetric complex structures. The solution of the equations is robust and the errors are well understood. Currently, the method solves moderately small antenna problems on the order of a few wavelengths. Of course, faster and larger computers can solve larger problems, especially if the analyst has patience. FDTD handles microstrip antennas with their complex layering of dielectrics, including a ﬁnite ground plane without the use of complex Green’s functions required of frequency-domain solutions. The interaction of antennas with the near environment, such as the effect of the head on cellular telephone handsets, can be solved. In this case the complex electromagnetic properties of the head can be described as cells each with different electrical properties. In addition to giving a solution to the radiation pattern and allowing characterization of the communication system, it can provide insight into the radiation safety concerns of users. The method handles the solution of the interaction of antennas with the human body in a straightforward manner for prediction of biomedical applications, such as electromagnetic heating for cancer treatment. Learning to apply the technique, whether formulating your own routines or using a commercial code, will yield insight for design. The method can produce time-domain animated displays of the ﬁelds that show radiation centers and where the ﬁelds propagate, but the user must learn to interpret these new displays. It will be worth your effort to learn this task. The time-domain responses using impulse signals can produce solutions over a wide band of frequencies when converted to the frequency domain using the discrete Fourier transform (DFT). The only drawback is the computer run time required. 2-6.1 Implementation By using a direct implementation of Maxwell’s curl equations in the time domain, you do little analytical processing of the equations. No vector potential or Green’s

77

FINITE-DIFFERENCE TIME-DOMAIN METHOD

functions are developed as in frequency-domain methods. Although the antenna may be volumetrically complex and contain many different materials, the method yields sparse matrices rather than the dense matrices produced by moment methods. It is a direct solution that does not require the inversion of large matrices and includes only nearest-neighbor interactions. Having only nearest-neighbor interactions means that it is possible to run problems on parallel machines. You need to embed the antenna in a rectangular region and divide it into rectangular cubical cells with sizes ranging from 10 to 20 samples per wavelength at the highest frequency where analysis is desired. The outer surfaces contain absorbing boundaries to eliminate reﬂections that would produce errors. Formulating absorbing boundary conditions has been a signiﬁcant part of the method. You need to locate a solution surface between the absorbing boundaries and the antenna outer surface where we compute currents by using the equivalence theorem. The DFT of the time response determines the radiation pattern at a given frequency after the equivalent currents are found. If you need the pattern amplitude in only a few directions, the time-domain radiation can be found directly: for example, the gain in one direction. We can formulate some problems in one or two dimensions if they possess symmetry instead of the three-dimensional rectangular cube. The solution time is reduced dramatically, and the time-animated presentation may provide sufﬁcient insight when the radiation pattern is found in two dimensions. Because this is a time-domain analysis, we need to excite the structure with a pulse. You use the pulse frequency power response to normalize the patterns and compute gain. When the formulation includes the material losses, the efﬁciency of the antenna can be found since the dissipation in the inner cells prevents the radiation from reaching the outer surface. 2-6.2 Central Difference Derivative Numerical derivatives have greater potential for errors than integrals, but FDTD uses them to reduce Maxwell’s differential curl equations to simple difference equations. A second-order accurate formula for a derivative can be found by using central differences instead of using the difference between the value at a location or time and the value at the next point in a sequence of evenly spaced points: f (u0 + u/2) − f (u0 − u/2) ∂f = + O( u)2 ∂u u

(2-49)

We can use ﬁnite differences to solve the curl equations provided that we use electric and magnetic ﬁelds spaced at half intervals because each is related to the derivatives of the other ﬁeld and we want to use central differences to reduce error. Because Maxwell’s equations involve time derivatives, we need to calculate the electric and magnetic ﬁelds at interspersed half time intervals. 2-6.3 Finite-Difference Maxwell’s Equations Consider Maxwell’s curl equations in the time domain, including lossy materials: ∂H 1 = − (∇×E − M + σ ∗ H) ∂t µ ∂E 1 = − (∇×H − J + σ E) ∂t ε

(2-50)

78

RADIATION STRUCTURES AND NUMERICAL METHODS

Equations (2-50) contain the source currents J and M and include losses due to conducting dielectric material σ and magnetic material losses σ ∗ . Both equations have the same form, with only an interchange of symbols. Expanding the curl operator, we get the following equation for the x-component of the magnetic ﬁeld: 1 ∂Ey ∂Ez ∂Hx ∗ = − − Mx − σ Hx (2-51) ∂t µ ∂z ∂y The x-component of the electric ﬁeld has the same form but with the interchanges H → E, E → H , M → J , and σ ∗ → σ . You obtain the equations for the y- and z-components by a cyclic variation (repeating pattern of interchanges) x → y → z → x → y, and so on. For example, the equations are reduced to two dimensions by leaving out the y-component. FDTD calculates the ﬁeld at discrete times and locations on a grid. The ﬁelds can be represented as an indexed function using integers: f (i x, j y, k z, n t) = f (i, j, k, n) Because we use central differences [Eq. (2-49)], for derivatives, and the magnetic (electric) ﬁeld is found from the space derivative of the electric (magnetic) ﬁeld, the magnetic and electric ﬁelds need to be spaced a half-space interval apart. The time derivative becomes f (i, j, k, n + 12 ) − f (i, j, k, n − 12 ) ∂f (i, j, k, n) = ∂t t and means that the electric and magnetic components are interspersed at t/2 times that which produces a leapfrog algorithm. We substitute these ideas into Eq. (2-51) to derive the time-stepping equation for one component: Hx (i − 12 , j, k, n + 1) =

1 − σ ∗ (i − 12 , j, k) t/2µ(i − 12 , j, k) 1 + σ ∗ (i − 12 , j, k) t/2µ(i − 12 , j, k) +

Hx (i − 12 , j, k, n)

t/µ(i − 12 , j, k) 1 + σ ∗ (i − 12 , j, k) t/2µ(i − 12 , j, k)

Ey (i − 12 , j, k + 12 , n + 12 ) − Ey (i − 12 , j, k − 12 , n + 12 ) z −

Ez (i − 12 , j + 12 , k, n + 12 ) − Ez (i − 12 , j − 12 , k, n + 12 ) y

− Mx (i − 12 , j, k, n + 12 )

(2-52)

FDTD uses similar equations for the other components [18,19]. Yee’s Cell Figure 2-15 shows one cubic cell and the components of the ﬁelds. When we consider the upper face, we see that the magnetic ﬁeld components are spaced a

FINITE-DIFFERENCE TIME-DOMAIN METHOD

79

Hy

z Ez Hx

Hx Hz Hy Hz

Ey

Hz Ex Hx (i,j,k)

Hy y

x

FIGURE 2-15 Unit cell of a Yee space lattice showing time and space separation of electric and magnetic ﬁelds in a cell. (From [15], Fig. 1, 1966 IEEE.)

half space interval from the central electrical ﬁeld and the arrows show the direction of ﬁelds. Although it would appear that the electric ﬁeld is different on the upper and lower face along the z-axis, the method assumes that the ﬁeld is constant throughout the cell. The magnetic ﬁelds shown are at the center of adjoining cells. A leapfrog solution uses stored values of the electric ﬁelds to calculate the magnetic ﬁelds at a half time interval later and stores these values. In the second step the solution takes another half time step and uses the stored values of the magnetic ﬁelds to calculate the electric ﬁelds. The method gains stability by using the half time steps and by solving for both electric and magnetic ﬁelds. Although the ﬁelds are a half time step out of synch, we can average between the two half time steps to produce simultaneous ﬁelds at a point, but we only need to do this when calculating equivalent currents on the surface used for far-ﬁeld pattern calculations. 2-6.4 Time Step for Stability You need to pick the time step to produce a stable solution. Consider a plane wave traveling through the cubes. If the time step is too large, the wave can pass through more than one cell for each time step. At that point the solution cannot follow the actual wave propagation and fails. We must reduce the time step until it is less than the Courant condition or the wave propagation rate. Consider the fastest-moving wave in the problem, usually free space, and for equal sides to the cube, we compute the time step from the velocity and cell length: x v t ≤ √ d

(2-53)

The cell length is x and the number of dimensions is d. The time step must be lower for conducting materials (σ > 0) to produce a stable solution. The magic step uses

80

RADIATION STRUCTURES AND NUMERICAL METHODS

the equality and produces the most stable solutions. If you pick unequal sides to the rectangular cell, Eq. (2-53) is modiﬁed. 2-6.5 Numerical Dispersion and Stability FDTD analyses produce solutions that fail to propagate through the cells at the proper phase velocity in all directions. The propagation velocity depends on the cell size in wavelengths; it has a frequency-dependent component. You need to consider this numerical dispersion because it affects accuracy. Because the waves travel at different velocities in different directions, the dispersion problem increases for large structures where many time steps must be taken. After many steps, signals disperse because they have taken different routes and fail to add together in the correct phase. Finer cells solve the problem, but the computation requirements grow rapidly. The equation for the propagation constant can be found from considering the FDTD formulation to produce the following equation for three-dimensional problems: ky y 2 kz z 2 1 ω t 2 kx x 2 1 1 1 = + + sin sin sin sin c t 2 x 2 y 2 z 2 (2-54) The factor kx is the FDTD propagation constant in the cells along the x-axis, only approximately the same as kx , the actual propagation constant in the structure. The yand z-axes have similar problems. If you take the limit as cell length approaches zero, u → 0, and so on, then sin(au)/u → a. Because t → 0 as the cell size shrinks for the solution still to satisfy the Courant limit, Eq. (2-54) reduces to the expression ω 2 = kx2 + ky2 + kz2 (2-55) c Equation (2-55) is the normal propagation constant equation for a plane wave in space and shows that the cell propagation constants converge to the correct values as the cell size shrinks. If you formulate a problem in one or two dimensions, you remove terms from the right side of Eq. (2-54) to determine the dispersion relationship. Absorbing boundary conditions (ABCs) can cause numerical instabilities. ABCs approximate inﬁnite space to simulate radiation by the antenna into space. FDTD problems must be placed in a ﬁnite number of cells because each cell requires computer storage. Every FDTD problem uses a ﬁnite number of cells for the ABCs with more cells required in the directions of maximum radiation. ABCs degrade as the number of time steps increases and eventually leads to numerical instabilities. A lively research on ABCs has produced good ones, but be aware that most have been found to produce problems at some point. If you write your own analyses, you will need to ﬁnd appropriate ones. Commercial codes will give their limitations. At one time, ABCs limited solution dynamic range, but ABCs are now available that produce reﬂection coefﬁcients from 10−4 to 10−6 . Numerical dispersion limits the dynamic range as well. Remember that the antenna will be modeled with small cubes that limit the resolution of the results. The errors of modeling lead to solution errors that limit the dynamic range. 2-6.6 Computer Storage and Execution Times The antenna to be analyzed is modeled by a set of cubic cells. Choosing an appropriate number is an art. Similarly, it will be necessary to have a meshing program. Using a

FINITE-DIFFERENCE TIME-DOMAIN METHOD

81

two-dimensional model will greatly reduce computer storage and run time. Remember that our purpose should be to gain insight unavailable from measurements. The calculations require the storage of three components of both the electric and magnetic ﬁelds in each Yee cell. Because we solve the problem in the time domain, the components are only real numbers, unlike frequency responses, which use complex numbers for each component. The material properties of the cells can be indicated with short 1byte integers provided that there are no more than 256 different ones. Single-precision storage of the components requires 30 bytes for each cell; double-precision storage requires 54 bytes. A three-dimensional problem with 200 cells on a side contains 8 M cells and would need 240 Mbytes of storage for single-precision and 432 Mbytes for double-precision components. At each time step approximately 10 ﬂoating-point operations (ﬂops) are needed for each component in each cell. We must run the time steps until the input pulse has peaked and died out in each cell. This takes about 10 times the number of cells in the longest direction (maximum number along one axis). The three-dimensional problem with 200 cells on a side runs for 2000 time steps and requires 60 ﬂops times the number of cells. The solution needs 2000 × 8 M × 60 ﬂops = 960 Gﬂops for completion. 2-6.7 Excitation We specify the excitation of an antenna in the time domain since FDTD operates in the time domain. If all we need is a single-frequency solution, a ramped sinusoidal waveform can be applied. The waveform is tapered from zero in about three cycles and the FDTD solution steps continue until a steady state is reached. It is more efﬁcient to use a waveform that gives a wide-frequency-range response after performing a discrete Fourier transform on the radiating boundary to compute equivalent currents used at a given frequency. The computer storage and run times are the same for the wideband response as the single-frequency response. A suitable wide-bandwidth excitation is the differentiated Gaussian pulse shown in Figure 2-16: (t/τp )2 − 1 t (2-56) Vinc (t) = −V0 exp − τp 2 We calculate the frequency response of the differentiated Gaussian pulse from the Fourier transform of Eq. (2-56): √ (ωτp )2 − 1 (2-57) Vinc (ω) = −j ω 2π τp2 V0 exp − 2 The spectrum of Eq. (2-57) peaks for ωp = 1/τp . Figure 2-17 gives the normalized frequency response and shows that the −20-dB-level normalized frequency extends from 0.06 to 2.75. For example, if we wanted to center the frequency response at 10 GHz, the normalizing pulse time is easily found: τp =

1 = 1.592 × 10−11 s = 15.92 ps 2π(10 × 109 )

A check of Figure 2-17 shows that the antenna frequency response could be found from 2 to 22 GHz with only a 10-dB loss in dynamic range compared to the response at 10 GHz. A single time response computation yields a wide-frequency-range response.

82

RADIATION STRUCTURES AND NUMERICAL METHODS

Normalized Voltage

Gaussian Pulse

Differentiated Gaussian Pulse

Normalized Time

Amplitude, dB

FIGURE 2-16 Differentiated Gaussian pulse time response used in FDTD analysis.

Gaussian Pulse Differentiated Gaussian Pulse

Normalized Frequency

FIGURE 2-17

Differentiated Gaussian pulse normalized frequency response.

A sinusoidal modulated Gaussian pulse produces a narrow-bandwidth excitation useful in visualization because the bandwidth of the pulse can be controlled:

(t/τp )2 Vinc (t) = V0 exp − sin ω0 t 2

(2-58)

FINITE-DIFFERENCE TIME-DOMAIN METHOD

83

The unmodulated Gaussian pulse shown in Figure 2-16 has a low-pass frequency response: √ (ωτp )2 Vinc (ω) = 2πτp V0 exp − (2-59) 2 Figure 2-17 gives the low-pass frequency response of the Gaussian pulse with a −4.37dB response at ωp = 1/τp . The sinusoidal modulation centers the frequency response of the Gaussian pulse around ω0 , and the convolution of the two frequency responses produces a two-sided response of the Gaussian pulse.

2-6.8 Waveguide Horn Example [19] The literature contains solutions for the patterns of a number of antennas. Figure 2-18 shows the meshing of a commercial standard gain horn analyzed and compared to measurement. The horn operates from 8.2 to 12.4 GHz. The horn has a radiating aperture that is 110 mm wide and 79 mm high and a bell length of 228 mm. The 51-mm length of the input waveguide and the details of the feed probe were included in the model. Placing a perfectly magnetic conductor through the midsection of the horn uses symmetry to halve the number of cells to a uniform mesh of 519 × 116 × 183 Yee cells. Ten cells were used on the outside for the ABCs around the sides of the horn and 40 cells for the front ABCs in the maximum radiation direction. The model placed 20 cells between the edge of the horn and the equivalent current surface used for pattern calculations. The longest side of the grid determined the number of time steps at 10 times the number of cells = 5190 time steps. The model contains approximately 11 M Yee cells that require 330 Mbytes of computer storage. Assuming that the problem

Probe

WR-90 Waveguide

FDTD Cells Shown 2X

Coaxial Line a

b a

lw D b b

a

FIGURE 2-18 FDTD model of a standard gain horn. (From [17], Fig. 7.17, 1998 Artech House, Inc.)

84

RADIATION STRUCTURES AND NUMERICAL METHODS

(a)

(b)

FIGURE 2-19 FDTD calculated electric ﬁeld in the vertical symmetrical plane of a standard gain horn: (a) early time with a pulse in the throat; (b) pulse leaving the mouth of the horn. (From [17], Fig. 7.20, 1998 Artech House, Inc.)

takes 60 ﬂops per cell for each time step, the solution required 3.43 Tﬂops of computer calculations. The initial calculation used a differentiated Gaussian pulse excitation with τp = 15.9 ps that centered the response at 10 GHz. The calculation produced patterns that matched measurements. A second calculation used a sinusoidal modulated Gaussian pulse with the time constant 79.6 ps. This pulse time constant gives a normalized frequency of 2 GHz for the Gaussian pulse. The −3-dB frequency is 0.83 times the normalizing frequency. The pulse is centered at 10 GHz with a 3-dB bandwidth of 3.32 GHz. Figure 2-19 shows the ﬁelds when the pulse reached the horn aperture. Note the high ﬁelds in front of the horn and the amount of ﬁelds still radiating beyond and behind the aperture. By using a sinusoidal modulated pulse, the visual display contains nulls that improve its clarity. 2-7 RAY OPTICS AND THE GEOMETRIC THEORY OF DIFFRACTION Ray optics can give you a good physical feel for radiation and spur design ideas, but we need to question the accuracy of their use. Ray optics or geometric optics (GO) methods come from the design of lens and optical reﬂectors where the wavelength is very short compared to the size of the object being analyzed, whereas we may be interested in analyzing or designing an antenna on a structure only a few wavelengths in size. Below we show that GO is essentially correct over most of the radiation sphere and that by using elements of the geometric theory of diffraction [GTD (UTD)], the pattern prediction can be improved. In this case improvement means that we will increase the area of the radiation pattern that becomes more accurate. You will discover that it takes

RAY OPTICS AND THE GEOMETRIC THEORY OF DIFFRACTION

85

an increasing amount of effort to improve small areas of the pattern prediction, and at some point you should decide that it fails to give enough improvement to justify the work. Your real design purpose is to determine antenna dimensions that produce the desired antenna response. Of course, as the expense of the antenna increases, your customer may demand better predictions of the ﬁnal result, and then the cost of a better analysis is justiﬁed. You need to accept a new approach. Even though a part of the pattern prediction shows errors, obvious discontinuities, it only means that the pattern is inaccurate in directions near them and that over most of the radiation sphere the prediction is essentially correct. Discussion of this method begins with simple examples given in two-dimensional space that introduce the ideas behind GO and GTD. These examples can ignore the details of rotation of polarization directions because the waves are either polarized with the electric ﬁeld normal to the page or located in the plane of the page. We consider radiation blockage by objects, the reﬂection of rays by the objects, and the diffraction of rays around edges that ﬁlls in the pattern in the shadow regions and across the boundary of the last reﬂected ray. After the discussion of simple examples, the key points of GTD will be given for use in three-dimensional problems. This involves the rotation of coordinate systems so that ray polarizations line up with planes of incidence for reﬂections, with edges for diffraction and curvature directions on curved surfaces that shed rays around the object into the shadow. You will need to investigate the references if you want to develop your own routines, but this discussion will introduce you to the topic and give you an appreciation of the method so that you can use available computer programs and understand their limitations. GO uses ray methods to approximate electromagnetics. It is exact only in the limit of zero wavelength (inﬁnite frequency), but we gain useful insight from it at any frequency. It will not give good results close to physical boundaries; but when we include the GTD, the results are accurate down to one-wavelength sizes and are useful at λ/4 sizes. GO gives us physical insight when we deal with reﬂectors. We must consider three aspects to use GO fully: (1) ray reﬂections, (2) polarization, and (3) amplitude variations along the ray path and through reﬂections. 2-7.1 Fermat’s Principle Rays travel through a medium at the speed of light determined by the index of refrac √ tion: n = εr µr . We deﬁne the optical path length as C n dl, where C is a prescribed path in space. Fermat’s principle determines the paths of rays between two points. It states that the optical path length is stationary along a valid ray path. An expression is stationary when its ﬁrst derivatives are zero and the optical path is a minimum (or maximum). We use Fermat’s principle to trace ray paths through reﬂection or refraction by searching for the minimum optical path lengths. We can ﬁnd more than one possible ray between points because Fermat’s principle requires only a local minimum. When we exclude the boundaries of lenses, regions of homogeneous medium, rays travel in straight lines. 2-7.2 H -Plane Pattern of a Dipole Located Over a Finite Strip Figure 2-20 illustrates the geometry of this problem and the various regions of the analysis. The diagram shows the end of the dipole rod with the two rods located

86

RADIATION STRUCTURES AND NUMERICAL METHODS

I RB

RB

l/4 II

II 3l/8

5l/8 SB

SB III

FIGURE 2-20 GTD example using a two-dimensional model of a dipole located over an asymmetrical ground plane.

normal to the page. The dipole pattern is omnidirectional in the page with the electric ﬁeld directed normal to the page. When we trace rays from the dipole to the ﬁnite strip, we discover two signiﬁcant directions on both sides of the strip. The dashed boundaries labeled RB (reﬂection boundary) are the directions of the last rays reﬂected from the strip. Similarly, the dashed boundaries labeled SB (shadow boundary) are the last rays of radiation not blocked by the strip. The radiation in region I results from the sum of the direct radiation from the dipole plus the radiation reﬂected by the strip. Only direction radiation from the dipole occurs in the two parts of region II. Finally, region III is totally blocked from any radiation by a direct or reﬂected ray. This region receives rays diffracted around the edges. If we add the direct and reﬂected rays in an analysis, we obtain the pattern given in Figure 2-21, which also traces the actual pattern. The pattern, using only the direct and reﬂected rays, accounts for the phasing between the direct radiation from the dipole and an image dipole located below the strip. If you compare the two traces on Figure 2-21, you see that the two patterns are similar near θ = 0, but the direct plus reﬂected ray pattern has discontinuities at the SBs and RBs. Figure 2-22 gives the results for the same analysis, but using a 5λ-wide ground-plane strip. When using the larger strip, the two patterns match to about 80◦ , and in the second case the simple analysis is correct over most of the forward semicircle. Simple geometric optics gives good results for large objects provided that you realize the patterns will contain discontinuities. Removing the discontinuities requires extra effort. A discontinuity in the pattern cannot exist because shadow and reﬂection boundaries occur in free space. It takes a material boundary to produce a discontinuous ﬁeld. But, for example, the tangential electric ﬁeld must be continuous across even material boundaries. Edge diffraction solves the discontinuity problem. Figure 2-23 gives the pattern of the edge diffraction for both edges normalized to the total pattern. The edge diffraction has matching discontinuities to the sum of the direct and reﬂected rays at the SBs and RBs. The UTD (uniform theory of diffraction) technique [20, p. 55] calculated these diffractions. When these diffractions are added to the direct and reﬂected ray radiation, the total pattern given in Figure 2-21 is obtained. The dipole, its image in the ground plane, and the two edge diffractions form a four-element array where each element has a unique pattern. Adding edge diffractions to the geometric optics ﬁelds removes the discontinuities and allows calculation of the pattern behind the strip ground plane.

RAY OPTICS AND THE GEOMETRIC THEORY OF DIFFRACTION

87

Total Pattern

Direct + Reflected

FIGURE 2-21 H -plane pattern of a dipole over asymmetrical ground using direct and reﬂected rays compared only to a full solution for the 1λ ground plane of Figure 2-20.

Total

Direct + Reflected

FIGURE 2-22 H -plane pattern of a dipole over symmetrical ground using direct and reﬂected rays compared only to a full solution for a 5λ ground plane.

2-7.3 E -Plane Pattern of a Rectangular Horn Figure 2-24 illustrates the cross section of a horn or, in this case, a two-dimensional approximation to a horn. The waveguide feeds the horn and produces a uniform aperture distribution in the E-plane. In this model the direct GO radiation is a constant wedge signal as shown in Figure 2-25 ranging between −15◦ and +15◦ . The reﬂected pattern

88

RADIATION STRUCTURES AND NUMERICAL METHODS

FIGURE 2-23 GTD edge diffraction of an H -plane pattern for an asymmetrical 1λ ground plane under a dipole.

combines with the direct radiation and produces the same pattern. Figure 2-25 also shows the diffraction patterns from the two edges. These peak along the plates and exhibit a discontinuity at the same angle as the GO ﬁeld. Each diffraction pattern has a discontinuity on one side at 90◦ because the mouth of the horn blocks the diffraction from the opposite edge. When we add the diffracted ﬁelds to the GO ﬁeld, the pattern shown in Figure 2-26 is obtained. By just adding the three components, we obtain an accurate pattern of the horn over most of the angles of the plot. At 90◦ we see discontinuities in the pattern caused by not considering enough terms in the GTD calculation. You need to realize that these discontinuities only cause pattern errors at nearby angles. The majority of the pattern is correct. We need another term to correct the pattern near 90◦ . The blockage of the diffraction from one edge by the mouth of the horn causes a secondary diffraction at that edge. We call this double diffraction. Some available programs do not implement double diffraction because the general three-dimensional double diffraction takes considerable calculation due to the extensive ray tracing required. In these cases you must accept the pattern discontinuities. Some programs calculate double diffraction as an option, but turning on this option will slow the calculations. Figure 2-27 gives the pattern when double diffraction is included. Double diffraction reduces the discontinuity at 90◦ , but a small discontinuity remains. Adding triple diffraction would reduce this further, but the pattern area affected by the small discontinuity has shrunk. A new discontinuity near 60◦ appeared in the pattern after adding double diffraction at the

RAY OPTICS AND THE GEOMETRIC THEORY OF DIFFRACTION

89

FIGURE 2-24 Geometry of a two-dimensional model of a rectangular horn used for GTD analysis.

mouth of the horn. We could continue to add another term to remove this one or just accept it. 2-7.4 H -Plane Pattern of a Rectangular Horn The tangential electric ﬁelds vanish at the walls of the two-dimensional horn in the H -plane. This affects the GO ﬁeld and produces the following equation for them: E GO = cos

π tan θ e−j kR √ 2 tan α R

(2-60)

Equation (2-60) includes the phasing term and square-root spreading factor of a twodimensional ﬁeld. The horn walls tilt from the centerline by the angle α. Figure 2-28 plots the GO ﬁeld and shows that it vanishes at the walls. We do not expect edge diffraction because the ﬁeld vanishes at the edges, but Figure 2-28 shows diffraction patterns that peak in the direction of the walls. We call this new term slope diffraction. This new type requires another set of coefﬁcients not identical to the edge (or wedge) diffraction coefﬁcients. While the amplitude of the edge diffraction is proportion to the ﬁeld incident on the edge, the amplitude of slope diffraction is proportional to the derivative of the ﬁeld in the direction normal

90

RADIATION STRUCTURES AND NUMERICAL METHODS

FIGURE 2-25 E-plane pattern of a rectangular horn with a GO term (solid curve) and edge diffractions (dashed curves).

FIGURE 2-26 Combination of GO and edge diffractions in the E-plane pattern of a rectangular horn.

to the edge. We ﬁgure the same geometric factors for both edge and slope diffraction but now must calculate the normal derivative of the incident electric ﬁeld. Figure 2-29 plots the H -plane pattern of the horn. The pattern fails to predict a pattern behind it. The E-plane diffraction produces a back hemisphere pattern for a real horn, but our two-dimensional model does not include the E-plane. 2-7.5 Amplitude Variations Along a Ray Power decreases in a general ray as the distance from the source increases. If we expand the constant-phase surface (eikonal) about the ray in a Taylor series, we obtain a surface described by its radii of curvature [20, p. 55]. The maximum and minimum

RAY OPTICS AND THE GEOMETRIC THEORY OF DIFFRACTION

91

FIGURE 2-27 E-plane pattern of a rectangular horn combining GTD terms of direct GO, edge diffractions, and double diffractions between edges.

FIGURE 2-28 H -plane pattern of a rectangular horn with a GO term (solid curve) and edge slope diffractions (dashed curves).

values lie in the orthogonal principal planes. These radii of curvature determine the amplitude spread of the wave from point to point on the ray. We compute the ratio of differential areas about the ray at two locations as dA2 ρ1 ρ2 = dA1 (ρ1 + d)(ρ2 + d)

(2-61)

92

RADIATION STRUCTURES AND NUMERICAL METHODS

FIGURE 2-29 H -plane pattern of a rectangular horn by GTD analysis by combining direct GO ﬁeld and edge slope diffraction.

r1 dA1 r2

dA2

d

Principal Planes S

FIGURE 2-30 Astigmatic ray.

where ρ1 and ρ2 are the principal radii of curvature and d is the distance between two points on the ray (Figure 2-30). The electric ﬁeld variation along the ray becomes E0 e

−j kd

ρ1 ρ2 (ρ1 + d)(ρ2 + d)

(2-62)

for the astigmatic ray spreading from unequal radii of curvature. When d = −ρ1 or d = −ρ2 , GO fails because it predicts an inﬁnite power density. We call these locations caustics. Remember that the ray always has differential area and never has any real area as implied by Figure 2-30. We have three special cases of the astigmatic ray: 1. Spherical wave, ρ1 = ρ: E0 e−j kd

ρ ρ+d

(2-63)

93

RAY OPTICS AND THE GEOMETRIC THEORY OF DIFFRACTION

2. Cylindrical wave, ρ1 = ∞: E0 e−j kd 3. Plane wave, ρ1 = ρ2 = ∞:

ρ ρ+d

E0 e−j kd

(2-64)

(2-65)

The plane wave does not spread but has constant amplitude as distance changes. Both cylindrical and plane waves require inﬁnite power, and they are therefore nonphysical, but we ﬁnd them convenient mathematically. 2-7.6 Extra Phase Shift Through Caustics We cannot determine the ray amplitude at a caustic but can determine its amplitude and phase on either side. Passage through a caustic causes an extra phase shift to the ray [21, p. 31]. The denominator factors in the square root of Eq. (2-62) produce a 180◦ sign change when the ray distance factor d passes through either ρ1 or ρ2 . The square root changes 180◦ to +90◦ (ej π/2 ) or −90◦ (e−j π/2 ), depending on the direction of movement along the ray. When tracing a ray moving through a caustic in the direction of propagation, you multiply by ej π/2 . The ﬁeld is multiplied by e−j π/2 for a ray traced in the opposite direction of propagation. 2-7.7 Snell’s Laws and Reﬂection We derive Snell’s laws of reﬂection and refraction from Fermat’s principle. The two laws of reﬂection are given as: 1. The incident ray, the reﬂected ray, and the normal of the reﬂecting surface at the point of reﬂection lie in the same plane. 2. The incident and reﬂected rays make equal angles with the surface normal. Implicit in Snell’s laws is the idea that locally the wavefront behaves like a plane wave and that the reﬂector can be treated as a plane. Given the direction of the incident ray S1 , reﬂected ray S2 , and the reﬂector normal n, Snell’s laws of reﬂection can be expressed vectorially [22]: n×(S2 − S1 ) = 0

n·(S1 + S1 ) = 0

(2-66)

We combine Eq. (2-66) to determine the ray directions before or after reﬂection: S1 = S2 − 2(S2 · n)n

S2 = S1 − 2(S1 · n)n

(2-67)

Snell’s law of refraction can also be expressed vectorially as n×(n2 S2 − n1 S1 ) = 0 where n1 and n2 are the index of refractions in the two mediums.

(2-68)

94

RADIATION STRUCTURES AND NUMERICAL METHODS

2-7.8 Polarization Effects in Reﬂections The electric ﬁeld is orthogonal to the ray direction (a free-space wave) and is described by a two-dimensional polarization space (Section 1-11). We can describe polarization in any conveniently rotated two-dimensional basis vectors in the plane with the ray vector as its normal. We will use a ray-ﬁxed coordinate system that changes direction after a reﬂection: (2-69) Ei = ai|| Ei|| + ai⊥ Ei⊥ where ai|| is a unit vector in the plane of incidence and ai⊥ is perpendicular to the plane of incidence. We compute ai⊥ from the normal to the plane n at the reﬂection point and the incident ray unit vector Si : ai⊥ =

Si × n |Si × n|

ai||

ai⊥

=

(2-70)

× Si

After reﬂection, we calculate the output ray-ﬁxed polarization vectors using the output ray Sr : ar⊥ = ai⊥ and ar|| = ar⊥ × Sr Ei|| is the incident electric ﬁeld in the direction of ai|| and Ei⊥ is in the direction of ai⊥ . Of course, the direction of unit vector a|| changes from incident to reﬂected rays. The electric ﬁeld parallel to the reﬂector surface must vanish on the conductor surface: Er⊥ = −Ei⊥

(2-71)

where Er⊥ is the reﬂected ﬁeld along ai⊥ . We calculate the reﬂection properties of E|| from the corresponding magnetic ﬁelds parallel to the surface: Hr|| = Hi||

(2-72)

By combining Eqs. (2-71) and (2-72), we obtain the dyadic relation for the ray-ﬁxed coordinate system: Ei|| 1 0 Er|| = (2-73) 0 −1 Er⊥ Ei⊥ where Er|| and Ei⊥ are the reﬂected ﬁeld components. At each reﬂection we rotate the polarizations to align ai⊥ with the normal to the plane of incidence. We can express Eq. (2-73) as a dyadic in terms of the incident and reﬂected wave polarization vectors = ai ar − ai ar . Of course, the alternative method is to describe polarizations in a R || || ⊥ ⊥ ﬁxed three-dimensional coordinate system, but it requires a 3 × 3 reﬂection matrix. 2-7.9 Reﬂection from a Curved Surface A wave reﬂected from a curved surface changes its radii of curvature and principal planes. The ﬁeld along the reﬂected ray is given by ρ1r ρ2r Er (s) = Ei0 ·R e−j ks (2-74) r (ρ1 + s)(ρ2r + s)

RAY OPTICS AND THE GEOMETRIC THEORY OF DIFFRACTION

95

where s is the distance along the ray from the reﬂection, ρ1 and ρ2 the reﬂected ray the reﬂection dyadic. Ei0 the incident ray electric ﬁeld. For radii of curvature, and R a ﬂat surface we use images of the incident ray caustics for ρ1r and ρ2r , but in general, ρ1r and ρ2r become 1 1 = ρ1r 2

1 1 + i ρ1i ρ2

+

1 1 = ρ2r 2

1 f1

1 1 + i ρ1i ρ2

+

1 f2

(2-75)

where f1 and f2 are generalized focal lengths of the surface. The spreading factor of Eq. (2-74) simpliﬁes in the far ﬁeld:

ρ1r ρ2r ≈ r (ρ1 + s)(ρ2r + s)

r r ρ1 ρ2 s

Kouyoumjian and Pathak [23] derived formulas for the focal lengths of a surface. We start with a surface with principal radii of curvature R1 and R2 with directions u1 and u2 at the point of reﬂection. For an incident ray with principal axes deﬁned by unit vectors xi1 and xi2 , we deﬁne a matrix relation between the incident ray and surface principal curvature directions:

xi · u θ = 1i 1 x2 · u1

xi1 · u2 xi2 · u2

(2-76)

where the determinant is |θ | = (xi1 · u1 )(xi2 · u2 ) − (xi2 · u1 )(xi1 · u2 ). Given the angle of incidence θ i , the following are the focal lengths: 2 2 2 2 θ21 + θ12 + θ11 1 cos θ i θ22 = + f1,2 |θ |2 R1 R2 2 2 2 2 2 θ21 − θ12 − θ11 1 1 1 4 cos θ i θ22 1 1 − i + − i + ± 2 |θ |2 R1 R2 ρ1i ρ2 ρ1i ρ2

1/2 2 2 2 2 2 θ21 + θ12 + θ11 θ22 4 cos2 θ i 4|θ |2 + + − (2-77) |θ |4 R1 R2 R1 R2 With a single reﬂection, we need not compute the direction of the principal axes. We need only the focal lengths. Multiple reﬂections require knowledge of the reﬂected-ray principal plane directions. Deﬁne the following matrices to determine the directions of the principal axes after reﬂection:

1 ρ1i Qi0 = 0

0 1 ρ2i

1 C0 = R1 0

0

1 R2

Qr = Qi0 + 2(θ −1 )T C0 θ −1 cos θi br1 = xi1 − 2(n · xi1 )n

br2 = xi2 − 2(n · xi2 )n

96

RADIATION STRUCTURES AND NUMERICAL METHODS

where n is the surface normal at the reﬂection point. One principal-axis direction is xr1

r Q22 − 1/ρ1r br1 − Qr12 br2 = 2 Qr22 − 1/ρ1r + (Qr12 )2

(2-78)

We derive the other from the cross product of Eq. (2-78) and the reﬂected ray unit vector: xr2 = −Sr × xr1 (2-79) We must reapply Eqs. (2-75) through (2-79) for every reﬂection. We use Eqs. (2-75) through (2-79) for analysis, but except for computer optimizations, they cannot be applied directly to synthesis. If we limit the reﬂectors to ﬁgures of rotation, the radii of curvature are given by the meridians and parallels and these problems reduce to two dimensions. Similarly, a cylindrical reﬂector fed with a cylindrical wave [Eq. (2-64)] reduces the problem to two dimensions. The incident and reﬂected waves remain in the single plane chosen for the reﬂector analysis. 2-7.10 Ray Tracing Tracing rays through a reﬂector system is conceptually straightforward. Where a ray strikes a reﬂector, we compute the normal to the surface. By using Eq. (2-67), we solve for the reﬂected-ray direction. Equation (2-73) determines the polarization effects when we express the incident and reﬂected rays in the ray-ﬁxed coordinates. We use geometric arguments to determine the amplitude variation along the ray through the reﬂection instead of the general expressions given above. We experience difﬁculty when we try to discover the reﬂection points for given ﬁeld and source points. No analytical expressions exist for calculating the reﬂection point of a general surface. The usual computer routines search for the minimum optical path length (Fermat’s principle) without using Eq. (2-67), since a local minimum will satisfy this equation. 2-7.11 Edge Diffraction Keller [24] extended the idea of reﬂection to edge diffraction by applying a generalized Fermat’s principle to the rays. Figure 2-31 illustrates the rays in edge diffraction and the associated polarization directions. The ﬁgure shows the edge vector at the diffraction point. The vector cross product between the edge vector and the incident ray points in the direction of the incident plane normal. We measure the diffraction angle of incidence in this plane between the incident ray and the edge normal. Because diffraction obeys a generalized Fermat’s principle, the diffracted ray exits at the same angle, similar to the reﬂected ray angles. The diffracted rays lie in a cone with the edge vector as its axis. The diffracted rays spread the incident power into a cone. Figure 2-31 shows a particular diffracted ray and how we determine the diffracted ray exiting plane. We deﬁne diffracted ray polarization in terms of the incident and diffracted planes. The vectors are parallel and perpendicular to the two planes. Given the edge unit vector e, you compute the incident perpendicular polarization vector: aφ =

e × S sin β0

(2-80)

RAY OPTICS AND THE GEOMETRIC THEORY OF DIFFRACTION

97

e^

s^

Efd

Ed

Plane of ^ ^ Diffraction (s, e)

b0

QE

E if′

Edge-Fixed Plane of Incidence (s^′, e^)

E ib′0

b0

^ s′

Edge (a)

f

^

f

^

f′

^

n

Re f Bo lectio und n ary

f′ O

QE w ado y Sh undar Bo

Plane ⊥ e^ at QE

np

(b)

FIGURE 2-31 Ray-ﬁxed coordinates related to edge-ﬁxed coordinates at the edge diffraction point on a curved edge by showing planes of incidence and diffraction. (From [25], Fig. 5, 1974 IEEE.)

where S is the incident ray and β0 is the angle between the edge tangent and the incident ray. The diffracted ray perpendicular polarization is similar to the incident ray aφ = −

e×S sin β0

(2-81)

where S is the diffracted ray unit vector. We have the following vector relations for diffraction: |e × S| = |e × S | and e · S = e · S (2-82)

98

RADIATION STRUCTURES AND NUMERICAL METHODS

We determine the parallel polarization vector along the ray-ﬁxed coordinates by the following cross products: aφ × aβ0 = S

and aφ × aβ0 = S

(2-83)

By using ray-ﬁxed coordinates, the diffraction matrix reduces to 2 × 2. When β0 = π/2, the parallel polarization components are parallel to the edge and the electric ﬁeld vanishes: Eβ0 + Eβ0 = 0. Acoustics calls this the soft boundary condition (Dirichlet); it operates on the parallel polarization components. The perpendicular components satisfy the hard boundary condition (Neumann). At a diffraction point Qe we describe diffraction by the matrix equation

i d Eβ (Qe ) Eβ0 (s) ρ −Ds 0 0 e−j ks (2-84) = i 0 −Dh Eφd (s) s(s + ρ) Eφ (Qe ) where s is the distance from the diffraction point. Diffraction locates one caustic on the diffraction point. We compute the second caustic distance ρ from the incident ray radius of curvature in the plane of incidence ρei and the edge curvature unit vector nˆ e : 1 nˆ e · (ˆs − sˆ ) 1 = i − ρ ρe a sin2 β0

(2-85)

where a is the edge radius of curvature. When a → ∞ (straight edge), the second term of Eq. (2-85) vanishes. A number of factors determine the wedge diffraction coefﬁcients. The diffracting edge factors include (1) the angle between the faces, (2) the edge curvature, and (3) the curvature of the faces. The ray angle factors are (1) the incident angle relative to the edge tangent, (2) the diffraction angle to the shadow boundary, and (3) the angle to the reﬂection boundary. The diffraction coefﬁcients peak at the shadow and reﬂection boundaries. UTD formulation uses characteristic lengths associated with incident and diffracted ray caustics. These many factors are beyond the current discussion. 2-7.12 Slope Diffraction The spatial rate of change of the ﬁeld normal to the edge produces slope diffraction, an added ﬁeld component. This ray optics term also satisﬁes the generalized Fermat’s principle with geometry determined by Eqs. (2-80) through (2-83), and (2-85). The slope diffraction equation has the same form as Eq. (2-84):

d i Eβ (Qe ) Eβ0 (s) ρ −es 0 0 = (2-86) e−j ks 0 −eh Eφd (s) s(s + ρ) Eφi (Qe ) where the diffraction coefﬁcients es,h are related to the ﬁeld derivative normal to the surface: 1 ∂Ds,h ∂ es,h = (2-87) j k sin β0 ∂φ ∂n The term ∂/∂n of Eq. (2-87) indicates the derivative of the incident ﬁelds given in the vector of Eq. (2-86). Equation (2-87) has the term ∂Ds,h /dφ for the soft and hard slope diffraction terms returned from a subroutine; it is only a notational derivative.

RAY OPTICS AND THE GEOMETRIC THEORY OF DIFFRACTION

99

2-7.13 Corner Diffraction Every structural discontinuity diffracts waves. We derive edge diffraction from an inﬁnite wedge where the wedge terminations (corners) produce diffracted rays. Recall from Section 2-4.2 that PTD added currents at edges to handle the effect of not having an inﬁnite surface; the formulation for corner diffraction uses equivalent currents to derive these coefﬁcients. We handle edge diffraction from each edge as always. Since each corner arises from two edges, we compute separate corner diffraction for each edge, two terms per corner. Whereas edge diffraction is bound to a cone, corner diffraction radiates in all directions. The edge must be visible from both the source and receive points before corner diffraction contributes. We must include corner diffraction in any three-dimensional problem. As the source and receiver become farther and farther away from the object, corner diffraction contributions dominate over edge diffractions since it is derived from equivalent currents. 2-7.14 Equivalent Currents GTD fails to predict ﬁelds at caustics. In many cases we consider these points unimportant, but for those cases where we need the ﬁelds, equivalent currents provide the answer. We derive equivalent currents from edge diffraction, which then replaces it and we use them instead of edge diffraction for all pattern points. The use of currents reduces the problem to a PO solution and line integrals are required. We relate the incident ﬁelds expressed in the ray ﬁxed to equivalent currents: √ 2j i Eβ Ds 2πk ej π/4 ηk 0 √ 2j i Eφ Dh 2πk ej π/4 M= 0 k I =

(2-88) (2-89)

The soft and hard diffraction coefﬁcients Ds,h depend on the source and receiver positions. Since we calculate the ﬁelds using vector potentials or dyadic Green’s functions, the formulation has no caustics. They are only associated with a geometric optics solution. Equivalent currents allow the calculation of the ﬁelds directly behind a reﬂector near the axis. The GTD solution produces a caustic as all points along the rim “light up” for an axisymmetrical design. PTD uses equivalent currents in a similar but different way to calculate correct ﬁelds in the same region. Equivalent currents derived from the diffraction coefﬁcients produce the entire solution, since the reﬂector blocks the incident ﬁeld. In PO we continue to include the direct ﬁeld and the induced current radiation on the reﬂector, but add the PTD current radiation. Realize that slope diffraction also adds to the equivalent currents. 2-7.15 Diffraction from Curved Surfaces [26, 27] In one analytical approach to surface-wave radiation we postulate waves bound to a surface that radiate only from discontinuities. Surface waves on inﬁnite structure do not radiate but attenuate exponentially away from the surface, because they are bound to it. We can formulate GTD as radiating from discontinuities, and this produces an approach

100

RADIATION STRUCTURES AND NUMERICAL METHODS

for ﬁelds radiated on the shadowed side of a curved body. The continuous discontinuity of the curved surface causes power to be radiated at every point in the shadow region. These waves radiate tangentially from a wave traveling along a geodesic and bound to the surface. Surface waves require a dielectric coating or a corrugated surface to slow and bind the wave to the surface. The surface curvature slows and binds the wave to the surface without the need for a dielectric or corrugated surface coating. The wave that propagates along the surface sheds power in rays tangentially to it. The rays travel along a surface geodesic from the attachment point to the radiation point. The geodesic curve is a minimum distance path on the surface between two points. In differential geometry it has a broader meaning, but for our purpose, the minimum distance deﬁnition will serve. The curved surface diffraction satisﬁes a generalized Fermat’s principle (minimum distance) as do all other terms of GTD. The best approach uses another ray-ﬁxed coordinate along the surface where the vectors are normal and tangential to the surface at both the attachment and radiation (shedding) points. Curved surface diffraction considers three types of problems with different formulations. Two of them start with an antenna mounted on the surface. We either calculate the pattern in the presence of the curved object or calculate the coupling to a second antenna also mounted on the curved object. The third case determines the ﬁeld scattered for a source located off the surface. All three use the ray-ﬁxed coordinates. We start with the surface normal nˆ and the tangent vector ˆt directed along the geodesic path. A vector cross product deﬁnes the third direction of the local coordinate system. We ˆ and the three vectors form a triad: nˆ × bˆ = ˆt. On a general use the surface binormal b, surface all three vectors change direction as the wave moves along the geodesic. We use the term torsion for a path with a changing binormal. A soft dyadic diffraction coefﬁcient is used with ﬁelds aligned with the attachment point binormal and the tangential shedding point binormal. We apply the hard dyadic diffraction coefﬁcient ﬁelds aligned along the normal vectors. No formulas exist for computing the attachment and shedding points on a general curved surface given the source and receive points. We usually start with a known diffraction and ﬁnd other points by incrementing along the curve by small steps.

REFERENCES 1. L. Diaz and T. A. Milligan, Antenna Engineering Using Physical Optics, Artech House, Boston, 1996. 2. B. F. Harrington, Time-Harmonic Electromagnetic Fields, McGraw-Hill, New York, 1961. 3. P. S. Hacker and H. E. Schrank, Range distance requirements for measuring low and ultralow sidelobe antenna patterns, IEEE Transactions on Antennas and Propagation, vol. AP-30, no. 5, September 1982, pp. 956–966. 4. K. Pontoppidan, ed., Technical Description of Grasp 8, Ticra, Copenhagen, 2000 (selfpublished and available at www.ticra.com). 5. J. H. Richmond, A reaction theorem and its application to antenna impedance calculations, IRE Transactions on Antennas and Propagation, vol. AP-9, no. 6, November 1961, pp. 515–520. 6. R. F. Harrington, Field Computation by Moment Methods, Macmillan, New York, 1968; reprinted by IEEE Press, New York, 1993.

REFERENCES

101

7. J. H. Richmond, Radiation and scattering by thin-wire structures in homogeneous conducting medium, IEEE Transactions on Antennas and Propagation, vol. AP-22, no. 2, March 1974, p. 365 (see also ASAP wire code). 8. N. N. Wang, J. H. Richmond, and M. C. Gilreath, Sinusoidal reactance formulation for radiation from conducting structures, IEEE Transactions on Antennas and Propagation, vol. AP-23, no. 3, May 1975, pp. 376–382. 9. B. M. Kolundzija and A. R. Djordjevic, Electromagnetic Modeling of Composite Metallic and Dielectric Structures, Artech House, Boston, 2002. 10. A. R. Djordjevic et al., AWAS for Windows Version 2.0: Analysis of Wire Antennas and Scatterers, Artech House, Boston, 2002. 11. A. C. Ludwig, Wire grid modeling of surface, IEEE Transactions on Antennas and Propagation, vol. AP-35, no. 9, September 1987, pp. 1045–1048. 12. A. W. Glisson and D. R. Wilton, Simple and efﬁcient numerical methods for problems of electromagnetic radiation and scattering from surfaces, IEEE Transactions on Antennas and Propagation, vol. AP-28, no. 5, September 1980, pp. 563–603. 13. S. M. Rao, D. R. Wilton, and A. W. Glisson, Electromagnetic scattering by surfaces of arbitrary shape, IEEE Transactions on Antennas and Propagation, vol. AP-30, no. 3, May 1982, pp. 409–418. 14. B. M. Kolundzija et al., WIPL-D: Electromagnetic Modeling of Composite Metallic and Dielectric Structures, Artech House, Boston, 2001. 15. P. -S. Kildal, Foundations of Antennas, Studentlitteratur, Lund, Sweden, 2000. 16. C. Scott, The Spectral Domain Method in Electromagnetics, Artech House, Boston, 1989. 17. K. S. Yee, Numerical solution of initial boundary value problems involving Maxwell’s equations in isotropic media, IEEE Transactions on Antennas and Propagation, vol. 14, no. 3, May 1966, pp. 302–307. 18. K. S. Kunz and R. J. Luebbers, The Finite Difference Time Domain Method for Electromagnetics, CRC Press, Boca Raton, FL, 1993. 19. J. G. Maloney and G. S. Smith, Modeling of antennas, Chapter 7 in A. Taﬂove, ed., Advances in Computational Electrodynamics: The Finite-Difference Time-Domain Method, Artech House, Boston, 1998. 20. D. J. Struik, Differential Geometry, Addison-Wesley, Reading, MA, 1950. 21. D. A. McNamara, C. W. I. Pistorius, and J. A. G. Malherbe, Introduction to the Uniform Geometrical Theory of Diffraction, Artech House, Boston, 1990. 22. F. S. Holt, in R. E. Collin and F. J. Zucker, eds., Antenna Theory, Part 2, McGraw-Hill, New York, 1969. 23. R. Kouyoumjian and P. Pathak, The dyadic diffraction coefﬁcient for a curved edge, NASA CR-2401, June 1974. 24. J. B. Keller, Geometrical theory of diffraction, Journal of the Optical Society of America, vol. 52, 1962, pp. 116–130. 25. R. G. Kouyoumjian and P. H. Pathak, A uniform geometrical theory of diffraction for an edge in a perfectly conducting surface, Proceedings of IEEE, vol. 62, no. 11, November 1974, pp. 1448–1461. 26. P. H. Pathak, W. D. Burnside, and R. J. Marhefka, A uniform GTD analysis of the diffraction of electromagnetic waves by a smooth convex surface, IEEE Transactions on Antennas and Propagation, vol. AP-28, no. 5, September 1980. 27. P. H. Pathak, N. Wang, W. D. Burnside, and R. G. Kouyoumjian, A uniform GTD solution for the radiation from sources on a convex surface, IEEE Transactions on Antennas and Propagation, vol. AP-29, no. 4, July 1981.

3 ARRAYS

We begin with arrays of antennas before discussing particular antenna elements to show the relationship between antenna size and shape and the resulting pattern characteristics. We ignore the feed network design initially and assume that the proper array feed distribution will be obtained. At ﬁrst, we assume a distribution of point sources and compute the approximate array pattern. Working with simple models provides insight rather than accuracy, and later we consider element pattern and interaction. In Chapter 12 we discuss feed network design and analysis in the discussion of phased arrays. The chapter begins with a mathematic description of an array and gives various assumptions used to simplify the expressions. We analyze a simple two-element array to gain insight into the radiation phenomenon and how far-ﬁeld patterns can be found with simple arguments. The discussion of a uniformly spaced linear array shows the Fourier series relationship between array layout and the pattern space given in sin(angle) space. The principal idea is that pattern beamwidth shrinks as the array length increases. If we space the elements too far apart, multiple beam peaks or grating lobes form in the pattern, and we show how to control these grating lobes and their relationship to maximum scan angle, array layout, and element spacing. Phased arrays scan the beam by controlling the relative phasing between the elements. We extend the linear array to planar layouts that produce narrow beams in both principal planes. The planar array design is unchanged from the methods for linear arrays, but the grating lobe analysis shows their unique properties, as they sometimes form outside the plane of scan. We can divide the phased array into pieces to form multiple scanning beams, but the beam shape is determined by the segment size and shape used for each beam. By adding amplitude control the phased array can form multiple beams with beamwidths determined by the entire size of the array. Each element in an array receives a portion of the power radiated by the other elements on transmittal, or scatters power into neighboring elements in reception. The Modern Antenna Design, Second Edition, By Thomas A. Milligan Copyright 2005 John Wiley & Sons, Inc.

102

ARRAYS

103

radiation from each antenna excites currents on its neighboring elements that also radiate, and we associate the total pattern with the antenna input. In an array the effective element patterns change due to this scattering. Because of reciprocity, which says that transmit and received patterns are identical, we can analyze the problem either way. This leads to mutual coupling, which we describe and analyze by mutual impedance (admittance, or scattering) matrices. This phenomenon causes the input impedance of the elements to change as we scan the array. The mutual coupling can lead to scan blindness when the feed reﬂection coefﬁcient grows due to mutual coupling, and the array totally reﬂects the signal into the feed network. If we want the exact pattern designed for, we must compensate the feeding coefﬁcients for the mutual coupling. A discussion of array gain gives two methods of calculation. First, the effective area and the associated gain of a planar array cannot exceed its area when we include the extra half-element spacing area provided by the edge elements. When we space the elements so that their individual effective areas no longer overlap, array gain is the element gain multiplied by the number of elements. We can calculate gain by adding up the input power instead of integrating the pattern to compute total radiated power. We relate input power of elements to the self- and mutual resistances to determine gain of linear and planar arrays using realistic elements. The chapter ends with a discussion of three-dimensional arrays using arbitrarily oriented elements. We add this analysis to the simple array formula to handle the polarization of rotated antennas. Related to this problem is the pointing of an antenna on a positioner. We apply rotation matrices to both problems. An array radiates or receives from two or more antennas at the same frequency. To calculate the ﬁeld radiated from arrays we add the electric ﬁelds radiated from each element. The amplitudes and phases of each antenna, determined by the feed network, give us extra degrees of freedom to shape the pattern and design shifts from radiating elements to the feed network. A single antenna radiates an electric ﬁeld with both polarization components: E = Eθ (θ, φ)θˆ + Eφ (θ, φ)φˆ where Eθ and Eφ are the two complex components (amplitude and phase) referred to some point on the antenna. If we move the antenna or the phase reference point, we only change the antenna radiated phase. We assume that the movement is small enough that the radiation approximation can still be used. Given r as the location of the antenna relative to the phase reference point, the added phase component is ej k·r , where k = 2π/λ(sin θ cos φ xˆ + sin θ sin φ yˆ + cos θ zˆ ) and r = x xˆ + y yˆ + z zˆ is the location of the antenna; k · r is the phase distance from the antenna to the reference plane through the reference point and is deﬁned by the radiation (receiving) direction. The electric ﬁeld radiated from the moved antenna becomes ˆ j k·r [Eθ (θ, φ)θˆ + Eφ (θ, φ)φ]e We assume that nearby objects do not alter the patterns in the movement, but we can alter element patterns if necessary. Suppose that we have an array of antennas located at points r 1 , r 2 , and so on. We obtain the total pattern by adding the electric ﬁelds radiated from each: E=

N i=1

ˆ j k·r i [Eθi (θ, φ)θˆ + Eφi (θ, φ)φ]e

(3-1)

104

ARRAYS

Bringing the antennas close together will change the patterns of each because every antenna will block the radiation of the others and the distribution of currents on the elements may be changed. The shape of small resonant antennas limits the possible distribution of currents, but the magnitude and phase may be changed due to the coupling. We make various approximations to Eq. (3-1). Changes in the patterns due to nearby antennas are ignored, and isolated element patterns are used. We assume initially a certain amplitude and phase distribution on the elements and ignore the problem of the feed network. Polarization reduces to a single term for equally polarized elements, such as dipoles, slots, or horns. If the antennas have identical element patterns, we can separate Eq. (3-1) into a product. ˆ E = [Eθ (θ, φ)θˆ + Eφ (θ, φ)φ] Ei ej k·r i (3-2) where Eθ and Eφ are the normalized patterns of the single element. Ei is the electric ﬁeld of the ith element, including the amplitude and phase of the feed distribution. Equation (3-2) describes pattern multiplication that separates the pattern into an element pattern and an array factor. The method requires that all antennas have the same pattern and be orientated in the same direction. The array factor represents the pattern from an array of isotropic pattern antennas. Because array factors can be calculated by hand, we ﬁnd them useful for gaining insight. We leave calculations using Eqs. (31) and (3-2) to the computer. The element patterns themselves could be arrays and we could use pattern multiplication to synthesize planar and volumetric arrays from linear arrays.

3-1 TWO-ELEMENT ARRAY Consider two elements lying on the z-axis and spaced a distance d centered on the origin (Figure 3-1). If we rotate the isotropic pattern antennas around the z-axis, the problem remains unchanged, which means that all great-circle (constant φ) patterns are identical. On the z-axis, the element phase constant becomes ej kz cos θ . For simple line arrays we can locate pattern nulls and peaks by simple arguments. Example Two elements are spaced λ/2 and have equal amplitudes and phases. Locate the nulls and peaks. The phase reference planes can be placed at any convenient point. Consider the pattern at θ = 90◦ . We place the reference plane through the axis of the array. The added phase factor is zero for both elements and we just add components. The equal element phases add to give a beam peak. If we place a second reference plane through the top element, the wave radiated from the bottom element travels across the array λ/2 to the reference plane. Increasing the distance propagated decreases phase and it changes by −180◦ . The two out-of-phase signals cancel to produce a pattern null. The array has symmetry about the x –y plane, which means that the array will have the same pattern above and below the symmetry plane. We denote this conﬁguration an even-mode array. Figure 3-2 plots this pattern with a solid line. You should repeat the example for an odd-mode array (phases 0◦ and 180◦ ) and convince yourself that the null occurs at θ = 90◦ and the beam peak occurs at θ = 0◦ (180◦ ), plotted in

TWO-ELEMENT ARRAY

105

z d/2 d cos q 2 Radiation

q

Reference plane

−d cos q 2 −d/2

FIGURE 3-1

Two-element array on a z-axis.

Figure 3-2 as a short-dashed curve. The solid and short-dashed curves have the same directivity. Example Suppose that the two elements are spaced λ/4, with the top element phase −90◦ and the bottom element phase 0◦ . Locate the beam peak and pattern null. We start by placing a reference plane through the top element. The wave radiated from the bottom element travels across the array, and its phase decreases by 90◦ . Both radiated waves have the same phase (−90◦ ) at the reference plane and add in phase for a beam peak. Consider a second plane through the bottom element. The wave from the top element loses 90◦ propagating across the array and the two waves are 180◦ out of phase and cancel for a null. The second example is an end-ﬁre array. Figure 3-2 illustrates the end-ﬁre pattern with a long-dashed curve. All three patterns on the ﬁgure have the same directivity. The phase distribution of an end-ﬁre array matches those of a wave traveling in the direction of the maximum. In these examples unequal amplitudes would limit the null depth to the difference. Varying the element phases while maintaining equal amplitudes changes the null directions. Consider a general two-element array with equal amplitudes and a phase difference between them. We split the phase shift into equal parts. The top-element phase is −δ/2 and the bottom-element phase is δ/2. When we apply Eq. (3-2) with an isotropic element pattern, we obtain the following electric ﬁeld using Euler’s identity:

δ πd E(θ ) = 2E0 cos cos θ − λ 2

e−j kr r

(3-3)

106

ARRAYS

Even Mode l/2 Spacing

Odd Mode l/2 Spacing

Endfire l/4 Spacing

FIGURE 3-2 Two isotropic element array pattern: even-mode λ/2 spacing (solid curve); odd-mode λ/2 spacing (short-dashed curve); end-ﬁre λ/4 spacing (long-dashed curve).

θ is measured from the z-axis. If we spaced the elements along the x-axis and found the pattern in the x –z plane, we substitute sin θ for cos θ in Eq. (3-3). In Chapter 4 we sample continuous distributions and position the elements along the x- or y-axis. Pattern peaks occur when the argument of the cosine is nπ, the nulls when it is (2n − 1)π/2. δ λ cos θmax = nπ + (3-4) 2 πd π δ λ cos θnull = (2n − 1) + (3-5) 2 2 πd If we subtract either Eq. (3-4) or (3-5) evaluated at two peaks or nulls, we get the same equation: λ cos θ1 − cos θ2 = (n1 − n2 ) (3-6) d Figure 3-3 illustrates the pattern of an equally phased two-isotropic-pattern-element array spaced 5λ along the z-axis. Because array symmetry makes the patterns on the right and left sides the same, we consider only one side. The wide element spacing

TWO-ELEMENT ARRAY

107

5l Spaced Array

Array - 10 dB Added to Isotropic

FIGURE 3-3 Two-isotropic-element array-spaced 5λ pattern (solid curve); added central element 10 dB higher power than array (dashed curve).

allows six solutions to Eq. (3-4) from 0 to 90◦ for the pattern peaks and ﬁve solutions for Eq. (3-5) over the same range for the nulls because the magnitude of cos θ is limited to 1. We call the multiple beams grating lobes. We usually choose the main beam and call the others grating lobes, but they are just all lobes of the array. Figure 3-3 shows that we must space the elements close together to prevent grating lobes. With a greater number of elements in the array, the amount of beam movement due to element phasing adds another factor to the prediction of when grating lobes form. The amount of phase scanning determines the maximum spacing allowed without the formation of grating lobes. The n = 0 lobe forms at θ = 90◦ and we compute the n = 1 mode direction from Eq. (3-4): θ = cos−1 ( 15 ) = 78.46◦ . When we substitute these angles into Eq. (3-3), we calculate a relative phase of 180◦ between them. The lobes have a phase of zero for n even and 180◦ phase for n odd in the far-ﬁeld approximation. Remember we remove the exponential and 1/R factors from Eq. (3-3) for the far-ﬁeld pattern. The actual phase of any real point depends on the distance from the center of the array. The dashed curve in Figure 3-3 shows what happens if we add the array pattern to an isotropic radiator in the center. For a peak response of the array −10 dB relative to the isotropic antenna, we get the 5.7-dB peak-to-peak ripple shown by using Scale 1-8. The array pattern either adds or subtracts from the isotropic radiator pattern. The angular ripple rate is half that of the array lobes. Below we see that a two-element array

108

ARRAYS

spaced at an integer multiple of λ/2 has a 3-dB greater gain than a single element. We feed half the power of the array into each element. By adding these factors we calculate the array element level to be −16 dB below the main central radiating antenna. When we mount an antenna over a ﬁnite ground plane, the diffraction from the edges creates a two-element array. A 5λ-wide ground plane would produce the same pattern ripple angular rate as shown in Figure 3-3. You will often observe a similar-amplitude ripple in measured antenna patterns. Note the minimum angular distance between the peak and minimum responses in the pattern. The extra signals occur along the line in the pattern plane perpendicular to this direction. Use Eq. (3-6) to determine the distance between the array elements and you should be able to identify the structure causing the ripple. The scattering point could be on the test ﬁxture. Consider whether the mounting structure will be different in the ﬁnal conﬁguration. You can calculate the effect from a single diffraction point by forming an array using the baseline of the primary radiator and the diffraction point. Both conﬁgurations produce the same angular ripple rate. The ripple peak occurs along that array axis, but Figure 3-3 shows that the angular ripple rate will be reduced along this end-ﬁre direction of the θ = 0 axis. If you make a careful consideration of the angular rates, in various pattern planes, you should be able to discover the cause. Always consider unexpected sources of diffraction. You can consider the ripple using its beamwidth. To produce a symmetrical pattern about zero, we use sin θ instead of cos θ in Eq. (3-3), which means that the array lies along the x-axis. The −3-dB angle for the two-element uniform amplitude array can be found from Eq. (3-3): πd π sin θ3 dB = λ 4

θ3 dB = sin−1

λ 4d

(3-7)

The beamwidth is twice Eq. (3-7). For large d we can approximate sin X ≈ X and beamwidth = λ/2d. The 5λ spaced array has a beamwidth of 5.7◦ (0.1 rad). We can look at a 5λ-wavelength ground-plane example that has a large-amplitude element compared to the edge diffraction as two 2.5λ-spaced two-element arrays where one element has a high amplitude. Each two-element array produces a pattern with an 11.4◦ beamwidth the value of the composite pattern in Figure 3-3. We often mount an antenna in the center of a ground plane for measurement and observe patterns similar to Figure 3-3. If in the actual application the antenna is mounted off center, we need to add the patterns of arrays formed on both sides of the ﬁnite ground plane. The ﬁnal pattern will be the composite pattern from each array and be more complicated than the simple case given above. We calculate average radiation intensity by an integral: Uavg = The directivity is

4E02 η

π/2 0

cos2

δ πd cos θ − sin θ dθ λ 2

|2Emax |2 Umax = Uavg 1 + sin(2πd/λ) cos δ/(2πd/λ)

(3-8)

where Emax = cos[(πd/λ) cos θmax − δ/2]. When d ≥ λ/2, Emax = 1. Figure 3-4 shows the directivity versus spacing for the special cases δ = 0◦ and δ = 180◦ (even and odd modes). The directivity varies because each antenna receives power from the other. The

LINEAR ARRAY OF N ELEMENTS

109

Odd-mode array (0°, 180°)

4

Directivity, dB

3

2

Even-mode array (0°, 0°)

1

0.5

1.0

1.5 Spacing, l

2.0

2.5

FIGURE 3-4 Directivity of even- and odd-mode two-isotropic-element arrays.

combination of the input power and the power transferred between elements changes with spacing. 3-2 LINEAR ARRAY OF N ELEMENTS Suppose that there are N isotropic radiators equally spaced along the z-axis and fed with equal amplitudes. We assign a ﬁxed phase shift δ between progressive elements. The array factor ﬁeld is sin(N ψ/2) (3-9) N sin(ψ/2) where ψ = kd cos θ + δ [1, p. 258]. We use this to plot a universal radiation pattern for the array (Figure 3-5) for two to 10 elements. The abscissa ψ is plotted in degrees (360◦ is substituted for 2π in k). Both ends of the plot are lines of symmetry. The plot is periodic (period 360◦ ). We see that the level of the ﬁrst sidelobe (N = 2 has no sidelobe) decreases as N decreases but approaches a limit of 13.3 dB of the continuous aperture. Figure 3-6 demonstrates the periodic pattern for N = 6 and shows a projection to a polar pattern when the progressive phase between elements is zero and the elements are spaced λ/2. We can plot similar curves for other array distributions; all have a period of 360◦ . Figure 3-6 illustrates the use of a circle diagram, a method of constructing a polar pattern from the universal pattern such as Eq. (3-9) for the uniform-amplitude distribution. An array can be analyzed as a sampling of the continuous distribution that produces a Fourier series of the distribution. A Fourier series has multiple responses. In Chapter 4 we design large arrays by sampling continuous distributions. The pattern angle of an array is measured either from the axis using cosine of pattern angle or is measured broadside using sine. You should become comfortable with either notation since the sine and cosine of angles involves only a complementary operation of the angles.

110

ARRAYS

−2 −4

3

4

N=2

−6 Amplitude, dB

−8 6

−10

3 4

8

−12

8 10

−14

6

5

10

5 7

−16 8

−18 −20

FIGURE 3-5

6

0

20

40

60

80 100 120 kd cos q + d

140

160

180

ψ-space pattern of linear arrays with a uniform amplitude distribution.

360 sin q Spacing/l Visible Region

FIGURE 3-6 Circle diagram of a six-element uniform-amplitude array with λ/2 spacing.

Since cos θ (or sin θ ) is limited to ±1, the region along the abscissa of the universal pattern used (the visible region) is found from the range of ψ: −360◦ d +δ λ

to

360◦ d +δ λ

LINEAR ARRAY OF N ELEMENTS

111

The circle diagram is constructed by ﬁrst drawing a circle the same diameter as the visible region below the universal diagram centered at δ, the progressive phase shift between elements. Figure 3-6 has a δ = 0. Since the element spacing is λ/2, the range is ±180◦ . The polar pattern radius equals the amplitude of the universal pattern. Both the universal pattern and the polar pattern use a logarithmic (dB) scale from 0 to −40 dB. Projecting points vertically from the universal pattern to the visible region performs the cosine or sine operation, and the polar pattern becomes the real pattern in space. We project each point vertically until it intersects the dashed visible region circle in two places and then draw lines from these points to the center. After you project the nulls and peaks of the universal pattern to the dashed circle, it is easy to sketch the polar pattern. The circle diagram helps us visualize patterns and the effects of scanning, but no one would do serious design with it. Second, it is useful only for small arrays because large arrays produce unwieldy diagrams. When the spacing between elements is greater than λ/2, the visible region widens to include more than one periodic main lobe and the array has multiple beams. To have a beam centered at θ1 , set the progressive phase difference between elements: δ=

−360◦ d cos θ1 λ

(3-10)

−360◦ d λ

(3-11)

End ﬁre (θ1 = 0) occurs when δ=

We can use Figure 3-5 to compute beamwidth angles of arrays. Table 3-1 is a list of the ψ-space angles of the 3- and 10-dB levels. Example A six-element equally spaced uniform array has spacings of λ/2 and zero progressive phase shift between elements (δ = 0◦ ). Calculate the 3-dB beamwidth. We read from Table 3-1 the value ψ3 dB = 26.90◦ . Because the pattern is symmetrical in ψ space (Figure 3-6), the second ψ3 dB is −26.9◦ . kd cos θ1,2 + δ = ±ψ3 dB ±26.9◦ 360◦ λ ◦ cos θ1,2 = cos θ1,2 = ±26.90 λ 2 180◦ ◦ ◦ θ1 = 81.4 θ2 = 98.6 TABLE 3-1 ψ-Space Angles of 3- and 10-dB Levels of an Equal-Amplitude Distribution Array (deg) N

3 dB

10 dB

N

3 dB

10 dB

N

3 dB

10 dB

2 3 4 5 6 7 8 9 10

90.00 55.90 40.98 32.46 26.90 22.98 20.07 17.81 16.02

143.13 91.47 67.63 53.75 44.63 38.18 33.36 29.62 26.64

11 12 13 14 15 16 17 18 19

14.55 13.33 12.30 11.42 10.65 9.98 9.39 8.87 8.40

24.21 22.18 20.47 19.00 17.74 16.62 15.64 14.77 14.00

20 24 28 32 36 40 50 64 100

7.980 6.649 5.698 4.985 4.431 3.988 3.190 2.492 1.595

13.29 11.08 9.492 8.305 7.382 6.643 5.314 4.152 2.657

112

ARRAYS

Remember that θ is measured from the axis of the array (z-axis) and the 3-dB beamwidth is the difference (17.2◦ ). On Figure 3-6 the visible region ranges between −180◦ and +180◦ . There are four sidelobes in the visible region (Figure 3-6). Since an array samples a continuous aperture distribution, the continuous distribution is Nd long. We can estimate beamwidth by using a uniform amplitude distribution: ◦

HPBW = 50.76

λ ◦ = 16.92 Nd

This formula approximates the array beamwidth reasonably. Example A six-element array has a progressive phase shift δ of 90◦ between elements. Compute the 10-dB beam edge angles for λ/2 spacing. Figure 3-7 shows the circle diagram analysis of this example. The line to the center of the polar pattern has been shifted to 90◦ and the pattern spans 360◦ of the linear scale. By projecting the nulls and peaks to the circle below, the pattern can easily be sketched. ◦

ψ10 dB = ±44.63

(Table 3-1)

kd cos θ1,2 = ±ψ10 dB − δ

360 sin q Spacing/l Phase Shift

Visible Region

FIGURE 3-7 Six-element uniform-amplitude array with λ/2 spacing scanned with 90◦ progressive phase shift between elements.

LINEAR ARRAY OF N ELEMENTS

113

Solving for cos θ1,2 , we have ±ψ10 dB − δ ±44.63◦ − 90◦ = kd 180◦ ◦ ◦ ◦ θ1 = 104.6 θ2 = 138.4 beamwidth = 33.8

cos θ1,2 =

There are ﬁve sidelobes in the visible region (Figure 3-7). Equation (3-9) gives the beam maximum direction: cos θ0 =

−δ −90◦ = kd 180◦

◦

θ0 = 120

The main beam is no longer symmetrical about the beam peak. The 3-dB pattern angles are 110.5◦ and 130.5◦ . The beamwidth (3-dB beamwidth = 20◦ ) increases with scan angle. What element spacing would result in this beamwidth for broadside radiation (δ = 0◦ )? 360◦ ◦ d cos θ1 = 26.90 (Table 3-1) λ On solving for spacing, we have d 26.9◦ = λ 360◦ cos θ1 Remember that the beam is centered on θ = 90◦ , so that θ1 = 90 − 20/2 = 80◦ . 26.9◦ d = = 0.431 λ 360◦ cos 80◦ The effective spacing has been reduced by approximately the cosine of the scan angle from θ = 90◦ , broadside: d ◦ cos 30 = 0.433 λ The accuracy of the cosine relation increases with more elements. Example Determine the progressive phase shift between elements for an end-ﬁre array with 0.3λ element spacing and compute beamwidth for a uniform distribution array with ﬁve elements. Figure 3-8 illustrates this example using the circle diagram. End ﬁre occurs when [Eq. (3-11)] −360◦ (0.3λ) ◦ δ= = −108 λ This is the progressive phase shift for all distributions with 0.3λ element spacing for an end-ﬁre pattern. Table 3-1 gives the ψ-space angle, ψ3 dB = ±32.46◦ . Substituting in the expression for ψ, we have 360◦ (0.3λ) ◦ cos θ1,2 = ±32.46 λ ±32.46◦ + 108◦ cos θ1,2 = 360◦ (0.3) 140.46 cos θ1 = = 1.301 108

cos θ2 =

75.46 = 0.699 108

114

ARRAYS

Visible Region

360 sin q Spacing/l Phase Shift

FIGURE 3-8 Five-element uniform-amplitude array scanned to end ﬁre.

θ1 is in invisible space, since | cos θ | ≤ 1; θ2 = 45.6◦ . Symmetry about the z-axis supplies us with the second angle θ1 = −45.6◦ and beamwidth is the difference: 91.2◦ . The end-ﬁre array samples a traveling-wave distribution. The continuous uniform distribution phased for end ﬁre with the same length has a 90.4◦ beamwidth. Remember that we have been dealing with isotropic pattern antennas. For example, broadcast towers, seen from above, approximate isotropic antennas in the horizontal plane. The patterns of the individual antennas modify the results of isotropic antenna arrays. In small arrays the element pattern is quite signiﬁcant, but the beamwidths of large arrays are determined mainly by the array factor. The beamwidths calculated for array factors approximate the actual beamwidths only when the elements have signiﬁcant patterns. We must rely on computer solutions of speciﬁc cases, including the element pattern, for better results. 3-3 HANSEN AND WOODYARD END-FIRE ARRAY [2] The end-ﬁre array directivity increases if the sum of the progressive phase shifts between elements is decreased by approximately π. The equivalent traveling-wave velocity slows in the structure relative to free space. The progressive phase shift between elements becomes δ = −kd −

2.94 π −kd − N N

rad

(3-12)

PHASED ARRAYS q

115

q Hansen and Woodyard end fire

Normal end fire

10 dB

20 dB

30 dB

FIGURE 3-9 Patterns of a normal end ﬁre and a Hansen and Woodyard end-ﬁre array of isotropic elements.

where N is the number of elements in the array. The normal end-ﬁre progressive phase shift between elements, δ = −kd, places one edge of the visible region at the origin of ψ space. This method shifts the edge to a lower portion of the curve. The universal radiation curve peak (Figure 3-5) shifts into invisible space and the sidelobes rise in proportion to the new beam peak, but the beamwidth narrows. Equation (3-12) holds strictly only for large arrays, but the directivity increases for all arrays when it is applied. Example Suppose that eight elements are spaced λ/4 apart with a uniform amplitude distribution. Compare the two endﬁre designs. The two patterns are compared in Figure 3-9. The results are as follows: The beamwidth decreases, and the directivity increases by 2.5 dB. The sidelobes rise to 9 dB from 13 dB.

3-4 PHASED ARRAYS Suppose that a wave approaches at an angle to the axis of an array located on the z axis (Figure 3-10). The wave reaches the top element ﬁrst and progresses down the array in succession. If the signals are added directly, they will cancel each other to some extent because they have a progression of phases. Figure 3-10 shows the results of adding a series of time delays to equalize the path lengths in the lines where the position zi along the axis determines the time delay τi for incident angle θ0 : τi =

zi cos θ0 + τ0 c

116

ARRAYS

t1 Time delay t2

t3

t4

t5

Incident wave

t6

FIGURE 3-10 Linear array scanned with time-delay networks.

and velocity of light c. We add an arbitrary time delay τ0 to keep all time delays, τi , positive. This feed network is frequency independent, as we vary the progression of time delays to scan the beam. Phase shifters replace the time-delay networks in phased arrays. They provide equivalent beam scanning at a single frequency. To scan to an angle θ0 , the required phase shift is 2π − z cos θ0 modulo 2π (rad) λ −360◦ ◦ z cos θ0 modulo 360 (deg) λ for elements located on the z-axis. For a general space array we must counteract the phase difference to the reference plane, ej k·r , for the direction of scan so that the phases of all elements are zero. To scan in the direction (θ0 , φ0 ), we must add a phase factor to every element, depending on its position. The phase factor on each element of a general space array is e−j k0 · r (3-13) where k0 =

2π (sin θ0 cos φ0 xˆ + sin θ0 sin φ0 yˆ + cos θ0 zˆ ) λ

GRATING LOBES

117

is the vector propagation constant in the direction of the beam and r is the element location. Adding this phase factor to the element phases causes the product of the exponential factors [Eq. (3-2)] to be 1 at the scan angle, and the components Ei add in the scan direction. Using phase shifters limits the frequency bandwidth. Given a ﬁxed phase shift over a small frequency range, increasing the frequency scans the beam toward broadside: θ =

π f2 − f1 − θ0 tan f2 2

(3-14)

rad

where θ0 is the scan angle [3]. Limiting the allowable scanning with frequency to plus or minus one-fourth of the local beamwidth deﬁnes the bandwidth of the array. When the beam is scanned to 30◦ off the axis, the bandwidth is related directly to the beamwidth at broadside (θ = 90◦ ): ◦

bandwidth(%) beamwidth (deg) at θ0 = 30

The beam shifts less with frequency near broadside, since the tangent factor in Eq. (314) approaches zero. A general estimate is given by bandwidth(%)

beamwidth (deg) 2 cos θ0

(3-15)

where the broadside beamwidth is used. Example Given an array with 100 elements spaced at λ/2, determine the bandwidth when scanned to 45◦ . The beamwidth is estimated from the aperture width: HPBW =

50.76◦ ◦ 1 100( 12 )

bandwidth(%)

1 = 0.7% 2 cos 45◦

Any radar antenna would have a broader beamwidth because the sidelobes need to be reduced, but this is a good ﬁrst estimate. The bandwidth can be increased by feeding subarrays with time-delay networks. The subarrays continue to be scanned with phase shifters. Only a few time-delay networks are needed, and the subarray beamwidth determines the bandwidth. In Chapter 12 we discuss the problems caused by using subarrays. 3-5 GRATING LOBES Phased arrays vary the progressive phase by Eq. (3-13) to scan the beam. When the array element spacing is greater than λ/2, the appearance of secondary beam peaks (grating lobes) limits the scan angle. The grating lobe attains full amplitude when d −1 λ (1 + cos θgr ) = 1 −1 (3-16) θgr = cos λ d

118

ARRAYS

360 sin q Spacing/l Phase Shift

Visible Region

FIGURE 3-11 Ten-element array with 3λ/4 spacing scanned to 26◦ , showing the onset of a grating lobe.

Example The spacing of the elements of an array is 0.75λ. Determine the scan angle when the grating lobe is full amplitude. 4 ◦ − 1 = 70.5 θgr = cos−1 3 At this point the grating lobe is the same amplitude as the main beam. The lobe does not appear suddenly, but it grows as the visible region shifts and starts including the second periodic main lobe. Figure 3-11 shows the grating lobe formation for an array with 0.75λ element spacings on a circle diagram. The dashed circle of the visible region spans more than one beam of the universal radiation pattern of the uniform amplitude array. Arrays with element spacing greater than λ always have grating lobes (multiple main beams), but the pattern of the antenna elements may reduce the grating lobes to acceptable levels and allow a wide element spacing. 3-6 MULTIPLE BEAMS An array can form multiple beams. Equation (3-13) gives the phase coefﬁcients to multiply each element feed voltage Ei to scan it to a given angle. The array will form

MULTIPLE BEAMS

119

a second beam if we add a second distribution: Ei e−j k2 ·ri . The distribution Ei remains constant for both beams. We add the two distributions to obtain both beams:

Ei (e−j k1 ·ri + e−j k2 ·ri )

(3-17)

This multiplies the distribution Ei by a second distribution whose amplitudes and phases are functions of the antenna position and the scan angles of the two beams. Each beam uses the entire array to form its beam. In a phased array both phase and amplitude must be varied to achieve multiple beams. An array, which can only vary phase, must be divided into subarrays to form multiple beams, but its beamwidths will depend on the subarray widths. We can produce unequal beams with different amplitude distributions and pattern shapes if needed. We can add as many beams as necessary by including the distribution element factors with the scanning phase coefﬁcients in Eq. (3-17). The element feeding coefﬁcients become the sum. Example Compute the feed coefﬁcients of a 15-element array with λ/2 spacings and a uniform distribution scanned to 45◦ and 120◦ from the z-axis. First center the array on the z-axis. The elements are located at zi =

(−8 + i)λ 2

To scan to 45◦ , the element phase factors are 360◦ 1 (−8 + i)λ exp(−j kzi cos 45 ) = exp −j √ λ 2 2 ◦

◦

phase factors are ej 90 (−8+i) . The ninth-element (z9 = λ/2) To scan to 120◦ , the element ◦ ◦ −j 127.3 j 90 phase factors are e and e . Assume a voltage magnitude of one-half for each uniform-amplitude-distribution beam so that the center element has a magnitude of 1. We sum◦ the distributions to calculate the feeding coefﬁcient of the ninth element:0.32ej 161.4 . When converted to decibel ratios, Table 3-2 lists the feeding coefﬁcients for the array. We can estimate both beamwidths from Table 3-1. ψ3 dB = 10.65◦ :

cos θ1,2

±10.65◦ + 127.28◦ 180◦ = ±10.65◦ − 90◦ 180◦

θ1 = 40◦ , θ2 = 49.6◦ θ1 = 116.2◦ , θ2 = 124◦

The pattern (Figure 3-12) has these beams. The gain of each beam depends on the feed network. If a single input supplies power to two beams, each beam can receive only half the input power and gain reduced 3 dB for both beams. Butler matrices [4] and Blass beamforming networks [5] supply an input for each beam. The inputs are isolated from each other and the transmitter power in each port feeds only one beam, and therefore the full array gain is available to each input. Similarly, we can place a receiver on each port and use the full effective area for each.

120

ARRAYS

TABLE 3-2 Feeding Coefﬁcients for a Dual-Beam 15-Element Array Beamwidth (dB)

Angle (deg)

Element

Element

Beamwidth (dB)

Angle (deg)

1 2 3 4 5 6 7 8

−2.38 −8.59 −0.01 −11.49 −1.64 −1.99 −9.91 0.00

130.48 111.84 −86.80 74.56 55.92 −142.72 161.36 0.0

9 10 11 12 13 14 15

−9.91 −1.99 −1.64 −11.49 −0.01 −8.59 −2.38

161.36 142.72 −55.92 −74.56 86.80 −111.84 −130.48

FIGURE 3-12 Fifteen-element linear array pattern with simultaneous beams at θ = 45◦ and 120◦ .

We will delay the important topics of array synthesis and sidelobe reduction until after we have discussed aperture distributions. A trade-off is made between the beamwidth and the sidelobe levels. The beamwidth narrows only by putting more power into the sidelobes. 3-7 PLANAR ARRAY The linear array only controls the pattern in one plane; it depends on the element pattern to control the beam in the other plane. Planar arrays can control the beam shape in both planes and form pencil beams. Whereas a linear array can only scan in

PLANAR ARRAY

121

a single plane, a planar array can scan to any angle in the upper hemisphere. Most planar arrays rely on the element pattern or ground plane to eliminate the backlobe on the opposite side of the plane. The planar array has N − 1 nulls that can be used to control the pattern where N is the total number of elements. A simple feed distribution uses the product of two linear arrays. This eliminates many degrees of freedom of the array because an M × N array would be determined by M − 1 + N − 1 nulls when we could have used M × N − 1 possible nulls. Figure 3-13 shows the spherical pattern of a uniformly spaced 8 × 8 planar array where all elements are fed the same amplitude where a 90◦ beamwidth element eliminates the backlobe. The pattern along either principal axis shows the steady sidelobe reduction that starts with −13.2 dB. Diagonal plane sidelobes are the product of the sidelobes in the principal planes. The ﬁrst sidelobe in the diagonal plane is down 26.4 dB. An array feed distribution, not a product of two linear arrays, can yield more equal sidelobes in all pattern planes. Figure 3-14 illustrates the pattern of the rectangular array when the element feeding coefﬁcients are phased to scan the beam along one principal plane. The main beam broadens in the plane of scan as the effective array length is reduced but stays narrow in the orthogonal plane. More sidelobes appear behind the main beam. We see a large sidelobe growing on the horizon that will become a grating lobe when the array is scanned further. The sidelobes in the plane orthogonal to the plane of scan move with the main beam but roll into a cone that becomes tighter with increased scan.

FIGURE 3-13 Spherical radiation pattern of an 8 × 8-element uniform-amplitude and spaced square planar array.

122

ARRAYS

FIGURE 3-14 Spherical radiation pattern of an 8 × 8-element square-planar array scanned along a principal plane. 2.5 2 1.5 1

ky-space

0.5 0 −0.5 −1 −1.5 −2 −2.5 −2.5

−2

−1.5

−1

−0.5

0

0.5

1

1.5

2

2.5

kx-space

FIGURE 3-15 Contour plot of the pattern of a 4 × 4-element square array in kx ky -space showing multiple beams and sidelobes.

Figure 3-15 shows a contour plot of the universal pattern of a 4 × 4 element rectangular array. We denote this universal pattern kx ky space because the principal axes have sin θ factors similar to the universal pattern of a linear array. The array for Figure 3-15 has its y-axis element spacing 1.5 times wider than the x-axis spacing. The diagram axes extend until multiple beams show on the ﬁgure. The main beams correspond

PLANAR ARRAY

123

to the center of the large “squares.” The visible region on the ﬁgure is a unit circle with its center at the negative scan direction (−kx0 , −ky0 ). This technique mirrors the circle diagram of the linear array where the visible region is given by a linear region centered at the negative scan direction. You should notice that the diagonal sidelobes have smaller amplitudes than the principal plane sidelobes. We move the unit circle as the array scans and the diagram shows those locations of scan that have multiple beams (grating lobes). A grating analysis simpliﬁes the diagram of Figure 3-15 to the main beam locations. When we place the two axes of the planar array at an angle instead of orthogonal, we form a triangular array. Figure 3-16 gives the positions of a hexagon array made with equilateral triangles. We derive the characteristics of this array from a linear transformation of the rectangular array [6, p. 11-23ff]. Because the array has six-way symmetry, Figure 3-17 the pattern of a uniform-amplitude 61-element hexagon array shows the same six-way symmetry in the ring sidelobe around the main beam. If we collapsed the hexagonal distribution to a line in one plane, the distribution has a taper that reduces the sidelobes. The sidelobe amplitudes of the uniform hexagonal array are lower that the principal-plane sidelobes of the rectangular uniform array. Figure 3-18 plots the spherical pattern of the hexagon array when scanned to 36◦ . The ﬁrst ring sidelobe has a distorted six-fold symmetry. Similar to the scanned rectangular array (Figure 3-14), the hexagon array moves more sidelobes into visible space in the area opposite the scanned main beam. Figure 3-14 showed a grating lobe entering visible space, but the hexagon array pattern in Figure 3-18 does not. The grating lobes of a rectangular array can be found from a linear array when it is scanned along one of the principal axes, but the hexagon array requires a more elaborate analysis. When we scan the rectangular array off the principal axes, we can no longer use the grating lobe analysis of linear arrays.

FIGURE 3-16 Position of elements in a hexagonal planar array.

124

ARRAYS

FIGURE 3-17 Spherical radiation pattern of a 61-element hexagonal array.

FIGURE 3-18 Spherical radiation pattern of a 61-element hexagonal array scanned along a principal plane.

GRATING LOBES IN PLANAR ARRAYS

125

3-8 GRATING LOBES IN PLANAR ARRAYS The circle diagrams of the linear array can be used in principal planes of a rectangular array to compute grating lobes. For the planar array we use a sin θ pattern space to see the periodicity of the grating lobes and to analyze scanning in planes other than the principal axes. The visible region is now limited to a unit circle in kx ky -space where kx = sin θ cos φ and ky = sin θ sin φ. kx is the pattern in the x –z plane, and ky is the pattern in the y –z plane. Figure 3-19a shows the array layout, and Figure 3-19b shows the corresponding kx ky -plane grating lobe diagram. We reduce a contour plot of the pattern response similar to Figure 3-15 to only the main beams for analysis of grating lobes. The full contour plot is too busy. The narrower x-axis array spacing compared to the y-axis spacing leads to wider-spaced grating lobes in the kx -plane than in the ky -plane. Beam scanning corresponds to movement of the unit circle in kx ky -space. Each small circle in kx ky -space is a main beam in pattern space. When the unit circle encloses more than one kx ky -circle, the pattern has multiple main beams or grating lobes. The kx ky -plane diagram could also include sidelobe peaks or the contour plot of the array pattern and illustrate the pattern change with scan. The kx ky -plane is the two-dimensional Fourier transform of the distribution that becomes the periodic twodimensional Fourier series because the distribution is discrete. Increasing frequency or relative spacing between elements increases the unit circle diameter on an existing kx ky -diagram in a manner similar to the circle diagram. When we scan the beam, we move the unit-circle center in kx ky -space. We use Eq. (3-13) to locate the unit-circle center on the diagram: k(sin θ0 cos φ0 , sin θ0 sin φ0 ). The off-center circle on Figure 3-19 corresponds to a scanned beam and encloses two main beams. In this case the grating lobe does not lie in the scan plane and would fail to show in a simple pattern cut through the scan plane. A rectangular array produces a rectangular grating lobe diagram, while other periodic arrays lead to more complicated grating lobe diagrams. Figure 3-20 shows the layout of the hexagonal array and the corresponding grating lobe diagram. The hexagonal array (or equilateral triangular array) can be found from a linear transformation of the rectangular array. The grating lobe diagram can be found from the transformation as well. The spacing along the x-axis A1 corresponds to the y

k-y Main Beam l/A2 Visible Region, Broadside

A2

x λ/A1

A1 Scanned Beam (a)

k-x

Grating Lobe

(b)

FIGURE 3-19 (a) Grating lobe diagram of a rectangular array; (b) distribution in k-space.

126

ARRAYS

l/(a1 sin α) Grating Lobe a2

a a1

Visible Region Broadside Beam λ/(a2 sin a) Scanned Beam

(a)

(b)

FIGURE 3-20 (a) Grating lobe diagram of a hexagonal array; (b) distribution diagram.

vertical spacing B2 on the grating lobe diagram, and the spacing along the diagonal of each diagram are related. In both cases the corresponding axes on the two diagrams are perpendicular: λ λ B2 = and B1 = A1 sin α A2 sin α The angle between the triangular axes α is 60◦ for the hexagonal array. By allowing a grating lobe when the beam is scanned to 90◦ , we can determine the maximum element spacing without grating lobes: λ = 2 or A1 sin 60◦

1 A1 = = 0.577 λ 2 sin 60◦

Figure 3-20 shows the visible region unit circle with the beam broadside to the plane and then scanned to 36◦ for an element spacing of λ. When scanned, the unit circle encloses three lobes. The three lobes do not lie in a plane. Figure 3-21 gives the

FIGURE 3-21 Spherical radiation pattern of hexagon-array grating lobes.

SCAN BLINDNESS AND ARRAY ELEMENT PATTERN

127

spherical pattern of the array when scanned to 36◦ and shows the three lobes in the pattern. The array sidelobes have been reduced by sampling a circular Taylor distribution with the array so that the lobes show clearly. In Chapter 4 we discuss the use of continuous aperture distributions to determine the feed amplitudes of planar arrays. 3-9 MUTUAL IMPEDANCE Antennas in an array couple to each other because they receive a portion of the power radiated from nearby elements. This affects the input impedance seen by each element, which depends on the array excitation. We scan a phased array by changing the feeding coefﬁcients, and this changes the element input impedance called the scan impedance. To ﬁrst order, the coupling or mutual impedance is proportional to the element pattern level along the array face, and we reduce coupling by using narrower-beamwidth elements. Mutual coupling can be represented by an impedance, admittance, or scattering parameter matrix. The ﬁrst element of an N -element array has the impedance equation V1 = Z11 I1 + Z12 I2 + Z13 I3 + · · · + Z1N IN If we know the radiation amplitudes, we calculate the ratio of the currents: I2 I3 IN V1 = I1 Z11 + Z12 + Z13 + · · · + Z1N I1 I1 I1 The effective or scan impedance of the ﬁrst element is Z1 =

V1 I2 I3 IN = Z11 + Z12 + Z13 + · · · + Z1N I1 I1 I1 I1

(3-18)

It depends on the self-impedance and the excitation of all the other antennas. Scan impedance was formerly called active impedance, but this led to confusion. The power into the ﬁrst element is I2 I3 IN ∗ ∗ P1 = Re(V1 I1 ) = I1 I1 Re Z11 + Z12 + Z13 + · · · + Z1N (3-19) I1 I1 I1 By knowing the feeding coefﬁcients and the mutual impedances, we can compute the total input power and gain. In general, every antenna in the array has different input impedances. As the feeding coefﬁcients change in a phased array to scan the beam, so will the impedance of elements. The scan impedance change with scan angle causes problems with the feed network. We can repeat the same arguments for slots using mutual conductance, since magnetic currents are proportional to the voltage across each slot. 3-10 SCAN BLINDNESS AND ARRAY ELEMENT PATTERN [7, pp. 339–355; 8, pp. 365–366] Large arrays made from elements with wide beamwidths can exhibit scan blindness. When a phased array is scanned, at certain angles the input reﬂection coefﬁcient of every element rapidly increases to 1. The array fails to radiate and forms a pattern

128

ARRAYS

null. Mutual coupling between elements causes the change of the scan impedance, which leads to scan blindness, which is complicated and difﬁcult to predict accurately except where the array structure supports a surface wave. One approach says that scan blindness occurs when a grating lobe ﬁrst enters from invisible space and radiates along the surface of the array. In this case we solve Eq. (3-16) for the scan angle of the grating lobe: λ | cos θgr | = − 1 (3-20) d The angle in Eq. (3-20) is measured from the array plane (or axis). Scan blindness occurs approximately at this angle, but it can be reduced to only a dip in the pattern if the array is small or the mutual coupling between the elements is small, because they have narrow beams. The grating lobe causes a large increase in mutual coupling. Arrays made with antennas that can support surface waves, such as microstrip patches on dielectric substrates, can exhibit scan blindness when the electrical distance between the elements equals the surface-wave propagation phase shift: | cos θgr | =

λ ksw λ − = −P d k d

(3-21)

P is the relative propagation constant with a value > 1 for a surface wave (Section 101). Scan blindness will occur at an angle near this value because of the complicated nature of the coupling addition in the array. We can build a small portion of the array and determine where scan blindness will occur. Feed the center element and load all others with the feeder resistance. Each element in an array will couple to its neighboring elements and we can associate the combination of the element radiation and the coupled radiation of the neighbors when loaded to the element. We call this the array element pattern or scan element pattern (formerly called the active pattern). Elements near the edges will have different effective patterns, but in a ﬁrst-order solution we assume the pattern of the center element for all and calculate the total pattern as the product of the element pattern and the array factor. The array element pattern will exhibit dips where scan blindness will occur in the full array. Because it is only a small portion of the array, the full scan blindness will not occur. You should build a small array and test for scan blindness whenever it is a possibility. For example, arrays that scan to large angles off broadside using broad-beamwidth elements need to be tested with a small array before building the complete array. 3-11

COMPENSATING ARRAY FEEDING FOR MUTUAL COUPLING

Mutual coupling (impedance) is a measure of how much one antenna receives radiation from its neighbors in the array. Each element radiation changes the effective excitation on its neighbors. In a large array not requiring exact patterns, the effects average out. But when the array is small or you try to achieve low sidelobes, mutual coupling must be compensated for in the array. Small antenna elements such as dipoles or slots are resonant structures that radiate in only one mode. Mutual coupling only changes the element excitation, not the shape of the current distribution on the element. In this case we measure or calculate the mutual coupling matrix and use it to compute element

ARRAY GAIN

129

excitation to achieve the desired excitation [9]. Find the coupling matrix by adding the identity matrix to the S-parameter matrix of the antenna coupling: C=I+S

(3-22)

We compute the new feed excitation from the desired excitation and matrix inverse of Eq. (3-22): Vrequired = C−1 Vdesired (3-23) Because we assumed a single mode distribution on the antenna elements, S is independent of scanning and Eq. (3-23) gives the compensation for all scan angles. The compensation can be applied to the received signal in an adaptive array by matrix multiplication in digital signal processing. The effects of these operations have been illustrated [10]. Without compensation adaptive arrays, such as the MUSIC algorithm, only generate small peaks, whereas compensation produces the expected large peaks. Compensation for multimode elements starts with a moment solution [11] and uses the pattern characteristics to solve for the feeding coefﬁcients. We use the pattern desired to compensate the feeding coefﬁcients. We start with a matrix between the pattern response and the currents on all the antenna elements found from a moment method solution, where each array element has multiple current segments: A(k) = FI

(3-24)

A(ki ) is an element of the column matrix that gives the pattern response at an angle given by ki = xˆ sin θi cos φi + yˆ sin θi sin φi + zˆ cos θi or a given pattern angle (θi , φi ). The elements of the matrix F are the isotropic element phase terms, ej ki ·rj , and I is the column vector of the currents on the segments. We calculate excitation voltages by inverting the mutual impedance matrix: I = Z−1 V

(3-25)

We substitute Eq. (3-25) into Eq. (3-24) and note that the matrix V has only q nonzero terms corresponding to the feed points. We specify q pattern points, which reduces F to q × M for M current segments. The vector V has M − q zero elements and we delete the corresponding columns in the matrix product FZ−1 . This reduces the matrix to q × q, denoted B: A(k) = BV This uses the nonzero element V . We solve for the feeding coefﬁcients by inverting the matrix B found from q pattern points: V = B−1 A(k)

(3-26)

Choosing good pattern points is an art that requires pattern evaluation to verify whether the ﬁnal pattern is acceptable. 3-12 ARRAY GAIN We can use the mutual impedance concept to determine the effective input power of every element and thereby avoid having to integrate the pattern to calculate average

130

ARRAYS

radiation intensity. We represent the circuit relations of two antennas by a two-port impedance matrix: V1 Z11 Z12 I1 = V2 Z21 Z22 I2 Symmetrical elements across the diagonal of the matrix are equal for antennas satisfying reciprocity. The total input power is given by Pin = Re(V1 I1∗ ) + Re(V2 I2∗ ) The general N -element array has an N × N matrix and N terms in the input power sum. Given the feed coefﬁcients, we have a relation between different Ii . For our two elements, I2 = I1 ej δ and V1 = (Z11 + Z12 ej δ )I1 The power into the ﬁrst element is Re(Z11 + Z12 ej δ )I1 I1∗ By symmetry, the power into the second antenna is the same. The total input power to the array is Re(Z12 ej δ ) ∗ Pin = 2 Re(Z11 )I1 I1 1 + Re(Z11 ) The factor Re(Z11 )I1 I1∗ is the power into an isolated element: 4πE02 /η. The average radiation intensity (100% efﬁcient antenna) is Pin /4π: gain = directivity =

|2E(θmax )|2 |2E(θmax )|2 /η = Pin /4π 1 + [Re(Z12 ej δ )]/[Re(Z11 )]

(3-27)

By comparing Eqs. (3-27) and (3-8), we can identify sin(2πd/λ) cos δ R12 cos δ Re(Z12 ej δ ) = = Re(Z11 ) R11 2πd/λ sin(2πd/λ) R12 (d) = R11 2πd/λ We can use this mutual impedance ratio to compute directivity of arrays of isotropic elements of any number. Example Calculate the directivity of a linear array of three equally spaced isotropic elements with equal amplitudes and phases. The powers into the elements are 4πE02 R12 (d) R12 (2d) P1 = P3 = + 1+ η R11 R11 2 4πE0 32 E02 2R12 (d) Umax = P2 = 1+ η R11 η

ARRAY GAIN

131

The total power into the array is found from the sum: 4πE02 4R12 (d) 2R12 (2d) Pt = P1 + P2 + P3 = + 1+ η R11 R11 directivity =

Umax 9 = Pt /4π 3 + [4R12 (d)/R11 ] + [2R12 (2d)/R11 ]

The directivity of the general N -element equally spaced linear array, excited by equal-amplitude and equal-phase signals, is easily found by extending the development: directivity =

N 2 (element directivity) N−1 N +2 (N − M)[R12 (Md)/R11 ]

(3-28)

M=1

The directivity attained in an array depends on the particular mutual impedance terms of the radiators. The equation above only handles uniform-amplitude linear arrays. We can extend the idea of mutual resistance to calculate input power to a general planar array consisting of identical elements and determine gain. By using a two-element array spaced along the x-axis we can integrate the pattern to compute directivity and from that determine the ratio of mutual resistance to selfresistance of the elements versus element spacing: R12 (x) element directivity = R11 2π

2π 0

π 0

Ee2 (θ, φ) cos2

πx λ

cos φ sin θ sin θ dθ dφ − 1

(3-29) Equation (3-29) uses the normalized element pattern in the integral. By using an axisymmetrical element pattern, we calculate the ratio of resistances at a number of different distances and interpolate on the table for the directivity (gain) analysis of a planar (linear) array. If the element pattern is not symmetrical, the normalized resistance must be calculated for a number of φ. Given the element excitations Ei with elements located at the vector locations xi , we can derive an equation similar to Eq. (3-28) for directivity of a planar array:

directivity = N i=1

2

N

Ei (element directivity)

i=1 N [R12 (|xi − xj |)/R11 ]Re[Ej /Ei ]|Ei |2

(3-30)

j =1

Figure 3-22 illustrates the directivity calculated from Eq. (3-30) for linear arrays with realistic elements, such as a microstrip patch with 90◦ beamwidths, as the element spacing is varied. The graph shows directivity reduction when the element spacing exceeds λ and grating lobes form a more pronounced characteristic as the number of elements increases. When the second grating lobe occurs for wider element spacing, the directivity exhibits only minor variations. Increasing the element directivity (decreased beamwidth) reduces variation because the element pattern reduces the grating lobe. We use Eq. (3-30) with a planar array to obtain Figure 3-23. This array consists of 217 elements arranged in a hexagonal pattern, with amplitudes found from sampling a circular Taylor distribution (Sections 4-18 and 4-19) to lower the sidelobes. The 30-dB circular Taylor distribution reduces the gain by 0.6 dB relative to a uniform distribution

132

ARRAYS

32 elements, 60° Beamwidth 32 elements, 90° Beamwidth

Directivity, dB

16 elements, 60° Beamwidth 16 elements, 90° Beamwidth 8 elements, 60° Beamwidth 8 elements, 90° Beamwidth 4 elements, 60° Beamwidth 4 elements, 90° Beamwidth 2 elements, 60° Beamwidth 2 elements, 90° Beamwidth

Element Spacing, l

FIGURE 3-22 Directivity of a uniform-amplitude line array versus element spacing for 60◦ and 90◦ beamwidth elements.

16.2 dB Gain, 30° Beamwidths

Directivity, dB

12.8 dB Gain, 45° Beamwidths

10.4 dB Gain, 60° Beamwidths

8.6 dB Gain, 75° Beamwidths 7.3 dB Gain, 90° Beamwidths

Element Spacing Normalized to 0.6l

FIGURE 3-23 Directivity of a 217-element uniform-amplitude hexagonal array for various element beamwidths versus spacing.

due to the amplitude taper across the array. Initially, the element spacing is 0.6λ, but the element spacing has been allowed to grow in Figure 3-23 to show the effect. Figure 3-23 also illustrates the effect of increasing the element gain on the gain of a planar array. When we space the elements less than λ, increasing the element gain has no effect on array gain because the effective area of an antenna with a 90◦ beamwidth exceeds the area between elements and collects all power incident on the array. If we increase the element gain, the effective areas of the elements overlap and they share the incident power. On Figure 3-23 the curves overlap for element spacing less than

ARRAYS USING ARBITRARILY ORIENTED ELEMENTS

133

about λ. At the lower end of Figure 3-23 the gain increases by 6 dB as the element spacing doubles. This shows that increasing the element gain will have no effect on array gain when the present element covers the area associated with it. When the grating lobe enters from invisible space as the element spacing increases, those arrays with narrower-beam elements suppress the lobes and continue the general gain increase. The array directivity (gain) drops as the element spacing increases for the wide-beamwidth element with 90◦ beamwidth because of the grating lobe. At a large element spacing the array gain becomes N times the element gain. We determine array gain from the array area and the amplitude taper for closely spaced elements. For wide element spacing, we calculate gain from the product of the number of elements and the element gain. Figure 3-23 shows a smooth transition between the two regions. 3-13 ARRAYS USING ARBITRARILY ORIENTED ELEMENTS When we mount arrays on vehicles, the elements are pointed in arbitrary directions. Although Eq. (3-1) will calculate the pattern of any array, the element patterns are usually measured in a coordinate system in a different orientation than in the array. The idea of an array factor times the element pattern collapses and an analysis must rotate the pattern direction into the coordinates of each element. We will use coordinate rotations on the elements not only to specify them, but to calculate the pattern of the array. In a later chapter we use the same concept to point a feed antenna at a reﬂector. We rotate the pointing direction into the coordinate system of the orientated antenna to determine what direction angles to use for the element pattern. We do this by using a 3 × 3 rotation matrix on rectangular components: X (3-31) [Xrotated Yrotated Zrotated ] = [rotation matrix] Y Z A similar problem is rotating an object. Both cases use the same matrix. To rotate an object we multiply the vector by the rotation matrix from the left to compute the rotated coordinates. Rotating a position is given by the equation Xrotated [Xold Yold Zold ][rotation matrix] = Yrotated (3-32) Zrotated The rotation matrix can be found from the directions of the unit vectors when rotated. It is given by rotated X-axis (3-33) rotation matrix = rotated Y -axis rotated Z-axis The method uses 3 × 3 matrices to perform the rotation by a multiplication with a position or direction vector. Rotation about the X-axis is given by 1 0 0 0 cos A sin A 0 − sin A cos A

134

ARRAYS

rotation about the Y -axis is given by cos B 0 sin B

0 1 0

− sin B 0 cos B

and rotation about the Z-axis is given by cos C sin C − sin C cos C 0 0

0 0 1

We use products of these axis rotations to reorient an object or pointing direction. Consider the rotation of a position by the product of three rotation matrices: Xrotated [Xold Yold Zold ]R1 R2 R3 = Yrotated Zrotated The logical approach is to multiply the 3 × 3 matrices, R1 , R2 , and R3 , before multiplying by the position vector. When we postmultiply R1 by R2 , it rotates the axis of rotation of R1 . The postmultiplication by R3 rotates the rotation axis R2 and R1 is rotated once more. We can take the rotations one by one from left to right and use the rotation matrices about each of the principal axes provided that we convert the column vector back to a row vector after each multiplication. A convenient way to deﬁne the orientation of objects in space is to use spherical coordinate angles, since they are the same as pattern angles. We line up the matrices from right to left in this case. When rotating the coordinate system about an axis, the other axes change direction. The next rotations are about these new axes. The three rotations are often called the Euler angles. We use the following three rotations for spherical coordinate pointing: 1. Z-axis rotation = φ 2. New Y -axis rotation = θ 3. New Z-axis rotation: aligns the polarization of the antenna The last rotation takes some thought because the ﬁrst two rotations have altered the orientation of the antenna. When calculating the pattern of the array for a particular direction, ﬁrst compute rectangular components of the direction vector and the two polarization vectors. Multiply the direction vector by k(2π/λ) and take the dot (scalar) product with the position vector to calculate phasing of a particular element. You need to determine the pattern direction in the rotated antenna’s coordinate system found by using Eq. (3-31). Multiply the rotation matrix by the unit direction vector placed to the right. When you convert the output vector to spherical coordinates, you obtain pattern coordinates of the rotated antenna. Both the pattern components of the rotated element and the unit polarization vectors are needed. In the next operation you rotate the prime coordinate polarization unit vectors into the rotated element coordinate system using the same operation as the direction vector.

REFERENCES

135

You calculate ﬁnal radiated components by projecting the rotated prime coordinate polarization vectors on the element pattern unit polarization vectors: Eθ = Eθ,element θˆelement · θˆrotated + Eφ,element φˆ element · θˆrotated Eφ = Eθ,element θˆelement · φˆ rotated + Eφ,element φˆ element · φˆ rotated

(3-34)

Since we measure element patterns on antenna positioners, it is convenient to consider positioners as a series of coordinate system rotations. REFERENCES 1. C. A. Balanis, Antenna Theory, Analysis and Design, 2nd ed., Wiley, New York, 1997. 2. W. W. Hansen and J. R. Woodyard, A new principle in directional antenna design, Proceedings of IRE, vol. 26, March 1938, pp. 333–345. 3. T. C. Cheston and J. Frank, Array antennas, Chapter 11 in M. I. Skolnik, ed., Radar Handbook, McGraw-Hill, New York, 1970. 4. J. L. Butler, Digital, matrix, and intermediate frequency scanning, Section 3 in R. C. Hansen, ed., Microwave Scanning Antennas, Vol. III, Academic Press, New York, 1966. 5. J. Blass, The multidirectional antenna: a new approach to stacked beams, IRE International Convention Record, vol. 8, pt. 1, 1960, pp. 48–50. 6. Y. T. Lo and S. W. Lee, eds., Antenna Handbook, Vol. II, Van Nostrand Reinhold, New York, 1993. 7. R. J. Mailloux, Phased Array Antenna Handbook, Artech House, Boston, 1994. 8. P.-S. Kildal, Foundations of Antennas, Studentlitteratur, Lund, Sweden, 2000. 9. H. Steyskal and J. S. Herd, Mutual coupling compensation in small array antennas, IEEE Transactions on Antennas and Propagation, vol. AP-38, no. 12, December 1990, pp. 1971–1975. 10. A. G. Derneryd, Compensation of mutual coupling effects in array antennas, IEEE Antennas and Propagation Symposium, 1996, pp. 1122–1125. 11. B. J. Strait and K. Hirasawa, Array design for a speciﬁed pattern by matrix methods, IEEE Transactions on Antennas and Propagation, vol. AP-18, no. 1, January 1971, pp. 237–239.

4 APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

Continuous apertures and arrays share similar characteristics. We compute the radiation pattern of the aperture by using the Fourier transform. Array sampling of an aperture distribution leads to a Fourier series analysis for its pattern. We rely on our familiarity with signal processing to give us insights into these processes and their characteristics. We apply aperture theory to the analysis of horns, lens, and reﬂector antennas, but it also describes array antennas. Since we can design antennas only approximately to produce particular aperture distributions, we often realize them by sampling with an array. We start with aperture efﬁciencies developed from the Huygens source approximation of Section 2-2. We apply this method to horns, lens, and reﬂector antennas for both synthesis and tolerance analysis. The uniform and cosine distributions occur naturally in horns and simple resonant antennas. We use aperture distributions to realize bounds on antenna characteristics given size and excitation distribution. Taylor developed an aperture distribution based on Dolph’s use of the Chebyshev polynomials to produce the narrowest beamwidth for a speciﬁed sidelobe level for an array. The Chebyshev array design produces equal-amplitude sidelobes that we discover to be undesirable for large arrays because the equivalent aperture distribution peaks at the ends and the average value of the sidelobes limits the directivity to 3 dB above the sidelobe level. Large edge peaking of the distribution requires a feed network containing a large ratio of coupling values. Mutual coupling between elements causes unwanted excitation for a large ratio of element amplitudes and we lose control. Our usual practice is to sample a Taylor distribution for large arrays. The distribution has limited edge peaking, and large arrays can realize high gains. Aperture distribution synthesis involves manipulating pattern nulls to achieve desired characteristics. Taylor used the zeros of the Chebyshev array to alter the positions of the inner nulls of the uniform distribution to lower sidelobe levels. Elliott extended this Modern Antenna Design, Second Edition, By Thomas A. Milligan Copyright 2005 John Wiley & Sons, Inc.

136

AMPLITUDE TAPER AND PHASE ERROR EFFICIENCIES

137

idea to iterate the positions of these nulls to produce a linear aperture that radiates individually speciﬁed sidelobes. Schelkunoff developed a transformation between the pattern of an array and a polynomial where we combine the roots (or zeros) of the array polynomial in the complex plane with a mapped pattern variable that traverses the unit circle to analyze array patterns. We synthesize arrays by manipulating these polynomial zeros in the complex plane. Similar to Elliott’s method of null positioning for the continuous linear aperture, Orchard (and Elliott) developed an iterative method applied to array polynomial zeros to synthesize arrays. The method allows us to specify sidelobes individually and to shape the main beam pattern by moving some zeros off the unit circle. When designing shaped beams, improved synthesis by the Orchard method reduces our use of both array sampling of the Woodward continuous aperture method and direct Fourier series synthesis for linear arrays, but both earlier methods give us insight. We consider the design of series feeding where elements are fed directly from a transmission line for a linear array or continuous linear aperture. This requires speciﬁcation of the couplers or loading of the transmission line along the array because a portion of the power is extracted at each position with the remaining power dissipated in a load. We repeat aperture analysis for circular apertures to show limitations of large reﬂector antennas and for use in sampling with an array. For planar arrays, we reduce many rectangular apertures to the product of two linear distributions. A Chebyshev-type planar array with equal sidelobes can be designed so that the sidelobes in the diagonal planes are not reduced unnecessarily. Convolution synthesis of planar arrays allows manipulation of the pattern zeros in groups of smaller arrays similar to the Schelkunoff method. Finally, we consider aperture blockage and phase errors that lead to gain reduction and increased sidelobes.

4-1 AMPLITUDE TAPER AND PHASE ERROR EFFICIENCIES When we use the Huygens source approximation, we calculate power radiated by summing (integrating) the magnitude squared of the electric ﬁeld in the aperture and dividing by the impedance of free space. The average radiation intensity is the radiated power divided by the area of a unit sphere, 4π. To complete the calculation, we compute the maximum radiation intensity by dividing the maximum of the magnitude squared of Eq. (2-24) by the impedance of free space and directivity (Umax /Uavg ) becomes 2 j k · r Ee ds π(1 + cos θ )2 s max λ2 |E|2 ds

(4-1)

s

Equation (4-1) can be used for directivity in any pattern direction, including the maximum of the numerator integral. An aperture with a uniform amplitude and phase distribution has directivity 4πA/λ2 , where A is the area. We separate directivity reductions into individual terms due to

138

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

aperture ﬁeld amplitude and phase variations, and we express the general aperture directivity as 4πA directivity = 2 · ATL · PEL λ where ATL is the amplitude taper efﬁciency (loss) and PEL is the phase error efﬁciency (loss). Only amplitude variations contribute to ATL, and only phase variations determine PEL. We start with a uniform phase distribution in the aperture where the beam peak occurs normal to the aperture (θ = 0◦ ) and PEL = 1. We obtain uniform phase ﬁelds by using |E| in Eq. (4-1): directivity =

4π λ2

2

|E| ds

s

=

|E|2 ds

4πA · ATL λ2

s

where kx = ky = 0 on the boresight (θ = 0◦ ). On solving for ATL, we derive

2 |E| ds

s

ATL =

(4-2)

A

|E| ds 2

We have forced a constant phase everywhere in the aperture to separate out the amplitude taper effects. We account for nonuniform phase with PEL. The phase error efﬁciency can be found from PEL(θ, φ) =

directivity(θ, φ) (4πA/λ2 ) · ATL

where we use directivity (θ , φ) and PEL (θ , φ) depends on the pattern direction (θ, φ): 2 j k · r Ee ds (1 + cos θ )2 s PEL(θ, φ) = 2 4 |E| ds s

k = k(sin θ cos φ xˆ + sin θ sin φ yˆ + cos θ zˆ ) For an aperture in the x –y plane, k · r = k(x sin θ cos φ + y sin θ sin φ)

(4-3)

AMPLITUDE TAPER AND PHASE ERROR EFFICIENCIES

139

We determine maximum PEL to relate it and ATL to directivity. Traditionally, we use the boresight value (θ = 0◦ ) and Eq. (4-3) reduces to 2 E ds s PEL = 2 |E| ds

(4-4)

s

Unless speciﬁed, PEL will be Eq. (4-4) and we use Eq. (4-3) for scanned apertures. Equations (4-2) and (4-4) separate the effects of amplitude and phase variations in the aperture on the directivity at the boresight. If these efﬁciencies are expressed in decibels, the directivity becomes directivity(dB) = 10 log

4πA + ATLdB + PELdB λ2

Expressed in decibels, the efﬁciencies are called losses: amplitude taper loss (ATL) and phase error loss (PEL). It is important to remember that these are the losses at the boresight. A linear phase taper across the aperture scans the beam, but Eq. (4-4) predicts the boresight loss, which could be a null of the pattern. ATL is independent of phase variations that cause squinting of the beam. 4-1.1 Separable Rectangular Aperture Distributions If the distribution in a rectangular aperture is separable, E(x, y) = E1 (x)E2 (y) the efﬁciencies also are separable. and PEL = PELx PELy

ATL = ATLx ATLy

(4-5)

Given a rectangular aperture with an x-axis excursion of ±a/2, ATLx = a

a/2

2 |E1 (x)| dx

−a/2 a/2

−a/2

(4-6) |E1 (x)| dx 2

2 a/2 E1 (x) dx −a/2 PELx = 2 a/2 |E1 (x)| dx −a/2

The formulas for the y-axis are the same except for the substitution of y for x.

(4-7)

140

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

4-1.2 Circularly Symmetrical Distributions If a circular aperture has a circularly symmetrical distribution, we easily reduce Eqs. (4-2) and (4-4) to

a

2 0

ATL = a2

a

2 |E(r)|r dr (4-8) |E(r)| r dr 2

0

2 a E(r)r dr PEL = 0a 2 |E(r)|r dr

(4-9)

0

where a is the radius. We need a short word on formulas using integrals. They look formidable and seem to have little immediate practical use. In the catalog of distributions to follow, results will be given. A general distribution must be solved by numerical integration. One of the Newton–Cotes methods, such as Simpson’s rule or the Rhomberg integration, can be used when evenly spaced values are known. With a known function for the distribution, we use the Gauss–Legendre technique, whereby the method selects the required function values. It is sometimes easier to calculate the integrals numerically instead of writing routines for special functions that arise with circular apertures. Exact expressions are ideal; but unless a distribution is forced by a mode on the structure, it is difﬁcult to achieve the exact distribution. We need only approximations to the accuracy of practical interest.

4-2 SIMPLE LINEAR DISTRIBUTIONS We assume that rectangular apertures have separable distributions so that we can deal with one coordinate at a time. We compute the pattern in the plane containing the line. By drawing the pattern in kx (or ky )-space, we can calculate patterns independent of the aperture size in a way similar to that used for arrays in Chapter 3. In Chapter 2 we derived the kx -space pattern for a uniform distribution: a sin(kx a/2) kx a/2

(4-10)

where a is the aperture width and kx = k sin θ cos φ. We suppress cos φ and consider only patterns in the φ = 0◦ plane. Figure 4-1 shows the k sin θ space pattern of a uniform distribution. The pattern does not repeat at 2π intervals (radians) as the array does, but the sidelobes continue to decrease at a rate of 1/x. The ﬁrst sidelobe is 13.2 dB below the peak. The aperture size a, along with the scanning variable sin θ0 , determines the visible region in Figure 4-1. It ranges between ±ka/2 centered on ka/2 sin θ0 , since the maximum value of sin θ = 1.

SIMPLE LINEAR DISTRIBUTIONS

141

10

Pattern level, dB

20

30

40

50

60 −5 −25 −100

−4 −20 −80

−3 −15 −60

−2 −10 −40

−1 −5 −20

1 5 20

0

2 10 40

3 15 60

4 20 80

5 25 100

ka (sin q − sin ) 0 2

FIGURE 4-1 kx -space pattern of uniform line-source distribution.

Example An aperture is four wavelengths long. Determine the number of sidelobes between θ = ±90◦ when sin θ0 = 0 (boresight). The maximum value in (k sin θ )-space is 2π 4λ = 4π or λ 2

12.57

There are three sidelobes on each side of the main beam (Figure 4-1) in the visible region. The ﬁrst sidelobe occurs when ka/2 sin θ1 = 4.5, or θ1 = sin−1

4.5λ ◦ = 21 aπ

We found the half-power beamwidth in Chapter 2: HPBW = sin−1

0.4429λ a

(4-11)

valid when we ignore the obliquity factor, (1 + cos θ )/2. When we approximate x = sin x (radians) for small angles, we obtain ◦λ

HPBW = 50.76

a

(4-12)

We use this as the standard and describe other HPBW by their beamwidth factors. The beamwidth factor of the uniform distribution is 1.00. We also consider the null

142

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

beamwidth (BWnull ) of the distribution. The ﬁrst null occurs at ±π in the k sin θ pattern: λ ◦λ BWnull = 2 sin−1 ≈ 114.59 a a We also establish a beamwidth factor for the null beamwidth. When we scan the beam to a direction θ0 , the visible region centers at πa/λ sin θ0 in (k sin θ )-space. Example

Compute beam edges when θ0 = 30◦ and a = 6λ for a uniform distribution. a (sin θ1,2 − sin θ0 ) = ±0.4429 λ ±0.4429 + 0.5 sin θ1,2 = 6 ◦ ◦ θ1 = 35.02 θ2 = 25.23

The beamwidth is the difference, 9.79◦ . If we take the beam center as the average between the 3-dB beam edges, we get 30.12◦ for the beam center. By using the cosine of the beam center times the aperture size, we get 5.19λ, the projected aperture dimension. On substituting this in Eq. (4-11), we calculate HPBW = 9.79◦ . The actual beam peak is at θ = 30◦ , but the pattern is asymmetrical about θ0 . Other simple geometrical distributions on a linear aperture follow the same Fourier transform relation as the uniform distributions with differing transforms in (k sin θ )space. Table 4-1 lists the properties of some common distributions. Example Compute the beamwidth of a 7λ aperture with a cosine distribution. From Table 4-1, the beamwidth factor = 1.342. The taper increases the beamwidth over that of a uniform distribution:

50.76λ 0.4429λ ◦ ◦ HPBW = 1.342 = 9.73 or HPBW = 2 sin−1 1.342 = 9.74 a a We can add distributions and calculate the pattern from the sum of the transforms. Adding a pedestal (uniform distribution) to the cosine-squared distribution decreases TABLE 4-1 Common Linear Distribution Characteristics Distribution Uniform Triangular Cosine Cosine2

fx sin(kx a/2) kx a/2 sin(kx a/4) 2 kx a/4 cos(kx a/2) 2π 2 π − (kx a)2 sin(kx a/2) (kx a/2)[1 − (a/2λ)]2

First Sidelobe (dB)

HPBW Factor

BWnull Factor

ATR (dB)

13.2

1.000

1.000

0

26.5

1.439

2.000

1.25

23.0

1.342

1.5

0.91

31.5

1.625

2.000

1.76

SIMPLE LINEAR DISTRIBUTIONS

143

the beamwidth and the sidelobes of the cosine-squared distribution. The aperture distribution is given by E(x) = PD + (1 − PD) cos2

πx a

|x| ≤

a 2

where PD is the voltage pedestal level. The ﬁrst sidelobe of the uniform distribution lies within the null beamwidth of the cosine-squared distribution. The phase of sidelobes with respect to the main beam alternates between 180◦ and 0◦ , and the sidelobe of the pedestal subtracts from the main lobe. The second sidelobe of the pedestal occurs in almost exactly the same k-space location as the ﬁrst sidelobe of the cosine-squared distribution. These lobes cancel each other to some extent. Table 4-2 gives the required pedestal measured relative to the peak of the distribution for a given maximum sidelobe level. The minimum sidelobes (43.2 dB) occur for a pedestal level of −22.3 dB. At lower pedestal levels, the sidelobes rise and the beamwidth factor increases at a constant rate as the pedestal level decreases. The amplitude taper efﬁciency of the cosine squared on a pedestal is ATL =

2(1 + PD)2 3 + 2PD + 3PD2

(4-13)

ratio

Amplitude distributions based on simple functions have limited use. The uniform and cosine distributions or close approximations occur naturally, but the others must be forced on an aperture. An array can sample a distribution to achieve results similar to those for an aperture. A sampled cosine squared on a pedestal is handy for quick tolerance studies of array feed networks, but is far from optimum. Table 4-2 lists the pedestal to achieve a given sidelobe level for this distribution. We consider distributions that allow close control of sidelobes and achieve minimum beamwidths. The rate of decrease of the far-out sidelobe depends on the functional relation of the distribution at the edges [1]. If α is the exponent of the distribution approximation xeα , where xe is the distance from the edge, then the sidelobes decay as U −(1+α) , where U TABLE 4-2 Pedestal Level to Achieve a Given Maximum Sidelobe Level for a Cosine Squared on a Pedestal Distribution Sidelobe (dB)

Pedestal (dB)

Beamwidth Factor

ATL (dB)

30 32 34 36 38 40 42 42.7a 43.2b

−12.9 −14.2 −15.7 −17.3 −18.7 −20.0 −21.4 −21.9 −22.3

1.295 1.325 1.357 1.390 1.416 1.439 1.463 1.471 1.476

0.79 0.89 0.99 1.10 1.18 1.25 1.32 1.34 1.36

a b

Hamming distribution. Minimum sidelobe level.

144

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

is a linear function of the k-space variable. Both the triangular and cosine distributions have α = 1, and the far-out sidelobes decay as 1/U 2 . The cosine-squared distribution sidelobes decay as 1/U 3 , since α = 2. In the case of a cosine squared on a pedestal, the edge functional relation is a step (pedestal, α = 0) and the far-out sidelobes decay as 1/U . The sidelobes of the pedestal eventually overtake the cosine-squared distribution sidelobes, decreasing as 1/U 3 . To achieve uniform sidelobes, α must be −1, which occurs only when the distribution edges are Dirac delta functions, which requires inﬁnite energy in the aperture or design reduction to discrete sources (an array). We must accept a trade-off between radiated power in the main beam and in the sidelobes. When we narrow the main beam in a ﬁxed size aperture, more power radiates in the sidelobes. We achieve minimum beamwidth in the main beam when all the sidelobes radiate the same power (maximum radiated power in the sidelobes for a given level) and all sidelobes are at the same level. This case leads to the Dolph–Chebyshev array [2], impossible to duplicate in a continuous aperture. 4-3 TAYLOR ONE-PARAMETER LINEAR DISTRIBUTION [3] The uniform distribution has k-space zeros at (Figure 4-1) ±nπ, n = 1, 2, 3, . . . . Taylor deﬁnes a new variable U to replace k sin θ : sin πU πU

(4-14)

where U = (a/λ)(sin θ − sin θ0 ) and a is the aperture width. The nulls (zeros) are then located at integer values of U . Taylor adjusted the inner zeros of the uniform distribution to lower the sidelobes while retaining the outer zeros at their locations in the uniform distribution. The zeros are modiﬁed by a parameter B the boundary between the two regions in U -space: Un = n 2 + B 2 (4-15) The pattern has different expressions in two regions: √ B2 − U 2 sinh π √ |U | ≤ B π B2 − U 2 F (U ) = √ sin π U 2 − B 2 |U | ≥ B √ π U 2 − B2

(4-16a) (4-16b)

The high value of Eq. (4-16a) at the boresight depresses the sidelobes of the uniform distribution, and the parameter B controls all the parameters of the distribution. We compute B from the desired sidelobe level (SLR) by an iterative solution of the equation SLR = 13.26 + 20 log

sinh πB πB

(4-17)

Scale 4-1 gives the Taylor single-parameter distribution B for a given sidelobe level. The aperture distribution over the range −0.5 to 0.5 is given by the equation I0 [πB 1 − (2x)2 ] (4-18) I0 (πB)

TAYLOR ONE-PARAMETER LINEAR DISTRIBUTION

145

Design Sidelobe Level, dB

Taylor Single Parameter, B

SCALE 4-1 Taylor single-parameter B for a sidelobe level.

Design Sidelobe Level, dB

Taylor Single Parameter Edge Taper, dB

SCALE 4-2 Taylor single-parameter edge taper for a given sidelobe level.

Design Sidelobe Level, dB

Taylor Single Parameter Taper Loss, dB

SCALE 4-3 Taylor single-parameter amplitude taper loss for a given sidelobe level.

Design Sidelobe Level, dB

Taylor Single Parameter HPBW Factor

SCALE 4-4

Taylor single-parameter HPBW factor for a given sidelobe level.

Using Eq. (4-18), we calculate aperture edge taper as a function of sidelobe level, given by Scale 4-2. By inserting the expression for the aperture distribution [Eq. (4-18)] into Eq. (4-6), we calculate amplitude taper loss as a function of sidelobe level (Scale 4-3). The HPBW factor can be found from Eq. (4-16) or read easily from Scale 4-4. Figure 4-2 compares the U -space patterns of the Taylor one-parameter and uniform distributions. Synthesizing aperture distributions and arrays concentrates on the placement of pattern nulls. The one-parameter distribution scaled the locations of the nulls (zeros) by using Eq. (4-15). You should notice that the nulls approach those of the

146

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

30-dB Taylor One-Parameter Line Distribution

Pattern, dB

Uniform Distribution

U = a/l sin q

FIGURE 4-2 U -space pattern of 30-dB Taylor one-parameter linear distribution versus uniform distribution.

uniform distribution as U increases. Except for a shift near U = 0, the pattern falls off at a 1/U rate for far-out sidelobes. You can use the one-parameter Taylor distribution to estimate the characteristics of a linear distribution for a given sidelobe level. A comparison of this distribution to the cosine squared on a pedestal (Table 4-3) shows that it is not as efﬁcient for moderate sidelobe levels. The cosine squared on a pedestal distribution achieves low sidelobes by canceling sidelobes in two distributions and cannot be extended to any sidelobe level, whereas the one-parameter distribution can produce designs for any sidelobe level. More important, it demonstrates the systematic use of U -space pattern null placement for design. Taylor improved on this distribution by considering the zeros of the Dolph–Chebyshev array to ﬂatten the ﬁrst few sidelobes of the pattern response and achieved a more efﬁcient distribution.

TABLE 4-3 Comparison Between the Taylor One-Parameter Distribution and Cosine Squared on Pedestal Linear Distribution for Selected Sidelobe Levels Distribution

Pedestal (dB)

ATR (dB)

HPBW Factor

30-dB 30-dB 36-dB 36-dB 40-dB

−21.13 −12.9 −28.49 −17.3 −32.38

0.96 0.79 1.30 1.10 1.49

1.355 1.295 1.460 1.390 1.524

one-parameter cos2 + pedestal one-parameter cos2 + pedestal one-parameter

TAYLOR n LINE DISTRIBUTION

147

4-4 TAYLOR n LINE DISTRIBUTION [1] The Taylor n line-source distribution modiﬁes the location of the inner pattern zeros (nulls) of a uniform distribution to approximate the Dolph–Chebyshev array. The distribution contains a pedestal α = 0 and retains the 1/U fall-off for the far-out sidelobes. We can modify any number of inner zeros of the pattern to approximate the uniform sidelobe-level array, but we force the aperture voltage to peak at the ends in approximating the Dirac delta functions. We limit the number of altered zeros to keep the distribution practical. After a point, shifting more zeros reduces beamwidth negligibly. We manipulate the location of pattern zeros to obtain desired patterns. Both aperture and array syntheses depend on zero locations. The number of array elements determines the number of independent zeros (n − 1), but a continuous aperture has an inﬁnite number of independent zeros. Practical consideration of the distribution edge shape limits the number, but we are free to move zeros. For a given aperture size, we can move zeros out of the invisible region into the visible region and narrow the main beam as much as we want while maintaining low sidelobes. The invisible region represents stored energy in the aperture. When a zero moves out of the invisible region, the amount of stored energy and the Q of the antenna increase. The overall efﬁciency of the antenna decreases while the antenna becomes more and more narrowband. We call these arrays superdirective because their directivity exceeds that of a uniform distribution. The Taylor line-source distribution retains the zeros in the invisible region and prevents superdirectivity. There is no limit to the directivity achievable on paper for a given aperture, but the theoretical distributions are unrealizable except for very small levels of superdirectivity. The costs of superdirectivity are decreased bandwidth and efﬁciency. We will modify the location of the ﬁrst n − 1 pairs of inner nulls to lower the sidelobes. Choosing the zeros symmetrically about the origin of U -space gives us a constant phase distribution. We remove the inner zeros by dividing them out of the uniform distribution U -space pattern: sin πU n−1 πU 1 − U 2 /N 2 N=1

We then add new nulls Un without becoming superdirective: n−1

U2 1− 2 sin πU N=1 UN F (U ) =

n−1 U2 πU 1− 2 N=1 N

(4-19)

Because we want to approximate the DoIph–Chebyshev response, we choose the inner zeros from the array: A2 + (N − 12 )2 UN = n A2 + (n − 12 )2

N = 1, 2, . . . , n − 1

(4-20)

148

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

where A relates to the maximum sidelobe level: cosh πA = b

(4-21)

in which 20 log b = sidelobe level. Equation (4-19) gives us the U -space (k-space) pattern of the distribution with modiﬁed zeros. We determine the aperture distribution by expanding it in a Fourier cosine series: E(x) =

∞

Bm cos 2mπx

|x| ≤ 0.5

(4-22)

m=0

where the aperture size has been normalized. We calculate the pattern of the distribution from the Fourier transform: 1/2 1/2 j kx x f (kx ) = E(x)e dx or f (U ) = E(x)ej 2πU x dx −1/2

−1/2

We substitute Eq. (4-22) for E(x) and reverse the order of summation and integration: ∞

f (U ) =

Bm

m=0

1/2

−1/2

cos 2mπx cos 2πU x dx

(4-23)

Since the aperture function is an even function, the odd-function part of the integral is zero, as reﬂected in Eq. (4-23). We calculate coefﬁcients Bm by matching the patterns at integer values of U . The integral [Eq. (4-23)] is zero unless U = m: B0 = f (0)

Bm = f (m) 2

m = 1, 2, . . . , n − 1

Since we have only modiﬁed the location of the ﬁrst n − 1 zeros of the U -space pattern, f (m) = 0 for m ≥ n and the Fourier cosine series has only n components: E(x) = f (0) + 2

n−1

f (m) cos 2mπx

(4-24)

m=1

The coefﬁcients are given by f (0) = 1

n−1 (−1)m 1 − m2 /UN2 N=1 f (m) = n−1 −2 1 − m2 /N 2

m = 1, 2, . . . , n − 1

(4-25)

N=1,N =m

Equation (4-19) computes the U -space pattern of the Taylor distribution but requires L’Hospital’s rule at integer values of U . The ﬁnite number of coefﬁcients Bm makes Eq. (4-23) more convenient since the integral is easily solved: n−1 sin(πU ) 1 sin[π(U − i)] sin[π(U + i)] Bi + + f (U ) = B0 πU 2 i=1 π(U − i) π(U + i)

(4-26)

TAYLOR n LINE DISTRIBUTION

149

Example Design the Taylor line-source distribution with 30-dB maximum sidelobes and n = 6. We use Eq. (4-21) to calculate A: b = 1030/20 = 31.6228 A=

cosh−1 b = 1.3200 π

We substitute this constant into Eq. (4-20) to compute the ﬁve (n − 1) nulls: No. Null UN

1

2

3

4

5

1.4973

2.1195

2.9989

3.9680

4.9747

The ﬁrst null value gives us the BWnull factor (1.4973). The null beamwidth has been increased almost 50% relative to the uniform distribution. The coefﬁcients of the Fourier cosine aperture distribution are found from Eqs. (4-24) and (4-25) (Table 4-4). Coefﬁcients of the series are normalized so that the distribution is 1 at x = 0, and the amplitude distribution is found by plotting the Fourier cosine series. We calculate the U -space pattern by using Eq. (4-26). We calculate the half-power point and compare it to the uniform distribution to determine the HPBW factor, 1.2611. By using Eq. (4-6), we calculate ATL = 0.66 dB for the distribution. The U -space plot (Figure 4-3) of the example above shows the 30-dB sidelobe level. The ﬁrst sidelobe is at 30 dB, and lobes after that fall away from 30 dB. With a higher value of n, the ﬁrst unchanged zero, more sidelobes would be nearer 30 dB. The dashed curve gives the pattern of a uniform distribution. Notice that the inner ﬁve nulls have been shifted to lower the sidelobes. At the sixth null and higher, the Taylor distribution has the same nulls as the uniform distribution. The n distribution has a narrower beamwidth than the one-parameter distribution (Figure 4-2) and a higher taper efﬁciency of 0.66 dB versus 0.96 dB. Figure 4-4 shows the normalized aperture voltage for 30-dB-maximum sidelobe Taylor distributions. The one-parameter design produces a lower pedestal than the two n designs. The n = 20 design voltage peaks as it approaches the edge. This peaks because the Taylor n distribution approximates the Dolph–Chebyshev array that peaks at the edge of the array. The amplitude taper efﬁciency was calculated for a number of designs and is given in Table 4-5. The corresponding beamwidth factors are listed in Table 4-6 together with TABLE 4-4 Fourier Cosine Series Coefﬁcients for Taylor Distribution: 30 dB, n = 6 No. 1 2 3 4 5 6

Bm

Bm Normalized

Function

1.0000 0.5733 −0.0284 −0.000213 0.005561 −0.003929

0.64672 0.37074 −0.01838 −0.000138 0.003597 −0.002541

1 cos 2πx cos 4πx cos 6πx cos 8πx cos 10πx

150

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

30-dB Taylor Line Distribution

Pattern, dB

Uniform Distribution

U = a/l sin q

Normalized Voltage

FIGURE 4-3 U -space pattern of 30-dB Taylor n = 6 linear distribution versus uniform distribution.

12 One Parameter

6

Normalized Aperture

FIGURE 4-4 A 30-dB Taylor linear aperture distribution comparison.

the null beamwidth factors (location of ﬁrst zero in U -space) in Table 4-7. ATL depends on the sidelobe level (Table 4-5) more than the number of modiﬁed zeros. Both the 20and 25-dB sidelobe levels show that there is an optimum number of zeros. The edge of the distribution peaks toward the Dirac delta function and reduces the amplitude taper efﬁciency. More than three modiﬁed zeros are needed to reduce the sidelobes

TAYLOR n LINE DISTRIBUTION

151

TABLE 4-5 Amplitude Taper Losses of Taylor Line-Source Distributions Sidelobe Level (dB)

TABLE 4-6

n

20

25

30

35

40

45

50

4 5 6 7 8 10 12 16 20

0.17 0.15 0.15 0.15 0.16 0.19 0.24 0.35 0.46

0.43 0.41 0.39 0.37 0.36 0.34 0.34 0.35 0.27

0.69 0.68 0.66 0.65 0.63 0.61 0.59 0.57 0.56

0.95 0.93 0.92 0.91 0.90 0.88 0.86 0.84 0.82

1.16 1.15 1.15 1.14 1.13 1.11 1.09 1.07

1.37 1.36 1.36 1.35 1.34 1.32 1.30

1.56 1.55 1.55 1.54 1.53 1.51

Beamwidth Factor of Taylor Line-Source Distribution Sidelobe Level (dB)

n

20

25

30

35

40

45

50

4 5 6 7 8 10 12 16 20

1.1043 1.0908 1.0800 1.0715 1.0646 1.0545 1.0474 1.0381 1.0324

1.1925 1.1837 1.1752 1.1679 1.1617 1.1521 1.1452 1.1358 1.1299

1.2696 1.2665 1.2611 1.2555 1.2504 1.2419 1.2353 1.2262 1.2203

1.3367 1.3404 1.3388 1.3355 1.3317 1.3247 1.3189 1.3103 1.3044

1.4065 1.4092 1.4086 1.4066 1.4015 1.3967 1.3889 1.3833

1.4733 1.4758 1.4758 1.4731 1.4695 1.4628 1.4576

1.5377 1.5400 1.5401 1.5379 1.5326 1.5280

TABLE 4-7

Null Beamwidth Factor of Taylor Line-Source Distributions Sidelobe Level (dB)

n

20

25

30

35

40

45

50

4 5 6 7 8 10 12 16 20

1.1865 1.1696 1.1566 1.1465 1.1386 1.1270 1.1189 1.1086 1.1023

1.3497 1.3376 1.3265 1.3172 1.3095 1.2978 1.2894 1.2783 1.2714

1.5094 1.5049 1.4973 1.4897 1.4828 1.4716 1.4632 1.4518 1.4444

1.6636 1.6696 1.6671 1.6632 1.6569 1.6471 1.6392 1.6277 1.6200

1.8302 1.8347 1.8337 1.8306 1.8231 1.8161 1.8051 1.7975

1.9990 2.0031 2.0032 1.9990 1.9934 1.9835 1.9760

2.1699 2.1739 2.1740 2.1705 2.1623 2.1553

below 40 dB; hence, the blanks represent unrealizable designs. The beamwidth factor (Table 4-6) reduces with increasing n, but it depends mainly on the sidelobe level. Example Compute beamwidths and ATL of an 8λ-wide aperture with n = 8, 40-dB sidelobes, and a Taylor line-source distribution design.

152

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

From Table 4-6, HPBW = From Table 4-7, BWnull =

1.4066(50.76◦ ) ◦ = 8.92 8

1.8306(114.59◦ ) ◦ = 26.22 8

From Table 4-5, ATL = 1.14 dB. A square aperture with the same distribution in both directions has directivity = 10 log

4πA − 2ATL = 26.77 dB λ2

4-5 TAYLOR LINE DISTRIBUTION WITH EDGE NULLS Rhodes [4] has shown that it is impossible to have a step discontinuity of the ﬁelds at the edge of a physical aperture. Given the radius of curvature of the edge, ρ, the ﬁeld varies as C2 d ρ C1 2 d Es ∼ ρ

Ed ∼

polarized perpendicular to the aperture edge polarized parallel to the aperture edge

where C1 and C2 are constants and d is the distance from the edge. Without the possibility of an edge pedestal, a traditional Taylor line source cannot be realized with a physical aperture. We can sample the distribution with an array or closely approximate it, but we cannot achieve the exact distribution. A Taylor distribution with a null at the edge can be realized in an aperture. Rhodes [5] extended the Taylor line source by modifying the U -space pattern zeros of the cosine distribution. Since α = 1, the far-out sidelobes drop off as 1/U 2 and the distribution is zero on the edges. The zeros of the cosine distribution occur at (N + 1/2)π

N = 1, 2, 3, . . .

k space

When the Taylor U -space variable is used, the modiﬁed U -space pattern becomes n−1 1 − U 2 /UN2 N=1 f (U ) = n−1 2 [1 − (2U ) ] (1 − U 2 /(N + 12 )2 ) cos πU

(4-27)

N=1

We remove the inner n − 1 zeros at N +

1 2

and substitute new ones given by

UN = ±(n +

1 ) 2

A2 + (N − 12 )2 A2 + (n − 12 )2

N = 1, 2, . . . , n − 1

(4-28)

TAYLOR LINE DISTRIBUTION WITH EDGE NULLS

153

When we compare Eqs. (4-28) and (4-20), we see that the nulls are shifted by (n + 12 )/n between the two Taylor distributions. When n is large, the nulls are close to the same for the two distributions. To determine the amplitude distribution in the aperture, we expand the aperture ﬁelds in a Fourier cosine series, E(x) =

Bm cos(2m + 1)πx

|x| ≤ 0.5

(4-29)

Like the Taylor line source, there are only n terms in the series whose coefﬁcients are found by equating the pattern from the Fourier transform of Eq. (4-29) to Eq. (4-27). The coefﬁcients are given by n−1 2 1 − 14 /UN2 B0 = n−1 N=1 1 − 14 /(N + 12 )2 N=1

n−1 (−1)m (m + 12 ) 1 − (m + 12 )2 /UN2 N=1 Bm = n−1 [1 − (2m + 1)2 ] 1 − (m + 12 )2 /(N + 12 )2

m = 1, 2, . . . , n − 1

N=1,N =m

(4-30)

The U -space pattern can be found using the coefﬁcients Bm : f (U ) = C0

n−1

Bi

sin[π(U − i − 12 )] π(U − i − 12 )

i=0

C0 = 2

n−1 i=0

Bi

+

sin[π(U + i + 12 )]

π(U + i + 12 )

(4-31)

sin[π(i + 12 )] π(i + 12 )

Example Design the Taylor line-source distribution with edge nulls for 30-dB maximum sidelobes and n = 6. We use Eq. (4-21) to calculate A: b = 1030/20 = 31.6228 A=

cosh−1 b = 1.32 π

the same as the pedestal edge Taylor line-source distribution. We substitute this constant into Eq. (4-28) to compute the ﬁve modiﬁed nulls: No. Null UN

1

2

3

4

5

1.6221

2.2962

3.2488

4.2987

5.3892

The null locations have increased by (n + 12 )/n = 6.5/6 = 1.0833 from the pedestal Taylor line-source design. The null beamwidth factor has also increased by this factor

154

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

as well. The coefﬁcients of the Fourier cosine aperture distribution are found from Eq. (4-30) Table 4-8. The normalized coefﬁcients sum to 1 at x = 0. Equation (4-27) determines the U space pattern given the nulls. On ﬁnding the half-power point and comparing it with the uniform distribution half-power point, we compute the beamwidth factor: 1.3581. Tables 4-9 to 4-11 give results for this Taylor line source. As n increases, the results approach the result of the pedestal Taylor line source. Since the maximum sidelobe of the cosine distribution is 23 dB, a distribution must have peaking toward the edges to raise the sidelobes above that level. In all distributions the voltage approaches zero linearly at the edges. TABLE 4-8 Fourier Cosine Series Coefﬁcients for Taylor Distribution with Edge Nulls: 30 dB, n = 6 Bm

Bm Normalized

Function

0.50265 0.023087 0.017828 −0.010101 0.007374 −0.003245

0.94725 0.04351 0.02220 −0.02075 0.01390 −0.006116

cos πx cos 3πx cos 5πx cos 7πx cos 9πx cos 11πx

No. 1 2 3 4 5 6

TABLE 4-9 Amplitude Taper Losses of a Taylor Line Source with Edge Null Distribution Sidelobe Level (dB) n

25

30

35

40

45

50

4 6 8 12 16 20

0.86 0.67 0.56 0.45 0.41 0.39

1.13 0.97 0.87 0.74 0.68 0.64

1.36 1.24 1.14 1.02 0.96 0.92

1.55 1.47 1.39 1.28 1.22 1.17

1.71 1.66 1.60 1.51 1.45 1.41

1.84 1.84 1.79 1.71 1.66 1.62

TABLE 4-10 Beamwidth Factor of a Taylor Line Source with Edge Null Distribution Sidelobe Level (dB) n

25

30

35

40

45

50

4 6 8 12 16 20

1.3559 1.2666 1.2308 1.1914 1.1705 1.1576

1.4092 1.3581 1.3242 1.2850 1.2635 1.2502

1.4815 1.4407 1.4097 1.3716 1.3500 1.3363

1.5443 1.5153 1.4882 1.4522 1.4308 1.4170

1.5991 1.5831 1.5608 1.5276 1.5068 1.4930

1.6470 1.6448 1.6280 1.5984 1.5785 1.5649

155

ELLIOTT’S METHOD FOR MODIFIED TAYLOR DISTRIBUTION

TABLE 4-11 Null Beamwidth Factor of a Taylor Line Source with Edge Null Distribution Sidelobe Level (dB) n

25

30

35

40

45

50

4 6 8 12 16 20

1.5184 1.4371 1.3913 1.3431 1.3182 1.3031

1.6980 1.6221 1.5755 1.5242 1.4971 1.4805

1.8715 1.8060 1.7604 1.7075 1.6786 1.6605

2.0374 1.9875 1.9450 1.8918 1.8616 1.8424

2.1949 2.1656 2.1284 2.0765 2.0455 2.0254

2.3433 2.3395 2.3097 2.2610 2.2298 2.2091

4-6 ELLIOTT’S METHOD FOR MODIFIED TAYLOR DISTRIBUTION AND ARBITRARY SIDELOBES [6, pp. 162–165] Elliott’s method separates the distribution nulls into right- and left-hand values in U space that allows different sidelobe levels in the two regions. By applying a differential expression, the null positions in U -space can be found from the solution of a set of linear equations to produce designs with arbitrary sidelobes. Consider Eq. (4-19) and factor the null location term:

U2 U U U U 1− 2 = 1+ 1− = 1+ 1− UN UN UNL UNR UN We associate UNL with a nulls on the left side of the origin and UNR with the right side or a positive pattern angle. If we also separate the term in the denominator of Eq. (4-19), we can independently pick the number of nulls to be moved on either side of the pattern:

F (U ) = C0

nL −1

nR −1 (1 + U/UN ) (1 − U/UN ) N=1 N=1 nL −1 nR −1 πU (1 + U/N ) (1 − U/N )

sin πU

N=1

(4-32)

N=1

Equation (4-32) allows different Taylor distributions on the two sides. We add a normalization factor C0 when we use different distributions. The pattern peak will shift off zero for unbalanced distributions. Since the two sides are not independent, a simple selection of the two levels will not produce the desired sidelobes. Table 4-12 lists the U -space locations of the pattern peaks and sidelobe level for a design with 35- and 30-dB sidelobes. The left distribution lowered the sidelobes on the right and the right one raised the left sidelobes. A few manual iterations produced suitable left and right distributions to give the desired sidelobes. The main beam shifts a little bit. A linear progressive phase shift across the aperture can shift the pattern to broadside. We expand the aperture distribution in a complex exponential series similar to Eq. (4-22): n R −1 E(x) = Bi e−j 2πix |x| ≤ 0.5 (4-33) i=−nL +1

156

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

TABLE 4-12 Modiﬁed Taylor Distribution Sidelobes for Independent Left and Right Sidelobe Design Using n = 6 for Both Sides Left-Side−35-dB U -Space

Right Side−30-dB Sidelobe (dB)

Left-Side−36-dB U -Space

Right-Side−28.6-dB Sidelobe (dB)

−5.4849 −4.4905 −3.5275 −2.6313 −1.8997 −0.0511 1.7546 2.5372 3.4697 4.4584 5.4711

−35.70 −35.06 −34.66 −34.44 −34.39 0 −31.05 −31.45 −32.00 −32.74 −33.75

−5.4874 −4.4976 −3.5399 −2.6510 −1.9293 −0.0758 1.7141 2.5119 3.4544 4.4498 5.4676

−36.00 −35.44 −35.11 −34.97 −34.99 0 −30.05 −30.55 −31.19 −32.02 −33.13

We calculate the coefﬁcients by the same method used for Eq. (4-25): B(0) = 1

nR −1 nL −1 (−1)|m| N=1 (1 + m/UN ) (1 − m/UN ) N=1 B(m) = n −1 nR −1 L − (1 + m/N ) (1 − m/N ) N=1,N =m

N=1,N =m

m = −nL + 1, . . . , −1, 1, 2, . . . , nR − 1

(4-34)

We derive the pattern from the integral of the ﬁnite complex exponential: f (U ) = C0

n R −1 i=−nL +1

Bi

sin[π(U − i)] π(U − i)

(4-35)

We include the normalization factor C0 for unequal left and right sidelobes. We control the sidelobes by adjusting the location of the nulls in the U -space pattern. We can iterate the null positions to produce individually selected sidelobes. The peak of each sidelobe given in Table 4-12 was found through a one-dimensional search between pairs of nulls. A search based on the Fibonacci numbers [7, p. 280] computes the peak with the minimum number of evaluations of the pattern using Eq. (4-35). p We denote the pattern peaks by Um starting with the peak between −nL and U−nL +1 , those between nulls, the peak near 0, and the last peak between UnR −1 and nR for nL + nR − 1. We adjust the U -space nulls by the differentials δUN found from the solution of a matrix equation. The terms of the matrix are the differential term of a Taylor series expansion of the numerator of Eq. (4-32) evaluated at the pattern peaks: p

am,n

Um /UN2 = p 1 − Um /UN

am,0 = 1

N = −nL + 1, . . . , −1, 1, . . . , nR − 1 (4-36)

157

ELLIOTT’S METHOD FOR MODIFIED TAYLOR DISTRIBUTION

The vector of differential nulls is δU = [δU−nL +1 , . . . , δU−1 , δC/C0 , δU1 . . . , δUnR −1 ]T where δC/C0 is the change in the pattern normalization. We form a vector using the ratio of the desired pattern peak fd (U ) to the actual pattern fa (U ) with terms p p [fd (Um )/fa (Um )] − 1. We solve the matrix equation for the null shifts: [am,n ]

−1

T p fd (Um ) = [δUN ] p −1 fa (Um )

(4-37)

We calculate new distribution nulls UN + δUN , substitute the new nulls into Eq. (4-34) to determine the new expansion coefﬁcients Bm , and evaluate the pattern using Eq. (4-35) between the new nulls to compute new pattern peaks. We iterate the process until the sidelobe levels are satisfactory. Notice that f (U ) is a voltage. The Taylor linear distribution produces a pattern with only approximately equal sidelobes. Table 4-13 lists the iteration to produce a distribution with a pattern that has ﬁve 30-dB sidelobes. The solution starts with a 30-dB, n = 6 Taylor distribution. In two iterations the method found a distribution with exactly the desired sidelobes. Table 4-14 gives the results of repeating the example of Table 4-12 of the design for 35and 30-dB sidelobes. This method can produce a linear distribution with individually TABLE 4-13

Iteration of Distribution Nulls for a Pattern with 30-dB Sidelobes

Taylor Distribution Null 1.4973 2.1195 2.9989 3.9680 4.9747

First Iteration

Second Iteration

U -Space

Sidelobe

Null

U -Space

Sidelobe

Null

U -Space

Sidelobe

1.7557 2.5387 3.4709 4.4591 5.4718

−30.22 −30.46 −30.89 −31.53 −32.48

1.4708 2.0827 2.9490 3.9075 4.9145

1.7258 2.4987 3.4215 4.4072 5.4424

−30.00 −29.99 −29.96 −29.86 −29.63

1.4729 2.0859 2.9541 3.9152 4.9242

1.7284 2.5027 3.4274 4.4147 5.4471

−30.00 −30.00 −30.00 −30.00 −29.99

TABLE 4-14

Iteration for 35- and 30-dB Sidelobes in Linear Distribution

Left-Side−36-dB U -Space −5.4874 −4.4976 −3.5399 −2.6510 −1.9293 −0.0758 1.7141 2.5119 3.4544 4.4498 5.4676

Right-Side−28.6-dB Sidelobe (dB) −36.00 −35.44 −35.11 −34.97 −34.99 0 −30.05 −30.55 −31.19 −32.02 −33.13

Second Iteration Null

U -Space

Sidelobe (dB)

Null

−4.9964 −4.0169 −3.0845 −2.2583 −1.7008

−5.4798 −4.4859 −3.5348 −2.6555 −1.9410 −0.1037 1.6615 2.4475 3.3839 4.3839 5.4313

−35.00 −35.00 −35.00 −35.00 −35.00 0 −30.00 −30.00 −30.00 −30.00 −29.98

−4.9778 −4.0043 −3.0829 −2.2670 −1.7143

1.4495 2.0883 2.9801 3.9574 4.9699

1.4023 2.0249 2.9049 3.8780 4.9001

158

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

selected sidelobes. It may be necessary to design an intermediate distribution if the change in sidelobes is too great for the simple iteration scheme to converge. We will design linear arrays by a similar technique of manipulating pattern nulls to produce arbitrary sidelobes.

4-7 BAYLISS LINE-SOURCE DISTRIBUTION [8] The Bayliss distribution produces a pattern null on a boresight while controlling the height of the sidelobes. The second dashed curve in Figure 4-5 below is a Bayliss difference pattern also designed to give 30-dB sidelobes when combined with the Taylor distribution. As in the Taylor distribution, the ﬁrst few sidelobes are nearly the same height, to minimize the beamwidth of the two beams split about the boresight. A monopulse tracking system uses an auxiliary pattern with a boresight null coincident with the beam peak of the main pattern. The tracking system drives the antenna positioner until the signal in this difference channel nulls so that the main channel (sum) points at the emitter or radar target. The accuracy of the pointing angle is improved, since a null is a more exact direction than the broad sum pattern peak. Noise and receiver sensitivity, along with the slope of the difference pattern, limit the tracking accuracy. Stronger signals can be tracked farther into the null. Because the phase of a pattern shifts by 180◦ when passing through a null, phase relative to the sum pattern (a reference signal) can be used to give direction. Without monopulse or some other sequential lobing technique, such as conical scan, radar cannot track effectively.

10

20

Pattern, dB

Taylor 30 dB

Bayliss 28 dB

30

40

50

60 −20

0 ka (sin q − sin ) 0 2

FIGURE 4-5

Taylor and Bayliss line distributions to give 30-dB sidelobes (n = 6).

20

BAYLISS LINE-SOURCE DISTRIBUTION

159

Any odd-function distribution produces a null on a boresight. A uniform distribution that switches phase by 180◦ in the center has the best amplitude taper efﬁciency but high sidelobes (10 dB). These high sidelobes allow interfering or noise signals to enter the receiver. The Bayliss distribution adjusts the inner nulls of the U -space pattern to lower the sidelobes. Adjusting the zeros to correspond to the Dolph–Chebyshev array does not lower the sidelobes to the same level as it did in the Taylor distribution. Further adjustments of the four inner zeros are required. Bayliss found the proper location through a computer search. We locate the zeros by ξN2 (n + 12 ) N = 1, 2, 3, 4 A2 + n2 UN = (4-38) 2 2 A + N 1 N = 5, 6, . . . , n − 1 (n + 2 ) A2 + n2 By using the U -space pattern, we have n−1

1 − U 2 /UN2

f (U ) = U cos πU n−1N=1 1 − U 2 /(N + 12 )2

(4-39)

N=0

The coefﬁcients were ﬁtted to polynomials depending on the sidelobe level. Given S = |sidelobe level(dB)|: A = 0.3038753 + S(0.05042922 + S(−0.00027989 + S(0.343 × 10−5 − S(0.2 × 10−7 ))))

(4-40a)

ξ1 = 0.9858302 + S(0.0333885 + S(0.00014064 + S(−0.19 × 10−5 + S(0.1 × 10−7 ))))

(4-40b)

ξ2 = 2.00337487 + S(0.01141548 + S(0.0004159 + S(−0.373 × 10−5 + S(0.1 × 10−7 ))))

(4-40c)

ξ3 = 3.00636321 + S(0.00683394 + S(0.00029281 + S(−0.161 × 10−5 )))

(4-40d)

ξ4 = 4.00518423 + S(0.00501795 + S(0.00021735 + S(−0.88 × 10−6 )))

(4-40e)

The location of the pattern peak was also ﬁtted to a polynomial: Umax = 0.4797212 + S(0.01456692 + S(−0.00018739 + S(0.218 × 10−5 + S(−0.1 × 10−7 ))))

(4-41)

We obtain the aperture distribution by a Fourier sine series having only n terms: |x| ≤ 0.5 (4-42) E(x) = Bm sin(m + 12 )2πx

160

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

where

n−1 1 − (m + 12 )2 /UN2

(−1)m (m + 12 )2 Bm = 2j

N=1

n−1

(4-43)

1 − (m + 12 )2 /(N + 12 )2

N=0,N =m

The phase constant (−j ) has little effect on the coefﬁcient Bm except to balance the phase ±90◦ about the null. Example Design a Bayliss distribution with 30-dB sidelobes and n = 6. Use Eq. (4-40) to compute the coefﬁcients. A = 1.64126

ξ1 = 2.07086

ξ2 = 2.62754

ξ3 = 3.43144

ξ4 = 4.32758

We substitute these constants into Eq. (4-38) to calculate the ﬁve (n − 1) nulls: No. Null UN

1

2

3

4

5

2.1639

2.7456

3.5857

4.5221

5.4990

Equation (4-41) computes the beam peak of the split-beam pattern in U space: Umax = 0.7988

ka sin θmax = πUmax = 2.5096 2

where a is the aperture width. We substitute these zeros into Eq. (4-43) to determine the coefﬁcients of the Fourier sine series of the aperture distribution (Table 4-15). By evaluating the series across the aperture, the coefﬁcients can be normalized to give a maximum aperture voltage of 1. We use Eq. (4-39), after substituting the zeros, to evaluate the pattern. The 3-dB pattern points can be found by searching the pattern: ka sin θ1 = 1.27232 2

ka sin θ2 = 4.10145 2

Figure 4-5 contains the plot of a Bayliss distribution (n = 6) designed to have sidelobes 30 dB below the Taylor distribution with 30-dB sidelobes. The losses to the difference pattern are about 2 dB higher than the sum pattern. We design the Bayliss TABLE 4-15 Fourier Cosine Series Coefﬁcients for Bayliss Distribution: 30 dB, n = 6 No. 0 1 2 3 4 5

Bm

Bm Normalized

Function

0.13421 0.081025 −0.0044151 0.001447 −0.0003393 −0.000014077

0.85753 0.51769 −0.028209 0.0092453 −0.0021679 −0.00008994

sin sin sin sin sin sin

πx 3πx 5πx 7πx 9πx 11πx

BAYLISS LINE-SOURCE DISTRIBUTION

161

distribution to have 28-dB sidelobes. If designed for 30-dB sidelobes as in the example above, then, relative to the sum Taylor distribution, the sidelobes would be 32 dB down from the sum pattern peak. The last nulls show that the unmodiﬁed zeros of the Taylor distribution occur at ±nπ, whereas the unmodiﬁed zeros of the Bayliss distribution occur at ±(n + 12 )π. By using Eq. (4-6), we calculate amplitude taper efﬁciency of the pattern at the beam peak. When we evaluate the phase error efﬁciency by using Eq. (4-7), the result is zero because of the boresight null. We use Eq. (4-3) to evaluate the phase error efﬁciency at the beam peak:

PEL =

a/2 −a/2

2 E(x)ej k sin θmax x dx

a/2 −a/2

(4-44)

2 |E(x)| dx

Table 4-16 lists results of calculations on Bayliss distributions with n = 10 for various sidelobe levels. Lower sidelobe levels produce higher distribution losses and push the beam peak out. The position of the beam peak is independent of n, since the ﬁrst four zeros are ﬁxed by Eq. (4-40). Like the Taylor distribution, the sidelobe level determines most of the parameters of the Bayliss distribution. Changing n has less effect than it has for the Taylor distribution. The values of parameters for distribution with n = 10 will differ little from those in Table 4-16. Example Compute the beam peak and beam edges for an 8λ-wide aperture excited in a Bayliss distribution with n = 10 and 30-dB sidelobes. 8λ 2π sin θmax = 2.5096 λ 2 2.5096 sin θmax = 8π 1.263 sin θ1 = 8π ◦ θmax = 5.73

TABLE 4-16 Parameters

4.071 8π ◦ ◦ θ1 = 2.88 θ2 = 9.32 sin θ2 =

Characteristics of a Bayliss Line-Source Distribution with n = 10 3-dB Edge

Sidelobe Level (dB)

Beam Peak, ka/2 sin θmax

ka/2 sin θ1

ka/2 sin θ2

ATL (dB)

PEL (dB)

20 25 30 35 40

2.2366 2.3780 2.5096 2.6341 2.7536

1.140 1.204 1.263 1.318 1.369

3.620 3.855 4.071 4.270 4.455

0.50 0.54 0.69 0.85 1.00

1.81 1.90 1.96 2.01 2.04

162

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

4-8 WOODWARD LINE-SOURCE SYNTHESIS [9] In the preceding sections, methods to determine distributions that give the minimum beamwidth for speciﬁed sidelobe levels were discussed. Some applications require shaped beams extending over a range of angles. The Woodward synthesis samples the desired k-space pattern at even intervals to determine the aperture distribution. No integrals are required to compute coefﬁcients. The technique is based on the scanned pattern of a uniform amplitude distribution. Express the pattern in terms of U -space so that when scanned to U0 , it becomes sin π(U − U0 ) π(U − U0 ) with the nulls of the pattern occurring at integer values of U − U0 . U=

a sin θ λ

U0 =

a sin θ0 λ

The visible region extends between +a and −a, centered about U0 . Figure 4-6 shows two patterns, scanned to U0 = 1 and U0 = 2. The peak of the curve scanned at U0 = 2 occurs at one of the nulls of the pattern scanned to U0 = 1. If we allow only integer values of U0 , the pattern scanned to U0 solely determines the pattern at the point U0 in U -space. The two curves (Figure 4-6) in the regions below U = 0 and above U = 3 cancel each other to some extent when the distributions are

0.9 0.8 0.7

Voltage amplitude

0.6 0.5 0.4 0.3 0.2 0.1 0 −0.1 −0.2 −0.3 −10

−8

−6

FIGURE 4-6

−4

−2

0 a sin q l

2

4

6

Scanned uniform distributions: U0 = 1 and U0 = 2.

8

163

WOODWARD LINE-SOURCE SYNTHESIS

added. We form the aperture distribution from a sum of 2a/λ + 1 independent sample points of scanned apertures: N

E(x) =

Ei e−j (i/a)x

(4-45)

i=−N

where N = integer part (a/λ). Each term is a uniform amplitude distribution scanned to an integer value of U . The amplitudes Ei are determined by the sample values of the U -space pattern at those points. Example Design a 10λ aperture with a constant beam between θ = 0◦ and θ = 30◦ . The nonzero portion of the U -space pattern extends from U1 = 10 sin 0◦ = 0 to U2 = 10 sin 30◦ = 5. When we sample the U -space pattern, we discover six nonzero terms: i Ei

0

1

2

3

4

5

0.5

1.0

1.0

1.0

1.0

0.5

At U1 = 0 and U2 = 5, we use the average value. The aperture distribution is 0.5 + e−j x/a + e−j 2x/a + e−j 3x/a + e−j 4x/a + 0.5e−j 5x/a The U -space pattern of this distribution (Figure 4-7) shows some ripple in the beam and the reduction to 6 dB at the beam edges. If we increased the sample level at the edges, U = 0 and U = 5, the pattern would increase to that level.

−10

Level, dB

−20

−30

−40

−50

−60 −10

−8

−6

−4

−2

0 a sin q l

2

4

6

8

10

FIGURE 4-7 U -space pattern of Woodward–Lawson sampling for constant beam from 0 to 30◦ (10λ aperture).

164

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

A cosecant-squared power pattern can be designed by the same method as in the preceding example. When an antenna with this pattern on the ground points its maximum toward the horizon, it delivers a constant signal to an aircraft that maintains a constant altitude. The pattern falloff matches the range decrease as the aircraft ﬂies toward the antenna. The voltage pattern is given by E = E0

sin θmax sin θ

where θmax is the angle of the pattern maximum. In U -space this becomes E(U ) = E0

Um U

The amplitudes of the scanned apertures decrease as 1/U . Example Design a 10λ aperture with a cosecant-squared pattern from θ = 5◦ to θ = 70◦ with the maximum at 5◦ . There are 2a/λ + 1 possible sample points (21). The nonzero portion of the U -space pattern extends from Umin = 10 sin 5◦ = 0.87 to Umax = 10 sin 70◦ = 9.4. We sample only at integer values of U , which gives us nine nonzero terms: Um = 0.8716. The coefﬁcients are given in Table 4-17. The sum [Eq. (4-45)] for this distribution contains nine terms. E(x) =

9

Ei e−j (i/a)x

i=1

Figure 4-8 shows the amplitude and phase of this aperture distribution. The pattern obtained by summing the scanned aperture distributions (Figure 4-9) shows ripple about the desired pattern. Increasing the aperture size increases the number but does not change the level of ripples. The aperture distribution (Figure 4-8) has a negative phase slope to scan the beam off broadside. 4-9 SCHELKUNOFF’S UNIT-CIRCLE METHOD [10] Schelkunoff’s unit-circle method consists of the manipulation of the zeros (nulls) of the array pattern to achieve a desired pattern for a line array. The method is similar to designing networks by specifying the placement of poles and zeros in the complex plane, but the array has only zeros to manipulate. We can use the representation to describe any uniformly spaced array. TABLE 4-17 Woodward Synthesis Coefﬁcients of 10λ Cosecant-Squared Pattern i

Ei

i

Ei

i

Ei

1 2 3

0.8716 0.4358 0.2905

4 5 6

0.2179 0.1743 0.1453

7 8 9

0.1245 0.1089 0.0968

165

SCHELKUNOFF’S UNIT-CIRCLE METHOD

0.9 0.8 0.7 Amplitude (V)

0.6 0.5 0.4 0.3 0.2 0.1 0 −5

−4

−3

−2

1 −1 0 Aperture position (a)

2

3

4

5

−180 −5

−4

−3

−2

−1 0 1 Aperture position (b)

2

3

4

5

180 150 120 90

Phase (deg)

60 30 0 −30 −60 −90 −120 −150

FIGURE 4-8 Aperture distribution of Woodward–Lawson sampling for cosecant-squared pattern (10λ aperture): (a) aperture amplitude distribution; (b) aperture phase distribution.

Consider a uniformly spaced array along the z-axis with the pattern angle θ measured from the axis. The array response will be symmetrical about the z-axis. If we deﬁne the variable ψ = kd cos θ + δ, where δ is a ﬁxed progressive phase shift between elements, d the element spacing, and k the wave number (2π/λ), the pattern of the array is given by E = I0 + I1 ej ψ + I2 ej 2ψ + I3 ej 3ψ · · · (4-46)

166

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

−10

Pattern levels, dB

−20

−30

−40

−50

−60 −10

−8

−6

−4

−2

0 a sin q l

2

4

6

8

10

FIGURE 4-9 U -space pattern of 10λ aperture Woodward–Lawson sampling for a cosecant-squared pattern.

where Ii , a phasor, is the excitation of the ith element in the array. We simplify the notation further by deﬁning W = ej ψ (4-47) We then write Eq. (4-46) as E = I0 + I1 W + I2 W 2 + I3 W 3 + · · · + IN−1 W N−1

(4-48)

where N is the number of elements in the array. We use the ﬁrst element as our phase reference point. This array factor (isotropic elements) is a polynomial with N − 1 roots (zeros) for N elements. We denote the roots as Wi and rewrite Eq. (4-48) as E = E0 (W − W1 )(W − W2 ) · · · (W − WN−1 ) We can ignore the normalization E0 and compute array pattern magnitude as |E(W )| = |W − W1 ||W − W2 | · · · |W − WN−1 | where |W − Wi | is the distance from the root Wi to W in the complex plane. W is limited to the unit circle [Eq. (4-47)] because it always has unit value. Both the spacing of the elements and the progressive phase shift δ determine the limits of the phase of W : ◦

θ =0

◦

θ = 180

ψs = kd + δ

start

ψf = −kd + δ

ﬁnish

(4-49)

SCHELKUNOFF’S UNIT-CIRCLE METHOD

167

q q = 0°

w = ejy

ys y

yf q = 180°

FIGURE 4-10 Unit circle in the W -plane.

As θ increases, ψ decreases and W progresses in a clockwise rotation along the unit circle (Figure 4-10). We have no 2π limitation on either ψs or ψf . The element spacing determines the number of times W cycles the unit circle as θ varies from 0 to 180◦ . If ψs − ψf , 2kd, exceeds 2π, there is a possibility of more than one main beam (grating lobes). The zeros Wi , suppress the pattern when W moves close to one or more of them. The pattern rises to form a lobe when W is far from the zeros. The main-beam peak occurs at the point with the maximum product of the distances from the zeros. Whenever W passes through that point, another main beam forms. A uniformly fed array has the W -space polynomial f (W ) =

1 − WN 1−W

for N elements

The zeros of f (W ) are the N zeros of W N = 1 with the zero at W = 1 removed: Wi = ej 2πi/N . These are spaced uniformly on the unit circle. Figure 4-11 shows the unit circle diagram of a 10-element array fed with uniform phase and amplitude. W starts at −1 since d = λ/2, and it progresses clockwise around the unit circle one revolution to the same point as θ varies from 0 to 180◦ . At θ = 90◦ , the product of the distances from the zeros is a maximum. A lobe forms within the space between each pair of zeros. As W moves from the start to the main beam at W = 1, it starts at a zero and passes through four additional zeros. These zeros Wi correspond to the nulls in the pattern from θ = 0◦ to θ = 90◦ . An equal number of nulls occur as W moves through the range θ = 90 to 180◦ . A uniform-amplitude end-ﬁre array can be represented on the same unit-circle diagram. With antenna elements spaced λ/4, the excursion from start ψs to ﬁnish ψf is only π(2kd). A progressive phase shift δ of −kd through the array forms an end-ﬁre pattern. From Eq. (4-49), ψs = 0◦ and ψf = 180◦ . The end-ﬁre array pattern has only ﬁve nulls, including the null at θ = 180◦ as θ ranges from 0 to 180◦ , since only ﬁve zeros occur in the visible region.

168

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

FIGURE 4-11 Unit-circle representation of a 10-element array with λ/2 spacings.

The Hansen and Woodyard increased-directivity end-ﬁre array corresponds to a shift in the start and stop locations on the unit circle. The excursion from start to ﬁnish remains π determined by element spacing. Equation (4-49) calculates the start: ψs = 90◦ − 108◦ = −18◦ . The pattern has ﬁve nulls from θ = 0 to 180◦ . A binomial array has all its zeros at W = −1 and its pattern has no sidelobes, since they occur for points on the unit circle between zeros. Only one beam forms as W traverses the unit circle. The W -space polynomial is f (W ) = (W + 1)N−1 . For an array of given size we can manipulate the location of the nulls either to reduce sidelobes or to place pattern nulls. We reduce a sidelobe by moving the zeros on both sides of it closer together, but either the main-lobe beamwidth increases or the other sidelobes rise. We form a null in the array pattern by moving one of the zeros to that point on the unit circle corresponding to W at the null angle. Given a desired null θn , Wi = ej (kd cos θn +δ)

(4-50)

Equation (4-50) gives the phase angle kd cos θn + δ of the zero required on the unit circle in W -space. In the case of an end-ﬁre array in which the spacing between elements is less than λ/2, we can shift zeros from invisible space into visible space to narrow the beam and reduce sidelobes. We thereby form large lobes in invisible space that represent energy storage in the array. The large energy storage reduces the bandwidth and efﬁciency of the array. This super-directivity method has limited success, although we can produce beautiful patterns on paper. Example Design a four-element array of broadcast towers to give nearly uniform coverage for θ = ±45◦ with nulls at θ = 270◦ and 135◦ [11, p. 69]. We will align the array with θ = 0◦ to obtain symmetry for the ±45◦ requirement. We actually need only three elements, since only two nulls are speciﬁed. Using λ/4 spacings, we set δ = −90◦ to get an end-ﬁre array. Equation (4-50) gives the zeros of

SCHELKUNOFF’S UNIT-CIRCLE METHOD

169

the polynomial required for the pattern nulls. W1 : W2 :

360◦ λ ◦ ◦ ◦ cos(270 ) − 90 = −90 λ 4 360◦ λ ◦ ◦ ◦ cos(135 ) − 90 = −153.64 ψ= λ 4 ψ=

We determine the polynomial from the roots: ◦

◦

f (W ) = (W − e−j 90 )(W − e−j 153.64 ) ◦

= W 2 + 1.6994W ej 53.18 + ej 116.36

◦

We normalize the phase to the ﬁrst element of the array [constant term of f (W )]: ◦

◦

f (W ) = W 2 e−j 116.36 + 1.6994W e−j 148.18 + 1 At this point the polynomial representation of the array f (W ) does not include the progressive phase factor δ = −90◦ . We add the factor to the polynomial by adding −90◦ to the phase of the second element (W term) and −180◦ to the third element (W 2 term): ◦ ◦ f (W ) = W 2 e−296.36 + 1.6994W e−j 148.18 + 1 The coefﬁcients of the polynomial are the voltage (or current) components of the array. No null develops at θ = 180◦ because the two available nulls (N − 1) were used. Adding the fourth element gives us the freedom to improve the response ﬂatness in the ±45◦ region of θ . Figure 4-12 shows a unit-circle representation and pattern to give a nearly equal ripple response between ±45◦ and the required nulls. We increase

FIGURE 4-12 Four-element linear array with pattern nulls at θ = 90, 135, and 180◦ . The elements are spaced at 0.35λ to give a ﬂat response ±45◦ .

170

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

TABLE 4-18 Four-Element 0.35λ Spaced-Array Coefﬁcients for Uniform Beam ±45◦ No. 1 2 3 4

Amplitude (dB)

Phase (deg)

−9.50 −4.11 −4.11 −9.11

0.0 −103.3 138.4 35.1

the spacing to 0.35λ and place the pattern nulls at 90◦ , 135◦ , and 180◦ . W starts at +1 on the unit circle or ψs = 0◦ and determines δ: ψs = 0 = kd + δ

or

δ = −kd = −

360◦ ◦ 0.35λ = −126 λ

We compute the phase of the zeros from Eq. (4-50): ◦

◦

◦

◦

ψ1 = 360 (0.35) cos(90 ) − 126 = −126 ◦

◦

◦

◦

◦

◦

◦

◦

ψ2 = 360 (0.35) cos(135 ) − 126 = −215.1 ◦

ψ3 = 360 (0.35) cos(180 ) − 126 = −252

(144.9 ) ◦

(108 )

By following the same steps as above, we compute the phase and amplitude of the array elements (Table 4-18). 4-10

DOLPH–CHEBYSHEV LINEAR ARRAY [2]

The Chebyshev polynomials have equal ripples in the region x = ±1, and the amplitude varies between +1 and −1. Outside that region the polynomial value rises exponentially: −1 (−1)m cosh(m cosh |x|) x < −1 −1 Tm (x) = cos(m cos x) −1 ≤ x ≤ 1 −1 x>1 cosh(m cosh x) The order of the polynomial m equals the number of roots. Dolph devised a method of relating the Chebyshev polynomials to the array factor polynomial for a broadside array. We scale the polynomial to make the equal-ripple portion the sidelobes and the exponential increase beyond x = 1 becomes the main beam. Take an array fed symmetrically about the centerline that has either 2N + 1 or 2N elements. We expand the array factor in a polynomial with factors cos(ψ/2), where ψ = kd cos θ + δ. The beam peak occurs when ψ = 0. If we make this correspond to a value x0 , where the Chebyshev polynomial has a value R, the sidelobes will be equal to the ripple at the level 1/R. By substituting x = x0 cos(ψ/2), we use the Chebyshev polynomial for the array polynomial with Tm (x0 ) = R

or

x0 = cosh

cosh−1 R m

(4-51)

DOLPH–CHEBYSHEV LINEAR ARRAY

171

where 20 log R is the desired sidelobe level in decibels. The zeros of Tm (x) are given by (2p − 1)π xp = ± cos (4-52) 2m By using the equation xp = x0 cos(ψ/2) = x0 (ej ψ/2 + e−j ψ/2 ), we calculate the angles of the symmetrical zeros in the W -plane: ψp = ±2 cos−1

xp x0

(4-53)

Both values of xp [Eq. (4-52)] give the same ψp pair. Given the zeros in the W -plane, we multiply out the root form of the polynomial to calculate feeding coefﬁcients of the array. Example Design a 10-element array with 25-dB sidelobes. The array has nine nulls, so we pick m = 9 for the Chebyshev polynomial. [Eq. (4-51)]

R = 1025/20 = 17.7828

x0 = 1.0797

We need only the ﬁrst ﬁve zeros, since they are symmetrical about zero. We calculate them from Eq. (4-52), divide them by x0 , and use Eq. (4-53) for their angles on the unit circle of the W -plane (Table 4-19). We multiply out the root form of the polynomial for the voltage (current) feeding coefﬁcients of the array. Because the roots are symmetrical about the real axis, all phase angles are zero. We obtain the following coefﬁcients: Nos.

1, 10

2, 9

3, 8

4, 7

5, 6

Coefﬁcient (dB)

−8.07

−5.92

−2.84

−0.92

0.0

Figure 4-13 shows the unit-circle representation and pattern of the array with λ/2 spacing. We can estimate the beamwidth of a Chebyshev array by using a beamwidth broadening factor and the beamwidth of a same-length uniformly fed array [12]. The beamwidth broadening factor is given by

2 2 −1 2 2 f = 1 + 0.632 cosh (cosh R) − π R TABLE 4-19 Chebyshev Polynomial Roots and W -Plane Roots for 10-Element 25-dB Sidelobe Array p

Xp

ψp (deg)

1 2 3 4

0.9848 0.6428 0.3420 0.0

±48.41 ±106.93 ±143.06 180.00

(4-54)

172

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

FIGURE 4-13 Ten-element Chebyshev array designed for 25-dB sidelobes.

Equation (4-54) is valid in the range of sidelobe levels from 20 to 60 dB and for scanning near broadside. Example Compute the broadside beamwidth of a Dolph–Chebyshev array with 61 elements, a 30-dB sidelobe level, and λ/2 spacings. Equation (4-54) gives the value 1.144 for f using R = 1030/20 . We estimate the beamwidth of the uniform array from HPBW = 50.76◦ λ/N d = 1.66◦ , where d is the element spacing: ◦

◦

HPBWarray = (f )HPBWuniform = 1.144(1.66 ) = 1.90

We use the beam-broadening factor to estimate the array directivity: D=

2R 2 1 + (R 2 − 1)f λ/N d

(4-55)

Example We calculate the directivity of the 61-element array above from Eq. (4-55). D = 52.0 (17.2 dB). If we take its limit as N d → ∞, Eq. (4-55) becomes 2R 2 . An inﬁnite Dolph– Chebyshev array has a gain 3 dB more than the sidelobe level.

4-11

VILLENEUVE ARRAY SYNTHESIS [13]

Villeneuve devised a method similar to the Taylor distribution that modiﬁes the n − 1 inner zeros of a uniform amplitude array to lower sidelobes. Since the positions of the outer zeros remain ﬁxed, the outer pattern sidelobes decrease as 1/U . The uniform distribution W -plane zeros are located uniformly around the unit circle except for W = 1: 2πp ψp = (4-56) Ne

ZERO SAMPLING OF CONTINUOUS DISTRIBUTIONS

173

The inner zeros correspond to the Chebyshev zeros [Eq. (4-53)] except that we multiply them by a constant factor α dependent on the number of elements, the sidelobe level, and n: nπ (4-57) α= −1 Ne cos (1/x0 ) cos (2n − 1)π/2m The order of the Chebyshev polynomial m = Ne /2. We use Eq. (4-51) to compute x0 . Example Design a 10-element Villeneuve array containing 10 elements for 25-dB sidelobes and n = 4. We determine α = 1.00653 using Eq. (4-57). The inner three W -plane zeros are found by multiplying the Chebyshev zeros by α, which occur in pairs, and the next three zeros are found from the uniform amplitude array using Eq. (4-56): ψp :

±48.73

± 73.82

± 107.63

± 144

180

Figure 4-14 illustrates the W -plane and pattern of the 10-element Villeneuve array: Nos.

1, 10

2, 9

3, 8

4, 7

5, 6

Coefﬁcient (dB)

−8.44

−5.85

−2.91

−0.91

0.0

The sidelobes drop off instead of staying constant: −25.08, −25.19, −25.43, −26.14. 4-12 ZERO SAMPLING OF CONTINUOUS DISTRIBUTIONS [14] We sample continuous distributions, such as the Taylor line source, for large arrays. By using that method, we avoid the numerical difﬁculties of multiplying out long polynomials. When a small array samples an aperture distribution, its pattern fails to follow the pattern of the distribution. We improve the pattern by matching the zeros of the array to the distribution nulls. The array ψ-space pattern repeats at 2π intervals, but the k-space pattern of the aperture has no repeat. We space elements by λ/2 to

FIGURE 4-14 Ten-element Villeneuve array designed for 25-dB sidelobes, n = 4.

174

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

span the total ψ-space nonrepeating region. We then equate an array with λ/2 spacings to an aperture of the same length regardless of the actual spacings between elements. Since the array samples a continuous distribution, the aperture is N d long, where d is the distance between array elements and we consider the array element to be sampling d/2 on both sides of its location. Consider the U -space pattern of a uniform aperture distribution: sin πU/πU . The aperture zeros occur at integer values of U . The corresponding zeros of the uniformly fed array are Wi = ej 2πi/N , where i = 1, 2, . . . , N − 1. The Taylor distribution modiﬁes the location of the zeros of the uniform distribution to Ui , and the sampled zeros of the array must move to follow this pattern: Wi = ej 2πUi /N

(4-58)

Example Given a Taylor line source with 30-dB sidelobes and n = 6, compute the zeros of an array with 12 elements to sample the distribution. The array spans 12/2 in U -space. We calculate zeros of the distribution from Section 4-4 and the angles of the array zeros from Eq. (4-58): Ui

±1.473

±2.1195

±2.9989

±3.9680

±4.9747

6

ψi

±44.19

±63.58

±89.97

±119.04

±149.24

180

We multiply out the root form of the polynomial to compute the array feeding coefﬁcients. The array has 30-dB ﬁrst sidelobes. A straight sampling of the distribution gives an array whose sidelobes exceed 30 dB. Figure 4-15 shows the unit-circle diagram of a zero-sampled Taylor line source with 25-dB sidelobes and n = 5. The method places the zeros on the unit circle close enough together to limit the sidelobe peaks to less than 25 dB when W for a given pattern direction lies between the zeros. The array has higher sidelobes than the equivalent

FIGURE 4-15 Twelve-element array designed by zero-sampling 25-dB Taylor distribution: pattern of normal array (solid curve); pattern with null ﬁlling by moving three zeros off unit circle (dashed curve and triangles).

FOURIER SERIES SHAPED-BEAM ARRAY SYNTHESIS

175

aperture, but closer to the speciﬁed 25 dB, because the ﬁnite array cannot control sidelobes as well as the continuous aperture. Aperture

25.29

25.68

26.39

27.51

29.63

12-Element Array

25.03

25.07

25.18

25.44

26.41

The dashed plot of Figure 4-15 illustrates pattern behavior when W -space zeros are moved off the unit circle. We can ﬁll pattern nulls and generally shape the pattern. When we place all zeros on the unit circle in the complex plane, it can be proved that the array excitations will have amplitude symmetry about the centerline. Moving the zeros off the unit circle disturbs this symmetry. We can eliminate all pattern nulls by moving all the W -plane zeros off the unit circle. If we start with a uniformly fed array and move all the zeros to the same radius, the distribution taper across the array will be linear in decibels. In the next two sections we explore techniques for moving the zeros systematically to produce shaped patterns from an array. 4-13 FOURIER SERIES SHAPED-BEAM ARRAY SYNTHESIS The preceding methods seek the narrowest beamwidths for a given sidelobe level. Arrays can also produce shaped beams. We discussed the Woodward line-source method for shaped beams in Section 4-8. We obtain good approximations by sampling the line-source distribution with an array. Beyond sampling a line source, we can apply Fourier series to design an array directly. An array for a shaped beam must be much larger than is required for the beamwidth. The extra size of the array gives us the degrees of freedom necessary for beam shaping. Increasing the array size increases the match between the speciﬁed and the actual beam shape. Because the array pattern is periodic in k-space, we can expand the pattern in a Fourier series. The array pattern for a symmetrically fed array is given by either f (ψ) = 1 + 2

m In n=1

or f (ψ) = 2

m In n=1

I0

I0

cos

cos

2nψ 2

(2n − 1)ψ 2

N odd

(4-59)

N even

(4-60)

where m = (N − 1)/2 (odd) or m = N/2 (even) with ψ = kd cos θ + δ. Equations (459) and (4-60) are Fourier series expansions of the pattern in ψ-space. The elements farthest from the centerline produce the highest harmonics in the series. In an asymmetrically fed array, we express Eqs. (4-59) and (4-60) as a sum of exponential terms: m an ej nψ N odd n=−m f (ψ) = (4-61) m j (2n−1)ψ/2 −j (2n−1)ψ/2 an e + a−n e N even n=1

176

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

Suppose that we have a desired pattern in k-space given by fd (ψ). We expand it in an inﬁnite Fourier series of the same form as Eq. (4-61) with m = ∞. We equate the ﬁrst m coefﬁcients of the two Fourier series to approximate the desired pattern. As in any Fourier series method, we solve for the coefﬁcients by using the orthogonality of the expansion functions when integrated over a period: π 1 fd (ψ)e−j nψ dψ N odd 2π −π an = (4-62) π 1 −j (2n−1)ψ/2 fd (ψ)e dψ N even 2π −π π 1 fd (ψ)ej (2n−1)ψ/2 dψ N even (4-63) a−n = 2π −π We determine the array coefﬁcients directly from the Fourier series coefﬁcients. Example Design a 21-element array with λ/2 element spacing with a constant beam 2b wide centered in ψ-space. We use Eq. (4-62) to compute coefﬁcients an : b 1 sin nb an = e−j nψ dψ = 2π −b πn Suppose that the constant beam is 45◦ at broadside: 67.5◦ ≤ θ ≤ 112.5◦ . Then b=

360◦ λ ◦ ◦ cos 67.5 = 68.88 λ 2

We can ignore the constant factor i/π and expand to compute the array coefﬁcients (Table 4-20). The method fails to some extent when we try it on arrays with spacings greater than λ/2. The integral does not cover the total visible region. We can, however, use it with TABLE 4-20 Fourier Series Synthesis Coefﬁcients of 21-Element Array for Pattern of Figure 4-16 n 0 ±1 ±2 ±3 ±4 ±5 ±6 ±7 ±8 ±9 ±10

an

Amplitude (dB)

Phase (deg)

1.0000 0.9328 0.3361 −0.1495 −0.2488 −0.0537 0.1336 0.1209 −0.0240 −0.1094 −0.0518

0.00 −0.60 −9.47 −16.50 −12.08 −25.40 −17.48 −18.35 −32.40 −19.22 −25.72

0 0 0 180 180 180 0 0 180 180 180

FOURIER SERIES SHAPED-BEAM ARRAY SYNTHESIS

177

spacings less than λ/2 with good results. As we increase the number of elements in the array, the match to the desired pattern improves. Of course, tapering the desired pattern reduces the higher harmonics and the subsequent need for more elements. Example Suppose that we want to scan the beam of the 21-element array with λ/2 element spacing to 60◦ with a 45◦ beamwidth. The beam edges are 37.5◦ and 82.5◦ . We could calculate coefﬁcients by integrating Eq. (4-62) directly with this requirement, but we can use δ, the progressive phase shift between elements, to simplify the problem. The beam edges in ψ-space are 180◦ cos(37.5◦ ) + δ 142.8◦ + δ

and 180◦ cos(82.5◦ ) + δ 23.49◦ + δ

We pick δ to center the beam in ψ-space: b = 142.8◦ + δ, −b = 23.49◦ + δ. On solving, we have δ = −83.15◦ and b = 59.65◦ . We use the formula sin(nb)/πn to compute coefﬁcients of the array and then add the progressive phase shift through the array. Figure 4-16 shows the array pattern. When we scan the beam to end ﬁre, we must account for the symmetry about θ = 0◦ . Because we limit the spacings to less than λ/2 to prevent grating lobes, we have an unspeciﬁed region of ψ-space that we can choose in any convenient manner. Example

Design a 21-element end-ﬁre array with a 90◦ beamwidth and 0.30λ spacings.

FIGURE 4-16 Twenty-one-element array designed by the Fourier series method to scan to 60◦ with a 45◦ beamwidth.

178

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

For an end-ﬁre array we pick δ = −kd = −108◦ . This places the edge of the visible region on the ψ-space origin. We are free to specify the invisible region that will be included in the integral [Eq. (4-62)]. We specify the invisible region as the mirror image of the portion in the visible region and solve for b: ◦

◦

◦

◦

−b = 360 (0.3) cos(45 ) − 108 = −31.63

We use the sin(nb)/πn formula to calculate array coefﬁcients and then apply the progressive phase shift δ to the coefﬁcients obtained to get the proper phase to scan to end ﬁre. We cannot control the sidelobes of an array designed using Fourier series expansion. The initial speciﬁcation calls for no sidelobes. Sampling a Woodward linear aperture with an array also fails to give control of the sidelobes. The Woodward linear distribution cannot control sidelobes; it provides only ease of design. In the next section we explore a method with direct control of sidelobes of an array.

4-14

ORCHARD METHOD OF ARRAY SYNTHESIS [15]

In Section 4-6 we manipulate the nulls of a continuous linear distribution to control the sidelobes of the radiated pattern individually. In Section 4-9 we show that the nulls of the linear aperture pattern can be related directly to the roots of Schelkunoff polynomial representation of the linear array pattern in W -space. The unit circle method gives us a tool for array synthesis expanded in the Orchard method for the design of arrays with arbitrary patterns. We apply an iterative technique on the W -space zeros to produce the pattern desired. We control all the sidelobes individually and produce shaped patterns for the main beam. The ﬁnite size of the array limits the control of the main beam shape as we saw in the Fourier series expansion method. Each array element corresponds to a term in the Fourier series expansion. We start with the Schelkunoff transformation of the array pattern: f (W ) = C0

N

(W − Wn )

(4-64)

n=1

A normalization constant C0 has been added. We write Wn = exp(an + j bn ). Expansion of Eq. (4-64) produces the feeding coefﬁcients of an array with N + 1 elements: W = ej ψ

with ψ = kd cos θ + δ

θ is measured from the array axis. The effect of δ on the unit circle method is to rotate the starting and ﬁnishing points when varying W to calculate the pattern using Eq. (4-64). An equally valid method is to rotate the zeros about the origin of the complex plane, which leaves the ψ-space pattern shape unchanged. When designing a shaped beam, we need to rotate the main beam peak to the proper location to calculate the amplitudes because our speciﬁcation will be in terms of the pattern angle θ relative to the peak.

ORCHARD METHOD OF ARRAY SYNTHESIS

179

Figure 4-11 illustrates that the pattern amplitude is the product of distances from each zero to the pattern W point. Expansion of Eq. (4-64) in terms of the product of distances to W gives |f (W )|2 = C02

N

[1 − 2ean cos(ψ − bn ) + e2an ]

(4-65)

n=1

The Orchard method requires the speciﬁcation of each sidelobe and additional values located at the minimum ripple points in the shaped region. For a single-beam unshaped pattern, we only specify sidelobes, and all an will be zero since all zeros Wn will be on the unit circle. We restrict the array to λ/2 spaced elements when applying the method so that the entire unit circle is used in the pattern. An array with N zeros has N pattern peaks which lie between the zeros in the W -plane. When we include the normalization constant C0 to specify the main beam peak and all the zeros, we have N + 1 unknowns to ﬁnd. Without loss of generality we specify the last zero as WN = −1 or ψ = π to reduce the number of unknowns to N . Since we rotate the zeros after we determine the proper zero spacing for speciﬁed sidelobes, we place the main beam between WN−1 and WN = −1. Before starting the iteration technique, we generate a list of sidelobe levels with the main beam as the last one. The method expands the pattern in a multiple-variable Taylor series using bn , an , and the normalization constant as variables. To facilitate calculating the partial derivatives, we express Eq. (4-65) in decibels: N−1

10 ln[1 − 2ean cos(ψ − bn ) + e2an ] + 10 log10 [2(1 + cos ψ)] + C ln(10) n=1 (4-66) The second term of Eq. (4-66) is due to the zero WN = −1 and C is the normalization constant of the main beam. The logarithm to the base 10 has been expressed as a natural logarithm for the calculation of derivatives: G=

∂G Mean [ean − cos(ψ − bn )] = ∂an 1 − 2ean cos(ψ − bn ) + e2an

(4-67)

∂G Mean sin(ψ − bn ) =− ∂bn 1 − 2ean cos(ψ − bn ) + e2an

(4-68)

∂G =1 ∂C

(4-69)

The variable M = 20/ln(10). The multiple-variable Taylor series involves three types of terms: G(bn , an , C) = G0 (bn0 , an0 , C0 ) +

N−1 n=1

+

N−1 n=1

∂G (bn − bn0 ) ∂bn

∂G (an − an0 ) + (C − C0 ) ∂an

(4-70)

180

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

Every nonzero value of an ﬁlls in the pattern null at ψ = bn . If we specify the desired pattern amplitude at every sidelobe peak, the main beam, and at points between the sidelobe peaks equal to the number of nonzero an , we form a square matrix equation. The solution gives the changes in bn , an , and C. Since we expanded Eq. (4-66) as a linear approximation, the solution of Eq. (4-70) gives only an approximate solution. In a few iterations the method converges and we obtain an acceptable pattern. Suppose that the shaped pattern is limited to a range in W -space so that there are only L nonzero an . Given the desired pattern Sm (ψm ) at ψm and the current pattern G0 (ψm ), one row of the matrix is

∂G(ψm ) ∂G(ψm ) ∂G(ψm ) ∂G(ψm ) ,..., , ,..., ,1 ∂b1 ∂bN−1 ∂a1 ∂aL

We need N + L rows or pattern points to solve Eq. (4-70) for changes in bn , an , and C: [δb1 , . . . , δbN , δa1 , . . . , δaL , δC]T We require a search routine to locate the pattern peaks between the pattern nulls or minima between peaks in the shaped region for given values of bn and an after we normalize to the current pattern peak. We subtract these from the levels desired: [S(ψ1 ) − G0 (ψ1 ), . . . , S(ψN+L ) − G0 (ψN+L )]T After solving the square matrix equation, we update the W -plane zeros: b1 = b1 + δb1 ... bN−1 = bN−1 + δbN−1 .. . a1 = a1 + δa1 .. . aL = aL + δaL C = C0 + δC The iteration alters the beam peak and its location. The pattern peak is normalized after iteration, and for a shaped pattern a new zero rotation is found to line up the beam peak for the pattern-shaping function. Example Design an eight-element array with its beam peak at 90◦ and speciﬁed sidelobes before the peak of 25, 30, and 25 dB and 20, 25, and 30 dB after the peak. The sidelobes values begin with the ﬁrst sidelobe after the peak and rotate to the peak: −20 − 25 − 30 − 25 − 30 − 25 0

181

ORCHARD METHOD OF ARRAY SYNTHESIS

FIGURE 4-17 Eight-element array designed using Orchard synthesis for individually speciﬁed sidelobes, λ/2 spacings.

The solution converges in four iterations after starting with uniformly spaced zeros on the unit circle. Figure 4-17 shows the unit-circle zeros on the left and the corresponding pattern on the right with λ/2 element spacing: W -Space Zero (deg)

178.14

142.72

99.26

58.01 −62.18

−96.89 −130.37

Pattern Null (deg)

8.25

37.54

56.53

71.20

122.57

110.21

136.41

The feeding coefﬁcients for the ﬁnal design are given in Table 4-21. Although the Orchard method requires the elements to be spaced λ/2 during synthesis, the completed design can be used at another element spacing. Figure 4-18 gives the unit-circle diagram of the same array with a 0.7λ element spacing. The range of W now exceeds 2π and the sidelobe regions of the unit circle have been used more than once. Sidelobes 3 and 4 occur twice in the pattern. Of course, if we scan the array too far, the pattern would have grating lobes. Figure 4-19 plots the pattern of an end-ﬁre array with λ/4-element spacing using the same zeros. Only a portion of the unit circle is used, and not all sidelobes are realized. Figure 4-20 illustrates the end-ﬁre case with the elements spaced so that the ﬁnal position of W occurs at a null. The pattern contains all six sidelobes. The unit-circle analysis mirrors that of TABLE 4-21 Synthesis

Coefﬁcients of Eight-Element Array of Figure 4-17 Designed by Orchard Magnitude (dB)

Phase (deg)

Element

Element

Magnitude (dB)

Phase (deg)

1 2 3 4

−8.69 −3.90 −1.06 0

8.70 3.22 1.29 4.91

5 6 7 8

0 −1.06 −3.90 −8.69

3.79 7.41 5.48 0

182

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

FIGURE 4-18 Eight-element array designed using Orchard synthesis for individually speciﬁed sidelobes, 0.7λ spacings.

FIGURE 4-19 Eight-element array designed using Orchard synthesis for individually speciﬁed sidelobes, λ/4 spacings scanned to end ﬁre.

FIGURE 4-20 Eight-element array designed using Orchard synthesis for individually speciﬁed sidelobes, 0.42λ spacings scanned to end ﬁre.

183

ORCHARD METHOD OF ARRAY SYNTHESIS

the circle diagram in Chapter 3, where increasing the element spacing increases the visible region. In this case the visible region corresponds to rotation about the unit circle. Expansion of Eq. (4-64) produces the array feeding coefﬁcients independent of element spacing, and the progressive phase shift between elements δ affects phase but not amplitude. The four examples given in Figures 4-17 to 4-20 have the same sequence of feed magnitudes. We can use Orchard synthesis to generate a difference pattern similar to the Bayliss line distribution and control all the sidelobes. A difference pattern has two main beams. Using the same example of an eight-element array, we modify the sidelobe list to include side-by-side main beams. We eliminate the −25-dB lobe next to the original main beam from the values above: −20

− 25

− 30

− 25

− 30

0

0

When we apply the synthesis by placing the last main beam at 90◦ , we obtain a pattern with two main beams with the null between them at 101.6◦ , corresponding to a W plane null at −36.3◦ . We rotate all W -plane zeros by 36.3◦ to place the null between the two main beams at 90◦ . Figure 4-21 shows the W -plane and polar pattern for the ﬁnal design. Note the placement of the W -plane zero at W = +1. Table 4-22 lists the feeding coefﬁcients.

FIGURE 4-21 Difference pattern array using eight-elements designed by Orchard synthesis.

TABLE 4-22 Coefﬁcients of Eight-Element Difference Pattern Array of Figure 4-21 Designed by Orchard Synthesis Element

Magnitude (dB)

Phase (deg)

Element

Magnitude (dB)

Phase (deg)

1 2 3 4

−6.32 −0.35 0.0 −6.91

5.39 1.28 0.8 7.05

5 6 7 8

−6.91 0.0 −0.35 −6.32

178.35 184.6 184.12 180

184

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

We ﬁll in the null between the different lobes to form a ﬂat-topped beam for the eight-element array and use a constant-amplitude shaping function for the pattern desired. The beamwidth of the ﬂat lobe is determined by the lobe spacing, and only certain sizes are possible. Remember that an array is a Fourier series approximation to the pattern desired. With only eight elements the match is poor between the pattern desired and the approximate pattern. We use one nonzero an to move the W -plane zero off the unit circle that forms the pattern null between the two beams and add another pattern speciﬁcation: −20

− 25

− 30

− 25

− 30

0

0

−1

The last number gives the pattern level at the null relative to the shaped pattern level. This last term uses Eq. (4-67) for its columns. The constant beam design uses a 22◦ wide beam centered at 90◦ for the pattern shape function. We start with an = 0.01 before iterating. The iteration using the matrix equation computes a1 = 0.4435, which can be either positive or negative without changing the pattern. Rotation of the W -plane zeros placed the zero for minimum ripple along the positive real axis and produced a symmetrical pattern about θ = 90◦ . Figure 4-22 contains the ﬁnal design W -space zeros and polar pattern. The iterations produced the sidelobe levels speciﬁed (Table 4-23).

FIGURE 4-22 Flat-topped beam eight-element array designed by Orchard synthesis. TABLE 4-23 W -Plane Zeros of Eight-Element Flat-Topped Beam of Figure 4-22 Designed by Orchard Synthesis W -Space Zero (deg)

W -Space Radius

Pattern Null (deg)

165.51 123.99 84.33 0 −91.65 −115.39 −161.18

1.0 1.0 1.0 0.6418 1.0 1.0 1.0

23.15 46.46 62.06 90 120.61 129.87 153.57

ORCHARD METHOD OF ARRAY SYNTHESIS

185

TABLE 4-24 Coefﬁcients of Eight-Element Array for Flat-Topped Beam of Figure 4-22 Designed by Orchard Synthesis Element

Magnitude (dB)

Phase (deg)

Element

Magnitude (dB)

Phase (deg)

1 2 3 4

−12.95 −10.78 −24.69 −7.90

−174.39 178.85 167.92 −1.47

5 6 7 8

−1.15 0.0 −2.26 −9.10

1.54 4.37 3.95 0.0

The radius of the fourth term could be 1/0.6418 = 1.5581 without affecting the pattern result. Inserting the zeros into Eq. (4-64) and expanding the polynomial produces the feeding coefﬁcients (Table 4-24). The Fourier series example for a constant beam centered at 60◦ with a 45◦ beamwidth using 21 elements spaced λ/2 (Figure 4-16) was repeated using Orchard synthesis. Fourier series synthesis could not control the sidelobes. First, we need to ﬁgure out how many array lobes cover the shaped pattern region. Place the zeros uniformly around the unit circle in the W -plane and determine how many of the roots are within the beam. For a 21-element array six beams and ﬁve zeros lie in the ψ = π cos θ angular region of the constant beam found using Eq. (4-71): beams =

N (cos θmin − cos θmax ) 2

(4-71)

The solution to Eq. (4-71) is an integer given N as the number of W -plane zeros. All sidelobes were set at −30 dB and the ripple at −0.9 dB below the constant beam: Lobes

1–14

15–20

21–25

Sidelobe (dB)

−30

0.0

−0.9

Figure 4-23 gives the ﬁnal result of the synthesis, an improvement over Figure 4-16, with its uncontrolled sidelobes.

FIGURE 4-23 Twenty-one-element array designed by Orchard synthesis to scan to 60◦ with a 45◦ beamwidth.

186

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

We must consider the element excitations. The zeros not lying on the unit circle can be either inside or outside the circle and produce the same pattern. Different combinations of zero locations lead to different element amplitudes. The last example has ﬁve zeros displaced from the unit circle, which produces 25 = 32 combinations. We need to check the amplitude distribution that results from each case (Table 4-25). Arrays with a large range of amplitudes are difﬁcult to produce. In some cases the range of amplitude available is limited, such as waveguide slot arrays. Mutual coupling between elements also makes it difﬁcult to achieve the desired low amplitudes on some elements because nearby elements will excite them, and compensation for mutual coupling may prove difﬁcult. Figure 4-23 shows one of the combinations of root placements that produced the minimum amplitude variation in the array. The Fourier series synthesis gave an amplitude variation of 32.4 dB, whereas the Orchard synthesis variation is 13.29 dB. This synthesis produced better patterns with less amplitude variation. Decreasing the ripple depth increases the amplitude variation of the array. Csc2 θ cos θ Pattern This pattern produces constant round-trip signals versus the elevation angle for radar. The pattern from the array axis is given by csc2 (θ − 90◦ ) cos(θ − 90◦ ). The peak occurs beyond 90◦ and decreases for greater angles. The shaped pattern function requires the rotation of the W -plane zeros at each step so that the pattern peak calculated from the zeros occurs at the proper angle. The changing zero locations move the beam peak location at each iteration. Example Design a 16-element csc2 (θ − 90◦ ) cos(θ − 90◦ ) beam array to operate from 100 to 140◦ and have 30-dB sidelobes. Equation (4-71) determines that ﬁve beams cover the pattern region and sets the number of nonzero an as 4. The 16-element array has 15 zeros, with the ﬁrst 10 speciﬁed as −30 dB, ﬁve for the shaped-beam region, and four for the minima between the shaped-beam peaks. We specify the shaped-beam lobes relative to the shape levels. The last lobe is the beam peak.

TABLE 4-25 Coefﬁcients of 21-Element Array for Flat-Topped Beam of Figure 4-23 Designed by Orchard Synthesis Element 1 2 3 4 5 6 7 8 9 10 11

Amplitude (dB) −11.85 −6.89 −6.23 −11.34 −5.12 0.0 −0.01 −6.33 −9.73 −2.09 −1.75

Phase (deg) −65.47 −146.85 126.55 4.64 −158.19 108.08 23.97 −65.70 86.93 −4.11 −70.12

Element 12 13 14 15 16 17 18 19 20 21

Amplitude (dB) −3.48 −3.01 −2.97 −4.93 −7.51 −9.64 −9.44 −8.06 −9.26 −13.29

Phase (deg) −115.31 −158.73 134.01 49.59 −40.53 −111.80 −159.41 142.99 73.66 0.0

187

ORCHARD METHOD OF ARRAY SYNTHESIS

Lobe

11

12

13

14

15

16

17

18

19

Amplitude (dB)

1.0

0.8

0.6

0.4

0.2 −0.9 −0.7 −0.5 −0.2

Allowing the ripple to increase in the lower levels of the shaped pattern region decreases the range of element amplitudes. The method converged in 11 iterations to the design given in Figure 4-24. All 24 = 16 combinations of an placements inside and outside the unit circle were checked (Table 4-26). The amplitude variation ranged from 11.47 to 25.47 dB. Figure 4-25 illustrates the design repeated with eight elements (Table 4-27). Although the sidelobes could be controlled at −30 dB, the shaped pattern region shows less pattern control than with 16 elements. Extensions to the Orchard method make various improvements. By adding balancing zeros inside and outside the unit circle in the W -plane, the feeding coefﬁcients of the array can be made real with only 0 or 180◦ phases [16]. This adds elements to the array and changes the shape of the beam somewhat. The coefﬁcients are real only if the pattern is symmetrical about θ = 90◦ . To implement the method you add a term

FIGURE 4-24 Sixteen-element array with csc 2 θ cos θ pattern designed by Orchard synthesis. TABLE 4-26 Coefﬁcients of 16-Element Array for csc2 θ cos θ Beam of Figure 4-24 Designed by Orchard Synthesis Element

Magnitude (dB)

Phase (deg)

Element

Magnitude (dB)

Phase (deg)

1 2 3 4 5 6 7 8

−11.47 −9.84 −8.07 −4.79 −2.65 −2.04 −0.82 0.0

−149.27 −100.16 −69.72 −40.25 −0.56 34.20 65.82 107.07

9 10 11 12 13 14 15 16

−1.19 −2.71 −3.53 −7.82 −9.46 −4.30 −4.55 −8.72

149.09 −177.54 −131.64 −52.50 100.34 −157.29 −76.99 0.0

188

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

FIGURE 4-25

Eight-element array with csc 2 θ cos θ pattern designed by Orchard synthesis.

TABLE 4-27 Coefﬁcients of Eight-Element Array for csc2 θ cos θ Beam of Figure 4-24 Designed by Orchard Synthesis Element

Magnitude (dB)

Phase (deg)

Element

Magnitude (dB)

Phase (deg)

1 2 3 4

−11.62 −8.61 −12.62 −2.61

130.02 −170.78 179.21 −173.92

5 6 7 8

0.0 −2.05 −5.73 −11.54

−122.98 −79.05 −44.44 0.00

to Eqs. (4-67) and (4-68) for the extra elements located off the unit circle. A design of a ﬂat-topped beam centered at 90◦ using the balanced zeros produced a design with more than 30 dB of variation between the elements similar to a Fourier series expansion that had about the same range of amplitudes. The range of amplitudes in the array can be reduced by placing all the zeros off the unit circle [17, p. 124]. We give up the nulls between the lobes and must now search a large set of possible solutions to select a design with the least amplitude variation. A genetic algorithm sorts through the large set of zero combinations inside/outside all that satisfy the pattern requirements to discover the best design.

4-15

SERIES-FED ARRAY AND TRAVELING-WAVE FEED SYNTHESIS

A series-fed array uses couplers along a line that distribute power to the elements from a single transmission line. A single wave travels along the line with each element removing a portion of the power. A matched load absorbs the remaining power at the end to prevent the reﬂection of a wave traveling toward the source end. A second backward traveling wave would produce another beam with reduced amplitude indistinguishable from a sidelobe. The coupling could be a physical coupler or it could be just a series or shunt load across the transmission line. Waveguide slots are an example of loads on a transmission line. An array using couplers can have phase shifters between

SERIES-FED ARRAY AND TRAVELING-WAVE FEED SYNTHESIS

189

the couplers and the elements to form a phased array. A second conﬁguration for a phased array places phase shifters in the transmission line between the couplers. This case uses the simple control of identical phase shifters set to the same value to scan the beam. The phase shifters are the progressive phase δ along the array used for scanning. The array distribution is given by the sequence of radiated powers, Pi . The traveling wave or nonresonant array dissipates a ratio of the input power R in the load: N

Pi = Pin (1 − R)

i=1

N We normalize the distribution to the sum of radiated power: P0 = Pi . The elei=1 ment power becomes Pi (1 − R)/P0 and we use the normalized power distribution to calculate coupling values: C1 = P1

remaining power = 1 − P1

The coupling to the second element removes power from the remaining power: C2 =

P2 1 − P1

remaining power = 1 − P1 − P2

The general expression is Ci =

1−

Pi i−1 j =1

Pj

(4-72)

If the element electrical model consists of a shunt conductance on a transmission line, such as waveguide slots, the power radiated by each slot = |Vinc |2 gi and the normalized gi = Ci . Similarly, an electrical model of an element as a series resistance on a transmission line can be solved in a similar manner. Power radiated = |Iinc |2 ri and the normalized ri = Ci . Some array feeders have signiﬁcant losses between the elements and we must account for these losses when designing the couplers. Suppose that the feeder has identical losses Lf = 1 − 10−attenuation/10 between couplers. The power balance equation becomes N N (j − 1)Pj Lf Pi Pin RPin (N − 1)Lf RPin j =2 i=1 = + + + total load losses to load antennas losses to antennas N N Lf (j − 1)Pj + Pi j =2 i=1 Pin = 1 − R − (N − 1)Lf R As before, we must normalize the power at each element to the input power Pi /Pin . The coupling to the ﬁrst element is C1 = P1 and the power left is 1 − P1 . The transmission medium attenuates the signal between the ﬁrst and second elements and we compute the power at the second element = (1 − P1 )(1 − Lf ). We determine the coupling value from the ratio P2 C2 = (1 − P1 )(1 − Lf )

190

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

and the element removes P2 power. The remaining power travels to the next element but is attenuated by (1 − Lf ). The power removed, P2 , is subtracted from the power at that point and the remaining power is attenuated before reaching the next extraction: P3 [(1 − P1 )(1 − Lf ) − P2 ](1 − Lf ) P4 C4 = {[(1 − P1 )(1 − Lf ) − P2 ](1 − Lf ) − P3 }(1 − Lf ) C3 =

etc.

The total loss due to attenuation is found from the sum of the normalized powers: loss(dB) = 10 log

N

Pi

i=1

Continuous Traveling Wave As the wave propagates along the antenna, it loses power continuously. The slots or holes must radiate more and more of the remaining power if the distribution is to be uniform. In general, the holes or slots must load the waveguide increasingly as the wave travels to the termination. The power at any point in the guide is z P (z) = P0 exp −2 α(z) dz (4-73) 0

where P0 is the power at z = 0 and α(z) is the attenuation distribution (nepers/length). Suppose that we have a desired amplitude distribution, A(z) (voltage):

L

Pin =

L

|A(z)|2 dz +

0

ρL (z) dz + Pload

(4-74)

0

where Pload is the power lost in the termination, |A(z)|2 the radiated power distribution, and ρL (z) the ohmic loss in the walls. Let the power into the termination be a ratio of the input power Pload = RPin ; then Pin =

1 1−R

L

|A(z)|2 + ρL (z) dz

(4-75)

0

The power anywhere along the leaky wave antenna is

L

P (z) = Pin −

|A(z)|2 + ρL (z) dz

(4-76)

0

We differentiate this to get dP (z) = −[|A(z)|2 + ρL (z)] dz

(4-77)

We differentiate Eq. (4-73) to relate α(z) to P (z): 1 dP (z) = −2α(z) P (z) dz

(4-78)

CIRCULAR APERTURES

191

We substitute Eq. (4-75) into Eq. (4-76) for Pin . By combining Eqs. (4-76) and (4-77) into Eq. (4-78), we derive the required attenuation distribution [18, p. 153]: α(z) =

L

[1/(1 − R)]

1 |A(z)|2 2

z

|A(z)|2 + ρL (z)dz −

0

(4-79) |A(z)|2 + ρL (z)dz

0

If we assume a lossless transmission line, ρL (z) = 0 and Eq. (4-79) simpliﬁes. Example Design the attenuation distribution for a uniform distribution along a lossless transmission-line leaky wave antenna. Substitute A(z) = 1 and ρL (z) = 0 into Eq. (4-79) and perform the integrations: α(z) =

1 2

[L/(1 − R)] − z

=

1 (1 2

− R) L[1 − z(1 − R)/L]

Given R = 0.05 (5% of the power into the load) for a structure with length 10λ. The initial and ﬁnal attenuation constants are 0.95 = 0.0475 Np/λ or 0.413 dB/λ 20 0.95 αf (L) = = 0.95 Np/λ or 8.25 dB/λ 2LR αi (0) =

We reduce the variation between the initial and ﬁnal values by dissipating more power in the termination. Given R = 0.1, αi (0) = 0.045 Np/λ or αf (L) = 0.45 Np/λ or

0.39 dB/λ 3.9 dB/λ

If we take the ratio of the attenuations at the ends, we have α(L)/α(0) = 1/R. We can normalize Eq. (4-79) to the interval ±2 and use the linear distributions given above where x = z/L and ρL (z) = 0. Figure 4-26 shows the attenuation distribution for a Taylor distribution with 30-dB sidelobes and n = 8 for various levels of power dissipation in the load. Table 4-28 lists the bounds on α(x)L for various Taylor distributions. Changing the number of modiﬁed zeros has only a minor effect on the bounds. A cos2 on a pedestal distribution with a 30-dB sidelobe level has very similar bounds on the attenuation. The 40-dB sidelobe level design requires a greater variation of attenuation than the 30-dB cases. Long structures may not be able to provide the low levels of radiation above the ohmic losses for an effective design. In all cases we decrease the attenuation range on an antenna by decreasing the antenna efﬁciency though absorbing more power in the termination. 4-16 CIRCULAR APERTURES Many common apertures conform to circles. The two-dimensional Fourier transform relation for the pattern holds for any aperture rim shape and becomes for the circle 2π a E(r , φ )ej kr sin θ cos(φ−φ ) r dr dφ (4-80) f (θ, φ) = 0

0

192

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

Input power dissipated: 5% 25

10%

Attenuation × length, dB

20

15% 15

20% 25%

10

5

0

0.2

0.4

0.6

0.8

1

Normalized length

FIGURE 4-26 Leaky wave attenuation distribution for Taylor distribution with 30-dB sidelobes, n = 8. TABLE 4-28 Maximum and Minimum Normalized Attenuation α(z )L of a Leaky Wave Taylor Distribution 30 dB n=6

n = 12

40 dB, n = 8

Termination Power (%)

Maximum

Minimum

Maximum

Minimum

Maximum

Minimum

5 6 8 10 12 15 20 25

27.08 25.38 22.70 20.66 19.00 17.00 14.42 12.42

0.59 0.58 0.57 0.56 0.55 0.53 0.50 0.46

26.04 24.54 22.06 20.08 18.46 16.52 14.04 12.09

0.63 0.63 0.61 0.60 0.59 0.57 0.53 0.50

31.61 29.63 26.52 24.12 22.18 19.81 16.78 14.44

0.12 0.12 0.12 0.11 0.11 0.11 0.10 0.10

CIRCULAR APERTURES

193

where a is the radius, r the radial coordinate, and φ the angle coordinate of the aperture point. The integral leads to a kr -space. When the distribution has circular symmetry, the φ integral can be evaluated easily, which reduces Eq. (4-80) to

a

f (kr ) = 2π

E(r )J0 (kr sin θ )r dr

(4-81)

0

where J0 (x) is the zeroth-order Bessel function of the ﬁrst kind. All great-circle patterns (constant φ) are identical. For a uniform distribution, f (kr ) =

2J1 (ka sin θ ) ka sin θ

plotted in Figure 4-27. The zeros occur at the zeros of J1 (x). The 3-dB pattern point of the uniform distribution is ka sin θ1 = 1.6162 −1

HPBW = 2 sin

sin θ1 = 0.5145

λ D

(4-82)

0.5145λ D

−10

Pattern level, dB

−20

−30

−40

−50

−60 −5 −25 −100

−4 −20 −80

−3 −15 −40

−2 −10 −20

−1 −5 −10

0

1 5 20

2 10 40

3 15 60

4 20 80

ka (sin q − sin0)

FIGURE 4-27 kr -space pattern of uniform circular aperture distribution.

5 25 100

194

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

where D is the diameter. For large apertures we can approximate sin θ by θ (rad). Converted to degrees, the half-power beamwidth becomes ◦

HPBW = 58.95

λ D

(4-83)

Example Compute the beamwidth of a uniform distribution circular aperture with 10.5λ diameter. The beamwidths are found from Eqs. (4-82) and (4-83): HPBW = 2 sin−1 HPBW =

0.5145 ◦ = 5.62 10.5

58.95◦ ◦ = 5.61 10.5

The ﬁrst zero of J1 (x) gives the k-space pattern null point. ka sin θnull = 3.8317 BWnull = 2 sin−1

1.2197λ ◦ λ 139.76 D D

(4-84)

We can also deﬁne a null beamwidth factor and relate the beams of other distributions to the uniform circular distribution beamwidth [Eq. (4-84)]. All other circular distributions relate to Eq. (4-82) or (4-83) through a beamwidth factor. The uniform distribution has a unity beamwidth factor.

4-17

CIRCULAR GAUSSIAN DISTRIBUTION [19]

A truncated Gaussian distribution has a simple functional relation: E(r) = e−ρr

2

|r| ≤ 1

(4-85)

We can easily calculate the edge taper through the conversion between logarithms: edge taper(dB) = 8.686ρ

(4-86)

We determine amplitude taper efﬁciency by substituting Eq. (4-85) into Eq. (4-8) and carrying out the integrations: ATL =

2(1 − e−ρ )2 ρ(1 − e−2ρ )

(4-87)

Table 4-29 lists designs for various sidelobe levels in terms of the single parameter: edge taper. Equation (4-86) relates the parameter ρ to the edge taper. Example Estimate the beamwidth of the pattern radiated from a circular distribution with a 13-dB edge taper and radius of three wavelengths.

HANSEN SINGLE-PARAMETER CIRCULAR DISTRIBUTION

TABLE 4-29

195

Circular-Aperture Gaussian Distribution, e −ρr (|r| < 1) 2

Sidelobe Level (dB)

Edge Taper (dB)

ATL (dB)

Beamwidth Factor

20 22 24 25 26 28 30 32 34 35 36 38 40

4.30 7.18 9.60 10.67 11.67 13.42 14.93 16.23 17.32 17.81 18.75 21.43 24.42

0.09 0.24 0.41 0.50 0.59 0.76 0.92 1.06 1.18 1.23 1.34 1.65 2.00

1.0466 1.0800 1.1109 1.1147 1.1385 1.1626 1.1839 1.2028 1.2188 1.2263 1.2405 1.2820 1.3296

We use linear interpolation in Table 4-29 to determine the beamwidth factor. From Eq. (4-82), 1.1568(0.5145) ◦ = 11.38 HPBW = 2 sin−1 6 From Eq. (4-83), ◦

HPBW = 58.95

1.1568 6

◦

= 11.36

The amplitude taper efﬁciency is calculated from Eq. (4-87): 13 = 1.497 8.686 2(1 − e−1.497 )2 = 0.847 ATL = 1.497(1 − e−2.993 ) ρ=

(−0.72 dB)

We obtain the same value by interpolating in Table 4-29. Sidelobes below 40 dB are difﬁcult to obtain with this distribution. The inner sidelobes continue to decrease with a decreasing edge level, but the outer lobes fail to reduce and dominate over the ﬁrst few sidelobes. Table 4-29 results from a search because no direct method exists for computing the edge taper for a speciﬁed sidelobe level.

4-18 HANSEN SINGLE-PARAMETER CIRCULAR DISTRIBUTION [20, 21] This distribution leads directly from sidelobe level to a single parameter H that relates through closed-form expressions to all other distribution parameters. The pattern of a uniform distribution is modiﬁed close in to the main beam. By using the U -space

196

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

variable of Taylor, we have U = (2a/λ) sin θ , where a is the radius. The pattern has different expressions in two regions: √ 2I1 (π H 2 − U 2 ) π√H 2 − U 2 f (U ) = √ 2J1 (π U 2 − H 2 ) √ π U2 − H2

|U | ≤ H

(4-88a)

|U | ≥ H

(4-88b)

I1 (x) is the ﬁrst-order modiﬁed Bessel function of the ﬁrst kind. The high function value of Eq. (4-88a) at the boresight reduces the sidelobes of the uniform distribution [Eq. (4-88b)], 17.57 dB, below the level at U = H . The sidelobe level is 2I1 (πH ) SLR = 17.57 + 20 log (4-89) πH Given the sidelobe level [positive (dB)], we use Eq. (4-89) in an iteration scheme to determine H . The aperture distribution is given by E(r) = I0 (πH 1 − r 2 )

|r| ≤ 1

(4-90)

where I0 is the zeroth-order modiﬁed Bessel function of the ﬁrst kind. Equation (4-8) can be integrated for this circularly symmetrical distribution [Eq. (4-90)] to derive the amplitude taper efﬁciency: ATL =

4I12 (πH ) π2 H 2 [I02 (πH ) − I12 (πH )]

(4-91)

Table 4-30 lists the parameters of the Hansen distribution for various sidelobe levels. At the top, Tables 4-29 and 4-30 are very similar. Any sidelobe level can be achieved with this distribution, subject to tolerance problems generated by any low-sidelobe design. The distribution is not optimum, but it is convenient.

4-19

TAYLOR CIRCULAR-APERTURE DISTRIBUTION [22]

Similar to the line source, the Taylor circular-aperture distribution modiﬁes inner zeros of the uniform amplitude and phase circular-aperture k-space pattern to approximate the Dolph–Chebyshev distribution. By use of the variable πU = ka sin θ the uniform distribution pattern is found to be J1 (πU )/πU . We remove n − 1 inner zeros and add those of the Dolph–Chebyshev distribution: n−1 J1 (πU ) (1 − U 2 /UN2 ) N=1 f (U ) = n−1 πU (1 − U 2 /SN2 ) N=1

(4-92)

TAYLOR CIRCULAR-APERTURE DISTRIBUTION

TABLE 4-30

197

Hansen Single-Parameter Circular-Aperture Distribution

Sidelobe Level (dB)

H

Edge Taper (dB)

ATL (dB)

Beamwidth Factor

20 22 24 25 26 28 30 32 34 35 36 38 40 45 50

0.48717 0.66971 0.82091 0.88989 0.95573 1.08027 1.19770 1.30988 1.41802 1.47084 1.52295 1.62525 1.72536 1.96809 2.20262

4.49 7.79 10.87 12.35 13.79 16.59 19.29 21.93 24.51 25.78 27.04 29.53 31.98 38.00 43.89

0.09 0.27 0.48 0.60 0.72 0.96 1.19 1.42 1.64 1.75 1.85 2.05 2.24 2.68 3.08

1.0484 1.0865 1.1231 1.1409 1.1584 1.1924 1.2252 1.2570 1.2876 1.3026 1.3174 1.3462 1.3742 1.4410 1.5039

Given a zero of J1 (x), J1 (x1N ) = 0, let x1N = πSN . By retaining approximately the same number of zeros in the visible region as in the uniform distribution, we avoid superdirectivity. The new zeros UN are modiﬁed zeros of the uniform distribution: A2 + (N − 12 )2 UN = Sn A2 + (n − 12 )2

(4-93)

where A relates to the maximum sidelobe level, cosh πA = b and 20 log b = sidelobe level(dB). Equation (4-93) is the same as Eq. (4-20) except for the scaling constant Sn , the nth zero of J1 (x) divided by π. Equation (4-92) gives the U -space pattern of the new distribution. We expand the aperture distribution in a Fourier–Bessel series: E(r) =

n−1

Bm J0 (πSm r)

r≤1

(4-94)

m=0

We compute coefﬁcients Bm by transforming the Fourier–Bessel series [Eq. (4-94)] into U -space and comparing the far-ﬁeld pattern with Eq. (4-92). As indicated in Eq. (4-94), the series contains only n nonzero terms: B0 = 1 Bm =

−

n−1

J0 (πSm )

(1 − Sm2 /UN2 )

N=1 n−1

N=1,N =m

(1 − Sm2 /SN2 )

m = 1, 2, . . . , n − 1

(4-95)

198

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

Example Design a Taylor circular-aperture distribution with 30 dB maximum sidelobes and n = 6. We use Eq. (4-21) to calculate the constant A: b = 1030/20 = 31.6228 cosh−1 b = 1.32 A= π We substitute this value into Eq. (4-93) to compute the ﬁve nulls: No. Null UN

1

2

3

4

5

1.5582

2.2057

3.1208

4.1293

5.1769

The ﬁrst null of the uniform distribution occurs at x11 x11 = 3.83171 S1 = = 1.2197 π We use this with the location of the ﬁrst zero to determine the null beamwidth factor: BWnull =

U1 1.5582 = = 1.2775 S1 1.2197

The coefﬁcients of the Fourier series [Eq. (4-95)] are given in Table 4-31. Figure 4-28 contains the k-space pattern.

10 Taylor 30 dB

Bayliss 27.4 dB

Pattern level, dB

20

30

40

50

60 −20

0 ka (sin q − sin q0)

20

FIGURE 4-28 Taylor and Bayliss circular aperture distributions to give 30-dB sidelobes (n = 6).

TAYLOR CIRCULAR-APERTURE DISTRIBUTION

199

Tables 4-32 to 4-34 list the characteristics for a few designs of the circular Taylor distribution. Table 4-32 shows that for each sidelobe level there is an optimum n. As the sidelobes are lowered, the optimum value of n increases. The blanks are unsuitable designs. The beamwidth factor (Table 4-33) and the null beamwidth factor (Table 4-34) continue to decrease as n increases at a given sidelobe level. In all three tables the values depend primarily on the sidelobe level. TABLE 4-31 Fourier–Bessel Series Coefﬁcients for Taylor Distribution: 30 dB, n = 6 Bm

Bm Normalized

Function

1.0000 0.93326 0.038467 −0.16048 0.16917 −0.10331

0.53405 0.49841 0.01808 −0.08570 0.09035 −0.05517

1 J0 (x11 r) J0 (x12 r) J0 (x13 r) J0 (x14 r) J0 (x15 r)

No. 0 1 2 3 4 5

TABLE 4-32 Amplitude Taper Losses of Taylor Circular-Aperture Distribution (dB) Sidelobe Level (dB) n

25

30

35

40

45

50

4 6 8 12 16 20

0.30 0.28 0.43 1.03 1.85

0.71 0.59 0.54 0.62 0.86 1.20

1.14 1.03 0.94 0.86 0.87 0.94

1.51 1.48 1.40 1.28 1.22 1.20

1.84 1.88 1.82 1.71 1.64 1.60

2.23 2.21 2.12 2.05 2.01

TABLE 4-33 Beamwidth Factor of Taylor Circular-Aperture Distribution Sidelobe Level (dB) n

25

30

35

40

45

50

4 6 8 12 16 20

1.0825 1.0504 1.0295 1.0057 0.9927

1.1515 1.1267 1.1079 1.0847 1.0717 1.0634

1.2115 1.1957 1.1796 1.1580 1.1451 1.1367

1.2638 1.2581 1.2457 1.2262 1.2137 1.2054

1.3095 1.3149 1.3067 1.2899 1.2782 1.2701

1.3666 1.3632 1.3499 1.3391 1.3314

200

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

TABLE 4-34 Null Beamwidth Factor of Taylor Circular-Aperture Distribution Sidelobe Level (dB)

4-20

n

25

30

35

40

45

50

4 6 8 12 16 20

1.1733 1.1318 1.1066 1.0789 1.0643

1.3121 1.2775 1.2530 1.2244 1.2087 1.1989

1.4462 1.4224 1.4001 1.3716 1.3552 1.3442

1.5744 1.5654 1.5470 1.5197 1.5029 1.4920

1.6960 1.7056 1.6928 1.6680 1.6514 1.6402

1.8426 1.8370 1.8162 1.8003 1.7890

BAYLISS CIRCULAR-APERTURE DISTRIBUTION [8]

We can also design a Bayliss distribution (difference pattern) for circular apertures. This gives us the pattern necessary for monopulse tracking along one axis. The U space pattern has modiﬁed zeros to produce nearly equal sidelobes close in to the main lobes: n−1 (1 − U 2 /UN2 ) (4-96) f (U, φ) = cos φπU J1 (πU ) N=1 n−1 (1 − U 2 /µ2N ) N=0

where UN are the new zeros and πµN are zeros of J1 (πU ). Bayliss lists those zeros µN (Table 4-35). The inner zeros have been removed and replaced by new ones, UN . We compute the zeros in a manner similar to that used for a linear distribution (Section 4-7): ξN2 µ n A2 + n2 UN = A2 + N 2 µn A2 + n2

N = 1, 2, 3, 4 (4-97) N = 5, 6, . . . , n − 1

The four inner zeros had to be adjusted to achieve the desired sidelobe level. Bayliss found these through a computer search. The values for ξN and A can be found through the polynomial approximations [Eq. (4-40)]. TABLE 4-35 Bessel Function Zeros, J1 (πµN ) N 0 1 2 3 4

µN 0.5860670 1.6970509 2.7171939 3.7261370 4.7312271

N

µN

N

µN

N

µN

5 6 7 8 9

5.7345205 6.7368281 7.7385356 8.7398505 9.7408945

10 11 12 13 14

10.7417435 11.7424475 12.7430408 13.7435477 14.7439856

15 16 17 18 19

15.7443679 16.7447044 17.7450030 18.7452697 19.7455093

BAYLISS CIRCULAR-APERTURE DISTRIBUTION

201

Like the Taylor circular aperture distribution, the aperture distribution is expanded in a ﬁnite-length Fourier–Bessel series: E(r, φ) = cos φ

n−1

Bm J1 (πµm r)

r ≤1

(4-98)

m=0

where the coefﬁcients are found by transforming Eq. (4-98) and comparing it with a U -space pattern [Eq. (4-96)]. The coefﬁcients are given by n−1 (1 − µ2m /UN2 ) µ2m N=1 Bm = j J1 (πµm ) n−1 (1 − µ2m /µ2N )

(4-99)

N=0,N =m

Example Design a Bayliss circular-aperture distribution with 30-dB sidelobes and n = 6. We start with Eq. (4-40) to compute coefﬁcients A and ξN : A = 1.64126

ξ1 = 2.07086

ξ2 = 2.62754

ξ3 = 3.43144

ξ4 = 4.32758

We substitute these constants into Eq. (4-97) along with the zeros from Table 4-35 to calculate the modiﬁed zeros: N

1

2

3

4

5

UN

2.2428

2.8457

3.7163

4.6868

5.6994

We use the zeros in Eq. (4-96) to calculate the pattern. The U -space pattern peak can be found by using Eq. (4-41): Umax = 0.7988

ka sin θmax = πUmax = 2.5096

where a is the aperture radius. The coefﬁcients of the Fourier–Bessel series are found from Eq. (4-99) (Table 4-36). The normalized coefﬁcients give an aperture distribution peak of 1. The 3-dB pattern points can be found by searching the pattern: ka sin θ1 = 1.3138

ka sin θ2 = 4.2384

TABLE 4-36 Fourier–Bessel Series Coefﬁcients for Bayliss Distribution: 30 dB, n = 6 No. 0 1 2 3 4 5

Bm

Bm Normalized

Function

0.62680 0.50605 −0.06854 −0.0028703 0.014004 −0.011509

1.2580 1.0157 −0.03415 −0.005761 0.028106 −0.02310

J1 (πµ0 r) J1 (πµ1 r) J1 (πµ2 r) J1 (πµ3 r) J1 (πµ4 r) J1 (πµ5 r)

202

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

TABLE 4-37 Characteristics of a Bayliss Circular-Aperture Distribution, n = 10 3-dB Beam Edge

Sidelobe Level (dB)

Beam Peak, ka/2 sin θmax

ka/2 sin θ1

ka/2 sin θ2

ATL (dB)

PEL (dB)

20 25 30 35 40

2.2366 2.3780 2.5096 2.6341 2.7536

1.165 1.230 1.290 1.346 1.399

3.700 3.940 4.160 4.363 4.551

1.47 1.15 1.32 1.62 1.95

1.80 1.89 1.96 2.01 2.05

Figure 4-28 contains a plot of a Bayliss circular-aperture distribution (n = 6) designed to have sidelobes 30 dB below those of the Taylor distribution with 30-dB sidelobes. The losses to the difference pattern are about 2.6 dB higher than the sum pattern. The amplitude taper efﬁciency is calculated from 2 n−1 4 Bm J1 (πµm r) r dr m=0 0 AT L = 2 1 n−1 2 π 4 Bm J1 (πµm r) r dr

1

(4-100)

m=0

0

where the integrals over φ have been separated and evaluated. An integral expression for the phase error efﬁciency can be found similarly by evaluating the separable cos φ integrals along the coordinate φ = 0, the peak:

PEL(U ) =

2 Bm J1 (πµm r)J1 (πU r)r dr m=0 0 2 1 n−1 4 Bm J1 (πµm r) r dr

2π

1

0

n−1

(4-101)

m=0

Table 4-37 lists the parameters of Bayliss circular-aperture distributions with n = 10 and various sidelobe levels. The optimum design for n = 10 occurs for 25-dB sidelobes.

4-21

PLANAR ARRAYS

We design planar arrays with nearly circular boundaries by sampling circular distributions. Given enough sample points in the array, a distribution such as the circular Taylor will be modeled adequately to produce a similar pattern. We can use pattern multiplication to combine the designs for linear arrays into a planar array, but in the special case of a square array, a true Chebyshev design can be obtained in all planes. A technique has been developed to allow the synthesis from pattern nulls provided that some of the possible nulls are not speciﬁed. We are still left with the problem of specifying the numerous nulls possible with a planar array. Chebyshev Array [23] When we combine two Dolph–Chebyshev linear arrays through pattern multiplication, it produces a pattern that has lower sidelobes than

203

CONVOLUTION TECHNIQUE FOR PLANAR ARRAYS

those speciﬁed in all planes except the principal ones along the axes. These designs give beamwidths in the diagonal planes that are wider than necessary. The pattern deviates from the optimum because sidelobes are suppressed more than necessary. We use a technique on a square array to produce equal sidelobes in all constant φ cuts around the array. The array is square in the number of elements, but different spacings along the axes can produce a rectangular array. We expand the pattern in a single Chebyshev polynomial: TL−1 (x0 cos ψ1 cos ψ2 )

(4-102)

where ψ1 = kdx cos θ cos φ + δ1 and ψ2 = kdy cos θ sin φ + δ2 L = 2N or L = 2N + 1 for L2 elements in the array. We compute x0 from Eq. (4-51) for a given sidelobe level. The pattern for an odd number of elements in each row and column is E(θ, φ) =

N+1 N+1

εm εn Imn cos 2(m − 1)ψ1 cos 2(n − 1)ψ2

L = 2n + 1

m=1 n=1

where εm = 1 for m = 1 and εm = 2 for m = 1. Similarly, E(θ, φ) = 4

N N

Imn cos(2m − 1)ψ1 cos(2n − 1)ψ2

L = 2n

m=1 n=1

The element excitations Imn are given by Imn

2 N+1 N+1 (q − 1)π 2 (p − 1)π cos = εp εq TL−1 x0 cos L p=1 q=1 L L × cos

2π(m − 1)(p − 1) 2π(n − 1)(q − 1) cos L L

L = 2N + 1 (4-103)

or

Imn

1 1 2 π π p− q− N N 4 2 2 = TL−1 x0 cos cos L p=1 q=1 L L

1 1 p− 2π n − 2π m − 2 2 cos × cos L

1 1 q− 2 2 L

L = 2N (4-104)

4-22 CONVOLUTION TECHNIQUE FOR PLANAR ARRAYS We may synthesize a desired pattern through multiplication of two or more simpler patterns. Because patterns derive from Fourier transforms of distributions in space, the distribution to produce the product of two simpler patterns is the convolution of the

204

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

simpler distributions [24, p. 30]. We ﬁnd it easier to synthesize by using a few elements and then build up patterns through multiplication. Consider the convolution of a linear array with another linear array on the same axis. We describe the array as a distribution consisting of weighted impulse functions, δ(x − xi ): N1 a1i δ(x − xi ) A1 (x) = i=1

where a1i are the feeding coefﬁcients and xi are the locations for an N1 -element array. To determine the array that gives the product of two array patterns, we convolve the second array, A2 (x), with the ﬁrst: A1 (x) ∗ A2 (x) =

A1 (τ )A2 (x − τ ) dτ

(4-105)

We evaluate a function at the argument of the impulse function when we convolve the two arrays [25, p. 237]. Equation (4-105) reduces to A1 (x) ∗ A2 (x) =

N2 N1

a1i a2i δ(x − xi − xj )

(4-106)

i=1 j =1

Example Consider the two 2-element arrays in Figure 4-29 and the graphical solution of the convolution. Figure 4-29a shows the location of the elements in the arrays on the x-axis. To perform the convolution, we reﬂect the x-axis of one array and move it across the other array while performing the integral at each location x, the coordinate of the convolution. We have a net result to the integral only when two impulse functions are aligned, x = xi + xj . We have four elements in the resulting array (Figure 4-29c). If the elements are equally spaced in the two arrays, two elements will sum into one. Patterns are the result of a three-dimensional Fourier transform. For a general array with element locations ri , we must perform a convolution along all three axes to ﬁnd the distribution that gives the product of the patterns of two simpler distributions. For the general array, Eq. (4-106) becomes A1 (r) ∗ A2 (r) =

a1i a2j δ(r − ri − rj )

(4-107)

where r is the location vector and ri and rj are the locations of elements in the two arrays. A rectangular array can be described as the convolution of a linear array on the xaxis with a linear array on the y-axis. When y = yj there is a string of values x = xi that satisfy the impulse argument [Eq. (4-107)]. These are the locations of the elements. We step through all values of yj until the entire array is formed. Equation (4-107) gives the feeding coefﬁcient of each element a1i a2j since no two elements of the convolution are in the same place. The pattern of the rectangular array is the product of the linear array patterns along the axes.

CONVOLUTION TECHNIQUE FOR PLANAR ARRAYS

205

a11 A1 (x) x11

a21

a12

A2 (x)

x21

x12

a22

x22

(a)

A1 (t) t

x

A2 (x − t) t

a11a22

x12 + x21 a12a21

x21 + x22

a11a21

x11 + x22

x11 + x21

(b)

a21a22 (c)

FIGURE 4-29 Convolution of two linear arrays: (a) separate arrays; (b) graphical convolution; (c) convolution.

Given an array, we compute the pattern from a Fourier transform containing N terms each of which corresponds to one element. Ignoring the element pattern, we have N E= ai ej k·ri (4-108) i=1

The array has N − 1 independent nulls (zeros) in the pattern. Given the set of nulls kj we can substitute them into Eq. (4-108) to form a matrix equation in N − 1 unknowns ai . We must normalize one coefﬁcient, ai = 1, to solve the set of equations for the

206

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

feeding coefﬁcients:

j k1 ·r1 e a2 ej k2 ·r1 a3 [B] . = − .. .. . j k ·r aN e N −1 1

(4-109)

where bij = ej ki ·rj +1 . We ﬁnd the direct solution of Eq. (4-109) unwieldy for a large array. We can subdivide the array into smaller arrays whose convolution is the total array and use pattern multiplication. We reduce the number of nulls we need to specify in the synthesis of an array. Convolution can be used in the synthesis of planar arrays by using a rhombic array (four elements) as the basic building block [26] (Figure 4-30). If we convolve two identically shaped rhombic arrays, we obtain a nine-element (three on a side) array (Figure 4-30b). By continuing to convolve the resulting array with other rhombic arrays, we can build up a large array in the shape of the rhombus. Each rhombic array has three pattern nulls without the symmetry of the linear array about some axis. The rhombic array has symmetry only about the plane of the rhombus. We build up an array of N + 1 by N + 1 elements through the convolution of N rhombic arrays. The original array has (N + 1)(N + 1) − 1 independent nulls. The convolution of N rhombic arrays reduces the number of independent nulls to 3N . Similarly, when we use the convolution of two linear arrays to form a square array, N + 1 by N + 1, the number of independent nulls is 2N , or N for each array. We denote a single rhombic array as RA1 and the convolution of two rhombic arrays as RA2 . The number of elements on each side of an RAN array is N + 1. We can convolute a rhombic array with a linear array to form an M × N array (M > N ). Denote the linear array by LN , where the array has N + 1 elements. The planar array P AM,N becomes P AM,N = LM−N ∗ RAN−1 (4-110) We specify 3(N − 1) nulls in space for the rhombic arrays and M − N nulls about the axis of the array. Like all convolutions, the pattern is the product of the individual array patterns.

(a)

(b)

FIGURE 4-30 Rhombic array with its convolution: (a) rhombic array RA1 ; (b) convolution of two rhombic arrays RA2 .

CONVOLUTION TECHNIQUE FOR PLANAR ARRAYS

207

This method allows the speciﬁcation of nulls in space with other than linear symmetry. Second, it reduces the required speciﬁcation of nulls. Third, it provides a method for synthesis of triangularly or hexagonally spaced elements. Example Consider the six-element rectangular array shown in Figure 4-31a. It can be broken down into the convolution of a four-element rectangular array (rhombic) and a two-element linear array from Eq. (4-110): P A3,2 = L1 ∗ RA1 Pick the three nulls of the rhombic array at θ

90◦

90◦

90◦

φ

110◦ −60◦ 180◦

We measure the pattern nulls from the normal of the plane containing the rhombus and the x-axis (φ). For an array of broadcast towers, the nulls point toward the horizon. We restrict θ to less than or equal to 90◦ . We substitute the positions of the elements (Figure 4-31b) and the nulls into Eq. (4-109) to solve for the feeding coefﬁcients of the rhombic array (Table 4-38).

(a)

(b)

(c)

FIGURE 4-31 Rectangular array from convolution of rhombic and linear arrays: (a) six-element rectangular array; (b) rhombic array; (c) linear array. TABLE 4-38 Coefﬁcients of Rhombic Array for Horizon (θ = 90◦ ) Nulls at φ = 100◦ , −60◦ , and 180◦ Element

Amplitude (dB)

Phase (deg)

1 2 3 4

0.00 4.12 0.00 4.12

0.0 −79.2 −109.2 −30.2

208

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

TABLE 4-39 Coefﬁcients of Six-Element Rectangular Array with Pattern of Figure 4-32b Element

Amplitude (dB)

Phase (deg)

1 2 3 4 5 6

0.00 8.13 4.12 0.00 8.13 4.12

0.0 −88.5 177.2 147.2 −124.3 −30.1

We pick the single null of the two-element array at 135◦ . This null has symmetry about the axis of the array. With the ﬁrst element at zero phase, we pick the element phase to cancel the ﬁrst element voltage when θ = 135◦ : ◦

◦

◦

◦

phase = 180 − 360 (0.3)cos 135 = 256.37

When we convolute the two arrays, we obtain the feeding coefﬁcients from Eq. (4-107) (Table 4-39). The elements in the center that result from two convolutions have summed feeding coefﬁcients producing a six-element array. Figure 4-32 shows the patterns of the convolution. We obtain the six-element array pattern (Figure 4-32b) by multiplication of the patterns of the individual subarrays (Figure 4-32a).

4-23

APERTURE BLOCKAGE

Blocking an aperture reduces the gain and raises the sidelobes. The blockage either scatters the aperture power in unwanted directions in a broad pattern or is just an area without ﬁelds. Scattered blockage causes higher sidelobes and greater loss than the

10 dB

Rhombic array

10 dB

20 dB

20 dB

30 dB

30 dB

Pattern of convolution

Linear array

(a)

(b)

FIGURE 4-32 Patterns of the convolution of a rhombic and a linear array to form the six-element rectangular array of Figure 4-31.

APERTURE BLOCKAGE

209

nonexcitation blockage. Scattered blockage has the same power input as the unblocked aperture, but ﬁelds scattered off the blockage do not contribute signiﬁcantly to the maximum ﬁeld. Compared to the unblocked aperture, the blockage efﬁciency becomes 2 j k·r Ee ds blocked max blockage efﬁciency = 2 j k·r Ee ds unblocked

scattered

(4-111)

max

We use Eq. (2-16) to compute the directivity of each distribution by using the total power radiated from the unblocked aperture [denominator of Eq. (2-16)] for the blocked aperture. A centrally blocked circular aperture with a uniform distribution has the blockage efﬁciency (1 − b2 )2 (scattered), where b is the normalized blockage radius. The blockage of a circular Gaussian distribution has a simple blockage function: blockage efﬁciency =

e−ρb − e−ρ 1 − e−ρ

2

The second type of blockage is an area without ﬁelds. The blockage does not waste power in the aperture. When we take the ratio of the two directivities, we must account for the power in each aperture: 2 j k·r Ee ds |E|2 ds blocked unblocked max blockage efﬁciency = 2 Eej k·r ds |E|2 ds unblocked

max

blocked

nonexcitation

(4-112) The blockage of a uniformly excited centrally blocked circular aperture where the center is not excited reduces the directivity only by the area lost from the aperture, 1 − b2 (nonexcitation). In a sense, nonexcitation blockage is not a true loss; it is a loss of potential radiation aperture. Table 4-40 lists the blockage losses of centrally blocked circular apertures calculated by Eq. (4-111), the more severe case. The uniformly excited aperture is affected least by blockage. All points are equally important. The tapered distributions suffer more loss with increased taper toward the edge. The lists for different tapered distributions track each other fairly closely, and any one of them gives a good estimate of the blockage loss. Blockage causes sidelobes. In the case of scattered blockage the exact sidelobes cannot be found without an analysis of the scatterer. A Cassegrain reﬂector would need a geometric theory of diffraction (GTD) analysis to locate the directions of scattering from the subreﬂector. We can handle the nonexcitation blockage in a general fashion. Consider the aperture to be broken into two radiating apertures. The ﬁrst is the unblocked aperture; the second is the blockage. If we take the blockage aperture to be 180◦ out of phase with respect to the unblocked aperture distribution, the sum gives us the blocked distribution. We use this analysis as an approximation to scattered blockage with the realization that scattering may produce unpredicted lobes.

210

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

TABLE 4-40 Blockage Losses of Circular-Aperture Distributions (dB) Central Blockage (%) 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

Taylor

Hansen

Uniform

Gaussian 12-dB Edge

30 dB, n = 6

40 dB, n = 6

30 dB

40 dB

0.02 0.03 0.04 0.06 0.07 0.09 0.11 0.13 0.15 0.17 0.20 0.22 0.26 0.29 0.32 0.36 0.39 0.43 0.47 0.52 0.56

0.04 0.06 0.08 0.10 0.13 0.16 0.19 0.23 0.27 0.32 0.36 0.41 0.47 0.52 0.58 0.65 0.71 0.78 0.86 0.94 1.02

0.04 0.06 0.08 0.10 0.13 0.16 0.20 0.24 0.28 0.32 0.37 0.42 0.48 0.54 0.60 0.67 0.74 0.81 0.88 0.96 1.05

0.05 0.08 0.11 0.14 0.18 0.22 0.26 0.31 0.37 0.43 0.49 0.56 0.63 0.71 0.79 0.88 0.97 1.07 1.17 1.27 1.38

0.05 0.07 0.09 0.12 0.16 0.19 0.23 0.28 0.33 0.38 0.43 0.49 0.56 0.63 0.70 0.77 0.86 0.94 1.03 1.12 1.22

0.07 0.09 0.13 0.17 0.21 0.26 0.32 0.38 0.44 0.51 0.59 0.67 0.76 0.85 0.95 1.06 1.17 1.28 1.40 1.53 1.66

We can calculate an upper bound on the sidelobes easily. Assume that the blockage distribution is uniform and compared to the main aperture, produces a broad, ﬂat beam. Since the blockage aperture ﬁelds are 180◦ out of phase from the unblocked aperture ﬁelds, their radiation subtracts from the main beam and adds sidelobes 180◦ out of phase with respect to the main lobe. The sidelobe due to the blockage is proportional to the area: sidelobe level = 20 log b. This formula estimates values much higher than are realized. Table 4-41 lists the sidelobes of a centrally blocked Taylor circular aperture distribution with 40-dB design sidelobes. They are far less than predicted by the upper bound.

TABLE 4-41 Sidelobe Level Due to Central Blockage of a Circular Aperture with Taylor Distribution (40 dB, n = 6) Blockage (% of Diameter)

Sidelobe Level (dB)

Blockage (% of Diameter)

Sidelobe Level (dB)

Blockage (% of Diameter)

Sidelobe Level (dB)

7 8 9 10 11 12

34.5 32.8 31.3 29.8 28.5 27.3

13 14 15 16 17 18

26.1 25.6 24.2 23.3 21.7 21.7

19 20 21 22 23 24

21.1 20.4 19.7 19.1 18.5 18.0

QUADRATIC PHASE ERROR

211

Ludwig [27] has found distributions to reduce the sidelobes of blocked apertures. The ﬁrst sidelobe can be reduced only a little, but the outer sidelobe levels can be controlled. In many applications one high sidelobe next to the main beam is acceptable. A Taylor distribution for circular apertures with a zero edges value, like Section 4-5 for linear apertures, reduces the far-out sidelobes. A second aperture function with a doughnut distribution also reduces all but the ﬁrst sidelobe. Reducing the edge taper of the blockage distribution lowers the blockage-caused sidelobes. Sachidananda and Ramakrishna [28] use a numerical optimization technique to reduce the sidelobes of a blocked aperture for both the sum and difference patterns of a monopulse excitation. They start with the Taylor and Bayliss circular-aperture distribution functions [Eqs. (4-94) and (4-98)]. The coefﬁcients Bm are determined through the numerical optimization, which restrains the sidelobes while optimizing the monopulse tracking coefﬁcients and sum pattern gain. 4-24 QUADRATIC PHASE ERROR A linear phase error function scans the aperture beam with some loss of gain because of the shrinking of the projected aperture in the direction of the main beam. Quadratic phase error (order 2) does not scan the beam but causes loss and a change in the sidelobe levels and the depth of the nulls between them. This phase error arises mainly from defocusing when the source of radiation appears as a point source. A feed axially displaced from the focus of a parabolic reﬂector produces quadratic phase error in the aperture. The ﬂare angle of a horn changes the distance from the assumed point source in the throat to different points in the aperture at the end of the ﬂare. We can approximate the phase distribution as quadratic. We express the quadratic phase error in a line-source aperture as e−j 2πS(2x/a)

2

linear:

|x/a| ≤ 0.5

(4-113a)

where S is a dimensionless constant, cycles and a is the aperture width. Similarly, the circular-aperture phase is circular: e−j 2πSr

2

r ≤1

(4-113b)

where r is the normalized radius. We use Eq. (4-7) with the linear-source aperture phase error [Eq. (4-113a)] and use Eq. (4-9) with the quadratic phase error [Eq. (4-113b)] in a circularly symmetrical aperture distribution to compute phase error loss:

PELx =

PEL =

2 2 E(x)e−j 2πS(2x/a) dx −a/2 2 a/2 |E(x)| dx a/2

linear

(4-114)

−a/2

2 E(r)e r dr 0 2 1 |E(r)|r dr 1

−j 2πSr 2

0

circular

(4-115)

212

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

A few distributions have simple formulas for the phase error efﬁciency when excited with quadratic phase error [29]: PELx =

uniform linear:

√ & 1 % 2 √ C (2 S) + S 2 (2 S) 2S

(4-116)

where C(t) and S(t) are the Fresnel integrals, tabulated functions: uniform circular: PEL = Circular Gaussian(e−ρr ) : 2

PEL =

sin πS πS

2 (4-117)

ρ 2 [1 − 2e−ρ cos(2πS) + e−2ρ ] (4-118) [ρ 2 + (2πS)2 ](1 − e−ρ )2

We use numerical integration for the general distribution. Table 4-42 lists quadratic phase error losses for various linear-aperture distributions. We will use the lists for uniform and cosine distributions to evaluate the gains of rectangular horns. The effect of quadratic phase error decreases as the distribution taper increases. Table 4-43 lists results for a few circularly symmetrical aperture distributions. Quadratic phase error raises the sidelobes of low-sidelobe antennas. Figure 4-33 shows the effects on a circular Taylor distribution with 35-dB design sidelobes. The ﬁrst lobe increases, and the null between the main lobe and the ﬁrst sidelobe disappears as the quadratic phase error increases. A source antenna spaced a ﬁnite distance, as on

TABLE 4-42 Quadratic Phase Error Loss of Linear-Aperture Distributions (dB) Cycles, S 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00

Uniform

Cosine

Cosine

Cosine2 + 19.9-dB Pedestal

0.04 0.15 0.34 0.62 0.97 1.40 1.92 2.54 3.24 4.04 4.93 5.91 6.96 8.04 9.08 9.98 10.60 10.87 10.80 10.50

0.02 0.07 0.16 0.29 0.45 0.65 0.88 1.14 1.43 1.75 2.09 2.44 2.82 3.20 3.58 3.95 4.31 4.65 4.97 5.25

0.01 0.04 0.09 0.16 0.25 0.36 0.49 0.64 0.80 0.97 1.16 1.36 1.57 1.79 2.01 2.23 2.46 2.69 2.91 3.13

0.02 0.07 0.16 0.28 0.44 0.63 0.84 1.08 1.34 1.62 1.90 2.19 2.48 2.76 3.04 3.29 3.52 3.73 3.92 4.09

2

213

QUADRATIC PHASE ERROR

TABLE 4-43

Cycles, S

Quadratic Phase Error Loss of Circular-Aperture Distributions (dB)

Uniform

0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00

0.04 0.14 0.32 0.58 0.91 1.33 1.83 2.42 3.12 3.92 4.86 5.94 7.20 8.69 10.46 12.62 15.39 19.23

Gaussian 12-dB Edge 0.03 0.13 0.29 0.53 0.82 1.20 1.64 2.16 2.76 3.44 4.22 5.08 6.04 7.10 8.24 9.44 10.66 11.81 12.75 13.36

Taylor

Hansen

30 dB

40 dB

30 dB

40 dB

0.04 0.15 0.33 0.59 0.93 1.36 1.86 2.46 3.16 3.95 4.86 5.88 7.01 8.25 9.56 10.87 12.01 12.80

0.03 0.11 0.26 0.46 0.72 1.03 1.41 1.84 2.33 2.87 3.47 4.11 4.79 5.50 6.21 6.91 7.56 8.14 8.62 8.99

0.03 0.11 0.25 0.45 0.70 1.01 1.38 1.81 2.30 2.85 3.46 4.16 4.85 5.63 6.43 7.26 8.09 8.88 9.60 10.20

0.02 0.08 0.19 0.34 0.53 0.76 1.03 1.34 1.69 2.08 2.50 2.95 3.43 3.94 4.46 4.98 5.51 6.03 6.53 6.99

25

30

Level, dB

35

40

45

l/16 Maximum - aperture phase error distance =

2D l

2

l/32

l/64

0

50 6

7

8

9

10

2pa sin q q

FIGURE 4-33 Effects of quadratic phase error on 35-dB circular Taylor distribution (n = 6).

214

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

an antenna measurement range, feeds the aperture with a quadratic phase error. The source would have to be spaced 8D 2 /λ to measure the sidelobe level within 0.5 dB. Low-sidelobe antennas require greater distances than the usual 2D 2 /λ for accurate sidelobe measurement [30]. 4-25 BEAM EFFICIENCY OF CIRCULAR APERTURES WITH AXISYMMETRIC DISTRIBUTION From Eq. (1-27) we can derive an approximate formula for axisymmetric distributions that depends on the normalized variable kr (or U ). For large apertures we can approximate cos θ ≈ 1 in the main beam, integrate the φ integral to obtain 2π, and incorporate the (ka)2 directivity factor into the integral:

kr1

ATL · PEL beam efﬁciency =

|f (kr )|2 kr dkr

(4-119)

0

2|f (0)|2 u1 |f (U )|2 U dU ATL · PEL

=

(4-120)

0

2|f (0)|2

where kr is the factor (2πa sin θ )/λ, U (the Taylor distribution factor) is (2a sin θ )/λ, and a is the aperture radius. U1 and kr1 correspond to the beam edge θ1 . The integrals of Eqs. (4-119) and (4-120) cause underestimations of beam efﬁciency for small apertures when we ignore the cos θ factor, which should divide the argument of the integral. Table 4-44 lists beam edges in kr -space (2πa sin θ )/λ for various distributions along with the beam efﬁciency at the null beam edge. We can use it to determine the aperture size required for a given beam efﬁciency beamwidth speciﬁcation. Example Calculate the aperture radius to give a 90% beam efﬁcient beamwidth of 5◦ for the distribution: parabolic on 12-dB pedestal. TABLE 4-44 Beam Efﬁciencies of Axisymmetric Circular-Aperture Distributions

Distribution Uniform Parabolic Parabolic + 12-dB pedestal Taylor 30 dB, n = 6 30 dB, n = 10 40 dB, n = 6 Hansen 30 dB 40 dB

kr = 2πa sin θ/λ Speciﬁed Beam Efﬁciency (%)

Null, kr

Beam Efﬁciency at Null (%)

80

85

90

95

3.83 5.14 4.58

83.7 98.2 96.4

2.82 2.81 2.60

4.71 3.03 2.81

5.98 3.31 3.10

3.75 3.64

4.90 4.74 6.00

96.2 91.4 99.5

2.65 2.76 2.90

2.88 3.06 3.13

3.19 3.65 3.42

3.82

5.37 6.64

99.3 99.9

2.79 3.17

3.01 3.42

3.28 3.73

3.69 4.19

3.85

REFERENCES

215

From Table 4-43, 5◦ 2πa sin λ 2 3.10 a = = 11.31 λ 2π sin(5◦ /2)

kr (90%) = 3.10 =

The beam edge has cos 2.5◦ = 0.999, which justiﬁes the approximation in Eq. (4-119). REFERENCES 1. T. T. Taylor, Design of line source antennas for narrow beamwidth and low sidelobes, IEEE Transactions on Antennas and Propagation, vol. AP-4, no. 1, January 1955, pp. 16–28. 2. C. L. Dolph, A current distribution for broadside arrays which optimizes the relationship between beamwidth and sidelobe level, Proceedings of IEEE, vol. 34, June 1946, pp. 335–348. 3. R. C. Hansen, Linear arrays, Chapter 9 in A. W. Rudge et al., eds., The Handbook of Antenna Design, Vol. 2, Peter Peregrinus, London, 1982. 4. D. R. Rhodes, On a new condition for physical realizability of planar antennas, IEEE Transactions on Antennas and Propagation, vol. AP-19, no. 2, March 1971, pp. 162–166. 5. D. R. Rhodes, On the Taylor distribution, IEEE Transactions on Antennas and Propagation, vol. AP-20, no. 2, March 1972, pp. 143–145. 6. R. S. Elliott, Antenna Theory and Design, Prentice-Hall, Englewood Cliffs, NJ, 1981. 7. D. A. Pierre, Optimization Theory with Applications, Wiley, New York, 1969. 8. E. T. Bayliss, Design of monopulse antenna difference patterns with low sidelobes, Bell System Technical Journal, vol. 47, May–June 1968, pp. 623–650. 9. P. M. Woodward, A method of calculating the ﬁeld over a plane aperture required to produce a given polar diagram, Proceedings of IEE, vol. 93, pt. IIIA, 1947, pp. 1554–1558. 10. S. A. Schelkunoff, A mathematical theory of linear arrays, Bell System Technical Journal, vol. 22, 1943, pp. 80–107. 11. J. Kraus, Antennas, McGraw-Hill, New York, 1950. 12. R. S. Elliott, Beamwidth and directivity of large scanning arrays, Microwave Journal, vol. 6, no. 12, December 1963, pp. 53–60. 13. A. T. Villeneuve, Taylor patterns for discrete arrays, IEEE Transactions on Antennas and Propagation, vol. AP-32, no. 10, October 1984, pp. 1089–1092. 14. R. S. Elliott, On discretizing continuous aperture distributions, IEEE Transactions on Antennas and Propagation, vol. AP-25, no. 5, September 1977, pp. 617–621. 15. H. J. Orchard, R. S. Elliott, and G. J. Stern, Optimising the synthesis of shaped beam antenna patterns, IEE Proceedings, vol. 132, pt. H, no. 1, February 1985, pp. 63–68. 16. Y. U. Kim and R. S. Elliott, Shaped-pattern synthesis using pure real distributions, IEEE Transactions on Antennas and Propagation, vol. AP-36, no. 11, November 1988, pp. 1645–1648. 17. F. Ares-Pena, Application of genetic algorithms and simulated annealing to some antenna problems, Chapter 5 in Y. Rahamat-Samii and E. Michielssen, eds., Electromagnetic Optimization by Genetic Algorithms, Wiley, New York, 1999. 18. C. H. Walters, Traveling Wave Antennas, Dover, New York, 1970. 19. G. Doundoulakis and S. Gethin, Far ﬁeld patterns of circular paraboloidal reﬂectors, IRE National Convention Record, pt. 1, 1959, pp. 155–173.

216

APERTURE DISTRIBUTIONS AND ARRAY SYNTHESIS

20. R. C. Hansen, Circular aperture distribution with one parameter, Electronic Letters, vol. 11, no. 8, April 17, 1975, p. 184. 21. R. C. Hansen, A one-parameter circular aperture with narrow beamwidth and low sidelobes, IEEE Transactions on Antennas and Propagation, vol. AP-24, no. 4, July 1976, pp. 477–480. 22. T. T. Taylor, Design of circular apertures for narrow beamwidth and low side lobes, IEEE Transactions on Antennas and Propagation, vol. AP-8, no. 1, January 1960, pp. 17–22. 23. F. I. Tseng and D. K. Cheng, Optimum scannable planar arrays with an invariant side-lobe level, Proceedings of IEEE, vol. 56, no. 11, November 1968, pp. 1771–1778. 24. D. Steinberg, Principles of Aperture and Array System Design, Wiley, New York, 1976. 25. D. K. Cheng, Analysis of Linear Systems, Addison-Wesley, Reading, MA., 1959. 26. S. R. Laxpati, Planar array synthesis with prescribed pattern nulls, IEEE Transactions on Antennas and Propagation, vol. AP-30, no. 6, November 1982, pp. 1176–1183. 27. A. C. Ludwig, Low sidelobe aperture distributions for blocked and unblocked circular apertures, IEEE Transactions on Antennas and Propagation, vol. AP-30, no. 5, September 1982, pp. 933–946. 28. M. Sachidananda and S. Ramakrishna, Constrained optimization of monopulse circular aperture distribution in the presence of blockage, IEEE Transactions on Antennas and Propagation, vol. AP-31, no. 2, March 1983, pp. 286–293. 29. A. W. Love, Quadratic phase error loss in circular apertures, Electronics Letters, vol. 15, no. 10, May 10, 1979, pp. 276, 277. 30. P. S. Hacker and H. E. Schrank, Range requirements for measuring low and ultralow sidelobe antenna patterns, IEEE Transactions on Antennas and Propagation, vol. AP-30, no. 5, September 1982, pp. 956–966.

5 DIPOLES, SLOTS, AND LOOPS

A dipole is a conductive rod usually split in the center and fed from a balanced transmission line that carries equal and oppositely ﬂowing currents. Not all dipoles are split and fed in the center because currents can be excited on it electromagnetically or it can be shunt fed. The dipole length determines possible current distributions in modes, and when we place a continuous rod near an antenna radiating a linear polarization component directed along the rod, it excites a standing-wave current on the rod. The amount excited on the rod depends on how close its length is to resonance and the antenna spacing. Of course, the continuous rod loads the fed antenna through mutual coupling. We can feed the continuous rod from a coax line by attaching the outer conductor to the center and then connecting the center conductor away from the center in a shunt feed. A slot is a narrow-width opening in a conductive sheet. When excited by a voltage across the narrow dimension it appears to radiate from an equivalent magnetic current ﬂowing along the long dimension that replaces the voltage (or electric ﬁeld) across it. Most slots, similar to dipoles, have a ﬁnite length with either short or open circuits at both ends. The voltage along the slot forms a standing wave. Of course, magnetic currents are ﬁctitious, and real electric currents ﬂow in the conductive sheet around the slot. These currents do not have a simple distribution and are difﬁcult to use for analysis, so we use simpler magnetic currents, although when analyzing a slot using the method of moments, we model the conductors around the slot and calculate patterns, reaction, and so on, from these real currents. Initial slot calculations assume that the conductive sheet is inﬁnite, similar to the analysis of dipoles situated in free space. Complete analysis of the dipole requires analysis in the presence of the mounting conﬁguration. Similarly, full analysis of slots includes the effects of the ﬁnite sheet and scattering from the objects around it.

Modern Antenna Design, Second Edition, By Thomas A. Milligan Copyright 2005 John Wiley & Sons, Inc.

217

218

DIPOLES, SLOTS, AND LOOPS

After considering ideal cases, we analyze the effects of ﬁnite ground planes, nearby scatterers, and the interaction between dipoles and slots. The batwing antenna presents an unusual case where the antenna at ﬁrst glance looks like a dipole but actually radiates from a combination of a slot and a ﬁnite dipole structure. Another interesting case is the waveguide slot. Currents ﬂow on the inside surfaces of a waveguide, and the ﬁnite current skin depth prevents it from reaching the outside. The metal walls shield the currents and prevent the loss of power by radiation. When we cut a slot in the wall, the internal currents ﬂow out the slot and onto the outside of the waveguide and radiate. The excitation and length of the slot relative to the internal currents determine the amount radiated. Similarly, the slots load the waveguide as a transmission line because of the loss of power. Our analysis starts with a dipole in free space or a slot on an inﬁnite conductive sheet. The two problems are duals. Dipoles radiate from a standing-wave electric (real) current, whereas the slot radiates from a standing-wave magnetic current. We use the same mathematics for both patterns. By the Babinet–Booker principle of complementary structures, we relate the input impedance of one to the other. Both structures radiate the same pattern but differ in polarization. Dipoles and slots share the same analysis through duality, so we develop them together. Singly and in arrays, they satisfy many antenna needs. Although they share a dual analysis, they have unique feeding requirements. We discuss baluns for dipoles and waveguide slot excitations as practical implementations. In Chapter 2 we presented the analysis of a small loop excited with a uniform current (Section 2-1.2). The loop current was replaced with a small magnetic current element ﬂowing along the normal to the plane of the loop. Multiple turns and ferrite loading increase the efﬁciency of loops and produce a more useful antenna. Exciting a uniform current on a loop is a difﬁcult task that offers little practical beneﬁt. The loops discussed will have standing-wave electric currents excited on them determined by feeding methods. The natural balun used to excite a small loop produces a standingwave current with zero current at the point where the two sides are connected to form the loop. A resonant length loop of about one wavelength perimeter radiates a dipole pattern from a standing-wave current. The quadriﬁlar helix consists of two loops twisted around a common axis. The twist produces currents that radiate circular polarization from each loop. Analysis shows that the currents are standing wave. Feeding a dipole or loop requires a balun to prevent current ﬂow either along the outside of a coaxial feeder or excitation of unbalanced currents along a two-wire line. The current ﬂowing along the outside of the coax or unbalanced currents on the twowire line radiate in unwanted directions or radiate undesired polarization. When we design an antenna without considering or knowing its ﬁnal mounting, we produce an uncontrolled situation without a balun. Our initial conﬁguration may work without a balun, but the antenna may fail to produce the desired pattern in the ﬁnal location. If you control the installation completely, you can reduce your design effort and may be able to eliminate the balun.

5-1 STANDING-WAVE CURRENTS Think of a dipole as a diverging two-wire transmission line. The characteristic impedance increases as the wave approaches the open-circuited ends. The slot is the

STANDING-WAVE CURRENTS

219

Current S Voltage

Open circuit

FIGURE 5-1 Standing wave.

dual of a strip dipole. A voltage excited across the slot propagates along a slotline toward short-circuited ends. Each type of transmission line reﬂects the incident wave from the terminations. The combination of two waves traveling in opposite directions creates a standing wave on the line. The current and voltage are 90◦ out of phase and 90◦ out of space phase (Figure 5-1). Current and voltage change places on the short-circuited termination of the slot. The dipole is not a uniform transmission line, but we can approximate the current as a standing wave with the current vanishing on the ends. The slot voltage is a standing wave also vanishing on the ends. The standing waves for a center-fed dipole or slot are expressed as follows: Dipole L I = I0 sin k −z 2 L I = I0 sin k +z 2

Slot L V = V0 sin k −z 2 L V = V0 sin k +z 2

z≥0

(5-1)

z≤0

The voltage distribution on the slot is equivalent to a magnetic current. We calculate radiation from the linear sinusoidal current distributions by the vector potentials: electric (slot) (Section 2-1.2) and magnetic (dipole) (Section 2-1.1). Figure 5-2 gives typical sinusoidal distributions for various lengths. The currents match at the feed point and vanish on the ends. Consider the pattern of the 2λ dipole at θ = 90◦ . We can assume that it is a continuous array and sum the ﬁelds from each z q I (V)

l/2

Dipole

Slot

I (V)

l

I (V)

3l/2

Electrical length

FIGURE 5-2 Sinusoidal distributions.

2l

220

DIPOLES, SLOTS, AND LOOPS 100

5

80

4

60

3

40

Beam peak, q

Directivity, dB

Beam peak

Directivity 2

20

1

0.5

1.0

1.5

2.0

0 3

2.5

Dipole length, l

FIGURE 5-3 Dipole (slot) directivity and beam peak versus length.

portion along the axis. The equal positive and negative portions of the standing-wave current sum to zero and produce a pattern null normal to the axis. By integrating Eqs. (2-5) and (2-10), we compute far ﬁelds for radiators centered on the z-axis through the far-ﬁeld conversion [Eqs. (2-1) and (2-9)] [1, p. 82]: Eθ = j η

I0 −j kr cos(kL/2 cos θ ) − cos(kL/2) e 2πr sin θ

dipole

(5-2)

where L is the total dipole length. Using the Y = 0 plane as the slot ground plane, the far-ﬁeld magnetic ﬁeld is found as Hθ =

±j V0 −j kr cos(kL/2 cos θ ) − cos(kL/2) e η2πr sin θ

slot

(5-3)

where L is the total slot length. We apply the upper sign for Y > 0 and the lower sign for Y < 0. The electric ﬁeld of the slot is found from Eφ = −ηHθ . Equations (5-2) and (5-3) have the same pattern shape and directivity. We integrate the magnitude squared of Eqs. (5-2) and (5-3) to determine the average radiation intensity. Joined with the maximum radiation intensity, we calculate directivity (Figure 5-3) versus length. 5-2 RADIATION RESISTANCE (CONDUCTANCE) The far-ﬁeld power densities, Poynting vectors, are given by 2 |Eθ | dipole Sr = η slot |Hθ |2 η

221

RADIATION RESISTANCE (CONDUCTANCE)

where η is the impedance of free space (376.7 ). When these are integrated over the radiation sphere to compute the power radiated, the results contain either |I0 |2 (dipole) or |V0 |2 (slot), the maximum sinusoidal current (voltage). We deﬁne the radiation resistance (conductance) as Pr |I0 |2 Pr Gr = |V0 |2 Rr =

dipole (5-4) slot

Figure 5-4 is a plot of the radiation resistance of each versus length [2, p. 157]. The input resistance differs from the radiation resistance because it is the ratio of the input current (voltage) to the power radiated: kL 2 kL Vi = V0 sin 2 Ii = I0 sin

dipole (5-5) slot

Combining Eqs. (5-4) and (5-5), we ﬁnd that Rr sin (kL/2)

Ri =

2

Gr Gi = 2 sin (kL/2)

dipole (5-6) slot

500 Dipole input

Resistance, Ω

400

Slot

300 Dipole

200

100

0

Slot input

0.5

1

1.5 Length, λ

2

2.5

FIGURE 5-4 Dipole and slot radiation and center-fed input resistances.

3

222

DIPOLES, SLOTS, AND LOOPS

The input resistances (Figure 5-4) differ from the radiation resistances by Eq. (5-6). The input resistance of a one-wavelength dipole is large but not inﬁnite, as shown; it depends greatly on the diameter and input region. If we take the product of the radiation or input resistances, we determine that Rdipole Rslot =

η2 4

(5-7)

one of the consequences of the Babinet–Booker principle [3]. The input resistance depends on the current at the input [Eq. (5-6)]. When the standing-wave current is high and the voltage is low, the input resistance is moderate. A center-fed half-wavelength dipole has the same input resistance as radiation resistance, since the current maximum occurs as the input. On the other hand, a center-fed halfwavelength slot has a current minimum (voltage maximum) at its input, which gives it high input resistance. When both are a full wavelength long, the dipole standingwave current is at a minimum and the slot standing-wave current is at a maximum (Figure 5-2). The dipole has a high input resistance and the slot has a low input resistance. We can lower the input resistance by feeding at a high current point, but we may excite a distribution different from that expected. A short dipole looks like a capacitor at the input. As the length increases, the radiation resistance grows and the capacitance decreases. Just before the length reaches λ/2, the capacitance becomes zero. The exact length at which the antenna resonates (zero reactance) depends on the diameter of the elements and the input gap. A good starting point is 95% of a half wavelength. Beyond the resonant length, the dipole becomes inductive. The impedance of a thin half-wavelength dipole is 73 + j 42.2 , whereas the resonant-length dipole resistance is about 67 . The slot looks like an inductor when short. Think of it as a short-length short-circuited shunt slotline stub. The inductance increases as its length increases and the slot resonates like the dipole, just short of λ/2. Additional resonances occur at longer lengths. Increasing the frequency is equivalent to increasing the length for the thin dipole.

5-3 BABINET–BOOKER PRINCIPLE [3; 4, p. 337] A strip dipole and a slot are complementary antennas. The solution for the slot can be found from the solution to an equivalent dipole by an interchange of the electric and magnetic ﬁelds. Not only the pattern but also the input impedance can be found. Figure 5-5 shows two such complementary structures. Babinet’s principle of optical screens (scalar ﬁelds) states that given the solutions to the diffraction patterns of a screen, Fi , and the screen’s complement, Fc , the sum equals the pattern without the screen. Booker extended Babinet’s principle to vector electromagnetic ﬁelds. Strict complementation of an electric conductor requires a nonexistent magnetic conductor. Booker solved this problem by using only perfectly conducting inﬁnitesimally thin screens and by interchanging the electric and magnetic ﬁelds between the screen and its complement. If we take two such complementary screens and perform line integrals over identical paths to compute the impedance of each, we obtain the result Z1 Zc =

η2 4

(5-8)

DIPOLES LOCATED OVER A GROUND PLANE

223

Slot

FIGURE 5-5

Complementary screens.

where Z1 is the input impedance of the structure, Zc the input impedance of the complementary structure, and η the impedance of free space (376.7 ). Equation (5-8) extends Eq. (5-7) to the total impedance and includes mutual impedances as well as self-impedances. Certain antennas, such as ﬂat spirals, are self-complementary—an exchange of the spaces and conductors leaves the structure unchanged except for rotation. For a twoarm structure, η2 Z02 = or Z0 = 188 4 Rumsey [5, p. 28] extended these ideas to antennas with more than two conductors to determine the input impedances in various feeding modes. We must relate ﬂat-strip dipoles to normal round-rod dipoles to use the available results for round dipoles. The diameter of an equivalent round rod equals one-half the strip width of the ﬂat structure. Consider a thin dipole with its near λ/2 resonance of 67 . We calculate equivalent slot impedance from Eq. (5-8): Zslot =

376.72 = 530 4(67)

A half-wavelength slot impedance is Zslot =

376.72 = 363 − j 211 4(73 + j 42.5)

The λ/2 dipole is inductive when it is longer than a resonant length, whereas the slot is capacitive.

5-4 DIPOLES LOCATED OVER A GROUND PLANE We analyze a dipole over a ground plane as a two-element array of the dipole and its image. The ground plane more than doubles the gain of the element by limiting the radiation directions. We can expect a change in the input impedance as the dipole interacts with its image. A vertical dipole excites currents in the ground plane, when transmitting, equivalent to its image. The image is vertical (Figure 5-6) and has the same phase as the dipole (even mode). The impedance of the dipole becomes

224

DIPOLES, SLOTS, AND LOOPS

FIGURE 5-6

Ground-plane images.

Z = Z11 + Z12 . Z12 is the mutual impedance between the dipole and its image spaced 2H , where H is the center height of the dipole over the ground plane. The array radiates its maximum in the direction of the ground plane. The dipole also radiates its maximum pattern along the ground plane given by Umax

η|I0 |2 = (2π)2

kL 1 − cos 2

2 (5-9)

where L is the dipole length. The radiated power of the single dipole is R12 Pin = R11 |I0 |2 1 + R11 The two-element array increases the ﬁeld over a single element by 2 and the radiation intensity by 4: 4Ud,max 4η[1 − cos(kL/2)]2 directivity = = Pin /4π (R11 + R12 )π We used only the power into the dipole, since no source is connected to the image. Figure 5-7 is a plot of the directivity of a vertical dipole versus height over the ground plane. A horizontal dipole and its image (Figure 5-6) form an odd-mode two-element array (Section 3-1). The input impedance of the dipole becomes Z11 − Z12 for the odd-mode array. The value of the mutual impedance Z12 approaches that of the self-impedance Z11 as the two dipoles move close together. The input impedance approaches zero as the distance from the dipole to ground plane shrinks. The input impedance of all oddmode array elements decreases as the elements approach each other. The two-element odd-mode array produces a null along the ground plane. The beam peak occurs normal to the ground plane (θ = 0◦ ) when the distance between the dipole and its image is less than λ/2 or H ≤ λ/4. The pattern bifurcates after that height is exceeded. The maximum radiation from the array is

UA,max

2πH 4 sin2 λ = 4

λ 4 λ H ≥ 4

H ≤

θmax = cos−1

λ 4H

DIPOLE MOUNTED OVER FINITE GROUND PLANES

225

9 Horizontal

Directivity, dB

8.5 Vertical

8.0

7.5

7.0

6.5

0

FIGURE 5-7

0.5

1.0 Height above ground plane, l

1.5

2

Directivity of half-wavelength dipoles over a ground plane.

The dipole pattern [Eq. (5-2)] increases the radiation intensity. The total input power into the single dipole becomes Pin = |I0 |2 (R11 − R12 ) UA,max Ud,max directivity = Pin /4π After inserting the various terms, we obtain the directivity of a horizontal dipole over ground: 4η sin2 (2πH /λ)[1 − cos(kL/2)]2 π(R11 − R12 ) directivity = 2 4η[1 − cos(kL/2)] π(R11 − R12 )

λ 4 λ H ≥ 4

H ≤

Its plot is included in Figure 5-7.

5-5 DIPOLE MOUNTED OVER FINITE GROUND PLANES Most conﬁgurations have a dipole mounted over a ﬁnite ground plane. You can calculate the ﬁnal pattern by using GTD, PO, or MOM, or you can measure the pattern using the actual ground plane. Analyses produce idealized patterns, and measurements contain errors due to the presence of the positioner mounting. If the ﬁnal system requires

226

DIPOLES, SLOTS, AND LOOPS

E-Plane

10l 2l Dia H-Plane

l Diameter Disk

FIGURE 5-8 Dipole spaced λ/4 over disk ground planes with 1λ, 2λ, and 10λ diameters.

exacting patterns, it has no margin and will fail. In this section we consider dipoles with idealized ground planes and give you ideas about the ﬁnal performance or spur you to use the ground plane purposely as a design parameter. Figure 5-8 shows the result of a PO analysis of a dipole mounted λ/4 above ﬁnite disks 1, 2, and 10λ in diameter. The E-plane pattern contains a pattern null at 90◦ due to the dipole pattern. The ground plane restricts the broad H -plane pattern as pattern angles approach 90◦ and reduces the backlobe more and more as it increases in size. At one wavelength the disk increases the gain of the antenna from the 7.5 dB given in Figure 5-7 to 8.1 dB. We can size the ground plane to produce small gain increases. We can analyze ﬂat-plate reﬂectors from three perspectives. In the ﬁrst, plates restrict radiation directions and thereby increase directivity. Waves polarized parallel with the surface must vanish on the reﬂector surface and cause a greater restriction of the beam. We see this effect in Figure 5-7, which shows horizontal dipoles having greater directivities than vertical dipoles for close spacing over a ground plane. In the second method we use aperture theory to analyze the reﬂector by using an aperture plane and integrate the ﬁelds or evaluate illumination losses. If the phase of the ﬁelds on the aperture varies rapidly, we must either take ﬁne increments in numerical integration or evaluate only around areas of stationary phase. Third, we can replace the reﬂector with images and restrict the valid pattern region. In GTD this method is combined with diffractions to smooth the ﬁeld across shadow and reﬂection boundaries. In Section 5-4 we analyzed the pattern and gain of a dipole mounted over an inﬁnite ground plane by the method of images. The antenna and its image formed a two-element

DIPOLE MOUNTED OVER FINITE GROUND PLANES

227

TABLE 5-1 Results of a GTD Analysis of a Horizontal λ/2 Dipole λ/4 Over a Limited Square Ground Plane (H -Plane) Ground-Plane Size (λ)

Directivity (dB)

Front-to-Back Ratio (dB)

H -Plane Pattern Level at 90◦ (dB)

H -Plane Beamwidth (deg)

Phase Center (λ)

0.5 0.6 0.7 0.8 0.9 1.0 1.2 1.4 1.6 1.8 2.0 2.5 3.0 4.0 5.0 10

5.37 6.32 7.08 7.68 8.14 8.34 8.65 8.45 7.96 7.39 6.95 7.13 7.74 7.28 7.56 7.41

8.4 10.3 12.0 13.5 14.8 16.0 17.8 19.1 20.0 21.1 22.3 25.0 28.3 32.8 35.4 36

−6.3 −7.6 −8.8 −9.8 −10.6 −11.2 −12.0 −12.3 −12.2 −12.3 −12.4 −12.7 −13.8 −14.8 −16.2 −19.1

108.5 104.0 100.9 97.8 95.1 93.2 93.3 99.4 108.4 112.4 113.1 115.8 111.4 116.1 118.0 121.3

0.18 0.15 0.14 0.12 0.10 0.08 0.04 0.01 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0

array, but with real power into only one element. The imaging method gives limited information that can be ﬁlled with GTD methods. Table 5-1 lists the results of a GTD analysis of a half-wavelength horizontal dipole located λ/4 over a limited square ground plane. An inﬁnite ground plane and dipole combination has an inﬁnite front-toback (F/B) ratio with the ﬁelds vanishing in the ground-plane direction. By using the methods of Section 3-3, we calculate a 120◦ half-power beamwidth for the two-element half-wavelength spaced array of the dipole and its image. The F/B ratio increases as the reﬂector (ground plane) size increases. Unfortunately, F/B is only the ratio of two pattern angles. We could tune the size of the ground plane to produce a high F/B ratio for a nonsquare ground plane, but it holds for only a small range of angles. Figure 5-8 illustrates the general increase in F/B as the size of the ground plane increases. We expect zero ﬁelds at θ = 90◦ on an inﬁnite ground plane, and Table 5-1 shows a decrease of the ﬁelds with an increase of the ground plane. The half-power beamwidth cycles about 120◦ as the ground plane increases in size. Phase center is the apparent radiation center placed at the focus of a paraboloidal reﬂector when used as a feed. The phase center of the equivalent two-element array is located on the ground plane. As we decrease the ground plane, the effect of the image decreases and causes the phase center to move toward the dipole. In the limit of no ground plane, the phase center is on the dipole. Table 5-1 shows the small gain changes that occur as the relative phase of the ground-plane scattered ﬁelds and the dipole direct ﬁelds add in the far ﬁeld. The small ground plane at λ/2 square fails to signiﬁcantly limit radiation and gain drops. Peak gain occurs when the ground plane is 1.2λ square, but this result would not necessarily hold for a circular ground plane. In most applications the dipole cannot be mounted directly above the ground-plane center, but we can add a small ground plane to control

228

DIPOLES, SLOTS, AND LOOPS

H-Plane, 1l dia.

E-Plane, 10l dia. E-Plane, 1l dia.

H-Plane, 1l dia.

FIGURE 5-9 V-dipole spaced 0.35λ over and tilted 35◦ toward 1λ- and 10λ-diameter ground planes.

the pattern and then place the combination on a pedestal over the larger ground plane. Most cases should be analyzed or measured in the ﬁnal conﬁguration. The dipole E-plane pattern null can be reduced by tilting the two poles down toward the ground plane. Figure 5-9 illustrates the calculated pattern of a tilted element dipole above a ﬁnite disk ground plane. The feed point of the dipole has been raised to 0.35λ to allow for the 35◦ tilt to the poles. Tilt and ground-plane height give additional parameters to control the pattern of the dipole mounted over a ﬁnite ground plane. For example, a horizontal dipole located λ/2 over an inﬁnite ground plane forms an odd-mode (0◦ , 180◦ ) two-element array using the dipole and its image. The simple ray-tracing argument given in Section 3-1 predicts a pattern null at zenith. But when placed over a ﬁnite ground plane, the fainter image fails to produce a complete null. We sometimes mount a dipole spaced away from a metal cylinder that provides a ground plane to restrict radiation. The curved ground plane allows greater radiation around the cylinder when rays spread as they scatter from it. Figure 5-10 shows the horizontal plane pattern for a vertical dipole mounted near a 1λ-diameter cylinder for spacing of 0.25λ, 0.4λ, 0.5λ, and 0.75λ. When we space a dipole λ/2 above a large ﬂat ground plane, the pattern has a null normal to the plane. The cylinder is unable to generate a full image of the dipole to produce this null, but the pattern does dip 11.2 dB from the peak. A dipole spaced 3λ/4 over a ground plane produces a threelobed pattern that we can see in Figure 5-10 except that the cylinder can produce only 8-dB dips. If we mount the dipole over a 2λ-diameter cylinder, the pattern is similar to Figure 5-10 except that F/B increases and the nulls have greater depths. Table 5-2 summarizes pattern results for vertical dipoles mounted over small cylinders.

DIPOLE MOUNTED OVER FINITE GROUND PLANES

229

0.4l l/2 3l/4

l/4 Spacing

FIGURE 5-10 Horizontal plane pattern for a vertical dipole mounted near a 1λ-diameter cylinder at 0.25λ, 0.4λ, 0.5λ, and 0.75λ distances.

TABLE 5-2

Dipole Mounted Over a Cylinder Aligned with a Cylinder Gain (dB)

Height (λ) Over Cylinder

Cylinder Diameter (λ)

At 0

0.25

0.25 0.50 1.0 2.0 0.25 0.50 1.0 2.0 0.25 0.50 1.0 2.0 0.25 0.50 1.0 2.0

3.5 6.1 6.7 7.3 3.2 3.6 2.2 2.2 0.5 −2.9 −5.9 −8.7 3.3 5.0 5.1 6.4

0.4

0.5

0.75

◦

At 180◦

Gain Peak

Peak Angle

−2.1 −2.7 −6.1 −10.7 0.3 −1.3 −5.3 −9.7 0.9 −1.8 −4.2 −8.5 −0.2 −0.9 −3.2 −6.8

3.6 6.1 6.7 7.3 4.9 6.0 5.1 6.0 5.2 4.8 5.2 5.9 3.4 4.7 4.6 5.2

0 0 0 0 64 62 60 54 80 80 76 70 102 102 98 90

230

DIPOLES, SLOTS, AND LOOPS

l/4

0.4l l/2 3l/4 Spacing

FIGURE 5-11 Horizontal dipole mounted over a vertical 1λ diameter pole at 0.25λ, 0.4λ, 0.5λ, and 0.75λ distances.

TABLE 5-3 Dipole Mounted Over a Cylinder Perpendicular to a Cylinder Gain (dB) Height (λ) Cylinder Gain Peak Peak Angle Over Cylinder Diameter (λ) At 0◦ At 180◦ Perpendicular to Plane Perpendicular to Plane 0.25

0.4

0.5

0.75

0.25 0.50 1.0 2.0 0.25 0.50 1.0 2.0 0.25 0.50 1.0 2.0 0.25 0.50 1.0 2.0

3.8 6.6 7.1 7.0 3.0 2.9 0.6 1.2 0.0 −4.2 −5.6 −6.7 3.5 5.3 5.7 6.4

−2.6 −4.4 −8.2 −8.6 −0.2 −2.5 −7.1 −7.3 0.5 −2.8 −5.7 −6.3 −0.4 −1.6 −4.3 −5.7

3.8 6.6 7.1 7.4 4.9 6.5 6.4 7.3 5.1 5.2 6.6 7.3 2.7 4.3 5.1 5.8

0 0 0 30 46 48 50 54 54 54 56 60 66 66 66 70

CROSSED DIPOLES FOR CIRCULAR POLARIZATION

231

To complete the analysis, the dipole was rotated so that its axis is perpendicular to the pole. Figure 5-11 illustrates the patterns calculated for a horizontally polarized dipole mounted above a vertical pole. We expect a pattern null at 90◦ in this horizontal plane due to the dipole polarization null, but the dipole induces curved currents on the cylinder that radiate and ﬁll in these nulls. The null due to the dipole does narrow the pattern in the horizontal plane compared to Figure 5-10, and in many cases peak radiation occurs in the vertical plane. Table 5-3 lists the characteristics of the horizontal dipole mounted over a vertical pole for various dipole spacing above the pole and its diameter.

5-6 CROSSED DIPOLES FOR CIRCULAR POLARIZATION We produce a circularly polarized antenna by placing two dipoles along the x- and y-axes over a ground plane and feeding them with equal amplitudes and quadrature phase (0◦ and −90◦ for RHC). Without the ground plane the combination radiates LHC in the −z direction. The ground plane changes the sense of circular polarization of the wave radiated in the −z-direction and it adds with the direct radiated wave. The dipoles are fed from either dual folded baluns that produce two separate inputs or by a split coax balun connecting both dipoles in shunt. The shunt connection requires differing lengths for the dipoles to produce the 90◦ phase difference that we call the turnstile conﬁguration. The dual-feed antenna uses either a quadrature hybrid equal-amplitude power divider to feed the two ports or an equal phase and amplitude power divider with an extra line length on one of the two ports. The hybrid power divider feed produces an antenna with a wide impedance and axial ratio bandwidth. The hybrid power divider has two inputs that provide ports for both RHC and LHC polarizations. The signals reﬂected from the two equal-length dipoles when fed from one port of the hybrid reﬂect into the second port due to the phasing in the hybrid coupler. When measuring at one port of the hybrid, the impedance bandwidth is quite broad because the reﬂected power is dissipated in the load on the other port. This dissipated power lowers the efﬁciency of the antenna, a hidden loss unless you measure the coupling between the inputs of the hybrid. The second conﬁguration, using the extra line length, produces an antenna with a narrowed axial ratio bandwidth and a wider impedance bandwidth compared to a single dipole. The extra 180◦ round-trip total signal path in one arm causes the equal reﬂections to cancel. Figure 5-12 gives the circularly polarized pattern from a pair of crossed dipoles over a ground plane with a perfect feed. The E-plane dipole null limits the angular range of good circular polarization. We improve the circular polarization by raising the dipoles a little and tilting them down to widen the E-plane beamwidth. Figure 5-12 shows the pattern for the tilted dipole pair and illustrates the improved cross polarization and the wider beamwidth. The placement on a ﬁnite ground plane complicates this result somewhat and will require extra design effort. Turnstile feeding exploits the impedance properties of the dipole to shift the relative phase between two different dipoles when shunt connected to the same port. When we shorten a dipole below resonance, its impedance is capacitive and its current has positive phase relative to the resonant-length dipole, while the lengthened dipole has an inductive reactance and a negatively phased current. We determine the lengths of

232

DIPOLES, SLOTS, AND LOOPS

FIGURE 5-12 Crossed dipoles fed for circular polarization: (a) λ/4 height and 0◦ tilt; (b) 0.3λ height and 30◦ tilt.

the two dipoles by a perturbation technique using the Q of the resonant circuit of the dipole. Q is related to the VSWR bandwidth: BW =

VSWR − 1 √ Q VSWR

Q=

VSWR − 1 √ BW VSWR

(5-10)

We derive the lengths of the two dipoles in terms of the resonant (zero reactance)-length dipole, L0 : L0 Lx = √ 1 + 1/Q

Ly = L0 1 +

1 Q

RHC polarization

(5-11)

A dipole of 0.014λ diameter located 0.3λ above a ground plane and tilted down 30◦ has a resonant length of 0.449λ. The 2 : 1 VSWR bandwidth for 70 is 18.3% or a Q of 3.863 by using Eq. (5-10). When we insert this Q in Eq. (5-11), we calculate the two lengths for a turnstile design: Lx = 0.400λ and Ly = 0.504λ for RHC polarization. The +x and +y poles are fed from the same port. Figure 5-13 plots the Smith chart of this design. The trace on a Smith chart rotates clockwise for increasing frequency. The cusp in the trace is the frequency with the best axial ratio, which did not occur at the frequency of best match. Nevertheless, the 2 : 1 VSWR bandwidth of the antenna has increased to 41.5% because the combined reactance of the two dipoles cancels over a large frequency range. At center frequency the pattern is similar to Figure 5-12 except that the patterns in the two planes have slightly different beamwidths due to the dipole lengths. When the frequency shifts off center, the axial ratio degrades. The axial ratio bandwidth is far less than the impedance bandwidth, and the design gives a 16.4% 6-dB axial ratio bandwidth. An axial ratio of 6 dB produces 0.5-dB polarization loss similar to the 0.5 dB reﬂected power loss of 2 : 1 VSWR. This illustrates the importance of considering not only the impedance bandwidth but also the pattern characteristics over the frequency band. We can increase the beamwidth of the turnstile dipole located over a ground plane by adding a notched cone under it. Figure 5-14 illustrates the arrangement of the slightly

CROSSED DIPOLES FOR CIRCULAR POLARIZATION

233

Finish

Best Axial Ratio

Start

FIGURE 5-13 Smith chart response of a turnstile dipole pair Lx = 0.400λ and Ly = 0.504λ mounted 0.30λ over a ground plane with 30◦ tilt.

FIGURE 5-14 Turnstile dipole mounted over a notched cone on a ﬁnite circular ground plane with radial line chokes to reduce the backlobe.

less than λ/4-long notches in a 45◦ cone with the turnstile dipoles located about λ/4 above the ground plane. A split-tube coaxial balun feeds the two dipoles sized as a turnstile with dipoles of longer and shorter length. The upper feed jumper excites RHC radiation. The dipoles excite magnetic currents in the slots that radiate a broad pattern to ﬁll in the E-plane nulls of the dipoles. On an inﬁnite ground plane the horizontal

234

DIPOLES, SLOTS, AND LOOPS

FIGURE 5-15 Pattern of a turnstile dipole mounted over a notched cone with a 0.75λ ground plane and two radial chokes.

polarization must vanish along the ground plane, and the RHC and LHC components would be equal at 90◦ similar to the pattern shown in Figure 5-12. By using a ﬁnitesize ground plane, the horizontal component does not vanish, and a wide beamwidth is obtained with circular polarization at 90◦ as shown in Figure 5-15, which uses a 0.75λdiameter ground plane and 0.5λ-base-diameter cone. To reduce the backlobe below the ground plane, two short-circuited radial transmission line chokes were placed around the edge to form a soft surface. We size the inner radius so that the transmission line produces an open-circuit impedance at the outer rim that reduces the edge diffraction and the backlobe [6, p. 88]. From a PO perspective the radial line choke is a slot that supports a magnetic current loop. This example illustrates that slots or notches can be used to shape the patterns of small antennas.

5-7 SUPER TURNSTILE OR BATWING ANTENNA [7] The super turnstile or batwing antenna was developed for TV transmitter antennas. The antenna combines a slot with a dipole batwing to produce an antenna with a wide impedance bandwidth. Figure 5-16 shows the normal conﬁguration, with four wings placed around a central support metal mast. Each wing connects to the mast at the top and bottom with a metal-to-metal connection. The inner vertical rod and the support

SUPER TURNSTILE OR BATWING ANTENNA

235

FIGURE 5-16 Super turnstile or batwing antenna using an open rod construction.

V-Pol. H-Pol.

FIGURE 5-17 Elevation pattern of a single bay of a super turnstile antenna showing horizontal and vertical polarization components.

236

DIPOLES, SLOTS, AND LOOPS

Reflection, dB

mast form a two-line slot fed by a jumper located at the center of each wing. To produce an omnidirectional pattern about the mast, a feed power divider located inside the mast phases the inputs for circular polarization (0◦ , 90◦ , 180◦ , 270◦ ). The antenna radiates horizontal polarization in the horizontal plane but radiates cross-polarization that increases with elevation (depression) angle as shown in Figure 5-17. A four-wing antenna produces a horizontal plane pattern ripple of about 1.5 dB. Adding more wing antennas around a larger central mast reduces the ripple. The extraordinary characteristic of the antenna is its impedance bandwidth. Figure 5-18 gives the return loss frequency response for a wire frame antenna. The 1.1 : 1 VSWR bandwidth is about 35%; if adjusted to 1.25 : 1 VSWR, the antenna has a 51% bandwidth. You make small adjustments to the spacing between the mast and the inner rod to tune the VSWR. Table 5-4 lists the parameters of batwing antennas with both wire frame and solid panel wings. The solid panels lower the input impedance to 75 from the 100 of the wire frame antenna. Using an antenna with only two wings

Adjusted for 1.25:1 VSWR

Adjusted for 1.1:1 VSWR

Normalized Frequency

FIGURE 5-18 Super turnstile wire frame antenna return-loss response adjusted for 1.1 : 1 and 1.25 : 1 VSWR. TABLE 5-4 Dimensions of a Super Turnstile Antenna in Wavelengths for Four Wings Center-Fed for Circular Polarization Parameter

Wire Frame

Impedance () Height Wing upper Wing middle Gap Rod diameter Mast diameter

100 0.637 0.2254 0.0830 0.0169 0.0508 0.0847

Solid Wing 75 0.637 0.229 0.0847 0.0216 0.0847

CORNER REFLECTOR

237

changes the input impedance from the value for an antenna with four wings because the close coupling between the wings alters the impedance. It depends on the feeding mode. This holds for any antenna with close coupling: for example, a spiral antenna. You must feed it in the operating mode to measure the correct input impedance. The transmitter antenna will consist of a number of these antennas stacked vertically to produce a narrow pattern directed at the horizon. 5-8 CORNER REFLECTOR [8, p. 328] The usual corner reﬂector (Figure 5-19) has a dipole located between two ﬂat plates that limit directions of radiation. The angle between the reﬂectors can be any value, but 90◦ seems to be the most effective. On paper, decreased angles give better results, but only marginally. We could consider the ﬂat plate as a limiting case. The tangential electric ﬁelds must vanish at the surface of the ﬂat plates. We discover a greater restriction, since the ﬁelds can only decrease gradually in the limited space between the ground planes and the dipole. Most of the power is concentrated in lower-order spherical modes. In the limit of zero vertex distance, the single mode possible restricts the beamwidth to 45◦ in the H -plane. We analyze the 90◦ corner reﬂector as an array by using the three images of the dipole in the ground planes (Figure 5-19) plus the real dipole. The array factor of the array of dipole and images is 2 j kd cos θ e + e−j kd cos θ − (ej kd sin θ sin φ + e−j kd sin θ sin φ )

Length

d d d Height

Images

FIGURE 5-19 A 90◦ corner reﬂector.

238

DIPOLES, SLOTS, AND LOOPS

In the H -plane, φ = 90◦ and we evaluate terms to get 4[cos(kd cos θ ) − cos(kd sin θ )]2 where d is the distance from the vertex to the dipole and θ is the H -plane pattern angle from the axis. We must multiply this by the pattern of the dipole to obtain the radiation intensity. We consider only the H -plane, where the maximum radiation intensity is found from Eq. (5-9): U = 4[cos(kd cos θ ) − cos(kd sin θ )]2

η|I0 |2 (2π)2

1 − cos

kL 2

2 (5-12)

where η is the impedance of free space, I0 the dipole current, and L the dipole length. The radiated power of the single dipole is √ (5-13) Pin = |I0 |2 [R11 + R12 (2d) − 2R12 ( 2 d)] where R11 is the self-resistance of the dipole and R12 (x) is the mutual resistance function between the dipole and its images. The directivity is found from directivity(θ ) =

4πU (θ ) Pin

(5-14)

We combine Eqs. (5-12) and (5-13) into Eq. (5-14) to compute directivity of the 90◦ corner reﬂector with inﬁnite sides: directivity(θ ) =

4η[1 − cos(kL/2)]2 [cos(kd cos θ ) − cos(kd sin θ )]2 √ R11 + R12 (2d) − 2R12 ( 2 d)

(5-15)

Table 5-5 gives the directivity, beamwidth, and impedance of a 90◦ corner reﬂector fed from a dipole 0.42λ long and 0.02λ in diameter. We must shorten the dipole further than a free-space dipole length at resonance to compensate for the mutual coupling between dipoles. Directivity increases as the vertex distance decreases, but the effects of superdirectivity cause the efﬁciency and gain to fall as the vertex is approached. The antenna has a 50- input impedance for d = 0.37λ. This point shifts when we increase the dipole’s diameter to increase its bandwidth. Kraus gives the following guidelines for the size of the sides. Each plate should be at least twice the length of the dipole-to-vertex distance, and the plate height (the dipole direction) should be at least 0.6λ. To evaluate those guidelines, a GTD analysis was performed on various combinations (Table 5-6) with d = 0.37λ. The H -plane beamwidth decreases with an increase in plate length. After about 1.5λ sides, the H -plane beamwidth ﬂuctuates about 45◦ as the sides increase. Even with 5λ sides the beamwidth is below 45◦ . The E-plane beamwidth ﬂuctuates with the plate height. The directivity was estimated from the beamwidths. In one case—1.5λ sides and 1.5λ high—the estimated directivity exceeds the directivity of the inﬁnite-side case. The edge diffractions add to the reﬂected and direct radiation of the rest of the antenna. Refer to Section 2-4.2 for an example using PO to analyze a corner reﬂector. Similar to inﬁnite plate analysis, the reaction of the image dipoles in the ﬁnite plates can be used to ﬁnd the input impedance and gain of the antenna. When we analyze the corner

CORNER REFLECTOR

239

TABLE 5-5 Characteristics of a 90◦ Corner Reﬂector with Inﬁnite Sides and 0.42λ Dipole Vertex Directivity Beamwidth Input Impedance Distance (λ) (dB) (deg) () 0.30 0.32 0.34 0.36 0.37 0.38 0.40 0.42 0.44 0.46 0.48 0.50 0.52 0.54 0.56 0.58 0.60

12.0 12.0 11.9 11.9 11.9 11.8 11.8 11.7 11.7 11.6 11.5 11.4 11.4 11.3 11.2 11.1 10.9

44.7 44.6 44.5 44.3 44.2 44.1 43.9 43.6 43.3 42.9 42.4 41.8 41.1 40.2 39.2 38.1 36.8

29.1 − j 1.1 34.9 + j 0.4 40.9 + j 1.1 47.0 + j 0.8 50.0 + j 0.3 53.0 − j 0.5 58.8 − j 2.8 64.1 − j 6.0 68.8 − j 10.0 72.7 − j 14.9 75.7 − j 20.3 77.7 − j 26.2 78.6 − j 32.2 78.4 − j 38.4 77.0 − j 44.3 74.6 − j 49.8 71.3 − j 54.8

TABLE 5-6 Results of a GTD Analysis of a 90◦ Corner Reﬂector with Finite Sides and Vertex Distance 0.37λ Beamwidth

Side Length (λ)

Plate Height (λ)

E-Plane

H -Plane

F/B (dB)

Estimated Directivity (dB)

0.75 1.00 1.50 0.75 1.00 1.50 0.75 1.00 1.50 5.00

0.75 0.75 0.75 1.00 1.00 1.00 1.50 1.50 1.50 5.00

70.4 73.6 72.6 60.2 61.0 58.5 53.4 51.6 48.2 68.8

97.4 72.4 50.8 91.6 62.8 46.0 81.6 60.0 42.6 43.4

18.4 17.3 18.2 23.4 22.7 23.8 34.0 39.0 46.3 63.5

7.7 8.8 10.0 8.5 10.1 11.4 9.3 11.0 12.6 10.8

reﬂector using GTD, the method does not determine input impedance and gain must be estimated from the patterns. We can use the method of moments to analyze the corner reﬂector. One preferred construction method is to use rods for the reﬂector so that the antenna has minimum wind loading. Figure 5-20 illustrates a corner reﬂector made with only six rods on each side. Figure 5-21 gives the pattern of this antenna from a moment method calculation. This small antenna produces excellent results. We can use the angle of the sides as a design parameter. A geometric optics analysis that uses images restricts the angle, but nothing stops the antenna from working for

240

DIPOLES, SLOTS, AND LOOPS

FIGURE 5-20 Corner reﬂector constructed from 0.6λ-long rods spaced 1/6λ with a dipole spaced 0.37λ from the vertex.

E-Plane H-Plane

FIGURE 5-21 Pattern of a corner reﬂector made from 0.6λ-long rods spaced 1/6λ with dipole 0.37λ from vertex.

241

CORNER REFLECTOR

TABLE 5-7 Corner Reﬂector with Varying Angle H -Plane 0.9λ-Wide Plates Connected to a 0.2λ Central Plate, 1λ E -Plane Width, Dipole 0.3λ Above the Central Plate Beamwidth

Side Angle

E-Plane

60 55 50 45 40 35 30 25 20

58.1 57.1 56.3 55.8 55.4 55.4 55.6 56.2 57.4

Beamwidth

H -Plane

Gain (dB)

F/B (dB)

Side Angle

E-Plane

H -Plane

Gain (dB)

F/B (dB)

59.4 56.0 52.3 49.1 46.8 45.5 45.6 47.9 53.7

9.1 9.9 10.5 10.8 11.1 11.2 11.0 10.7 10.2

21.7 22.6 23.0 23.4 22.8 24.3 24.9 25.5 20.3

15 10 5 0 −5 −10 −15 −20 −25

59.2 61.6 64.7 67.8 70.2 71.5 71.8 71.7 71.8

65.9 83.3 99.2 108.6 117.0 125.8 135 143.8 152.2

9.5 8.7 7.8 7.5 7.2 6.9 6.6 6.3 5.4

19.3 26.3 25.1 23.6 22.0 16.5 19.0 17.8 16.1

Dipole 0.9l Side Plate Hinge 0.2l Center Plate

FIGURE 5-22

Corner reﬂector with variable-angle side plates and a center ﬂat plate.

arbitrary side angles. It is convenient to have a small plate between the tilted sides for the mounting brackets, and these side plates could be mounted on hinges and rotated to vary the H -plane beamwidth. Table 5-7 lists the parameters of a corner reﬂector 1λ along the E-plane, a central plate 0.2λ wide in the H -plane, and sides 0.9λ long where the side angle is varied. The dipole is located 0.3λ above the central plate. We measure the side plate angle from the plane containing the small central ground plane; zero corresponds to a ﬂat plane ground plane and 45◦ the usual corner reﬂector. Negative side-plate angle means that the side plates are tilted behind the central plate away from the dipole. Figure 5-22 illustrates the H -plane cross section of this corner reﬂector with 30◦ side plates. We should not design corner reﬂectors with large sides since the gain is limited. The gain of paraboloid reﬂectors of the same size soon exceeds that of a corner reﬂector. A 2λ-diameter paraboloid reﬂector at 50% efﬁciency has a gain of 13 dB, and its gain exceeds that of a corner reﬂector. Any corner reﬂector with a vertex angle given by 180◦ /N , where N is an integer, can be analyzed by the method of images. Corner reﬂectors with N greater than 2 have only marginally higher gains. The 90◦ corner reﬂector gives the best result for a given amount of material. Elkamchouchi [9] adds a cylindrical surface between the plates centered on the vertex. This surface adds another set of images within the cylinder. The images increase the gain by about 2 dB and decrease the frequency dependence of the impedance.

242

DIPOLES, SLOTS, AND LOOPS

10l dia. Disk

2l dia.

1l dia.

FIGURE 5-23 A λ/4 monopole located on 1λ-, 2λ-, and 10λ-diameter disk ground planes.

5-9 MONOPOLE A monopole consists of a single conductor fed out of a ground plane from the center conductor of a coax. When we include its image (Figure 5-6), the monopole equates to a dipole for analysis. The ﬁelds vanish below the ground plane and restricting the ﬁelds to the upper hemisphere doubles the gain over a dipole, since only half the input power of the dipole is needed to produce the same ﬁeld strength. The input impedance decreases to half that of the equivalent dipole. We can form the image of the voltage source feeding the monopole in the ground plane. The voltage across the input of the equivalent dipole is twice that of the monopole to produce the same current. Therefore, the impedance of the monopole is half the impedance of the dipole. The large value of edge diffraction greatly limits the F/B ratio of a monopole when it is placed on a ﬁnite ground plane. Figure 5-23 shows the pattern of a monopole when placed on 1λ-, 2λ-, and 10λ-diameter circular ground planes. The back radiation can be reduced by placing the monopole over a ground plane with circular corrugations that forms a soft surface at the edge when the corrugations are slightly deeper than λ/4 [10]. When the corrugations are less than λ/4, the ground plane can support surface waves. 5-10

SLEEVE ANTENNA [8, p. 422; 11, Chap. 5; 12; 13, p. 278]

A sleeve around the monopole (Figure 5-24) moves the virtual antenna feed up the monopole. The bandwidth increases because the current at the feed point remains nearly

SLEEVE ANTENNA

Virtual Feed Point

l/4

243

l/2

I Sleeve

I

I

Image Feed Point

Images

Coax

FIGURE 5-24 Sleeve monopole and current distributions.

constant over a wide band. Currents at the input for the case when the monopole is a quarter-wavelength long and when it is a half-wavelength long are about the same (Figure 5-24). The input resistance remains constant as the frequency changes. The sleeve shields possible radiation from the internal currents while the currents on the outside of the sleeve radiate. The pattern changes little from that of an unshielded monopole. The internal structure is available as a series-matching stub and a transformer to broadband the antenna. Design consists of adjusting the parts until a suitable compromise input impedance match is achieved over the band. Dipole sleeve antennas (Figure 5-25) require symmetrical sleeves on the arms to maintain the symmetry of the currents. It is equivalent to feeding the antenna in two places. The balun is made an integral part of the base. In both antennas, strips or rods can replace the total coaxial sleeve [14]. The currents on the rods cancel the radiation from the currents on the internal feeder. Figure 5-26 illustrates an open-sleeve dipole using two rods designed to be mounted over a ground plane. The antenna is fed from a folded balun that consists of a grounded vertical coax with one pole connected to the outer shield and a matching tube connected to the second pole. The center conductor jumps across the gap to the second pole. Following are the design dimensions in wavelengths normalized to the lower-frequency band edge: Dipole length Sleeve length Dipole-to-sleeve spacing Input taper

0.385 0.2164 0.0381 0.056

Dipole diameter Sleeve diameter Dipole height above ground

0.0214 0.0214 0.1644

Figure 5-27 plots the return-loss response of the antenna for various conﬁgurations and models of the antenna. The dipole without the sleeves has its best return loss over a narrow band centered at a normalized frequency of 1.05. The sleeves have little effect on this response at the low-frequency end. Adding sleeves produces a second resonance, which combines with the lower one to produce a broad bandwidth. An initial method of moments analysis used constant-diameter rods for the antenna, and Figure 5-27 shows the poor impedance match response of the antenna. A key element of the experimental

244

DIPOLES, SLOTS, AND LOOPS

Sleeve Dipole with Folded Balun

Sleeve Dipole

Type II Sleeve Balun

FIGURE 5-25 Sleeve dipoles.

FIGURE 5-26 Open-sleeve dipole with conical input taper.

(a) Single Dipole

Reflection, dB

(b) With Sleeves, No Taper (d) Large Ground (e) 1l Ground Plane

(c) Free Space with Sleeves

Normalized Frequency

FIGURE 5-27 Return-loss response of an open sleeve dipole: (a) dipole without sleeves; (b) open sleeve antenna; (c) open sleeve antenna with tapered input; (d) open sleeve antenna with tapered input located λ/4 over ground plane; (e) open sleeve antenna over a 1λ-diameter ground plane.

CAVITY-MOUNTED DIPOLE ANTENNA

245

antenna is the tapered input. Adding this feature to the model produced the improved broad response of the experimental antenna. The constant-diameter model response shows a notable capacitive term on a Smith chart, and the tapered input produced the necessary inductance to reduce this effect. If the antenna is located in free space, the impedance response improves as shown. Figure 5-27 points out the importance of analyzing an antenna in its operating environment. The dot-dashed curve illustrates the response when the antenna was mounted over a one-wavelength-square ground plane. The ﬁnite ground plane produces a small but noticeable change in the input impedance. The effects of small changes in the analytical model warn us that we cannot expect antennas to match their models exactly and that small mechanical details can be used to improve performance. An open-sleeve antenna can be made using a wire cage. Since the diameters of the dipole and sleeve rods are large, the weight can be reduced by using a circular array of wires for each conductor. The effective diameter of the cage, deff , is given as deff = d

nd0 d

1/n or

1 d0 = d n

deff d

n

(5-16)

The diameter of the individual wires is d0 , the cage diameter is d, and n is the number of wires.

5-11 CAVITY-MOUNTED DIPOLE ANTENNA A dipole can be placed in a cup, and the assembly can be ﬂush-mounted in a ground plane. The antenna shown in Figure 5-28 has disk sleeves located above and below the dipoles to stretch the bandwidth over a 1.8 : 1 range [15]. Following are the dimensions normalized to the dipole length: D = 2.57 L T = 0.68 L

H = 0.070 L G = 0.40 L

S = 0.505 L

The operating range is 0.416λ ≤ L ≤ 0.74λ. The antenna cavity ranged from 0.28λ to 0.50λ deep and can no longer be considered thin. The cup antenna has a nearly constant gain (±0.5 dB) of 10.5 dB over the band. Mounting the antenna in a cavity opens up new possibilities, because extra parameters are added to the design. At the low-frequency end, the cavity diameter is 1.07λ, which grows to 1.90λ at the high end. We can use a dipole in a cup as a reﬂector feed. Excellent pattern and impedance response is obtained with the dipole mounted in a truncated cone cup with a 0.88λ aperture diameter, a 0.57λ-diameter base, and a 0.44λ depth [16, pp. 106–108]. The dipole is foreshortened to 0.418λ for an element diameter of 0.013λ and mounted 0.217λ above the base to achieve a 21% 2 : 1 VSWR bandwidth for a single element. When we use a cross-polarized pair fed from a hybrid coupler to radiate CP, the impedance match at the input port improves. The signals reﬂected from the two dipoles add in phase at the isolated port and cancel at the input port. The load dissipates the reﬂected power, and the antenna through the hybrid presents an excellent impedance match.

246

DIPOLES, SLOTS, AND LOOPS

FIGURE 5-28 Cavity-mounted sleeve dipole antenna.

RHC Pol.

LHC Pol.

FIGURE 5-29 Circular polarization response of a crossed dipole mounted 0.217λ above the bottom of a truncated cone 0.44λ deep with a 0.88λ aperture and a 0.57λ base.

FOLDED DIPOLE

247

TABLE 5-8 Illumination Losses When Pattern of Figure 5-29 Feeds a Paraboloidal Reﬂector Loss (dB) f/D

Average

Maximum

0.36 0.38 0.40 0.42 0.44 0.46 0.48 0.50 0.52

1.69 1.60 1.54 1.50 1.49 1.50 1.52 1.55 1.60

1.74 1.66 1.65 1.65 1.68 1.72 1.77 1.83 1.91

Figure 5-29 plots its pattern when excited for CP. The cross-polarization is about 30 dB below the peak co-polarization response over the entire 10-dB beamwidth cone. It has the following illumination losses when the antenna is used as a paraboloidal reﬂector feed (see Section 8-2); for f/D = 0.44 and averaged over the 21% bandwidth: spillover loss = 0.72 dB

amplitude taper loss = 0.65 dB

cross-polarization loss = 0.12 dB Table 5-8 demonstrates the broad optimum reﬂector f/D for a phase center 0.02λ inside the aperture plane, where we position it at the reﬂector focus.

5-12 FOLDED DIPOLE A half-wavelength folded dipole increases the input impedance of a normal dipole fourfold while radiating the pattern of a single dipole. With the two elements closely coupled, we analyze the antenna using even and odd modes (Figure 5-30). The even mode divides the antenna into separate dipoles because the magnetic wall halfway between them is a virtual open circuit. The input current to the even mode becomes Ie =

V 2(Z11 + Z12 ) l/2

+

V/2

−

Magnetic

+ Wall

+

− V/2

=

V/2

−

Electric

+ Wall

−

+ V/2

FIGURE 5-30 Folded dipole analysis modes.

V

−

248

DIPOLES, SLOTS, AND LOOPS

where Z11 is the self-impedance of one of the dipoles and Z12 is the mutual impedance between the closely coupled dipoles. The odd mode reduces the antenna to the series connection of two nonradiating λ/4 stubs: Io =

V j Z0 tan(kL/2)

where Z0 is the characteristic impedance between the two rods. The input current is the sum of the even- and odd-mode currents. Near L = λ/2, the odd-mode current is quite small because its input impedance is an open circuit, and the input impedance is then determined by the even mode only: Zin =

V = 2(Z11 + Z12 ) Ie

For closely coupled lines, Z11 = Z12 and the input impedance becomes Zin = 4Z11 , where Z11 is the self-impedance of the dipole. Higher input impedance levels can be obtained by adding more elements. A second method of altering the step ratio from 4 is to use unequal feed and shorted element diameters [17,18]. Given a driven element radius a1 , parasitic element radius a2 , and center-to-center spacing b, Hansen [18] gives a convenient formula for the step-up ratio (1 + γ 2 ): γ =

5-13

cosh−1 [(v 2 − u2 + 1)/2v] cosh−1 [(v 2 + u2 − 1)/2uv]

where u = a2 /a1 and v = b/a1

(5-17)

SHUNT FEEDING [19, p. 118]

Shunt feeding grows out of the folded dipole. The T-match (Figure 5-31) starts as a folded dipole when the taps are at the ends. As the taps move toward the center, the impedance of the dipole dominates at ﬁrst, since the admittance of the shunt stub in the odd mode is small and the input impedance is capacitive. At some point, as the taps move toward the center, the inductive admittance of the stub will cancel the capacitive admittance of the dipole and produce antiresonance with its high input resistance. The location and magnitude of this peak resistance depends on the diameters of the rods in the T-match section and the diameter of the radiator. The input resistance decreases as we continue to move the tap point toward the center after the feed location passes the antiresonance point. The input impedance is inductive and match is achieved by using symmetrical series capacitors. The T-match is fed from a balanced line. The center short on the dipole allows the direct connection of the dipole to ground. Direct connection of broadcast towers (monopoles) to ground gives some lightning protection because the transmitter is capacitively connected to the tower. Shunt feeding with a T-match enables solid conductors, such as the skin of an aircraft, to be excited as a dipole. Horizontal shunt-fed dipoles can be connected directly to vertical towers with a metal-to-metal connection to increase the strength of the antenna to withstand adverse weather conditions. A gamma match (Figure 5-31) can be fed from an unbalanced coax line. The shield of the coax connects to the shorted center of the dipole while the center conductor

DISCONE ANTENNA

Tee

Match

249

Gamma Match

Balanced Line Coax

FIGURE 5-31 Shunt-fed dipoles.

taps into one side of the solid rod. Moving the tap away from the center increases the input resistance. The inductive reactance is series-tuned with a capacitor. Both of these connections reduce the bandwidth of the antenna as the input impedance is raised because the combination of the series capacitor and the shunt inductive stub increases the stored energy and Q of the antenna. 5-14 DISCONE ANTENNA The discone antenna (Figure 5-32) is a modiﬁcation of the dipole where the upper pole becomes a disk and the lower pole turns into a cone. We feed the antenna by locating a coax in the center of the cone and by connecting its outer shield to the lower cone at its top while we extend the coax center conductor and connect it to the disk. We obtain an antenna with a wide impedance bandwidth and a dipolelike pattern. As frequency increases the pattern peak moves toward the cone and gives a downward-pointing pattern. Figure 5-33 shows the pattern of a discone antenna at the design frequency and at two, three, and four times this frequency. The antenna produces less useful patterns as frequency increases. The antenna that gives the patterns in Figure 5-33 has a VSWR less than 3 : 1 from 1 to 10 times the design frequency. The cone upper diameter determines the high-frequency end of good impedance match. Typical slant length dimensions versus cone angle are as follows [20, pp. 128–130]: Total Cone Angle

25

35

60

70

90

Slant Length (λ)

0.318

0.290

0.285

0.305

0.335

The upper disk diameter equals 0.7 times the lower cone diameter. The spacing between the top of the cone and the upper disk equals 0.3 times the diameter of the upper cone. The diameter of the upper cone determines the upper frequency limit, but practice shows that the antenna patterns are good only over a 4 : 1 to 4.5 : 1 frequency range. The impedance bandwidth is much wider than the pattern bandwidth. To reduce weight and wind loading, the cone and disk can be made from rods, with a typical implementation having at least eight.

250

DIPOLES, SLOTS, AND LOOPS

FIGURE 5-32 Discone antenna with coaxial feed with a center conductor connected to the upper disk and a shield connected to the lower cone.

Freq. = 2

Freq. = 4 Freq. = 3

Freq. = 1

FIGURE 5-33 Elevation pattern of a 60◦ discone antenna at normalized frequencies = 1, 2, 3, and 4.

BALUNS

251

5-15 BALUNS [21; 22, pp. 167–180] A balun properly connects a balanced transmission line to an unbalanced transmission line. Simple arguments about impedances to the balanced and unbalanced modes of the three-wire transmission lines explain its operation. Considering one of the lines of a transmission line as ground misleads us. A ground plane under the transmission-line feeder becomes the third conductor of a three-wire line. Currents ﬂowing in the ground plane can unbalance the currents in the feeder. A balanced three-wire transmissionline mode carries equal and opposite currents in the feeder lines. The capacitances per unit length of the two lines to ground are the same. Coax is an example of an unbalanced line structure (Figure 5-34). The inner conductor has no direct capacitance to ground. The two-wire line shown in Figure 5-34 is a balanced line having equal capacitances to ground, but we must judge a balanced line by the currents, not just the physical structure. Before we analyze baluns, we must consider the fundamental modes of a threewire transmission line. Figure 5-35 shows circuit representations of the modes without showing the ground conductor. Equal loads terminate ports 3 and 4. The even mode applies equal voltages on ports 1 and 2 and forms a magnetic wall between the conductors where the magnetic ﬁeld vanishes to produce a virtual open circuit. The unbalanced mode—equal current directions—is associated with the even mode. Equal and opposite voltages on ports 1 and 2 form the odd mode and set up an electric wall between the conductors. The electric wall is a virtual short circuit. The odd mode excites equal and opposite currents—balanced mode—on the two lines. When the loads on ports 3 and 4 are unequal, the modes separate according to the voltages, even and odd, or the currents, unbalanced and balanced. Dipoles present loads between the lines and not to ground. Balanced line

Unbalanced line

Ground

FIGURE 5-34

+V

−V

Ground

Physically balanced and unbalanced transmission lines.

l 1

l Electrical Wall

2 l

4 3

Balanced Mode (Odd Mode)

V

V

1

Magnetic Wall

2 l

4 3

Unbalanced Mode (Even Mode)

FIGURE 5-35 Balanced and unbalanced modes on a three-wire transmission line.

252

DIPOLES, SLOTS, AND LOOPS

Unbalanced mode circuits radiate. Only closely spaced equal and opposite currents, the balanced mode, cancel the far-ﬁeld radiation from the currents on the feed lines. The radiating feeder line adds radiation components to the antenna. These components can radiate unwanted polarizations and redirect the beam peak of the antenna (squint). In reception, the unwanted currents excited on the feeder by passing electromagnetic waves reach the receiver terminals without a balun to block them. We analyze baluns by using either transmitting or receiving antennas, depending on convenience, because reciprocity applies to baluns as well as antennas. We detect balance problems from pattern squint and cross polarization. An impedance-measuring setup can detect some balance problems. Radiating unbalanced currents cause changes in the impedance. The radiation shows when the impedance changes as ﬁngers are run over the coax line from the equipment. If we feed a dipole from a coax without a balun, the current on the outer conductor splits between the dipole conductor and the outside of the conductor. Patterns and impedance measurements detect this current. Unbalanced currents on the arms of the dipole and feeder currents cause pattern squint, but the cross-polarization radiated is usually a greater concern. 5-15.1 Folded Balun A folded balun (Figure 5-36) allows the direct connection of a coax line to the dipole. A dummy coax outer conductor is connected to the pole fed from the center conductor. It runs alongside the feeder coax for λ/4 and connects to ground. The other pole connects directly to the shield of the feeder coax. The outer conductor of the coax and the extra line are two lines in a three-wire line with ground. We analyze the structure by using balanced (odd) and unbalanced (even) modes. Unbalanced-mode excitation

Balanced Output to Dipole Arms

l/4

Unbalanced Coax

FIGURE 5-36 Folded balun.

BALUNS

Ys

Ya

253

Antenna Ys = −jY0 cot kl

Coax

Quarter Wavelength Line with a Short Circuit Termination

FIGURE 5-37

Folded balun equivalent circuit (balanced mode).

at the dipole forms a magnetic wall through the ground connection between the two coax shields. The circuit reduces to a single line with an open circuit at the ground connection. The open circuit transforms through the quarter-wavelength line to a short circuit at the dipole. Any unbalanced currents induced on the dipole or the coax outer conductor are shorted at the input. Balanced-mode excitation at the dipole forms an electric wall through the ground connection. The balanced-mode circuit of the two coax shields is a λ/4 short-circuited stub connected in shunt with the dipole (Figure 5-37). We analyze the frequency response from Figure 5-37. The bandwidth of the balun, although narrow, exceeds the bandwidth of the dipole. The Roberts balun [23] design adds an open-circuited stub λ/4 long inside the dummy coax of the folded balun. Instead of connecting the center conductor of feeding coax to the outer shield, we connect it to the open-circuited stub. The equivalent circuit for the balanced mode includes the short-circuited stub of the folded balun plus the open-circuited stub. The two reactances shift in opposite directions as frequency changes and produce a dual resonance we see as a loop on the Smith chart plot of impedance. The frequency bandwidth increases to almost 3 : 1, a more suitable choice for wide-bandwidth antennas. 5-15.2 Sleeve or Bazooka Baluns An outer jacket shields the outer conductor of the coax feeder in a sleeve balun (Figure 5-38). The sleeve and outer conductor of the coax form a series stub between the coax feeder and ground when the cup is short circuited to the coax outer conductor. The λ/4 stub presents a high impedance to the unbalanced currents at the top of the cup (Figure 5-39). A second sleeve below the ﬁrst one and directed away from the dipole further prevents currents excited on the coax from reaching the input. When the frequency shifts, the connection to ground through the sleeve unbalances the transmission line. This balun is inherently narrowband. Adding a stub to the center conductor (Figure 5-40) increases the bandwidth because the stubs track each other when the frequency changes. Figure 5-39 demonstrates the circuit diagrams of the two types of sleeve baluns. The type II sleeve balun has matching series stubs on the outputs. The lines remain balanced at all frequencies, but the stubs limit the bandwidth of efﬁcient operation. Marchand [21] adds an open-circuited λ/4 stub inside the matching type II extra shorted stub of the sleeve balun and connects it to the coax center conductor in the same manner as the Roberts balun. The Roberts balun is a folded balun version of the Marchand compensated sleeve balun. The coaxial dipole is a variation of the sleeve or bazooka balun. We rotate the right pole in Figure 5-38 until it is vertical and remove the left pole. We turn over the

254

DIPOLES, SLOTS, AND LOOPS

l/4

FIGURE 5-38 Sleeve or bazooka balun. Coax Inner Conductor Unbalanced

Balanced line

Inside of Outer Conductor

Outside of Outer Conductor

Zs

Zs

l/4

l/4

Sleeve

Type II Only Zg

FIGURE 5-39

Schematic of types I and II sleeve or bazooka baluns. Balanced Mode

Unbalanced Mode

l/4

l/4

FIGURE 5-40 Type II sleeve balun.

sleeve and connect the short-circuit end to the outer conductor of the coax. The sleeve becomes the second pole of the dipole. The short-circuited stub at the bottom of the dipole between the outer conductor of the coax and sleeve transforms to an open-circuit impedance at the end of the lower pole. This prevents current ﬂow farther down the coax. Some references call this a sleeve dipole, which should not be confused with the

BALUNS

255

sleeve dipole used to increase the impedance bandwidth. The coaxial dipole has the inherently narrow bandwidth of the bazooka balun, but is a convenient construction. 5-15.3 Split Coax Balun [24, p. 245] A split coax balun allows the connection of both arms of a dipole to the outer shield of the coax that maintains symmetry to the dipole arms. Its rigidity helps to overcome vibration problems. Slots cut in the outer shield (Figure 5-41) enable the coax line to support two modes and make it equivalent to a three-wire line. A shorting pin excites the TE11 mode in the slotted coax (Figure 5-42) to feed the dipole in the balanced mode. Analysis of a split coax balun is similar to that of a folded balun. The ends of the slots are equivalent to the ground connection of the two coax shields of the folded balun. A virtual open circuit forms at the ends of the slots in the unbalanced (even) mode. It transforms to a short circuit at the dipole and shorts the unbalanced mode at the input. The virtual short circuit at the end of the slots in the balanced mode transforms to an open circuit at the input. Figure 5-37 gives its circuit diagram. Symmetry improves the performance of a split coax balun over a folded balun. The shorting pin is used only to excite the TE11 mode to feed the dipole arms. The extra wire length of the center conductor jumper of the folded balun introduces phase shift to the second arm and squints the beam. For that reason, the split coax balun is a better high-frequency balun. The phase shift problem of the jumper also occurs with the “inﬁnite” balun of the log-periodic antenna. Dipole Arm Shorting Pin

l/4

Slot Dipole Arm

FIGURE 5-41

TEM

Split coax balun. (From [24], Fig. 8-5, 1948 McGraw-Hill.)

TE11

Pin E=0

FIGURE 5-42 Coaxial transmission-line modes in a split coax balun. (From [24], Fig. 8-6, 1948 McGraw-Hill.)

256

DIPOLES, SLOTS, AND LOOPS

5-15.4 Half-Wavelength Balun A half-wavelength balun (Figure 5-43) works by cancellation of the unbalanced-mode currents at the input to the coax. The impedance transforms by a factor of 4 from unbalanced- to balanced-mode ports. In the unbalanced (even) mode, equal voltages are applied to the two output ports. When the voltage wave on the upper line propagates through λ/2, its phase changes by 180◦ . This signal cancels the signal connected directly to the coax center conductor. A load across a balanced-mode transmission line has a virtual short circuit halfway through it. The load on each balanced-mode line is 2Z0 , where Z0 is the coax characteristic impedance. The load on the end of the λ/2-long line is transformed by the transmission line to the identical impedance when it circles the entire Smith chart. The two loads, each 2Z0 , are connected in shunt at the coax input and combine to Z0 . A balanced-mode impedance of 4Z0 transforms to Z0 at the coax input. The λ/2long cable can be rolled up for low frequencies. The balun transforms 300- input impedances of folded dipoles to 75 by using RG-59 cable (75 ). 5-15.5 Candelabra Balun A candelabra balun (Figure 5-44) transforms the unbalanced-mode impedance fourfold to the balanced-mode port. The coax cables on the balanced-mode side connect in series, whereas those on the unbalanced-mode side connect in parallel. We can divide the balanced-mode impedance in two and connect each half to a 2Z0 impedance transmission line. These lines then connect in shunt at the unbalanced-mode port. The unbalanced-mode currents short out at the input to the 2Z0 coax lines in the same manner as does the folded balun. More lines can be stacked in series and higher-impedance transformations obtained, but construction becomes more difﬁcult. 5-15.6 Ferrite Core Baluns Ferrite cores can be used to increase the load impedance to unbalanced-mode currents and reduce them. At low frequencies (l/2

lg/2

TEM Input

Ground Planes

FIGURE 5-55 Stripline series slot.

SHALLOW-CAVITY CROSSED-SLOT ANTENNA

269

the cavity. The cavity reactance slope limits the bandwidth of the stripline-fed slot to a few percent. Increasing the impedance of the waveguide cavity transmission line reduces the reactance slope contributed by the cavity. We increase the bandwidth by using greater distances between the ground planes and thereby increase the waveguide transmission-line impedance. In general, greater volumes for an antenna increase the impedance bandwidth. Rotating the slot relative to the stripline feeding line reduces its load on the transmission line. The waveguide top wall series slot relation [Eq. (5-34)] applies in this case. The slot maintains its polarization while the nonradiating stripline center conductor approaches the slot at an angle. Rotated slots in waveguide must be paired symmetrically to reduce cross-polarization. A longitudinal array [37] can be made by placing all the slots on the centerline of a boxed stripline. Either edge plating or a series of plated-through holes forms a waveguide structure that supports only the TE10 mode. Slots placed on the centerline (as in Figure 5-59, slot c) fail to interrupt the waveguide mode currents. The stripline meanders below and varies the excitation by changing the angle between the slot and the stripline center conductor. The slight loading of each slot excites very little of the parallel-plate mode that causes unwanted slot coupling. Both traveling-wave and resonant linear arrays are possible. See Section 5-26 for a discussion of slot arrays.

5-22 SHALLOW-CAVITY CROSSED-SLOT ANTENNA We can feed the slot in Figure 5-55 by exciting the cavity in an odd mode from two points on opposite sides of the slot. To be able to excite both polarizations, we divide the slot in two and rotate the two parts in opposite directions by 45◦ to form a cross. We use a square cavity to maintain symmetry and replace the shorting pins with solid walls (Figure 5-56c). Since we feed across the diagonal between the crossed slots, we excite both slots. The sum of the ﬁelds radiated from the two slots is polarized in the direction of the diagonal. We increase the radiation conduction by lengthening the crossed slots to the maximum, which lowers the Q (increased bandwidth). The cavity compensates for the slot susceptance to obtain resonance. A crossed-slot antenna was built [38] with the following dimensions: Cavity edge Cavity depth Slot length

0.65λ 0.08λ 0.915λ

The measured 2 : 1 VSWR bandwidth was 20.8%. √ The bandwidth exceeded that of a microstrip patch of the same thickness by about 2. Lindberg [39] found that the resonant length of the slot depends on the cavity depth and requires some experimental adjustment. King and Wong [41] added ridges (Figure 5-56b) to increase the bandwidth. Antennas with ridges need a larger cavity width and a longer slot than the unridged design. The ridges can be stepped as shown to increase the bandwidth. Adding ridges gives us extra parameters to adjust for best input match performance. The following design with uniform ridges produces a 58.7% 2.5 : 1 VSWR bandwidth with a double resonance curve.

270

DIPOLES, SLOTS, AND LOOPS Feed Probes (4)

16

16 (a) b d

5.5

h1 h2

(b)

Feed Probes (4) (0.35 in from Corner) 180° Hybrid l1

L

l1

l1

l1

l2

l2 180° Hybrid 90° Hybrid (c)

FIGURE 5-56 Shallow-cavity crossed-slot antenna: (a) cavity with ridge; (b) cavity with ridge; (c) typical slot conﬁguration. All dimensions are in inches. (From [38], Fig. 2, 1975 IEEE.)

Cavity edge Slot length Ridge width Cavity thickness

0.924λ 1.3λ 0.087λ 0.115λ

Slot width, W2 Ridge height Feed width, W1

0.058λ 0.076λ 0.144λ

Both the ridge and slot shapes can be varied to improve the performance. As fed in Figure 5-56c, the antenna radiates circular polarization on a boresight. Near the horizon (90◦ from the boresight), the polarization reduces to linear as we enter the null of one of the slots.

5-23

WAVEGUIDE-FED SLOTS [24, p. 291; 40, p. 95]

Waveguide is an ideal transmission line for feeding slots. Although its impedance cannot be deﬁned uniquely, all possible candidates—voltage and current, power and

RECTANGULAR-WAVEGUIDE WALL SLOTS

271

current, or power and voltage—yield high values that match the high values of impedance of half-wavelength slots. Waveguide provides a rigid structure with shielded ﬁelds. The slots couple to the internal ﬁelds and allow the easy construction of linear arrays fed from traveling waves or standing waves in the waveguide. By controlling the position of the slots in the walls, the amplitude of the slot excitation can be controlled. The waveguide ﬁelds excite a slot when the slot interrupts the waveguide wall currents. When excited, the slot loads the waveguide transmission line. We make the following assumptions about the wall slots. 1. The slot width is narrow. When a slot grows in width, we must either consider it to be an aperture in the wall or assume that it is excited by interrupting currents in two coordinate directions. 2. The slot is a resonant length and its length is near λ/2. The waveguide environment, the wall thickness, and the position in the wall all affect the resonant length. In most cases, experiments must determine the resonant length. 3. The electric ﬁeld is directed across the narrow width of the slot and varies sinusoidally along its length and is independent of the excitation ﬁelds. This reiterates assumptions 1 and 2. An aperture radiates the polarization of the incident ﬁelds, but resonant-length slots can be excited only with a sinusoidal voltage standing wave. The slot direction determines polarization. 4. The waveguide walls are perfectly conducting and inﬁnitely thin. Even though the walls have thickness, the difference has a small effect on the general form of the slot excitation formulas. As in the case of the resonant length, experiments determine a few values from which the rest must be interpolated, or the values provide the constants for more elaborate models. 5-24 RECTANGULAR-WAVEGUIDE WALL SLOTS The lowest-order mode (TE10 ) in a rectangular waveguide has the following ﬁelds [41, p. 69]: Ey = E0 sin(kc x)e−j kg z kg E 0 sin(kc x)e−j kg z ωµ kc E 0 cos(kc x)e−j kg z Hz = − j ωµ

Hx = −

(5-24)

where kc = π/a, kg2 = kc2 − k 2 , and a is the guide width with cutoff wavelength λc = 2a. We can separate TE10 -mode rectangular waveguide ﬁelds into two plane waves that propagate at an angle to the axis and reﬂect from the two narrow walls. We denote as ξ the angle of the waves measured from the centerline of the waveguide or with respect to the wall. We relate the waveguide propagation to this angle: ξ = sin−1 (λ/λc )

(5-25)

At high frequencies, ξ → 0 and the waves travel straight through the guide as though the walls are not there. As the wavelength approaches cutoff, ξ → 90◦ and the waves

272

DIPOLES, SLOTS, AND LOOPS

reﬂect back and forth between the sidewalls instead of propagating down the guide. This angle factors into the expressions for slot loading to the waveguide transmission line and can be related to propagation: λ λ = cos ξ 1 − (λ/λc )2 c phase velocity Vph = and group velocity = c cos ξ cos ξ λ = cos ξ relative propagation constant P = λg guide wavelength, λg =

For analysis we divide the ﬁelds bouncing down the waveguide into z-directed ﬁelds of the axial wave moving down the guide and x-directed ﬁelds of the transverse wave, a standing wave between the two narrow walls. A standing wave causes a 90◦ separation of the voltage and currents in a transmission line as shown in Figure 5-1. The phase of the currents excited in the waveguide walls due to the ﬁelds will be 90◦ relative to the electric ﬁeld. The wall currents Js are determined by J = n × H, where n is the unit normal to the wall. When we apply this boundary condition to the walls, we obtain the following wall currents: Sidewalls: Jy = −j

E0 kc −j kg z e ωµ

Bottom wall (y = 0 ): Js =

E0 −j kg z e [kg sin(kc x)ˆz + j kc cos(kc x)ˆx] ωµ

Js =

−E0 −j kg z e [kg sin(kc x)ˆz + j kc cos(kc x)ˆx] ωµ

(5-26)

Top wall (y = b):

Equation (5-26) shows that transverse wave currents are 90◦ out of phase with respect to the electric ﬁeld E0 . The current alternates between the two types of current as the wave propagates down the waveguide. In the case of a standing wave along the z-axis caused by a short circuit, the axial wave currents are 90◦ out of phase with the electric ﬁeld across the waveguide (Figure 5-1). The peak amplitude of the transverse wave currents occurs at the same point as the electric ﬁeld in a standing wave along the z-axis, since both are 90◦ out of phase with the axial wave currents. The sidewalls Jy have only transverse wave currents. The top and bottom broad walls have both xdirected transverse wave and z-directed axial wave currents. Figure 5-57a shows the direction and amplitude distribution of these transverse waves. Slots interrupting these

RECTANGULAR-WAVEGUIDE WALL SLOTS IJI

273

IJI

J J

E

J

E J

Traveling Wave Currents

Transverse Standing Wave Currents (a)

(b)

FIGURE 5-57 TE10 -mode rectangular waveguide wall currents: (a) transverse wave currents; (b) axial wave currents.

Axial Wave Current

Transverse Wave Current

Slot

Slot Short

FIGURE 5-58 Short-circuited waveguide axial and transverse wave currents and the location of longitudinal wall slots.

currents are shunt loads to the waveguide. In an axial wave along the z-axis, these transverse waves propagate in the z-axis direction. Equation (5-26) shows that the transverse wave currents are 90◦ phase with respect to the axial wave currents. Figure 5-58 shows the two types of currents along the zaxis when the guide has a short circuit at its end. When measuring slots that interrupt transverse wave currents, we need to place the waveguide short at λg /4 or 3λg /4 away from the slot. This locates the peak of the transverse wave currents ﬂowing around the waveguide walls at the slot shown in Figure 5-58 because the axial wave currents are at a minimum. The second consideration is the shunt load on the waveguide. The λg /4 section of waveguide transforms a short circuit on the end of the waveguide (to the axial wave currents) to an open circuit at the slot. From a voltage point of view the susceptance of the shorted stub is at a minimum. We place the short circuit at

274

DIPOLES, SLOTS, AND LOOPS

λg /2 from the last slot for a series loading slot that interrupts the axial wave currents. This locates the current maximum at the slot and causes maximum interaction with the waveguide ﬁelds. Figure 5-58 illustrates the placement of the next slot λg /2 down the guide at the next current maximum. Figure 5-57a indicates the transverse wave current ﬂow and we see that the currents ﬂow toward the centerline, producing currents 180◦ out of phase on the two sides of the centerline. The two slots in Figure 5-58 are excited by oppositely directed currents that add 180◦ phase shift between the slots. This phase shift compensates for internal standing-wave current phasing of 180◦ due to the λg /2 spacing. Longitudinal top and bottom wall slots cut x-directed transverse shunt currents. The central slot c, located at a current null, fails to be excited. We use this nonradiating slot to insert a traveling probe to measure VSWR. When moved off center, slots d and e cut x-directed currents and are excited. The shunt conductance has the relation g = g1 sin2

πx a

(5-27)

where x is the distance from the guide centerline. Shunt currents on either side of the centerline of the top or bottom wall (Figure 5-57a) have different directions. Besides any traveling-wave phase, slots d and e (Figure 5-59) are 180◦ out of phase. Top-wall longitudinal slots generate no cross-polarization, since all maintain the same orientation. We relate the peak conductance g1 to the direction of the waves in the guide [41]: g1 = 2.09

a cos2 [(π/2) cos ξ ] b cos ξ

(5-28)

Equation (5-28) indicates that the conductance increases for a given spacing off the centerline as the frequency approaches cutoff and ξ → π/2. We cannot use Eqs. (5-27) and (5-28) for design because they do not include the wall thickness and we need to determine the exact length for resonance. The resonant length depends on the spacing from the centerline. Fortunately, the coupling between longitudinal slots is small enough that measurements can be made on single slots. Elliott suggests a measurement plan for longitudinal slots [3]. We build a series of slotted waveguides each containing a single slot at different distances from the centerline. Seven cases are sufﬁcient to generate a curve for design. We need

d

x′ q b

c

q

e g

x′

h

f

a

Shunt Loads

Series Loads

FIGURE 5-59 TE10 -mode rectangular waveguide wall slots.

RECTANGULAR-WAVEGUIDE WALL SLOTS

275

to locate a sliding short circuit farther down the waveguide and adjust it until the standing-wave current peaks at the slot to produce maximum radiation and conductance. With a network analyzer we measure the conductance normalized with respect to the waveguide impedance. Initially, we machine the slots too short, measure the results, and then machine longer slots using the same guides and remeasure until they pass through resonance. Since the manufacturing cost of test slots is high and they require careful measurements, analytical methods of determining slot parameters become attractive. FEM programs can model the details of the slot, the waveguide, and the wall thickness. A number of runs similar to those of the measurements allows design curves to be created. Sidewall slots (Figure 5-59) interrupt shunt transverse waves. Slot a fails to cut surface currents and is not excited. By tilting slot b, currents are cut. The sidewall slot conductance is given for θ < 30◦ by g = g0 sin2 θ

(5-29)

where g0 is the peak conductance. Note that the sidewall slots must cut into the top and bottom walls to achieve a resonant length. The peak conductance can be related to the direction of the waves in the waveguide [Eq. (5-25)] [1, p. 82]: g0 = 2.09

a sin4 ξ b cos ξ

(5-30)

Equation (5-30) shows the relationship of the slot load conductance versus the frequency. As frequency increases, ξ decreases and the conductance falls off as the fourth power of the sine of the angle. The complete theory of Stevenson gives the conductance for an arbitrary tilt [42]: a sin4 ξ g = 2.09 b cos ξ

sin θ cos[(π/2) cos ξ sin θ ] 1 − cos2 ξ sin2 θ

2 (5-31)

Tilting the slots to interrupt currents introduces cross-polarization components in the array pattern. We alternate the direction of tilt to reduce cross-polarization. Two things prevent the total cancellation of cross-polarization. First, the amplitude taper of the array changes the amplitude from element to element and the ﬁelds do not cancel. Alternating the tilt of the slots symmetrically about the centerline in an array with an even number of elements prevents cross-polarization on the boresight. Off the boresight, the array effect of the spaced elements introduces a cross-polarization pattern, since cross-polarization is not canceled at each element. Although Eqs. (5-29) and (5-30) give the slot conductance, they cannot be used for design. They assume an inﬁnitely thin wall and ignore the high level of radiation along the waveguide wall. These slots readily couple to neighboring slots. The effective conductance needs to include the mutual conductance. For these slots we build a series of slotted waveguides containing a group of slots all tilted to the same angle and cut so that they are a resonant length. This means that we will ﬁrst need to build the slots about 5% shorter than resonance length, make measurements, and then machine the slots longer and repeat the measurements to ﬁnd the resonant length. We space the slots at the same distance as will be used in the ﬁnal design and either place a

276

DIPOLES, SLOTS, AND LOOPS

short-circuit beyond the last slot to produce a maximum current at all slots or load the waveguide to form a nonresonant array. We measure the load of the group of slots on the waveguide transmission line using a network analyzer and divide the conductance by the number of slots to get an incremental conductance. This conductance is larger than the one measured on a single slot. We ﬁt the group of measurements to a curve that replaces Eq. (5-29) for design. Axial z-directed waves (Figure 5-57b) peak in the center of the broad walls and taper to zero at the edges. They remain zero on the sidewalls. When centered, transverse slots f and g (Figure 5-59) interrupt the maximum current. When moved off center, g, their series loading to the waveguide drops: R = R0 cos2

πx a

(5-32)

The maximum resistance is related to the direction of the waves in the waveguide: R0 = 2.09

a sin2 ξ 2 π cos sin ξ b cos3 ξ 2

(5-33)

An evaluation of Eq. (5-33) shows that the resistance increases as frequency approaches cutoff for a given location of the slot, a result similar to that for other slot conﬁgurations. The mutual coupling between these series slots is high. We perform incremental resistance experiments similar to the procedure used for sidewall slots to discover the true values of resistance versus offset. Rotating the broadwall transverse slot, h, reduces the z-axis directed current interrupted. When the slot is centered, equal and opposite shunt currents are cut by the slot and the slot fails to present a load to shunt currents: R = R0 cos2 θ

(5-34)

We can excite slots a and c by probe coupling into the waveguide. A probe placed next to the slot and extending into the guide feeds the slot. The longer the probe, the more it disturbs the waveguide ﬁelds to excite the slot. Probes placed on opposite sides of the slots induce ﬁelds 180◦ out of phase with respect to each other.

5-25

CIRCULAR-WAVEGUIDE SLOTS

Figure 5-60 shows the transverse wave and axial wave currents of the circular waveguide TE11 dominant mode. Slots may be placed successfully only at the current maximums without affecting the polarization of the internal wave. A longitudinal slot placed halfway between the current maximums, 45◦ , interrupts only shunt transverse waves. Since any polarization is possible in the circular waveguide, analytically we divide the incident wave into two waves. One is polarized in the direction of the slot; the other is polarized perpendicular to the slot axis. The wave polarized perpendicular to the slot location has its current maximum at the slot and it removes power from the wave. The other wave produces a current null on the slot. When we combine the two ﬁelds after the slot, the unloaded wave is larger and the combined wave rotates its polarization toward the slot. Circumferential slots interrupting axial wave currents

CIRCULAR-WAVEGUIDE SLOTS

J

J

IJI

E J

E

277

IJI

J

Transverse Standing Wave Currents

Traveling Wave Currents

FIGURE 5-60 TE11 -mode circular waveguide wall currents.

also cause polarization rotation of the wave when not centered 90◦ from the electric ﬁeld direction. Slots placed at the maximum of transverse currents cut them when rotated about the axis of the waveguide. Like rectangular-waveguide sidewall slots, the slots oriented perpendicular to the guide axis, circumferential, do not load the waveguide. Rotating the slot increases the shunt load on the waveguide. Slots placed at the maximum of the axial wave cut z-directed currents. Field probes can monitor the internal ﬁelds of the waveguide through a longitudinal slot without causing radiation from the slot. When the slot is rotated away from the axis direction, it interrupts series axial wave currents, loads the waveguide, and radiates. Coaxial TEM-mode transmission line and TM01 -mode circular waveguide have the same outer wall currents (Figure 5-61). Slots can be excited and load the waveguide only by interrupting these axial wave currents. In Figure 5-61, slot a fails to cut currents and is not excited. VSWR measuring probes use this slot. Slots b and c interrupt the currents and series-load the guide. Slot c, whose total length is resonant, is excited by the small portion in the center cutting z-directed currents. We can probe feed slot a, but the probe shunt loads the waveguide or TEM coax that would be series loads on the waveguide if they directly interrupted the axial wave currents.

IJI

Traveling Wave Currents

a

b

c

Wall Slots

FIGURE 5-61 Coax or TM01 -mode circular waveguide wall currents and slots.

278

5-26

DIPOLES, SLOTS, AND LOOPS

WAVEGUIDE SLOT ARRAYS [4, p. 402]

Waveguide slot arrays can produce low sidelobe antennas for pencil beams with good aperture efﬁciency. Array fabrication requires close manufacturing tolerances to achieve the desired amplitude distribution because random errors in manufacture produce unwanted sidelobes and raise the general sidelobe level. Producing these arrays is an art requiring careful analysis of all slot interactions, slot dimensioning determined from models and measurements, and precision machining and assembly. An array consists of a set of waveguides loaded with slots and joined with a corporate feed into the total array. The corporate feed can also be a slotted array feeding the individual waveguides that contain the radiating slots. Aperture size and distribution determine the beamwidth and sidelobes in the various planes. We divide arrays of slots into two groups: nonresonant, excited by traveling waves, and resonant, excited by standing waves. Waves either travel along the guide into a terminating load or reﬂect from a short and set up standing waves along the z-axis (Figure 5-58). Traveling-wave currents excite the slots as they pass, and slots may be placed anywhere relative to the load. The distance between slots and the propagation constant determine the relative phases. Standing waves set up a ﬁxed sinusoidal current pattern along the waveguide axis at a given frequency. The standing-wave phase is either 0◦ or 180◦ . Slots placed in the current nulls of standing waves interrupt no currents and fail to be excited by the waveguide. We can vary the amplitude by the z-axis placement of the slots. The termination determines the array type. Do not confuse transverse waves that produce shunt currents and z-axis standing waves caused by a short-circuit termination. Both traveling and standing waves on the z-axis have shunt currents. Standing waves (resonant array) produce beams normal to the array axis. A resonant array maintains its beam direction when frequency changes, but the standing-wave pattern shifts and changes the excitation of the slots (Figure 5-62). The amplitudes of the slots farthest from the short circuit change the most, since the standing waves have shifted farther. The length of the resonant array determines its bandwidth. The pattern shape changes because distribution and input impedance change as the loads change when the standing-wave currents shift. Nonresonant array (traveling-wave) beam directions are functions of the propagation constant of the wave exciting the slots. Changing the frequency shifts the beam direction. If the load on the end reﬂects a wave, another beam forms from the reﬂected traveling wave. The second beam appears at the same angle to the axis of the waveguide

Frequency Shift by 10%

Waveguide Axis

Slot

Half Guide Wavelength Slot Slot

Short Slot

FIGURE 5-62 Standing-wave currents in resonant array relative to slots and after 10% frequency shift.

279

WAVEGUIDE SLOT ARRAYS

as the ﬁrst but measured from the −z-axis. The ﬁrst-pass radiated power and return loss of the load determine the level of this second beam relative to the ﬁrst. Both resonant and nonresonant waveguide slot arrays use resonant-length slots. We space the slots λg /2 apart in the resonant array, as shown in Figure 5-58. We place the slots at alternating positions about the centerline of the broadwall or at alternating tilt angles in the sidewall to give the additional 180◦ phase shift to produce a broadside beam. The admittances of the slots of the resonant array add at the input because the λg /2 spacing produces a complete rotation around the Smith chart. In nonresonant arrays a traveling wave is used to excite the slots. We space the slots at other than λg /2 distances and terminate the waveguide with a load. We assume a matched system throughout the antenna in a ﬁrst-order analysis suitable for most designs. The beam of most nonresonant slot arrays is designed to backﬁre at an angle to broadside. 5-26.1 Nonresonant Array [43] In a nonresonant waveguide, slot array resonant-length slots are used in a travelingwave antenna terminated at the end with a load. The antenna radiates at an angle to the normal of the waveguide face determined by wave velocity and slot spacing. We vary the slot loading along the waveguide so that each slot radiates the proper amount of the remaining power. A termination absorbs the power remaining after the last slot. With a mismatched termination the reﬂected power radiates a second lower-amplitude beam as the wave travels to the source. We design with either shunt- or series-loading slots. A shunt slot radiates the power |V |2 gi /2, where gi is the normalized slot conductance. Similarly, a series slot radiates the power |I |2 ri /2, where ri is the normalized slot resistance. We normalize the conductance or resistance to a per unit length function: g(z) or r(z). The attenuation equation (4-78) becomes 1 dP = −g(z) or P (z) dz

− r(z)

(5-35)

Equation (5-35) modiﬁes the normalized attenuation equation (4-79) [24, p. 291]: g(z)L =

|A(z)|2

L

[1/(1 − R)] 0

z

|A(z)|2 dz −

(5-36) |A(z)|2 dz

0

where the aperture runs ±L/2 and R is the ratio of the input power absorbed by the termination. A(z) is the normalized aperture distribution on the interval ± 21 . We change to r(z)L in Eq. (5-36) for series-loading slots. Equation (5-36) assumes light loading by the slots so that the waveguide transmission line is matched at all points. This approximation improves as the length increases. Equation (5-36) is the same as Eq. (4-79) except for a constant. We divide the values in Table 4-28 or Figure 4-26 by 4.34 to calculate normalized conductance (resistance) of shunt (series) slots times the array length. Each slot provides the loading over the spacing between slots: gi =

d/2

−d/2

where d is the spacing of the slot at zi .

g(z) dz g(zi ) d

280

DIPOLES, SLOTS, AND LOOPS

We space the slots at other than λg /2. At λg /2 spacings, all reﬂections from the mismatches (slots) add in phase at the input. The small mismatches from each slot add with various phase angles for element spacing different from λg /2 and cancel each other to some extent to give a good input match over a reasonable bandwidth. When we increase the array length, we can no longer ignore the waveguide losses. The slot conductances become very small and radiate power on the same order as the losses. We modify Eq. (5-36) to include the losses as in Eq. (4-79), and the slot conductance increases to compensate for the ohmic losses in the walls. A small slot conductance is difﬁcult to achieve with longitudinal broadwall slots because one edge of the slot must be over the centerline of the waveguide wall and the results become unpredictable. The achievable conductances limit possible distributions in a slotted waveguide array. Mutual coupling between slots changes the distribution and we must modify the slot offsets to account for mutual coupling using Eq. (3-23). If we specify the radiating power of each slot in a discrete sequence Pi , we modify Eq. (5-36). The integrals become summations, since |A(z)|2 = δ(z − id)Pi where d is the slot spacing, δ(x) the Dirac delta (impulse) function, and Pi the power coefﬁcient of the ith slot. The power radiated is N

z 0

|A(z)|2 dz

0

i=1

The integral reduces to

L

Pi = Pin (1 − R) =

|A(z)|2 dz is the power radiated by the preceding slots. Equation (5-36) gi = ri =

1−

Pi i−1 n=1

Pn

(5-37)

Dissipating more power in the termination decreases each Pi and the required conductance (resistance) range of the slots. We alternate the locations of longitudinal slots about the centerline of the broadwall to add 180◦ phase shift between elements. Similarly, sidewall slot directions are alternated along the array. The additional phase shifts cause backﬁre of the beam in most cases. The element spacing, as well as the traveling-wave phase velocity, determines the beam direction. The phasing equation in the array factor for beam peak becomes kd cos θ + 2nπ = P kd − π, where θ is measured from the array axis, P is the relative propagation constant (P < 1), and n is an arbitrary integer. We solve for the beam peak direction and the necessary spacing to get a particular beam direction: (n + 12 )λ −1 θ = cos P− (5-38) d n + 12 d = λ P − cos θmax We usually work with n = 0 because using n > 0 produces multiple beams.

(5-39)

WAVEGUIDE SLOT ARRAYS

281

Example Compute slot spacing to produce a beam at θ = 135◦ in a waveguide of width 0.65λ. Calculate the relative propagation constant from the general equation for a waveguide. 2 λ P = 1− λc For λc = 2a,

P =

1−

1 1.3

2 = 0.640 =

λ λg

From Eq. (5-39), using n = 0, we determine spacing in free space: d/λ = 0.371. The waveguide spacing is given by d d = P = 0.371(0.640) = 0.237 λg λ If we use n = 1, then d/λ = 1.11, which radiates an additional beam at θ = 79◦ for n = 0 [Eq. (5-38)]. Beams enter visible space at cos θ = −1 (180◦ ) and move toward end ﬁre (θ = 0) as the spacing increases. We calculate the region of single-beam operation from Eq. (539). The minimum d/λ occurs when θ = 180◦ for n = 0, and the maximum occurs when θ = 180◦ for n = 1: d 1.5 0.5 ≤ ≤ (5-40) 1+P λ 1+P We substitute the upper bound into Eq. (5-38) and use n = 1 to derive the minimum angle of single-beam operation: 1+P θmin = cos−1 P − 3

(5-41)

Example Determine the minimum scan angle (toward end ﬁre) for P = 0.6, 0.7, 0.8, and 0.9 that has a single beam. We substitute these values into Eq. (5-41) to ﬁnd: P θmin

0.6

0.7

0.8

0.9

86.2◦ 82.3◦ 78.5◦ 74.5◦

If we scan to θ = 90◦ , the spacing becomes λg /2 and the mismatches from each slot add to the input and produce a resonant array. The array with a forward ﬁring beam has a slot spacing greater than λg /2. Given a waveguide with P = 0.8, we use Eq. (5-39) to calculate spacing to give beams at 80◦ and 100◦ : 0.5 d = = 0.798 and λ 0.8 − cos 80◦ d d = P = 0.639 and λg λ

0.5 d = = 0.514 λ 0.8 − cos 100◦ d d = P = 0.411 λg λ

282

DIPOLES, SLOTS, AND LOOPS

A nonresonant array has a backﬁre beam that scans toward broadside as frequency (and P ) increases. Hansen [44, p. 90] gives the slope of the beam shift with frequency change: d sin θ 1 f = − sin θ (5-42) df P where f is the frequency. 5-26.2 Resonant Array We space the slots at λg /2 and terminate the waveguide end with a short circuit either λg /4 or 3λg /4 from the last one for shunt loading slots in a resonant array. The beam radiates broadside to the array. The 2 : 1 VSWR bandwidth of the array is approximately 50%/N for N elements in the array. The antenna is narrowband. The admittances of all elements add at the input. To have a matched input, N g = 1, where gi is the i=1 i normalized slot conductance. If we deﬁne Pi as the normalized power radiated by the ith slot, then N gi = Pi where Pi = 1 i=1

5-26.3 Improved Design Methods The methods given above ignore the interaction of slots and their effect on the transmission line. We can describe the array as a loaded transmission line and consider the interactions of the slots by accounting for the transmission-line mismatches [45, pp. 9–11]. We ignore the mutual coupling for longitudinal broadwall slots because it is small, but sidewall slots have high mutual coupling and require an adjustment of the effective slot impedance. We use an incremental admittance, found from the measured change in admittance, when one slot is added to the array or total conductance of the array divided by the number. This accounts somewhat for the mutual coupling. Elliott and Kurtz [46] relate the self-admittance of a longitudinal broad-wall slot, measured or calculated, to the mutual admittance of the array of slots found from equivalent dipoles. They use Babinet’s principle and the mutual impedance of equivalent dipoles. The method requires solution of a set of 2N equations in the location and length of the slots to give the desired excitation while accounting for mutual coupling. Their formulation ignores slot interaction in the waveguide beyond the ﬁrst-order mode. Elliott [47] extends this method to the analysis and design of nonresonant arrays. Of course, when we design a planar array, the slots between waveguide sticks couple readily and we need to account for the mutual coupling between them. The voltage excitation needs to be adjusted to account for this coupling or the desired distribution will not be achieved. Dielectric loaded waveguide arrays require additional analysis because the approximation of a piecewise sinusoidal distribution, such as dipole current, fails to model the slot distribution adequately. Elliott [48] uses a slot distribution E(x) = cos

πx 2b

where b is the length. Mutual impedances between dipoles that have the wrong distribution are not used; instead, the active admittances are found from forward and

REFERENCES

283

back scattering between the slots directly. The method still requires the solution of 2N equations for the slot lengths and locations. REFERENCES 1. R. F. Harrington, Time-Harmonic Electromagnetic Fields, McGraw-HiIl, New York, 1961. 2. C. A. Balanis, Antenna Theory, Analysis and Design, 2nd ed., Wiley, New York, 1997. 3. H. G. Booker, Slot aerials and their relation to complementary wire aerials, Proceedings of IEE, vol. 92, pt. IIIA, 1946, pp. 620–626. 4. R. S. Elliott, Antenna Theory and Design, Prentice-Hall, Englewood Cliffs, NJ, 1981. 5. V. H. Rumsey, Frequency Independent Antennas, Academic Press, New York, 1966. 6. W. H. Watson, Wave Guide Transmission and Antenna Systems, Oxford University Press, London, 1947. 7. R. W. Masters, Super-turnstile antenna, Broadcast News, vol. 42, January 1946. 8. J. D. Kraus, Antennas, McGraw-Hill, New York, 1950. 9. H. M. Elkamchouchi, Cylindrical and three-dimensional corner reﬂector antennas, IEEE Transactions on Antennas and Propagation, vol. AP-31, no. 3, May 1983, pp. 451–455. 10. S. Maci et al., Diffraction at artiﬁcially soft and hard surfaces by using incremental diffraction coefﬁcients, IEEE AP-S Symposium, 1994, pp. 1464–1467. 11. E. L. Bock, J. A. Nelson, and A. Dome, Very High Frequency Techniques, McGraw-Hill, New York, 1947. Chapter 5. 12. A. J. Poggio and P. E. Mayes, Pattern bandwidth optimization of the sleeve monopole antenna, IEEE Transactions on Antennas and Propagation, vol. AP-14, no. 5, September 1966, pp. 623–645. 13. W. L. Stutzman and G. A. Thiele, Antenna Theory and Design, Wiley, New York, 1981. 14. H. E. King and J. L. Wong, An experimental study of a balun-fed open-sleeve dipole in front of a metallic reﬂector, IEEE Transactions on Antennas and Propagation, vol. AP-19, no. 2, March 1972, pp. 201–204. 15. J. L. Wong and H. E. King, A cavity-backed dipole antenna with wide bandwidth characteristics, IEEE Transactions on Antennas and Propagation, vol. AP-21, no. 5, September 1973, pp. 725–727. 16. A. Kumar and H. D. Hristov, Microwave Cavity Antennas, Artech House, Boston, 1989. 17. Y. Mushiake, An exact step-up impedance ratio chart of folded antenna, IRE Transactions on Antennas and Propagation, vol. AP-2, 1954, p. 163. 18. R. C. Hansen, Folded and T-match dipole transformation ratio, IEEE Transactions on Antennas and Propagation, vol. AP-30, no. 1, January 1982, pp. 161–162. 19. The ARRL Antenna Book, American Radio Relay League, Inc., Newington, CT, 1974. 20. R. A. Burberry, VHF and UHF Antennas, Peter Peregrinus, London, 1992. 21. N. Marchand, Transmission-line conversion transformers, Electronics, December 1941, pp. 142–145. 22. W. L. Weeks, Antenna Engineering, McGraw-Hill, New York, 1968. 23. W. K. Roberts, A new wide-band balun, Proceedings of IRE, vol. 45, December 1957, pp. 1628–1631. 24. S. Silver, ed., Microwave Antenna Theory and Design, McGraw-Hill, New York, 1949. 25. J. W. Duncan and V. P. Minerva, 100 : 1 Bandwidth balun transformer, Proceedings of IRE, vol. 48, February 1960, pp. 156–164. 26. B. A. Munk, Baluns, Chapter 23 in J. D. Kraus and R. J. Marhefka, Antennas, McGrawHill, New York, 2002.

284

DIPOLES, SLOTS, AND LOOPS

27. P. K. Park and C. T. Tai, Receiving antennas, Chapter 6 in Y. T. Lo and S. W. Lee, eds., Antenna Handbook, Van Nostrand Reinhold, New York, 1993. 28. A. Alford and A. G. Kandoian, Ultrahigh frequency loop antennas, AIEE Transactions, vol. 59, 1940, pp. 843–848. 29. A. J. Fenn, Arrays of horizontally polarized loop-fed slotted cylinder antennas, IEEE Transactions on Antennas and Propagation, vol. AP-33, no. 4, April 1985, pp. 375–382. 30. T. Tsukiji and S. Tou, On polygonal loop antennas, IEEE Transactions on Antennas and Propagation, vol. AP-28, no. 4, July 1980, p. 571. 31. W. C. Wilkinson et al., Two communication antennas for the Viking lander spacecraft, 1974 IEEE Antennas and Propagation Symposium Digest, vol. 12, June 1974, pp. 214–216. 32. C. C. Kilgus, Multielement fractional turn helices, IEEE Transactions on Antennas and Propagation, vol. AP-16, no. 4, July 1968, pp. 499–500. 33. C. C. Kilgus, Resonant quadriﬁlar helix, IEEE Transactions on Antennas and Propagation, vol. AP-17, no. 3, May 1969, pp. 349–351. 34. A. A. Oliner, Equivalent circuits for discontinuities in balanced strip transmission line, IRE Transactions on Microwave Theory and Techniques, vol. MTT-3, March 1955, pp. 134–143. 35. J. S. Rao and B. N. Ras, Impedance of off-centered stripline fed series slot, IEEE Transactions on Antennas and Propagation, vol. AP-26, no. 6, November 1978, pp. 893, 894. 36. J. Van Bladel, Small hole in waveguide wall, Proceedings of IEE, vol. 118, January 1971, pp. 43–50. 37. P. K. Park and R. S. Elliott, Design of collinear longitudinal slot arrays fed by boxed stripline, IEEE Transactions on Antennas and Propagation, vol. AP-29, no. 1, January 1981, pp. 135–140. 38. H. E. King and J. L. Wong, A shallow ridged-cavity cross-slot antenna for the 240- to 400MHz frequency range, IEEE Transactions on Antennas and Propagation, vol. AP-23, no. 5, September 1975, pp. 687–689. 39. C. A. Lindberg, A shallow-cavity UHF crossed-slot antenna, IEEE Transactions on Antennas and Propagation, vol. AP-17, no. 5, September 1969, pp. 558–563. 40. R. C. Hansen, ed., Microwave Scanning Antennas, Vol. II, Academic Press, New York, 1966. 41. R. F. Harrington, Time-Harmonic Electromagnetic Fields, McGraw-Hill, New York, 1961, p. 69. 42. A. F. Stevenson, Theory of slots in rectangular waveguides, Journal of Applied Physics, vol. 19, January 1948, pp. 24–38. 43. A. Dion, Nonresonant slotted arrays, IRE Transactions on Antennas and Propagation, vol. AP-7, October 1959, pp. 360–365. 44. R. C. Hansen, in A. W. Rudge et al., eds., The Handbook of Antenna Design, Vol. II, Peter Peregrinus, London, 1983. 45. M. J. Ehrlich, in H. Jasik, ed., Antenna Engineering Handbook, McGraw-Hill, New York, 1961. 46. R. S. Elliott and L. A. Kurtz, The design of small slot arrays, IEEE Transactions on Antennas and Propagation, vol. AP-26, no. 2, March 1978, pp. 214–219. 47. R. S. Elliott, On the design of traveling-wave-fed longitudinal shunt slot arrays, IEEE Transactions on Antennas and Propagation, vol. AP-27, no. 5, September 1979, pp. 717–720. 48. R. S. Elliott, An improved design procedure for small arrays of shunt slots, IEEE Transactions on Antennas and Propagation, vol. AP-31, no. 1, January 1983, pp. 48–53.

6 MICROSTRIP ANTENNAS

Microstrip antennas are planar resonant cavities that leak from their edges and radiate. We can utilize printed circuit techniques to etch the antennas on soft substrates to produce low-cost and repeatable antennas in a low proﬁle. The antennas fabricated on compliant substrates withstand tremendous shock and vibration environments. Manufacturers for mobile communication base stations often fabricate these antennas directly in sheet metal and mount them on dielectric posts or foam in a variety of ways to eliminate the cost of substrates and etching. This also eliminates the problem of radiation from surface waves excited in a thick dielectric substrate used to increase bandwidth. As electronic devices continue to shrink in size, the antenna designer is pushed to reduce the antenna size as well. Cavity antennas use valuable internal volume, but we have the conﬂict that restricting the volume limits impedance bandwidth. Bandwidths widen with increased circuit losses (material losses) or by efﬁcient use of the restricted volume. Bounds on bandwidth can be found by enclosing the antenna in a sphere and expanding the ﬁelds into TE and TM spherical modes [1,2]. Each mode radiates, but it requires more and more stored energy as the mode number increases. Decreasing the volume increases the Q value of each mode and a sum, weighted by the energy in each mode, determines the overall Q value. Antennas that use the spherical volume efﬁciently and reduce power in the higher-order modes have the greatest bandwidths. A single lowest-order mode puts an upper bound on bandwidth, given the size of the enclosing sphere. Greater volumes have potential for greater bandwidth provided that the energy in higher-order spherical modes is restricted. Increasing material losses or adding small resistors increases bandwidth beyond the single-mode bound [2]. We discover that increasing the volume of ﬂush antennas increases the impedance bandwidth provided that the radiation mode on the structure can be maintained. Thicker substrates develop greater bandwidths, but they increase the possibility of higher-ordermode excitation and surface-wave losses. Losses limit the lower bound of bandwidth Modern Antenna Design, Second Edition, By Thomas A. Milligan Copyright 2005 John Wiley & Sons, Inc.

285

286

MICROSTRIP ANTENNAS

as we reduce the thickness because efﬁciency degrades to a point where the bandwidth remains constant. Microstrip consists of a metal strip on a dielectric substrate covered by a ground plane on the other side. Unlike stripline, the single ground plane shields the circuit on only one side, but normal packaged microstrip—within a receiver, for example—has a second shielding ground plane to reduce circuit interactions. The dielectric substrate retains most of the power because the shielding ground plane is spaced a few substrate thicknesses away. Removing the shield in antenna applications allows radiation from resonant cavities. We also discover feeding circuits etched on the substrate radiate to some extent, but their radiation is comparatively small. Arrays of antennas can be photoetched on the substrate, along with their feeding networks, and microstrip provides easy connections to active devices and allows placement of preamps or distributed transmitters next to the antenna elements. Diode phase-shifter circuits etched in the microstrip form single-board phased arrays. Microstrip circuits make a wide variety of antennas possible through the use of the simple photoetching techniques. The vast literature on microstrip antennas concentrates on the microwave circuit analysis of the internal parts of the antenna used to control the internal modes. Designers have increased the bandwidth of the antenna by coupling to multiple resonators, such as vertically stacked or coplanar coupled patches or by using internal slots and apertures. These multiple resonators increase the impedance bandwidth, and in the best cases the antenna continues to radiate the same pattern. As antenna designers we need to concentrate ﬁrst on obtaining the desired pattern while working to increase the impedance bandwidth. Simple microstrip antennas have much larger pattern bandwidths than impedance bandwidths, but as more resonators are added to increase the impedance bandwidth, spreading in the horizontal plane alters the radiated pattern and we must return to concentrate on the pattern. Microstrip patch antennas consist of metal patches large with respect to normal transmission-line widths. A patch radiates from fringing ﬁelds around its edges. Impedance match occurs when a patch resonates as a resonant cavity. When matched, the antenna achieves peak efﬁciency. A normal transmission line radiates little power because the fringing ﬁelds are matched by nearby counteracting ﬁelds. Power radiates from open circuits and from discontinuities such as corners, but the amount depends on the radiation conductance load to the line relative to the patches. Without proper matching, little power radiates. The edges of a patch appear as slots whose excitations depend on the internal ﬁelds of the cavity. A general analysis of an arbitrarily shaped patch considers the patch to be a resonant cavity with metal (electric) walls of the patch and the ground plane and magnetic or impedance walls around the edges. The radiating edges and fringing ﬁelds present loads along the edges. In one analysis [3] the patch effective size is increased to account for the capacitive susceptance of fringing ﬁelds, and the radiation admittance is ignored to calculate resonant frequency. The far ﬁeld is integrated to compute radiated power and the equivalent radiation conductance. The second method [4] is to retain the patch size but satisfy boundary conditions into a loaded wall whose load is determined by radiation and fringing ﬁelds. Assuming a constant electric ﬁeld from the ground plane to the substrate allows solutions in terms of modes TM to the substrate thickness. Boundary conditions determine possible modes and correspond to the dual TE modes of waveguides having electric walls. Patches in the shape of standard coordinate system

MICROSTRIP ANTENNA PATTERNS

287

axes, such as rectangular and circular, give solutions in terms of tabulated functions. Numerical techniques used for arbitrarily shaped waveguides can be applied to patches with nonstandard shapes. We consider only rectangular and circular patches. 6-1 MICROSTRIP ANTENNA PATTERNS We start our discussion of patches with their pattern characteristics. It is difﬁcult to separate a discussion of pattern from the internal construction consideration, but we will only brieﬂy discuss the internal structures that affect the pattern. The small size of microstrip antennas limits control of the pattern and we must use arrays of patches to control its pattern seriously. Rectangular and circular are the most common shapes for microstrip antennas and they radiate similar broad patterns. When we load the cavity to shrink its size, it radiates wider beamwidth patterns that lower directivity (gain). Antennas that couple to coplanar patches to increase the impedance bandwidth will radiate narrower beams, but the basic patch has a wide beamwidth. If we couple to multiple coplanar patches, we can expect the pattern to narrow or vary its shape as the mixture of modes on the various patches changes over the frequency range of operation. Patches consist of metal plates suspended over large ground planes. We excite the cavity in a variety of ways that we discuss later. Electric currents ﬂow on the plate and on the ground plane around the antenna, and these radiate. If we use vertical probes to excite the antenna from coaxial lines, the currents ﬂowing on these radiate and add to the pattern. We can reduce the antenna size by adding vertical shorting plates (quarter-wave patches) or shorting pins near the feed pins (compact patches), and these also radiate from the current ﬂow on them. Remember that the patch radiates from real electric currents, although the distribution is complicated. We simplify the problem of computing patch radiation by using magnetic currents along the edges. Figure 6-1 illustrates the fringing electric ﬁelds around the edges of square and circular patch antennas excited in the lowest-order cavity modes. The arrow sizes indicate the magnitude of the ﬁelds. The square patch has nearly uniform ﬁelds along two edges we call the width, and a sinusoidal variation along the other two edges, called the resonant length. The ﬁelds vanish along a virtual electrically short-circuited plane halfway across the patches. On either side of the short-circuit plane, the ﬁelds are directed in opposite directions. Looking from above the ﬁelds along the width, both edges are in the same direction. The circular patch fringing ﬁelds distribution varies as cos φ, where the angle φ along the rim is measured from the peak electric ﬁeld. Magnetic currents found from the fringing electric ﬁelds can replace the electric currents located on the patch and the surrounding ground plane for pattern analysis. Figure 6-2 shows the distribution of magnetic currents around the edges, with the size of the arrowhead indicating magnitude. Our use of magnetic currents around the patch perimeter reduces the pattern calculation to equivalent slots. A two-element array consisting of slots with equivalent uniform magnetic currents produces the E-plane radiation of a rectangular patch. To √ ﬁrst order, the slots are spaced λ/2 εr and we can determine the pattern from the equivalent two-element array. The magnetic currents along the resonant length sides individually cancel because the current changes direction halfway across the edge. The currents also cancel from side to side. These cancellations eliminate pattern contributions to the E- and H -plane patterns. The slot length determines the H -plane pattern.

288

MICROSTRIP ANTENNAS

(a)

(b)

FIGURE 6-1 Fringing electric ﬁelds around microstrip patches: (a) square; (b) circular. (From L. Diaz and T. A. Milligan, Antenna Engineering Using Physical Optics, Figs. 3.12 and 3.19, 1996 Artech House, Inc.).

The H -plane of the slot has the same pattern as the E-plane of a dipole and produces a null along its axis. Figure 6-3 illustrates the pattern of a patch on an inﬁnite ground plane using a free-space substrate. The two-element slot array in the E-plane generates a null along the ground plane because the elements are spaced λ/2. The H -plane dashed curve shows the null along the ground plane due to the polarization of the slots. The light curves give the Huygens source polarization (Section 1-11.2) patterns in the diagonal planes. The antenna radiates cross-polarization (dashed curve) in this plane from the combination of separated magnetic currents along the resonant-length sides and from the unbalance in the beamwidths in the principal planes. When we design a microstrip patch on a dielectric substrate, the size reduction brings the two slots closer together and widens the E-plane beamwidth and eliminates its null along the ground plane. Figure 6-4 illustrates the pattern of a patch designed for a substrate with εr = 2.2. The H -plane pattern retains its null along the ground plane due to the slot pattern. The cross-polarization of the Huygens source in the diagonal plane increases because of the increased difference between the beamwidths of the principal plane patterns. Table 6-1 gives the directivity of a square and circular patch on an inﬁnite ground plane found by integrating the pattern. The range of directivity

MICROSTRIP ANTENNA PATTERNS

289

(a)

(b)

FIGURE 6-2 Equivalent magnetic currents on the edges of microstrip patches: (a) square; (b) circular.

of a patch is limited. Increasing the width of a rectangular patch increases directivity by shrinking the H -plane beamwidth. We gain some control of the pattern by placing the patch on a ﬁnite ground plane. Figure 6-5 shows the pattern of a square patch on a 2.21 dielectric constant substrate when located on circular disks 5λ, 2λ, and 1λ in diameter. On a 5λ ground plane, edge diffraction adds ripple to the pattern. As the ground plane increases, the angular separation between the ripples decreases, due to the increased array size of the radiation from the two edges. The H -plane pattern widens signiﬁcantly for 1λ- and 2λ-diameter ground planes, as the limited ground plane can no longer support the currents that make the patch edge radiate like a slot. Although the principal-plane beamwidths are more nearly equal for the patch on the 2λ-diameter disk, the cross-polarization in the diagonal plane increases relative to the pattern on the inﬁnite ground plane. The 1λ ground plane increases the gain of the patch by about 1 dB relative to the patch on an

290

MICROSTRIP ANTENNAS

E-Plane

X-Pol Diagonal Plane Diagonal Plane H-Plane

FIGURE 6-3 Patterns of microstrip patch on a free-space substrate mounted on an inﬁnite ground plane.

H-Plane X-Pol

E-Plane

Diagonal

FIGURE 6-4 Patterns of microstrip patch on a dielectric substrate εr = 2.2 over an inﬁnite ground plane. TABLE 6-1 Estimated Directivity of Square and Circular Microstrip Patches on a Large Ground Plane Dielectric Constant

Square Patch (dB)

Circular Patch (dB)

1.0 2.0 3.0 4.0 6.0 8.0 10.0 16.0

8.4 7.7 7.2 7.0 6.7 6.5 6.4 6.3

9.8 7.6 6.7 6.2 5.8 5.5 5.4 5.1

MICROSTRIP ANTENNA PATTERNS

291

X-Pol., Diag.

H-Plane

E-Plane Diag.

(a)

H-Plane

X-Pol., Diag, E-Plane

Diag.

( b)

FIGURE 6-5 Patterns of microstrip patches with dielectric substrate εr = 2.2 mounted over ﬁnite circular ground planes: (a) 5λ diameter; (b) 2λ diameter. (continued )

292

MICROSTRIP ANTENNAS

H-Plane

X-Pol., Diag. E-Plane

(c)

FIGURE 6-5 (continued ) (c) 1λ diameter.

RHC Principal Planes Diag. LHC

Diag. Planes

FIGURE 6-6 Circularly polarized patch mounted on a 1λ-diameter ground plane.

MICROSTRIP PATCH BANDWIDTH AND SURFACE-WAVE EFFICIENCY

293

inﬁnite ground plane. In this case the edge diffractions add constructively to narrow the beamwidths. We can take advantage of the nearly equal E- and H -plane patterns in the forward hemisphere to produce a pattern with excellent circular polarization over the entire hemisphere when we feed the patch for circular polarization. Figure 6-6 gives the circular polarization pattern when the patch is fed in two spots with equal signals phased 90◦ apart. The cross-polarization is 13 dB below the co-polarization at θ = 90◦ in the principal planes and −7 dB in the diagonal plane. We retain these excellent polarization characteristics over a large ground plane if we place the ﬁnite ground plane on a 1λ or greater pedestal above the ground plane. 6-2 MICROSTRIP PATCH BANDWIDTH AND SURFACE-WAVE EFFICIENCY Microstrip patches radiate from the currents induced on the patch or equivalently, the magnetic currents around the periphery of the patch and from surface waves induced in the dielectric slab. The surface waves radiate when they reach the edges of the substrate and their radiation contributes to the normal patch radiation. The fringing ﬁelds from the patch to the ground plane readily excite the lowest-order surface-wave TM0 mode that has no low frequency cutoff. Any thickness dielectric slab supports this mode. We can control the surface-wave radiation by limiting the substrate area or by adding etched photonic bandgap patterns to the open areas of the substrate, but generally, surface waves are undesirable. As the substrate thickness or dielectric constant increases, the ratio of the power in surface waves increases. When we calculate the microstrip patch antenna impedance bandwidth, we must include the directly radiated power and the surface-wave power. For most cases we consider surface-wave radiation as reducing radiation efﬁciency, but for a single patch on a substrate with limited area, its radiation can add constructively. We eliminate surface waves by using metal plate patches without dielectric substrates or low-density foam supports of the patch. Surface waves are bound to the dielectric similar to any transmission line except that the ﬁeld decays exponentially in the direction normal to the surface. Because the surface wave is excited along the ﬁnite edges of the patch, it spreads in the horizontal√plane. The radiation spreads like a two-dimensional wave and the ﬁelds decay as 1/ r, where r is the horizontal distance from the edge. This is a far-ﬁeld approximation, and close to the edge it is a near-ﬁeld problem. Unfortunately, these surface waves increase the coupling between patches fabricated on the same substrate. Simple formulas have been derived for the impedance bandwidth of rectangular patches that include the surface-wave loss [5]. Since substrates can be both electric and magnetic, we deﬁne the index of refraction of a patch substrate that includes both √ parameters: n = εr µr . The idea is that the ratio of space-wave radiation to surfacewave radiation can be found for any small antenna mounted on the substrate and we can then apply it to a patch. By integrating the power density in the radiation from a horizontal Hertzian (incremental) dipole spaced the substrate thickness over a ground plane, we obtain the space-wave radiated power in closed form given the substrate thickness h and the free-space propagation constant k: PRh k 2 (kh)2 · 20 µ2r C1 C1 = 1 −

1 0.4 + 4 n2 n

(6-1)

294

MICROSTRIP ANTENNAS

We express the current on the patch as an integral of Hertzian dipoles. The surfacewave power generated in the substrate by the Hertzian dipole can be simpliﬁed when the substrate is thin: 1 3 h 2 3 3 PSW = k (kh) · 15πµr 1 − 2 (6-2) n We deﬁne the surface-wave radiation efﬁciency by the ratio of radiated power to total power: Ph 4C1 (6-3) ηSW = h r h = 2 3 4C + 3πkhµ PR + PSW 1 r (1 − 1/n ) We relate the power radiated by a patch to a Hertzian dipole by integrating the surface current on the patch consisting of a distribution of small dipoles to calculate the total space-wave power of the patch: PR = PRh m2eq = PRh

2 JS dx dy

(6-4)

S

For a rectangular patch the ratio of PR to PRh m2eq , p, can be approximated by a simple formula given the resonant length L, the width W , and the propagation constant k: p =1−

0.16605(kW )2 0.02283(kW )4 0.09142(kL)2 + − 20 560 10

(6-5)

The 2 : 1 VSWR of the rectangular patch is related to the quality factor Q that includes the space- and surface-wave radiations: BW = √

16C1 p 1 h W = √ 2Q 3 2 ηSW εr λ0 L 1

(6-6)

Figure 6-7 plots the 2 : 1 VSWR bandwidth given by Eq. (6-6) for common substrates versus the free-space thickness in wavelengths and includes the radiation due to surface waves. The surface-wave radiation found using Eq. (6-3) becomes a signiﬁcant part of the total radiation as the substrate thickness increases or the dielectric constant increases, as shown in Figure 6-8 of the surface-wave loss. For a single resonator circuit model for a patch, Eq. (6-6) computes bandwidth from the Q and the allowable input VSWR: BW =

VSWR − 1 √ Q VSWR

or

Q=

VSWR − 1 √ BW VSWR

(6-7)

We determine bandwidth at different VSWR levels by manipulating Eq. (6-7): √ BW2 (VSWR2 − 1) VSWR1 √ = (6-8) BW1 (VSWR1 − 1) VSWR2 Quality factor Q is another way of expressing efﬁciency. The Q used in Eq. (6-6) is the combination of the space-wave radiation QR and the surface-wave radiation QSW : 1 1 PR + PSW 1 1 = + = = Q Qrad QR QSW ωWT

2:1 VSWR Bandwidth, %

MICROSTRIP PATCH BANDWIDTH AND SURFACE-WAVE EFFICIENCY

295

4.5 2.2

6.0

2.94 9.8 1.0

Substrate Thickness, l (Free Space)

Surface Wave Loss, dB

FIGURE 6-7 2:1 VSWR bandwidth of square microstrip patches versus substrate thickness in free-space wavelengths, including surface-wave radiation.

2.2 2.94 4.5 9.8

6.0

Substrate Thickness, l (Free Space)

FIGURE 6-8 Surface-wave loss of microstrip patches versus substrate thickness for common substrate dielectric constants.

WT is the energy stored in the patch and the surface wave and ω = 2πf , the radian frequency. Equation (6-3) can be expressed in terms of Q: ηSW =

Q Qrad = QR QR

296

MICROSTRIP ANTENNAS

The surface wave is not a dissipation loss, but potentially an uncontrolled radiation. Dielectric and conductor losses increase the impedance bandwidth of the patch, but reduce its gain. We express these losses as Q to compute patch efﬁciency. Given the dielectric loss tangent, tan δ, and the patch conductivity σ , we have two more Q terms that reduce the overall Q of the patch in terms of impedance bandwidth: Qd =

1 tan δ

and Qc = h πf µ0 σ

(6-9)

The total quality factor QT is found from the sum of the inverses: 1 1 1 1 1 = + + + QT QR QSW Qd Qc

(6-10)

If we attempt to fabricate a patch on a thin substrate, Qd and Qc become commensurate with the radiation Qrad and efﬁciency suffers. The impedance bandwidth increases due to the dissipation in the microstrip patch. Figure 6-7 does not include these losses. Dielectric Slab Surface Wave We consider the dielectric slab surface wave because it can be excited not only by a microstrip patch but by any wave that passes across it. The slab binds a portion of the wave and releases it when it diffracts at its edges. The surface-wave device slows the wave velocity of this wave relative to the spacewave signal, and when it radiates from the edges it no longer adds in phase with the space wave. The surface-wave ﬁelds decrease exponentially in the direction normal to the surface, and the exponential rate increases as the binding increases and the wave propagates more slowly. A dielectric slab on a ground plane will support TM modes when thin and TE modes when thick. The TM mode is polarized normal to the slab surface, whereas the TE mode is polarized parallel to the slab surface. A TM mode requires an inductive surface such as a corrugated ground plane to bind the wave. While corrugations prevent propagation between the slots, the wave propagates in the dielectric slab by bouncing between the two interfaces at an angle with respect to the surfaces. The second surface can be either free space or a conductor. To solve for the ﬁelds, we equate not only the wave impedance at the boundary but the propagation constants in the two regions as well. We deduce the grounded dielectric slab solution from a slab twice as thick in free space that has an odd-mode electric ﬁeld excitation on the slab sides. The center becomes a virtual short circuit for the odd-mode excitation. We divide the space around the slab into three regions: 1 above the slab, 2 in the slab, and 3 below the slab and then derive the ﬁelds from potential functions [6, p. 129]: −2πbx ψ1 = A1 exp exp(−j kz z) λ 2πpx x sin λ ψ2 = A2 (6-11) exp(−j kz z) cos 2πpx x λ 2πbx ψ3 = ±A1 exp exp(−j kz z) λ

MICROSTRIP PATCH BANDWIDTH AND SURFACE-WAVE EFFICIENCY

297

where the sign of ψ3 depends on satisfying continuous tangential ﬁelds across the lower slab boundary. The center of the coordinate normal to the slab (x) is the slab center. Equating the propagation constants and x-directed wave impedances produces transcendental equations in the transverse propagation constant in the slab px : πpx a tan 2 2 ωa πpx a πpx a λ (6-12) (ε1 µ1 − ε0 µ0 ) − = ±B0 2 λ λ cot πpx a λ where B0 = µ0 /µ1 for TE waves. B0 = ε0 /ε1 for TM waves, ω is the radian frequency (2πf ), a is the slab thickness, and ε1 and µ1 are the permittivity and permeability of the slab. We solve for px [Eq. (6-12)] numerically or graphically and use πp 2 ωa 2 πb x (ε1 µ1 − ε0 µ0 ) − (6-13) a= λ 2 λ to determine attenuation constant b and the relative propagation constant P of the slab surface wave: kz = P k = k 1 + b2 or P = 1 + b2 (6-14) For the TM0 mode we can use an approximate expression for P instead of solving Eq. (6-12) numerically when the slab is thin [7]: P2 = 1 +

(εr µr − 1)2 (2 ka)2 (εr µr )2

(6-15)

Equation (6-12) has an inﬁnite number of solutions, corresponding to the multiple values of the tangent and cotangent functions. Order 0 corresponds to the tangent function from 0 to 90◦ ; order 1 corresponds to the cotangent function from 90 to 180◦ ; and so on. Even-mode orders use the tangent function, and odd-mode orders use the cotangent function. We deﬁne the cutoff frequency as the point where α = 0, the transition between attached and detached waves: 2a ε1 µ1 λc = −1 (6-16) n ε0 µ 0 The cutoff frequency for the zeroth-order mode is zero. Only the TM0 mode has odd symmetry, required for the grounded slab. The grounded slab supports even-order TM modes and odd-order TE modes. Equation (6-12) coupled to Eq. (6-13) has been solved numerically to generate Tables 6-2 and 6-3. Table 6-4 lists the thicknesses of a slab in free space supporting the TM0 mode for a given P . The grounded slab is one-half the thickness of the values in Table 6-2. Similarly, Table 6-3 lists the thicknesses for the TE1 mode. Equation (6-16) can be solved for the minimum thickness to support the TE1 mode. Below that thickness the waves do not bind to the surface. Besides microstrip patches, we feed these surfaces from either a small horn or a parallel-plate transmission line. We match the feed polarization to the mode on the slab, but the slab binds only part of the power. The rest radiates directly from the feed or reﬂects to the feed input. We can feed an ungrounded slab by centering it in a waveguide. The TE10 waveguide mode excites the TE0 slab mode when the mode velocity determining thickness is in the H -plane. Like the grounded slab with the TM0

298

MICROSTRIP ANTENNAS

TABLE 6-2 Thickness (λ0 ) of a Dielectric Slab Supporting a TM0 Modea Dielectric Constant P 1.001 1.002 1.005 1.01 1.02 1.04 1.06 1.08 1.10 1.12 1.14 1.16 1.18 1.20 1.25 1.30 1.35 1.40 a

2.21

2.94

4.50

6.00

9.80

0.02699 0.03672 0.05792 0.08162 0.1147 0.1607 0.1956 0.2253 0.2520 0.2770 0.3012 0.3251 0.3493 0.3741 0.4426 0.5282 0.6492

0.02152 0.03041 0.04784 0.06713 0.09355 0.1289 0.1545 0.1752 0.1930 0.2088 0.2233 0.2369 0.2499 0.2625 0.2934 0.3250 0.3593 0.3986

0.01831 0.02574 0.04032 0.05623 0.07746 0.1046 0.1231 0.1374 0.1491 0.1590 0.1677 0.1756 0.1827 0.1894 0.2045 0.2182 0.2314 0.2444

0.01839 0.02420 0.03744 0.05195 0.07094 0.09444 0.1099 0.1215 0.1307 0.1384 0.1450 0.1508 0.1560 0.1607 0.1712 0.1803 0.1887 0.1966

0.03446 0.04710 0.06316 0.08180 0.09331 0.1015 0.1078 0.1129 0.1171 0.1208 0.1240 0.1269 0.1329 0.1380 0.1424 0.1463

Use half-thickness for a slab on a ground plane.

TABLE 6-3 Thickness (λ0 ) of a Dielectric Slab Supporting a TE1 Modea Dielectric Constant P 1.001 1.002 1.005 1.01 1.02 1.04 1.06 1.08 1.10 1.15 1.20 1.25 1.30 a

2.21

2.94

4.50

6.00

9.80

0.4469 0.4720 0.4829 0.4961 0.5164 0.5494 0.5790 0.6078 0.6368 0.7140 0.8046 0.9182 1.0712

0.3689 0.3709 0.3765 0.3843 0.3962 0.4150 0.4313 0.4465 0.4613 0.4982 0.5372 0.5802 0.6291

0.2743 0.2774 0.2770 0.2810 0.2873 0.2968 0.3049 0.3122 0.3191 0.3356 0.3518 0.3683 0.3856

0.2260 0.2272 0.2302 0.2330 0.2373 0.2438 0.2492 0.2540 0.2585 0.2690 0.2790 0.2890 0.2992

0.1701 0.1705 0.1717 0.1736 0.1761 0.1797 0.1825 0.1851 0.1874 0.1928 0.1978 0.2026 0.2073

Use half-thickness for a slab on a ground plane.

mode, the TE0 mode has no cutoff frequency for a free-space slab. Table 6-4 lists the slab thicknesses for a given relative propagation constant for the TE0 mode. The surface-wave power was found in terms of the relative propagation constant P [7]:

PSW = n2

15πk 2 n2 µ3r (P 2 − 1) √ n4 (P 2 − 1) P2 − 1 1 + kh 1 + + 2 √ n − P2 n2 − P 2 P2 − 1

(6-17)

RECTANGULAR MICROSTRIP PATCH ANTENNA

TABLE 6-4

299

Thickness (λ0 ) of a Dielectric Slab Supporting a TE0 Mode Dielectric Constant

P 1.001 1.002 1.005 1.01 1.02 1.04 1.06 1.08 1.10 1.15 1.20 1.25

2.21

2.94

4.50

6.00

9.80

0.01274 0.01684 0.02649 0.03772 0.05409 0.07872 0.09935 0.1184 0.1368 0.1833 0.2348 0.2968

0.00994 0.01174 0.01666 0.02341 0.03345 0.04823 0.06027 0.07104 0.08113 0.1051 0.1290 0.1542

0.00661 0.00920 0.01289 0.01846 0.02640 0.03275 0.03832 0.04343 0.05507 0.06595 0.07661

0.00410 0.00638 0.00905 0.01286 0.01839 0.02276 0.02656 0.03002 0.03779 0.04489 0.05168

0.00250 0.00364 0.00514 0.00729 0.01040 0.01284 0.01494 0.01684 0.02106 0.02483 0.02835

We combine Eq. (6-1) for the space-wave power with Eq. (6-17) for the surface-wave power to calculate efﬁciency in the same manner as Eq. (6-3). The results are similar. 6-3 RECTANGULAR MICROSTRIP PATCH ANTENNA Although design equations will be given below for single-layer rectangular and circular patches, serious design work should use one of the excellent available commercial design codes [8]. Their use reduces the need to modify the ﬁnal dimensions using a knife to remove metal or metal tape to increase the patches. Antennas can be built with tuning tabs, but the labor to trim these increases cost. Tuning tabs are unsuitable for arrays when the input port to individual antennas cannot be accessed. As we add layers to increase bandwidth, a cut-and-try method becomes extremely difﬁcult, and numerical methods are a necessity. Rectangular patch antennas can be designed by using a transmission-line model [9] suitable for moderate bandwidth antennas. Patches with bandwidths of less than 1% or greater than 4% require a cavity analysis for accurate results, but the transmissionline model covers most designs. The lowest-order mode, TM10 , resonates when the effective length across the patch is a half-wavelength. Figure 6-9 demonstrates the patch fed below from a coax along the resonant length. Radiation occurs from the fringing ﬁelds. These ﬁelds extend the effective open circuit (magnetic wall) beyond the edge. The extension is given by [10] εeff + 0.300 W/H + 0.262 = 0.412 H εeff − 0.258 W/H + 0.813

(6-18)

where H is the substrate thickness, W the patch nonresonant width, and εeff the effective dielectric constant of a microstrip transmission line the same width as the patch. A suitable approximation for εeff is given by [11] εeff

εr + 1 εr − 1 10H −1/2 + = 1+ 2 2 W

(6-19)

300

MICROSTRIP ANTENNAS

l 2√ eff ∋

Feed E

X

E

W

A

Polarization

A

Patch Shorting Pin

L ∆

Center Pin Soldered to Path E

AA Ground Soldered Coax

FIGURE 6-9 Coax-fed microstrip patch antenna.

where εr is the substrate dielectric constant. The transmission-line model represents the patch as a low-impedance microstrip line whose width determines the impedance and effective dielectric constant. A combination of parallel-plate radiation conductance and capacitive susceptance loads both radiating edges of the patch. Harrington [6, p. 183] gives the radiation conductance for a parallel-plate radiator as πW (kH )2 G= 1− ηλ0 24

(6-20)

where λ0 is the free-space wavelength. The capacitive susceptance relates to the effective strip extension: W B = 0.01668 (6-21) εeff H λ Example Design a square microstrip patch antenna at 3 GHz on a 1.6-mm substrate with a dielectric constant of 2.55 (woven Teﬂon ﬁberglass). The patch will be approximately a half-wavelength long in the dielectric.

RECTANGULAR MICROSTRIP PATCH ANTENNA

301

Assume at ﬁrst that the width is λ/2. W =

2f

c √

εr

= 31.3 mm

By Eq. (6-19), εeff = 2.405. On substituting that value into Eq. (6-18), we obtain the effective cutback on each edge; = 0.81 mm. The resonant length is L=

2f

c √

εeff

− 2 = 30.62 mm

When we use this length as the width (square patch) to calculate the effective dielectric constant, we obtain 2.403, very close to the initial value. We can iterate it once more and obtain 30.64 mm for the resonant length. The input conductance of the patch fed on the edge will be twice the conductance of one of the edge slots [Eq. (6-20)]: 30.64 mm [2π(1.6)/100]2 G= 1− = 2.55 mS 120(100 mm) 24 R=

1 = 196 2G

A microstrip feeding line can be attached to the center of one of the radiating edges but 50- transmission lines become inconveniently wide on low-dielectric-constant substrates. More convenient, 100- narrower lines have about the same low loss and are generally used in feed networks. To transform the 196- input resistance of the example above to 100 , we use a 140- quarter-wavelength transformer. The bandwidth of the transformer far exceeds that of the antenna. In the example above, we have a square patch. Why doesn’t the antenna radiate from the other two edges? We can equally well say that the patch is a transmission line in the other direction. The equal distances from the feed point to the nonradiating edges produce equal ﬁelds from the patch to ground. Equal ﬁelds on the edges set up a magnetic wall (virtual open circuit) through the centered feed line and create a poor impedance match to the feed. We expand the radiating edge ﬁelds in an odd mode, since the power traveling across the patch loses 180◦ of phase. The odd mode places a virtual short circuit halfway through the patch. A shorting pin through the center (Figure 6-9) has no effect on radiation or impedance, but it allows a low-frequency grounding of the antenna. The patch can be fed by a coax line from underneath (Figure 6-9). The impedance varies from zero in the center to the edge resistance approximately as Ri = Re sin2

πx L

0≤x≤

L 2

(6-22)

where Ri is the input resistance, Re the input resistance at the edge, and x the distance from the patch center. The feed location does not signiﬁcantly affect the resonant frequency. By using Eq. (6-22), we locate the feed point given the desired input impedance: L −1 Ri x = sin (6-23) π Re

302

MICROSTRIP ANTENNAS

Compute the 50- feed point in the example above: 30.64 −1 sin x= π

50 = 5.16 mm 196

The feed pin currents add to the pattern by radiating a monopole pattern. Figure 6-10 shows this radiation for a patch using a free-space substrate where the E-plane radiating edges are spaced λ/2. The pattern of Figure 6-10 has a null along the ground plane in the E-plane, but the monopole radiation increases the radiation along the ground plane. On one side the radiation adds and on the other it subtracts from the E-plane pattern to form a null tilted above the ground plane. The H -plane pattern now contains crosspolarization. We can reduce the monopole radiation by feeding the patch at a second port located an equal distance from the center on the opposite side. This requires an external feed network that divides the power equally between the two ports with a 180◦ phase difference. The problem with this feed arrangement is that signiﬁcant power is coupled between the two feeds in the equivalent microwave circuit of the patch. The estimated value of −6 dB coupling between the ports causes a portion of the input power to be dissipated in the second port. At this level the patch efﬁciency drops 1.25 dB. We can reduce the monopole radiation by coupling to a second short-circuited probe to the patch instead of directly feeding it. The gap between the second probe and the patch is adjusted until the antenna radiates minimum cross-polarization in the H -plane. This uses the microstrip patch as the feed network, and the second probe has no resistive load to dissipate power. The feed probe across the microstrip patch substrate is a series inductor at the input. Higher-order modes excited in the patch by this feeding method add to the inductive component of the antenna. Below resonance, the antenna is inductive and has nearzero resistance. As the frequency increases, the inductance and resistance grow as the parallel resonance is approached. Above the resonant frequency, the antenna is capacitive as the impedance sweeps clockwise around the Smith chart (Figure 6-11) and ﬁnally back to a slight inductive component near a short circuit. Increasing the

E-Plane H-Plane, X-Pol H-Plane

FIGURE 6-10 Pattern of coax-fed, microstrip patch including feed pin radiation for free-space substrate.

RECTANGULAR MICROSTRIP PATCH ANTENNA

303

OverCoupled UnderCoupled Critically Coupled

FIGURE 6-11 Smith chart frequency response of under-, critically, and overcoupled patches as the feed point moves toward one radiating edge of a rectangular patch.

input resistance by changing the feed point causes the resonant frequency response circle to grow on the Smith chart and cross the resistance line at a higher level. We call the left-hand curve the undercoupled case because the sweep of the curve fails to enclose the center of the chart. The center curve is critically coupled and the right curve is the overcoupled case. This general impedance response also holds for circular patches. We use these terms for all resonance curves when they sweep around or toward the Smith chart center from any peripheral point. Figure 6-12 plots the Smith chart for a design with a patch on a 0.05λ-thick substrate with dielectric constant 1.1 that includes the inductance of the feed pin. The response locus lies above the real axis and is always inductive. We can tune this impedance locus by adding a series capacitor at the input with a reactance −j 50 at the center frequency. The series capacitor moves the locus down until it sweeps around the center of the chart in an overcoupled response. Figure 6-13 shows implementation of the capacitor as a disk on the end of the feed pin. The pin passes through a hole in the patch so that the only connection is through the capacitor disk. The disk can be placed below the patch on a separate substrate in a multiple-layer construction. Other conﬁgurations use an annular ring capacitor etching in the patch at the feed point for small capacitors. Adding to this a series inductor and adjusting the series capacitor improves the impedance match over a larger frequency range, as shown in Figure 6-14, where the locus encircles the origin [12]. The patch with the single added

304

MICROSTRIP ANTENNAS

W/Pin Inductance

Added Series Capacitor

FIGURE 6-12 Impedance improvement by adding a series capacitor to a patch on a thick substrate. Patch

Disk Capacitor Feed Pin

Coax Input

FIGURE 6-13 Cross section of a probe-fed patch with an added series capacitor.

series capacitor has a 9.1% 10-dB return-loss bandwidth while adjusting the series capacitor, and adding a series inductor increases the impedance bandwidth to 15.4%. Matching networks have limited ability to add resonances to broadband the impedance match, but construction becomes difﬁcult. Later, we will obtain extra resonances by adding antenna elements. We can feed patches from the edge by using an inset microstrip line as shown in Figure 6-15, where the gap on either side of the microstrip line equals its width. A FDTD analysis shows that the inset disturbs the transmission line or cavity model

305

RECTANGULAR MICROSTRIP PATCH ANTENNA

w/Pin Inductance

Series Capacitor Series Inductor Matching Network

FIGURE 6-14 Impedance response of a patch with a two-element matching network.

Substrate

Patch Feed Inset

FIGURE 6-15

Inset-fed square patch.

and increases the impedance variation with distance compared to a coaxial probe feed given a patch resonant length L and feed position x from the center [13]: Ri = Re sin4

πx L

0≤x≤

L 2

(6-24)

306

MICROSTRIP ANTENNAS

Equation (6-24) is an approximate solution because at x = 0, the resistance remains ﬁnite. We locate the feed from the equation using a radian angle measure: L −1 Ri 1/4 x = sin (6-25) π Re Compute the 50- feed point in the example above: 30.64 −1 50 0.25 x= = 7.71 mm sin π 196 The inset distance (7.3 mm) is less than the distance of the probe (9.8 mm) from the edge. Aperture Feed [14,15] A microstrip patch is a planar resonant cavity with opencircuited sidewalls that leak power in radiation. We can also think of the rectangular patch operating in the lowest-order mode as a low-impedance transmission line with end susceptance and radiation conductance. Both models predict a resonant structure with signiﬁcant Q. Resonant cavities are readily excited by coupling to a transmission line through an aperture or by direct feeding from a transmission line. The Q of the resonant cavity limits the excitation ﬁelds to one of the modes. We can expand the excitation in the cavity modes, but the lowest-order mode is usually the most signiﬁcant and contains most of the stored energy. We generally consider the voltage distribution in a patch with its null plane located halfway across the patch through the center. Whether we consider it as a cavity or a transmission line the standing-wave voltage has a standing-wave current associated with it. This current is out of phase with the voltage and its peak occurs along the virtual short circuit through the centerline. Along the resonant length the current has a cosinusoidal distribution that vanishes at the radiating edges in a single half-cycle for the lowest-order mode. The current has a uniform distribution along the patch width. We produce maximum coupling to a patch through a slot by distorting the currents in the ground plane of the patch where they are maximum in the center of the patch. To ﬁrst order the currents ﬂow along the resonant length. This means that we align the slot perpendicular to the current ﬂow for maximum excitation in the same manner as slots in waveguides (Section 5-24). To excite the slot we pass a microstrip transmission line across it perpendicularly. This leads to a three-layer structure. The patch is located on the top layer. Its ground plane contains a coupling aperture usually placed under the center of the patch for maximum coupling. The third layer contains a microstrip transmission using the same ground plane as the patch and located under the center of the slot to maximize coupling. Figure 6-16a shows an exploded view of the patch, ground with its aperture, and the microstrip transmission line ﬂipped over relative to the patch. Figure 6-16b gives the general parameters associated with the slot aperture. Although xos and yos are usually zero to maximize coupling, the patch current distribution tells us how the coupling varies with slot location. Because the current in the ground plane is uniform across the patch width W , coupling is independent of xos until the slot starts to overlap the edge of the patch. The cosinusoidal distribution current distribution along the resonant length direction L means that the coupling falls off slowly as yos is moved off zero. The sign of yos does not matter because the distribution is an even function. The slow variation of current near the patch center means that the slot location has a loose tolerance.

RECTANGULAR MICROSTRIP PATCH ANTENNA

b ε r

307

db

Ground Plane with Aperture

a ε r

da

(a ) Lp

Wap

Wf

y Xos Wp

Ls

Lap

yos

x ( b)

FIGURE 6-16 Aperture feed of square patch. (From [15], Fig. 1, 1986 IEEE.)

The microstrip transmission line excites the slot (aperture) from a standing-wave with its maximum current located at the slot. We maximize the standing-wave current by either using a shorting via from the microstrip line to the ground plane or by using a quarter-wave open-circuited transmission line stub of length Ls . Ls will be less than a quarter-wave in the effective dielectric constant of the microstrip line because the open-circuit end has fringing capacitance and its capacitance must overcome the higherorder modes of the microstrip patch, which load the input inductively. The reactance of the stub, a series load to the input, is given by the equation ZS = −j Z0 cot(keff LS ) where Z0 is the characteristic impedance of the microstrip feed line, keff the effective propagation constant in the microstrip substrate, and Ls the stub length ≈0.22λeff . We increase the coupling to the patch resonant cavity by increasing the aperture size. Figure 6-17 shows the Smith chart variation with aperture size as the coupling

308

MICROSTRIP ANTENNAS

e

rtur

g asin

Ape

e

Incr

FIGURE 6-17 Effect of aperture size on coupling to a patch where larger openings move the response to the right.

varies left to right as undercoupled, critically coupled, and overcoupled. When we increase the bandwidth, we lower the Q, and the coupling aperture size must increase. Waterhouse [8] suggests starting with a slot about one-half the patch width and using a commercial code to analyze the response while adjusting dimensions before fabrication. We control the rotational position on the Smith chart by varying the open-circuited stub length. Shorter lengths, below λ/4, increase the capacitive reactance and the coupling loop will rotate around a constant-resistance circle with its diameter determined by the aperture size, as shown in Figure 6-18. Figure 6-19 gives aperture shapes in order of increasing coupling. The longer slot of (b) compared to slot (a) increases coupling. Widening the aperture as in (c) increases coupling relative to (a). The H-shaped slot has a more uniform distribution along the horizontal slot and increased coupling. The bowtie and hourglass apertures increase coupling from a consideration of increased path length around the opening. The smooth curve of the hourglass reduces current discontinuities at the edges and increases coupling [16, pp. 158–159]. Aperture feeding eliminates the vertical pin structure in the microstrip patch and eases construction but at the cost of a multiple-layer etching. The elimination of the vertical pin removes the added monopole pattern, which increases cross-polarization. When the patch is edge fed, whether directly or inset, the substrate for good patch radiation does not match the one needed for good microstrip lines. With an aperturefed patch, each structure can use its optimum substrate, because they are independent and connected only through the aperture. As we try to feed broadband patches, the

RECTANGULAR MICROSTRIP PATCH ANTENNA

309

FIGURE 6-18 Effect of varying length of an open-circuited stub in an aperture-fed patch when critically coupled.

0.55

0.15 1.1

1.4

(a)

(b)

1.1 (c) 0.55

0.15 1.1 (d )

1.1 (e)

1.1 ( f)

FIGURE 6-19 Aperture shapes to increase coupling and bandwidth. (From [16], Fig. 4-29, 2003 Artech House, Inc.)

Q decreases and the aperture size grows. This slot, although below a resonant size, increases its radiation and decreases the front-to-back ratio because it radiates equally on both sides. One solution is to enclose the microstrip line in a box to prevent slot radiation on the back side. If we use a high-dielectric-constant substrate for the microstrip line, the coupling through the aperture remains high, but the second ground plane of the microstrip will reduce the coupling. The slot aperture adds a pole to the patch circuit that can be used to broadband the impedance response. To use this pole effectively, we must increase the aperture size until it becomes a signiﬁcant radiator.

310

MICROSTRIP ANTENNAS

6-4 QUARTER-WAVE PATCH ANTENNA When operation is in the lowest mode, a virtual short circuit forms through a plane centered between the two radiating edges. We can make an antenna by using half the patch and supplying the short circuit (Figure 6-20). The E-plane pattern broadens to that of a single slot. The resonant length is about a quarter-wavelength in the dielectric of the substrate. We use the effective dielectric constant εeff of a microstrip line of patch width W and given by Eq. (6-18) to determine the resonant length L of the quarter-wave patch: L λ − (6-26) = √ 2 4 εeff We can implement the short circuit with a series of pins or etched vias between the ground plane and the patch. These add an inductive component to the transmission-line model of the antenna. The effective shorting plane occurs further along the transmission line. The equivalent extra length l is found from the parallel-plate circuit model of a row of evenly spaced pins [17, p. 104]. Given the pin center spacing S, their radius √ r, and the wavelength in the dielectric λd = λ0 / εr , we compute the patch-length reduction from the equation 2 S S S 2πr 2 l = ln (6-27) + 0.601 − 2π 2πr S λd We have only the conductance and susceptance of a single edge that doubles the resonant resistance at the edge as compared with the full patch. It becomes difﬁcult to feed the antenna from microstrip because this raises the quarter-wavelength transformer impedance and requires narrower lines. We can increase the edge width to reduce the edge input resistance, but the antenna is usually fed from underneath. Equation (6-23) gives the approximate feed location measured from the short circuit. The resonant frequency shifts slightly as the feed point moves. During tuning for impedance match, the length of the cavity will have to be adjusted to maintain the desired resonant frequency. Quarter-wave and full-patch antennas have the same Q. A half-patch antenna has half Input Coax l 4√ eff ∋

∆

Feed Pin

L/2

E E E

Substrate

FIGURE 6-20 Quarter-wave patch.

Shorting wall

QUARTER-WAVE PATCH ANTENNA

311

the radiation conductance but only half the stored energy of a full-patch antenna. Its bandwidth is approximately the same as that of the full patch. Example Design a half-patch antenna at 5 GHz on a 0.8-mm-thick substrate (εr = 2.21) with a radiating width of 0.75λ. The edge width is 0.75(300 mm)/5 = 45 mm. By using Eq. (6-19), we compute the effective dielectric constant in the cavity: εeff = 2.16. Equation (6-18) gives us the cutback for fringing ﬁelds: = 0.42 mm. The resonant length becomes L λ − = 10.20 − 0.42 = 9.78 mm = √ 2 4 εeff The radiation conductance from the single edge is [Eq. (6-20)] G=

45 = 6.25 mS 120(60)

or

R = 160

The 50- feed point is found from Eq. (6-23): 19.56 −1 x= sin π

50 = 3.69 mm 160

where x is the distance from the short. The short circuit of this antenna is quite critical. The low impedance of the microstrip cavity raises the currents in the short circuit. Without a good low-impedance short, the antenna will detune and have spurious radiation. If the antenna is made from a machined cavity, careful attention must be paid to the junction between the top plate and the cavity to assure good electrical contact. Figure 6-21 shows the calculated pattern of a quarter-wave patch on a free-space substrate 0.04λ thick on an inﬁnite ground plane. The antenna radiates primarily from the single edge located opposite the short-circuited edge. A vertical probe feeds the

E-Plane

H-Plane X-Pol.

H-Plane

FIGURE 6-21 Pattern of a quarter-wave patch on a free-space substrate.

312

MICROSTRIP ANTENNAS

E-Plane H-Plane X-Pol.

H-Plane

(a)

H-Plane X-Pol E-Plane

H-Plane

(b)

FIGURE 6-22 Pattern of a quarter-wave patch mounted on (a) 2λ- and (b) 10λ-diameter ground planes.

CIRCULAR MICROSTRIP PATCH

313

antenna directly. The E-plane has a broad, nearly constant pattern. Radiation from current on the probe and shorting pins adds to positive angle radiation of the equivalent single slot while subtracting in the opposite direction. The H -plane pattern (dashed curve) retains its null along the ground plane. The light-line curve gives the cross-polarization in the H -plane. The feed probe and shorting pin currents produce a pattern similar to that from a monopole. The model that uses equivalent magnetic currents fails to predict the high radiation from these currents. When the quarter-wave patch is mounted on a ﬁnite ground plane, it exhibits behavior similar to that of a monopole. Figure 6-22a,b plots the pattern when it is mounted on 2λ- and 10λ-diameter ground planes. These show a monopole-type pattern, where radiation spreads readily behind the ground plane. Currents ﬂowing in the feed pin and shorting wall distort the E-plane and cause asymmetry. The magnetic currents ﬂowing along the side slots no longer cancel as in the square patch and increase cross-polarization. If we close off the nonradiating edges with metal walls, the walls convert the parallel-plate line into a waveguide and we use the waveguide propagation constant to calculate the quarter-wavelength cavity depth. The slot ﬁelds vanish on the ends and establish a sinusoidal slot distribution. We can offset the feed toward both the back wall and the sidewall to reduce the input impedance. The peak voltage (minimum current and peak resistance) occurs at the slot center. Figure 6-23 illustrates the pattern of the waveguide quarter-wave patch on an inﬁnite ground plane. The sidewalls reduce the monopole radiation, and the H -plane cross-polarization is reduced compared to a quarter-wave patch. When mounted on a 2λ-diameter disk, centered on the feed pin, the pattern (Figure 6-24) exhibits lower-level radiation in the backlobe because the monopole pattern has been reduced. The high radiation level at the disk edges still causes considerable edge diffraction in the E-plane. 6-5 CIRCULAR MICROSTRIP PATCH In some applications, a circular patch ﬁts in the available space better than a rectangular one. In a triangularly spaced array, they maintain a more uniform element

E-Plane H-Plane X-Pol

H-Plane

FIGURE 6-23 Pattern of a quarter-wave waveguide patch.

314

MICROSTRIP ANTENNAS

H-Plane E-Plane

H-Plane X-Pol

FIGURE 6-24 Quarter-wave waveguide patch mounted on a 2λ-diameter disk.

environment. No suitable transmission-line model presents itself, and the cavity model must determine the resonant frequency and bandwidth. The cutoff frequencies of TE modes of circular waveguides give the resonant frequencies of circular patch antennas. The patch with its magnetic walls and TM modes is the dual of the waveguide. The resonant frequencies are given by fnp =

Xnp c √ 2πaeff εr

(6-28)

where Xnp are the zeros of the derivative of the Bessel function Jn (x) of order n, as is true of TE-mode circular waveguides. The term aeff is an effective radius of the patch [18]: 2H πa ln + 1.7726 (6-29) aeff = a 1 + πaεr 2H where a is the physical radius and H is the substrate thickness. Using the effective radius gives the resonant frequency within 2.5%. We combine Eqs. (6-28) and (6-29) to determine radius to give a particular resonant frequency: Xnp c (6-30) aeff = √ 2πfnp er

CIRCULAR MICROSTRIP PATCH

315

Since a and aeff are nearly the same, we can iterate Eq. (6-29) to compute a, the physical radius [19, p. 119]: a=

aeff 1 + 2H /πaεr ln (πa/2H ) + 1.7726

(6-31)

We start by using aeff for a in Eq. (6-31), which converges rapidly. The lowest-order mode, TM11 , uses X11 (1.84118) and produces a linearly polarized ﬁeld similar to a square patch. The TM01 mode (X01 = 3.83171) produces a monopole-type pattern from a uniform edge fringing ﬁeld. Example Design a circular microstrip patch antenna (TM11 mode) at 3 GHz on a 1.6-mm substrate that has a dielectric constant of 2.55 (woven Teﬂon ﬁberglass). We calculate the effective radius from Eq. (6-30): aeff =

1.84118(300 × 109 mm/s) = 18.35 mm √ 2π(3 × 109 Hz) 2.55

The physical radius will be slightly less. By using aeff in the denominator of Eq. (631), we obtain a physical radius: a = 17.48 mm. We can then substitute this back into Eq. (6-31) and obtain a = 17.45 mm. Equation (6-31) converges in two iterations to a reasonable tolerance, since another iteration gives the same value. Actually, a single iteration gives the value within 0.2% on a formula accurate to only 2.5%. The ﬁelds of the TM11 mode produce a virtual short circuit at the center of the patch. We can reinforce the short circuit with a pin soldered between the patch and ground. The radial line along which the feed is placed determines the direction of the linear polarization. The nonuniform radiation along its edge gives a larger edge impedance than the square patch. Experience shows that the 50- feed point is located from the center at about one-third the radius. Experiments, actual or numerical, will be required to locate the proper point. Use a network analyzer with a Smith chart display to measure the input impedance. If the resonance circle swings around the origin, the impedance is too high (overcoupled). Move the feed toward the center. A scalar return-loss display cannot give you the direction of movement required. Like the rectangular patch, mismatching the impedance at center frequency to about 65 increases the bandwidth slightly. Derneryd [20] gives an approximate expression for the radial impedance variation: Rin = Re

J12 (kε ρ) J12 (kε a)

(6-32)

where Re is the edge resistance, ρ the radial distance, and J1 the Bessel function of the √ ﬁrst kind. kε is the propagation constant in the substrate dielectric constant: kε = k εr . Figure 6-25 gives the 2 : 1 VSWR bandwidth of a circular patch on various substrates as a function of the substrate thickness. It has a slightly smaller bandwidth than that of a square patch because it has a smaller volume. The curves on Figure 6-25 include surface-wave radiation (or losses).

316

MICROSTRIP ANTENNAS

2:1 VSWR Bandwidth, %

2.94

6.0

2.2

9.8

4.5 1.0

Substrate Thickness, l (Free Space)

FIGURE 6-25 2 : 1 VSWR bandwidth of circular microstrip patches versus substrate thickness in free-space wavelengths, including surface-wave radiation.

6-6 CIRCULARLY POLARIZED PATCH ANTENNAS Figure 6-26 show methods of achieving circular polarization with square patches fed with two inputs. The patches are fed by equal signals 90◦ out of phase. The branchline hybrid (Figure 6-26a) consists of four transmission lines connected in a square. The hybrid shown (100- system) produces equal outputs 90◦ out of phase at center frequency. The two inputs produce patterns with opposite senses of circular polarization. Both the VSWR and axial ratio bandwidths far exceed the singly fed patch bandwidth. Reﬂections due to the patch mismatch are routed to the opposite input. Patch input reﬂections, undetected at the input, reduce the efﬁciency of the antenna by the same amount as the singly fed patch mismatches. The antenna can be fed from below in two places by using a coupled line hybrid, but it suffers from the same efﬁciency problem. The cross-fed antenna (Figure 6-26b) splits the signal to feed both edges. A quarterwavelength-longer line provides the extra 90◦ phase shift to give circular polarization. Shifting the impedance from one input through a quarter-wavelength line before adding the two in shunt cancels some of the reﬂection from the second line and increases the impedance bandwidth. The impedance bandwidth approximately doubles compared to the singly fed patch. The 6-dB axial ratio bandwidth roughly equals the singly fed square-patch bandwidth. The polarization loss (0.5 dB) of a 6-dB axial ratio equals the 2 : 1 VSWR mismatch loss. The antennas in Figure 6-27 use asymmetries to perturb the resonance frequencies of two possible modes and achieve circular polarization [21]. The approximately square patches have been divided into two groups: type A, fed along the centerline, and type B, fed along the diagonal. All these antennas radiate RHC. We can understand the operation of these patches from an analysis of the turnstile dipole antenna (Figure 6-28). The orthogonal dipoles could be of equal length and fed from a 90◦ hybrid to achieve

CIRCULARLY POLARIZED PATCH ANTENNAS

l/4

317

100 Ω

RHC 100 Ω

l/4

LHC 70.7 Ω

(a)

−90°

LHC

∆ ∆ + l/4

0° Cross-fed patch

l/4

(b)

FIGURE 6-26 Dual-fed circularly polarized patch antennas: (a) branchline hybrid fed; (b) cross-fed patch.

FIGURE 6-27 Classes of perturbed microstrip patches to generate circular polarization from a single feed. (From R. Garg et al., Microstrip Patch Handbook, Fig. 8-15, 1999 Artech House, Inc.)

circular polarization (like the patch in Figure 6-26a). Instead, the lengths are changed to shift the phase of each dipole by 45◦ at resonance. If we lengthen the dipole beyond resonance, the input impedance becomes inductive. The current becomes I=

V V (R2 − j X2 ) = R2 + j X2 R22 + X22

318

MICROSTRIP ANTENNAS

FIGURE 6-28 Turnstile dipole antenna.

The radiated ﬁeld phase decreases relative to the resonant-length dipole. Shortening the dipole from resonance increases the far-ﬁeld phase. We adjust the lengths until the phase difference of the radiated ﬁelds is 90◦ and the susceptances from the two dipoles cancel at center frequency. The combination of the two modes produces a Smith chart response with a small loop or kink (see Figure 5-13). The best circular polarization occurs at the frequency of the kink, and the response degrades below and above this frequency. The axial ratio bandwidth is far less than the impedance bandwidth, because the combination of the two modes causes a cancellation of transmission-line reﬂections from the two modes and increases the impedance bandwidth. The phase required for good circular polarization changes rapidly. We denote the total change in area S to achieve two resonances for a normal patch area of S and it is proportional to the Q. A type A patch, fed along the square patch axis, requires less area change than a type B patch, fed along the diagonal: type A:

1 S = S 2Q

type B:

S 1 = S Q

(6-33a,b)

We achieve the same effect with a patch by perturbing the lengths of a square patch and feeding both polarizations. An input along the diagonal (type B) feeds all edges in two separate resonances. The ratio of the edge lengths is found in terms of Q by a perturbation technique [4]. We rearrange Eq. (6-33b) to derive the ratio of these lengths: b 1 =1+ (6-34) a Q We calculate resonant frequencies for the two lengths from Eq. (6-34): f0 1 f2 = f0 1 + f1 = √ Q 1 + 1/Q

(6-35)

COMPACT PATCHES

319

Q is related to the VSWR bandwidth by Eq. (6-7). The 3-dB axial ratio bandwidth of the antenna is limited to 35%/Q or 35% of the frequency difference between f1 and f2 . Example Compute resonant lengths for a corner-fed patch on a 1.6-mm substrate with εr = 2.55 at 3 GHz. We have λ = 100 mm and thickness/λ = 0.016. From Figure 6-7 we read the 2 : 1 VSWR bandwidth: 1.61%. From Eq. (6-7) we calculate Q: Q=

1

√ = 43.9 0.0161 2

We use Eq. (6-35) to determine the resonant frequencies: 3 = 2.966 GHz 1 + 1/43.9 f2 = 3 1 + 1/43.9 = 3.034 GHz

f1 = √

By using the techniques of Section 6-3, we calculate the resonant lengths: a = 30.27 mm, b = 31.01 mm. All perturbations by small areas in a circular patch can only be type A feeding. The perturbation equations are related to the circular patch separation constant X11 (1.84118): 1 S 2 S = type B: = (6-36) type A: S X11 Q S X11 Q A circular patch perturbed into an elliptical patch radiates circular polarization when fed on a 45◦ diagonal from the major or minor axis and produces type B feeding. The ratio of major to minor axes is related to Q [4]: b 1.0887 =1+ a Q with resonant frequencies f0 f1 = √ 1 + 1.0887/Q

and f2 = f0 1 +

1.0887 Q

(6-37)

We compute Q by using Eq. (6-7) and read the bandwidths from Figure 6-25 for circular patches. Use the techniques of Section 6-5 to calculate the physical radius of the major and minor axes from the frequencies [Eq. (6-35)]. 6-7 COMPACT PATCHES The desire to produce small patches for cellular telephone handset use has lead to the development of compact designs. The ideal antenna is one whose location the user is unaware of and which is as small as possible. Because most signals arrive at the user

320

MICROSTRIP ANTENNAS

after many bounces and edge diffractions, polarization is arbitrary. We do not need to control the radiation pattern or its polarization carefully and it opens up a range of possibilities. Shorting pins placed close to the feed pin reduce the patch size to about 1 λ on a side, but its polarization is poorly controlled. If we can force the current to take 8 a longer path along the resonant-length path, we can shrink the overall size. We etch notches in the patch to make the current wander or use various spiral-wound networks on a ﬂat substrate. Three-dimensional solutions consist of folding a patch by using the vertical direction or some sort of winding around a cylinder. Many variations on these ideas appear in the literature and in collections of these ideas [16,22,23]. Adding a shorting pin closely spaced to the vertical feed pin (Figure 6-29) greatly reduces the resonant frequency of a given-size patch and produces a compact patch [24]. The idea is to make the current ﬂow over a longer path from the feed point to the radiation site; in other words, the transmission line has been folded to make the path longer in the resonant cavity. We use this concept for all compact patches. In this conﬁguration the resonant wavelength is found from the patch perimeter. Given the width W and the length L of the patch on a dielectric substrate εr , the resonant wavelength is given by √ λ0 = 4 εr (L + W ) (6-38) which reduces to a square patch λ/8 on a side. This patch has one-half the length and one-fourth the area of a quarter-wave path, with its short circuit along an entire √ edge. The circular shorting pin compact patch resonant diameter equals 0.14λ0 / εr . Making a patch this small produces highly inductive input impedance, which we can see by looking at the Smith chart of a coaxial probe-fed patch (Figure 6-12). The curve sweeps clockwise as the frequency increases. At low frequency (or small size) the patch is highly inductive. Figure 6-12 shows that using thicker substrates to increase bandwidth makes the patch impedance even more inductive. The shorting pin next to the feed pin forms a transmission line with it and adds a capacitive component to the input impedance that counteracts the patch and feed pin inductance. As the shorting pin is moved farther away from the feed pin, the capacitance decreases and the shorting pin becomes an inductive component, as it is in the quarter-wave patch. y

y

(xps, yps) R

x

x

W

(xp, yp)

(xp, yp)

(xps, yps) L

(a) z q Shorting Post εr

d

(b)

FIGURE 6-29 Compact patch with a shorting pin near the feed. (From [24], Fig. 1, 1998 IEEE.)

COMPACT PATCHES

321

The recommended position of the shorting pin is 80 to 90% of the distance from the center to the outer edge and a diameter of 0.008λ. You will need to iterate the position of the feed probe, usually one-half the diameter of the shorting pin, to achieve an impedance match. Table 6-5 lists the bandwidth achieved versus substrate thickness on foam, εr = 1.07 [25, p. 207]. Figure 6-30 gives the calculated pattern of a shorting pin compact patch on 0.034λ free-space substrate. The broad E-plane pattern has 10-dB dip on the broadside matched by the H -plane Eφ component. The large current in the shorting pin produces a significant monopole pattern seen in the Eθ radiation in the H -plane. This small antenna is a combination of a top-loaded monopole and a patch. Thinner antennas have a lower pattern dip broadside to the substrate because the monopole is shorter. The planar inverted F antenna (PIFA) is similar electrically to the shorting pin compact patch. We move the shorting pin to one corner and often make it a small shorting plate. We locate the feed pin close to the small shorting plate to again form a transmission line whose capacitance with the feed pin counteracts the inductive component of the small patch. We use Eq. (6-38) to determine its resonant wavelength. If we rotate the coordinates so that the shorting plate and diagonal lie on the x-axis, we obtain the pattern response of Figure 6-30. Since there is practically no difference between the two antennas, Table 6-5 gives the bandwidth of the PIFA versus thickness [26]. TABLE 6-5 Bandwidth of a Single Shorting Pin Compact Patch Thickness (λ0 ) 0.01 0.02 0.03 0.04 0.05 0.06

Bandwidth, 2 : 1 VSWR (%)

Feed-to-Pin Center Distance (λ0 )

1.6 2.2 2.7 3.4 4.3 5.7

0.0071 0.0076 0.0081 0.0085 0.0101 0.0135

E-Plane

H-Plane X-Pol

H-Plane

FIGURE 6-30 Pattern of a compact patch.

322

MICROSTRIP ANTENNAS

(b)

(a)

FIGURE 6-31 Reduced-size microstrip patches using meandered current paths. (From [22], Fig. 1-3, 2002 John Wiley & Sons, Inc.)

Modest size reduction can be obtained by making the currents ﬂow along a longer path along the resonant length. Figure 6-31 shows two planar antennas where slits cut from the width sides and disrupting the resonant-length path cause wandering of the current. The bowtie patch also makes the current path longer. These antennas radiate normal patch patterns with broader beamwidths in the E-plane because the notches bring the radiating edges closer together. The antennas in Figure 6-32 shrink the resonant length by folding the antennas vertically. The total length along the path is approximately λ/2, but the radiating edges are closer together. A large number of variations using slots have been investigated and offer interesting approaches to both shrink the patch size and produce dual-frequency antennas by using both the patch mode and slot radiation [22]. Ground Plane

Bent Edge

Air-Substrate Thickness ( a) Ground Plane

Folded Edge (b) Ground Plane

Double-Folded Edge (c)

FIGURE 6-32 Folding microstrip patches to reduce size. (From [22], Fig. 1-4, 2002 John Wiley & Sons, Inc.)

DIRECTLY FED STACKED PATCHES

323

6-8 DIRECTLY FED STACKED PATCHES Figure 6-7 illustrates the limited impedance bandwidth achievable from a singleresonator microstrip patch. When we increase the substrate thickness to widen the bandwidth, the antenna excites more surface waves (Figure 6-8) difﬁcult to control, and we accept them as losses. In Section 6-3 we discussed the use of external circuit elements to improve the impedance response. These have limited usefulness, although the simple series capacitor input to overcome the inductance of a long feed probe and the inductive nature of the higher-order modes is easily implemented. These external elements add poles to the resonant circuit to increase bandwidth. We can increase the number of poles by adding antenna elements instead. One solution is to couple to additional patches located around the fed patch on the same substrate surface. This increases the antenna size and reduces pattern beamwidths. This solution is difﬁcult to use in an array because the large spacing between elements produces grating lobes. Stacking patches vertically above the driven patch and coupling to them electromagnetically produces the best solution in terms of pattern response. The disadvantage of this approach is the additional fabrication cost. Our discussion of aperture-coupled patches in Section 6-3 points out that large apertures also add resonant poles that can increase the bandwidth. These added resonant elements complicate the design and call for the application of analytical tools instead of a cut-and-try approach. Although either patch in a two-element stacked patch design can be fed, feeding the lower element produces a design with minimum feed pin inductance. Aperture coupling through the ground plane feeds the lower patch directly as well. If we use an edge feed, we want the input transmission line to be as narrow as possible to reduce radiation by feeding the lower patch. Initially, we consider the probe-fed stacked patch [27]. A coaxial probe feeds directly a lower substrate of thickness d1 and dielectric constant εr1 through a hole in the ground plane. Figure 6-12 shows that the feed probe adds inductance for a thick substrate and the resonant loop is located on the upper inductive portion of the Smith chart. When we couple the lower patch to an upper patch with thickness d2 and dielectric constant εr2 , its circuit response becomes more inductive. We need to start with the impedance locus of the lower patch to be capacitive without the upper patch. This can be achieved by using an overcoupled feed. Figure 6-11 illustrates the overcoupled patch whose impedance locus sweeps around the origin of the Smith chart. The inductance of the feed probe rotates these curves clockwise around the center of the chart and the overcoupled response has signiﬁcant capacitive reactance when it sweeps around the origin. If we matched the lower patch critically, upward movement of the locus due to the coupled patch would reduce the impedance bandwidth. Figure 6-33 illustrates these design steps. Figure 6-33b shows that increasing the lower patch thickness leads to a longer feed probe that sweeps across the center from a more inductive portion of the Smith chart. Adding the second patch fails to increase the bandwidth relative to the thinner optimum lower patch. The thickness of the upper patch substrate d2 controls the tightness of the resonant loop. A greater thickness d2 produces a tighter loop in the Smith chart response that leads to a lower VSWR over a narrower bandwidth. Remember that we cannot use Eq. (6-7) to determine the bandwidth for different VSWR levels because we now have multiple resonators. If we use a foam upper substrate, the dielectric constant and thickness of the lower substrate determines the surface-wave efﬁciency. Waterhouse [27] used a dielectric

324

MICROSTRIP ANTENNAS Single layer + Stacked

+

+ ++ +

++

+ + + +

+

+++ ++ ++ + + ++ + + ++ + +

(a)

Single layer + Stacked

+ + + + + + + + + ++

+ ++ + ++ ++ ++ ++ +++ + + + + + + + + + + + +++ + + ++ + + ++ + + + +

(b)

FIGURE 6-33 Effect of coupling to a second patch: (a) overcoupled single lower patch response forms resonant loop with the second coupled patch; (b) increasing lower patch thickness causes rotation on a Smith chart and lower bandwidth. (From [27], Fig. 3, 1999 IEEE.)

constant of 2.2 for the lower substrate with a thickness of 0.04λ0 and a foam upper substrate 0.06λ0 to achieve optimum bandwidth with acceptable surface-wave losses. The lower patch was overcoupled so that it swept through the 250- resistance point at resonance. Since the impedance locus sweeps clockwise on the Smith chart as frequency increases, this resonant point should be slightly below the lower end of the desired frequency band. We adjust the second substrate thickness to move the resonant loop on the Smith chart in the vertical direction. As we increase the size of the upper patch, the loop moves around an arc in the clockwise direction, which we use to center the impedance response on the Smith chart for optimum bandwidth. This method produces impedance bandwidths of around 25%. The pattern bandwidth exceeds this bandwidth and we expect little change in pattern over this frequency range. Another successful stacked patch fed from a coaxial probe is the hi-lo conﬁguration, in which a high dielectric substrate (εr1 = 10.4) is used for the lower substrate and a foam (εr2 = 1.07) for the upper substrate [25, pp. 178–182]. The upper patch captures the surface wave of the lower patch and greatly improves the overall efﬁciency by radiating this power in a space wave. Although the two patches have different sizes, the coupling remains sufﬁcient to produce a broadband antenna with impedance bandwidths approaching 30%. In this design the lower patch is designed for the high dielectric of the lower substrate with little consideration for the upper patch except for making it a little overcoupled. The upper patch can be designed using the substrate thickness and dielectric constant assuming that the high dielectric substrate acts as the ground plane. When we mount the upper patch over the smaller lower patch, small adjustments must be made to the dimensions to achieve a 50- impedance match. The example given used a lower substrate thickness of 0.032λ0 with εr1 = 10.4, and by Figure 6-8 would have −1.3-dB surface-wave loss. Locating the second patch on a 0.067λ0 -thick foam substrate directly over the ﬁrst patch reduced the surface-wave loss to better than −0.7 dB over the entire band.

APERTURE-COUPLED STACKED PATCHES

325

6-9 APERTURE-COUPLED STACKED PATCHES The discussion on aperture feeding of a patch in Section 6-3 stated that we can utilize the aperture as another resonator to broadband the antenna. Figure 6-34 shows the stacked patch antenna fed from an aperture. In this implementation we make the coupling slot long enough to be one of the resonators, which increases the number of resonators to three: the aperture, the lower patch, and the upper patch. We must use element spacing to control coupling because frequencies control resonator sizes. By careful control of parameters two loops will form in the Smith chart response of impedance and be made to wrap tightly around the center of the chart [28] as shown in Figure 6-35b. We form these loops by coupling resonators. Undercoupling produces small tight loops; overcoupling produces large loops. Figure 6-35 illustrates the effect of aperture size. The left Smith chart shows undercoupling between the aperture and the lower patch by the small left loop. We increase the coupling by increasing the aperture slot length (Figure 6-35b) or by increasing the

Patch 2(PL2, PW2, Layer N2)

Layer N(εrN, dN, tan dN)

Layer N2(εrN2, dN2, tan dN2) Patch 1(PL1, PW1, Layer N1)

Layer N1(εrN1, dN1, tan dN1) Layer 1(εr1, d1, tan d1)

Ground Plane Aperture(SL, SW) Feedline(Wf, doff, Lstub)

Feed Substrate (εrf, df, tan df)

FIGURE 6-34 Construction of a resonant aperture coupled dual patch in exploded view. (From [28], Fig. 1, 1998 IEEE.)

(a)

(b)

(c)

FIGURE 6-35 Effect on increasing slot length, SL, of apertures in stacked dual patches: (a) SL = 8 mm; (b) SL = 10 mm; (c) SL = 12 mm. (From [28], Fig. 4, 1998 IEEE.)

326

MICROSTRIP ANTENNAS

lower patch size or reducing the lower patch thickness. The best results have been obtained by having the lower-frequency (left) loop determined by the lower patch and aperture. Overcoupling the aperture to the lower patch produces the impedance locus of Figure 6-35c. We control the upper loop size by varying the upper patch size, the relative size between the two patches, and the upper substrate thickness. The lower patch size is a critical parameter because it affects the coupling and size of both loops while shifting their center frequencies. For ﬁxed sizes of the other two resonators, decreasing the lower patch size decreases the coupling to the aperture while increasing the coupling to the upper patch. Increasing aperture size increases coupling to the aperture and decreases coupling to the upper patch. By remembering that overcoupling produces larger Smith chart loops, we determine in which direction to change parameters by observing changes in analytical results on the Smith chart to produce optimum designs. Because the slot aperture is one of the three resonators, we cannot vary its length to determine coupling to the lower patch. The overcoupled large slot produces high resistance at the microstrip input. We can lower this impedance by offset feeding the slot or by using a wide transmission line. A single offset line will unbalance the ﬁelds in the slot and lead to unbalanced excitation of the patches. This unbalanced excitation on the patches increases cross-polarization. The dual balanced offset feeding shown in Figure 6-36, where we join the two lines in a reactive power divider, both lowers the resistance and balances the patch excitation. A design using rectangular patches for a single linear polarization achieved a 67% 2 : 1 VSWR bandwidth [28]. The only signiﬁcant problem with the design is the poor front-to-back ratio, which is reduced to 6 dB at the upper frequencies as the aperture radiation increases. Placing a reﬂector patch below the microstrip feed line, it can be sized to reduce the F/B ratio by forming a Yagi–Uda antenna with the stacked patch [29]. Figure 6-37 illustrates an exploded view of a dual polarized aperture stacked patch. The potential bandwidth shrinks because we lose width as a parameter with square patches to optimize impedance. The key element of this design is the feed crossed slot [30]. The crossed-slot feeding aperture is located on a ground plane shared by microstrip networks located below and above the aperture. Each network consists of a reactive power divider to raise the impedance of the feed lines and allow offset feeding of the slot for each polarization. The balanced feed reduces cross-polarization and cross coupling between the two ports that would occur in both the crossed slot

2doff

Low Impedance Line Lstub

50Ω Line

Lstub

100Ω Line

50Ω Line (a)

(b)

FIGURE 6-36 Impedance matching for resonant aperture dual stacked patches: (a) wide transmission line; (b) dual offset feed. (From [28], Fig. 3, 1998 IEEE.)

PATCH ANTENNA FEED NETWORKS

Patch 2 (W2, L2)

327

εr8, h8

W

εr7, h7

L

εr6, h6 Patch 1 (W1, L1)

εr5, h5

Feed 2 (W2F, L2STUB)

εr4, h4

Slot 1 (W1SL, L1SL) Slot 2 (W2SL, L2SL)

L1STUB

εr3, h3 εr2, h2

Feed 1 (W1F, L1STUB) εr1, h1 Reflector (W1REF, L1REF) Reflector (W2REF, L2REF)

FIGURE 6-37 Exploded view of construction of a dual polarized aperture-fed stacked patch utilizing a crossed strip reﬂector. (From [8], Fig. 3.6.22, 2003 Kluwer Academic Publishers.)

and the patch elements. This shows that the slot that couples to a patch resonator can be fed by a microstrip line located either below or above the slot. The ground plane between the two networks for each polarization eliminates direct coupling between the microstrip networks and symmetrical feeding reduces coupling in the slot. Because we use long slots to feed the lower microstrip patch in an overcoupled excitation, direct coupling of the upper microstrip to the lower patch is minimal in comparison. We use thin substrates of moderate dielectric constant (εr = 2.2) to support the etched patches and foam layers between to separate the patches to increase bandwidth and control coupling. Figure 6-37 shows a crossed dipole used as a reﬂector element below the microstrip feed lines to reﬂect direct radiation from the crossed slot that reduces the F/B ratio. 6-10 PATCH ANTENNA FEED NETWORKS Patch antenna arrays may be fed from below (Figure 6-9) by using a stripline distribution network. The connections between the boards greatly complicate the assembly. A connection made vertically from the center strip of a stripline unbalances the ﬁelds and induces parallel-plate modes. Shorting pins between the ground planes suppress this mode. It is far easier to etch the feed network on the microstrip and use either edge feeds or aperture feeds with the network located below the patch layer. Feed networks radiate very little in comparison with the patches when etched on the same substrate because radiation from fringing ﬁelds on the two sides of the microstrip lines cancel each other except at discontinuities (corners and steps). Consider the equally fed array (Figure 6-38). Equal amplitude and phase feeding generates virtual magnetic walls between the patches as shown. We can join the edges

328

MICROSTRIP ANTENNAS Magnetic Walls

l/4 l/4

Transformer 50 Ω

100 Ω 70.7 Ω 100 Ω

FIGURE 6-38 Equally fed microstrip patch array.

between the patches without effect, since the midpoint remains a virtual open circuit and the separate patches join into a continuous strip. The feeds must be spaced close enough together to prevent grating lobes and to provide uniform amplitude along the edges. These antennas can be wrapped around missiles to provide omnidirectional coverage about the roll axis. To eliminate pattern ripple, feeds must be spaced about every 0.75λ in a circular array. The resistance at each feed at resonance will be the combination of the radiation conductances from the portions of the edges between the magnetic walls. Figure 6-38 illustrates an equally fed four-element array. Starting from the patch, a quarter-wavelength transformer reduces the roughly 200- impedance to 100 . Two 100- lines join in shunt to 50 at their juncture. A 70.7- quarter-wavelength line transforms the 50 back to 100 . We continue this sequence for any 2N -array for reactive power dividers at each junction. Equal path lengths from the input excite them with equal phases. Arrays with the number of elements different from 2N -are possible, but they require more difﬁcult feed networks. A 100- system was picked because 50- lines on low-dielectric-constant substrates are quite wide. The reactive power divider (Figure 6-38) has more bandwidth than the patch while it is matched at the input but not at its outputs. The network can be analyzed by using even and odd modes and shows that the output return loss is 6 dB, and it provides only 6 dB of isolation between outputs. The power reﬂected from a damaged antenna distributes to the other elements of the array and produces an effect greater than that of just a missing element. Making power dividers with isolation resistors reduces this problem, but we cannot justify the added difﬁculty of mounting resistors when both good etchings and low probability of damage make them unnecessary. We must be wary of coupling between different parts of the feed network. We want to pack the feed network into the smallest area, but coupled signals between the lines produce unexpected anomalies. Distinguishing direct radiation from the feed and coupling redistribution is difﬁcult. Although couplings are predictable, they appear as random errors when we cannot perform a full analysis. Unfortunately, the coupling between microstrip lines falls off quite slowly. Table 6-6 lists the coupling and peak errors for 100- lines; those of 50- lines are very similar. We read the amplitude and phase errors from Scales 1-8 and 1-9.

SERIES-FED ARRAY

329

TABLE 6-6 Peak Feed Errors Due to Microstrip Coupling for 100- Lines (εr = 2.4) Spacing/Substrate Thickness 1.0 2.0 3.0 4.0 5.0

Coupling (dB)

Amplitude Error (dB)

Phase Error (deg)

16 23 28 32 35

1.5 0.7 0.4 0.22 0.12

9.0 4.0 2.3 1.5 1.0

6-11 SERIES-FED ARRAY If we reduce the width of the patch, the radiation conductance is insufﬁcient to match the input. We can use the microstrip patch as a transmission line and connect a line opposite the feed to lead to other patches (Figure 6-39). If we space the patches by halfwavelengths, the impedances of the patches will add in phase at the input, because it rotates once around the Smith chart in λ/2. The characteristic impedance of the connecting lines has no effect at center frequency. The junction of the transmissionline feeder and the patch introduces excess phase shift. In arrays of a few elements, the extra phase shift can be ignored, but arrays with a large number of elements, or when we design for critical amplitude taper, must account for δ. Of course, traveling-wave or resonant arrays can be designed. The frequency dispersion of the traveling-wave array can be used to frequency-scan the beam. Various experimental methods have been devised to measure the parameters of the series array. Metzler [31] performed experiments on uniform-width element arrays to determine the radiation conductance and excess phase shift. Measuring the transmission loss through the array as a network with input and output connectors determines the radiation conductance of the patches. An empirical equation was obtained:

W G = 0.0162 λ0

1.757 0.033 ≤

W ≤ 0.254 λ

(6-39)

where G is the total radiation conductance of each patch, with half from each edge. Measurement of the beam direction of the uniform traveling-wave array determines the excess phase shift in each patch. Jones et al. [32] model the patch (Figure 6-39) with extensions due to the fringing ﬁelds as a transmission line: L + 2 long. The other excess phase shift, due to the step, is modeled as extensions to the input lines (δ). Jones et al. perform measurements on single elements to establish these lengths. is found from the resonant frequency of the √ patch: L + 2 = λ/2 εeff , where εeff is given by Eq. (6-19). When the transmissionline phase is measured through the patch at resonance, the excess phase beyond π is equated to a phase shift length in the narrow feeder lines: 2δ =

λN φexcess 2π

where λN is the wavelength in the narrow line.

330

MICROSTRIP ANTENNAS

(a)

(b)

(c)

FIGURE 6-39 Series-fed patch and its equivalent circuit. (From [32], Fig. 2, 1982 IEEE.)

When designing the array, we vary the widths of the patches to achieve the desired √ amplitude taper. The voltage distribution at each patch is given by V g, where g is the patch conductance. Standing-wave (resonant) arrays require that the sum of the conductances be equal to the input conductance desired. We have some latitude when we feed the array through a quarter-wavelength transformer. The nonresonant array requires a matched load on the end to prevent standing waves. We must pick the ratio of the power dissipated to the radiated power that gives us an extra parameter with which to optimize the design. We control the beam direction by spacing the elements to achieve the phase shift required. 6-12

MICROSTRIP DIPOLE [33]

As the width W of a patch narrows, the input impedance increases. When the width approaches that of a microstrip feed line, either the patch fails to be a resonator or the

331

MICROSTRIP DIPOLE

feed line becomes very narrow in trying to transform the impedance. The microstrip dipole solves these problems by having a coupled line feeder. The dipole is a halfwavelength strip whose width equals that of a microstrip feed line. A line etched on a substrate below feeds the dipole by coupling into the strip (Figure 6-40a). The equivalent circuit (Figure 6-40b) transforms the high impedance of the dipole through the unequal coupled lines. By varying the coupling, we can change the input impedance at resonance. Best results occur for quarter-wavelength overlap where the equivalent stubs (Figure 6-40b) do not contribute reactance. We vary the coupling by changing the thickness of the substrate between the strips or by offsetting the lower strip. The dipole radiates as a narrow patch and not as a dipole. No pattern nulls appear along the axis of the strip, but they occur more strongly in the direction of the equivalent magnetic currents of the edges. The H -plane pattern becomes quite broad for the narrow strip width. The feed distribution circuit is etched on the substrate below the dipoles. With the feed circuit on a separate level, we have greater freedom in the feed network design to excite desired distributions. Also, because the dipoles are small, we can use density tapering of the dipoles to that end. Proper design requires measurement [34] Overlap Region Microstrip Feed Line

L/2

L

W

Ground Plane

Substrates (a)

a

Zoo = vLa =

Cb vF

b

Zoo = vLb =

q a

Cb vF

q

q

a

Zoe, Zoo 1

a

=

1

a

Zoe − Zoo = vLab = Cab vF 2

2

2 b

b

Zoe, Zoo

(b)

F = CaCb + CaCab + CbCab

FIGURE 6-40 (a) Microstrip dipole; (b) equivalent circuit. [(b) From G. L. Matthaei et al., Microwave Filters, Impedance Matching Networks, and Coupling Structures, 1980 Artech House, Inc.]

332

MICROSTRIP ANTENNAS

to obtain the desired effect, since mutual coupling will change the distribution by changing the active impedance of each dipole. The feed network must compensate for the coupling. 6-13

MICROSTRIP FRANKLIN ARRAY [35]

An electrically long line with a standing wave on it fails to radiate on broadside because the many cycles cancel each other. We obtain a pattern with many nulls and lobes. By folding the lines with out-of-phase standing-wave currents close together, we can prevent their radiation. The other portions are free to radiate (Figure 6-41a). The Franklin array consists of straight sections λ/2 long connected by λ/4 shorted stubs. The standing-wave currents on the straight portions add in phase. We can construct a microstrip version (Figure 6-41b). Half-wavelength lines act as radiators (patches). We connect them with half-wavelength lines folded into stubs so that the counteracting standing-wave currents do not radiate. The straight lines are narrow patches. The total radiation conductance of each strip is 1 W 2 G= (6-40) 45 λ for narrow strip widths W , where λ is the free-space wavelength. Using lines for the stubs whose impedance is twice the radiating strip impedance reduces unwanted internal reﬂections. The two stubs add in shunt. Since the antenna is quite narrowband and the length of the lines between patches is a half-wavelength long, the impedance of these connecting arms has a secondary effect. Example Design an eight-wavelength array at 10 GHz. There are 16 patches in the array. The radiation conductances add for elements spaced at λ/2 intervals. For a 100- input, each patch supplies a conductance 0.01/16. We solve Eq. (6-40) for the width: 0.01(45) W =λ = 0.168λ 16 l/2

l/2

l/2

l/2

l l

l/4

l (a)

L

l/2

W

Connecting line

Patch

(b)

FIGURE 6-41 (a) Dipole and (b) microstrip Franklin arrays.

MICROSTRIP ANTENNA MECHANICAL PROPERTIES

333

If we use Eq. (6-39) from the series patch, we obtain W = 0.157λ, within the range of the empirical formula. For 10 GHz, W = 4.71 mm. On an 0.8-mm substrate (εr = 2.21), W/H = 5.89 and the impedance of the strip radiator Z0 = 44.01 . We need to ﬁnd the effective dielectric constant of the strip to determine the patch length and impedance. From Eq. (6-19), εr = 1.97. We calculate the cutback from each end by using Eq. (6-18); = 0.40 mm. Each radiating strip is 300 × 109 L= − 2(0.40) = 9.88 mm √ 1010 (2) 1.97 √ The radiating-strip impedance is Z0 / εeff = 31.3 . We need 62.6- connecting lines in the stubs to achieve the broadest bandwidth. With so few radiators, we could use 100- connecting arms with little change in bandwidth and have more reasonable connecting arm widths: 0.71 mm. The example shows that the microstrip Franklin array works best for high frequencies or long arrays. The elements are narrow, and the interconnecting arms are thin.

6-14 MICROSTRIP ANTENNA MECHANICAL PROPERTIES A microstrip patch antenna has very desirable mechanical properties. It can withstand tremendous shock and vibration. Because the antenna is on a solid substrate, the patch cannot ﬂex, and small changes in the substrate thickness have only a minor effect on the resonant frequency. The commonly used soft substrate (Teﬂon and ﬁberglass) has a good damped resilience. Microstrip patch antennas have been used to telemeter data from artillery shells and high-velocity rockets, which have high shock and vibration levels. The repeatability of the dimensions of the patches depends only on the etcher’s art. Complicated shapes and feed networks are produced as cheaply as simple ones. The antennas can withstand exposure to high temperatures when covered by a radome made of the same soft dielectric as the substrate. The cover protects the metal patches but has only a minor effect on the resonant frequency [36]. High temperatures on the surface of the radome or ablation fail to change the resonance signiﬁcantly because the radome itself has only a minor effect. Variation in the dielectric constant of the substrate from lot to lot causes problems with repeatability. The narrowband antennas require measurement of the dielectric constant of each lot, and sometimes of each sheet, to get the center frequency desired. A series of etching masks can be made to cover the expected range. The antennas can be tuned with inductive shorting pins or capacitive screws, but tuning is prohibitive when the number of elements in an array is large. Careful quality control of the dielectric constant is the answer. Close monitoring of the etching process may also be needed to prevent excessive undercutting. Temperature variations can be a problem with thin substrates when the bandwidth is narrow. The patch and substrate size grow when the temperature rises, but they are overshadowed by the change in dielectric constant of soft substrates. Instead of decreasing the resonant frequency because of the increased patch size, a lowered dielectric constant raises the center frequency. Whenever we need more bandwidth than a microstrip patch can provide, we must turn to cavity antennas. We increase the antenna volume by penetrating the vehicle for the cavity, but we gain a design parameter.

334

MICROSTRIP ANTENNAS

REFERENCES 1. R. F. Harrington, Effects of antenna size on gain, bandwidth, and efﬁciency, Journal of Research, NBS, D, Radio Propagation, vol. 64D, January–February 1960, pp. 1–12. 2. R. C. Hansen, Fundamental limitations in antennas, Proceedings of IEEE, vol. 69, no. 2, February 1981, pp. 169–173. 3. Y. T. Lo et al., Study of microstrip antenna elements, arrays, feeds, losses, and applications, Final Report, RADC-TR-81-98, Rome Air Development Center, Rome, NY, June 1981. 4. K. R. Carver and E. L. Coffey, Theoretical investigation of the microstrip antenna, Technical Report PT-00929, Physical Science Laboratory, New Mexico State University, Las Cruces, NM, January 1979. 5. D. R. Jackson and N. G. Alexopoulas, Simple approximate formulas for input resistance, bandwidth, and efﬁciency of a resonant rectangular patch, IEEE Transactions on Antennas and Propagation, vol. 39, no. 3, March 1991, pp. 407–410. 6. R. F. Harrington, Time-Harmonic Electromagnetic Fields, McGraw-Hill, New York, 1961. 7. D. M. Pozar, Rigorous closed-form expressions for the surface wave loss of printed antennas, Electronics Letters, vol. 26, no. 13, June 21, 1990, pp. 954–956. 8. R. B. Waterhouse, ed., Microstrip Patch Antennas, A Designer’s Guide, Kluwer Academic, Boston, 2003. 9. R. E. Munson, Conformal microstrip antennas and microstrip phased arrays, IEEE Transactions on Antennas and Propagation, vol. AP-22, no. 1, January 1974, pp. 74–78. 10. E. O. Hammerstad, Equations for microstrip circuit design, Proceedings of the 5th European Micro-strip Conference, Hamburg, Germany, September 1975, pp. 268–272. 11. M. V. Schneider, Microstrip lines for microwave integrated circuits, Bell System Technical Journal, vol. 48, May–June 1969, pp. 1421–1444. 12. D. A. Paschen, Practical examples of integral broadband matching of microstrip antenna elements, Proceedings of the 1986 Antenna Applications Symposium, Monticello, IL. 13. T. Samaras, A. Kouloglou, and J. N. Sahalos, A note on impedance variation of a rectangular microstrip patch antenna with feed position, IEEE Antennas and Propagation Magazine, vol. 46, no. 2, April 2004. 14. D. M. Pozar, Microstrip antenna aperture-coupled to a microstrip-line, Electronics Letters, vol. 21, no. 2, January 1985, pp. 49–50. 15. P. L. Sullivan and D. H. Schaubert, Analysis of an aperture coupled microstrip antenna, IEEE Transactions on Antennas and Propagation, vol. AP-34, no. 8, August 1986, pp. 977–984. 16. G. Kumar and K. P. Ray, Broadband Microstrip Antennas, Artech House, Boston, 2003. 17. J. R. James, P. S. Hall, and C. Wood, Microstrip Antennas: Theory and Design, Peter Peregrinus, London, 1981. 18. L. C. Shen et al., Resonant frequency of a circular disk, printed circuit antenna, IEEE Transactions on Antennas and Propagation, vol. AP-25, no. 4, July 1977, pp. 595–596. 19. I. J. Bahl and P. Bhartia, Microstrip Antennas, Artech House, Dedham, MA, 1980. 20. A. G. Derneryd, Analysis of the microstrip disk antenna element, IEEE Transactions on Antennas and Propagation, vol. AP-27, no. 5, September 1979, pp. 660–664. 21. J. L. Kerr, Microstrip antenna developments, Proceedings of the Workshop on Printed Circuit Antennas, New Mexico State University, Las Cruces, NM, October 1979, pp. 3.1–3.20. 22. K.-L. Wong, Compact and Broadband Microstrip Antennas, Wiley, New York, 2002. 23. K.-L. Wong, Planar Antennas for Wireless Communications, Wiley, New York, 2003.

REFERENCES

335

24. R. B. Waterhouse, S. D. Targonski, and D. M. Kokotoff, Design and performance of small printed antennas, IEEE Transactions on Antennas and Propagation, vol. AP-46, no. 11, November 1998, pp. 1629–1633. 25. R. B. Waterhouse, ed., Microstrip Patch Antennas: A Designer’s Guide, Kluwer Academic, Boston, 2003. 26. T. Taga and K. Tsunekawa, Performance analysis of a built-in planar inverted-F antenna for 800 MHz band portable radio units, IEEE Transactions on Selected Areas in Communication, vol. SAC-5, no. 5, June 1987, pp. 921–929. 27. R. B. Waterhouse, Design of probe-fed stacked patches, IEEE Transactions on Antennas and Propagation, vol. AP-47, no. 12, December 1999, pp. 1780–1784. 28. S. D. Targonski, R. B. Waterhouse, and D. M. Pozar, Design of wide-band aperture-stacked patch microstrip antennas, IEEE Transactions on Antennas and Propagation, vol. AP-46, no. 9, September 1998, pp. 1245–1251. 29. S. D. Targonski and R. B. Waterhouse, Reﬂector elements for aperture and aperture coupled microstrip antennas, IEEE Antennas and Propagation Symposium Digest, Montreal, Quebec, Canada, July 1997, pp. 1840–1843. 30. J. R. Sanford and A. Tengs, A two substrate dual polarized aperture coupled patch, IEEE Antennas and Propagation Symposium Digest, 1996, pp. 1544–1547. 31. T. Metzler, Microstrip series arrays, Proceedings of the Workshop on Printed Circuit Antennas, New Mexico State University, Las Cruces, NM, October 1979, pp. 20.1–20.16. 32. B. B. Jones, F. V. M. Chow, and A. W. Seeto, The synthesis of shaped patterns with seriesfed microstrip patch arrays, IEEE Transactions on Antennas and Propagation, vol. AP-30, no. 6, November 1982, pp. 1206–1212. 33. D. A. Huebner, An electrically small dipole planar array, Proceedings of the Workshop on Printed Circuit Antennas, New Mexico State University, Las Cruces, NM, October 1979, pp. 17.1–17.16. 34. R. S. Elliott and G. J. Stern, The design of microstrip dipole arrays including mutual coupling, parts I and II, IEEE Transactions on Antennas and Propagation, vol. AP-29, no. 5, September 1981, pp. 757–765. 35. K. Solbach, Microstrip-Franklin antenna, IEEE Transactions on Antennas and Propagation, vol. AP-30, no. 4, July 1982, pp. 773–775. 36. I. J. Bahl et al., Design of microstrip antennas covered with a dielectric layer, IEEE Transactions on Antennas and Propagation, vol. AP-30, no. 2, March 1982, pp. 314–318.

7 HORN ANTENNAS

Horn antennas have a long history, traced in part in the collection of papers by Love [1] together with papers on every other horn topic. Horns have a wide variety of uses, from small-aperture antennas to feed reﬂectors to large-aperture antennas used by themselves as medium-gain antennas. Horns can be excited in any polarization or combination of polarizations. The purity of polarization possible and the unidirectional pattern make horns good laboratory standards and ideal reﬂector feeds. Horns also closely follow the characteristics predicted by simple theories. Horns are analyzed using a variety of techniques. Barrow and Chu [2] analyzed a sectoral horn, ﬂaring in only one plane, by solving the boundary value problem in the wedge. They expanded the ﬁelds in terms of Hankel functions in cylindrical coordinates. The ﬁelds form an equiphase surface over a cylindrical cap to which the Kirchhoff–Huygens equivalent current method [Eq. (2-23)] can be applied to compute the pattern. Similarly, Schorr and Beck [3] use spherical Hankel and Legendre functions to analyze conical horns. The integration surface consists of a spherical cap. Schelkunoff and Friis [4] use the mouth of the horn as the aperture and approximate the phase distribution as quadratic. Both aperture theories have the same valid pattern range. The method predicts patterns accurately in the area in front of the aperture. The error increases as the plane of the aperture is approached. The predicted pattern remains continuous and gives no indication of its increasing error. GTD methods [5] predict the pattern both in back and in front of the aperture while providing estimates of the error in the predictions. Most of the details needed for design can be obtained from the aperture theory. Only GTD predicts sidelobes accurately, since no assumption of zero ﬁelds outside the horn aperture is made. Figure 7-1 shows the general horn geometry. The input waveguide can be either rectangular or circular (elliptical). W is the width of a rectangular aperture, and a is the radius of a circular aperture. The distance from the junction of the projected sides Modern Antenna Design, Second Edition, By Thomas A. Milligan Copyright 2005 John Wiley & Sons, Inc.

336

RECTANGULAR HORN (PYRAMIDAL)

FIGURE 7-1

337

General geometry of a horn.

to the aperture is the slant radius R. The distance along the centerline from the aperture to the waveguide is the axial length. We derive the aperture ﬁeld amplitude from the input waveguide mode while the phase distribution is approximately quadratic across the aperture. We assume that the aperture ﬁelds radiate in spherical waves from the projected juncture of the sides, and the extra distance along the sides compared with the distance to the center of the aperture is given by =R−

R2 − a2

= R 1 −

1−

a R2 2

a2 ≈R 1− 1− 2R 2

=

a2 W2 = 2R 8R

We divide by wavelength to obtain the dimensionless constant S of the quadratic phase distribution: W2 a2 S= = = (7-1) λ 8λR 2λR Since the semiﬂare angle θ0 of most practical horns is small, we use the quadratic phase error approximation.

7-1 RECTANGULAR HORN (PYRAMIDAL) The rectangular horn ﬂares out of a rectangular or square waveguide with ﬂat metal walls. Figure 7-2 shows the horn geometry. The slant radiuses along the sides will be unequal, in general. The input waveguide dimensions are width a and height b.

338

HORN ANTENNAS

FIGURE 7-2 Rectangular horn geometry.

The aperture has width W in the H -plane and height H in the E-plane. Each aperture coordinate has its own quadratic phase distribution constant: Se =

H2 8λRe

Sh =

W2 8λRh

(7-2)

The TE10 mode of the lowest-order waveguide mode has the ﬁeld distribution Ey = E0 cos

πx a

Combining these ideas, the aperture electric ﬁeld is approximated by 2 πx 2y 2 2x exp −j 2π Se + Sh Ey = E0 cos W H W

(7-3)

The ratio of the electric and magnetic ﬁelds approaches the impedance of free space for large apertures. In this case we use the Huygens source approximation and need only the electric ﬁeld with Eq. (2-24) to ﬁnd the pattern. Small-aperture horns require Eq. (2-23) with an arbitrary ratio of the magnetic and electric ﬁelds. We compute the E-plane pattern by using a uniform aperture distribution and the H plane pattern from a cosine distribution. Both have a quadratic phase error. Figures 7-3 and 7-4 plot the E- and H -plane universal patterns in U -space of the Taylor distribution with S as a parameter. We can use them to determine the pattern of a general rectangular horn. Example Compute the pattern level at θ = 15◦ in the E- and H -planes of a horn with the following measured dimensions: Aperture: W (H -plane) = 28.9 cm, H (E-plane) = 21.3 cm Input waveguide: width a = 3.50 cm, height b = 1.75 cm

RECTANGULAR HORN (PYRAMIDAL)

339

0.9

0.8

S = 0.6

Relative Field Strength

0.7 S= 0.6

H2 8lRe

0.4 0.5

0.5 0.3 0.4 0.2

0.3

0.1 0.2 0 0.1 0 1

2 H/l sin q

3

4

FIGURE 7-3 E-plane universal pattern of a rectangular, TE10 mode.

The slant distance from the aperture to the waveguide along the center of each plate of the ﬂare was measured: Dh = 44.8 cm and De = 44.1 cm. We calculate the slant radius from similar triangles: Rh W = Dh W −a

H Re = De H −b

(7-4)

Slant radius: Rh = 50.97 cm, Re = 48.05 cm The frequency is 8 GHz (λ = 3.75 cm). Using Eq. (7-2), we compute Sh = 0.55 and Se = 0.31. We use Figures 7-3 and 7-4 to determine the universal pattern ﬁeld intensity (voltage): W H sin θ = 2.0 sin θ = 1.47 λ λ The ﬁelds from the ﬁgures are 0.27 (H -plane) and 0.36 (E-plane). We must include the obliquity factor of the Huygens source element pattern: (1 + cos θ )/2 to obtain the proper pattern level. At θ = 15◦ , the obliquity factor is 0.983. We calculate the pattern level in decibels from 20 times the logarithm of the product of the ﬁeld intensity from

340

HORN ANTENNAS

0.9 S=1 0.8 S=

Relative Field Strength

0.7

W2 8lRh

0.8

0.6 0.6 0.5

0.4

0.3 0.4 0.2 0.2 0.1 0 0 1

FIGURE 7-4

2 W/l sin q

3

4

H -plane universal pattern of a rectangular, TE10 mode.

the ﬁgures and the obliquity factor: H -plane : −11.5 dB

E-plane : −9 dB

We can calculate gain of this horn by using aperture efﬁciencies: H -plane (cosine) (Table 4-1) : 0.91 dB

E-plane (uniform) : 0 dB

These values hold for all rectangular horns excited by the TE10 mode. The quadratic phase distributions give us the phase error loss. From Table 4-42 we interpolate these losses: Sh = 0.55 cosine distribution

PEL = 2.09 dB

Se = 0.31 uniform distribution

PEL = 1.50 dB

The directivity is given by directivity = 10 log

4πW H − ATLh − ATLe − PELh − PELe = 22.9 dB λ2

The aperture efﬁciency is 35.5%, since ATL + PELh + PELe = 4.5 dB.

(7-5)

RECTANGULAR HORN (PYRAMIDAL)

341

We usually equate gain to directivity, since the wall losses are very small. Of course, for millimeter-wave horns we must include wall losses. Although we can use Table 4-42 along with the ﬁxed-amplitude taper loss of 0.91 dB to determine the aperture efﬁciency of a rectangular horn, Schelkunoff and Friis [4] give the following closed-form equation for the directivity: directivity =

8Rh Re {[C(u) − C(v)]2 + [S(u) − S(v)]2 }[C 2 (z) + S 2 (z)] WH

where 1 u= √ 2

√

λRh W +√ W λRh

and

x

C(x) = 0

1 v=√ 2

πt 2 dt cos 2

√

λRh W −√ W λRh

x

S(x) =

sin 0

z= √

H 2λRe (7-6)

πt 2 dt 2

are the Fresnel integrals. Closed-form expressions for these integrals are available [6]. 7-1.1 Beamwidth We can use Figures 7-3 and 7-4 to compute beamwidths. The 3-dB point corresponds to 0.707 and the 10-dB point to 0.316 on the graphs. Table 7-1 lists the 3- and 10dB points (values of U ) for different quadratic phase constants S in the H -plane. Similarly, Table 7-2 lists the E-plane points. The tables are more convenient than the graphs. Because we remove the obliquity factor to get a universal pattern, we must modify the beamwidths found by using the tables. We ﬁnd the beamwidth, but then we must add the obliquity factor to the beamwidth level. The beamwidth is found for

TABLE 7-1

Rectangular-Horn H -Plane Beamwidth Points, TE10 Mode (W/λ) sin θ

(W/λ) sin θ

S

3 dB

10 dB

S

3 dB

10 dB

0.00 0.04 0.08 0.12 0.16 0.20 0.24 0.28 0.32 0.36 0.40 0.44 0.48

0.5945 0.5952 0.5976 0.6010 0.6073 0.6150 0.6248 0.6372 0.6526 0.6716 0.6951 0.7243 0.7609

1.0194 1.0220 1.0301 1.0442 1.0652 1.0949 1.1358 1.1921 1.2700 1.3742 1.4959 1.6123 1.7143

0.52 0.56 0.60 0.64 0.68 0.72 0.76 0.80 0.84 0.88 0.92 0.96 1.00

0.8070 0.8656 0.9401 1.0317 1.1365 1.2445 1.3473 1.4425 1.5320 1.6191 1.7071 1.7991 1.8970

1.8062 1.8947 1.9861 2.0872 2.2047 2.3418 2.4876 2.6246 2.7476 2.8618 2.9744 3.0924 3.2208

342

HORN ANTENNAS

TABLE 7-2 Rectangular-Horn E -Plane Beamwidth Points, TE10 Mode (H /λ) sin θ

(H /λ) sin θ

S

3 dB

10 dB

S

3 dB

10 dB

0.00 0.04 0.08 0.12 0.16 0.20

0.4430 0.4435 0.4452 0.4482 0.4527 0.4590

0.7380 0.7405 0.7484 0.7631 0.7879 0.8326

0.24 0.28 0.32 0.36 0.40 0.44

0.4676 0.4793 0.4956 0.5193 0.5565 0.6281

1.4592 1.5416 1.6034 1.6605 1.7214 1.8004

a lower pattern level than speciﬁed. Since the beamwidth levels are close, we use a relation by Kelleher [7, p. 12–5] with good results: BW2 level2 (dB) (7-7) = BW1 level1 (dB) Example Compute 3- and 10-dB beamwidths for the horn in the preceding example. We have Sh = 0.55 and Se = 0.31. From Tables 7-1 and 7-2,

θh3

W sin θh3 λ W sin θh10 λ 28.9 W = λ 3.75 = 6.34 θh10

= 0.851 = 1.8726 = 7.707 = 14.06

H sin θe3 = 0.4915 λ H sin θe10 = 1.588 λ H 21.3 = = 5.68 λ 3.75 θe3 = 4.96 θe10 = 16.24

We consider the obliquity factor, (1 + cos θ )/2, at these angles, and apply Eq. (7-7) to reduce the beamwidths found. ◦

BWh3 = 12.68 at 3.03 dB ◦

BWe3 = 9.92 at 3.02 dB

◦

BWh3 x = 12.62 at 3.01 dB ◦

BWe3 = 9.89 at 3.01 dB

◦

BWh10 = 27.94 at 10 dB

◦

BW10 = 32.2 at 10 dB

BWh10 = 28.12 at 10.13 dB BWe10 = 32.48 at 10.18 dB

◦

◦

Including the obliquity factor has a very small effect on the results, but the effect increases with larger beamwidths (smaller apertures). Aperture theory fails for small horns because the beam is determined more by edge diffraction than the aperture ﬁelds. Empirical data were collected and reduced to simple formulas for small rectangular horns based on aperture size only [8, p. 46–22]: ◦

BW10e = 88

λ H

◦

◦

and BW10h = 31 + 79

λ W

RECTANGULAR HORN (PYRAMIDAL)

343

7-1.2 Optimum Rectangular Horn A rectangular horn has extra parameters, which we can use to design various optimum horns. Given a desired gain, we can design any number of horns with the same gain. Any optimum design depends on the requirements. Without any particular requirements, we will pick an antenna with equal E- and H -plane 3-dB beamwidths [9], but even this does not determine the design totally. If we pick a constant slant radius and vary the aperture width, the gain increases with increasing aperture width, but the quadratic phase error loss increases faster and produces a maximum point. The maximum occurs in the two planes at approximately constant phase deviations independent of the slant radius: Sh = 0.40

Se = 0.26

(7-8)

At these points we read the 3-dB points from Tables 7-1 and 7-2: H sin θ = 0.4735 λ

W sin θ = 0.6951 λ

On dividing these equations to eliminate the constant sin θ in the two planes, we derive the ratio of width to height to give a constant 3-dB beamwidth in the two planes for this optimum point: H = 0.68 (7-9) W The ratio depends on the beamwidth level. For 10-dB beamwidths, H = 1.00 W

(7-10)

These values of S determine the efﬁciency of the optimum horn. We read the PEL of the quadratic phase distribution from Table 4-42 by using a cosine distribution for the H -plane and a uniform distribution for the E-plane. The H -plane distribution has an ATL of 0.91 dB. PELh = 1.14 dB

PELe = 1.05 dB

or an efﬁciency of 49%: gain =

ATL = 0.91 dB

4πH W 0.49 λ2

We solve for H and W for a given gain, since we know the ratio between them [Eq. (7-9)]: W = λ

gain 4π(0.68)(0.49)

W = 0.489 gain λ

H = λ

gain(0.68) 4π(0.49)

H = 0.332 gain λ

(7-11)

344

HORN ANTENNAS

We combine Eqs. (7-8) and (7-11) to calculate slant radiuses: Rh = 0.0746 · gain λ

Re = 0.0531 · gain λ

(7-12a, b)

If, given gain, we use Eqs. (7-11) and (7-12) to design a horn, the dimensions will not be practical with an arbitrary input waveguide. The axial lengths from the waveguide to the aperture must be equal in the E- and H -planes, so the horn will meet the waveguide in a single plane. Given waveguide dimensions a and b, the axial lengths are 2 W −a W H −a H2 Rh2 − Re2 − Le = (7-13a, b) Lh = W 4 H 4 We have a choice between retaining the E- or H -plane slant radius given by Eq. (7-12) and forcing the other radius to give the same axial length. The primary factor affecting gain is the aperture dimensions, which we retain from Eq. (7-11). The slant radius is secondary. We retain the H -plane radius calculated from Eq. (7-12) and modify the E-plane radius. Modifying the H -plane radius would give us a second optimum design: H (H − b)2 (7-14) L2 + Re = H −b 4 To obtain the proper gain, we must iterate, since we cannot use both Eq. (7-12). Design the horn by using Eqs. (7-11), (7-12a), (7-13a), and (7-14). Calculate the gain from the dimensions and obtain a new design gain from Gd,new =

Grequired Gd,old Gactual

(7-15)

where Grequired is the required gain, Gactual is the actual gain, and Gd,old is the old design gain. Example Design a horn fed from WR-90 waveguide to have 22 dB of gain at 10 GHz. The waveguide dimensions are 2.286 cm × 1.016 cm (0.9 in. × 0.4 in.) and Greq = Gd = 1022/10 = 158.5. On substituting in Eq. (7-11), we calculate aperture dimensions: W = 18.47 cm and H = 12.54 cm. From Eqs. (7-12a) and (7-13a), Rh = 35.47 cm and L = 30.01 cm. To get the same axial length in the E-plane [Eq. (7-14)], Re = 33.25 cm. We now calculate gain and compare it with the gain required. The amplitude taper loss and phase error loss in the H -plane remain constant, since Sh is ﬁxed. ATL = 0.91 dB

PELh = 1.14 dB at Sh = 0.40

Calculate Se : H2 = 0.197 PELe = 0.60 dB (Table 4-42) 8λRe 4πH W Gactual (dB) = 10 log − ATL − PELh − PELe = 22.45 dB λ2 = 175.8 Se =

RECTANGULAR HORN (PYRAMIDAL)

345

We must pick a new design gain [Eq. (7-15)]: Gd,new =

158.5(158.5) = 142.9 175.8

A second iteration with this gain gives the following dimensions: W = 17.54 cm

H = 11.91 cm

L = 26.75 cm

Rh = 31.98 cm

Re = 29.84 cm Se = 0.198 PELe = 0.60 4πH W − 0.91 − 1.14 − 0.60 = 22.00 dB gain = 10 log λ2 We obtain the gain desired, but the 3-dB beamwidths are only approximately equal: H -plane: 13.66◦ , E-plane: 13.28◦ . Scales 7-1 to 7-3 provide the dimensions of the optimum rectangular horn for a given gain. During their generation, a waveguide with a 2 : 1 aspect was used, but they are close to the proper values for nearby aspects. They design horns to within 0.1 dB. These scales produce short rapidly ﬂaring horns for low-gain antennas. In these cases it is better to deviate from the optimum design that gives the lightest horn for a given gain and design a horn with a small value of S. Scales 7-4 to 7-6 give designs for S = 0.1 to be used for low-gain designs. Aperture Width, l

Optimum Rectangular Horn Gain, dB

SCALE 7-1 Aperture width of an optimum pyramidal horn for a 2 : 1-aspect waveguide.

Aperture Height, l

Optimum Rectangular Horn Gain, dB

SCALE 7-2 Aperture height of an optimum pyramidal horn for a 2 : 1-aspect waveguide. Aperture Length, l

Optimum Rectangular Horn Gain, dB

SCALE 7-3 Axial length of an optimum pyramidal horn for a 2 : 1-aspect waveguide.

346

HORN ANTENNAS Aperture Width, l

S = 0.1 Rectangular Horn Gain, dB

SCALE 7-4 Aperture width of a pyramidal horn with S = 0.1 connected to a 2 : 1 waveguide. Aperture Height, l

S = 0.1 Rectangular Horn Gain, dB

SCALE 7-5 Aperture height of a pyramidal horn with S = 0.1 connected to a 2 : 1 waveguide. Aperture Length, l

S = 0.1 Rectangular Horn Gain, dB

SCALE 7-6 Axial length of a pyramidal horn with S = 0.1 connected to a 2 : 1 waveguide.

7-1.3 Designing to Given Beamwidths The beamwidths in the two planes of a rectangular horn can be designed independently. The axial lengths in the two planes must be equal if the design is to be realizable, but the aperture width and height can be adjusted to give the desired beamwidths. We pick S in one plane and then vary S in the other plane to produce the required beamwidth and the same axial length as in the ﬁrst plane. Since the ﬁrst S is arbitrary, the design is not unique, but in only a limited range of S will designs be realizable. Example Design a rectangular horn for the following 10-dB beamwidths: 30◦ H plane and 70◦ E-plane at 7 GHz using a 3.5-cm × 1.75-cm waveguide. Since the H -plane has the narrower beamwidth and therefore the wider aperture, we use it to determine length. Pick Sh = 0.20 (an arbitrary choice). The obliquity factor at 15◦ is 0.15 dB. When using Table 7-1 we must design for wider than a 30◦ beamwidth to compensate for the obliquity factor: 10.15 ◦ ◦ BWd = (30 ) = 30.22 10 The horn width to provide that beamwidth is 1.0949 W = = 4.200 λ sin(BWd /2)

RECTANGULAR HORN (PYRAMIDAL)

W = 18.00 cm at 7 GHz

347

Rh = 47.25 [Eq. (7.2)]

Use Eq. (7-13a) to determine the axial length: Lh = L = 37.36 cm. Because the Eplane beamwidth is wider than the H -plane beamwidth, the E-plane aperture will be smaller. We try a smaller value of Se than Sh for our initial guess: Se = 0.04. The obliquity factor at 35◦ adds 0.82 dB to the pattern loss and requires a larger design beamwidth. 10.82 ◦ ◦ BWd = 70 = 72.82 10 H 0.7405 = = 1.248 (Table 7-2) H = 5.246 λ sin(BWd /2) We use Eqs. (7-2) and (7-13b) to calculate slant radius and axial length: Re = 20.84 cm and Le = 13.90 cm. The axial lengths in the two planes do not match, so we pick a smaller Se because the E-plane is shorter than the H -plane beamwidth. At Se = 0.02, H = 5.337 cm, Re = 41.54 cm, and Le = 27.86 cm. Le has doubled when Se changes from 0.04 to 0.02, but H changes by only 0.01 cm. We change our method and pick H = 5.33 cm and force Re to give the same axial length as the H -plane: Re = 55.69 cm [from Eq. (7-14)] or Se = 0.0149. 7-1.4 Phase Center We deﬁne the phase center as the point from which it appears that an antenna radiates spherical waves. Measurements show that the phase center is seldom a unique point in a plane, but depends on the pattern angle. The E- and H -plane phase centers will also be unequal in general. Usually, they are extremes, and the axial position varies elliptically between the planes. Even with the phase-center location fuzzy, it is a useful point. We place the phase center of a feed at the focus of a parabolic reﬂector to minimize the reﬂector aperture phase error loss. An aperture with a quadratic phase distribution appears to be radiating from a point behind the aperture. Without quadratic phase error (S = 0), the phase center is located at the aperture plane. Increasing S moves the phase center toward the apex of the horn. Muehldorf [10] has calculated the phase-center location as a function of S, and Table 7-3 summarizes his results. The phase center located inside the aperture is given as a ratio of the slant length. TABLE 7-3 Phase-Center Axial Location of a Rectangular Horn (TE10 Mode) Behind the Aperture as a Ratio of the Slant Radius S 0.00 0.04 0.08 0.12 0.16 0.20 0.24

H -Plane Lph /Rh

E-plane Lph /Re

0.0 0.0054 0.022 0.048 0.086 0.134 0.191

0.0 0.011 0.045 0.102 0.182 0.286 0.416

S 0.28 0.32 0.36 0.40 0.44 0.48 0.52

H -plane Lph /Rh

E-plane Lph /Re

0.258 0.334 0.418 0.508 0.605 0.705 0.808

0.572 0.755

348

HORN ANTENNAS

Example above.

Calculate the phase-center location for the beamwidth design example Sh = 0.20

Rh = 47.25 cm

Se = 0.015

Re = 55.69 cm

From Table 7-3, we interpolate H -plane phase center = 0.134(47.25 cm) = 6.33 cm E-plane phase center = 0.004(55.69 cm) = 0.22 cm The difference in the phase-center location in the two planes is 1.43λ. As in the example, antennas with widely differing beamwidths will have widely separated phase centers.

7-2 CIRCULAR-APERTURE HORN With a circular-aperture horn, we lose independent control of the beamwidths in the principal planes. The circular waveguide can support any orientation of the electric ﬁeld and thereby allows any polarization in the horn. We use the same aperture ﬁeld method as with the rectangular horn and the waveguide mode determines the aperture amplitude. The cone of the horn projects to a point in the feed waveguide where we assume a point source radiating to the aperture. The aperture phase is approximately quadratic. The waveguide ﬁelds are given by x11 ρ E0 Eρ = J1 cos φc ρ a E0 x11 x11 ρ sin φc J1 E φc = − (7-16) a a where J1 is the Bessel function, ρ the radial component in the waveguide, a the radius, and φc the cylindrical coordinate. x11 (1.841) is the ﬁrst zero of J1 (x). Equation (7-16) has its maximum electric ﬁeld directed along the φc = 0 plane. We add the quadratic phase factor to Eq. (7-16) and calculate the Fourier transform on the circular aperture to determine the far ﬁeld. The direction of the electric ﬁeld changes from point to point in the aperture. For a given direction (θ , φc ) we must project the ﬁelds in the aperture onto the θˆ and φˆ directions before integrating over the aperture:

2π a J1 (x11 ρ/a) θˆ žρˆ θˆ žφˆ c x11 x11 ρ sin φc Eθ = E0 cos φc − J ρ cos θ a 1 a cos θ 0 0

ρ 2 (7-17) × ρ exp j kρ sin θ cos(φ − φc ) − 2πS dρ dφc a

CIRCULAR-APERTURE HORN

x11 ρ sin φc φˆ · φˆ c Eφ = E0 a 0 0 ρ 2 × ρ exp j kρ sin θ cos(φ − φc ) − 2πS (7-18) dρ dφc a

2π

a

349

x J1 (x11 ρ/a) cos φc φˆ · ρˆ − 11 J1 ρ a

θˆ · ρˆ = cos θ (cos φ cos φc + sin φ sin φc ) θˆ · φˆ c = cos θ (sin φ cos φc − cos φ sin φc ) φˆ · ρˆ = cos φ sin φc − sin φ cos φc φˆ · φˆ c = cos φ cos φc + sin φ sin φc By a suitable change of variables in the integrals, universal radiation patterns can be generated for the E-and H -planes (Figures 7-5 and 7-6). The equality of S in the two planes ties the curves together. The axis is the k-space variable. We can calculate a few pattern points for a given horn with those curves if we remember to add the obliquity factor to the values taken from the curves.

0.9

S = 0.7

0.8 S=

a2 2lR

0.7 Relative Field Intensity

0.6 0.6 0.5 0.5 0.4 0.4

0.3

0.3 0.2

0.2

0.1 0.1 0 0 1

2

3

4

5

6 7 2pa sin q l

8

9

10

11

12

FIGURE 7-5 E-plane universal pattern of a circular, TE11 mode. (From T. Milligan, Universal patterns ease circular horn design, Microwaves, vol. 20, no. 3, March 1981, p. 84.)

350

HORN ANTENNAS

0.9

0.8 S=

a2 2lR

Relative Field Intensity

0.7 0.6

S = 0.7

0.5 0.6 0.4 0.5 0.3 0.4 0.2 0.2

0.1

0

0 1

2

3

4

5

6 7 2pa sin q l

8

9

10

11

12

FIGURE 7-6 H -plane universal pattern of a circular, TE11 mode. (From T. Milligan, Universal patterns ease circular horn design, Microwaves, vol. 20, no. 3, March 1981, p. 84.)

Example A horn has an aperture radius of 12 cm and a slant radius of 50 cm. Compute the pattern level at θ = 20◦ at 5 GHz. From Figures 7-5 and 7-6 we interpolate the pattern voltage level: H -plane level = 0.18 (−14.7 dB)

E-plane level = 0.22

(−13.1 dB)

The obliquity factor is 20 log[(1 + cos 20◦ )/2] = −0.3 dB, and the plane level at 20◦ becomes H -plane level = −15 dB E-plane level = −13.4 dB 7-2.1 Beamwidth Table 7-4 lists the 3- and 10-dB points from Figures 7-5 and 7-6. We can use them to compute beamwidths from dimensions. Example Calculate 10-dB beamwidths of the horn in the example above. S = 0.24, a = 12 cm, and λ = 6 cm.

351

CIRCULAR-APERTURE HORN

TABLE 7-4

Circular-Horn Beamwidths, TE11 Mode (2πa/λ) sin θ 3 dB

10 dB

S

E-Plane

H -Plane

E-Plane

H -Plane

ATL + PEL (dB)

0.00 0.04 0.08 0.12 0.16 0.20 0.24 0.28 0.32 0.36 0.40 0.44 0.48 0.52 0.56 0.60 0.64 0.68 0.72

1.6163 1.6175 1.6212 1.6273 1.6364 1.6486 1.6647 1.6855 1.7123 1.7471 1.7930 1.8552 1.9441 2.0823 2.3435 3.4329 4.3656 4.8119 5.1826

2.0376 2.0380 2.0391 2.0410 2.0438 2.0477 2.0527 2.0592 2.0676 2.0783 2.0920 2.1100 2.1335 2.1652 2.2089 2.2712 2.3652 2.5195 2.8181

2.7314 2.7368 2.7536 2.7835 2.8296 2.8982 3.0024 3.1757 3.5720 4.6423 5.0492 5.3139 5.5375 5.7558 6.0012 6.3500 7.6968 8.4389 8.8519

3.5189 3.5211 3.5278 3.5393 3.5563 3.5799 3.6115 3.6536 3.7099 3.7863 3.8933 4.0504 4.2967 4.6962 5.2173 5.6872 6.0863 6.4622 6.8672

0.77 0.80 0.86 0.96 1.11 1.30 1.54 1.82 2.15 2.53 2.96 3.45 3.99 4.59 5.28 5.98 6.79 7.66 8.62

From Table 7-4 we read the k-space values at 10 dB: 2πa sin θh = 3.6115 λ 2πa sin θe = 3.0024 E-plane k-space = λ 3.6115 3.0024 ◦ ◦ = 33.40 = 27.65 BWh = 2 sin−1 BWe = 2 sin−1 4π 4π H -plane k-space =

We must add the obliquity factor to the 10-dB universal pattern level:

BWh10 =

1 + cos 16.7◦ : 2 1 + cos 13.8◦ : 2 10 ◦ ◦ 33.40 = 33.10 10.18

0.18 dB

H -plane

0.13 dB

E-plane

BWe10 =

10 ◦ ◦ 27.65 = 27.48 10.13

We can also use Table 7-4 to design a horn to a given beamwidth, but we can design to only one plane. Any number of horns can be designed to a given beamwidth, since S is an independent parameter. Table 7-4 lists the combined amplitude taper loss and phase error loss as a function of S for the circular horn. With this table we can easily estimate the gain of a given horn or design a horn to a given gain.

352

HORN ANTENNAS

Example Compute the gain of a horn with a 12-cm aperture radius and 50-cm slant radius at 5 GHz. From the examples above we read S = 0.24 and λ = 6 cm: πD − GF where GF = ATL + PEL (dB) λ 24π = 20 log − 1.54 = 20.4 dB 6

gain = 20 log

(7-19)

Example Design a circular horn at 8 GHz with a gain of 22 dB. The quadratic phase distribution constant S is arbitrary. Pick S = 0.20. Rearrange Eq. (7-19) to ﬁnd the diameter: λ · 10(gain+GF)/20 π 3.75 · 10(22+1.30)/20 = 17.45 cm = π D2 = 50.77 cm R= 8λS

D=

(7-20)

We can determine an optimum circular horn in the sense of minimizing the slant radius at a given gain. When we plot gain as a function of aperture radius for a ﬁxed slant radius, we discover a broad region in which the gain peaks. By plotting a series of these lines with a voltage gain ordinate, we see that a single line corresponding to S = 0.39 can be drawn through the peaks. This is our optimum with GF = 2.85 dB (ATL + PEL). Example Design an optimum horn at 8 GHz with gain of 22 dB. From Eq. (7-20), D=

3.75 · 10(22+2.85)/20 = 20.86 cm π

R=

20.862 D2 = = 37.2 cm 8λS 8(3.75)(0.39)

The optimum is quite broad. A horn designed with S = 0.38 has a 0.07-cm-longer slant radius and a 0.25-cm-smaller aperture diameter. 7-2.2 Phase Center The phase-center location behind the aperture plane is a function of S. Table 7-5 lists the phase-center location as a ratio of the slant radius. As S increases, the phase center migrates toward the horn apex and the difference between the phase-center locations in the E- and H -planes increases. Example Use Table 7-5 to compute phase-center locations in the E- and H -planes for the circular horns of the preceding two examples. R = 50.77 cm S = 0.20 H -plane phase center = 0.117(50.77) = 5.94 cm E-plane phase center = 0.305(50.77) = 15.48 cm

CIRCULAR (CONICAL) CORRUGATED HORN

353

TABLE 7-5 Phase-Center Axial Location of a Circular-Waveguide Horn TE11 Mode Behind the Aperture as a Ratio of the Slant Radius S

H -Plane Lph /Rh

E-Plane Lph /Re

0.00 0.04 0.08 0.12 0.16 0.20 0.24

0.0 0.0046 0.018 0.042 0.075 0.117 0.171

0.0 0.012 0.048 0.109 0.194 0.305 0.416

S

H -Plane Lph /Rh

E-Plane Lph /Re

0.28 0.32 0.36 0.40 0.44 0.48

0.235 0.310 0.397 0.496 0.604 0.715

0.603 0.782 0.801 0.809 0.836 0.872

The phase centers differ by 2.5 wavelengths at 8 GHz. The optimum horn has the dimensions R = 37.2 cm S = 0.39 H -plane phase center = 0.471(37.2) = 17.5 cm E-plane phase center = 0.807(37.2) = 30.0 cm The phase centers of the optimum horn differ by 3.3 wavelengths. The difference increases with increasing S.

7-3 CIRCULAR (CONICAL) CORRUGATED HORN Normal smooth-wall horns present problems that can be eliminated by corrugating the walls. Many applications require dual linear or circular polarizations. The horn aperture must be square or circular and the beamwidths in the two planes are unequal. When the smooth-wall horn feeds a reﬂector, we have astigmatism (unequal phase centers in orthogonal planes). A horn has higher sidelobes in the E-plane than in the H -plane. Finally, the diffraction off E-plane walls causes backlobes. The aperture theory fails to predict them, but measurement or a GTD analysis shows them. Corrugating the walls can prevent all these problems. Figure 7-7 shows the cross sections of two types of corrugated horns. The smallﬂare horn (a) is nominally the corrugated horn, and the wide-ﬂare horn (b) is the scalar horn of Simmons and Kay [11]. Many papers on corrugated horns appear in Section VI of Love’s [1] collection of papers. Thomas [12] provides a good design summary in a topic with many papers. The corrugations that extend circumferentially should be cut normal to the slant radius as in (b), but they may be cut normal to the axis (a) for small ﬂare angles. Horns can be built either way, but when cut normal to the axis, the depth is different on the two sides. The corrugated wall presents the same boundary conditions to the electric and magnetic ﬁelds when it is capacitive (slots λ/4 to λ/2 deep). When excited in the transition between the smooth-wall waveguide and the corrugated-wall cone, the TE11 and TM11 waveguide modes, have equal phase velocities. The combination of these

354

HORN ANTENNAS

t w

p E11

g a1 q0 d1 R

a

∆ Rq0 d

w t

(a)

a0 q0 0

g

p

Rq0

∆ d1

R

d (b)

FIGURE 7-7 (a) Corrugated horn; (b) scalar horn. (From [12], 1978 IEEE.)

modes forms the hybrid mode HE11 when the mode phases are equal. When the modes are out of phase by 180◦ , the hybrid mode is denoted EH11 . The ratio of the modes is called γ , and γ = 1 for the balanced HE11 mode. γ = 0 corresponds to having only the TM11 mode and γ = ∞ to having only the TE11 mode. γ = 1 occurs when the corrugation depth is λ/4, but the horn parameters vary slowly with changing γ [13]. We consider only γ = 1. Changing γ has its biggest effect on the cross-polarization [12,14]. When γ = 1, the amplitude of the aperture ﬁelds is given by [15] x ρ 01 cos φc Eρ = J0 a (7-21) x ρ 01 sin φc Eφ = −J0 a where x01 = 2.405 is the ﬁrst zero of J0 (x), the Bessel function. The ﬁelds vanish at the walls and prevent edge diffractions that produce sidelobes and backscatter. The lower-order slope diffractions still produce small sidelobes and backlobes, but we get H -plane-type lobes in all planes. In amplitude the aperture ﬁelds are symmetrical about the axis and all patterns through the cone axis are identical. A Huygens source analysis of the aperture ﬁelds with a quadratic phase distribution produces Figure 7-8, valid when the 10-dB beamwidth is less than 74◦ [12]. For greater beamwidths the ﬂange changes the beamwidths of the small-aperture horn in the two planes and we should use the scalar horn. Table 7-6 lists the 3-, 10-, and 20-dB points

CIRCULAR (CONICAL) CORRUGATED HORN

355

S=1 0.9

0.9

0.8 0.8

S=

a2 2lR

0.7 Relative Field Intensity

0.7 0.6

0.5 0.6 0.4 0.5 0.3 0.4

0.2

0.3 0.1 0.2 0 2

4

0 6 2pa sin q l

8

10

12

FIGURE 7-8 Universal pattern of a circular corrugated horn: HE11 mode. TABLE 7-6 Mode

Resonant Circular Corrugated Horn Beamwidth Points (2πa/λ) sin θ , HE11

S

3 dB

10 dB

20 dB

ATL + PEL (dB)

S

3 dB

10 dB

20 dB

ATL + PEL (dB)

0.00 0.04 0.08 0.12 0.16 0.20 0.24 0.28 0.32 0.36 0.40 0.44 0.48

2.0779 2.0791 2.0827 2.0887 2.0974 2.1088 2.1234 2.1415 2.1637 2.1906 2.2231 2.2624 2.3103

3.5978 3.6020 3.6150 3.6371 3.6692 3.7129 3.7699 3.8433 3.9372 4.0572 4.2112 4.4090 4.6578

4.6711 4.6878 4.7405 4.8387 5.0061 5.3052 5.8451 6.3379 6.6613 6.9179 7.1534 7.3939 7.6633

1.60 1.62 1.66 1.73 1.83 1.96 2.12 2.30 2.52 2.76 3.04 3.34 3.68

0.52 0.56 0.60 0.64 0.68 0.72 0.76 0.80 0.84 0.88 0.92 0.96 1.00

2.3688 2.4411 2.5317 2.6469 2.7966 2.9946 3.2597 3.6061 4.0189 4.4475 4.8536 5.2331 5.5984

4.9532 5.2720 5.5878 5.8913 6.1877 6.4896 6.8134 7.1788 7.6042 8.0852 8.5773 9.0395 9.4701

7.9936 8.4261 8.9472 9.4352 9.8514 10.2337 10.6250 11.0735 11.6356 12.2658 12.8236 13.3059 13.7706

4.04 4.44 4.86 5.31 5.79 6.30 6.83 7.39 7.96 8.54 9.13 9.72 10.29

356

HORN ANTENNAS

from Figure 7-8. We use the table to ﬁnd beamwidths of given horns or design to given beamwidths. Table 7-6 also lists the sum of ATL and PEL(GF) for various S. We estimate gain or design to a given gain with this listing. Example Calculate 10-dB beamwidth and gain of a corrugated conical horn with an aperture radius of 12 cm and a slant radius of 50 cm at 5 GHz: S=

a2 122 = = 0.24 2λR 2(6)(50)

From Table 7-7 we read the k-space value at 10 dB for S = 0.24: 2πa sin θ = 3.7699 λ

or

◦

θ = 17.46

We include the obliquity factor, since the pattern loss will be greater than 10 dB at θ = 17.46◦ : 10 ◦ ◦ ◦ (1 + cos 17.46 )/2 : −0.20 dB BW10 = 34.92 = 34.57 10.20 A smooth-wall horn with the same dimensions has a similar H -plane beamwidth (33.4◦ ). We calculate gain from the distribution losses and aperture area: GF = ATL + PEL = 2.12 dB

gain = 20 log

πD − GF = 19.86 dB λ

The smooth-wall horn has about 0.5 dB more gain (20.4 dB). Example Design a circular corrugated-wall horn at 8 GHz with a gain of 22 dB. We use Eq. (7-20) with the GF from Table 7-6. Choose S = 0.20 (arbitrary): 3.75 cm · 10(22+1.96)/20 = 18.83 cm π D2 = 59.10 cm R= 8λS

D=

TABLE 7-7 Phase-Center Axial Location of a Circular Corrugated Horn (HE11 Mode) Behind the Aperture as a Ratio of the Slant Length S

Lp /R

S

Lp /R

0.00 0.04 0.08 0.12 0.16 0.20 0.24 0.28 0.32

0.0 0.005 0.020 0.045 0.080 0.124 0.178 0.240 0.310

0.36 0.40 0.44 0.48 0.52 0.56 0.60 0.64 0.68

0.386 0.464 0.542 0.614 0.673 0.718 0.753 0.783 0.811

CIRCULAR (CONICAL) CORRUGATED HORN

357

The phase center, like that of other horns, starts at the aperture for S = 0 (R = ∞) and moves toward the apex as S increases. Table 7-7 lists the phase-center location as a ratio of the slant radius. Because the aperture distribution is the same along all radial lines of the aperture, the phase center is the same in all planes. We measure some variation, since the balance between modes will not be perfect, and some higher-order modes will be generated. The phase center moves least over a frequency band for long horns (small S). A wide-ﬂare-angle horn has its phase center at the apex and is better described as a scalar horn. 7-3.1 Scalar Horn A scalar horn has a wide half-ﬂare angle, θ0 . Its beamwidth depends on the half-ﬂare angle. Since the phase distribution in the aperture is large, there is an optimum diameter for a given ﬂare angle. Table 7-8 lists the optimum diameter versus the ﬂare angle. The beamwidth is approximately a linear function of the half-ﬂare angle, θ0 , for the optimum horn: BW3 dB = 0.74θ0 BW10 dB = 1.51θ0

(7-22)

BW20 dB = 2.32θ0 A scalar horn has a wider bandwidth as a reﬂector feed than that of a small-ﬂare-angle corrugated horn, because the phase center is ﬁxed at the horn apex. 7-3.2 Corrugation Design The corrugations present a capacitive reactance to the passing wave. When a corrugated surface is inductive, it will support surface waves. The depth of corrugations must be between λ/4 and λ/2. Less than λ/4 or greater than λ/2, it is inductive. Between 3λ/2 and λ it will be capacitive again, but this second passband is seldom used. A quarterwavelength corrugation depth balances the two modes and gives the best results. The corrugations need be only λ/4 at the aperture. Before the aperture we ﬁnd it better to deepen the slots. Quarter-wavelength-deep corrugations mismatch the horn in the transition region, where the TM11 mode is generated from the TE11 mode and depths approaching λ/2 have the least effect on match. TABLE 7-8 Optimum Diameter of a Scalar Horn Half-Flare Angle, θ0 (deg)

Aperture Diameter (λ)

Half-Flare Angle, θ0 (deg)

Aperture Diameter (λ)

15 20 25 30 35 40

10.5 8.0 6.4 5.2 4.5 3.9

45 50 55 60 65 70

3.5 3.0 2.8 2.6 2.4 2.3

Source: [16, p. 429].

358

HORN ANTENNAS

To design for a particular band, limited to about 1.5 : 1 for a good match, we design with tapered corrugation depths. We make the corrugations λ/4 deep at the aperture at the low-frequency end. The high-frequency band edge determines the corrugation depth at just short of λ/2 in the throat. The horn needs at least four corrugations per wavelength along the slant radius. The high-frequency end determines the corrugation spacing. The ﬁrst few corrugations can be used to match the horn to the waveguide, and we can improve the match by shaping the corrugation widths [14]. The slots should be as wide as practical spacers will allow. Mechanical considerations, such as shock and vibration, will determine the limits on the thinness of spacers, but corrugations greatly increase the strength of the bell. The circular geometry of the horn changes the corrugation depth necessary for the balanced HE11 mode from λ/4. An empirical formula for the depth is given by [17] 1 λ d = exp ka > 2 (7-23) 4 2.5 ka We increase the depth slightly at the horn aperture. 7-3.3 Choke Horns We can place corrugations in the ﬂanges of small-aperture horns and design widebeamwidth antennas with good pattern symmetry and low cross-polarization. The choke horn (Figure 7-9) is the limit of a scalar horn opened to θ0 = 90◦ . The corrugations consist of concentric rings about the aperture and are generally a quarter-wavelength deep. The best location for the corrugated rings may not be in the same plane as the aperture but instead somewhat behind as reported for a feed for f/D = 0.3 reﬂector [18;19, pp. 200–209]. The design for the reﬂector feed [18] used four corrugations. James [20] and Kumer [21] show that using only one corrugation is quite effective. More corrugations improve the design but only add marginally. Small apertures need the corrugations to reduce the cross-polarization that peaks in the diagonal planes between the E- and H -planes. We usually assume a Huygens source in the aperture plane of the horn. This approximation collapses as we shrink the aperture to achieve wide beamwidths. In the limit

FIGURE 7-9

Choke horn.

CORRUGATED GROUND PLANE

359

we have only a slot analyzed from magnetic currents replacing the electric ﬁeld in the aperture. The magnetic ﬁeld is ignored in the slot while a Huygens source assumption is that the ratio of the electric to magnetic ﬁeld is the same as the impedance of free space. Waveguides have high wave impedances, which implies small magnetic ﬁelds. To calculate the far ﬁeld, Eq. (2-23) must be used with the actual ratio of ﬁelds in the aperture instead of Eq. (2-24), with its Huygens source approximation. Restricting the aperture dimensions to achieve wide beamwidths will limit the bandwidth as well as the cross-polarization isolation, because reducing volume raises Q. 7-3.4 Rectangular Corrugated Horns We can design rectangular horns with corrugated walls, but we only need to cut corrugations in the E-plane walls to produce a cosine distribution in the E-plane. Only for dual polarization do we need corrugations in the H -plane walls. We analyze the horn as having an H -plane distribution (cosine) in both planes and use the results of Section 7-1. The larger aperture dimension in the diagonal plane decreases the beamwidth slightly, but the rectangular horn still provides an acceptable design. Both planes have a linear amplitude taper loss of 0.91 dB. We use the cosine column of Table 4-42 for the quadratic phase error loss. We design beamwidths using Table 7-1. Equalizing the distributions in both planes of square horns results in equal phase centers, given by Table 7-3 (H -plane). Example Compute the gain of a square corrugated horn with an aperture width of 24 cm and a slant radius of 50 cm at 5 GHz. From Eq. (7-1), W2 242 S= = = 0.24 8λR 8(6)(50) We use the cosine column of Table 4-42 for the phase error loss: PELx = PELy = 0.42 (cosine). The amplitude taper loss is the same in both planes: 0.91 dB. 4πW 2 − ATLx − ATLy − PELx − PELy λ2 = 23.03 − 0.91 − 0.91 − 0.42 − 0.42 = 20.4 dB

Gain = 10 log

A circular corrugated horn with a diameter equal to the width and having the same slant radius has a gain of 19.9 dB or 0.5 dB less. The larger aperture area increases the gain over the circular horn. 7-4 CORRUGATED GROUND PLANE The corrugated surface (Figure 7-10) supports surface waves (TM) when the slot depth is less than λ/4 (inductive). As with the corrugated horn, we assume many slots per wavelength along the direction of propagation. The ﬁelds attenuate exponentially above the ends of the corrugations in a surface wave. We derive the ﬁelds from a potential function −2πbx ψ = A1 exp (7-24) exp(−j kz z) λ

360

HORN ANTENNAS x g

t d z

FIGURE 7-10 Corrugated ground plane.

above the corrugations, where A1 is an amplitude constant, x the distance out of the corrugations, and α(= 2πb/λ) the attenuation constant of the ﬁelds above the corrugations. We expand the ﬁelds and take the ratio of the z-directed electric ﬁeld to the y-directed magnetic ﬁeld to ﬁnd the wave impedance into the corrugated surface: 1 2πb 2 A1 2πb (k 2 − kz2 )ψ = − exp − exp(−j kz z) j ωε0 λ j ωε0 λ 2πb ∂ψ 2πb = A1 exp − Hy = − exp(−j kz z) ∂x λ λ √ j (2πb/λ) µ0 Ez j (2πb/λ) j (kb) = = √ = Z−x = η = j bη (7-25) √ Hy ωε0 ω ε0 µ 0 ε0 k Ez =

where η is the impedance of free space and b is related to α [Eq. (7-24)]. The structure must present this impedance to the wave. The corrugated surface is a parallel-plate transmission line to Ez , and it presents a per unit length impedance of Zc = j η tan kd

(7-26)

where d is the corrugation depth. We equate Eqs. (7-25) and (7-26) to determine the constant b: b = tan kd (7-27) We use Eq. (7-27) in Eq. (10-16) for the relative propagation constant: P = 1 + b2 = 1 + tan2 kd

(7-28)

We include the effect of the corrugation thickness by averaging between the parallelplate impedance and the zero impedance along the corrugation edges. Equation (7-28) becomes 2 g P = 1+ tan2 kd (7-29) g+t where g is the gap distance and t is the corrugation thickness. P increases without bound as the depth d approaches λ/4. The ﬁelds become tightly bound to the surface and attenuate rapidly to zero above the corrugations—the normal electric ﬁeld vanishes as in a corrugated horn E-plane wall. We design the depth of the corrugations by using d=

λ g+t tan−1 √ 2π g P2 − 1

(7-30)

CORRUGATED GROUND PLANE

361

When the corrugation depth approaches λ/4, the surface impedance [Eq. (7-26)] approaches inﬁnity and the tangential magnetic ﬁeld vanishes on the surface to create an artiﬁcial PMC (Section 2-3) for waves polarized along the z-axis. The reﬂection coefﬁcient is +1 instead of −1 for the PEC surface. Waves polarized along the y-axis encounter closely spaced corrugations that approximately produce a PEC surface with the usual metal wall reﬂection coefﬁcient of −1. Whereas a PEC reﬂects an incident circularly polarized wave with opposite sense of circular polarization, the artiﬁcial PMC (soft) surface [22, pp. 276–280] reﬂects the wave with the same sense of polarization. We can use these surfaces to shape the pattern of a wide-beamwidth circularly polarized antenna to narrow the beamwidth without generating the opposite sense polarization, which would be generated by metal walls. A ground plane covered with circular coaxial corrugations λ/4 deep reduces the edge diffraction that produces a large backlobe for a monopole antenna mounted in the center (Figure 5-23). The artiﬁcial soft wall causes the reduction of circumferential magnetic ﬁelds and the associated GTD diffraction (Section 2-7.11) [23]. This increases the forward gain by reducing the backlobe. It is not necessary to corrugate the entire top surface. Figure 7-11 illustrates a surface with only two coaxial corrugations around the outer rim. These reduce the backlobe for a dipole mounted over the ground plane without generating signiﬁcant cross-polarization from a pair of orthogonal dipoles fed for a circular polarization. Corrugating the entire surface would cause radiation of crosspolarization because the region below the dipole pair radiates oppositely sensed circular polarization. The corrugated surface reﬂects the same sense of circular polarization as incident. The PEC surface reverses the sense of circular polarization of the reﬂected wave and both waves add. The corrugations only reduce the backlobe. The choke horn uses the same type of corrugations to reduce the backlobe radiated from the small-diameter horn aperture. The corrugations can be placed radially below the ground plane by using shortcircuited radial transmission lines (Figure 7-12) and also reduce the backlobe. We

FIGURE 7-11

Ground plane with two coaxial corrugations to reduce edge diffraction.

FIGURE 7-12 Ground plane with short-circuited radial transmission-line corrugations.

362

HORN ANTENNAS

TABLE 7-9 Radial Transmission Outer Choke Depth at Resonance Outer Radius (λ)

Depth (λ)

Outer Radius (λ)

Depth (λ)

0.25 0.30 0.35 0.40 0.50 0.60

0.188 0.199 0.208 0.213 0.222 0.227

0.70 0.80 1.0 2.0 4.0

0.230 0.233 0.236 0.243 0.247

compute the reactance at the outer radius from the following equation, which uses Bessel and Neumann functions: X=j

ηb N0 (kr)J0 (kro ) − J0 (kr)N0 (kro ) 2πr J1 (kr)N0 (kro ) − N1 (kr)J0 (kro )

The short-circuited wall is located at radius ro and the spacing between the plates is b. Reactance X grows rapidly as r approaches resonance. For a large outer radius the difference r − ro approaches λ/4 but is less for a small radius. Table 7-9 gives the difference in radius versus the outer radius for resonance. Corrugations on the upper surface are more effective than radial corrugations, but the radial line chokes ﬁt easily behind a small ground plane. In both cases the corrugations enhance radiation behind the ground plane at frequencies below resonance λ/4 depths, because a surface wave is generated along the corrugations. Corrugated surfaces are useful structures because they can be used to enhance or reduce radiation, depending on their depth.

7-5 GAUSSIAN BEAM Corrugated horns and simple reﬂector feeds can be approximated with Gaussian beams. An inﬁnite circularly symmetrical Gaussian aperture distribution located in the x –y plane radiates a Gaussian beam along the z-axis. The radial exponent of the Gaussian distribution determines the spread of the wave as it propagates along the z-axis. We use the distribution to calculate the radiation pattern and then add the Huygens source (Section 2-2.2) for polarization. The analysis is divided into far- and near-ﬁeld approximations. The near-ﬁeld approximation consists of a paraxial wave. The Gaussian beam satisﬁes Maxwell’s equations by using the free-space Helmholtz equation and produces correct patterns when applied with physical optics (PO). The free-space Green’s function satisﬁes the Helmholtz equation: e−j kR /R. We derive the Gaussian beam from a point source placed at a complex position along the z-axis: z0 = −j b. A source at this position produces a Gaussian distribution in the z = 0 plane. 2 −ρ exp with ρ 2 = x 2 + y 2 W02

GAUSSIAN BEAM

363

W0 is the beam waist radius where amplitude has dropped by 1/e. We relate the waist radius W0 to the position b by [24, pp. 80–90] W02 =

2b k

where k =

2π λ

As the wave propagates along the z-axis, its amplitude retains the Gaussian distribution in the radial direction ρ but the waist spreads: z 2

W 2 (z) = W02 1 + b The waist surface is a hyperboloid with a ring focus at radius b located at z = 0. The wave amplitude reduces by the ratio of the waists and combines with the radial Gaussian distribution:

W0 ρ2 exp − 2 W (z) W (z) The phase of the paraxial (near-ﬁeld) wave has two terms. The ﬁrst is the normal z-directed wave phase exp (−j kz) and the second is a quadratic phase term that arises from the complex location of the point source at z = −j b. The quadratic phase term slant radius depends on the location along the z-axis: 2 b Rc (z) = z 1 + z The paraxial Gaussian beam has an additional slippage phase term ζ(z) = tan−1 (z/b). The paraxial Gaussian beam phase term is the sum

ρ2 + j ζ(z) exp −j kz − j k 2Rc (z) The constant phase (eikonal) surfaces between the hyperboloid amplitude surfaces are ellipsoids with a ring focus at radius b located at z = 0. At z = 0 the eikonal surface is planar. We combine the amplitude and phase terms for the complete paraxial Gaussian beam equation:

ρ2 θ W0 ρ2 − j E0 cos2 exp − 2 + j ζ(z) exp −j kz − j k 2 W (z) W (z) 2Rc (z) × (θˆ cos φ − φˆ sin φ)

(7-31)

The Huygens source polarization for an x-directed wave [Eq. (1-38)] and the obliquity factor [Eq. (2-14)] have been added to Eq. (7-31). We determine the constant E0 by equating the radiation between this paraxial beam and the far-ﬁeld expression for a Gaussian beam with a given input power. The recommended distance to equate the two representations is z = 200W02 /λ. We calculate the far-ﬁeld Gaussian beam by substituting the point source position into e−j kR /R and approximating R with a far-ﬁeld expression [25, pp. 96–106]: R=

x 2 + y 2 + z2 − b2 + j 2bz = r 2 − b2 + j 2br cos θ

(7-32)

364

HORN ANTENNAS

In the far ﬁeld we can ignore b2 and expand Eq. (7-32) in a Taylor series and retain the ﬁrst two terms, which reduces e−j kR /R to ekb cos θ e−j kr /r. We combine this term with the Huygens source radiation to produce the far-ﬁeld Gaussian beam equation for an x-directed linear polarization in the aperture normalized at θ = 0 to directivity: E(r, θ, φ) =

e−j kr P0 · directivity · η θ cos2 ekb(cos θ−1) (θˆ cos φ − φˆ sin φ) 4π 2 r

(7-33)

The directivity is found by integrating the pattern of Eq. (7-33): directivity =

4(2 kb)2 2(2 kb) − 2 + 1/(2 kb) − e−2(2 kb) /(2 kb)

(7-34)

Scale 7-7 gives the relationship between gain and the 10-dB beamwidth for a Gaussian beam. Given the beamwidth (BW) at a given level L(dB), we solve Eq. (7-33) for the complex-plane point source position b: b=

2 log[cos(BW/4)] + |L/20| k[1 − cos(BW/2)] log e

(7-35)

Scale 7-8 relates Gaussian beam half-depth of focus, b, to its 10-dB beamwidth, and Scale 7-9 gives the minimum waist diameter. We simplify the expression for the Gaussian beam for small angles by expanding cos θ in a Taylor series cos θ ≈ 1 − θ 2 /2, which reduces Eq. (7-33): e−j kr θ 2 E(r, θ, φ) = E0 cos2 e−(θ/θ0 ) (θˆ cos φ − φˆ sin φ) 2 r

(7-36)

10-dB Beamwidth (degrees)

Gaussian Beam Gain, dB

SCALE 7-7

Gaussian beam gain compared to a 10-dB beamwidth.

Half Depth of Focus b, l

Gaussian Beam, 10-dB Beamwidth (degrees)

SCALE 7-8

Gaussian beam half-depth of focus, b, compared to a 10-dB beamwidth.

RIDGED WAVEGUIDE HORNS

365

Minimum Waist Diameter 2W, l

Gaussian Beam, 10-dB Beamwidth (degrees)

SCALE 7-9

Gaussian beam minimum waist diameter compared to a 10-dB beamwidth.

The angle θ0 is the beam divergence [24, pp. 80–90], given by 2 θ0 = = kW0

2 kb

Equation (7-36) cannot be used beyond θ0 because it is based on a small-angle approximation. We can use a Gaussian beam to approximate the pattern of a corrugated horn [26, pp. 170–176]. The minimum waist is located behind the horn aperture Lp , the phase-center distance given the aperture radius a and the slant radius R: Lp =

R 1 + [2R/k(0.644a)2 ]2

(7-37)

Lp is the location of z = 0 of the Gaussian beam. The minimum waist radius W0 is given by W02 k 0.644a (7-38) b = W0 = 1 + [k(0.644a)2 /2R]2 2 For a 22 dB-gain corrugated horn, Eq. (7-38) produces a Gaussian beam with the same gain as the horn for S = 0.134. For different values of S, Eq. (7-38) gives only approximate Gaussian beams to match the gain of corrugated horns. The Gaussian beam has a 10-dB beamwidth of 27.5◦ and the corrugated horn has a beamwidth of 27.2◦ . The phase center of the Gaussian beam given by Eq. (7-37) is 2.44λ behind the aperture and the actual horn phase center is at 0.89λ. The Gaussian beam approximation ﬁnds the near-ﬁeld pattern of the corrugated horn because it includes the ﬁnite waist size instead of assuming a point source at the phase center of the horn. A PO analysis using the equivalent currents in the aperture [27, pp. 141–156] also ﬁnds the near-ﬁeld pattern but requires greater calculation effort.

7-6 RIDGED WAVEGUIDE HORNS Inserting ridges in the E-plane of a waveguide lowers the cutoff frequency compared to a waveguide of the same width. The ridges raise the cutoff frequencies of the next two higher modes and can produce a waveguide that operates over a 10 : 1 frequency range or more. If we use this as the input waveguide to a horn and taper the ridges until they do not block the horn aperture, the horn radiates a pattern similar to the smooth-wall horn. Near the aperture the horn can support many higher-order waveguide modes

366

HORN ANTENNAS

as frequency increases. The horn generates some higher-mode content to the ﬁelds which distorts the pattern over narrow frequency ranges, but for many applications such distortions are acceptable. Initial designs [28] used dual ridges for a single linear polarization, while later designs increased the number of ridges to four (quad-ridge) to allow for dual linear (or circular) polarization. Design concentrates on the input waveguide dimensions. We apply transverse resonance to the waveguide to calculate its cutoff frequencies. A rectangular waveguide with the electric ﬁeld parallel to the narrow wall can be considered as a parallel-plate transmission line with the wave traveling between the two narrow wall shorts at cutoff (see Section 5-24). The parallel-plate transmission-line impedance is ηb for a height of b (meters). The lowest-order mode cutoff frequency for a normal rectangular waveguide occurs when the width a = λ/2. The transverse resonance method considers half the width as a transmission line and cutoff occurs when the impedance at the centerline is an open circuit (odd-order mode) or a short circuit (even-order mode) (i.e., a/2 = N λ/4) for mode TEN0 . Of course, we ignore the impedance of the parallel-plate line because it is uniform. Figure 7-13a shows the cross section of a dual-ridged waveguide. The diagram illustrates feeding the waveguide with a coaxial line running through the center of one ridge. The center conductor extends across the gap to feed the second ridge. The center pin does not need to touch the second ridge but can be coupled capacitively. The transverse resonance circuit of a dual-ridged horn used to determine cutoff frequencies consists of two transmission-line segments with a shunt capacitor due to the step. The capacitance depends on the ratio of the heights α = b2 /b1 , where b2 < b1 [29]: C=

ε0 π

4α 1 + α2 α2 + 1 cosh−1 − 2 ln α 1 − α2 1 − α2

(7-39)

For the dual-ridged waveguide we analytically place a ground plane halfway across the waveguide E-plane and divide the waveguide into two half-height waveguides. Later we will consider the impedance, and the total impedance of the guide is these two

(a)

(b)

FIGURE 7-13 Coaxial feeds of ridged waveguides: (a) dual ridge; (b) quad ridge.

RIDGED WAVEGUIDE HORNS

367

half-height guides in series. Given the waveguide width a1 and height 2b1 , and the ridge width a2 with gap 2b2 , we solve for the cutoff frequency using a transcendental equation in admittance at the transition point between the two half-height waveguides. Odd-order TE modes have a virtual open circuit in the center of the ridge and a short circuit at the wall. Cutoff occurs for kc = 2π/λc = 2πfc /c for c equal to the speed of light [30]: tan(kc a2 /2) cot[kc (a1 − a2 )/2] + kc cC − =0 (7-40) ηb2 ηb1 We solve Eq. (7-40) numerically for kc for odd-order modes. The even modes have a virtual short circuit in the center, which leads to a similar equation for the cutoff number kc : cot[kc (a1 − a2 )/2] cot(kc a2 /2) + kc cC − =0 (7-41) − ηb2 ηb1 We use Eq. (7-40) to calculate the cutoff wavelengths of modes TE10 and TE30 and Eq. (7-41) to compute the cutoff wavelength of mode TE20 for given dimensions. We design the waveguide to have a suitable low-frequency cutoff with an impedance equal to the input coax, whose outer conductor is connected to one ridge, with the center conductor jumping the gap to feed the other. The impedance at an inﬁnite frequency is given by the equations Y∞

1 = kηb2

Z∞ =

ka2 sin ka2 b2 cos2 (ka2 /2) ka1 sin ka1 + + − 4 4 b1 sin2 (ka1 /2) 4 4

π b2 ka2 2b2 ln csc cos2 + λ 2 b1 2

1 Y∞

(7-42)

The impedance at a ﬁnite frequency increases: Z∞ Z0 = 1 − (fc /f )2

(7-43)

An approximate value for the gap can be found from the impedance of microstrip. The inﬁnite impedance equals slightly less than twice the impedance of microstrip the same width as the ridge with one-half the gap. The extra fringing capacitance between the sides of the ridges lowers the impedance compared to microstrip. You can use a microstrip line design program to ﬁnd an approximate gap and a few evaluations of Eq. (7-42) to determine the correct gap. Design for Z∞ because the impedance approaches Z∞ rapidly as frequency increases by Eq. (7-43) and ridged horns operate over a large bandwidth. Quad-ridged waveguide, illustrated in Figure 7-13b, requires modiﬁcations at the input to a horn. To achieve Z∞ = 50 , the gap must be reduced and the ridges made with a rooftop shape so that they ﬁt within each other. The capacitance between the ridges for one polarization is a series combination of the two capacitors to the ridges for the second polarization. Similar to the dual-ridged waveguide, we divide the

368

HORN ANTENNAS

waveguide along the centerline through the second set of ridges and analyze a singleridged waveguide. Given a square waveguide with width w, ridge width s, and gap g between the ridges of different polarizations, the equivalent single-ridged waveguide has parameters given by the expressions √ √ a1 = w + s( 2 − 1) a2 = s 2 b1 = w − s/2

b2 = g

(7-44)

For the quad-ridged waveguide the inﬁnite impedance equals slightly less than four times the impedance of microstrip the same width as the equivalent ridge a2 with onehalf the gap. We use the parameters of Eq. (7-44) in Eqs. (7-39) through (7-43) to ﬁnd the parameters of quad-ridged waveguide. Figure 7-13b demonstrates that the feed pin of one coax passes over the other to reduce coupling between them. The difference in distance to the waveguide shorting wall for the two coaxial lines produces different impedances for the two inputs. We can use the expressions above for circular waveguides. We design with a width equal to the diameter. The inﬁnite impedance is lower by the factor π/4. The cutoff frequency is about 1.25 times the cutoff frequency of the equivalent square waveguide [31]. Figure 7-14 gives a cross-sectional drawing of a ridged horn and demonstrates the key elements of design. A coax is fed through the center of one ridge and the center conductor jumps the gap and feeds the second ridge. We locate the coax close to the end of the ridge truncated before it reaches the waveguide back wall short circuit, leaving a small gap. Without ridges the waveguide is cutoff at the low-frequency end of the horn operation. Operating the waveguide below its cutoff frequency does not

FIGURE 7-14

Dual-ridged waveguide horn cross section.

RIDGED WAVEGUIDE HORNS

369

prevent the wave from reaching the back wall because the distance is short. The original horns [28] used waveguides in cutoff at the feed point at the lowest frequencies. By tapering the sidewalls the waveguide operates above cutoff in a short distance from the feed and the waves propagated to that region. Cutoff only means that a wave will not propagate in a long waveguide, but it attenuates as it moves along the waveguide. Figure 7-14 shows optional shorting pins between the back wall and the ridges. These prevent an additional resonance in impedance that may arise at a frequency when the height of a single ridge approaches λ/2. Not all designs need these pins. We space the ridges to form a transmission line matched to the feed coax at the input. A uniform section of ridged waveguide extends to the throat of the horn. The horn shown in Figure 7-14 uses an exponentially tapered ridge that has an additional linear taper with slope 0.02 [28] empirically found to improve the impedance match. It would seem that designing a classical tapered impedance transformer would give a better impedance match, but the simple exponential physical taper produces an excellent impedance match. The gain of the horn falls short of the equivalent open horn because multiple modes are excited and beamwidth broadens. In a dual-ridged horn the power concentrates between the ridges in the E-plane, and we can replace the H -plane sidewalls with a few rods. We space the rods close enough to block radiation at the lower frequencies and allow high-frequency radiation through the spaces. Since the ﬁelds are concentrated between the ridges at high frequencies, the side H -plane walls have little effect on the pattern. A quad-ridged horn requires solid walls. A circular quad-ridged horn was measured as a possible feed for a Cassegrain reﬂector from 6 to 18 GHz. The horn has a 13.2-cm aperture diameter and a 37.6-cm slant radius and operates from 2 to 18 GHz. Figure 7-15 plots the measured E- and H -plane 10-dB beamwidths along with the beamwidths of both smooth and corrugated wall horns of the same size. Neither smooth wall nor corrugated wall horns could be designed to operate over this wide bandwidth; they are shown only for comparison. The

FIGURE 7-15 Measured 10-dB beamwidths of a circular quad-ridged horn compared to the calculated beamwidths of smooth- and corrugated-wall horns.

370

HORN ANTENNAS

FIGURE 7-16 Measured directivity of a circular quad-ridged horn compared to those of smooth- and corrugated-wall horns.

quad-ridged horn has wider beamwidths in both planes compared to the other horns. This reduces the gain shown in Figure 7-16. Similar to the corrugated horn, the quadridged horn operates with multiple modes. We can determine the circular waveguide modes radiated by using physical optics analysis on the measured pattern. We radiate a plane wave into a circular aperture plane equal to the physical horn aperture and placed at the average phase center. Each plane wave, weighted by the pattern level and sin θ , excites Huygens source currents on the patches that cover the aperture by using Eq. (2-33). We normalize the currents to 1 watt and project the currents for each mode of a circular waveguide horn onto the incident wave currents by integrating over the aperture to determine their excitation levels bm : bm =

Ja ·J∗m dS

(7-45)

S

We use the aperture currents Ja and mode currents Jm in Eq. (7-45), where we take the complex conjugate of the vector for projection in the same manner as polarization calculations (section 1-11). We operate on the electric currents only because the magnetic currents are proportional to the electric currents for Huygens sources. Figure 7-17 plots the levels of the TE11 , TM11 , and diagonally oriented TE21 modes. TE11 and TM11 modes are also excited in a corrugated horn, but the level of the TM11 mode is approximately −5 dB relative to the TE11 mode. Further measurements of the horn show that it has approximately equal power in the TE11 and TM11 modes, all the way down to 2.7 GHz. Below that frequency the horn aperture will not support the TM11 mode and the pattern reverts to the TE11 mode only, which narrows the beamwidth. Analysis shows that the diagonally oriented TE21 mode peaks at an angle halfway around from the ridges and increases cross-polarization in the diagonal plane. The unmatched beamwidths in the E- and H -planes also increases the Huygens source

RIDGED WAVEGUIDE HORNS

371

TM11 TE11

TE12

FIGURE 7-17 Modal decomposition into circular waveguide modes of the measured pattern of a circular quad-ridged horn.

cross-polarization (section 1-11.2) in the diagonal plane. Square quad-ridged horns have similar modes. Measurements on a 63.5-cm-square aperture horn with a 140-cm slant length produced nearly equal levels of TE10 and TM12 modes, similar to the TE11 and TM11 circular modes in ﬁeld distribution. The TE10 and TM12 modes have approximately the same phase. The horn radiated the TE12 mode at the higher-frequency end of the band, which caused pattern distortion over a narrow frequency range. Both the TM12 and TE12 modes are excited by the electric ﬁeld between the ridges. The interplay of these three modes causes rapid changes in the beam shape as frequency changes. The horn exhibits these changes at the high end of the frequency band when all three modes exist with nearly equal power. Measurements on a dual-ridged horn produced patterns that reduced to the same three dominant modes as the square quad-ridged horn radiated and produced similar results. We fail to obtain a close match with the measured pattern of the quad-ridged horn by using the aperture currents beyond the 10-dB beamwidth for an aperture small in wavelengths. If we include currents excited along the outside of the horn bell in the physical optics analysis, we better match the measured pattern. This illustrates that the pattern of a horn is determined not only by aperture ﬁelds but also by the currents that ﬂow down the bell. Figure 7-18 shows the measured E- and H -plane patterns and the cross-polarization in the diagonal plane. The three-dimensional measured pattern plot in Figure 7-19 at 6 GHz shows the four cross-polarization lobes in the diagonal planes. The average pattern beamwidth matches a reﬂector with f/D = 1 and has an average illumination loss of 3 dB, with the value ranging from 2.5 to 4 dB. The average taper loss is 1.07 dB and the average spillover loss 1.08 dB. The cross-polarization exhibited in Figure 7-19 contributes an average 0.7 dB of loss. The phase-center location measurements show that the horn has up to 2λ astigmatism, which contributes 0.4 dB of loss when used as a reﬂector feed.

372

HORN ANTENNAS

FIGURE 7-18 Measured pattern of a circular quad-ridged horn.

Huygens Polarization, Co-Polarization

Huygens Polarization, Cross-Polarization

FIGURE 7-19 Spherical radiation pattern of a circular quad-ridged horn showing four-way symmetry of cross-polarization in diagonal planes.

7-7 BOX HORN [32, pp. 377–380] With a box horn (Figure 7-20), multiple waveguide modes are used to decrease the H -plane amplitude taper loss and axial length of the horn. We add the TE30 mode to

BOX HORN

373

FIGURE 7-20 Box horn.

the TE10 mode to reduce the cosine distribution taper of the H -plane. By phasing the modes 180◦ out of phase in the center of the aperture, the cos 3πx distribution subtracts from the TE10 -mode distribution in the center and adds in the region near the edges. A step in the width of a waveguide generates TEN0 modes when fed by the TE10 mode. Any modes not cut off by the waveguide will propagate to the aperture. If we maintain symmetry about the axis of the waveguide, only odd-order modes (TE30 , TE50 , etc.) will be generated. The width W of the waveguide (Figure 7-20) beyond the step determines the possible propagating modes: λc = 2W/N , where N is the mode number. If we limit the modes to the TE10 and TE30 modes in the aperture, the cutoff wavelength of the TE50 mode determines the maximum width: Wmax = 2.5λ. The TE30 -mode cutoff wavelength establishes the minimum width: Wmin = 1.5λ. Within this range, short horns with good aperture efﬁciency can be designed. We can ﬂare the E-plane to increase its aperture (Figure 7-20), but the limited axial length of the horn bounds the possible ﬂare without an excessive phase error loss. The H -plane can also be ﬂared, but ﬂaring it complicates the design for the proper length L. The step generates smaller amplitudes of higher-order modes with each increase in N . Small amounts of higher-order modes (TE50 , TE70 , etc.) will decrease the efﬁciency only marginally, since the mode amplitudes are small. The step generates modes in phase with the input TE10 mode, since the higher-order modes must peak in the center and subtract from the TE10 ﬁelds on the back wall of the larger waveguide section. The aperture distribution is a sum of TE10 and TE30 modes: Ey (x) = a1 cos

πx 3πx exp(−j k10 L) + a3 cos exp(−j k30 L) W W

(7-46)

where k10 and k30 are the propagation constants of the two modes. The amplitude distribution in the H -plane will be more nearly uniform if the phase between the modes is 180◦ . The modes travel from the step with different phase velocities, depending on their cutoff frequencies. We adjust the length L to give a 180◦ phase difference between the modes: (k10 − k30 )L = π where k10 = k 1 − (λ/2 W )2 and k30 = k 1 − (3λ/2 W )2 . We solve for the length: λ/2 L= 2 1 − (λ/2 W ) − 1 − (3λ/2 W )2

(7-47)

374

HORN ANTENNAS

TABLE 7-10 TE30 /TE10 (Voltages)

Box Horn Characteristics Ratio of Input Waveguide to Aperture

Linear ATLw (dB)

3 dB

10 dB

1.000 0.940 0.888 0.841 0.798 0.758 0.719 0.682 0.645 0.609 0.573 0.537 0.500 0.462 0.424

0.91 0.78 0.67 0.58 0.52 0.48 0.46 0.46 0.47 0.50 0.54 0.60 0.66 0.74 0.82

0.594 0.575 0.558 0.544 0.530 0.518 0.507 0.496 0.487 0.479 0.471 0.463 0.456 0.450 0.444

1.019 0.981 0.947 0.917 0.890 0.866 0.844 0.824 0.806 0.790 0.775 0.761 0.749 0.737 0.726

0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70

(W /λ) sin θ

The ratio of the modes generated by the step can be found from mode matching on the input waveguide aperture of width a:

a/2

cos(πx/a) cos(N πx/W )dx aN −a/2 = a/2 a1 cos(πx/a) cos(πx/W )dx

(7-48)

−a/2

where aN is the TEN0 mode amplitude. Table 7-10 lists the step dimensions needed to design to a given ratio of modes. The amplitude taper loss is a minimum at a3 /a1 = 0.32. The possible 3-dB beamwidths with a single mode, TE10 , range from 20 to 44◦ . Example Design a box horn with an H -plane 10-dB beamwidth of 50◦ . We pick a3 /a1 = 0.35. From Table 7-10, (W/λ) sin θ = 0.824. The obliquity factor at 25◦ adds 0.42 dB of loss. We must design with a wider 10-dB beamwidth. This is within the permissible range for only two modes in the aperture. We calculate the length to phase the modes by 180◦ by using Eq. (7-47): L = 1.451λ. The horn is shorter than the aperture width.

7-8 T-BAR-FED SLOT ANTENNA A T-bar-fed slot antenna (Figure 7-21) looks more like an open-circuited waveguide to coax transition than a slot. Like a slot, its pattern is very broad. The antenna has been designed experimentally [33, pp. 184–190] and those dimensions provide a good starting point. Table 7-11 lists two designs [33] referred to Figure 7-21. The aperture admittance is a combination of the radiation admittance and a capacitive

T-BAR-FED SLOT ANTENNA

375

b

a w x

x

Outline of Cavity G

b

Back Wall

b/2

F E

D H

Aperture a Front View

W Side View

FIGURE 7-21 T-bar-fed cavity slot antenna. (From [34], 1975 IEEE.)

TABLE 7-11 Dimensions for Two Antenna Designs Dimensions

Antenna 1

Antenna 2

b/a W/a x/a D/a I/a E/a F/a

0.323 0.323 0.118 0.118 0.059 0.118 0.057

0.226 0.295 0.113 0.090 0.045 0.090 0.045

susceptance. Behind the feed point, the length of short-circuited waveguide adds an inductive susceptance that grows as frequency decreases. The horizontal bar produces a capacitive susceptance at the input to counteract the back-wall susceptance. These susceptances track with frequency changes; each one decreases to maintain the sum near resonance. Later experimental work [34] revealed further properties of the antenna. Measurements on antenna 1 show that the lower-end 2 : 1 VSWR band edge occurs when a = 0.57λ and the upper end when a = 0.9λ. The bandwidth is about 1.6 : 1.

376

HORN ANTENNAS

Antenna 2 was reported [33] as having less bandwidth than antenna 1. When the round rod was replaced with a ﬂat strip, whose width across the guide was the same as the diameter of the rod, almost identical results were obtained. We have a choice. The ﬂat strip is an easier construction, but the round rod gives better mechanical support in all axes to withstand shock and vibration. The ﬂat strip adds to the freedom of design. The bandwidth potential increases when H is decreased while I is held constant. Newman and Thiele [34] found that when H was decreased, the nominal impedance level was raised. When the input impedance is plotted on the Smith chart, the locus is centered about a higher resistance. By adding a broadband impedance transformer on the input, we can achieve the higher bandwidth potential. Newman and Thiele achieved a nearly 2 : 1 VSWR bandwidth from a = 0.52λ to a = 1.12λ, or 2.3 : 1 bandwidth.

7-9 MULTIMODE CIRCULAR HORN [35] A step in the diameter of a circular waveguide generates a TM11 mode to satisfy the boundary conditions. The ﬁelds of the TM11 mode can be phased to cancel the ﬁelds from the TE11 mode at the edges of the aperture in the E-plane. The tapering of the ﬁelds in the aperture reduces the E-plane sidelobes while broadening the beamwidth. Equalizing the ﬁeld distributions in the two planes helps to bring the E- and H -plane phase centers closer together. The modes generated by the step are more complicated than those for the box horn. Symmetry eliminates generation of the unwanted modes: TM01 , TE21 , and TE01 . The step transition shifts the phase of the TM11 mode relative to the TE11 mode [36]. Since the waveguide modes have different phase velocities, they can be phased to produce the desired ﬁeld at the aperture. Although calculated information [36] is helpful, the designs must be completed empirically. The required phasing to achieve ﬁeld cancellation limits the bandwidth, but for narrowband applications a stepped horn is cheaper than a corrugated horn. Satoh [37] loads the ﬂare of a conical horn with a conical dielectric step to generate the TM11 mode. Symmetry prevents the excitation of unwanted modes. He places the step at a diameter where the TM11 mode can propagate. By using two steps, the bandwidth can be increased because the lengths can be adjusted to give perfect mode cancellation at two frequencies. We can replace the dielectric cone by metal steps each of which generates the TM11 mode and thereby achieve good results, in theory, at multiple frequencies.

7-10

BICONICAL HORN [4]

A biconical horn consists of two cones with a common vertex. The angle of the generating lines of the cones is measured from a common axis. The cones of the usual antenna have angles that sum to 180◦ . Spherical modes describe the ﬁelds between the cones, but we can use approximations with good results. The lowest-order mode is TEM between the cones and is easily excited by a coax line. The outer conductor connects to one cone, and the second cone feeds out of the center conductor. The electric ﬁeld of the TEM mode is polarized in the direction of the axis. The ﬁrst higher-order mode

BICONICAL HORN

377

has a circulating electric ﬁeld with the magnetic ﬁeld in the direction of the axis. It can be excited either from a TE01 -mode circular waveguide or by an array of slots on a cylinder. The distance between cones must be at least λ/2 at the point of excitation of the TE01 biconical mode. We approximate the distribution of the zeroth-order mode, TEM, as uniform along the axis. The ﬁrst-order mode, TE01 , distribution is approximately cosine along the axis. We calculate gain by using aperture distribution losses. We describe the horn with a slant radius along the generating line and a height between the ends of the cones. The expansion in spherical modes requires integration over a spherical cap at the aperture if a constant phase surface is used. We obtain good results by using a cylindrical aperture and a quadratic phase distribution. The antenna has circular symmetry about the z-axis that bounds the directivity to 2L/λ. We use linear distribution efﬁciencies to compute directivity (gain): 2L gain = 10 log (7-49) − ATLx − PELx λ The TEM mode has a uniform distribution, so we use the “uniform” column of Table 4-42 to calculate phase error loss. The uniform distribution has no amplitude taper loss. The cosine distribution of the ﬁrst-order mode requires an ATL = 0.91 dB and use of the cosine distribution quadratic phase error loss of Table 4-42. Given the height between the ends of the cones, H , and the slant radius R, we determine the quadratic phase distribution constant from S=

H2 8λR

(7-50)

Example Compute the gain of a biconical horn with a slant radius of 10λ and cone angles of 75◦ and 105◦ . H = 2R cos 75◦ = 5.176λ and S = 0.33. From Table 4-42, we read PELTEM = −1.76 dB

PELTE−01 = −0.79 dB

Vertical mode, TEM: gain = 10 log[2(5.176)] + PELTEM : 10.15 dB − 1.76 dB = 8.4 dB Horizontal mode, TE01 : gain = 10 log[2(5.176)] + PELTE−01 + ATLcosine : 10.15 dB − 0.79 dB − 0.91 dB = 8.45 dB We can calculate beamwidths by using the results of the rectangular horn, where we measure the angles from θ = 90◦ for the complementary-angled biconical horn. Example Compute the 3-dB beamwidths of the horn above. S = 0.33 and H = 5.176λ. Use Table 7-2 with the TEM mode and α as the angle from θ = 90◦ : H sin α = 0.5015 λ

◦

α = 5.56

378

HORN ANTENNAS

The obliquity factor is insigniﬁcant. ◦

HPBW = 11.1

TEM mode

Use Table 7-1 with the TE01 mode. H sin α = 0.6574 λ

◦

HPBW = 14.6

TE01 mode

The two modes have about the same gain, but the TE01 mode has a greater beamwidth. When we refer to Figures 7-3 and 7-4, we see that the TEM-mode horn has about 7-dB sidelobes and the TE01 -mode horn has practically no sidelobes. The sidelobes reduce the gain of the TEM mode with its narrower beamwidth.

REFERENCES 1. A. W. Love, ed., Electromagnetic Horn Antennas, IEEE Press, New York, 1976. 2. W. L. Barrow and L. J. Chu, Theory of electromagnetic horn, Proceedings of IRE, vol. 27, January 1939, pp. 51–64. 3. M. C. Schorr and E. J. Beck, Electromagnetic ﬁeld of the conical horn, Journal of Applied Physics, vol. 21, August 1950, pp. 795–801. 4. S. A. Schelkunoff and H. Friis, Antenna Theory and Practice, Wiley, New York, 1952. 5. P. M. Russo et al., A method of computing E-plane patterns of horn antennas, IEEE Transactions on Antennas and Propagation, vol. AP-13, no. 2, March 1965, pp. 219–224. 6. J. Boersma, Computation of Fresnel integrals, Mathematics of Computation, vol. 14, 1960, p. 380. 7. K. S. Kelleher, in H. Jasik, ed., Antenna Engineering Handbook, McGraw-Hill, New York, 1961. 8. D. G. Bodnar, Materials and design data, Chapter 46 in R. C. Johnson, ed., Antenna Engineering Handbook, 3rd ed., McGraw-Hill, New York, 1993. 9. E. H. Braun, Gain of electromagnetic horns, Proceedings of IRE, vol. 41, January 1953, pp. 109–115. 10. E. I. Muehldorf, The phase center of horn antennas, IEEE Transactions on Antennas and Propagation, vol. AP-18, no. 6, November 1970, pp. 753–760. 11. A. J. Simmons and A. F. Kay, The scalar feed: a high performance feed for large paraboloid reﬂectors, Design and Construction of Large Steerable Aerials, IEE Conference Publication 21, 1966, pp. 213–217. 12. B. M. Thomas, Design of corrugated conical horns, IEEE Transactions on Antennas and Propagation, vol. AP-26, no. 2, March 1978, pp. 367–372. 13. T. S. Chu and W. E. Legg, Gain of corrugated conical horn, IEEE Transactions on Antennas and Propagation, vol. AP-30, no. 4, July 1982, pp. 698–703. 14. G. L. James, TE11 to HE11 mode converters for small angle corrugated horns, IEEE Transactions on Antennas and Propagation, vol. AP-30, no. 6, November 1982, pp. 1057–1062. 15. P. J. B. Clarricoats and P. K. Saha, Propagation and radiation of corrugated feeds, Proceedings of IEE, vol. 118, September 1971, pp. 1167–1176. 16. A. W. Rudge et al., eds., The Handbook of Antenna Design, Vol. 1, Peter Peregrinus, London, 1982.

REFERENCES

379

17. G. L. James and B. M. Thomas, TE11 -to-HE11 corrugated cylindrical waveguide mode converters using ring-loaded slots, IEEE Transactions on Microwave Theory and Techniques, vol. MTT-30, no. 3, March 1982, pp. 278–285. 18. R. Wohlleben, H. Mattes, and O. Lochner, Simple small primary feed for large opening angles and high aperture efﬁciency, Electronics Letters, vol. 8, September 21, 1972, pp. 474–476. 19. A. D. Olver et al., Microwave Horns and Feeds, IEEE Press, New York, 1994. 20. G. L. James, Radiation properties of 90◦ conical horns, Electronics Letters, vol. 13, no. 10, May 12, 1977, pp. 293–294. 21. A. Kumer, Reduce cross-polarization in reﬂector-type antennas, Microwaves, March 1978, pp. 48–51. 22. P.-S. Kildal, Foundations of Antennas, Studentlitteratur, Lund, Sweden, 2000. 23. S. Maci et al., Diffraction at artiﬁcially soft and hard edges by using incremental theory of diffraction, IEEE Antennas and Propagation Symposium, 1994, pp. 1464–1467. 24. B. A. Saleh and M. C. Teich, Fundamentals of Photonics, Wiley, New York, 1991. 25. K. Pontoppidan, ed., Technical Description of Grasp 8, Ticra, Copenhagen, 2000 (self published and available at www.ticra.com). 26. P. F. Goldsmith, Quasioptical Systems, IEEE Press, New York, 1998. 27. L. Diaz and T. A. Milligan, Antenna Engineering Using Physical Optics, Artech House, Boston, 1996. 28. J. L. Kerr, Short axial length broadband horns, IEEE Transactions on Antennas and Propagation, vol. AP-21, no. 5, September 1973, pp. 710–714. 29. J. R. Whinnery and H. W. Jamieson, Equivalent circuits for discontinuities in transmission lines, Proceedings of IRE, vol. 32, no. 2, February 1944, pp. 98–114. 30. S. B. Cohn, Properties of ridge waveguide, Proceedings of IRE, vol. 35, no. 8, August 1947, pp. 783–788. 31. M. H. Chen, G. N. Tsandoulas, and F. W. Willwerth, Modal characteristics of quadrupleridged circular and square waveguide, IEEE Transactions on Microwave Theory and Techniques, vol. MTT-21, August 1974, pp. 801–804. 32. S. Silver, ed., Microwave Antenna Theory and Design, McGraw-Hill, New York, 1949. 33. A. Dome and D. Lazarno, Radio Research Laboratory Staff, Very High Frequency Techniques, McGraw-Hill, New York, 1947, pp. 184–190. 34. E. H. Newman and G. A. Thiele, Some important parameters in the design of T-bar fed slot antennas, IEEE Transactions on Antennas and Propagation, vol. AP-23, no. 1, January 1975, pp. 97–100. 35. P. D. Potter, A new horn antenna with suppressed sidelobes and equal beamwidths, Microwaves, vol. 6, June 1963, pp. 71–78. 36. W. J. English, The circular waveguide step-discontinuity mode transducer, IEEE Transactions on Microwave Theory and Techniques, vol. MTT-21, no. 10, October 1973, pp. 633–636. 37. T. Satoh, Dielectric loaded horn antenna, IEEE Transactions on Antennas and Propagation, vol. AP-20, no. 2, March 1972, pp. 199–201.

8 REFLECTOR ANTENNAS

The importance of reﬂector antennas cannot be overstated. Large-aperture antennas can be built only with reﬂectors or arrays and reﬂectors are far simpler than arrays. The arrays give us more degrees of freedom than is necessary in many applications. With plenty of room and slow scan rates, a reﬂector becomes a better design than an array. Of course, there can be many valid reasons for using an array in an application, but a reﬂector should always be considered. An array needs an elaborate feed network, whereas a reﬂector uses a simple feed and free space as its feed network. Most reﬂector designs require extensive calculations together with full characterization of the feed antenna. Many types of analysis have been developed. As with horn antennas, Love [1] has collected signiﬁcant papers on reﬂector antennas. In his classic book, Silver [2] provides the foundation for an analysis based on aperture theory and physical optics (induced currents on the reﬂector). Aperture theory or physical optics reduced to aperture theory is still used for most designs. Rusch and Potter fully develop aperture and physical optics theories for the design and analysis of both prime focus and dual-reﬂector (Cassegrain) antennas [3]. Other methods have been developed either to increase the range of valid patterns or to decrease the pattern calculation time so that optimization techniques can be applied. Wood [4] collects ideas for designing by using spherical wave expansions that allow for an overall system optimization using only a few terms. GTD methods [5,6] ﬁnd increasing applications as an analysis technique suitable for a full pattern analysis except at boresight. Improved methods of calculating the secondary pattern have been developed using aperture ﬁelds, such as FFT methods [7] and Jacobi–Bessel series [8]. Many of these techniques and hardware implementations of reﬂectors are summarized in a handbook [9, Chaps. 2 and 3]. Although all these methods are available, aperture theory and physical optics remain the main techniques of reﬂector design and analysis. Modern Antenna Design, Second Edition, By Thomas A. Milligan Copyright 2005 John Wiley & Sons, Inc.

380

PARABOLOIDAL REFLECTOR GEOMETRY

381

8-1 PARABOLOIDAL REFLECTOR GEOMETRY Figure 8-1 shows the geometry of the parabolic reﬂector. We form the reﬂector by rotating the ﬁgure about its axis or by moving the ﬁgure along an axis out of the paper to form a cylindrical reﬂector. Because the cylindrical reﬂector requires a line source, it is less important than the circularly symmetrical reﬂector fed from a single point source. A paraboloidal reﬂector transforms a spherical wave radiated by the feed located at its focus into a plane wave. Although the feed wave spreads from the focus, which reduces its amplitude, geometric optics predicts a plane wave reﬂection that remains constant. The reﬂected wave does not remain a plane wave but spreads because the ﬁelds must be continuous across the reﬂection boundary of the beam plane wave column because ﬁelds can be discontinuous only across physical boundaries. Nevertheless, we will use the aperture theory on the projected diameter to predict its performance. Since the reﬂected rays are parallel, we can place the aperture plane anywhere along the axis, but somewhat close in front of the reﬂector. The equations for the reﬂector surface are f cos2 (ψ/2) polar coordinates

ρ=

r 2 = 4f (f + z) rectangular coordinates

(8-1)

where f is the focal length, D the diameter, ρ the distance from the focus to the reﬂector, and ψ the feed angle from the negative z-axis. The reﬂector depth from the rim to the center is z0 = D 2 /16f .

Parabolic Reflector

^ n r

r y

Focus z

D F y0

FIGURE 8-1

Geometry of a parabolic reﬂector.

382

REFLECTOR ANTENNAS Feed Subtended Angle (degrees)

Parabola f/D

SCALE 8-1

Parabola f/D compared to a feed total subtended angle.

We eliminate the dimensions of the reﬂector by using the ratio f /D. The half subtended angle of the reﬂector, ψ0 , relates to f/D by ψ0 = 2 tan−1

1 4f/D

(8-2)

Scale 8-1 computes the total feed subtended angle from reﬂector f/D. When we place the aperture plane at the focus, the ray path distance becomes ρ + ρ cos ψ = 2ρ cos2

ψ = 2f 2

all ray path lengths are equal, and the aperture plane is a constant-phase surface (eikonal). The normal unit vector at a point on the reﬂector (r, z) is found from the feed angle: nˆ = − sin

ψ ψ rˆ + cos zˆ 2 2

At this point we need the radius of curvatures in the principal planes to apply Eq. (2-77) reﬂection from a curved surface: R1 in the r –z plane and R2 in the φ –z plane: R1 =

2f cos3 (ψ/2)

and R2 =

2f cos(ψ/2)

The spherical wave spreads from the feed as 1/ρ. At the surface of the reﬂector the wave curvature changes to a plane wave and propagates to the aperture plane at a constant amplitude. The spherical wave spreading multiplies the feed distribution by [Eq. (8-1)] cos2 (ψ/2) in the aperture. Then added edge taper = cos2

ψ0 2

voltage

(8-3)

Deeper reﬂectors (smaller f/D) have greater edge tapers than shallow reﬂectors (larger f/D). Scale 8-2 provides a quick calculation of the added edge taper due to spherical wave spreading. Example Calculate the edge taper of a paraboloidal reﬂector for f/D = 0.5 and an isotropic feed. From Eq. (8-2), ψ0 = 2 tan−1 12 = 53.13◦ . The edge taper is [Eq. (8-3)] edge taper = 20 log cos2

53.13◦ = −1.94 dB 2

PARABOLOIDAL REFLECTOR APERTURE DISTRIBUTION LOSSES

383

Added Edge Taper, dB

Parabola f/D

SCALE 8-2 Added edge taper due to a spherical wave from feed.

If the feed has its 10-dB pattern point directed toward the reﬂector edge, the aperture edge taper is 11.9 dB.

8-2 PARABOLOIDAL REFLECTOR APERTURE DISTRIBUTION LOSSES We manipulate Eq. (4-2) for ATL to eliminate the dimensions and relate the integrals to the feed pattern: 0

ATL =

2π

b 2π a

πa 2

a

Ea (r , φ ) r dr dφ

b

0

2

Ea (r , φ )2 r dr dφ

(8-4)

where a is the aperture radius, b the central blockage radius, and Ea (r , φ ) the aperture ﬁeld. We make the following substitutions into Eq. (8-4): r = ρ sin ψ = 2 sin

ψ ψ ψ f cos = 2f tan 2 2 2 cos (ψ/2) 2

ψ = ρdψ dr = f sec 2

(8-5)

2

The aperture ﬁeld is related to the feed pattern by Ea (r , φ ) =

E(ψ , φ ) ρ

These substitutions eliminate dimensions in Eq. (8-4):

2π 0

ATL = π[tan2 (ψ

0 /2)

−

ψ0 ψb

2 |E(ψ, φ)| tan(ψ/2) dψ dφ

tan2 (ψ

2π

b /2)] 0

ψ0 ψb

(8-6) |E(ψ, φ)| sin ψ dψ dφ 2

where ψb = 2 tan−1 [b/(2f )]. When we substitute the relations in Eq. (8-5) into Eq. (4-9) to eliminate dimensions in the integrals, we obtain an expression with only the feed

384

pattern:

REFLECTOR ANTENNAS

2π ψ0 2 E(ψ, φ) tan(ψ/2) dψ dφ ψ 0 PEL = b 2 2π ψ0 |E(ψ, φ)| tan(ψ/2) dψ dφ

(8-7)

ψb

0

PEL is the efﬁciency at the boresight. We modify Eq. (8-7) when we scan the beam to give off-boresight values as in Eq. (4-3). The amplitude taper efﬁciency (ATL) of Eq. (8-6) and the phase error efﬁciency (PEL) of Eq. (8-7) do not account for the total directivity loss of the aperture. The reﬂector does not intercept all the power radiated by the source and some of it spills over the edge. Spillover adds little to the pattern except as sidelobes, since usual feeds have small backlobes. We consider this spilled-over power as a loss (SPL):

2π

2π

0

SPL =

ψ0 ψb π

0

|E(ψ, φ)|2 sin ψ dψ dφ (8-8) |E(ψ, φ)| sin ψ dψ dφ 2

0

This expression for spillover includes the scattered portion of the central blockage efﬁciency, but not the loss of potential aperture. We include the remainder in the directivity calculation. We have ignored the cross-polarized power radiated by the source. We deﬁne crosspolarization efﬁciency (XOL) as XOL =

2π

0

π

2π 0

π

|EC (ψ, φ)|2 sin ψ dψ dφ

0

(8-9)

(|EC (ψ, φ)| + |EX (ψ, φ)| ) sin ψ dψ dφ 2

2

0

where Ec is the co-polarized ﬁeld and Ex is the cross-polarized ﬁeld. These polarizations correspond to Ludwig’s [10] third deﬁnition of cross-polarization. A Huygens source produces straight reﬂector surface currents when projected to the aperture plane. Including the cross-polarization efﬁciency gives us the true average radiation intensity as in Eq. (1-17). If we express the efﬁciencies as ratios, the directivity will be found from directivity =

π 2 λ

(Dr2 − Db2 )SPL · ATL · PEL · XOL (ratio)

(8-10)

where Dr is the reﬂector diameter and Db is the diameter of the central blockage. Equation (8-10) includes the nonscattered blockage loss of potential aperture. Equation (8-10) can be expressed in terms of decibel ratios: π 2 2 2 directivity = 10 log (Dr − Db ) + SPL(dB) + ATL(dB) λ + PEL(dB) + XOL(dB)

(8-11)

APPROXIMATE SPILLOVER AND AMPLITUDE TAPER TRADE-OFFS

385

Of course, all the decibel ratios of the efﬁciencies will be negative and subtract from the directivity calculated from the area. When measuring an actual feed, we can ignore the cross-polarized power. We measure the efﬁciency as the difference between directivity and gain. Actual directivity includes the co-polarizations and cross-polarizations in the average radiation intensity. If we ignore the cross-polarization, the measured efﬁciency decreases by the crosspolarization loss because the measured and true directivity differ by that loss. We must measure the cross-polarization pattern distribution of the feed if we want to calculate the cross-polarized secondary (reﬂector) pattern. When the cross-polarization pattern is not required, we save time without loss of accuracy by measuring only the co-polarized feed pattern. Equations (8-8) and (8-9) are by no means unique. We could include the crosspolarized power in the spillover calculation [Eq. (8-8)] and limit the integration limits in Eq. (8-9) to the reﬂector. A set of efﬁciency relations is correct when the equations account for all the power radiated by the feed. When we use calculated feed patterns, we must determine cross-polarization efﬁciency, since we can only estimate the efﬁciency due to material losses. The cross-polarization efﬁciency cannot be included as it is in measurements, and the division of cross-polarized power between Eqs. (8-8) and (8-9) is arbitrary.

8-3 APPROXIMATE SPILLOVER AND AMPLITUDE TAPER TRADE-OFFS We use the approximate pattern cos2N (ψ/2) for a feed pattern to establish trends. Of course, if the actual feed pattern distribution is available, we should use Eqs. (86) to (8-9). We obtain closed-form expressions when we substitute this pattern into Eqs. (8-6) and (8-8). Ignoring any central blockage, we get spillover efﬁciency = 1 − u2(N+1) amplitude taper efﬁciency =

4(N + 1)(1 − uN )2 ψ0 cot2 N 2 [1 − u2(N+1) ] 2

(8-12) (8-13)

where u = cos(ψ0 /2). We combine Eqs. (8-12) and (8-13) and plot their combination to ﬁnd the beamwidth for minimum loss. In Figure 8-2 the loss versus the 10-dB beamwidth for various f/D values is plotted. At narrow beamwidths little feed power spills over the reﬂector edge, but the reﬂector is underilluminated. Increasing the beamwidth improves the illumination but increases the spillover. The efﬁciency peaks when the feed 10-dB beamwidth is approximately the subtended angle of the reﬂector. Figure 8-2 shows a broad peak for any given f/D. Small changes in the beamwidth near the peak have no practical effect on the reﬂector’s gain. Scale 8-3 relates the average illumination loss reduction given the feed pattern level in the direction of the reﬂector rim for typical antennas. Example Estimate the amplitude taper loss for a reﬂector with f/D = 0.5 whose feed has a 10-dB edge taper. Compare the loss with that of the circular Gaussian and the Hansen single-parameter distributions: ψ0 = 2 tan−1 12 = 53.13◦ . The 10-dB beamwidth of the feed is then 106.26◦ . We modify Eq. (1-20) to compute the exponent N of the cos2N (ψ/2) feed

386

REFLECTOR ANTENNAS

Spillover + Amplitude Taper Loss, dB

0

F/D 2 1.5

1 0.8

0.6

1

0.5

0.4 0.3

2

3 0

50°

100°

150°

200°

Feed 10-dB Beamwidth

FIGURE 8-2 Sum of spillover and amplitude taper losses versus feed 10-dB beamwidth. Feed Level in Direction of Rim, dB

Mean Illumination Loss Change f /D : 0.3 - 0.6, dB

SCALE 8-3 Mean illumination loss change of a reﬂector given the feed pattern level in the rim direction.

pattern approximation, N=

log 0.1 = 10.32 2 log cos(106.26◦ /4)

From Eq. (8-13), u = cos(53.13◦ /2) = 0.894: 4(1 − 0.89410.32 )2 (11.32) ◦ cot2 (53.13 /2) = 0.864 10.322 (1 − 0.89422.64 ) = 10 log 0.864 = −0.63 dB

ATL(dB) =

The extra distance from the feed to the reﬂector edge compared with the center distance adds 1.94 dB and increases the aperture amplitude taper to 11.94 dB. We interpolate Table 4-29 for the circular Gaussian distribution and Table 4-30 for the Hansen singleparameter distribution to ﬁnd the following data: Gaussian ATL(dB) = −0.62 dB Sidelobe level = 26.3 dB Beamwidth factor = 1.142

Hansen ATL(dB) = −0.57 dB Sidelobe level = 24.7 dB Beamwidth factor = 1.136

PHASE ERROR LOSSES AND AXIAL DEFOCUSING

387

We multiply Eq. (4-83) by the beamwidth factor to estimate the reﬂector beamwidth: ◦

HPBW = 67.3

λ D

◦

and HPBW = 67

λ D

These compare well with the approximation, HPBW = 70◦ λ/D for a parabolic reﬂector. An integration of the aperture distribution for the far-ﬁeld pattern gives the following results: ◦

HPBW = 67.46

λ D

sidelobe level = 27 dB

8-4 PHASE ERROR LOSSES AND AXIAL DEFOCUSING All rays starting at the reﬂector focus travel the same distance through reﬂection to the aperture plane. The aperture plane is any convenient plane in front of the dish whose normal is the axis of the reﬂector. If we could build a feed with a unique phase center and place it at the focus of a perfect paraboloidal reﬂector, we would eliminate phase error loss in the aperture plane because it would have a constant phase. The feed, the positioning of the feed, and the reﬂector surface all contribute to the phase error loss. We discussed techniques for obtaining unique phase centers in the various planes for horns. Unlike smooth-wall horns, corrugated horns can have equal phase centers in all planes through the axis, but even their position will wander with changes in frequency. We measure the feed pattern distribution (amplitude and phase) to predict the contribution of the feed to the overall efﬁciency. From those measurements we deﬁne the practical phase center as the point on the feed leading to the minimum phase error loss when placed at the focus. The random and systematic phase error contributions can be measured directly on the feed and calculated numerically using Eq. (8-7). The feed phase center cannot always be placed at the focus. The phase-center location wanders with changes in frequency, and in any wideband application we expect axial defocusing. For example, the location of the phase center of a log-periodic antenna moves toward the apex as frequency increases. Figure 8-3 is a plot of the phase error loss due to axial defocusing. Each feed has its 10-dB beamwidth equal to the reﬂector subtended angle. Axial defocusing affects deep dishes (lower f/D) more than shallow dishes. We can estimate the axial defocusing phase error loss by approximating the distribution with a quadratic aperture phase distribution. Given z as the axial defocusing, the maximum phase deviation in cycles is z 1 −1 S= 1 − cos 2 tan λ 4f/D

(8-14)

We combine this with the quadratic phase error loss of the circular Gaussian distribution to estimate the loss. With z = λ we obtain a scaling factor for S (Scale 8-4) given z from Eq. (8-14). The scaling factor decreases with increasing f/D. Example Estimate the phase error loss for z = 2λ when f/D = 0.6 and the feed 10-dB beamwidth equals the reﬂector subtended angle.

388

REFLECTOR ANTENNAS

2 1 1.6 1.4

Phase Error Loss, dB

2

1.2

3

4

5

6

0.3

0.6 1

2

0.8

F/D = 1

3 4 Axial Defocusing, l

5

6

FIGURE 8-3 Paraboloidal reﬂector phase error loss due to axial defocusing of the feed.

Axial Defocusing Quadratic Phase Factor, S

Parabola f /D

SCALE 8-4

Quadratic phase factor S for axial defocusing of a paraboloidal reﬂector.

From Scale 8-4, S = 0.30(2) = 0.60. We use Eq. (8-3) to compute the edge taper: 1 ◦ = 45.2 2.4 ψ0 = −1.4 dB edge taper = 20 log cos2 2 ψ0 = 2 tan−1

An equivalent truncated Gaussian aperture distribution tapers to 10 dB + 1.4 dB = 11.4 dB

ρ=

11.4 = 1.31 8.69

We use Eq. (4-118) to calculate phase error efﬁciency of the truncated Gaussian distribution: PEL = 0.305 or PEL (dB) = −5.2 dB. This matches the value from Figure 8-3 found by integration of the actual distribution. The optimum feed beamwidth produces an average aperture edge taper of 11.8 dB. Scale 8-5 evaluates Eq. (4-118) for this taper. We detect axial defocusing by looking at the patterns of the reﬂector. Axial defocusing ﬁlls-in nulls between sidelobes. We adjust the feed location to maximize the null

ASTIGMATISM

389

Truncated (-11.8 dB) Circular Gaussian Distribution PEL, dB

Quadratic Phase Factor, S

SCALE 8-5 given S.

Truncated circular Gaussian distribution (−11.8 dB taper) phase error loss

depth, but antenna range errors and receiver sensitivity limit our ability to eliminate this defocusing. 8-5 ASTIGMATISM [11] Both the feed and the reﬂector can have astigmatism: unequal phase centers in different planes. We measure the feed by itself to discover its astigmatism. When the feed is mounted in the reﬂector, we detect astigmatism by the depth of nulls in the various pattern planes. A series of measurements can separate the feed and reﬂector astigmatism, but the feed must be able to move along the reﬂector axis and to rotate by 90◦ during the measurements. Move the feed along the axis to ﬁnd the locations that give maximum nulls. The extrema of the reﬂector focuses may not occur in the E- and H -planes and will require a search in the other planes. At this point we cannot separate the feed astigmatism from the reﬂector astigmatism. We rotate the feed and repeat the measurements. The feed phase center locations shift, and the reﬂector focuses remain ﬁxed. Simple manipulation of the data from the two measurements separates the two sources of astigmatism. The reﬂector can be shimmed to remove its astigmatism, or the feed phase centers can be matched to the reﬂector focuses. Figure 8-4 shows the magnitude

1.5 1

Phase Error Loss, dB

1 2 0.8 0.6

3 0.5 4 F/D = 0.3 0.4 5

6

FIGURE 8-4

1

4 5 2 3 Distance Between Phase Centers, l

6

Paraboloidal reﬂector phase error loss due to feed astigmatism.

390

REFLECTOR ANTENNAS

of phase error losses due to feed astigmatism. Astigmatism loss is not as severe as axial defocusing because in two planes the feed phase center is at the reﬂector focus. As is true of axial defocusing loss, deep dishes are affected more than shallow reﬂectors.

8-6 FEED SCANNING Moving the phase center of the feed off axis laterally scans the reﬂector beam to a limited extent without severe pattern problems. Figure 8-5 shows the k-space pattern effects of feed scanning. The sidelobes show the effects of coma (cubic phase errors) where the sidelobes on the boresight side grow and the sidelobes on the other side decrease. We call these coma lobes, although no new lobes are generated. In fact, we see one lobe disappearing as a vestigial lobe with increased scan (Figure 8-5). Suppose that the feed is offset from the axis by a distance d. We measure the offset angle ψS from the axis to a line from the feed to the reﬂector vertex: d = f tan ψS . We ignore the slight amplitude distribution change due to the small lateral offset. Referred to the focus, the movement produces a phase factor in the feed pattern: −kd sin ψ cos φc when the feed is moved along the negative x-axis. Equation (8-7) predicts only the boresight phase error loss. Like Eq. (4-3), we must calculate the phase error efﬁciency at any angle to determine the loss at the pattern peak:

PEL(θ, φ) =

2π 0

ψ0 0

2 E(ψ, φc ) tan(ψ/2)ej k2f tan(ψ/2) sin θ cos(φ−φc ) dψ dφc

2

(8-15)

|E(ψ, φc )| tan(ψ/2) dψ dφc When we include the offset along φ = 0, the phase factor becomes ψ exp j kf cos φc 2 tan sin θ − tan ψS sin ψ 2

0 2pa tan y = 8 s l

Pattern Level, dB

−10

ys Feed

−20

a

−30 −40 −50 −60

−20

−15

−10

−5

5 0 2pa sin q l

10

15

20

FIGURE 8-5 Feed-scanned paraboloidal reﬂector f/D = 0.5 and feed beamwidth = 60◦ .

FEED SCANNING

391

For large reﬂectors we make the approximations ψS ≈ tan ψS and θ ≈ sin θ . The pattern scale and the offset phase factor become kaθ and kaψS . A ﬂat plate would reﬂect the ray at an equal angle on the other side of the axis for an offset feed, but a curved reﬂector modiﬁes that result slightly. The offset factor in Figure 8-5 is 8, and the beam peak is at 7. We call the ratio of the beam maximum to offset angle the beam deviation factor (BDF) [12]: BDF =

θm 7 = ψS 8

θm = BDF · ψS

The BDF varies from less than 1 for a concave reﬂector to greater than 1 for a convex reﬂector. BDF equals 1 for a ﬂat reﬂector. Table 8-1 lists the BDF values for various f/D and Scale 8-6 gives the relationship. The BDF approaches 1 as f/D approaches inﬁnity (ﬂat plate). The approximate expression for BDF is BDF =

(4f/D)2 + 0.36 (4f/D)2 + 1

(8-16)

Feed scanning increases the phase error loss. When normalized to beamwidths of scan, a single loss curve can be drawn for each f/D (Figure 8-6). Scanning also raises the sidelobes. Table 8-2 gives the approximate level of the peak coma lobe for a given scan loss. It is almost independent of f/D.

TABLE 8-1 Feed-Scanned Paraboloidal Reﬂector Beam Deviation Factor f/D

BDF

f/D

BDF

0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75

0.724 0.778 0.818 0.850 0.874 0.893 0.908 0.921 0.930 0.938

0.80 0.85 0.90 1.00 1.10 1.20 1.40 1.60 1.80 2.00

0.945 0.951 0.957 0.965 0.970 0.975 0.981 0.986 0.989 0.991

Beam Deviation Factor

Parabola f/D

SCALE 8-6 Feed-scanned reﬂector beam deviation factor given f/D.

392

REFLECTOR ANTENNAS 0

Scanning Loss, dB

1.5 1.2 1.0 1

2 F/D 0.3

0.4

0.5

0.6

0.7

3 2

4

6

8 10 12 14 Beamwidths of Scan

16

20

FIGURE 8-6 Feed-scanning loss of a paraboloidal reﬂector.

TABLE 8-2 Sidelobe Level of a Feed-Scanned Paraboloidal Reﬂector Scanning Loss (dB)

Sidelobe Level (dB)

Scanning Loss (dB)

Sidelobe Level (dB)

0.50 0.75 1.00 1.25 1.50

14.1 12.9 11.9 11.2 10.6

1.75 2.0 2.5 3.0

10.1 9.7 9.0 8.5

Example A reﬂector with a 50λ diameter is feed-scanned to 6◦ . Compute the offset distance and scanning loss when f/D = 0.6. Use the approximation HPBW = 70◦ λ/D = 1.4◦ . The reﬂector is scanned 6/1.4 = 4.3 beamwidths: scanning loss (Figure 8-6) = 0.4 dB sidelobe level (Table 8-2) = 14.6 dB The angle between the axis and the feed point to vertex must be greater than the scan angle, since the reﬂector is concave: ψS =

θS 6◦ ◦ = = 6.61 BDF 0.908

(Table 8-1)

The offset distance is f tan 6.61◦ = 0.6(50λ) tan 6.61◦ = 3.48λ.

RANDOM PHASE ERRORS

393

The scalar analysis of this section gives only approximate results. Large feed scanning produces higher-order aberrations other than coma [13–15]. The optimum gain point moves off the focal plane but fails to follow the curve predicted from optics for reﬂectors extremely large in wavelengths [14]. The reﬂector f /D and illumination taper determine the maximum gain contour for feed-scanning a reﬂector. Vector analysis improves the match between calculated and measured results [15].

8-7 RANDOM PHASE ERRORS Reﬂector anomalies reduce the gain predicted from the feed analysis. We must specify reasonable manufacturing tolerances for the frequency of operation. It would appear that gain can be increased without bound by increasing the reﬂector diameter, but the tolerance problems of large reﬂectors limit the maximum gain. We consider only surface anomalies so small that on average the reﬂector retains its basic shape. The surface imperfections change the optical path length from the feed to the reﬂector aperture plane by δ(r, φ), which gives us 2π a 2 j δ(r,φ) E(r, φ)e r dr dφ 0 0 PEL = 2 2π a |E(r, φ)|r dr dφ 0

(8-17)

0

Cheng [16] bounds the phase error loss by using a limit on the integrals. Given a peak phase error of m (radians), the change in gain is bounded: 2 G m2 ≥ 1− G0 2

(8-18)

This gain loss estimate is too conservative, but it is useful as an upper bound. Ruze [17] improved the random surface error loss estimate by using a Gaussian distributed error correlated over regions. Dents or segments making up the reﬂector are correlated with the errors over a nearby region. The error at a point depends on the location of nearby points in the correlation region. The phase error efﬁciency becomes an inﬁnite series: 2

PEL = exp(−δ ) +

1 η

2C D

2

2

exp(−δ )

2 ∞ (δ )n n · n! n=1

(8-19)

where C is the correlation distance, D the diameter, and η the aperture efﬁciency 2 (ATL). δ is the mean-square phase deviation, given by 2

δ =

2π 0

a

0 2π

0

|E(r, φ)|δ 2 (r, φ)r dr dφ a |E(r, φ)|r dr dφ 0

(8-20)

394

REFLECTOR ANTENNAS

If we include the correlation distance, PEL decreases. The inﬁnite series [Eq. (8-19)] converges rapidly. When the correlation distance is small compared with the diameter, the phase error efﬁciency becomes −4πε0 2 2 PEL = exp = exp(−δ ) (8-21) λ where ε0 is the effective reﬂector tolerance. We use 4π instead of 2π because the wave travels to and from the reﬂector and the phase distance is twice the reﬂector tolerance. From Eq. (8-20) we derive the effective RMS tolerance: 2π a |E(r, φ)|ε2 (r, φ)r dr dφ 0 0 2 εo = (8-22) 2π a |E(r, φ)|r dr dφ 0

0

Ruze gives the distance ε in terms of the z-axis deviation z and the surface normal ε=

z 1 + (r/2f )2

n ε=

1 + (r/2f )2

We evaluate the constants in Eq. (8-21) and convert to a decibel ratio: ε 2 0 PEL(dB) = −685.8 λ

(8-23)

(8-24)

Example Compute the required reﬂector tolerance at 30 GHz to limit the RMS surface tolerance phase error loss to 1 dB. Using Eq. (8-24), we get ε0 1 = = 0.038 λ 685.8 At 30 GHz, λ = 1 cm and ε0 = 0.38 mm. We can also use Eq. (8-18), which gives the upper bound on surface error loss:

4πε0 G m= = 0.466 at 1 dB = 2 1− λ G0 ε0 = 0.037λ or ε0 = 0.37 mm at 30 GHz. Both methods give about the same answer in this case. Zarghamee [18] extended tolerance theory to include the effects of the surface error distribution. Some antennas have better support and construction in some areas and are more accurate in those areas. This improves the reﬂector performance. Zarghamee deﬁned a second variation of surface deviations by 2π a |E(r, φ)|[ε2 (r, φ) − ε02 ]r dr dφ 0 0 4 η0 = 2π a |E(r, φ)|r dr dφ 0

0

RANDOM PHASE ERRORS

395

The phase error efﬁciency becomes

−4πε0 PEL = exp λ

2 exp

πη 4 0

λ

The correlation of random errors increases the probable sidelobe level. The sidelobe level increases with the size of the correlation interval and decreases for larger aperture diameters. Increasing the amplitude taper of the distribution makes the aperture pattern more susceptible to random-error sidelobes, since increasing the taper is somewhat equivalent to decreasing the aperture diameter. Blockage and feed diffraction also limit the achievable sidelobe level in a reﬂector. A simple feed cannot carefully control the aperture distribution necessary for low sidelobes. Hansen [19, p. 74] discusses sidelobe limitations caused by random phase error in some detail. Paraboloidal reﬂectors can be made in an umbrella shape where the ribs are parabolic and wire mesh is stretched between them [20]. The gore shape causes phase error loss and their periodicity produces extra sidelobes. Given the number of gores NG and the focal length of the ribs fr , the surface is given by f (ψ) = fr

cos2 (π/NG ) cos2 ψ

where ψ is measured from the centerline between the ribs. We calculate the average focal length by integrating across the gore half-angle π/NG and dividing by π/NG : fav = fr

sin(2π/NG ) 2π/NG

(8-25)

We use Eq. (8-25) to calculate the rib focal length given the average focal length of the reﬂector. The peak sidelobe due to the periodic gores occurs at an angle θp found from the number of gores and the diameter D: λ (8-26) θp = sin−1 1.2NG πD Given the average f /D of the reﬂector, we determine the peak-to-peak phase deviation across the gore by the approximate equation =

800 − 500(f/D − 0.4) D λ NG2

(8-27)

Scale 8-7 lists the phase error loss for a feed edge taper of 10 dB. Increasing the feed taper decreases the phase error loss due to gore construction. When we use a 20-dB feed taper, the values given by Scale 8-7 reduce by 0.16 dB for 0.5 dB of loss, 0.31 dB for 1 dB, and 0.45 dB for 1.5 dB. The gain losses due to underillumination by the 20-dB edge taper feed exceed these values. Example Given a reﬂector with D/λ = 35 with a limit of 0.5 dB loss due to gore construction for f/D = 0.34, we discover that the allowable peak-to-peak phase error from Scale 8-7 is 124◦ . Using Eq. (8-27), we solve for the number of ribs: NG2 =

830 35 = 234 124

or

NG = 16

396

REFLECTOR ANTENNAS Phase Error Loss, dB

Peak-to-Peak Phase Error due to Gores (degrees)

SCALE 8-7 Phase error loss due to gore construction of a paraboloidal reﬂector.

We use Eq. (8-26) to compute the angle of the peak gore sidelobe, θp = 10.0◦ . Equation (8-27) shows that the phase deviation is proportional to frequency. If the frequency increases by 1.5 times, then by Eq. (8-27), 124◦ increases to 186◦ and we read 1.1 dB of loss from Scale 8-7 while the peak gore sidelobe becomes θp = 6.7◦ . 8-8 FOCAL PLANE FIELDS We improve the efﬁciency and pattern response of a reﬂector if we match the feed ﬁelds to the focal plane ﬁelds. GO assumes a point focus, but an actual focus is extended. We determine the reﬂector and feed efﬁciency from the ﬁeld match over the focal plane. When the reﬂector f /D value is large, we use the diffraction pattern of a circular aperture, the Airy function: J1 (krψ0 ) E= (8-28) krψ0 where ψ0 is the half subtended angle of the reﬂector (radians), r the radial coordinate, k the propagation constant, and J1 the Bessel function. In a more exact method the currents induced on the reﬂector (2n × H) and the magnetic vector potential are used to calculate the focal plane ﬁelds. As f /D decreases, the currents on the reﬂector interact and modify their distribution, but it is a secondary effect [21]. Iterative physical optics analysis (section 2-4) can ﬁnd these current modiﬁcations. We calculate the reﬂector efﬁciency from the ﬁeld match of the focal plane ﬁelds (E1 , H1 ) and the ﬁelds of the feed (E2 , H2 ) using Robieux’s theorem [4]:

η=

2 (E1 × H2 − E2 × H1 ) · dS S

4P1 P2

(8-29)

where P1 and P2 are the input powers to produce the ﬁelds and η is the efﬁciency. Equation (8-29) is the magnitude squared of Eq. (2-35), the reactance equation equivalence applied to Eq. (1-55) for the coupling between two antennas S21 . The ﬁnite size of the feed causes spillover. The extent of amplitude and phase mismatch between the two ﬁelds determines the efﬁciency. By illuminating the reﬂector with a crosspolarized wave, we compute the cross-polarization radiation level through its ﬁeld match [Eq. (8-29)]. We maximize efﬁciency [Eq. (8-29)] by conjugate-matching the focal plane ﬁelds with the feed ﬁelds. Corrugated horns can be designed by expanding the focal plane

FEED MISMATCH DUE TO THE REFLECTOR

397

ﬁelds in axial hybrid modes of the horn and mode matching [22,23]. Wood [4] expands the reﬂector and feed ﬁelds in spherical harmonics and matches them at a boundary. Both sets of ﬁelds can be approximated very well by just a few terms, and this method can handle dual-reﬂector and offset reﬂector systems as well as axisymmetric prime focus reﬂectors. We can feed the reﬂector with an array to match the focal plane ﬁelds [24–26]. The array samples the focal plane ﬁeld and conjugate-matches it so that the powers sum in phase. The array can form multiple beams and also correct reﬂector aberrations [24]. By using the multiple feeds of the array, coma can be reduced for scanned beams and efﬁciency improved. However, quantization of the array element locations and excitations, amplitude, and phase reduces efﬁciency and raises the sidelobe level [27]. We apply Eq. (1-55) for the coupling between two antennas to determine the feeding coefﬁcients of an array feed for a dish. Assume an incident ﬁeld distribution on the reﬂector that includes the incident wave direction and the desired aperture distribution for the reﬂector. Using physical optics, we calculate the currents induced on the reﬂector surface. If the reﬂector has signiﬁcant curvature so that the patches face each other, iterative PO can be used to account for their interaction. We calculate the ﬁelds radiated by each feed on the reﬂector surface and apply Eq. (1-55) to calculate coupling. This method applies the feed pattern to the calculation instead of the point matching used in a focal plane solution. Similar to scanning of an array, we use conjugative matching for the feed array elements to produce the beam desired. This method can determine array feed element amplitude and phase for any composite reﬂector aperture distribution that includes aperture distribution to control sidelobes or include multiple beams. The method reduces coma to the minimum possible with a given array. Analysis ﬁnds the array distribution desired, but we do not achieve this distribution merely by designing the feed network to produce these amplitudes and phases because the feed elements have signiﬁcant mutual coupling. We need to include the effect of the paraboloidal reﬂector when computing mutual coupling because the ﬁeld radiated by one feed induces currents on the reﬂector that couple to other feed elements. Below we show that the effect of the reﬂector diminishes as the reﬂector diameter increases. If the mutual coupling is signiﬁcant whether direct or due to the reﬂector, we need to apply the corrections given in Section 3-11 to adjust the feeding coefﬁcients of the array. 8-9 FEED MISMATCH DUE TO THE REFLECTOR The feed receives some of its transmitted power because it reﬂects from the parabola and returns as a mismatch at the feed terminals. We calculate the reﬂected ﬁeld at the feed by using surface currents and the magnetic vector potential. The only signiﬁcant contribution comes from areas near where the normal of the reﬂector points at the feed. Around every other point, the phase of the reﬂection varies rapidly and cancels and we need to consider only points of stationary phase. We calculate the reﬂection from each point of stationary phase from [2] Gf (ρ0 ) ρ1 ρ2

= −j e−j 2kρ0 (8-30) 4kρ0 (ρ1 + ρ0 )(ρ2 + ρ0 )

398

REFLECTOR ANTENNAS

where is the reﬂection coefﬁcient, ρ0 the distance to the stationary phase point, Gf (ρ0 ) the feed gain in the direction of ρ0 , and ρ1 and ρ2 the radiuses of curvature of the reﬂector at ρ0 . The vertex is the only point of stationary phase on a paraboloidal reﬂector: ρ1 = ρ2 = −2f and ρ0 = f . Equation (8-30) reduces to

= −j

Gf (0) −j 2kf e 2kf

(8-31)

Example Suppose that we have a reﬂector with f/D = 0.40. Compute reﬂector mismatch for a source with its 10-dB beamwidth equal to the reﬂector subtended angle. Half subtended angle [Eq. (8-2)] ψ0 = 2 tan(1/1.6) = 64◦ . By using the feed approximation cos2N (θ /2), we have N=

log 0.1 = 6.98 2 log cos(64◦ /2)

The feed gain at the boresight is N + 1 [Eq. (1-20c)]: [Eq. (8-31)]

| | =

λ 8λ = 1.59 4πf D

Increasing the reﬂector diameter in wavelengths decreases the reaction of the reﬂector on the feed. For example, given a 3-m reﬂector at 4 GHz, we calculate reﬂector reﬂection coefﬁcient as 0.04, or VSWR = 1.08. We can express the reﬂector reﬂection of a paraboloidal reﬂector as | | = V

λ D

(8-32)

and calculate Scale 8-8 of V versus f /D for feeds with 10-dB beamwidths equal to the reﬂector subtended angle. Higher reﬂector f /D values produce larger feed reﬂections, since the feed gain increases faster than the reduced area of the reﬂector seen from the feed. Narrowband corrections to these reﬂections can be designed by using a vertex plate (Silver [2]) or by designing sets of concentric ring ridges in the reﬂector (Wood [4]). The rings can match the feed at more than one frequency. By any of these methods, the free-space mismatch of the feed could be corrected for, but, of course, the feed itself can be mismatched to compensate for the reﬂector reaction.

Feed Reflection Coefficient Factor, V

Parabola f/D

SCALE 8-8 Feed reﬂection scale factor V given f/D.

OFFSET-FED REFLECTOR

399

Front/Back Increase K Factor, dB

Parabola f/D

SCALE 8-9 Paraboloidal reﬂector front-to-back ratio increase K given f/D.

8-10 FRONT-TO-BACK RATIO Figure 2-9 illustrates the pattern response of a paraboloidal reﬂector and shows that the pattern behind the reﬂector peaks along the axis. The diffractions from all points along the rim add in-phase along the axis and produce a pattern peak. We can reduce this rim diffraction by using a rolled, serrated, or castellated edge to reduce diffraction. An absorber-lined cylindrical shroud extending out to enclose the feed will greatly reduce back radiation, including spillover, and allows the close spacing of terrestrial microwave antennas with reduced crosstalk. For a normal truncated circular reﬂector rim, the following equation estimates the front-to-back ratio given the reﬂector gain G, the feed taper T , and feed gain Gf [28]: F/B = G + T + K − Gf

dB

(8-33)

The constant K, given by Scale 8-9, is related to f /D: K = 10 log 1 +

1 (4f/D)2

(8-34)

Example Estimate F/B for a reﬂector with f/D = 0.34 and 40 dB of gain. We read the feed subtended angle from Scale 8-1 to be 143◦ . A 10-dB edge taper feed has a gain of about 8.1, found from Scale 1-2. Using Eq. (8-33), we estimate F/B = 40 + 10 + 1.9 − 8.1 = 43.8 dB.

8-11 OFFSET-FED REFLECTOR Moving the feed out of the aperture eliminates some of the problems with axisymmetrical reﬂectors. Blockage losses and diffraction-caused sidelobes and cross-polarization disappear. We can increase the size of the feed structure and include more if not all of the receiver with the feed. For example, the reﬂector may be deployed from a satellite, with the feed mounted on the main satellite body. Figure 8-7 shows the offset-fed reﬂector geometry. We form the reﬂector out of a piece of a larger paraboloid. Every piece of the paraboloidal reﬂector converts spherical waves from the focus into a plane wave moving parallel with its axis. We point the feed toward the center of the reﬂector to reduce the spillover, but we still locate the feed phase center at the focus of the reﬂector. The aperture plane projects to a circle, although the rim shape is an ellipse. ψ0 is the angle from the axis of the parabola to

400

REFLECTOR ANTENNAS

ru

yc

D

ye

yf

D′

rL

ye

yo

H

yu

ƒ

FIGURE 8-7 Parameters of an offset-fed parabolic reﬂector.

the center of the cone of the reﬂector, and the reﬂector subtends an angle 2ψe about this centerline. Given the aperture plane diameter D and the height H of the center, we ﬁnd the lower rim offset D = H − D/2. From these parameters we determine the angle of the center of the rim cone from the z-axis: ψ0 = tan−1

16fH 2f (D + 2D ) −1 = tan 16f 2 + D 2 − 4H 2 4f 2 − D (D + D )

(8-35)

The half cone angle deﬁnes the rim: ψe = tan−1

8fD 2f D = tan−1 16f 2 + 4H 2 − D 2 4f 2 + D (D + D )

(8-36)

We direct the feed an angle ψf from the z-axis to the center of the projected diameter different from the angle ψ0 of the rim cone axis: ψf = 2 tan−1

H 2D + D = 2 tan−1 2f 4f

(8-37)

The rim lies in a plane at an angle ψc with respect to the z-axis: ψc = tan−1

2f 4f = tan−1 H 2D + D

(8-38)

401

OFFSET-FED REFLECTOR

The rim is an ellipse in this plane with major and minor axes given by ae =

D 2 sin ψc

and be =

D 2

(8-39)

The offset angle modiﬁes the f /D of the reﬂector: cos ψe + cos ψ0 f = D 4 sin ψe

(8-40)

We calculate the rim offset from the cone angles: D = 2f tan

ψ0 − ψe 2

(8-41)

Manufacturing an offset reﬂector requires speciﬁcation of the reﬂector when laid on its rim in the x –y plane so that the mold can be machined. We center the major axis of the reﬂector elliptical rim L = 2ae along the x-axis and the minor axis D along the y-axis. In this position the reﬂector depth d(x, y) is found from the expression [29] √ 2f L3 xD 2 L2 − D 2 D 2 (L2 − D 2 ) D 2 2 d(x, y) = 1+ + −y D(L2 − D 2 ) f L3 4f 2 L4 4 √ xD 2 L2 − D 2 −1− 2f L3

(8-42)

The deepest point of the reﬂector dmax occurs along the x-axis at xb : √ D 2 L2 − D 2 xb = − 16f L

where dmax =

D3 16f L

(8-43)

After measuring D, L, and dmax , we determine the offset focal length from the equation f =

D3 16Ldmax

(8-44)

We calculate the center height of the offset from H = 2f

L2 −1 D2

(8-45)

We calculate the reﬂector half cone angle ψe and the cone axis angle from the z-axis and ψ0 from the focal length f , ellipse major diameter L, and minor diameter D:

ψe ψ0

= tan−1

D D L L ∓ tan−1 −1+ −1− D2 4f D2 4f 2

2

(8-46)

402

REFLECTOR ANTENNAS

To align the reﬂector, we use the angle of the reﬂector rim major axis ψc = sin−1 (D/L) with respect to the z-axis and the radial distances from the lower and upper edges of the reﬂector in the offset plane, since the center offset H is not a distinguishable point:

ρU ρL

=

2 L D2 fL2 ± D + − 1 D2 16f D2

(8-47)

We analyze the offset reﬂector with the same tools as those used with the axisymmetric reﬂector: aperture ﬁeld, physical optics, and GTD. The asymmetry of the reﬂector to feed geometry introduces anomalies. Huygens sources no longer eliminate crosspolarization, because the source must be tilted. Symmetry prevents cross-polarization in the plane containing the x-axis (Figure 8-8), but cross-polarization for linear polarization increases in the plane containing the y-axis (symmetry plane) as f /D decreases

Diameter

2ye y0

x

Focal le

ngth

y

Focus

(a)

x

Diameter y 2ye y0 Focus Focal length (b)

FIGURE 8-8 Offset-fed paraboloidal reﬂector geometry: (a) perspective; (b) orthographic representation.

OFFSET-FED REFLECTOR

403

Amplitude, dB

E-Plane

Symmetrical Plane Cross-Polarization H-Plane

Pattern Angle (degrees)

FIGURE 8-9 Pattern of an offset-fed reﬂector with linearly polarized feed.

LHC

Amplitude, dB

RHC

Pattern Angle (degrees)

FIGURE 8-10 Pattern of an offset-fed reﬂector with circularly polarized feed.

(Figure 8-9). The Condon lobes move off the diagonal planes and into the plane containing the y-axis. The asymmetry along the x-axis tapers the amplitude distribution from a symmetrical feed, since the spherical wave travels farther to the outer edge of the reﬂector than to the lower edge. The offset-fed reﬂector geometry squints circularly polarized pattern peaks in the symmetrical (y-axis) plane without generating cross-polarization (Figure 8-10). An approximate formula for the squint is [9] ψs = sin−1

λ sin ψ0 4πf

(8-48)

404

REFLECTOR ANTENNAS

where ψs is the squint angle. Opposite senses of circular polarization squint in opposite directions and cause a problem with dual circularly polarized feed systems. In all cases increasing the f /D or the effective f /D through a subreﬂector reduces these problems. The subreﬂector should be kept out of the aperture of the main reﬂector. We can feed-scan the offset-fed reﬂector by moving the feed laterally along a line that lies perpendicular to the boresight of the feed (the line deﬁned by ψ0 ). We must modify the beam deviation factor (BDF): BDFoffset

fed

= BDFcenter

fed

(f/D)offset (f/D)center fed

(8-49)

Example Given an offset-fed reﬂector with ψ0 = 45◦ and ψe = 40◦ , compute the beam deviation factor. From Eq. (8-49), f cos 40◦ + 1 = = 0.687 D center fed 4 sin 40◦ cos 40◦ + cos 45◦ f = = 0.573 D offset fed 4 sin 40◦ From Table 8-1 we interpolate BDFcenter fed = 0.928, and we substitute the values into Eq. (8-49) to calculate BDFoffset fed = 0.774. We must laterally offset the feed farther than with a center-fed reﬂector to achieve the same feed scanning. Periscope Conﬁguration The periscope consists of an offset paraboloidal reﬂector with ψ0 = 90◦ with a long focal length fed by a paraboloidal reﬂector located at the focus. This eliminates the need to run a transmission line up a tower. Periscope antennas can be made using a ﬂat-plate reﬂector, but the long focal length means that the parabolic splash plate antenna has only a small deviation from ﬂat. The ﬂat plate is limited to a gain of only 6 dB more than the feed reﬂector for optimum conditions with a large plate. The gain of the offset paraboloidal reﬂector is determined by the diameter of the splash plate, not the feed reﬂector. Because the splash plate is in the near ﬁeld of the feeding reﬂector, gain is reduced by phase error, whereas spillover and amplitude taper losses also contribute to gain loss. Design starts with determining the splash reﬂector center height H required to clear obstacles along the transmission path. We calculate the splash reﬂector aperture diameter from the required gain and beamwidth. The periscope conﬁguration contributes to gain loss, but with proper selection of the feed paraboloidal reﬂector these losses are minor and can be compensated for by using a larger splash reﬂector. Having the splash reﬂector directly overhead corresponds to a parent reﬂector design with f/Dp = 0.25 and f = H /2. An analysis using a radial parabolic aperture distribution in the feed reﬂector determined that the optimum feed reﬂector diameter is found from the ratio of height to projected splash reﬂector aperture diameter Ds [30]: Df =

2λH αλH = Ds Ds

or

α=

Ds Df Ds Df F = λH Hc

(8-50)

Whereas α = 2 is the optimum dimensions at a particular frequency, we account for shift from the optimum with this factor. The parameter α is the frequency response

REFLECTIONS FROM CONIC SECTIONS

405

factor for frequency F and speed of light c. The illumination efﬁciency is the product of the feed reﬂector illumination efﬁciency and the periscope efﬁciency factor ηp : 4 1 − (1 − K)J0 (m) − K(2/m)J1 (m) ηp = (8-51) m2 (1 − K/2)2 J0 and J1 are Bessel functions, m = απ/2, and K = 1 − 10−[ET(dB)/20] for the feed reﬂector edge illumination taper ET(dB). Table 8-3 lists the added illumination loss of a periscope, given geometry using α in Eq. (8-50) for a 12-dB edge taper in a feeding reﬂector. Example A periscope antenna system placed a 3-m projected aperture splash reﬂector 30 m above the feed reﬂector to operate at 12 GHz (λ = 0.05 m). Using Eq. (8-50), we calculate the feed reﬂector diameter to be 0.5 m for α = 2. If we assume that the feed reﬂector has an efﬁciency of 60% (−2.22 dB), the efﬁciency of the splash reﬂector will be −2.6 dB: gain (dB) = 20 log

3π πDs − 2.6 = 20 log − 2.6 = 42.9 λ 0.05

The focal length f of the splash reﬂector is H /2 = 15 m. Since the angle of the splash reﬂector rim is 45◦ , L = Ds / sin(45◦ ) = 4.24 m. We determine the maximum depth of the reﬂector by using Eq. (8-43) to be 2.65 cm located 2.65 cm off center. The splash reﬂector increased the gain relative to the feed reﬂector by 7.4 dB. 8-12 REFLECTIONS FROM CONIC SECTIONS We use reﬂectors made from conic sections other than the parabola as subreﬂectors. The ellipse and hyperbola rotated about their axes to form solid ﬁgures that reﬂect incident spherical waves into spherical waves with different caustics (focal points). Reﬂectors formed by moving the ﬁgure along a line change the caustics of cylindrical waves. We consider only spherical waves, but we need only convert to cylindrical waves for cylindrical reﬂectors. All conic-section reﬂectors convert spherical waves from one focus into spherical waves directed toward the other focus. The ellipse has its two focuses located within the TABLE 8-3 Added Illumination Loss of a Periscope, Given Geometry Using α in Eq. (8-50) for 12-dB Edge Taper in a Feeding Reﬂector (dB) α 1.0 1.2 1.4 1.6 1.8 2.0

ηp

α

ηp

3.17 2.06 1.28 0.77 0.48 0.38

2.2 2.4 2.6 2.8 3.0 3.2 3.4

0.45 0.68 1.04 1.52 2.11 2.79 3.53

406

REFLECTOR ANTENNAS

ﬁgure. As we let one focus approach inﬁnity, the ellipse transforms into a parabola. If the focus is pushed through inﬁnity to the negative axis, the ﬁgure becomes a hyperbola located between the two focuses. Figure 8-11 shows the ray tracing for axisymmetrical conic section reﬂectors. A spherical source at one focus is reﬂected to the second focus by the reﬂector, although it is virtual (not actually reached) in some cases. We describe all conic sections with the same polar equation: ρ=

eP 1 − e cos θ

(8-52)

where P is the distance between the origin, the focus, to a line called the directrix (Figure 8-12). The eccentricity e is the ratio of the distance from the origin to a point on the curve to the distance from the same point to the directrix: r1 = er2 . In an ellipse, e < 1; in a parabola, e = 1; and in a hyperbola, e > 1. The distance between the focuses is 2P e2 (8-53) 2c = 1 − e2

2c

F1

F2

2b

Ellipse 2a

Parabola Focus

Hyperbola

Focus

Hyperbola

Focus Focus

Focus

FIGURE 8-11 Reﬂections from conic-section reﬂectors.

REFLECTIONS FROM CONIC SECTIONS

407

R2 Directrix

R1

q Focus

P

R= e=

eP 1 − e cos q R1 R2

FIGURE 8-12 Conic-section geometry.

A hyperbola with its axis containing the two focuses along the z-axis, located at ±c, intersects the z-axis at ±a and satisﬁes the equation r2 z2 − =1 a2 b2

where b2 = c2 − a 2

and e =

c >1 a

(8-54)

When we take the portion of the hyperbola along the +z-axis that intersects the axis at +a, we deﬁne the angles from the two focuses from the line between them because we place the feed at the left focus and locate a parabola focus at the right hyperbola focus. The left angle θ is the feed angle, and the right angle ψ is the parabola angle in a dual-reﬂector antenna. Given a point on the hyperbola, the distance from the left focus is ρ1 and the distance from the right focus is ρ2 : ρ1 =

b2 a(e2 − 1) = e cos θ − 1 e cos θ − 1

and ρ2 =

b2 a(e2 − 1) = e cos ψ + 1 e cos ψ + 1

(8-55)

We determine the radial position off the axis from either polar equation: r = ρ1 sin θ = ρ2 sin ψ

(8-56)

The two angles are related by the eccentricity e: (e + 1) tan

θ ψ = (e − 1) tan 2 2

(8-57)

At a given point on the hyperbola, the angle of the normal u relative to the radial line ρ1 is half the sum of the two angles, u = (θ + ψ)/2. We need the radius of curvatures in the principal planes to apply Eq. (2-77) for reﬂection from a curved surface: R1 in

408

REFLECTOR ANTENNAS

the r –z plane and R2 in the φ –z plane: R1 =

b2 a cos2 u

and R2 =

b2 a cos u

(8-58)

When we offset-feed a hyperboloid, a cone determines the rim and it lies in a planar ellipse. Similar to the offset paraboloid, we center the cone at an angle θ0 from the axis and deﬁne the cone by the half feed edge angle θe . We compute the distances from the focus to the upper and lower rim along the major axis of the rim ellipse: ρL =

b2 e cos(θ0 − θe ) − 1

and ρU =

b2 e cos(θ0 + θe ) − 1

We determine the major axis diameter (2ae ) of the elliptical rim from the triangle with sides ρL and ρU and angle 2θe between them: 2ae = ρL2 + ρU2 − 2ρL ρU cos 2θe (8-59) The minor axis diameter (2be ) is given by the equation

2be = (2ae )2 − (ρL − ρU )2

(8-60)

An ellipsoid can also be used as a subreﬂector in a dual-reﬂector antenna. Its equations are similar to the hyperboloid: r2 z2 + = 1 and b2 = a 2 − c2 a2 b2

with

e=

c 0

where Jn is the Bessel function and Hn(2) is the outward-traveling Hankel function, with Jn the derivative of Jn , and so on. Table 8-8 lists these factors for both polarizations versus strut radius. Using only physical optics we can determine accurately the blockage effects of struts that are at least 3λ in diameter. PO excites currents only on the visible half of the struts. For smaller-diameter struts, currents creep to the far side and alter the results. No matter how thin the struts, currents will be excited on them and affect the pattern. The physical optics analysis of a dual reﬂector includes currents excited on the struts a number of times. Assume that the antenna is transmitting. The feed illuminates the subreﬂector and the struts. The current excited on the struts also radiates a ﬁeld that illuminates the subreﬂector. If the strut blocks the path between the feed and the subreﬂector, PO analysis uses strut current to calculate its blockage. At this point we use the currents TABLE 8-8

IFRE and IFRH for a Circular Strut

a sin θ0

Re(IFRE )

Im(IFRE )

0.005 0.010 0.020 0.050 0.10 0.20 0.50 1.00 2.00

−5.148 −3.645 −2.712 −1.982 −1.641 −1.414 −1.215 −1.145 −1.092

14.088 6.786 3.964 2.003 1.225 0.758 0.381 0.255 0.160

Re(IFRH )

Im(IFRH )

−0.0001 −0.0005 −0.004 −0.054 −0.292 −0.552 −0.781 −0.858 −0.914

−0.0198 −0.050 −0.103 −0.272 −0.448 −0.374 −0.258 −0.188 −0.126

FEED AND SUBREFLECTOR SUPPORT STRUT RADIATION

419

on the subreﬂector to compute additional currents excited on the struts. These add to the ﬁrst set of strut currents. Radiation from the subreﬂector, all currents on the struts, and stray feed illumination add to illuminate the main reﬂector. The radiation from the main reﬂector current excites additional current on the subreﬂector and a third set of currents on the struts. We apply the far-ﬁeld Green’s function on the sum of all currents to calculate the pattern. PO currents are modiﬁed on thin struts to account for creeping-wave currents. We multiply the PO strut currents by the induced current ratio (ICR) to obtain equivalent currents suitable for predictions. The factor ICR includes a strut current distribution and a complex value: Js = 2n × Hinc · ICR(a, θ0 , φ ) (8-76) ICR depends on the strut radius a, the incidence angle θ0 with respect to the strut axis, and the angle around the strut φ from the direction of the plane wave and the incident wave polarization. We ﬁnd the incident magnetic ﬁeld Hinc at the point where the plane wave touches the strut and use the current excited at this point to calculate the current in a ring around the strut. Remember that the primary effect of the strut is to block the radiation from the main reﬂector that approximates a plane wave in the near ﬁeld where the struts are located. We solve for ICR by considering two-dimensional scattering of the strut cross section by a plane wave. By applying moment methods to a two-dimensional scattering problem, we can solve for the current distribution on any strut cross section, but here we consider only circular struts that have a closed-form solution [40, pp. 209–219]. To simplify the problem, consider a strut lying along the z-axis. For actual analysis you will need to rotate the strut into place and rotate the incident wave into the strut coordinate system to use ICR to calculate the current distribution. In two-dimensional space the incident wave is either TM or TE with respect to the z-axis. The TM wave has its electric ﬁeld in the plane containing the strut axis. A TE wave has its magnetic ﬁeld in this plane. The TM case produces the following equation from the scattering of a plane wave: ICRE zˆ = ICRTM (a, θ0 , φ )ˆz =

∞ zˆ e−j ka cos φ j m εm cos mφ πka sin θ0 m=0 Hm(2) (ka sin θ0 )

(8-77)

The phase factor e−j ka cos φ shifts the reference plane from the strut center to the attachment point. Equation (8-77) expands the current in a cos mφ Fourier series around the strut. ICRE has a complex value because Hm(2) has a complex value. As the strut radius a increases, ICR approaches 1 at the location φ = 0. For practical purposes ICR is 1 for a/λ > 1.5. Equation (8-77) gives the current distribution on the circular strut relative to the current excited at the initial contact point of the incident plane wave. Table 8-9 lists ICRE evaluated when φ = 0. The constant term grows rapidly as a → 0, with its imaginary part growing faster than the real part because it approximates the vector potential of a ﬁlamentary current element with its −j factor between the current and the ﬁeld [Eq. (2-1)]. Small struts have nearly constant current around their periphery. Equation (8-77) requires more and more terms as the strut diameter increases, and ﬁnally, a simple PO solution produces the same results. Table 8-9 lists the ICR factors versus the strut radius.

420

REFLECTOR ANTENNAS

TABLE 8-9 ICRE and ICRH PO Current Multipliers When φ = 0 for a Circular Strut a sin θ0 Re(ICRE ) Im(ICRE ) Re(ICRH ) Im(ICRH ) 0.002 0.004 0.01 0.02 0.05 0.10 0.50 1.00 1.50

−7.932 −4.530 −2.228 −1.353 −0.751 −0.479 −0.140 −0.076 −0.052

3.660 2.735 2.018 1.687 1.376 1.200 1.030 1.010 1.005

0.500 0.500 0.499 0.500 0.546 0.748 0.948 0.982 0.991

0.003 0.013 0.033 0.071 0.195 0.275 0.109 0.068 0.049

A TE incident wave produces similar results for scattering from a circular cylinder but has co- and cross-polarization terms: ICRH = ICRTE (a, θ0 , φ )φˆ =

∞ ˆ −j ka cos φ j φe j m εm cos mφ πka sin θ0 m=0 Hm(2) (ka sin θ0 )

(8-78)

Equation (8-78) has the same form as Eq. (8-77) and is expanded in the even function cos mφ as Eq. (8-77) with coefﬁcients using the derivative of the Hankel function. When the incident wave approaches the strut at an angle θ0 other than 90◦ , the strut scatters cross-polarization for a TE incident wave: JCRH = JCRTE zˆ =

j cos θ0 zˆ e−j ka cos φ (ka sin θ0 )2

∞

mj m ej mφ

m=−∞

Hm(2) (ka sin θ0 )

(8-79)

JFRH is an odd function around the perimeter of the strut with a zeroth term of zero. We can expand Eq. (8-79) in terms of sin mφ :

−2j cos θ0 zˆ e−j ka cos φ (I1 sin φ + j 2I2 sin 2φ − 3I3 sin 3φ − · · ·) JCRH = (ka sin θ0 )2 1 Im = (2) (8-80) Hm (ka sin θ0 ) Consider a plane wave incident on a straight strut at an angle θ0 to its axis. As the wave sweeps across the strut it excites current whose phase velocity is c/cos θ0 with respect to the strut axis. This is the same situation as a waveguide with two waves traveling back and forth between the sidewalls that produces a central phase velocity greater than c (Section 5-24). A thin strut with its constant current distribution around the circumference radiates a cone-shaped pattern peaked at an angle determined by the current phase velocity. The current is a fast or leaky wave radiator that radiates in a cone at an angle θ0 from the axis, while the length of the strut in wavelengths determines the narrowness of the radiation beamwidth. As the diameter of the strut increases, the peripheral current distribution alters the radiation level around the cone, but the peak radiation occurs along the cone determined by the incident angle.

DISPLACED-AXIS DUAL REFLECTOR

421

We use incident plane waves to derive strut blockage and scattering and to modify the formulation for spherical wave incidence. First divide the struts into coin sections. For a wave incident from a given point, we trace a ray from the point to the strut axis through a given coin section. We determine the incident magnetic ﬁeld and calculate the surface current density at the point of intersection of this ray and the strut. The intersection point is φ = 0. We apply ICRE , ICRH , and JCRH to calculate the currents around the coin cross section. This near-ﬁeld case does not radiate a strut cone pattern because it is not a plane wave incident on the strut. As the strut diameter grows, this method leads directly to a PO formulation for strut scattering. 8-15 GAIN/NOISE TEMPERATURE OF A DUAL REFLECTOR Collins [41] has developed a procedure for calculating the noise temperature of Cassegrain antennas pointed near the horizon. First, the diffraction pattern of the feed and subreﬂector combination is calculated. Some of the diffraction is added to the main-reﬂector spillover. At low elevation angles the antenna points about one-half of the spillover on the ground. It is a major noise temperature contribution, 12 (1 − SPL)TG , where TG is the ground temperature and SPL is the spillover efﬁciency (ratio). The scattered portion of the blockage produces wide-angle sidelobes, half of which see the ground. The gain is reduced by the spillover loss, and a uniform distribution for the blockage is assumed (ATL = 1): 1 Sb (SPL)TG 2 Sa where Sb is the blocked area and Sa is the total potential aperture. The main beam points toward the sky and collects noise, SPLηb ηm TS , where ηb is the blockage efﬁciency, ηm the ratio of the power in the main beam and the ﬁrst few sidelobes (ηm ≈ 0.99), and Ts the sky temperature. We include a group of minor contributors: 1 (1 − SPL)Ts 2

1 Sb (SPL)Ts 2 Sa

1 SPLηb (1 − ηm )(TG + Ts ) 2

Equation (1-56) can be used when the temperature distribution is known, but the procedure of Collins gives good, although slightly conservative results. Refer to Section 1-15 to calculate the gain noise temperature of the receiving system. 8-16 DISPLACED-AXIS DUAL REFLECTOR A displaced-axis dual reﬂector uses a paraboloidal main reﬂector with a ring focus that transforms the vertex into a ring. GO rays reﬂected from the main paraboloid miss the subreﬂector and reduce the blockage loss to a nonexcitation area instead of scattered blockage. This reﬂector achieves high aperture efﬁciency by using a subreﬂector that directs the higher feed radiation at the boresight to the outer rim of the parabola, where the differential area is the largest. For the moment, consider Figure 8-14 of the Gregorian dual reﬂector in two dimensions. The parabola and ellipse retain their reﬂecting properties because we can extend them out of the page into cylindrical reﬂectors and use a linear array as a feed. Mentally, remove the lower half of the parabola and the upper half of the ellipse. Rays

422

REFLECTOR ANTENNAS

from the left ellipse focus (feed) reﬂect from the remaining half of the lower ellipse to the upper half of the parabola, which transforms them into plane waves. If we ﬁx the right ellipse focus at the focus of the parabola, we can rotate the ellipse axis about the right focus without changing the ray tracing from a feed at the left focus. We use a slightly different portion of the ellipse determined by the rays traced to the edges of the remaining half parabola. Place a horizontal axis at the lower edge of the ellipse and rotate the ellipse axis until the feed focus is on this axis. Rotate the two-dimensional ﬁgure about this horizontal axis to form a three-dimensional reﬂector, and it becomes a displaced-axis dual reﬂector. Both the focus and vertex of the main reﬂector have become rings. The subreﬂector has a matching ring focus at the same diameter as that of the main vertex and a point focus at the feed (Figure 8-16). Now rays from the upper portion of the subreﬂector reﬂect to the upper portion of the parabola. Rays from the center of the subreﬂector terminate on the outer edge of the main reﬂector, while outer subreﬂector edge rays reﬂect to the ring vertex of the main reﬂector [42]. The reﬂector geometry has been found in closed form [43]. Given the main reﬂector diameter D, focal length f , diameter of the subreﬂector Ds , and feed half-subtended angle θ0 , the distance along the reﬂector axis from the vertex to the feed Lm is Lm =

fD Ds cos θ0 + 1 − D − Ds 4 sin θ0

(8-81)

We tilt the axis of the ellipse φ to collapse the ring focus to a point at the feed: tan φ =

2 (cos θ0 + 1)/ sin θ0 − 4f/(D − Ds )

Ring Focus

f

2c Feed

q f

Subreflector

Ring Vertex Ellipse

Parabola

FIGURE 8-16

Displaced-axis reﬂector antenna.

(8-82)

DISPLACED-AXIS DUAL REFLECTOR

423

The parameters of the ellipse are given by the equations Ds c= 4 sin φ

Ds and a = 8

4f cos θ0 + 1 + sin θ0 D − Ds

(8-83)

The half-subtended angle of the main reﬂector ψ0 is found from the normal parabola with the subreﬂector removed and the ring focus collapsed to a point [Eq. (8-2)]: ψ0 = 2 tan−1

D − Ds 4f

We compute the distance between the feed and the subreﬂector along the axis Ls from the geometry [45]: Ds Ls = 2c cos φ + 2 tan ψ0 We determine aperture power distribution A(r ) by tracing rays from the feed to the aperture radius r of the main reﬂector and by equating power in differential areas: P (θ ) sin θ dθ = A(r ) dr

(8-84)

Figure 8-17 gives the aperture distribution for a displaced-axis reﬂector designed for a main reﬂector f/D = 0.27 and an effective feff /D = 1.2 from the feed (θ0 = 23.54◦ ) for various feed edge tapers. The plot shows that increasing the feed edge taper increases the aperture power at larger radiuses but reduces the center amplitude. The center 20% of the diameter of the aperture is not excited, but this corresponds to only 4% lost area, or −0.18 dB. When calculating the amplitude taper loss [Eq. (4-8)], we

Feed Edge Taper

Normalized Aperture Voltage

10-dB 12-dB 14-dB 16 dB 18-dB

Normalized Aperture Radius

FIGURE 8-17 Aperture distribution in a displaced-axis reﬂector given feed edge taper for a particular antenna. (From [43], Fig. 3, 1997 IEEE.)

424

REFLECTOR ANTENNAS

TABLE 8-10 Illumination Losses of a Displaced-Axis Dual Reﬂector, f /D = 0.27, feff /D = 1.2 for Ds = 0.2D and 0.1D Versus a Feed Edge Taper Ds /D (%) 20

10

Edge Taper (dB)

SPL (dB)

ATL (dB)

Total (dB)

Sidelobe

SPL (dB)

ATL (dB)

Total (dB)

Sidelobe

10 11 12 13 14 15 16 17 18 19 20

0.434 0.341 0.268 0.212 0.167 0.132 0.105 0.083 0.066 0.052 0.041

0.476 0.455 0.442 0.434 0.432 0.436 0.444 0.456 0.472 0.491 0.513

0.910 0.796 0.710 0.646 0.600 0.568 0.548 0.539 0.537 0.543 0.555

15.2 14.9 14.5 14.2 14.0 13.7 13.5 13.3 13.1 12.9 12.7

0.434 0.341 0.268 0.212 0.167 0.132 0.105 0.083 0.066 0.052 0.041

0.411 0.377 0.350 0.330 0.316 0.308 0.304 0.306 0.311 0.320 0.333

0.845 0.718 0.619 0.542 0.483 0.440 0.409 0.388 0.377 0.372 0.374

18.3 17.8 17.3 16.8 16.4 16.0 15.6 15.3 14.9 14.6 14.3

use the full radius a = D/2, which accounts for the lost aperture center area. Table 8-10 lists the illumination losses of this reﬂector. The antenna has 88.2% aperture efﬁciency, including blockage loss for Ds /D = 0.2 and 91.8% for Ds /D = 0.1. These numbers do not include diffraction loss due to the subreﬂector and main reﬂector size in wavelengths or strut blockage. Similar to a Cassegrain reﬂector, increasing the subreﬂector diameter beyond the GO design by 1 to 2λ decreases the diffraction loss. Figure 8-17 shows the complete taper of the aperture to zero voltage at the edge. We can increase the aperture efﬁciency slightly by designing the antenna with an effective main reﬂector diameter slightly larger than the real diameter and produce a ﬁnite aperture edge taper at the cost of increased spillover past the main reﬂector. Table 8-11 lists the illumination losses for designs the same as Table 8-10 except that the effective main reﬂector is 2% greater. Four versions of displaced-axis reﬂectors have been derived from Gregorian and Cassegrain antennas [44]. One other case, the double-offset Cassegrain, crosses the feed illumination so that boresight feed amplitude reﬂects to the outer rim of the main reﬂector. This antenna, similar to the case covered above, has a high aperture efﬁciency, whereas the other two cases have modest aperture efﬁciencies. Equations to specify all four antennas are available [45]. The normal displaced-axis dual reﬂector has less sensitivity to feed axial defocusing than does a normal Cassegrain or Gregorian antenna, but it is more sensitive to lateral offset of the feed [46].

8-17

OFFSET-FED DUAL REFLECTOR

When we offset-feed a dual reﬂector, we can eliminate subreﬂector central blockage of the Cassegrain or Gregorian reﬂectors. This design adds parameters to give more convenient packaging that ﬁts in the available space, such as on a spacecraft. More important, by rotating the subreﬂector axis relative to the main reﬂector axis, we can

425

OFFSET-FED DUAL REFLECTOR

TABLE 8-11 Illumination Losses of a Displaced-Axis Dual Reﬂector, f /D = 0.27, feff /D = 1.2 for Ds = 0.2D, and 0.1D Versus a Feed Edge Taper (Effective Main Diameter = 102% Actual) Ds /D (%) 20

10

Edge Taper (dB)

SPL (dB)

ATL (dB)

Total (dB)

Aperture Taper (dB)

SPL (dB)

ATL (dB)

Total (dB)

Aperture Taper (dB)

10 11 12 13 14 15 16 17 18 19 20

0.455 0.352 0.280 0.225 0.181 0.147 0.120 0.099 0.083 0.070 0.060

0.371 0.352 0.340 0.335 0.334 0.339 0.349 0.363 0.380 0.401 0.425

0.816 0.705 0.621 0.559 0.515 0.486 0.469 0.462 0.463 0.471 0.485

10.6 10.4 10.1 9.9 9.6 9.5 9.3 9.1 9.0 8.8 8.7

0.443 0.350 0.278 0.222 0.178 0.144 0.117 0.096 0.079 0.066 0.056

0.304 0.271 0.246 0.227 0.215 0.208 0.206 0.209 0.216 0.227 0.241

0.746 0.621 0.524 0.449 0.393 0.352 0.323 0.305 0.295 0.293 0.297

13.5 12.5 11.5 10.6 10.4 10.2 10.0 9.8 9.7 9.5 9.4

Main Reflector

Subreflector

yf

qe qe a b

2c

2ye yu

Focus

Focus

Feed

Feed ( a)

(b)

FIGURE 8-18 Dual offset-fed Cassegrain reﬂector, including Mizugutch feed axis tilt: (a) feed and subreﬂector geometry; (b) dual reﬂector.

greatly reduce cross-polarization or beam squint of dual circularly polarized feeds in the offset reﬂector. Figure 8-18 illustrates the geometry of an offset-fed Cassegrain reﬂector, and Figure 8-19 shows the offset-fed Gregorian geometry. Refer to Figure 8-7 for the parameters of the offset main reﬂector. We point the feed at the subreﬂector center to reduce spillover and to equalize the amplitude distribution in the aperture of the

426

REFLECTOR ANTENNAS

Main Reflector Focus b a

Feed

yu qe

yf

Subreflector

qe

Focus Feed

(a)

(b)

FIGURE 8-19 Dual offset-fed Gregorian reﬂector, including Mizugutch feed axis tilt: (a) feed and subreﬂector geometry; (b) dual reﬂector.

main reﬂector. Similar to a displaced-axis dual reﬂector, we tilt the axis between the focuses of the subreﬂector relative to the main reﬂector axis β (α in Ticra [48, App. B]). The small amount of tilt to the subreﬂector axis converts the equivalent parabola of the dual reﬂector to an axisymmetric geometry [47]. It takes ﬁve parameters to specify the antenna if the Mizugutch angle requirement is applied to the feed tilt angle α (ψ0 Ticra) relative to the subreﬂector axis determined from the magniﬁcation M given the subreﬂector eccentricity e [Eq. (8-64)]: M tan

β α = tan 2 2

or

M tan

α ψ0 = tan 2 2

† for M =

e+1 e−1

(8-85)

The equation for M in Eq. (8-85) is used for both Cassegrain and Gregorian reﬂectors. We start the design with the diameter of the main reﬂector D because it determines the gain and beamwidth. Ticra [48, App. B] uses main reﬂector focal length f , the half distance between focuses of subreﬂector c, subreﬂector eccentricity e, and axis tilt β. Granet [49,50] supplies equations to calculate the reﬂector dimensions for 17 different sets of ﬁve input parameters. These sets of equations allow the direct application of various mechanical constraints to the design or electrical constraints, such as subreﬂector size to limit diffraction loss. All of Granet’s sets apply the Mizugutch relationship, because this small change should be applied to all designs. By tracing rays through the reﬂectors, the center offset H is found: H = −2f

tan(β/2) − M tan(α/2) 1 + M tan(β/2) tan(α/2)

(8-86)

Given H , D, and f , we compute main reﬂector parameters from Eqs. (8-35) to (8-47). We compute the half subtended angle of feed θe by tracing the ray to the upper rim of

HORN REFLECTOR AND DRAGONIAN DUAL REFLECTOR

427

the main reﬂector with feed angles ψU and α: ψU = −2 tan−1

2H + D 4f

1 ψU − β and θe = 2 tan−1 tan − α M 2

(8-87)

The feed subtended angle of the subreﬂector is 2θe . We calculate the rim ellipse of the subreﬂector determined by the cone axis with angle α and cone angle θe using Eq. (8-59) and (8-60).

8-18 HORN REFLECTOR AND DRAGONIAN DUAL REFLECTOR The horn reﬂector shown in Figure 8-20 consists of a pyramidal or conical input section excited with a rectangular or circular waveguide mode, respectively, that feeds an offset paraboloidal reﬂector. The beam exits horizontally. The horn reﬂector geometry is an offset reﬂector with offset angle ψ0 = 90◦ and center offset H = 2f the same as the periscope conﬁguration. Figure 8-21 gives the pattern of a 3-m-diameter reﬂector operating at 6 GHz (diameter = 60λ) with f = 3.215 m (ψe = 15◦ ). The antenna radiates cross-polarization −23 dB relative to the beam peak in the horizontal plane. It cannot be used for two channels with different polarizations, because similar to all offset-fed reﬂectors, circularly polarized beams squint right and left in this horizontal plane. The antenna radiates a signiﬁcant sidelobe 90◦ from the boresight in the horizontal plane that can be controlled using serrated-edge blinders [51]. A Dragonian dual reﬂector uses a hyperbola subreﬂector that curves toward the main reﬂector in a Cassegrain system. This produces a dual reﬂector with magniﬁcation

b d bc

FIGURE 8-20 Horn reﬂector with serrated side blinders. (From [51], Fig. 3, 1973 IEEE.)

428

REFLECTOR ANTENNAS

Amplitude, dB

Symmetrical Plane Cross-Polarization

Pattern Angle (degrees)

FIGURE 8-21 Pattern of a 3-m horn reﬂector at 6 GHz.

M < 1 and we need a feed antenna with a wider beamwidth than would be required for efﬁcient feed of the main reﬂector. We use long focal lengths for the main reﬂector, which ﬂattens its curve, so that the feed beamwidth can remain small. Jones and Kelleher [52] applied this Cassegrain arrangement to a horn reﬂector and located the feed horn in the middle of the paraboloidal main reﬂector. Dragone [53] derived the generalized Mizugutch criterion for multiple reﬂectors and showed that it could be applied to this Cassegrain system to eliminate cross-polarization. Figure 8-22 shows a Dragonian dual reﬂector fed by a corrugated horn feed located beyond the rim of the paraboloidal reﬂector to replace the 3-m horn reﬂector. Pattern analysis produces the same curves as in Figure 8-21 except that cross-polarization in the horizontal plane is eliminated. The design given in Figure 8-22 has f = 9.8 m, D = 3 m, θe = 20◦ , and tilts the subreﬂector axis by −73◦ to place the main reﬂector and subreﬂector in different quadrants. The Mizugutch criterion between the subreﬂector axis tilt and the feed tilt locates the feed axis at −97.5◦ relative to the main reﬂector axis and −24.5◦ relative to the subreﬂector axis. The parameters of the reﬂector were adjusted so that the plane wave radiated from the main reﬂector misses both the feed corrugated horn and the subreﬂector. All dimensions can be found using available equation sets [54] that require various sets of ﬁve inputs to totally specify the dual reﬂector. By using these equations, we discover that the hyperboloidal subreﬂector rim is an ellipse with 2.05- and 2.60-m diameters. To use the hyperbola close to the feed, specify a negative eccentricity and the equations curve the hyperbola toward the main reﬂector. Similarly, the equation for magniﬁcation produces a value of less than 1 for a negative eccentricity. Given the e = −1.832 for the reﬂector of Figure 8-22, M=

e+1 −1.832 + 1 = = 0.2938 e−1 −1.832 − 1

SPHERICAL REFLECTOR

429

Feed

Main Reflector Beam

Subreflector

FIGURE 8-22 Dragonian dual-reﬂector geometry.

We compute feed tilt relative to the subreﬂector axis to satisfy the Mizugutch criterion: β −73 ◦ α = 2 tan−1 M tan = 2 tan−1 0.2938 tan = −24.5 2 2 8-19 SPHERICAL REFLECTOR When we feed-scan a paraboloidal reﬂector, the pattern sidelobes develop coma and the beam shape generally degrades. Feed scanning is limited. In a spherical reﬂector a feed moved in an arc from the center of the sphere and sees the same reﬂector geometry if we discount the edge effects. Greater scanning is possible, but the spherical reﬂector fails to focus an incident plane wave to a point and requires more elaborate feeds. We can design many types of feeds for the spherical reﬂector. The reﬂector can be fed from a point source for large f /D by assuming that it is a distorted parabola [55,56]. It can be fed with a line source to follow the axis ﬁelds. Corrector subreﬂectors can be designed to correct the spherical aberrations [58]. Like the parabolic reﬂector, we can design arrays [24] to compensate for spherical aberrations and give multiple beams. Figure 8-23 shows the geometry and ray tracing of a spherical reﬂector illuminated by a plane wave. All rays intersect a radial line of the sphere (the axis) in the direction of the incident wave because the reﬂector has circular symmetry about all axes. The diagram traces rays hitting the outer portion of the reﬂector as passing through the axis closer to the vertex than do the rays reﬂected from areas closer to the axis. The reﬂector has a line focus. A distorted paraboloidal reﬂector with a line focus exhibits spherical aberration because the focal length depends on the radial distance from the axis of the reﬂection point. The spherical reﬂector has a cusplike caustic where GO

430

REFLECTOR ANTENNAS

Paraxial Focus

y y

R/2

z

H

R/2 z

y R/2

Center

Caustic

FIGURE 8-23 Ray tracing in a spherical reﬂector.

predicts inﬁnite ﬁelds. The second side of Figure 8-23 traces a single ray. We can easily solve the isosceles triangle for the results: R/2 z=

1 − H 2 /R 2 R2 H 2 = R2 1 − 2 4z

(8-88) (8-89)

where z is the location of the focus for a given ray. As H approaches zero, with rays near the axis, the reﬂected ray passes through the paraxial focus (z = R/2). We use Eq. (8-89) to ﬁnd the power distribution on the axis by using the conservation of power. The power in a differential area of the plane wave reﬂects into a differential length on the axis: dA = 2πH dH . We differentiate Eq. (8-89) implicitly: 2H dH =

R2 dz 2z3

The power distribution along the axis is Pz =

P0 R 3 8z3

(8-90)

where P0 is the power at the paraxial focus. The peak power occurs at the paraxial focus and drops by one-eighth (−9 dB) at the vertex. We determine the required length of the line source feed from the rotation angle ψ of the illuminated portion of the reﬂector: R(1/ cos ψ − 1) feed length = (8-91) 2

SPHERICAL REFLECTOR

431

Example If the half-rotation angle of the illuminated region is 30◦ , the feed length is 0.0774R from Eq. (8-91). The amplitude decreases by [Eq. (8-90)]

R3 Pz = = 0.65 (−1.9 dB) P0 8(R/2 + 0.0774R)3

The rays intersecting the axis are not at constant phase. The path length from the aperture plane through the reﬂector origin is path length =

R2 +z 2z

(8-92)

We can approximate Eq. (8-92) by a linear function if the feed length is short. Example The feed length is 0.0774R long; calculate the phase change required along the feed. The feed starts at the paraxial focus (z = R/2): path length =

R2 R + = 1.5R R 2

At z = R/2 + 0.0774R, the path length = 1.443R. The phase change is (2π/λ)R(0.0566). If we plot the phase change over the region of the feed, we can approximate the phase change by a linear function very accurately. The spherical reﬂector can be fed from a point source when the f /D is large [55]. The center of the reﬂector approximates a parabola. The optimum focal point is f =

!

1 R + R 2 − (D/2)2 4

(8-93)

The maximum phase path length error is [56] L 1 D 1 = λ 2048 λ (f/D)3

(8-94)

The approximate gain loss is L 2 G = 3.5092 G λ or

L PEL(dB) = 10 log 1 − 3.5092 λ

2 (8-95)

A path length deviation of 0.25λ reduces the gain by 1.08 dB. Example Determine the f /D value of a spherical reﬂector to limit L to 1/16λ for a reﬂector diameter of 50λ.

432

REFLECTOR ANTENNAS

By rearranging Eq. (8-94), we ﬁnd that f 16(50) = = 0.73 D 2048 8-20

SHAPED REFLECTORS

Shaped reﬂectors spread cylindrical or spherical waves into a desired pattern that depends on geometric optics. Shaped reﬂectors do not radiate patterns exactly as prescribed by GO. In all cases we must apply techniques such as aperture diffraction, induced currents, or geometric theory of diffraction (GTD) to compute the actual pattern. We consider only the ﬁrst-order GO for design, although analysis requires more elaborate techniques. We use two principles to design shaped reﬂectors. The ﬁrst is GO reﬂection expressed as a differential equation. The second is the conservation of power in ray tubes, which can be expressed either in terms of differential areas or integrals over sections of the feed and reﬂection patterns. We deﬁne two angles for the GO reﬂection equation. The feed points toward the reﬂector, and we measure its pattern angle ψ from an axis pointed toward the reﬂector. The reﬂector reradiates the incident feed pattern in a far-ﬁeld pattern whose angle θ is measured from the axis pointing away from the reﬂector. The differential equation of reﬂection is [2] θ +ψ dρ tan = (8-96) 2 ρdψ where ρ is the distance of the reﬂector from the feed. The edges of the reﬂector are deﬁned by angles ψ1 and ψ2 measured from the feed axis and reﬂect in directions θ1 and θ2 . We integrate this differential equation for a solution: ρ(ψ) = ln ρ0 (ψ1 )

ψ

tan ψ1

θ (ψ) + ψ dψ 2

(8-97)

where ψ1 is some initial angle of the feed and ρ(ψ1 ) is the initial radius vector locating the reﬂector at ψ1 . GO, the zero-wavelength approximation, is consistent at any size. All parabolic reﬂectors collimate spherical waves radiated from the focus regardless of size. Only by considering diffraction or currents induced on the reﬂector can we compute gain and beamwidth of the antenna. Example

From Eq. (8-97), determine the reﬂector surface to give θ (ψ) = 0. ln

ρ(ψ) = ρ0 (ψ1 )

ψ

tan ψ1

ψ ψ1 ψ dψ = −2 ln cos − ln cos 2 2 2

By the properties of the ln function, this becomes ln

ρ(ψ) cos2 (ψ1 /2) = ln ρ0 (ψ1 ) cos2 (ψ/2)

433

SHAPED REFLECTORS

By taking the exponential of each side, we get the polar equation of the reﬂector: ρ(ψ) = ρ0 (ψ1 )

cos2 (ψ1 /2) cos2 (ψ/2)

We let ψ1 = 0 and set ρ(ψ) = f , and we obtain the equation for the parabola [Eq. (8-1)]. The differential equation of reﬂection tells us only the shape of the reﬂector locally to produce a reﬂection in a direction θ for an incident angle ψ. We must still ﬁnd the power density in various directions. The ratio of differential areas gives us these power densities. Given the pattern of the feed Gf (ψ, φ) and the pattern of the reﬂection P (θ, φ), KP(θ, φ) dA(θ, φ) = Gf (ψ, φ) dAf (ψ, φ) (8-98) where dA(θ, φ) is the differential area of the reﬂection pattern, dAf (ψ, φ) the differential area of the feed pattern, and K a constant found by equating the total incident and reﬂected powers. Equation (8-98) is based on the assumption that reﬂections are 1 : 1 with the feed pattern. 8-20.1 Cylindrical Reﬂector Synthesis We feed cylindrical reﬂectors with line sources. The reﬂector determines the beam shape in one plane and the line-source distribution in the other. The problem reduces to designing a single two-dimensional curve moved along an axis to deﬁne the reﬂector. The power radiated by the feed is given by Gf (ψ) dψ. The reﬂector directs this power at an angle θ whose power density is P (θ ) dθ . These are proportional [Eq. (8-98)]: KP (θ ) dθ = Gf (ψ) dψ

(8-99)

At the limits of the reﬂector, feed angles ψ1 and ψ2 reﬂect to angles θ1 and θ2 . We calculate the constant K by equating the power in each pattern:

ψ2

ψ1

K=

Gf (ψ) dψ

θ2 θ1

(8-100) P (θ ) dθ

We integrate Eq. (8-99) to derive a formal solution usually arrived at numerically. By combining Eqs. (8-99) and (8-100), we eliminate K:

θ

θ1 θ2 θ1

ψ

=

ψ1 ψ2

P (θ ) dθ P (θ ) dθ

ψ1

Gf (ψ) dψ (8-101) Gf (ψ) dψ

We use Eq. (8-101) with a known feed pattern Gf (ψ) and a desired pattern function P (θ ) to establish the relation θ (ψ). We insert θ (ψ) into the differential equation for

434

REFLECTOR ANTENNAS

the reﬂection [Eq. (8-97)] and determine radial distance as a function of ψ: ρ(ψ) = ρ0 (ψ1 ) exp

ψ

tan ψ1

θ (ψ) + ψ dψ 2

(8-102)

With Eq. (8-102) we calculate the reﬂector coordinates to within a scale factor ρ0 (ψ1 ). 8-20.2 Circularly Symmetrical Reﬂector Synthesis The synthesis of circularly symmetrical reﬂectors is readily reduced to a twodimensional problem. We must assume that the feed pattern is also axisymmetrical. We describe the feed pattern by Gf (ψ) and the reﬂector pattern by P (θ ). The differential areas are sin ψ dψ and sin θ dθ . Equation (8-98) becomes KP (θ ) sin θ dθ = Gf (ψ) sin ψ dψ

(8-103)

We integrate Eq. (8-103) to ﬁnd the function θ (ψ):

θ

θ1 θ2 θ1

P (θ ) sin θ dθ = P (θ ) sin θ dθ

ψ

ψ1 ψ2 ψ1

Gf (ψ) sin ψ dψ (8-104) Gf (ψ) sin ψ dψ

We use Eq. (8-102) from the reﬂection differential equation with θ (ψ) to determine the polar equation of the reﬂector. The design of a shaped reﬂector can best be explained with an example. The cylindrical reﬂector synthesis follows parallel steps. Example Design a reﬂector to transform the feed pattern of Figure 8-24a into the pattern of Figure 8-24b. The required pattern drops by about 9 dB from 50◦ to 0◦ . We will use the feed pattern from 4◦ to 54◦ and design a reﬂector with a ring caustic. The feed pattern at 4◦ is reﬂected to 50◦ , and the feed pattern at 54◦ is reﬂected to 0◦ . The geometric optics rays cross somewhere in front of the reﬂector. We have the following limits: Feed: ψ1 = 4◦ , ψ2 = 54◦ Reﬂection: θ1 = 50◦ , θ2 = 0 We insert the patterns of Figure 8-24 into both sides of Eq. (8-104) and evaluate the ratio of the integrals. Table 8-12 gives the results of these integrals for θ and ψ as normalized in Eq. (8-104). Given ψ, we ﬁnd θ by equating integrals in the table. For example, follow the line from ψ = 28◦ (feed) to its integral value, match the values of the integrals, and determine the reﬂection angle θ = 42◦ . We trace a number of these through interpolation to generate Table 8-13 of reﬂection angles θ (ψ) for given feed angles. We use Table 8-13 of θ (ψ) in the integral of Eq. (8-102) to calculate the normalized polar equation of the reﬂector listed in Table 8-14. Figure 8-25 shows the reﬂector shape as well as the geometry of axisymmetric-shaped reﬂectors. There is a hole in the center because the procedure fails to specify that region. Note in Table 8-12 how

SHAPED REFLECTORS

435

( a)

(b)

FIGURE 8-24 Axisymmetrical reﬂector pattern transformation: (a) feed pattern; (b) reﬂector pattern.

436

REFLECTOR ANTENNAS

TABLE 8-13 Shaped Reﬂector Synthesis Reﬂection Angles for Given Feed Angles, θ (ψ) Feed Angle, ψ (deg)

Reﬂection Angle, θ (deg)

Feed Angle, ψ (deg)

Reﬂection Angle, θ (deg)

4 6 8 10 12 14 16 18 20 22 24 26 28

50.0 50.0 49.9 49.8 49.6 49.3 48.8 48.2 47.4 46.4 45.2 43.7 42.1

30 32 34 36 38 40 42 44 46 48 50 52 54

40.2 38.2 36.0 33.6 31.1 28.4 25.6 22.6 19.5 16.2 12.6 8.4 0.0

TABLE 8-14 Shaped Reﬂector Synthesis Normalized Polar Equation of a Reﬂector, ρ(ψ) Feed Angle, ψ (deg)

Normalized Radius, ρ

Feed Angle, ψ (deg)

Normalized Radius, ρ

Feed Angle, ψ (deg)

Normalized Radius, ρ

4 6 8 10 12 14 16 18 20

1.000 1.018 1.038 1.058 1.080 1.103 1.128 1.153 1.180

22 24 26 28 30 32 34 36 38

1.208 1.238 1.268 1.299 1.332 1.365 1.398 1.433 1.468

40 42 44 46 48 50 52 54

1.503 1.539 1.575 1.611 1.648 1.684 1.719 1.753

much of the inner portion of the reﬂector must reﬂect rays near 50◦ to achieve the high gain required. We could repeat the example and design without a ring caustic where rays from the feed near 54◦ reﬂect to 50◦ and would produce a ﬂatter reﬂector. Because diffraction effects spread the pattern, we could approximate the pattern of Figure 8-24b by designing a reﬂector to point the beam in a cone about the axis. If we take Eq. (8-97) and let θ (ψ) = θ0 , a constant, we get the surface ρ(ψ) = ρ0

cos2 [(ψ0 + θ0 )/2] cos2 [(ψ + θ0 )/2]

(8-105)

Example Estimate the directivity of a 40λ-diameter reﬂector shaped by Eq. (8-105) to scan in a cone to θ0 = 50◦ .

SHAPED REFLECTORS

437

q y

r(y)

y Feed

y1 q1

Reflector

y2 q2

FIGURE 8-25 Circularly symmetrical reﬂector designed with a caustic reﬂector.

Only half of the diameter is used for each side. The effective scanned aperture width becomes (40λ/2) cos 50◦ = 12.8λ. If we assume a uniform-amplitude aperture distribution, we obtain an upper bound. From Eq. (4-83), HPBW = 59◦ /12.8 = 4.6◦ . We use Eq. (1-24) to estimate the directivity: directivity =

2 = 32.5 (15 dB) cos(50 − 2.3 ) − cos(50◦ + 2.3◦ ) ◦

◦

The boresight gain of the aperture with a uniform distribution is 42 dB. Spreading the reﬂection into a cone greatly reduces gain. The shaped reﬂector above will have even less directivity because it has a greater edge taper. 8-20.3 Doubly Curved Reﬂector for Shaped Beams It is a common radar requirement to have a narrow beam in one plane and a shaped beam in the other. Such beams can be obtained from shaped cylindrical reﬂectors, but it is simpler to replace the line source with a single feed. We only specify the pattern in the principal planes denoted: θV , the shaped pattern coordinate, and θH , the pencil

438

REFLECTOR ANTENNAS

beam [2,59]. Similarly, we specify the feed antenna pattern in terms of ψV and ψH . For a given feed angle ψV , the reﬂected wave angle is θV . The only θH value allowed is zero. The reﬂector collimates the wave in the horizontal plane. This collimation requires a symmetrical reﬂector made from parabolic curves in the horizontal plane. We design the vertical curve only through the center of the reﬂector. For a given feed angle ψV (Figure 8-26a), all incoming rays at an angle θV must be reﬂected into the feed. The incoming rays form the x –z plane in Figure 8-26, and the reﬂector collimates these to the feed by a parabola in the plane. We call this parabola a rib of the reﬂector. Figure 8-26b shows the plane and two rays reﬂecting into the feed from a wave arriving at an angle θV in the x –z plane. For the beam to focus, the optical path lengths must be equal: BP + P O = AN + N O

(8-106)

Equation (8-106) establishes the curve of the rib in the x –z plane as a parabola with focal length θV (ψV ) + ψV f = ρc (ψV ) cos2 (8-107) 2 with the focus located on the z -axis. Using the parabolic ribs reduces the problem to the design of the central curve ρc (ψV ).

y

Central Curve z qV

yV z

Boresight

(a) y

Rib

Central Curve

Incoming Wave, x − z′ Plane z′

qV

P

qV r(yV)

N yV

Focus O

x

z

(b)

FIGURE 8-26

Doubly curved shaped reﬂector.

SHAPED REFLECTORS

439

The reﬂected and feed power densities modify Eq. (8-98) to KP (θV )dθV ρc (ψV ) dψH = Gf (ψV ) dψV dψH

(8-108)

We integrate Eq. (8-108) and normalize to the total power:

θV

θ1 θ2 θ1

P (θV ) dθV = P (θV ) dθV

ψV

ψ1 ψ2 ψ1

[Gf (ψV )/ρc (ψV )] dψV (8-109) [Gf (ψV )/ρc (ψV )] dψV

Equation (8-109) is similar to Eqs. (8-101) and (8-104) except that the feed pattern integral value depends on the radial distance to the central rib. We must know ρc (ψV ) before we can determine θV (ψV ), which will be required to compute ρc (ψV ) from the reﬂection differential equation [Eq. (8-102)]. The solution can be found only by an iterative process. We must assume a ρc (ψV ), solve for θV (ψV ), and use the result to compute a new ρc (ψV ). After a few iterations, the values of ρc (ψV ) converge. We use the normalized ρc with the foregoing ratio of integrals. We start with a parabola: cos2 (ψ1 /2) ρc (ψV ) = ρc (ψ1 ) cos2 (ψV /2) The surface generated by following the method may not be deﬁned uniquely. We pick a constant width for the reﬂector in the horizontal plane. We deﬁne the surface with a continuous series of parabolas each in a x –z plane determined by the reﬂection angle θV , which changes direction along the central rib. We must plot the curve of the vertical coordinate of the edge versus ψV to see if it is monotonic. If there are loops in the curve, the surface deﬁned is not unique. Given the width x, we calculate the vertical coordinate of the edge by the following development. The location of the rib on the central curve is given by ρc (ψV ) sin ψV . The rib is a parabola in the x –z plane with its focus given by Eq. (8-107). The z coordinate at the edge is z = x 2 /4f (ψV ). We determine the vertical dimension by projecting this point onto the y-axis: y = ρc (ψV ) sin ψV . Elliott [60, p. 500] points out that by following this method, one does not get the proper slope for reﬂection at all points, but we will get the desired pattern when we design for only small deviations from a pencil beam. The surface can be designed with or without a caustic depending on the reﬂection angles at the edges. Reﬂectors designed with caustic edge reﬂections have a better chance of being unique [60]. Carberry [61] presents a method of analysis that involves physical optics. When we apply these methods, we must subdivide the reﬂector into many patches because the phases of the currents change rapidly with position on the reﬂector, and the analysis must be repeated with ﬁner and ﬁner patches until the result converges. 8-20.4 Dual Shaped Reﬂectors We can design a dual-reﬂector antenna to produce an arbitrary phase and amplitude in the aperture plane by shaping both reﬂectors. By using both the conservation of power

440

REFLECTOR ANTENNAS

and the differential equations of reﬂection on the two surfaces, Galindo [62] derived a pair of differential equations in terms of the aperture radius. Runge–Kutta or any other suitable numerical method can be used to solve the simultaneous differential equations instead of an integration of the power equation. Williams [63] ﬁnds a solution to Cassegrain antennas within the restriction of equal amplitude and phase in the aperture plane by integration of the power equation. Collins [64] considers using a parabolic reﬂector for the main reﬂector, since the difference between the shaped main reﬂector and a parabola is small. He accepts a quadratic phase error in the aperture. Existing large reﬂectors can be retroﬁtted with a shaped subreﬂector to improve performance. For the method to work, an axisymmetric feed such as a corrugated horn is required. Galindo-Israel and Mittra [65] use a pair of reﬂectors offset from each other to transform a spherical wave from a feed antenna into a second spherical wave with a modiﬁed pattern amplitude. This combination of a feed with two reﬂectors can illuminate either prime focus paraboloidal reﬂectors or Cassegrain systems without modiﬁcation of the existing reﬂector surfaces. For example, a sec4 (θ/2) pattern can be realized from an ordinary pattern source to increase the aperture efﬁciency of the overall reﬂector system. The reﬂectors maintain equal GO path lengths for all rays, but they only approximate the desired pattern amplitude from the virtual focus. The procedure can be used to determine the contours of the reﬂectors along radial lines through numerical solution of differential equations. The equations develop from simplifying assumptions that depend on the extra degree of freedom introduced by the second reﬂector. In most cases, solution of the equations produces usable designs, although the method is not exact. Lee et al. [66] developed a method to shape offset-fed dual reﬂectors that reduces to the solution of a differential equation similar to that of the single-reﬂector design given above. The reﬂection properties of the subreﬂector determine the main reﬂector amplitude distribution to ﬁrst order. This method does not produce exact results but is close enough for engineering purposes. We start with a desired aperture power distribution P (r, φc ) and a known feed power pattern Gf (θ , φc ) given in the radial direction φc . Most cases use distributions independent of φc , but the design is performed along these planes. For a circularly symmetric design we only need to solve the differential equation along one plane, but the general case requires solutions along enough planes to allow splines along the coordinate φc to ﬁnd every point on both reﬂectors. A differential expression relates the feed power to the aperture power: Gf (θ , φc ) sin θ dθ = P (r, φc )r dr This leads to a ratio of integrals:

θ

−θe θe −θ e

Gf (θ ) sin θ dθ

Gf (θ ) sin θ dθ

"R R

P (r )r dr

R1

P (r )r dr

= " R12

(8-110)

Equation (8-110) covers the general case where the offset subreﬂector directs power from a lower angle −θe to an offset radius R1 that changes for each plane φc . For a circularly symmetric design, −θe = 0 and R1 = 0. Although many designs attempt to generate a uniform aperture distribution for the main reﬂector, we can substitute

SHAPED REFLECTORS

441

any distribution, such as a circular Taylor distribution to control the sidelobes into Eq. (8-110). Given the aperture distribution and the feed pattern, we calculate a table similar to Table 8-12 for each plane φc that gives the feed angle as a function of aperture radius. We interpolate on this table to determine every value. We start at the center of the subreﬂector described in spherical coordinates (ρ0 , 0, 0) relative to the axis of the subreﬂector centered at the feed focus. The subreﬂector axis may be tilted relative to the main reﬂector axis. The rectangular coordinates of the subreﬂector are (ρ sin θ cos φc , ρ sin θ sin φc , ρ cos θ ). The incident wave reﬂects to a point on the main reﬂector: (H ± R cos φc , ±R sin φc , z) using + Cassegrain, −Gregorian. We calculate the unit vector between the subreﬂector point and the main reﬂector. The normal vector on the subreﬂector is expressed as a differential: 1 1 ∂ρ 1 ∂ρ n= aθ − aφ (8-111) aρ − ρ ∂θ ρ sin θ ∂φc c

where =

1+

1 ∂ρ ρ ∂θ

2 +

1 ∂ρ ρ sin θ ∂φc

2

We apply both equations of Snell’s law [Eq. (2-67)] to the subreﬂector reﬂection and gather terms to form a pair of differential equations: QV ∂ρ = 2 ∂θ Q + U2

and

∂ρ U V sin θ = 2 ∂φc Q + U2

(8-112)

The terms of Eq. (8-112) are given by the expressions a cos θ cos φc + b sin θ sin φc − c sin θ ρ b cos φc − a sin φc U= ρ

Q=

V = L + a sin θ cos φc + b sin θ sin φc + c sin θ

(8-113)

a = H ± R cos φc − ρ sin θ cos φc b = ±R sin φc − ρ sin θ sin φc c = z − ρ cos θ where the vector (a, b, c) is from the subreﬂector to the main reﬂector and L = √ a 2 + b2 + c2 . We choose z = 0 as the aperture and equate path lengths along every ray. This gives an equation for the z-position of the main reﬂector: OL = ρ0 + L0 − z0 = ρ + L − z We solve for z: z=

a 2 + b2

1 + (ρ cos θ − ρ + OL) 2(ρ cos θ − ρ + OL) 2

(8-114)

442

REFLECTOR ANTENNAS

To solve for the reﬂector surfaces, we choose a starting point, usually the center of the reﬂector as the ﬁrst ray from the feed to the subreﬂector, and calculate the initial distance L0 between the subreﬂector and the main reﬂector to ﬁnd the path length. We select a polar plane φc and solve the left differential equation [Eq. (8-112)] for both surfaces using a Runge–Kutta numerical solution. We repeat this in a sufﬁcient number of planes φc to specify the surface totally. If the antenna is circularly symmetric, we solve the equation only once. For an offset dual reﬂector we can improve the cross-polarization by computing an equivalent subreﬂector using least squares and use its eccentricity to calculate the Mizugutch subreﬂector axis rotation.

8-21 OPTIMIZATION SYNTHESIS OF SHAPED AND MULTIPLE-BEAM REFLECTORS Silver [2] discusses using a linear array feed to shape the beam of a paraboloidal reﬂector. The method is quite empirical and involves the addition of a number of offset beams. A similar technique is used in three-dimensional radar, but the feeds are kept separate so that multiple beams can scan a larger area in a given time. An array feed provides the best solution to beam shaping in many cases. The number of elements in the array limits the number of variables to a ﬁnite set to which optimization techniques can be applied. A second method uses optimization to shape the reﬂector and possible subreﬂectors. This requires distortion functions on the reﬂectors. We start with conic-section reﬂectors and add distortions. These distortions can be global Zernike functions deﬁned over the total surface, or they could be localized functions such as B-splines [67]. A B-spline uses a grid of points on the reﬂector, but the spline coefﬁcients apply only over a limited area. In both cases we obtain a set of coefﬁcients used in the optimization algorithms. We have the choice of combining these coefﬁcients, or we can iterate between different sets of coefﬁcients. Optimization is an art. Because the reﬂector is an aperture antenna, we pick a set of directions in (u, v) = (sin θ cos φ, sin θ sin φ) space to evaluate the pattern. The number of points should exceed the number of coefﬁcients and be spaced close enough to fully describe the main-beam pattern: 0.5λ 0.25λ u and v ∼ to (8-115) D D We calculate the pattern power Pm (u, v) at these points and compare them to the desired pattern Pmd (u,v) using a suitable cost function. We weight each pattern direction ωm and use a summation cost [68] with a gradient minimization technique: F (x) =

M

|ωm (Pm (um , vm )) − Pmd (um , vm )|2

(8-116)

m=1

A second choice is a min–max optimization [69]. This algorithm minimizes the maximum error: max[ωm (Pm (um , vm )) − Pmd (um , vm )] (8-117)

REFERENCES

443

If we optimize the reﬂector shape, we express the distortion as B-splines speciﬁed at evenly spaced points across the aperture with the number determined by the maximum pattern angle θmax and the reﬂector diameter D [69]: Nx = Ny =

πD sin θmax +2 λ

(8-118)

Given a Zernike polynomial expansion with maximum azimuthal mode expansion Mmax and maximum polar mode index Nmax , we have similar mode number requirements: Mmax = Nmax =

πD sin θmax +2 λ

(8-119)

Shaping starts with a paraboloid main reﬂector whose beamwidth may be so narrow that a portion of the speciﬁed u–v space area may lie in the sidelobe region. In this case the optimization may become trapped because it cannot satisfy this area when changes effecting the main beam region positively affect the sidelobe region negatively. We must distort the main reﬂector before starting the optimization [69]. First surround the u–v space area of speciﬁed points with an ellipse centered at (u0 , v0 ) with major radius ω1 and minor radius ω2 tilted an angle α. Given a paraboloid with diameter D, focal length f , and center offset (x0 , y0 ), we deﬁne rotated coordinates on the aperture. x = (x − x0 ) cos α + (y − y0 ) sin α y = −(x − x0 ) sin α + (y − y0 ) cos α Using these coordinates, we alter the z-axis position of the reﬂector:

1 x2 + y2 + z = − 2 8f 2

ω1 x 2 + ω2 y 2 + u0 (x − x0 ) + v0 (y − y0 ) D

(8-120)

We have a choice with ω1 and ω2 because they can be both either positive or negative. Positive values ﬂatten the reﬂector while negative values cause a caustic reﬂection to broaden the beam. REFERENCES 1. A. W. Love, Reﬂector Antennas, IEEE Press, New York, 1978. 2. S. Silver, ed., Microwave Antenna Theory and Design, McGraw-Hill, New York, 1949. 3. W. V. T. Rusch and P. D. Potter, Analysis of Reﬂector Antennas, Academic Press, New York, 1970. 4. P. J. Wood, Reﬂector Antenna Analysis and Design, Peter Peregrinus, London, 1980. 5. C. A. Mentzer and L. Peters, A GTD analysis of the far-out sidelobes of Cassegrain antennas, IEEE Transactions on Antennas and Propagation, vol. AP-23, no. 5, September 1975, pp. 702–709. 6. S. W. Lee et al., Diffraction by an arbitrary subreﬂector: GTD solution, IEEE Transactions on Antennas and Propagation, vol. AP-27, no. 3, May 1979, pp. 305–316. 7. A. D. Craig and P. D. Simms, Fast integration techniques for reﬂector antenna pattern analysis, Electronics Letters, vol. 18, no. 2, January 21, 1982, pp. 60–62.

444

REFLECTOR ANTENNAS

8. V. Galindo-Israel and R. Mittra, A new series representation for the radiation integral with application to reﬂector antennas, IEEE Transactions on Antennas and Propagation, vol. AP25, no. 5, September 1977, pp. 631–641. 9. A. W. Rudge et al., eds., The Handbook of Antenna Design, Vol. 1, Peter Peregrinus, London, 1982. 10. A. C. Ludwig, The deﬁnition of cross polarization, IEEE Transactions on Antennas and Propagation, vol. AP-21, no. 1, January 1973, pp. 116–119. 11. J. R. Cogdell and J. H. Davis, Astigmatism in reﬂector antennas, IEEE Transactions on Antennas and Propagation, vol. AP-21, no. 4, July 1973, pp. 565–567. 12. Y. T. Lo, On the beam deviation factor of a parabolic reﬂector, IEEE Transactions on Antennas and Propagation, vol. AP-8, no. 3, May 1960, pp. 347–349. 13. J. Ruze, Lateral-feed displacement in a paraboloid, IEEE Transactions on Antennas and Propagation, vol. AP-13, no. 5, September 1965, pp. 660–665. 14. W. V. T. Rusch and A. C. Ludwig, Determination of the maximum scan-gain contours of a beam scanned paraboloid and their relation to the Petzval surface, IEEE Transactions on Antennas and Propagation, vol. AP-21, no. 2, March 1973, pp. 141–147. 15. W. A. Imbriale, P. G. Ingerson, and W. C. Wong, Large lateral feed displacements in a parabolic reﬂector, IEEE Transactions on Antennas and Propagation, vol. AP-22, no. 6, November 1974, pp. 742–745. 16. D. K. Cheng, Effect of arbitrary phase errors on the gain and beamwidth characteristics of radiation pattern, IEEE Transactions on Antennas and Propagation, vol. AP-3, no. 4, July 1955, pp. 145–147. 17. J. Ruze, Antenna tolerance theory: a review, Proceedings of IRE, vol. 54, no. 4, April 1966, pp. 633–640. 18. M. S. Zarghamee, On antenna tolerance theory, IEEE Transactions on Antennas and Propagation, vol. AP-15, no. 6, November 1967, pp. 777–781. 19. R. C. Hansen, ed., Microwave Scanning Antennas, Academic Press, New York, 1964. 20. W. V. T. Rusch and R. D. Wanselow, Boresight gain loss and gore related sidelobes of an umbrella reﬂector, IEEE Transactions on Antennas and Propagation, vol. AP-30, no. 1, January 1982, pp. 153–157. 21. W. H. Watson, The ﬁeld distribution in the focal plane of a paraboloidal reﬂector, IEEE Transactions on Antennas and Propagation, vol. AP-12, no. 5, September 1964, pp. 561–569. 22. T. B. Vu, Optimization of efﬁciency of reﬂector antennas: approximate method, Proceedings of IEE, vol. 117, January 1970, pp. 30–34. 23. B. M. Thomas, Theoretical performance of prime focus paraboloids using cylindrical hybrid modes, Proceedings of IEE, vol. 118, November 1971, pp. 1539–1549. 24. N. Amitay and H. Zucker, Compensation of spherical reﬂector aberrations by planar array feeds, IEEE Transactions on Antennas and Propagation, vol. AP-20, no. 1, January 1972, pp. 49–56. 25. V. Galindo-Israel, S. W. Lee, and R. Mittra, Synthesis of laterally displaced cluster feed for a reﬂector antenna with application to multiple beams and contoured patterns, IEEE Transactions on Antennas and Propagation, vol. AP-26, no. 2, March 1978, pp. 220–228. 26. B. Popovich et al., Synthesis of an aberration corrected feed array for spherical reﬂector antennas, IEEE/APS Symposium Digest, May 1983. 27. V. Mrstik, Effect of phase and amplitude quantization errors on hybrid phased-array reﬂector antennas, IEEE Transactions on Antennas and Propagation, vol. AP-30, no. 6, November 1982, pp. 1233–1236. 28. C. M. Knop, On the front to back ratio of a parabolic dish antenna, IEEE Transactions on Antennas and Propagation, vol. AP-24, no. 1, January 1976, pp. 109–111.

REFERENCES

445

29. M. Uhm, A. Shishlov, and K. Park, Offset-paraboloid geometry: relations for practical use, IEEE Antennas and Propagation Magazine, vol. 38, no. 3, June 1996, pp. 77–79. 30. R. F. H. Yang, Illuminating curved passive reﬂector with defocused parabolic antenna, 1958 IRE Wescon Convention Record, August 1958, pp. 260–265. 31. C. Granet, Designing axially symmetric Cassegrain and Gregorian dual-reﬂector antennas from combinations of prescribed geometric parameters, IEEE Antennas and Propagation Magazine, vol. 40, no. 2, April 1998, pp. 76–82. 32. C. Granet, Designing axially symmetric Cassegrain and Gregorian dual-reﬂector antennas from combinations of prescribed geometric parameters, part 2: minimum blockage condition while taking into account the phase-center of the feed, IEEE Antennas and Propagation Magazine, vol. 40, no. 3, June 1998, pp. 82–85. 33. W. V. T. Rusch, Phase error and associated cross polarization effects in Cassegrainian-fed microwave antennas, IEEE Transactions on Antennas and Propagation, vol. AP-14, no. 3, May 1966, pp. 266–275. 34. P.-S. Kildal, The effects of subreﬂector diffraction on the aperture efﬁciency of a conventional Cassegrain antenna: an analytical approach, IEEE Transactions on Antennas and Propagation, vol. AP-31, no. 6, November 1983, pp. 903–909. 35. A. M. Isber, Obtaining beam-pointing accuracy with Cassegrain antennas, Microwaves, August 1967, pp. 40–44. 36. W. V. T. Rusch and R. Wohlleben, Surface tolerance loss for dual-reﬂector antennas, IEEE Transactions on Antennas and Propagation, vol. AP-30, no. 4, July 1982, pp. 784–785. 37. A. F. Kay, Electrical design of metal space frame radomes, IEEE Transactions on Antennas and Propagation, vol. AP-13, no. 2, March 1965, pp. 188–202. 38. P.-S. Kildal, E. Olsen, and J. A. Aas, Losses, sidelobes, and cross polarization caused by feed-support struts in reﬂector antennas: design curves, IEEE Transactions on Antennas and Propagation, vol. AP-36, no. 2, February 1988, pp. 182–190. 39. W. V. T. Rusch et al., Forward scattering from square cylinders in the resonance region with application to aperture blockage, IEEE Transactions on Antennas and Propagation, vol. AP-24, no. 2, March 1976, pp. 182–189. 40. G. T. Ruck, ed., Radar Cross Section Handbook, Vol. 1, Plenum Press, New York, 1970. 41. G. W. Collins, Noise temperature calculations from feed system characteristics, Microwave Journal, vol. 12, December 1969, pp. 67–69. 42. B. E. Kinber, On two-reﬂector antennas, Radioengineering and Electronics, vol. 7, no. 6, 1962, pp. 973–980. 43. A. P. Popov and T. A. Milligan, Amplitude aperture-distribution control in displaced-axis two reﬂector antennas, IEEE Antennas and Propagation Magazine, vol. 39, no. 6, December 1997, pp. 58–63. 44. S. P. Morgan, Some examples of generalized Cassegrainian and Gregorian antennas, IEEE Transactions on Antennas and Propagation, vol. AP-12, no. 6, November 1964, pp. 685–691. 45. C. Granet, A simple procedure for the design of classical displaced-axis dual-reﬂector antennas using a set of geometric parameters, IEEE Antennas and Propagation Magazine, vol. 41, no. 6, December 1999, pp. 64–72. 46. T. A. Milligan, The effects of feed movement on the displaced-axis dual reﬂector, IEEE Antennas and Propagation Magazine, vol. 40, no. 3, June 1998, pp. 86–87. 47. Y. Mizugutch, M. Akagawa, and H. Yokoi, Offset dual reﬂector antenna, IEEE Symposium on Antennas and Propagation Digest, 1976, pp. 2–5. 48. P. H. Nielson and S. B. Sørensen, Grasp8 Software Users Manual, Ticra, Copenhagen, 2001.

446

REFLECTOR ANTENNAS

49. C. Granet, Designing classical offset Cassegrain or Gregorian dual-reﬂector antennas from combinations of prescribed geometric parameters, IEEE Antennas and Propagation Magazine, vol. 44, no. 3, June 2002, pp. 114–123. 50. C. Granet, Designing classical offset Cassegrain or Gregorian dual-reﬂector antennas from combinations of prescribed geometric parameters, part 2: feed-horn blockage conditions, IEEE Antennas and Propagation Magazine, vol. 45, no. 6 December 2003, pp. 86–89. 51. D. T. Thomas, Design of multiple-edge blinders for large horn reﬂector antennas, IEEE Transactions on Antennas and Propagation, vol. AP-21, no. 2, March 1973, pp. 153–158. 52. S. R. Jones and K. S. Kelleher, A new low noise, high gain antenna, IEEE International Convention Record, March 1963, pp. 11–17. 53. C. Dragone, Offset multireﬂector antennas with perfect pattern symmetry and polarization discrimination, Bell System Technical Journal, vol. 57, no. 7, September 1978, pp. 2663–2684. 54. C. Granet, Designing classical dragonian offset dual-reﬂector antennas from combinations of prescribed geometric parameters, IEEE Antennas and Propagation Magazine, vol. 43, no. 6, December 2001, pp. 100–107. 55. T. Li, A study of spherical reﬂectors as wide-angle scanning antennas, IEEE Transactions on Antennas and Propagation, vol. AP-7, no. 4, July 1959, p. 223–226. 56. R. Woo, A multiple-beam spherical reﬂector antenna, JPL Quarterly Technical Review, vol. 1, no. 3, October 1971, pp. 88–96. 57. A. W. Love, Spherical reﬂecting antennas with corrected line sources, IEEE Transactions on Antennas and Propagation, vol. AP-10, no. 5, September 1962, pp. 529–537. 58. F. S. Bolt and E. L. Bouche, A Gregorian corrector for spherical reﬂectors, IEEE Transactions on Antennas and Propagation, vol. AP-12, no. 1, January 1964, pp. 44–47. 59. A. S. Dunbar, Calculation of doubly curved reﬂectors for shaped beams, Proceedings of IRE, vol. 36, no. 10, October 1948, pp. 1289–1296. 60. R. S. Elliott, Antenna Theory and Design, Prentice-Hall, Englewood Cliffs, NJ, 1981. 61. T. F. Carberry, Analysis theory for the shaped beam doubly curved reﬂector antenna, IEEE Transactions on Antennas and Propagation, vol. AP-17, no. 2, March 1969, pp. 131–138. 62. V. Galindo, Design of dual reﬂector antennas with arbitrary phase and amplitude distributions, IEEE Transactions on Antennas and Propagation, vol. AP-12, no. 4, July 1964. pp. 403–408. 63. W. F. Williams, High efﬁciency antenna reﬂector, Microwave Journal, vol. 8, July 1965, pp. 79–82. 64. C. Collins, Shaping of subreﬂectors in Cassegrainian antennas, IEEE Transactions on Antennas and Propagation, vol. AP-21, no. 3, May 1973, pp. 309–313. 65. V. Galindo-Israel and R. Mittra, Synthesis of offset dual shaped subreﬂector antennas for control of Cassegrain aperture distributions, IEEE Transactions on Antennas and Propagation, vol. AP-32, no. 1, January 1984, pp. 86–92. 66. J. J. Lee, L. I. Parad, and R. S. Chu, A shaped offset-fed dual-reﬂector antenna, IEEE Transactions on Antennas and Propagation, vol. AP-27, no. 2, March 1979, pp. 165–171. 67. M. E. Mortenson Geometric Modeling, Wiley, New York, 1985. 68. C. C. Han and Y. Hwang, Satellite antennas, Chapter 21 in Y. T. Lo and S. W. Lee, eds., Antenna Handbook, Van Nostrand Reinhold, New York, 1993. 69. H.-H. Viskum, S. B. Sørensen, and M. Lumholt, User’s Manual for POS4, Ticra, Copenhagen, 2003.

9 LENS ANTENNAS

In lenses, as in parabolic reﬂectors, we utilize free space as a feed network to excite a large aperture. Because we locate the feed behind the aperture, the conﬁguration eliminates aperture blockage and allows direct connection of the feed to the transmitter or receiver. When frequencies are above microwaves, this feeding method removes lossy transmission lines that increase system noise. Lenses have only half the tolerance requirements of reﬂectors because the wave passes by the anomaly only once. In a reﬂector the wave path deviates by twice the distance as the wave travels to and from the reﬂector. At low microwave frequencies the lens is prohibitively heavy, but zoning and the use of artiﬁcial dielectrics reduce this problem. Both zoning and artiﬁcial dielectrics present mechanical stability problems and narrow the bandwidth. We organize the design of lenses by the available degrees of freedom. A single lens with a uniform dielectric has two surfaces and is equivalent to a dual reﬂector because each surface is a degree of freedom. We start our discussion with single-surface lenses where we eliminate one degree of freedom by making the second surface match either the incoming or outgoing wave. Shaping both surfaces lets us correct one lens anomaly. We can remove either coma to improve the feed scanning or design to convert a given feed pattern to a desired aperture distribution. Bootlace lenses have three possible degrees of freedom. They consist of back-to-back arrays with cables connecting the sides. Normally, we use the degrees of freedom of the bootlace lens to increase the number of focal points. We give up degrees of freedom in many designs to simplify the mechanical layout. Finally, we discuss the use of a variable index of refraction in the Luneburg lens. We base the design of lenses on geometric optics. Like parabolic reﬂectors, lenses have no inherent frequency bandwidth limitation. We are limited by the feeds and mechanical problems of large sizes. Because we borrow from optics, lenses have great high-frequency potential. Modern Antenna Design, Second Edition, By Thomas A. Milligan Copyright 2005 John Wiley & Sons, Inc.

447

448

LENS ANTENNAS

9-1 SINGLE REFRACTING SURFACE LENSES Single-surface lenses convert the wave type, such as spherical to planar, at one surface through refraction. The constant-phase surface (eikonal) of the wave type determines the shape of the second surface of the lens. The common lens converts an incident spherical or cylindrical wave to a plane wave. Conversion to a cylindrical wave requires a line source feed and cylindrical surfaces on the lens. Spherical waves use point feeds and axisymmetrical surfaces. Like the reﬂector, which also converts spherical waves from the feed to plane waves by geometric optics (GO), diffraction from an aperture determines the far-ﬁeld pattern. Consider the second or nonrefracting surface. If the surface toward the feed converts the wave type, the wave exits the second surface as a plane wave and it is a plane. Similarly, when the surface away from the feed converts the exiting wave to a plane wave, the inner surface toward the feed follows the incident wave eikonal, a cylindrical or spherical surface. Figure 9-1 shows the two types of single refracting surface lenses. We can determine the refracting surface shape by either of two different approaches. Snell’s law can be applied to the refracting surface, and the surface slope can be determined for each feed angle. Equivalently, we can apply Fermat’s principle to equalize the optical path length from the feed through the lens to an aperture plane. The designs are easily found [1,2]. For Figure 9-1a, ρ(ψ) =

(n − 1)f n cos ψ − 1

(9-1)

where n is the index of refraction, given by n=

√ εr µ r

(9-2)

and εr and µr are the relative permittivity and permeability of the lens medium. When n > 1, Eq. (9-1) describes a hyperbola with the feed at one focus. The distance from the feed to the hyperbola along the axis is f . The asymptotes of the hyperbola limit the collimated portion of the feed radiation: ψa = cos−1

1 n

(9-3)

We must limit the lens edge to angles less than ψa because the asymptotes imply an inﬁnite aperture size. Similar to the paraboloidal reﬂector, we have feed spillover, considered to be lost in sidelobes. But, for example, placing the lens at the aperture of a horn eliminates spillover. We calculate the lens diameter from D = 2ρ sin ψe =

2(n − 1)f sin ψe n cos ψe − 1

(9-4)

where ψe is the edge angle, subject to the restriction of Eq. (9-3). The surfaces of the lens in Figure 9-lb have the polar equations ρ1 = constant

ρ2 (ψ) =

(n − 1)f n − cos ψ

(9-5)

SINGLE REFRACTING SURFACE LENSES

449

r(y)

Feed

Focus Hyperbolical

Plane

(a)

r2(y) Feed y r1

Focus

Focus Spherical (Circular)

Elliptical

(b)

FIGURE 9-1 Single-surface lenses.

The inner surface must either be a circular cylinder (cylindrical lens) or be spherical (axisymmetrical lens). The outer surface ρ2 (ψ) is elliptical for n > 1. The junction of the circle and ellipse determines the feed angle limitation: cos ψe = n −

(n − 1)f ρ1

(9-6)

We can, of course, truncate the lens before the two curves meet. Equation (9-6) gives the limitation on ρ1 at the lens edge as well: ρ1 ≤

(n − 1)f n − cos ψe

(9-7)

Example Compute f and ρ1 at the edge for an elliptical lens (Figure 9-lb) with D = 10λ, n = 1.6 (polystyrene), and ψe = 50◦ . Solve Eq. (9-5) for f /D: f n − cos ψe = D 2 sin ψe (n − 1) D ρ1 ≤ 2 sin ψe

f = 10.41λ

(9-8) (9-9)

450

LENS ANTENNAS

If ρ1 remained constant to the center of the lens, it would be 3.88λ thick. In narrowband applications we can remove the multiple wavelength thicknesses by zoning and reduce weight and dielectric material loss. Lenses change the amplitude distribution of the feed in the aperture plane. We relate the feed pattern to the aperture distribution through the conservation of power in differential areas. For an axisymmetrical lens: F (ψ, φ) sin ψ dψ dφ feed power

=

A(r, φ)r dr dφ aperture power

(9-10)

where ψ is the feed angle and r, ρ sin ψ, is the aperture radial distance; F (ψ, φ) is the feed power pattern and A(r, φ) is the aperture power distribution: A(r, φ) sin ψ dψ = F (ψ, φ) r dr

(9-11)

For a cylindrical lens, we also equate differential area multiplied by the feed or aperture power: F (ψ, y) dψ dy = A(r, y) dr dy dψ A(r, y) = F (ψ, φ) dr

(9-12)

We substitute Eqs. (9-11) and (9-12) into Eq. (9-1) for the hyperbolical lens to calculate the aperture distribution relative to the feed power pattern. Axisymmetrical A(r, φ) (n cos ψ − 1)3 = 2 F (ψ, φ) f (n − 1)2 (n − cos ψ)

Cylindrical (n cos ψ − 1)2 A(r, y) = F (ψ, y) f (n − 1)(n − cos ψ)

(9-13a, b)

The ﬁeld distribution is the square root of Eq. (9-13). We substitute Eqs. (9-11) and (9-12) into Eq. (9-5) for the elliptical lens: Axisymmetrical (n − cos ψ)3 A(r, φ) = 2 F (ψ, φ) f (n − 1)(n cos ψ − 1)

Cylindrical (n − cos ψ)2 A(r, y) = F (ψ, φ) f (n − 1)(n cos ψ − 1) (9-14a, b) The hyperbolical and elliptical lenses concentrate the aperture power in different ways. The hyperbolical lens reduces the feed power directed toward the edges and produces an additional aperture taper. On the other hand, an elliptical single-surface lens increases the power toward the edges as compared with the center. Example For axisymmetrical lenses with ψe = 50◦ and n = 1.6, compute the edge taper due to the lenses.

451

ZONED LENSES

We divide Eq. (9-13b) with ψe = ψ by the same equation with ψ = 0 to determine the ratio of power at the edge to that at the center of the aperture (assuming an isotropic feed). We do the same calculation with Eq. (9-14a):

Ae = Ac

(n cos ψe − 1)3 (n − 1)2 (n − cos ψe ) (n − cos ψe )3 (n − 1)2 (n cos ψe − 1)

hyperbolical lens

(9-15)

elliptical lens

(9-16)

By substituting ψe and n, we compute the edge taper. hyperbolical lens: 0.038 (−14.2 dB)

elliptical lens : 7.14 (8.5 dB)

The increased taper of the hyperbolical lens reduces sidelobes, and the elliptical lens increases aperture efﬁciency by compensating for some of the feed antenna pattern taper to make the aperture distribution more uniform.

9-2 ZONED LENSES Lenses designed by the methods of Section 9-1 have bandwidth limitations determined only by the invariability of the dielectric constant. Zoning removes multiples-ofwavelength path lengths in the lens to reduce weight, to reduce the lens-induced amplitude taper, or to thin the lens. The act of changing dimensions by wavelengths implies narrowing the frequency bandwidth. We step the lens in either the nonrefracting or refracting surface. Stepping the nonrefracting surface (Figure 9-2a, b) has the least effect. The edges of the steps, parallel with the waves, will diffract waves and cause some change in the aperture ﬁelds, but GO predicts no effect. Stepping the refracting surface introduces losses either as misdirected feed power (Figure 9-2c, d) or as unexcited aperture (Figure 92e, f). But stepping the refracting surface reduces the lens-induced aperture taper by changing the focal lengths in various zones. Figure 9-2 shows the limits in the two types of refracting surface steps, since we could compromise between the directions and have both feed spillover and unexcited aperture. We can easily calculate the step dimensions in Figure 9.2a and b. We equate the path lengths inside and outside the dielectric along the step with a difference of or some integer multiple of λ: n = +λ inside outside The step becomes =

λ n−1

(9-17)

In Figure 9-2c–f we determine the change in focal lengths instead of the step dimensions. Zoning affects the optical path lengths in the center of the lens. We calculate the edge focal length from the unzoned case. The focal length increases by Eq. (9-17)

452

LENS ANTENNAS

l/(n − 1)

(a)

l/(n − 1)

( b)

Feed Dead Zones

(c)

(d )

Aperture Dead Zones

(e)

(f )

FIGURE 9-2 Zoning of single-surface lenses.

for each inner step of hyperbolical lenses and decreases by the same amount for each inner step of elliptical lenses. We derive the lens-induced taper relative to the center by using ratios of Eq. (9-13) or (9-14). For axisymmetrical lenses, fc2 (n cos ψ − 1)3 (9-18) hyperbolical A(r, φ) f 2 (n − 1)2 (n − cos ψ) = Ac f 2 (n − cos ψ)3 2 c 2 (9-19) elliptical f (n − 1) (n cos ψ − 1) where fc is the focal length in the center, f the focal length in the feed direction ψ, and Ac the central amplitude. Example Design an axisymmetrical hyperbolical lens (n = 1.6) by using the three types of zoning shown in Figure 9-2, using an aperture diameter of 30λ and a maximum feed angle of 35◦ with a 70◦ 10-dB beamwidth feed.

ZONED LENSES

453

The minimum allowable thickness is 0.5λ with a 0.3λ edge thickness. By working through the geometry, we obtain dimensions for the following cases: Figure 9-2a, nonrefracting surface zoning, we compute steps = λ/(n − 1) = 1.67λ. 1

2

3

4

12.418

10.009

7.403

4.256

Step Aperture Radius (λ)

We estimate the feed spillover loss from Eq. (8-12) for the 10-dB feed edge taper, 0.41 dB. Equation (9-15) gives the edge taper (9.72 dB), since the refracting surface is not zoned. Combined with the feed edge taper of 10 dB, we have a 19.72-dB amplitude taper in the aperture plane. We use Eq. (4-8) to calculate the amplitude taper loss (1.8 dB) for this axisymmetric distribution. Figure 9-2c, zoning in the refracting surface of the hyperbolical lens (parallel with outgoing waves), we compute the dimensions starting with the edge focal length found from a rearrangement of Eq. (9-4) (Table 9-1). The changing focal lengths in the zones alter the aperture amplitude distribution. The edge taper becomes [Eq. (918)] 6.24 dB and reduces the amplitude taper loss [Eq. (4-8)] to 1.19 dB. The portion of the feed directed to the dead zones refracts out of the aperture and forms sidelobes that reduce the aperture efﬁciency. We consider this as a second spillover loss (0.81 dB). Figure 9-2e, zoning in the refracting surface of the hyperbolical lens (parallel with the feed wave rays), again we start with the edge focal length and increase it by λ/(n − 1) at each step and determine the feed angles where the change in the focal length will give the minimum allowable thickness. The dimensions given in Table 9-2 were obtained. Since the focal lengths are the same as for Fig. 9-2c, the lens-induced edge taper is 6.24 dB. The dead zones in the aperture distribution increase the amplitude taper loss to 3.10 dB. These ring dead zones can be considered as radiating and adding to the fully excited aperture pattern. They radiate patterns with closely spaced lobes that raise the near sidelobes of the total antenna. The three designs are compared in Table 9-3. Example Similar to the example above, we can design zoned elliptical axisymmetrical lenses that have the same problems of feed angle or aperture dead zones. The edge taper of the elliptical lens counteracts some of the feed taper and reduces amplitude taper loss. The losses calculated for those designs are given in Table 9-4.

TABLE 9-1 Zoned Hyperbolical Lens, Figure 9-2c

Zone

Focal Length (λ)

Aperture Radius of Step (λ)

Thickness (λ)

Feed Dead Zone Angles (deg)

1 2 3 4 5

20.21 18.54 16.87 15.21 13.54

0 5.12 8.42 10.84 12.89

1.52 2.09 1.98 1.90 1.83

13.57–14.02 21.64–23.10 27.06–28.68 31.28–32.95

454

LENS ANTENNAS

TABLE 9-2

Zoned Hyperbolical Lens, Figure 9-2e

Zone

Focal Length (λ)

Feed Angle (deg)

Thickness (λ)

Aperture Dead Zones (λ)

1 2 3 4 5

20.21 18.54 16.87 15.21 13.54

0 13.57 21.64 27.06 31.28

1.51 2.25 2.41 2.60 2.83

0 4.70–5.12 7.66–8.42 9.77–10.84 11.48–12.89

TABLE 9-3 Aperture Illumination Losses of Three Hyperbolical Lenses, Figure 9-2 Design (Figure 9-2)

SPL (dB)

ATL (dB)

Sum (dB)

(a) (c) (e)

0.41 1.22 0.41

1.80 1.19 3.10

2.21 2.41 3.51

TABLE 9-4 Aperture Illumination Losses of Three Elliptical Lenses, Figure 9-2 Design (Figure 9-2)

SPL (dB)

ATL (dB)

Sum (dB)

(b) (d) (f)

0.41 1.43 0.41

0.06 0.14 1.43

0.47 1.57 1.84

Zoning reduces the frequency bandwidth. At the center frequency the optical path length difference between the central ray and edge ray is K − 1 for K zones. A common maximum allowed deviation of feed to aperture path is 0.125λ, which leads to a bandwidth of 25% B (9-20) K −1 For the ﬁve zone lenses of the examples above, Eq. (9-20) gives a 6% bandwidth. We determine the loss at band edge by tracing rays from the feeds-to-aperture plane and calculating the phase error loss using Eq. (4-9). The loss at band edge is about 0.3 dB for all the designs. Bandwidth is greatly underestimated by Eq. (9-20) if a greater phase error loss is allowed. The 1-dB phase error loss bandwidth is 45%/(K − 1). 9-3 GENERAL TWO-SURFACE LENSES Optical lens designs use either ﬂat or spherical surfaces, an approximation useful for long focal lengths. We design lenses exactly. In Section 9-1 we discussed lenses where the rays refracted at only one surface. The shape curve of these two lenses could be

GENERAL TWO-SURFACE LENSES

455

derived easily. In this section we pick the curve for one lens surface and use a numerical search for the second surface that generates a table of (r, z) values. We generate a cubic spline from the table and use it to calculate the surface normal vector and the radius of curvature useful in manufacture. The initial numerical search generates ray paths through the lens that produce a table of feed angle versus aperture radius. We apply Eq. (9-11) to compute the aperture distribution after we generate the cubic spline whose evaluation includes the derivative. If we specify the surface on the feed side, we start at the feed and trace rays to the inner surface. We know the inner surface normal vector because it is a speciﬁed surface. We compute the direction of the ray inside the lens given the incident medium index of refraction ni , no inside the lens, the surface unit normal n directed into the lens, and the incident ray unit vector Si . First determine if the ray will exit the incident ray medium, because if ni > no , it can act as a prism and have total reﬂection: Ra = n2o − n2i [1 − (Si · n)2 ] If Ra < 1, the ray is totally reﬂected. For Ra > 1, we determine the direction of the output ray unit vector So from the following operations [3, p. 355]: γ =

Ra − ni (n · Si )

ﬁnd the vector t = ni Si + γ n

So =

t |t|

(9-21)

For our case the normal vector, incident, and refracted rays are in two-dimensional space (z, r), because the lens has axisymmetry. To start, specify the z-axis distance from the feed to the outer rim of the lens, the initial radius, and the rim thickness along the initial internal ray direction. The feed-side position of the lens is (z1 , r1 ). Apply Eq. (9-21) and ﬁnd the outer lens point given the rim thickness tr , by tracing along the vector t to point (z2 , r2 ). Since the lens collimates the beam, we trace the ray to a plane z = z3 whose normal is the z-axis. Figure 9-3 illustrates the rim and internal ray paths for a lens with a spherical inner surface. We calculate the electrical path length from the feed located at z = zf to the output plane by including the index

(z1, r1)

(z2, r2)

Focus

Spherical Surface

FIGURE 9-3

Ray tracing in a single-surface lens with a speciﬁed spherical inner surface.

456

LENS ANTENNAS

of refraction n of the lens for the ray path length (PL) through the lens, and the input and output ray lengths: PL =

r12 + (z1 − zf )2 + n (r2 − r1 )2 + (z2 − z1 )2 + (z3 − z2 )

(9-22)

The design consists of stepping in r1 , tracing the ray to the inner lens surface, computing the direction of the ray internal by Eq. (9-21), and determining thickness tr when Eq. (9-22) minus the initial path length is zero. By Fermat’s principle of equal path length, the outer surface refraction direction adds to the inner surface refraction to produce parallel output rays. This procedure generates a table of (z2 , r2 ) pairs that we convert to a cubic spline. By using the cubic spline, we produce an evenly spaced table of (z2 , r2 ) for machining, and if necessary, a table of radius of curvature from the second derivative to assist machining operations. The next step is to calculate a cubic spline between the aperture radius r2 and the feed angle ψ because its output includes dψ/dr2 . By rearranging Eq. (9-11), we compute the aperture power distribution given the feed power pattern: A(r2 , φ) =

F (ψ, φ) sin ψ dψ r2 dr2

(9-23)

The design steps for a lens with the outer surface speciﬁed are similar except that we trace rays from the output plane backward to the feed point. Again, we start at the lens rim, use Eq. (9-21) to calculate the internal ray direction by using the −z-directed ray and the known surface normal to travel along this ray by the rim thickness to the inner surface. Equation (9-22) gives the electrical path length for this lens, as well. We repeat the root-ﬁnding procedure used above to determine a series of points (z1 , r1 ) along the surface by equating all electrical path lengths. We generate the same series of cubic splines to obtain machining dimensions and differential dψ/dr2 used in Eq. (9-23) for the aperture power distribution. Figure 9-4 shows the aperture distributions for various lenses with focus points located 1.5 times the radius below the lens. The curves include lenses of Section 9-1. The focal spot is not a singularity as drawn on the ﬁgures using geometric optics, but spreads due to the ﬁnite wavelength. We use Gaussian beams to evaluate the size of focal spots. For a lens with a collimated output, we assume a Gaussian beam on the output with minimum waist diameter 2W0 = D, the lens diameter, radiating into free space, and a matching Gaussian beam on the feed side which tapers to the focal plane. The lens transforms one Gaussian beam into another. The focal length f = zf and we determine the diameter of the focal spot 2W0 and the half depth of focus b from the lens F -number F# = f /D [4, pp. 91–95]: 2W0 =

4λ 4λ f F# = π π D

b=

8λ 2 F π #

(9-24)

A small modiﬁcation to Eq. (9-22) allows the design of lenses with a second focus at a ﬁnite position z = z3 on the axis: PL =

r12 + (z1 − zf )2 + n (r2 − r1 )2 + (z2 − z1 )2 + r22 + (z3 − z2 )2

(9-25)

457

Aperture Distribution, dB

GENERAL TWO-SURFACE LENSES

Normalized Radius

FIGURE 9-4 Aperture distribution for various single-surface lenses with f/D = 1.5.

Focus

Focus

FIGURE 9-5 Ray tracing in lens designed for two ﬁnite focuses.

We follow the same steps as in the design above except that we need to generate a table of feed angles given the output ray angle with respect to the z-axis. Figure 9-5 illustrates a typical design and shows the ray tracing. The lens designs noted above narrow the beamwidth. We design lenses to spread the beam by using a virtual focus located behind the output surface of the lens, as shown in Figure 9-6. Because the rays trace backward to the virtual focus, we change the sign of the last term in Eq. (9-25): 2 2 2 2 PL = r1 + (z1 − zf ) + n (r2 − r1 ) + (z2 − z1 ) − r22 + (z3 − z2 )2

(9-26)

The iterative design procedure should be modiﬁed to start in the center of the lens because the concave lens has a signiﬁcant outer thickness. We compute focal length f of the lens by the distance of the two focuses from the lens: 1 1 1 + = f zf z3

(9-27)

458

LENS ANTENNAS

Virtual Focus

Focus

Focus

(a)

Virtual Focus

(b)

FIGURE 9-6 Lenses designed with virtual focuses to widen beam: (a) spherical outer surfaces; (b) spherical inner surfaces.

To calculate the pattern we trace rays through the lens either to a planar surface output side of the lens or the actual surface. By using a cubic spline between the feed angle and the aperture position we calculate the amplitude due to spreading [Eq. (9-23)]. We replace the ﬁelds with currents and use physical optics to calculate the pattern. A second simpler approach uses a Gaussian beam approximation for lenses that accounts for electrical size. Both the input and output Gaussian beam have the same waist at the lens plane. Each Gaussian beam decreases in a hyperbola to the minimum waist at the location of the phase center or focus. The output does not pass through a narrow caustic, as shown in Figure 9-5, but reaches a ﬁnite-diameter waist related to the feed beamwidth and the lens diameter. At this point we design the lenses scaled to wavelengths. Given the feed beamwidth, we calculate the half depth of focus b by using Eq. (7-35). The lens magniﬁes the output beam waist compared to the input beam waist by M. First, we calculate an input magniﬁcation factor Mr : f (9-28) Mr = zf − f With b given in wavelengths, we compute the ratio of b to the shift of the feed relative to the focus: b r= (9-29) zf − f We combine Eqs. (9-28) and (9-29) to calculate the lens magniﬁcation M: M=√

Mr 1 + r2

(9-30)

SINGLE-SURFACE OR CONTACT LENSES

459

The output Gaussian beam propagates from the second focus with new parameters: b , the half depth of focus, waist W0 , and divergence angle θ0 [4]: b = M 2 b

W0 = MW0

θ0 =

θ0 M

(9-31)

Example The lens in Figure 9-5 was designed for a diameter of 20λ, zf = 15λ, and z3 = 25λ. We calculate the focal length from Eq. (9-27) as f = 9.375λ. A feed with a 68◦ 10-dB beamwidth is approximated by a Gaussian beam with b = 0.99λ (Scale 7-8) and 2W0 = 1.12λ (Scale 7-9). We use b to ﬁnd the magniﬁcation M by using Eqs. (9-28) to (9-30), M = 1.64. The Gaussian beam at the second focus has a minimum waist 2W0 = 1.64(1.12) = 1.84λ. Using Scales 7-7 to 7-9, we read the Gaussian beam output values: 10-dB beamwidth = 42◦ (gain = 18.4 dB). Scale 7-7 gives the gain of the feed as 14.3 dB. The half depth of focus increased from 0.99λ to 3.39λ. Example When we repeat the lens calculations for the lens of Figure 9-6a for a diameter = 20λ, zf = 15λ, and z3 = −5λ, we calculate f = −7.5. Starting with a feed 10-dB beamwidth = 80◦ , we use Scale 7-8 to ﬁnd b = 0.7λ and waist diameter 2W0 = 0.94λ. The negative focal length produces a magniﬁcation below 1: M = 0.333. This decreases b to 0.0775λ and 2W0 = 0.314. The output Gaussian beam gain drops to 6.6 dB from a feed gain of 12.9 dB.

9-4 SINGLE-SURFACE OR CONTACT LENSES We can alter the pattern of antennas with planar surfaces such as spirals and microstrip patches by placing a lens directly on the planar surface. The lens can be spaced a small distance away to avoid potential damage and have little pattern impact. The lens modiﬁes the original pattern of the antenna by using the refraction at the single output surface. Because the lens contacts the antenna, the pattern inside the lens is the same as radiated into free space except for the dielectric loading on the antenna. This loading shifts the operating frequency and in the case of a spiral improves its efﬁciency and widens its beamwidth (see Section 11-5.1). We design these antennas by ﬁrst generating a mapping between the feed angle ψ and the output angle θ . This could be as simple as θ (ψ) = constant or a function to generate a shaped beam, in the same manner as a shaped reﬂector (see Section 8-20). The relationship θ (ψ) enables calculation of the surface normal at every point along the outer surface. The surface normal is found from the gradient of the radial vector, and equating the two values produces a differential equation between the radius r and the feed angle ψ. Given the feed angle and the output angle at a point on the lens, we calculate the surface normal from n = nSi − So , where n is the index of refraction, Si the incident unit vector, and So the exiting ray unit vector. We normalize n to a vector v. The gradient of r(ψ) gives a second expression for the normal vector: ∇r(ψ) = ar + aψ

1 ∂r(ψ) r(ψ) ∂ψ

(9-32)

The coefﬁcient of aψ in Eq. (9-32) is the tangent of the angle α between the normal vector and the radial vector (incident ray). The tangent can be found from the unit

460

LENS ANTENNAS

vector v and the incident ray unit vector Si : tan α =

1 ∂r(ψ) (Si × v) = Si · v r(ψ) ∂ψ

(9-33)

The cross product between the incident ray and the unit vector v only has a z-axis component, because the incident ray and the surface normal lie in the x –y plane. We design the lens surface by solving the differential equation (9-33). A numerical technique such as the Runge–Kutta method easily solves the equation when we start at one feed angle and an arbitrary lens radius and step through feed angles. The method determines only the shape of the lens to an arbitrary size that we scale to the diameter desired. Figure 11-9a illustrates a lens designed to redirect all feed rays to θ = 0. Figure 9-7 illustrates the shape of a contact designed to redirect all rays between feed angles 0 through 60◦ to output rays at 30◦ . For a 3λ lens diameter and a feed 12-dB beamwidth of 120◦ , the lens spreads the beam to form a ﬂat-topped output beam shown in Figure 9-8. Contact lenses can greatly modify radiation from a feed with electrically small lenses. Exit Rays

Incident Rays

Feed

FIGURE 9-7 Contact lens designed to direct the beam in a cone at 30◦ .

FIGURE 9-8 Pattern of contact 3λ-diameter lens for a feed with a 12-dB beamwidth of 120◦ .

METAL PLATE LENSES

461

9-5 METAL PLATE LENSES The phase velocity of a wave in waveguide exceeds that of a wave in free space and produces a medium with an effective refractive index of less than 1. We can make a microwave lens by spacing parallel metal plates and feeding the lens with a wave polarized in the direction of the plates. For plates spaced a distance a, the index of refraction is 2 λ (9-34) n= 1− 2a where λ is the wavelength in the medium between the plates. The index of refraction is frequency dependent. The lens can be made polarization independent by forming an egg crate of orthogonal plates. We divide an arbitrary polarization into orthogonal polarizations normal to each set of plates. If we substitute n from Eq. (9-34) into Eq. (9-1), we obtain the equation of a front single-surface lens: (1 − n)f (9-35) ρ(ψ) = 1 − n cos ψ Equation (9-35) is an ellipse with f as the distance from the far focus of the ellipse to the center of the lens front surface. This surface refracts waves parallel with the axis and determines the second surface: a plane. The parallel plates constrain the waves parallel with the axis and prevent the design of an outer single-surface lens. The cutoff wavelength 2a and the possibility of higher-order modes restrain the range of n. The second-order-mode cutoff occurs when λ = a, and it limits n to 0.866 [Eq. (9-34)]. At cutoff, λ = 2a and n equals zero. Reasonable values lie between 0.3 and 0.7. The variation of n versus frequency limits bandwidth. When the phase variation in the aperture is limited to λ/8, the bandwidth is approximately [1]: bandwidth(%) =

λ 25n 1 − n (1 − n)t

(9-36)

where n is the center-frequency index of refraction and t is the maximum thickness. An acceptable bandwidth is underestimated by Eq. (9-36) through restriction of the band edge phase error. The elliptical surface increases the aperture distribution toward the edges. When we substitute Eq. (9-35) into Eqs. (9-11) and (9-12), we obtain the aperture amplitude distribution relative to the feed pattern: (1 − n cos ψ)3 A(r, φ) = 2 F (ψ, φ) f (1 − n)2 (cos ψ − n)

axisymmetrical

A(r, y) (1 − n cos ψ)2 = F (ψ, y) f (1 − n)(cos ψ − n)

cylindrical

(9-37a, b)

Example Design an axisymmetric parallel-plate lens with a diameter of 30λ, maximum feed angle of 35◦ , n = 0.625, and minimum thickness of λ. ◦

ρ(35 ) =

30λ = 26.15λ 2 sin 35◦

462

LENS ANTENNAS

Rearrange Eq. (9-35) to calculate the focal length at the edge: f =

(1 − n cos 35◦ )ρ(35◦ ) = 34.03λ 1−n

Equation (9-34) gives us the plate separation with a slight rearrangement: 1 = 0.64λ a= √ 2 1 − n2 The amplitude variation from the center to the edge caused by the ellipse is given by Eq. (9-37a): A(ψe ) (1 − n cos ψe )3 = = 4.26 (6.3 dB) A(0) (1 − n)2 (cos ψe − n) A feed with its 10-dB beamwidth equal to the subtended angle of the lens at the feed produces −3.7 dB edge taper in the aperture. The edge thickness is given by ◦

◦

t = f + 1 − ρ(35 ) cos(35 ) = 13.61λ Equation (9-36) predicts a bandwidth of 1.9%. A detailed analysis using ray tracing through the lens and varying n with changes in frequency predicts a 0.2-dB loss at this band edge. The 1-dB bandwidth is about 4.5%. Zoning a parallel-plate lens increases its bandwidth by limiting the maximum thickness, since the variation of the optical path length due to the varying n exceeds that due to the zoning. Figure 9-9 gives the central curve of the three possible types of zoning. The lens in Figure 9-9a only suffers loss due to diffractions from edges. The other two zoned lenses (Figure 9-9b, c) have dead zones in the aperture. These dead zones produce additional amplitude taper loss and high close-in sidelobes. The feed-side zoning has different focal lengths in each zone. The outer zone remains the same as the unzoned lens. The focal lengths of the inner zones are reduced by λ/(1 − n) at each step and the stepping reduces the amplitude taper of the lens by varying f : f 2 (1 − n cos ψ)3 A(ψ) = 2 c (9-38) A(0) f (1 − n)2 (cos ψ − n)

r(y) y Feed Elliptical

(a)

(b)

(c)

FIGURE 9-9 Central section of zoned parallel metal plate lenses.

SURFACE MISMATCH AND DIELECTRIC LOSSES

463

where fc is the central focal length and f is the focal length of the ellipse at ψ. The zoned lens bandwidth is approximately [1] BW =

25% K − 1 + [(1 + n)(1 − n)t/n]

(9-39)

where K is the number of zones and t is the maximum thickness. Equation (9-39) also is an underestimation of an acceptable loss-level bandwidth. Example The lens of the example above was zoned as in Figure 9-9a and c with ﬁve zones. The maximum thickness is 3.4λ. Equation (9-39) predicts a bandwidth of 3.4%. By tracing rays through the lens and applying Eq. (4-9) to calculate phase error loss, we predict a 10% bandwidth for a 1-dB loss. Zoning the nonrefracting surface has no effect on the aperture distribution except for edge diffractions that modify the ﬁelds slightly. Zoning the refracting surfaces causes aperture dead zones and reduces the lens-induced amplitude taper. The focal length of the ellipse at the edge remains at 34.03λ. The focal length of the center ellipse is reduced by 4λ/(1 − n) from the edge ellipse to 23.36λ. The edge taper [Eq. (9.38)] becomes 2.01 (3 dB). The aperture dead zones increase the loss by 2 dB. The bandwidth of a parallel-plate lens can be increased by a method of compounding lenses into a doublet [5]. We can make a lens by using a uniform waveguide length between the input and output surfaces and placing a phase shifter in each line to compensate for the optical path-length differences and produce an eikonal at the aperture plane. If we combine a refracting surface waveguide plate lens with a differential phase shift lens, we can match the aperture phase at two frequencies. This matching broadbands the antenna like an optical achromatic doublet. 9-6 SURFACE MISMATCH AND DIELECTRIC LOSSES The reﬂection coefﬁcient of a wave normally incident on a dielectric is =

1−n 1+n

(9-40)

valid for both dielectric and metal lenses. The actual reﬂection coefﬁcient of any ray depends on the angle of incidence and the polarization. Both surfaces of the lens have reﬂections, and these interact to produce the actual reﬂection. With plane surfaces, such as those assumed for radomes, we can analyze the combination of reﬂections using transmission-line mismatch equivalence. Since the reﬂections from second surfaces may not return to the same point as the incident waves and may have their caustic distances changed by the curved surfaces, transmission-line models of the two surfaces have limited use for lenses. Equation (9-40) gives us a reasonable approximation to expected mismatch, since one surface of the single refracting surface lens will be normal to the wave incident from the feed and reﬂect into the feed. The second surface fails to focus power back to the feed and has a minor effect. Example A lens with n = 1.6 or n = 0.625 has a reﬂection coefﬁcient magnitude of 0.23 [Eq. (9-40)] at one surface that focuses into the feed. This gives a feed mismatch if VSWR =

1 + 0.23 = 1.6 1 − 0.23

464

LENS ANTENNAS

the same as n (n > 1) or 1/n (n < 1). The mismatch loss becomes 1 − ||2 = 0.95 (0.2 dB). The wave may be matched to the surface by a quarter-wavelength transformer with an index of refraction n1/2 , but adding matching transformers narrows the bandwidth. The surface, which reﬂects power back to the feed, should be matched ﬁrst, since the second surface has a minor effect on feed mismatch. Also, the primary reﬂecting surface has normally incident waves and does not suffer from the need to vary the thickness to match waves off normal incidence. Transformers to match waves off normal incidence are polarization-sensitive. Simple methods can be used to reduce lens-caused feed mismatch [2]. The lens can be tilted to cause the reﬂection to miss the feed. Offsetting half the lens by λ/4 causes cancellation of the reﬂection from the two halves. Tilting does not reduce the mismatch loss but does produce backlobe power in the pattern. Similarly, the reﬂected power from the hyperbolical surface forms backlobes. These reﬂections reduce the antenna efﬁciency below that predicted by aperture theory alone. Cohn and Morita [2,6] developed methods of matching the surface of the lenses by removing some of the dielectric for a quarter wavelength. The surfaces are either corrugated, have arrays of holes, or have arrays of rods. With this method, the lens can be made from a single dielectric slab. The design depends on the angle of incidence and the polarization of the waves. The lens dissipates power by the attenuation constant of the material: 27.3n tan δ α= dB/length (9-41) λ where tan δ is the loss tangent of the dielectric. Waveguide losses reduce the power transmitted through metal plate lenses. Zoning eliminates material and its associated loss to improve efﬁciency, but for most materials, this effect is small. Artiﬁcial dielectrics [2] reduce excessive weight and material losses of lenses. We make them by embedding metal particles or plated microspheres in foam with a dielectric constant near 1. The metal parts may be strips or disks made from metal foil. Similarly, solid metal parts can be hollow. Since the effective dielectric constant depends on the size of the metal particles in wavelengths, lenses made from artiﬁcial dielectrics will be narrowband if the particles are large, but the use of plated microspheres dispersed in the foam reduces this problem.

9-7 FEED SCANNING OF A HYPERBOLOIDAL LENS [7] The hyperboloidal lens has no cross-polarization when fed from an electric dipole source. Kreutel [7] analyzed the effects of off-axis dipole sources on the pattern of the hyperboloidal lens. The coma increases more rapidly for the lens than for a paraboloidal reﬂector for the same scanning. Like the paraboloidal reﬂector, the hyperboloidal lens beam scans less than the deviation angle of the feed relative to the vertex and axis and has a beam deviation factor (Table 9-5). The scanning loss (Table 9-6) decreases with increasing n. The peak coma lobe (Table 9-7) limits the possible scan before unusable patterns are obtained. The paraboloidal reﬂector can be scanned further (Table 8-2) for the same coma.

DUAL-SURFACE LENSES

TABLE 9-5 f/D 0.8 1.0 1.2

Beam Deviation Factor for a Feed-Scanned Hyperboloidal Lens √ √ √ n= 2 n=2 f/D n= 2 n=2 f/D n= 2

465

0.75 0.80 0.83

0.84 0.87 0.89

1.4 1.6 1.8

0.86 0.89 0.92

0.91 0.92 0.94

2.0 2.5 3.0

0.93 0.95 0.97

n=2 0.95 0.96 0.98

TABLE 9-6 Scanning Loss for a Hyperboloidal Lens (dB) √ n=2 n= 2 Beamwidth of Scan f/D = 1 f/D = 2 f/D = 1 f/D = 2 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5

0.03 0.06 0.12 0.23 0.36 0.51 0.69 0.90 1.09

0.00 0.01 0.03 0.05 0.09 0.13 0.19 0.24 0.31 0.38 0.44

0.01 0.04 0.07 0.12 0.20 0.28 0.37 0.49 0.60 0.75

0.00 0.01 0.02 0.04 0.06 0.08 0.11 0.14 0.18 0.22 0.25

TABLE 9-7 Coma Sidelobe Level for a Scanned Hyperboloidal Lens (dB) √ n=2 n= 2 Beamwidth of Scan f/D = 1 f/D = 2 f/D = 1 0 1 2 3 4 5 6

20.9 17.6 15.2 13.1 11.3 9.8 8.8

18.5 17.5 16.5 15.5 14.8 14.0 13.3

19.5 17.6 15.7 14.2 12.9 11.5 10.4

9-8 DUAL-SURFACE LENSES The second surface of the lens offers an additional degree of freedom that can be used to control the pattern characteristics. Ruze [8] developed methods to reduce coma for feed-scanned cylindrically shaped metal plate lenses, which constrain the wave parallel with the axis. Both surfaces are used to satisfy focusing requirements. We will develop a method for axisymmetrical dielectric lenses to eliminate coma for small feed

466

LENS ANTENNAS

displacements. In a second design we can also use the second surface shape to control the amplitude distribution in the aperture plane. 9-8.1 Coma-Free Axisymmetric Dielectric Lens [9] The design of the coma-free axisymmetric antenna reduces to the numerical solution of a differential equation with side conditions to produce a collimated beam and satisfy the Abbe sine condition [10, p. 157]. Successful designs require a number of iterations, since the ultimate lens shape depends heavily on the initial conditions. Solution of the differential equation will sometimes diverge into unrealizable designs or fail to continue to satisfy the side conditions. A lens satisfying the Abbe sine condition is free of coma aberrations for small deviations of the feed from the axis. The deviations produce higher-order aberrations that eventually distort the beam with continued scanning, but coma is removed. For a lens focused at inﬁnity, the Abbe sine condition requires that the surface which refracts the waves parallel with the axis must be spherical with its center on the effective focus of the lens. The dielectric lens refracts waves parallel with the axis on the outer surface (away from the feed). Given the aperture radial component r, r = fe sin ψ

(9-42)

where fe is the effective focus and ψ is the feed angle. A waveguide plate lens satisﬁes the Abbe sine condition by having a spherical inner surface [6], since the waves are parallel with the axis in the lens because the waveguide plates constrain the wave to be parallel with the axis. The second surface must produce the conditions for a uniform phase in the aperture plane. The waveguide plate lens only has to equalize path lengths. In a dielectric lens, the inner surface must refract the waves in the proper direction to satisfy the Abbe sine condition, and it must be so placed as to equalize the path lengths from the feed to the aperture plane. The locations of both surfaces along the axis are varied to equalize the path lengths. Figure 9-10 shows the coordinates of the coma-free dielectric lens. The polar equation ρ (ψ) describes the inner surface and ψ is the angle of the refracted wave with the axis. The distance from the feed to the center of the lens inner surface is f , and T is the thickness. The coordinates (r, z) describe the outer lens surface, where z is the axis dimension and r is the aperture radial component. Snell’s law reduces to a differential equation at the inner surface: dρ n sin(ψ − ψ )ρ = dψ n cos(ψ − ψ ) − 1 where

(9-43)

r − ρ sin ψ z − ρ cos ψ

(9-44)

(fe − ρ) sin ψ z − ρ cos ψ

(9-45)

tan ψ = By use of Eq. (9-42), this reduces to tan ψ =

DUAL-SURFACE LENSES

467

(r,z) y′ r(y) D

y f

T

Feed

FIGURE 9-10 Coma-corrected dual-surface axisymmetric lens; n = 1.6, D = 35, f = 45, T = 6.5, fe = 49.

The requirement for equal optical path lengths to the aperture determines a quadratic equation in z: Az2 + Bz + C = 0 (9-46) where A = n2 − 1 B = 2(ρ − K) − 2n2 ρ cos ψ C = n2 ρ 2 cos2 ψ + n2 (fe − ρ)2 sin2 ψ − (ρ − K)2 K = T (n − 1) √ −B + B 2 − 4AC z= 2A Design consists of the numerical solution of the differential equation (9-43) subject to the conditions of Eqs. (9-45) and (9-46). Realizable solutions depend on the initial conditions. Most failures to produce a usable design occur in Eq. (9-46), which satisﬁes the requirement of equal aperture phase. Example Figure 9-10 shows a scale drawing of a realizable design for n = 1.6, focal distance f = 45, diameter D = 35, center thickness T = 6.5, and effective focal length fe = 49. Table 9-8 lists a few points of the solution obtained by a Runge–Kutta numerical method for the differential equation (9-43). The example above contains only relative dimensions. The solution is size and frequency independent, since it is obtained by geometric optics. We can zone the lens for a given frequency by cutting along ray paths. Each step is λ/(n − 1). Table 9-8, as in the example above, determines the ray paths through the lens. Zoning will produce either feed or aperture dead zones. The reduction of weight must be balanced with

468

LENS ANTENNAS

TABLE 9-8 Design of Figure 9-11 for a Coma-Free Lens Feed Angle, ψ (deg)

Inner Surface, ρ(ψ)

Horizontal Distance, z

Radius, r

Thickness Along Ray, T

0 5 10 15 20 20.92

45.00 45.18 45.70 46.59 47.63 47.61

51.50 51.19 50.27 48.71 46.28 45.40

0 4.27 8.51 12.68 16.76 17.50

6.50 6.20 5.29 3.77 1.59 1.06

the loss in efﬁciency to achieve some compromise. Because we used the degrees of freedom of the second surface to satisfy the Abbe sine condition, we lose control of the aperture distribution through the lens surfaces. Most practical designs produce an amplitude taper near that of the feed antenna. We must achieve low sidelobes, if required, through a tapered illumination from the feed. The feed pattern plays no part in the design and gives us degrees of freedom for amplitude taper. An antenna designed and built with a diameter of 32 wavelengths [9] showed no coma in a scanning of ±2 beamwidths. Like the paraboloidal reﬂectors, increasing the focal length for a given diameter allows greater scanning without signiﬁcant coma. 9-8.2 Speciﬁed Aperture Distribution Axisymmetric Dielectric Lens [11] We use the desired aperture amplitude distribution to specify the relation between the aperture radius r, and the feed angle ψ. Earlier, the Abbe sine condition established this relation. Given a feed power pattern F (ψ) and a required aperture distribution A(r), we relate the two through differential areas: F (ψ) sin ψ dψ = A(r)r dr where an axisymmetrical pattern is assumed. We derive the relation between ψ and r through normalized integrals as in Section 8-20:

ψ

0 ψm 0

F (ψ) sin ψ dψ = F (ψ) sin ψ dψ

r

0rm

A(r)r dr (9-47) A(r)r dr

0

In any particular design we generate a table, such as Table 8-12, of the feed angle versus its normalized feed pattern integral and the aperture radius versus its normalized aperture distribution integral. For a given feed angle ψ we equate the normalized integrals to compute the corresponding aperture radius. We generate a table of aperture radius versus feed angle, such as Table 8-13, using interpolation techniques. The design is very dependent on feed pattern, because changing the feed pattern alters the table. Low sidelobe aperture distributions require close tolerances and a good speciﬁcation of the feed pattern. Once we have the relation between ψ and r, the design follows steps similar to those taken in designing the lens for the Abbe sine condition. We solve the differential

DUAL-SURFACE LENSES

469

equation (9-43) numerically. We specify the aperture radius by the table generated for the transformation of the feed pattern into the aperture distribution. The requirement for equal path length through the lens determines the axis location z of the outer surface. Ax 2 + Bx + C = 0

x =z−f

A=n −1 2

B = 2[n2 (f − ρ cos ψ)] + ρ − K − f

(9-48)

C = [n(ρ cos ψ − f )]2 + (r − ρ sin ψ)2 − (f + K − ρ)2 √ −B + B 2 − 4AC K = T (n − 1) z=f + 4A where T is the central thickness and f is the axis focal length. A successful design requires a number of iterations, starting with different initial conditions. The differential equation solution will diverge to shapes unable to satisfy the equal-path-length side requirement with poor initial conditions. Each design has a narrow range of satisfactory initial conditions. In most cases, increasing the thickness increases the chance for a successful design. Example A lens was designed to transform the feed pattern of a conical corrugated horn into a circular Taylor distribution with 40-dB sidelobes (n = 8). The initial conditions were n = 1.6, diameter D = 32, focal distance f = 35, central thickness T = 9, and maximum feed angle ψm = 20◦ . A table of the feed angle and its normalized power pattern integral, along with the aperture radius and its normalized power distribution integral, was generated. A table of feed angle and the corresponding aperture radius follows from equating normalized integrals. This table is independent of the lens thickness but not the feed pattern. A Runge–Kutta numerical method is used to solve the differential equation (9-43) subject to the conditions imposed by the aperture radius table and Eq. (9-48) for equal optical path lengths. Figure 9-11 shows a design for a 36◦ 10-dB beamwidth feed horn (12-dB feed edge taper). The horn dimensions are aperture radius = 1.90λ and slant radius = 9λ for a maximum quadratic phase deviation S = 0.2. A few points of the design are listed in Table 9-9. Other antennas designed with small changes in the feed pattern beamwidth show signiﬁcant changes in the lens shape near the edges for a constant center thickness. Axisymmetric dielectric lenses can be designed to be independent of frequency because only relative sizes are speciﬁed. The lenses tend to be thick to allow room to satisfy the requirement for an equal optical path length. Zoning lens reduces weight while decreasing bandwidth. Low-sidelobe designs are inherently narrowband, since small changes in the beamwidth of the feed alter the aperture distribution and sidelobe levels. Zoning may not reduce the bandwidth significantly. An antenna designed and tested using the technique above revealed a number of requirements on the design [12]. The sidelobe levels exceeded the design speciﬁcation for three main reasons. First, the feed pattern was speciﬁed as sin(πU )/πU , an oversimpliﬁcation of the actual feed pattern. Realistic feed patterns must be used because small changes in the feed pattern require new designs. Second, the surfaces must be matched with quarter-wavelength sections to prevent reﬂections, unaccounted

470

LENS ANTENNAS

(r,z) y′ r

r(y) Feed

D

y ym

T

f

FIGURE 9-11 Dual-surface axisymmetric lens for circular Taylor aperture distribution (40-dB, nˆ = 8). Lens: n = 1.6, D = 32, f = 35, T = 9, ψm = 20◦ . Feed: conical corrugated horn, 36◦ 10-dB beamwidth. S = 0.2.

TABLE 9-9 Design of Figure 9-12 for a Speciﬁed Aperture Distribution Axisymmetric Dielectric Lens Feed Angle, ψ (deg)

Inner Surface, ρ(ψ)

Horizontal Distance, z

Radius, r

Thickness Along Ray, T

0 5 10 15 20

35.00 35.31 36.14 37.69 38.71

44.00 43.95 43.67 43.25 40.58

0 3.27 6.59 10.11 16.00

9.00 8.78 8.08 6.86 5.03

for in the design, that change the aperture distribution. Third, diffractions from the edges affect the distribution. Increasing the aperture diameter or using a low edge illumination feed reduces these effects. As designed, the lens exhibits severe coma when scanned by feed lateral offset. Since most lenses are quite thick, zoning can be used to approximate the Abbe sine condition on the inner surface. The lens refracts most of the rays parallel with the axis by the inner surface. If the zones approximate a spherical surface on the average, coma is reduced for the scanned beams. These coma-corrected lenses are useful for multibeam applications when each beam is fed from an offset feed.

9-9 BOOTLACE LENS The bootlace lens consists of a set of receiving antennas on a surface connected by cables to a set of transmitting antennas on the second surface (Figure 9-12). The cables constrain the path through the lens. We have three degrees of freedom with this lens: (1) input surface, (2) output surface, and (3) cable length. We can change the lens characteristics dynamically by placing phase shifters and/or attenuators in the lines

BOOTLACE LENS

471

y

Radiators Radiators F1 F2 x F2 F1

Focal Arc

C1

C2 Flexible Cables

FIGURE 9-12 Bootlace lens. (From [13], Fig. 3, 1965 IEEE.)

between the input and output radiators and scan one or more beams. The input and output surfaces are arrays, and we can generate multiple beams by placing more than one feed on a focal arc determined by the lens geometry. The simplest bootlace lens consists of a spherical input surface connected to a plane output surface by equal-length cables. This lens converts spherical waves radiated by the feed into plane waves at the output surface. The lens uses true time delay, which removes bandwidth limitations. Most bootlace lenses are line sources or twodimensional lenses fed by line sources. The general lens can have four focal points [13] placed symmetrically about the axis of the symmetric structure. The focal arc is chosen on a curve through the focal points to minimize defocusing when feeds are placed off the focal points. A feed at each point along the focal arc produces an output beam in a different direction. Because each feed uses the full aperture, it achieves the full array gain less the loss of projection of the aperture length in the beam direction. The number of focal points is reduced to three when the lens is further restrained. The Ruze design for a metal plate lens [8] has three focal points, since the waveguides between the surfaces travel in straight lines. There is one central axis focal point and two symmetrically placed focal points. The Rotman [14] lens loses one possible focal point because the output surface is limited to a straight line. A parallel-plate structure in the Rotman lens leads from the possible feed locations to the feed-side surface, which normally is excited by probes in the parallel-plate guide. The lens becomes a feed network that produces multiple beams whose directions depend on the location of the feed on the focal arc. Although perfect focusing is achieved at only three points, the phase error loss associated with points between them is small. Because the Rotman lens feed network is a true time-delay array feed network, we can achieve bandwidths greater than an octave from it. Rao [15] extends the design of bootlace lenses to three dimensions and shows that the number of focal points cannot be extended beyond four. Because the lens is not

472

LENS ANTENNAS

axisymmetric, it has different scanning capabilities in orthogonal planes. Rao designs lenses with two, three, and four focal points on a focal line. Decreasing the number of focal points in one plane increases the scanning capability in the orthogonal plane for a given phase error level.

9-10

LUNEBURG LENS [16, p. 545]

A Luneburg lens, a spherically symmetric lens with a variable index of refraction, radiates a beam in any direction for a feed located opposite the beam. We place the feed phase center either on the surface of the lens or a short distance away. We form multiple beams by feeding the lens at a number of places. Our only restriction is the blockage due to other feeds or support structures. We can rapidly scan a beam by moving a lightweight feed around the sphere or by switching between multiple feeds. When we place the feed on the outer surface of the sphere, the required index of refraction is

r 2 n= 2− (9-49) a where a is the lens outer radius and r is the inner radius. The dielectric constant n2 must vary between 2 in the center and 1 on the outer surface. Few feeds have their phase centers on a surface that can be mounted against a sphere. We can move the feed away from the surface by changing the variation of the index of refraction from that given by Eq. (9-49), but the required center index of refraction decreases as we move the feed away from the lens surface. We calculate the variation of the index of refraction from an integral equation, and the curves follow the general shape of Eq. (9-49). For the feed-to-sphere radius of 1.1, the proper center dielectric constant is 1.83 and it varies smoothly to 1 at the lens surface. Similarly, the center dielectric constant starts at 1.68 for the feed-to-sphere radius of 1.2. The lens changes the amplitude distribution in the aperture compared with the feed. Given the ratio of feed radius to lens radius ri , the aperture plane power distribution becomes F (ψ) A(r) = 2 (9-50) r1 cos ψ where ψ is the feed angle, F (ψ) the feed power pattern, and A(r) the aperture power distribution. Equation (9-50) shows that the lens refracts power toward the edge of the aperture. Lenses have been made by using a series of concentric spherical shells each with a constant dielectric. A minimum of 10 shells is needed for an adequate approximation of the required variation of the dielectric constant. REFERENCES 1. J. R. Risser, Chapter 11 in S. Silver, ed., Microwave Antenna Theory and Design, McGrawHill, New York, 1948. 2. S. B. Cohn, Chapter 14 in H. Jasik, ed., Antenna Engineering Handbook, McGraw-Hill, New York, 1961. 3. S. Cornbleet, Microwave Optics, Academic Press, London, 1976.

REFERENCES

473

4. B. E. A. Saleh and M. C. Teich, Fundamentals of Photonics, Wiley, New York, 1991. 5. A. R. Dion, A broadband compound waveguide lens, IEEE Transactions on Antennas and Propagation, vol. AP-26, no. 5, September 1978, pp. 751–755. 6. T. Morita and S. B. Cohn, Microwave lens matching by simulated quarter-wave transformers, IEEE Transactions on Antennas and Propagation, vol. AP-4, no. 1, January 1956, pp. 33–39. 7. R. W. Kreutel, The hyperboloidal lens with laterally displaced dipole feed, IEEE Transactions on Antennas and Propagation, vol. AP-28, no. 4, July 1980, pp. 443–450. 8. J. Ruze, Wide-angle metal-plate optics, Proceedings of IRE, vol. 38, no. 1, January 1950, pp. 53–59. 9. J. J. Lee, Numerical methods make lens antennas practical, Microwaves, vol. 21, no. 9, September 1982, pp. 81–84. 10. R. Kingslake, Lens Design Fundamentals, Academic Press, New York, 1978. 11. J. J. Lee, Dielectric lens shaping and coma-correcting zoning, part I: analysis, IEEE Transactions on Antennas and Propagation, vol. AP-31, no. 1, January 1983, pp. 211–216. 12. J. J. Lee and R. L. Carlise, A coma-corrected multibeam shaped lens antenna, part II: experiments, IEEE Transactions on Antennas and Propagation vol. AP-31, no. 1, January 1983, pp. 216–220. 13. M. L. Kales and R. M. Brown, Design considerations for two dimensional symmetric bootlace lenses, IEEE Transactions on Antennas and Propagation, vol. AP-13, no. 4, July 1965, pp. 521–528. 14. W. Rotman and R. F. Turner, Wide-angle microwave lens for line source applications, IEEE Transactions on Antennas and Propagation, vol. AP-11, no. 6, November 1963, pp. 623–632. 15. J. B. L. Rao, Multifocal three-dimensional bootlace lenses, IEEE Transactions on Antennas and Propagation, vol. AP-30, no. 6, November 1982, pp. 1050–1056. 16. R. S. Elliott, Antenna Theory and Design, Prentice-Hall, Englewood Cliffs, NJ, 1981.

10 TRAVELING-WAVE ANTENNAS

Traveling-wave antennas consist of transmission-line structures that radiate. We develop a uniﬁed theory for end-ﬁre line antennas because length and propagation constant along the structure determine most of their properties. To ﬁrst order, length determines gain and bandwidth. The size and shape of the structure produce secondary effects such as polarization nulls and narrower beamwidths. Most of these structures are slow wave-transmission structures that bind waves to it and radiate at discontinuities. We use surface-wave structures to radiate end-ﬁre beams and leaky wave structures to radiate beams at an angle to the axis of the line source. In both cases there are planar conﬁgurations that have their uses, but in this chapter we concentrate on long, thin geometries. We combine leaky wave line-source radiators, such as slotted rectangular waveguides, into planar arrays, but the line source remains the building block. We make traveling-wave antennas from structures that guide waves. Surface-wave structures bind the power to the transmission line and radiate from discontinuities such as bends or dimensional changes. In some cases we analyze the surface wave as radiating throughout its extent on the transmission line. Both methods provide insight. Leaky wave antennas carry waves internally, such as a waveguide, and radiate at openings that allow power to escape. The radiation mechanism differs in the two cases, but we use similar mathematics to describe both types. We may have trouble distinguishing the radiation mode because the structures may be similar because with small changes in structure, some antennas can radiate in either mode. We separate traveling-wave antennas from other antennas by the presence of a wave traveling along the structure, with most of its power propagating in a single direction. We divide antennas by their structure: line and planar. We usually analyze planar structures as being inﬁnite in the direction normal to the wave propagation. Similarly, we usually ignore the diameter of line sources in a ﬁrst-order analysis. The diameter is important for determining the mode structure, but to ﬁrst order we calculate patterns Modern Antenna Design, Second Edition, By Thomas A. Milligan Copyright 2005 John Wiley & Sons, Inc.

474

GENERAL TRAVELING WAVES

475

based on a thin line source since length and propagation constant determine the pattern and bandwidth. The width of a planar structure determines the pattern beamwidth in that plane. Increasing the diameters of the rods of line sources will decrease the pattern beamwidth and increase gain, but the effect is secondary. Only when we include the diameter can we make the transition to aperture-type structures considered to be radiating from the end. In this chapter we must consider unusual transmission-line structures. Properly designed dimensions provide the proper phase velocity to establish a single end-ﬁre beam for slow-wave antennas or to point the beam of a leaky wave antenna. We calculate some of the dimensions by analysis (an ever-expanding list), but we can also measure the velocities and leakage and proceed to design. 10-1 GENERAL TRAVELING WAVES A wave traveling in a single direction has a ﬁeld representation: E = E0 (z)e−kP z

(10-1)

where z is the direction of propagation, k the free-space propagation constant (wave number) 2π/λ, and P the relative propagation constant. E0 (z) describes the amplitude variation: P > 1 surface waves P < 1 leaky waves

(10-2)

For a planar structure in the y –z plane, we consider separable distributions: E = E0 (z)E1 (y)e−j kP z We compute the pattern from L f = 0

a −a

E0 (z)E1 (y)e−j kP z ej kz z ej ky y dz dy

(10-3)

where kz = k cos θ . Similarly, for circular distributions we have E = E0 (z)E1 (φ)e−j kP z and

L

f =

E0 (z)e

−j kP z j kz z

e

2π

dz

0

E1 (φc )aej ka sin θ cos(φ−φc ) dφc

(10-4)

0

where a is the radius. The second integral includes vector dot products to project the ring aperture ﬁelds on to the far-ﬁeld polarizations (see Section 7-2). We consider only the term along the z-axis, and we can consider the effect of the other coordinate separately. The pattern response is L f = E0 (z)e−j kz (P −cos θ) dz (10-5) 0

476

TRAVELING-WAVE ANTENNAS

We use the results of Chapter 4 with these separable distributions. Maximum gain comes from a uniform distribution reduced by the amplitude taper efﬁciency for tapered distributions. In Eq. (10-3) the y-axis distribution and size determine the gain factor as a product for aperture area. Equation (10-4) has a separable φ distribution that separates directivity into a product. We ignore these factors for now and concentrate on the z-axis pattern and associated directivity. Linear-rod antennas have increased directivity because of dipole φ distributions and their increased radius. A traveling wave with a uniform distribution has pattern response sin(ψ/2) ψ/2

where ψ = kL(P − cos θ )

(10-6)

for θ measured from the z-axis. The y or φ distribution determines the pattern in the other coordinate. For P > 1, a slow wave, the beam peak approaches θ = 0 when P → 1. The length bounds the range P for an end-ﬁre pattern peak. Leaky waves, P < 1, have a pattern peak when P = cos θ , or θmax = cos−1 P

(10-7)

The pattern peak approaches end ﬁre (θ = 0) as P → 1. By increasing P beyond 1, the directivity increases and reaches maximum value for a given P , depending on the length [1] 0.465 P =1+ (10-8) L Equation (10-8) is the Hansen and Woodyard criterion for increased directivity of a long end-ﬁre structure commonly approximated by [2] P =1+

1 2L

(10-9)

The phase increase of 180◦ [Eq. (10-9)] along the length gives the maximum directivity for a long structure with a uniform distribution. The amplitude distribution for most surface-wave devices (P > 1) peaks near the input and the taper reduces the gain by the amplitude taper efﬁciency [Eq. (4-8)]. We reduce the relative propagation constant from that given by Eq. (10-8) depending on the length [3]: P =1+

1 RL

(10-10)

where R = 6 at L = λ, diminishing to 3 from L = 3λ to L = 8λ and tapering to 2 [Eq. (10-9)] at L = 20λ. Zucker [4] uses R = 6 for the amplitude, which peaks by 3 dB at the input for all lengths. Equations (10-8) and (10-10) give designs with only small differences in gain. The value of P controls one edge of the visible region. Setting P = 0 centers the visible region about ψ = 0. End ﬁre occurs at P = 1. As we increase P beyond 1, the beam peak of the distribution in ψ space moves into invisible space and the sidelobe level increases. A progression of the distribution sidelobes becomes beam peaks as P increases. Since the amplitude difference between sidelobes decreases as the sidelobe

477

GENERAL TRAVELING WAVES

number increases, the sidelobe level of the pattern increases as the pattern degrades because P exceeds the value given by Eq. (10-9). Figure 10-1 shows the effects on directivity of varying P on an axisymmetrical traveling-wave antenna with a uniform amplitude distribution. For broadside radiation (P = 0) and for P near end ﬁre, the directivity is a constant value with scan: directivity =

2L λ

(10-11)

When the broadside conical beam is scanned until the cone joins into a single end-ﬁre beam, the directivity increases. For end ﬁre, P = 1: directivity =

4L λ

end ﬁre

(10-12)

SL λ

(10-13)

The directivity peaks for P given by Eq. (10-8): directivity =

10l

18

8l

Directivity, dB

16

6l 5l

14 4l 3l

12 2l

10

8

6 0.8

0.9

1.0

1.1

1.2

1.3

1.4

Relative Propagation Constant, P

FIGURE 10-1 Directivity of an axisymmetrical uniform-distribution traveling wave.

478

TRAVELING-WAVE ANTENNAS

20 HE11 Mode

18

Directivity, dB

16

Hansen-Woodyard Criterion

Helical Antenna

P=1

14 12 10 8 6 4 1

2

3

4

5

7 6 Length, l

8

9

10

11

12

FIGURE 10-2 Directivity of an end-ﬁre traveling-wave antenna.

L/λ

2

4

6

10

20

S

7.92

7.58

7.45

7.33

7.25

Figure 10-2 plots the maximum directivity of end-ﬁre structures versus length. For the case P = 1, Eq. (10-12) gives the directivity on the curve. The Hansen and Woodyard criterion increases the directivity as shown in Figure 10-2 for an inﬁnitesimal-diameter structure. The distribution on the ﬁnite diameter of the helical wire antenna in the axial mode increases the directivity over that for the Hansen and Woodyard increased directivity criterion. The hybrid mode with its linear polarization has a dipole null normal to the traveling-wave axis whose elemental pattern also increases directivity. Figure 10-3 is a plot of the corresponding beamwidths of those structures. Figure 10-2 sets an upper bound to the possible directivity of a small-diameter end-ﬁre travelingwave structure of given length. 10-1.1 Slow Wave A slow wave exists on an open transmission-line structure that binds the wave by slowing a passing wave and bending it in the direction of the structure. In the same manner, a lens bends waves toward regions of higher index of refraction (increased slowing). We designate x as the direction normal to a planar structure and the radial coordinate ρ as the direction normal to the cylindrical slow-wave structure. The relation between propagation constants in various directions is found in any electromagnetics text [5]: kz2 + kx2 = k 2 or kz2 + kρ2 = k 2 (10-14) Since x (or ρ) is unbounded, the waves must attenuate exponentially from the surface: α = j kx

or

α = j kρ

(10-15)

GENERAL TRAVELING WAVES

479

90

Beamwidth (degrees)

80 70 60 Hansen-Woodyard Criterion

50

P=1

40

HE11 Helical

30 20 10 1

2

3

4

5

7 6 Length, l

8

9

10

11

12

FIGURE 10-3 Beamwidth of a traveling-wave end-ﬁre antenna.

The z-directed propagation constant becomes kz2 = k 2 + α 2 = P 2 k 2

where P =

α2 1+ 2 = k

1+

λα 2π

2 (10-16)

P , the relative propagation constant, becomes a measure of the wave binding to the surface. We rearrange Eq. (10-16): α=

2π 2 2π 2 P − 1 (Np/λ) = 8.63 P − 1 (dB/λ) λ λ

As P increases, the wave is more tightly bound to the surface. Figure 10-4 is a plot of the distances normal to the surface of constant-ﬁeld contours versus P . The ﬁelds attenuate rapidly normal to the surface. For P → 1, the slow-wave structure only diffracts passing plane waves without capturing the power. This is the sense of a cutoff frequency for the structure. Most surface-wave antennas consist of three regions. The feed region launches the wave on the structure with P between 1.2 and 1.3 [4]. The structure tapers in a short section until P suitable for the length is reached. We design for a given phase shift along the entire length. For example, a long antenna would be designed so that the wave on the structure has an excess phase shift of 180◦ [Eq. (10-9)] over the traveling wave in free space. Near the end we sometimes taper the structure to reduce the end reﬂection given approximately by [4] P 2 − 1 (power). This end taper can be quite short and achieve good results.

480

TRAVELING-WAVE ANTENNAS 4

Distance, l

3

2 60 dB 50 dB 40 dB

1

30 dB 20 dB 6 dB 3 dB

1

10 dB

1.1

1.2

1.3

1.4

1.5

Relative Propagation Constant, P

FIGURE 10-4

Constant ﬁeld contours off the surface of a surface-wave structure.

10-1.2 Fast Waves (Leaky Wave Structure) Only closed structures such as waveguides support fast waves. An open structure requires a negative α [Eq. (10-16)] for fast waves, which implies an exponentially increasing wave away from the structure. The structure soon radiates all its power and no longer guides t