0% found this document useful (0 votes)
366 views372 pages

2019 Book CompactSlotArrayAntennasForWir

Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
366 views372 pages

2019 Book CompactSlotArrayAntennasForWir

Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 372

Signals and Communication Technology

Alan J. Sangster

Compact Slot
Array Antennas
for Wireless
Communications
Signals and Communication Technology
The series “Signals and Communications Technology” is devoted to fundamentals
and applications of modern methods of signal processing and cutting-edge
communication technologies. The main topics are information and signal theory,
acoustical signal processing, image processing and multimedia systems, mobile and
wireless communications, and computer and communication networks. Volumes in
the series address researchers in academia and industrial R&D departments. The
series is application-oriented. The level of presentation of each individual volume,
however, depends on the subject and can range from practical to scientific.

More information about this series at http://www.springer.com/series/4748


Alan J. Sangster

Compact Slot Array


Antennas for Wireless
Communications

123
Alan J. Sangster
Edinburgh, UK

ISSN 1860-4862 ISSN 1860-4870 (electronic)


Signals and Communication Technology
ISBN 978-3-030-01752-1 ISBN 978-3-030-01753-8 (eBook)
https://doi.org/10.1007/978-3-030-01753-8

Library of Congress Control Number: 2018959269

© Springer Nature Switzerland AG 2019


This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt from
the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, express or implied, with respect to the material contained herein or
for any errors or omissions that may have been made. The publisher remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
To Hamish
Whatever you choose to be, be the very best
you can. That way lies fulfilment.
Preface

At about the time when this book was being compiled, it was hardly controversial to
suggest that the predominant evolutionary course of electronic engineering had been
characterised by an inexorable drive toward the miniaturisation of components,
devices, and systems, facilitated by the unprecedented advances in integrated circuit
(IC) technology. The obvious example is the computer, the hardware for which
could fill a modest laboratory in the 1960’s, yet by 2015 a device of comparable
computational power was small enough to be held in the palm of the hand—Moore’s
law gives this trend a quantitive basis. However, for practitioners in the discipline of
radio frequency communications the ‘fly in the oitment’, to this generally desirable
miniaturisation drift, has been the interface with ‘free-space’, namely the antenna.
This has been particularly true in the burgeoning area of handheld mobile devices.
Unlike virtually all other components in an electronic circuit the antenna is
dimensionally constrained by the unyielding laws of physics. For reasons of radi-
ation efficiency, at least one major dimension (L) of the antenna has to be, in size,
close to, or greater than, a half of the free space wavelength (ko) at the frequency of
operation (L  k0 =2Þ. Needless to say, in the realms of portable and handheld
mobile communications devices, or in the domain of implantable devices for
monitoring medical conditions, circumventing this restraint has been a major focus
of recent compact antenna research. The problem can be simply stated. To create an
efficient compact antenna, at a prescribed frequency, any electrical dimension
which governs radiation strength must be preserved yet remain within the bounds of
a physical package that is required to be everywhere smaller. These contradictory
requirements are usually satisfied, with varying degrees of success, by securing
electrical length in a dimensionally restricted package through the adoption of
‘tricks’ such as meandering, corkscrewing and twisting the key dimensional ele-
ment of the shrunk antenna. However, not surprisingly, these distortions can have
deleterious effects on other important antenna properties and compensating tech-
niques are generally necessary to achieve viability.

vii
viii Preface

The purpose of this text lies in describing and providing design guidelines for
antennas which pursue compactness by recruiting the slot radiator as the funda-
mental building block. Initial research on the so called slot radiator can be traced
back in time to before the Second World War. It was strenuously propelled by
military pressures to advance the detection capabilities of radar. As a consequence
of this early ‘baptism’ the literature is comprehensive and extensive, requiring
patient time-consuming study by antenna designers, and by research engineers, with
a need to claim mastery of the topic. Recent progress has been largely triggered by
the burgeoning of modern communications techniques demanding increasingly
compact, lighter, more versatile antennas, offering large bandwidth to the user. For
the design engineer due diligence requires a thorough sifting of this literature,
which is inevitably time consuming and expensive when it is so extensive.
While the primary threads in the historical evolution of compact array antennas
are identified, the main focus is to provide an electrical engineering text which
summarises the extensive compact antenna related literature. More particularly, the
aim is to illuminate, in relation to the now ubiquitous slot radiator, how despite
significant antenna size reductions, seemingly defying physics, desirable radiation
pattern properties can be preserved. This leads to an examination of recent advances
in frequency selective surfaces or metamaterials which can, with caution, be used to
facilitate physics defying dimensional restrictions.
Slot radiators, excited from waveguide, have for some time provided basic
elements for creating array antennas most commonly for applications in radar and in
satellite communications. On the other hand, in stripline, microstrip or coplanar
waveguide formats slot radiators, either directly excited or through a patch transi-
tion, additionally form fundamental components in compact antennas and confor-
mal antennas, particularly in the realm of mobile communications. The range of
possible formats is briefly illustrated in Chap. 1.
In constructing the eleven chapters, focused both on already established and on
continuing developments in the realm of compact antennas for wireless commu-
nications, the intention has been to restrict the content to topics deemed salient to
antennas of the compact variety and particularly is this so in addressing theory,
modelling and design. This stance has been achieved by relegating some general,
but useful, background material to appendices where a copious bibliography is
provided directing the reader to standard texts which provide further elaboration.
The early chapters of the book (Chaps. 1–5) focus on the radiation mechanisms
and characteristics of the slot radiator when treated as an aperture antenna. The
progressive incorporation of the slot radiator into antenna technology developments
and in particular those directed toward compactness are both introduced and pre-
sented schematically in Chap. 1. This chapter also provides a guide to more detailed
explorations and expositions of compact antenna evolution which appear in later
chapters. In Chap. 2 the electromagnetic fundamentals of the radiation mechanism
are introduced using analogies from electrostatics and magnetostatics to construct
the key equations without undue mathematical toil. With these indispensable
equations established, Chap. 3 uses them to construct the now conventional
Preface ix

mathematical representations of the linear aperture which can be viewed as the


building block for electromagnetic modelling of more complex antenna geometries.
It also introduces the reader to basic antenna pattern forms, definitions and
trade-offs.
The reader is introduced to the method of moments in Chaps. 4 and 5 as
arguably the predominant route to efficient electromagnetic modelling of antenna
forms containing slot radiators. The fundamental Helmholtz inhomogeneous dif-
ferential equations for radiation generated by electric and magnetic current sources
are tabulated in Chap. 4, alongside the equivalent Green’s equations and the rele-
vant Green’s function solutions. Armed with these tools the moment method is
constructed. It is consolidated in Chap. 5 by applying it to the solution of two
typical slot radiation problems, namely the conventional shunt slot in rectangular
waveguide and the less common polarisation agile T-slot radiator in bifurcated
rectangular waveguide.
The second half of the book (Chaps. 6–11) delves into the topic of array
antennas. The evolving compact forms of such arrays, particularly the partially
reflective periodic surface varieties, have been attracting increasing interest from the
wireless communications community in recent years. To set the scene, the funda-
mental nature of arrays is first examined in Chap. 6, by analysing traditional linear
phased arrays comprising independent isotropic radiators. This is accomplished by
following a well established analysis route based on a polynomial representation of
such an array. The procedure permits the illumination of basic relationships
between array geometry, in particular element excitation levels and element spac-
ing, and performance specifications such as directivity, beamwidth, side-lobe levels,
grating lobes levels, mainbeam scanning and bandwidth limitations.
Slot array manifestations which are potentially much more compact and cost
effective than phased array alternatives are broached in Chaps. 7 and 8, with
travelling-wave array examples in the former, and scanned and leaky-wave array
versions in the latter. Design principles for conventional waveguide fed
travelling-wave arrays are presented in depth in Chap. 7 with the results used to
establish the basic trade-offs, firstly between slot orientation and positioning, and
secondly between achievable radiation patterns and the undesirable distortions
imposed by frequency deviation from the design value. While frequency deviation
is represented as a problem for conventional travelling-wave array antennas, this
need not be the case as is demonstrated in Chap. 8. There, it is shown that inten-
tional frequency scanning of the primary radiating beam of a travelling-wave array
can be significantly amplified by resorting to phase delay enhancement techniques.
The technology is shown to have major implications for radar. The reader is thereby
introduced to the periodic structure and its complex modal solutions and hence to
the importance of slow-waves and surface waves. It is demonstrated that the key to
the formulation of field solutions which satisfy periodic boundary conditions are the
Floquet harmonics. With this armoury, it is shown that the application of electro-
magnetic field boundary conditions at a periodic interface leads to an important
determinantal equation for computing the phase coefficient for the fundamental
x Preface

mode. A quasi-Green’s function for periodic waveguide is also developed. This


permits a slot radiator in a periodic waveguide feed to be modelled using the
moment method, in a manner not too unlike the slot discontinuity in conventional
waveguide. The chapter concludes with an examination of an antenna which is
physically similar to the travelling-wave array but is really quite different electro-
magnetically—namely the leaky-wave array. The differences between this array and
the conventional travelling-wave array are emphasised.
The state-of-the-art in the evolution for communication systems of compact
planar antennas, particularly those which utilize periodic array structures but
without any phased array connotations, is broached in Chaps. 9, 10 and 11.
Arguably, the least radical departure from established antenna practice is the planar
reflect-array which is addressed in Chap. 9. In essence, the reflect-array aims to
replace the parabolic reflector in a conventional high gain antenna with a planar
array geometry. To do this the surface is made to replicate the optics of the replaced
parabola by providing equivalent element-to-element phase profiling over the flat
digitized array surface.
A planar array antenna which arguably represents a more significant departure
from the preceding technology is provided in Chap. 10 by the introduction of
periodic surfaces into the design process. The substrates thus formed are variously
termed, frequency selective surfaces, partially reflective surfaces, textured surfaces
or occasionally metamaterials. In planar arrangements the periodic layer usually
forms a sandwich with a ground plane thus creating a parallel plate structure which
acts as either a Fabry-Perot resonator or as a leaky-wave array. The chapter outlines
the operational modes which lead to a high gain fixed beam antenna option or to a
frequency scanned leak-wave antenna alternative. Reconfigurability of the periodic
surface to achieve leaky-wave scanning at fixed frequency is also studied.
Finally in Chap. 11 an emerging technology is examined which could exercise
considerable influence on compact array antenna developments intended for
satellite communication systems and for devices incorporated into mobile plat-
forms. An old technique, initially developed for enhancing the radar visibility of, in
particular, small sea or ocean going vessels, has been re-engineered for incorpo-
ration into array antennas. Modern applications for this technology have grown out
of the need to create antennas which can automatically track a signal source when
relative movement exists between the transmitter and the receiver. A relevant
example is secure communications with a low earth orbit satellite.
Naturally all views, assertions, claims, calculations and items of factual infor-
mation contained in this book have been selected or generated by myself, and any
errors therein are my responsibility. However, the book would not have seen the
light of day without numerous personal interactions (too many to identify), with
family, with friends, and with colleagues at the Heriot-Watt University, on the topic
of array antenna developments to meet the burgeoning needs of wireless commu-
nications. So if I have talked to you on this topic, I thank you for your contribution,
and the stimulus it may have provided for the creation of this book. A special thank
you is extended to Dr. Hanyang Wang for figures in Chap. 3. I would, also,
Preface xi

particularly like to thank the members of staff, at the Heriot-Watt University library,
who have been very helpful in ensuring that I was able to access a wide range of
written material, the contents of some of which have been germane to the realisation
of this project.

Edinburgh, UK Professor Alan J. Sangster


C.Eng., FIET. B.Sc. (Eng), M.Sc., Ph.D.
Contents

1 Evolution of Compact Slot Antennas . . . . . . . . . . . . . . . . . . . . . . . 1


1.1 Free-Space Waves and Transmission Lines . . . . . . . . . . . . . . . 1
1.2 Plane Electromagnetic Wave in Free-Space . . . . . . . . . . . . . . . 2
1.3 TEM Transmission Lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3.1 TEM Transmission Line Fed Radiating
Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3.2 Radiating Element with Polarisation Agility . . . . . . . . 7
1.4 Waveguide Feed System . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.4.1 Trapped TE/TM Waveguide Modes . . . . . . . . . . . . . . 12
1.5 Waveguide Slot Radiators . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.5.1 Conventional Waveguide Slot Radiators . . . . . . . . . . 16
1.5.2 Unconventional Waveguide Slot Radiators . . . . . . . . . 20
1.5.3 Waveguide Cavity Backed Slot Radiators . . . . . . . . . 24
1.6 Planar Reflect-Arrays for Compact Antennas . . . . . . . . . . . . . 26
1.7 Compact Cavity Resonator Antennas . . . . . . . . . . . . . . . . . . . 28
1.8 Retro-Directive Array Antennas . . . . . . . . . . . . . . . . . . . . . . . 30
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2 Fundamentals of Electromagnetic Radiation . . . . . . . . . . . ....... 35
2.1 Theoretical Formulation of Radiation from a Slot . . . ....... 35
2.2 Reciprocity for Macroscopic Antennas . . . . . . . . . . . ....... 38
2.2.1 Reciprocity Theorems for Electromagnetic
Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.2.2 Inhomogeneous Vector Helmholtz Equation . . . . . . . . 41
2.3 Auxiliary Potential Functions . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.4 Electrostatics Analogy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2.5 Magnetostatics Analogy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.6 Hertzian Dipole . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

xiii
xiv Contents

2.7 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56


Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3 Compact Aperture Antennas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.1 Aperture Antenna Models . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.2 Auxiliary Potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.3 Radiation from a Linear Aperture . . . . . . . . . . . . . . . . . . . . . . 63
3.3.1 Huygen’s Principle and Equivalent Sources . . . . . . . . 63
3.3.2 Plane Wave Spectrum . . . . . . . . . . . . . . . . . . . . . . . . 65
3.4 Spectrum Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
3.4.1 Pattern Nulls and Sidelobes . . . . . . . . . . . . . . . . . . . . 70
3.4.2 Mainlobe Beamwidth . . . . . . . . . . . . . . . . . . . . . . . . 71
3.4.3 Pattern Gain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.5 Rectangular Aperture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.5.1 Uniformly Illuminated Rectangular Aperture . . . . . . . 74
3.5.2 Directivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3.6 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4 Computational Modelling Techniques for Slot Antennas . . . . . . . . 81
4.1 Introduction—Inhomogeneous Helmholtz Equation . . . . . . . . . 81
4.2 Common Green’s Functions in Electromagnetic Radiation . . . . 84
4.2.1 Green’s Function Solution for Hertzian Dipole . . . . . . 88
4.2.2 Modified Green’s Functions for Electromagnetic
Radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 90
4.2.3 Hertzian Dipole Adjacent to Conducting Ground
Plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.3 Concise Formulation of the Moment Method . . . . . . . . . . . . . 97
4.4 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
5 Moment Method Models of Compact Slot Antennas . . . . . . . . . . . . 101
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
5.2 Moment Method Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
5.3 Scattering for Longitudinal Slot in Waveguide . . . . . . . . . . . . 109
5.3.1 Computed and Measured Results . . . . . . . . . . . . . . . . 111
5.4 T-Shaped Slot in Bifurcated Waveguide . . . . . . . . . . . . . . . . . 115
5.4.1 Computed and Measured Results . . . . . . . . . . . . . . . . 121
5.5 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
6 Resonant and Non-resonant Linear Slot Arrays . . . . . . . . . . . . . . . 131
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
6.2 Uniform Linear Array of Isotropic Elements . . . . . . . . . . . . . . 132
Contents xv

6.2.1 Radiation Patterns . . . . . . . . . . . . . . . . . . . . . . . . . . 134


6.2.2 Broadside Array . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
6.2.3 End-Fire Array . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
6.2.4 Scanned Array . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
6.3 Array Design Using Theory of Polynomials . . . . . . . . . . . . . . 140
6.3.1 Optimum Element Spacing . . . . . . . . . . . . . . . . . . . . 147
6.3.2 Binomial Array . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
6.4 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
7 Conventional Waveguide-Fed Travelling-Wave Slot Arrays . . . . . . 151
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
7.2 Travelling-Wave Array—Design Principles . . . . . . . . . . . . . . . 154
7.2.1 Staggered Shunt Slot Array . . . . . . . . . . . . . . . . . . . . 155
7.2.2 Unstaggered Shunt Slot Array . . . . . . . . . . . . . . . . . . 158
7.2.3 Travelling-Wave Array Antenna Design . . . . . . . . . . 159
7.2.4 Resonant Array Design . . . . . . . . . . . . . . . . . . . . . . . 167
7.3 Some Slot Array Applications . . . . . . . . . . . . . . . . . . . . . . . . 169
7.4 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
8 Frequency Scanned and Leaky-Wave Linear Slot Arrays . . . . . . . . 179
8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
8.2 Frequency Scanned Slotted Waveguide Array . . . . . . . . . . . . . 179
8.2.1 Dielectric Loaded Waveguide . . . . . . . . . . . . . . . . . . 180
8.2.2 Sinuous Waveguide Feed . . . . . . . . . . . . . . . . . . . . . 180
8.2.3 Periodically Loaded Waveguide Feed . . . . . . . . . . . . 182
8.3 Transition from Uniform Waveguide to Slow-Waveguide
Feed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
8.3.1 Fast Waveguide Frequency Characteristics . . . . . . . . . 184
8.3.2 Floquet Harmonics . . . . . . . . . . . . . . . . . . . . . . . . . . 185
8.3.3 Slow Waveguide Frequency Characteristics . . . . . . . . 188
8.3.4 Approximate Solution for Phase Coefficient . . . . . . . . 190
8.3.5 Boundary Condition at Periodic Interface . . . . . . . . . . 194
8.3.6 Determinantal Equation . . . . . . . . . . . . . . . . . . . . . . . 196
8.4 Slot Radiators in a Periodic Waveguide Feed . . . . . . . . . . . . . 199
8.4.1 Frequency Scanned Linear Slot Array Employing
Periodic Waveguide Feed . . . . . . . . . . . . . . . . . . . . . 202
8.5 Uniform Leaky-Wave Antennas . . . . . . . . . . . . . . . . . . . . . . . 202
8.6 Periodic Leaky-Wave Antenna Arrays . . . . . . . . . . . . . . . . . . 208
8.7 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
xvi Contents

9 Compact Planar Reflectarrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215


9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
9.2 Passive Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
9.3 Reflectarrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
9.3.1 Reflection Phase Computation . . . . . . . . . . . . . . . . . . 226
9.4 Reflectarray—State of the Art . . . . . . . . . . . . . . . . . . . . . . . . 228
9.4.1 Microstrip Reflectarrays . . . . . . . . . . . . . . . . . . . . . . 229
9.4.2 All-Metal Reflectarrays . . . . . . . . . . . . . . . . . . . . . . . 232
9.4.3 3D-Printed Reflectarrays . . . . . . . . . . . . . . . . . . . . . . 237
9.5 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
10 Compact Planar Resonator Arrays . . . . . . . . . . . . . . . . . . . . . . . . . 243
10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
10.2 Cavity Resonator Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
10.3 PRS Antenna . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
10.3.1 Radiation Pattern Estimation . . . . . . . . . . . . . . . . . . . 247
10.4 Leaky-Wave Fabry-Perot Antenna . . . . . . . . . . . . . . . . . . . . . 249
10.4.1 Reciprocal Periodic Moment Method for CRA . . . . . . 249
10.4.2 CRA Radiation Characteristics . . . . . . . . . . . . . . . . . 259
10.5 CRA Thinning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260
10.6 Bandwidth Enhancement . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
10.7 Reconfigurable PRS for Fixed Frequency Scanning . . . . . . . . . 262
10.8 Beam Scanning in a Leaky-Wave CRA . . . . . . . . . . . . . . . . . 267
10.8.1 Linear Leaky-Wave Array . . . . . . . . . . . . . . . . . . . . . 267
10.8.2 Planar Leaky-Wave Array . . . . . . . . . . . . . . . . . . . . . 270
10.9 The Role of Metamaterials in Compact Slot Arrays . . . . . . . . . 272
10.9.1 Arrays as Passive Periodic Structures . . . . . . . . . . . . . 272
10.9.2 FSS as a Metamaterial . . . . . . . . . . . . . . . . . . . . . . . 274
10.9.3 Negative Refractive Index . . . . . . . . . . . . . . . . . . . . . 276
10.9.4 The Flat Lens . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280
10.10 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
11 Retro-directive Compact Array Antennas . . . . . . . . . . . . . . . . . . . . 285
11.1 Retro-directive Tracking—Basic Concept . . . . . . . . . . . . . . . . 285
11.2 Retrodirective Implementation Through Mixing . . . . . . . . . . . 286
11.3 Conjugation Employing PLL’s . . . . . . . . . . . . . . . . . . . . . . . . 288
11.4 Phase Conjugation Using IQ Processing . . . . . . . . . . . . . . . . . 290
11.5 Retro-directive Array Implementation . . . . . . . . . . . . . . . . . . . 291
11.6 Cross-slot Fed Microstrip Patch Antenna . . . . . . . . . . . . . . . . 293
11.7 Dual-Polarised Radiation from Apertures . . . . . . . . . . . . . . . . 293
11.8 Parasitically Fed Microstrip Patch Antenna . . . . . . . . . . . . . . . 297
Contents xvii

11.9 Phase Conjugated Active Antenna Element . . . . . . . . . . . . . . . 299


11.10 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302
Appendix A: Free-Space Electromagnetic Waves . . . . . . . . . . . . . . . . . . . 305
Appendix B: Basic Waveguide Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
Appendix C: Waveguide Green’s Function and Scattering
Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329
Appendix D: Green’s Function for Fin Loaded Waveguide . . . . . . . . . . . 335
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337
Abbreviations

A Amps
AC Alternating current
AMC Artificial magnetic conductor
BGAN Broadband global area network
BPF Bandpass filter
BW Bandwidth
BWFN Beamwidth for nulls
C Capacitance (Farads)
CBCPW Conductor backed co-planar waveguide
CEM Classical electromagnetism
CRA Cavity resonator antenna
CW Continuous wave
DC Direct current
DUT Device under test
EM Electromagnetism
EMW Electromagnetic wave
FEM Finite element method
FET Field effect transistor
FSS Frequency selective surface
HD High definition
HDMI High definition multimedia interface
HFSS High frequency structure simulator
HIS High impedance surface
HPBW Half power beamwidth
IF Intermediate frequency
IQ In-phase/quadrature-phase
L Inductance (Henries)
LEO Low earth orbit
LWA Leaky-wave antenna
MF Medium frequency

xix
xx Abbreviations

MKS Metre/kilogram/second
MMIC Monolithic microwave integrated circuit
MoM Method of Moments
PC Printed circuit
PCC Phase conjugating circuit
PMC Perfect magnetic conductor
PPL Phase locked loop
PRS Partially reflective surface
Q Quality factor (of resonance)
RF Radio frequency
RFID Radio frequency identification
RLP Ring loaded patch
SIW Substrate integrated waveguide
SLL Side lobe level
SP Square patch
SR Square ring
TE Transverse electric
TEM Transverse electromagnetic
TEN Transmission line equivalent network
TM Transverse magnetic
UWB Ultra-wideband
V Volt
VCO Voltage controlled oscillator
VHF Very high frequency
VLF Very low frequency
WG Waveguide
WLAN Wireless local area network
WPAN Wireless personal area network
List of Figures

Fig. 1.1 Representation of a simple sinusoidal plane electromagnetic


wave showing the fundamental orthogonal relationships
between the vector electric (E) and magnetic (H) fields
and the direction of propagation (k) . . . . . . . . . . . . . . . . . . . . 2
Fig. 1.2 Trapped TEM mode transmission lines systems . . . . . . . . . . . 5
Fig. 1.3 Slot radiator located in the ground plane of microstripline . . . 6
Fig. 1.4 Fundamental mode of operation for patch antenna, a top view
of patch antenna, b side view showing trapped E-field—note
edge radiation can be viewed as radiation from a two element
slot array, c voltage and current variation in patch and line . . . 7
Fig. 1.5 T-slot radiator fed from balanced microstrip line . . . . . . . . . . 8
Fig. 1.6 Mechanism for delivering polarisation agility based
on microstrip . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Fig. 1.7 Plane wave interference pattern . . . . . . . . . . . . . . . . . . . . . . . . 11
Fig. 1.8 Interference patterns of plane electromagnetic waves
reflecting off a flat perfectly conducting surface . . . . . . . . . . . 11
Fig. 1.9 TE10 mode field patterns for rectangular waveguide: width a,
height b, a E and H-fields in the y-z plane, b E-field in the x-y
plane, c E and H-fields in the x-z plane (E-field = green solid
lines: H-field = blue dashed lines: current density J = orange
chain-dotted lines) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Fig. 1.10 Brillouin diagram showing graphically the frequency
characteristics of the TE10 mode in rectangular waveguide
(orange line denotes b10 = ko and thus vp = c for all
frequencies: blue solid line = b10 for a rectangular waveguide
of width a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
Fig. 1.11 A hollow rectangular TE10 mode waveguide showing some
possible slot locations required to secure radiation. One
guide wavelength (kg10) of the modal current pattern
is depicted . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

xxi
xxii List of Figures

Fig. 1.12 Computed radiation pattern for a six element shunt


slot waveguide fed array (with acknowledgement
to www.remcom.com) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Fig. 1.13 Planar shunt slot resonant waveguide array, a isometric view,
b end sectional view . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
Fig. 1.14 Horizontally polarised resonant waveguide planar array
antenna formed from edge-slots . . . . . . . . . . . . . . . . . . . . . . . 20
Fig. 1.15 Omni-directional slot-doublet in rectangular waveguide . . . . . 21
Fig. 1.16 Simulated omni-directional radiation pattern in x-y plane
for slot-doublet radiator. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
Fig. 1.17 Dual-mode bifurcated rectangular waveguide structure
showing even mode to odd mode polarisation switching
mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Fig. 1.18 Polarisation agile waveguide cavity backed slot antenna . . . . . 24
Fig. 1.19 Dual-polarised cavity resonator antenna for RFID scanning
applications at 2.45 GHz (*4 cm wide by 7 cm long) . . . . . 26
Fig. 1.20 Evolution of ‘flat’ focused reflectors for incorporation into
compact reflectarray antennas with applications in
communication systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
Fig. 1.21 Schematic representation of a Fabry-Perot resonator antenna,
a end view of TE mode structure, b top view of TE mode
structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Fig. 1.22 Traditional corner reflectors showing retrodirective
operational mode. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
Fig. 2.1 Original problem of electromagnetic scattering by an aperture
in a conductive plane screen of infinite extent . . . . . . . . . . . . 36
Fig. 2.2 a Radiation source and b radiation sink . . . . . . . . . . . . . . . . . 37
Fig. 2.3 Equivalent circuit representation of transmit/receive antenna
combination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
Fig. 2.4 Electric field flux through a spherical surface containing a
point charge q . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
Fig. 2.5 Geometry for construction of retarded potentials. . . . . . . . . . . 46
Fig. 2.6 Magnetic field at P due to current element Idl0 . . . . . . . . . . . . 48
Fig. 2.7 Short current element at origin of Cartesian coordinate
system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
Fig. 2.8 Electric field contours for radiation from a short current
element at a specific instant in time (red = peaks; blue =
troughs in the wave) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
Fig. 2.9 Power and energy flows in near and far field regions for short
current element . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
Fig. 3.1 Examples of aperture antennas: a H-plane horn, b E-plane
horn, c sectoral horn, d shunt-slot in waveguide,
e edge-slot in waveguide, f dielectric rod, g horn fed
parabolic antenna . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
List of Figures xxiii

Fig. 3.2 Equivalence principles a for arbitrary surface S b for


conducting body S with aperture A . . . . . . . . . . . . . . . . . . . . 59
Fig. 3.3 Application of equivalence principle a original horn
structure b enclosing surface S c equivalent aperture
in infinite screen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
Fig. 3.4 Electromagnetic wave diffraction by a parallel sided
slit in a perfectly conducting membrane
(Huygen’s mechanism) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
Fig. 3.5 Plane wave diffraction at a narrow slit in an infinite perfectly
conducting screen—illustrating the diffraction pattern is a
spectrum of plane waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
Fig. 3.6 Cylindrical coordinates (r, w) of a field point P . . . . . . . . . . . 67
Fig. 3.7 Far-field patterns for apertures with the following field
distributions—uniform, cosine, cosine squared, cosine
squared on a 10 dB pedestal . . . . . . . . . . . . . . . . . . . . . . . . . . 69
Fig. 3.8 Geometry for two-dimensional aperture antenna analysis . . . . 73
Fig. 3.9 Rectangular aperture antenna embedded in a conducting
ground plane of infinite extent . . . . . . . . . . . . . . . . . . . . . . . . 75
Fig. 3.10 Three-dimensional E-field pattern for a rectangular
aperture (3ko  2ko) in a large conducting plane—constant
Ey-directed case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
Fig. 3.11 Three-dimensional E-field pattern for a square aperture
(3ko  3ko) in a large conducting plane—constant
Ex-directed case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
Fig. 4.1 Field and source coordinate system. . . . . . . . . . . . . . . . . . . . . 85
Fig. 4.2 Short current element Idl at origin of Cartesian system . . . . . 90
Fig. 4.3 Field and source regions imaged in an infinite and perfectly
conducting ground plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
Fig. 4.4 Current element above a perfectly conducting ground plane
and perpendicular to it a Converging rays at point P b rays
converging at infinity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
Fig. 4.5 Elevation plane field patterns in dB for a vertically orientated
current element located at different heights above the
perfectly conducting ground plane . . . . . . . . . . . . . . . . . . . . . 96
Fig. 5.1 a Waveguide fed compact slot radiator and b slot detail . . . . 102
Fig. 5.2 Field equivalence representation of slot radiator in thick
walled waveguide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
Fig. 5.3 Slot field approximations for a narrow (w/L < 0.1), near
resonant slot (L/k = 0.5) in the broadwall (T = 0) of
rectangular waveguide, as adopted in early radiation
computations [Solid line with diamond symbols = Oliner:
Dashed line with square symbols = Das] . . . . . . . . . . . . . . . . 107
xxiv List of Figures

Fig. 5.4 E-field distributions for a shunt slot radiator in the broadwall
of WG16 waveguide as predicted by the moment method
(freq = 9.375 GHz, L = 0.361 k, x1 = 6.43 mm, w = 1.58
mm, T = 1.27 mm) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
Fig. 5.5 Relative power radiated by a broadwall waveguide slot as a
function of slot length L, with the number of basis functions
as parameter (Freq = 10 GHz, a = 22.86 mm, b = 10.16 mm,
x1 = 2.00 mm, w = 1.5 mm, T = 0 mm) . . . . . . . . . . . . . . . . 111
Fig. 5.6 Reflection coefficient magnitude (|s11|) as a function of slot
length as predicted by trigonometric basis (Lyon) and by
pulse basis (Vu Khac) moment methods (WG16,
Freq = 9.375 GHz, T = 1.27 mm, x1 = 1.43 mm,
w = 1.58 mm) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
Fig. 5.7 Transmission coefficient magnitude (|s12|) as a function of slot
length as predicted by trigonometric basis (Lyon) and by
pulse basis (Vu Khac) moment methods (WG16,
Freq = 9.375 GHz, T = 1.27 mm, x1 = 1.43 mm,
w = 1.58 mm) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
Fig. 5.8 a Transmission coefficient magnitude for a rectangular
waveguide containing a longitudinal slot of changing length
(Freq = 10.48 GHz, T = 1.32 mm, x1 = 2.00 mm,
w = 1.5 mm) b Transmission coefficient phase for a
rectangular waveguide containing a longitudinal slot of
changing length (Freq = 10.48 GHz, T = 1.32 mm,
x1 = 2.00 mm, w = 1.5 mm). . . . . . . . . . . . . . . . . . . . . . . . . . 114
Fig. 5.9 a Transverse T-slot radiator in dual-mode rectangular
waveguide. b Transverse element of T-slot showing
coordinate system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
Fig. 5.10 Suggested electric field distributions in the transverse
component of the T-slot a L3 = 0, b 0 < L3 < L/2 . . . . . . . . . 117
Fig. 5.11 Transmission line based equivalent circuits for waveguide
slot radiators a conventional slot, b T-slot in bifurcated
waveguide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
Fig. 5.12 Resonant cavity set-up for T-slot scattering in dual mode
waveguide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
Fig. 5.13 Relative power radiated (Prad/Pinc) versus frequency for a
transverse T- slot in bifurcated waveguide in even-mode
(WG16 wall thickness T = 1.32 mm; slot width w = 1.5 mm;
slot length L = 16 mm; septum slot L3 = 5.48 mm) . . . . . . . . 124
Fig. 5.14 T-slot transmission coefficient magnitude a and phase b as a
function of slot length L (Even mode excitation; WG16;
T = 1.32 mm; L3 = 5.48 mm; bifurcation thickness = 1.5
mm; frequency = 9.89 GHz) . . . . . . . . . . . . . . . . . . . . . . . . . . 124
List of Figures xxv

Fig. 5.15 T-slot reflection coefficient magnitude a and phase b as a


function of slot length L (Even mode excitation; WG16;
T = 1.32 mm; L3 = 5.48 mm; bifurcation thickness = 1.5
mm; frequency = 9.89 GHz) . . . . . . . . . . . . . . . . . . . . . . . . . . 125
Fig. 5.16 Relative power radiated in dB as a function of clearance slot
depth (L3) for a symmetrically located T-slot—even mode
excitation (WG16; T = 1.32 mm; L3 = 5.48 mm; bifurcation
thickness = 1.5 mm; frequency = 9.89 GHz) . . . . . . . . . . . . . 125
Fig. 5.17 Moment method solution (solid lines) for T-slot compared
with finite element software (dashed lines) for deep
dual-mode waveguide (L = 2 mm, L3 = 13, 14, 15 mm,
w = 1.5 mm, t = 1.27 mm) . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
Fig. 6.1 Geometrical relationships for linear array with element
spacing d . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
Fig. 6.2 Principal lobe—broadside radiation . . . . . . . . . . . . . . . . . . . . . 135
Fig. 6.3 Principal lobe—end-fire array . . . . . . . . . . . . . . . . . . . . . . . . . 135
Fig. 6.4 Beamwidth variation with scanning (adapted from [2]) . . . . . . 139
Fig. 6.5 Gain as a function of d/ko for a uniform broadside array
(adapted from [3]) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
Fig. 6.6 Unit circle diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
Fig. 6.7 Range of w for broadside array (a = 0) . . . . . . . . . . . . . . . . . . 142
Fig. 6.8 Location of roots on circle diagram for four element
uniform array . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
Fig. 6.9 Far-field pattern for a four element uniform broadside array
with d = ko/2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
Fig. 6.10 Root location for four element end-fire array . . . . . . . . . . . . . 145
Fig. 6.11 Root locations for four element end-fire array with improved
far field pattern . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
Fig. 6.12 End-fire array pattern improvements by root adjustments . . . . 146
Fig. 6.13 Diagrams showing the range of w for optimum element
spacing. a Uniform broadside array, b uniform endfire
array. (The dotted arrows imply that the zeros occupy
the entire circle) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
Fig. 6.14 Roots locations on unit circle for binomial array . . . . . . . . . . 148
Fig. 7.1 Longitudinal (shunt) slot in TE10 mode rectangular
waveguide: a waveguide geometry, b equivalent circuit
for resonant slot with normalised g = Ys/Yo . . . . . . . . . . . . . . 153
Fig. 7.2 Resonant slot travelling-wave array antenna (a) and
equivalent transmission line model (b) . . . . . . . . . . . . . . . . . . 154
Fig. 7.3 Influence of slot spacing d on the principal beam locations
(/ degrees) for a staggered shunt slot travelling wave array
in WG16 waveguide at 10 GHz . . . . . . . . . . . . . . . . . . . . . . . 157
xxvi List of Figures

Fig. 7.4 Influence of slot spacing d on the principal beam locations


(/ degrees) for a unstaggered shunt slot travelling wave array
in WG16 waveguide at 10 GHz . . . . . . . . . . . . . . . . . . . . . . . 158
Fig. 7.5 Influence of slot spacing d on the principal beam locations
(/ degrees) for a unstaggered shunt slot travelling wave array
in WG16 waveguide at a 8 GHz and b 12 GHz . . . . . . . . . . . 160
Fig. 7.6 Shunt slot travelling wave array antenna—design example
a array configuration, b general transmission line model . . . . 161
Fig. 7.7 Continuous conductance profile for lossless, perfectly
matched, shunt-slot travelling-wave array antennas (with
acknowledgement to [9]). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
Fig. 7.8 Resonant form of the waveguide shunt slot array a schematic
representation, b circuit representation . . . . . . . . . . . . . . . . . . 168
Fig. 7.9 Pictorial representation of the slot excitation mechanism in
the resonant array . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
Fig. 7.10 Single layer planar slotted waveguide array antennas with
a co-phase corporate feed b anti-phase transverse linear
waveguide feed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
Fig. 7.11 Millimetre-wave slotted waveguide array antenna in substrate
based post-wall or SIW waveguide [18] . . . . . . . . . . . . . . . . . 174
Fig. 7.12 Corporate feed and laminated waveguide slot array
antenna . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
Fig. 7.13 Photograph of laminated hollow waveguide slot array a top
view, b side view [21] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
Fig. 8.1 Schematic demonstrating frequency scanning using a sinuous
waveguide feed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
Fig. 8.2 Double curvature parabolic reflector antenna illuminated
by a sinuous feed. Courtesy of NRPL Group:
www.radartutorial.com . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
Fig. 8.3 Planar array of slotted waveguide radiators end-excited
by a sinuous feed (on the left of the planar array structure)
(see [1]) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
Fig. 8.4 Frequency/phase diagrams (x-b diagrams) for a uniform
rectangular waveguide—TE10 mode = red curve: TE20 mode
= mauve curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
Fig. 8.5 Fin or iris loaded rectangular waveguide . . . . . . . . . . . . . . . . 185
Fig. 8.6 Frequency/phase diagram (x-b diagram) for rectangular
waveguide periodically loaded with metal fins—showing
fundamental plus first higher space harmonic (NB slow-wave
solutions to right of v = c line: fast-wave solutions occur in
region to the left of v = c line—note in an open structure (see
Chap. 10) waves in the fast region become leaky) . . . . . . . . . 189
List of Figures xxvii

Fig. 8.7 Periodic cell models showing E-wall and H-wall inserts.
a H-wall, b E-wall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
Fig. 8.8 Predicted and measured phase coefficients, as a function of
frequency, for a fin-loaded rectangular waveguide
a a = 72.14 mm, b = 12.24 mm, h = 21.8 mm, G = 4.76 mm,
t = 1.58 mm, b a = 64.5 mm, b = 12.24 mm, h = 26.5 mm,
G = 11.11 mm, t = 1.58 mm (curves = theory,
points = measured) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
Fig. 8.9 Fin-loaded feed waveguide for transverse slot radiator
in upper wall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
Fig. 8.10 Relative power radiated as a function of frequency for a
centred transverse slot in a waveguide feed (solid curves) and
a periodic waveguide feed (dashed curves) with slot length
l as parameter (l = 30, 40, 42, 46 mm: w = 4.5 mm)
Waveguide WG10: a = 72.14 mm, b = 34.04 mm
Slow-waveguide: a = 72.14 mm, b = 12.24 mm, h = 21.8 mm,
G = 4.7625 mm, t = 1.5875 mm. . . . . . . . . . . . . . . . . . . . . . . 201
Fig. 8.11 Beam scan as a function of frequency for a slow-wave
fin-loaded waveguide feed comprising six slots of the same
size, equi-spaced by 65.4 mm (solid curves = theory,
diamond markers = measurement) . . . . . . . . . . . . . . . . . . . . . 203
Fig. 8.12 Radiation from a leaky wave structure . . . . . . . . . . . . . . . . . . 204
Fig. 8.13 Guide wavelength and attenuation coefficient as a function of
slit width d in a rectangular waveguide with a = 22.86 mm
and b = 10.16 mm at a frequency of 9.4 GHz. (Dashed curves
= approximate solution: Solid curves = exact solution) . . . . . . 207
Fig. 8.14 Waveguide linear array antennas employing transverse
(series) slots. a Travelling-wave array b leaky wave array . . . 209
Fig. 8.15 Pictorial representation of the mutual coupling mechanism for
a shunt slots and b transverse or series slot in a linear
waveguide array antenna . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
Fig. 8.16 Attenuation coefficient as a function of frequency for the
dominant mode of a leaky ladder of slot radiators
(with L = a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
Fig. 8.17 Brillouin diagram for the dominant mode of a leaky ladder
of slot radiators: w/a = 0.0016, S/w = 5 . . . . . . . . . . . . . . . . . 212
Fig. 9.1 Plane wave illumination of passive periodic array of current
elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
Fig. 9.2 Electromagnetic wave scattering for FSS illuminated by a
plane wave . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
Fig. 9.3 a Focused reflector antenna, b unfocused slot/patch array,
c reflectarray focusing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
Fig. 9.4 Basic dimensions for 10 cm  10 cm reflectarray . . . . . . . . . 225
xxviii List of Figures

Fig. 9.5 Array phasing to form a broadside beam for a 30  30 square


reflectarray (0° = dark blue, 360° = brown) . . . . . . . . . . . . . . 226
Fig. 9.6 Typical reflection phase versus patch size for 28 GHz
reflectarray [3] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
Fig. 9.7 COMSOL model to determine the patch reflection phase
when embedded in an array . . . . . . . . . . . . . . . . . . . . . . . . . . 228
Fig. 9.8 Typical H-plane a and E-plane b radiation patterns
for a 77 GHz, 6 in. square reflectarray, formed from
0.005 in. thick Duroid substrate with 5776 linearly
polarized patch elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
Fig. 9.9 Square patch reflectarray fabricated in microstrip
and illuminated by a Cassegrain sub-reflector
(see Yoon [13]) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
Fig. 9.10 2-D schematics of possible all-metal reflectarray antennas
a grooved metal plate, b hollowed ‘pill-boxes’ raised from
surface of metal backing plate, c stepped or pixelated metal
plate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
Fig. 9.11 Multiple groove flat metal-only reflectarray—solution
options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
Fig. 9.12 Phase matching between all-metal reflect array and equivalent
parabolic reflector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
Fig. 9.13 Reflection phase computations for three all-metal reflectarray
options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
Fig. 9.14 Prototype all-metal reflectarray formed from 5961 rectangular
grooves machined into a 30 cm diameter base plate
(from [5]) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
Fig. 9.15 Cassegrain illuminated all-metal reflectarray with pixelated or
stepped periodic surface (from [19]) . . . . . . . . . . . . . . . . . . . . 238
Fig. 9.16 Plane wave illumination of 3D-printed reflectarray in off-set
receiver geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
Fig. 9.17 Phase and magnitude of reflection coefficient (q) for a
dielectric block in 3D-printed reflectarray . . . . . . . . . . . . . . . . 239
Fig. 10.1 Radiation mechanism for a partially reflective surface
on an infinite ground plane a E-plane (y-z plane);
b H-plane (x-z plane) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
Fig. 10.2 Ray optics representation of PRS cavity antenna . . . . . . . . . . 246
Fig. 10.3 PRS antenna near-field 2 mm above periodic surface.
Note the concentration of energy towards the mid-region
(gridded) for a surface modelled as unrestricted in x and y
(from [4]) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
List of Figures xxix

Fig. 10.4 Sketch depicting the fundamental difference between a PRS


antenna and a leaky-wave antenna a PRS antenna with
Fabry-Perot cavity formed by a PRS in parallel with a perfect
electrical conductor (PEC) or ground plane and separated by
a distance of k/2 b LWA employing a Fabry-Perot cavity
formed from a PRS above a high impedance surface
(HIS) and separated by a distance < k/2 (h = k/4 if the HIS
functions as a perfect magnetic conductor (PMC)) . . . . . . . . . 250
Fig. 10.5 Fabry-Perot resonator antenna a illuminated by a plane
wave from a distant test dipole at r, h, /b PRS comprising
a periodic array of slots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252
Fig. 10.6 Electromagnetic wave field components at CRA surface
when illuminated by a plane wave and transmission-line
equivalent circuits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
Fig. 10.7 Scanned E-plane (a), and H-plane (b) radiation patterns for a
slot based CRA with a dielectric filled resonator (er = 2.2) for
which L = 0.6 cm, w = 0.05 cm, p1 = 1 cm and p2 = 0.3 cm.
The CRA is resonant at 11 GHz for d = 0.85 mm. Over the
frequency range depicted the height (d) varies from d 
k=2 to d  k=3 (from [6]) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
Fig. 10.8 Schematic of a 2-D leaky-wave Fabry-Perot resonator
antenna fed by square patch at centre of cavity
(from [8]) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
Fig. 10.9 CRA with a phase agile HIS and b typical phase agile cell
incorporating varactor diode switches . . . . . . . . . . . . . . . . . . . 264
Fig. 10.10 Schematic of a high directivity PRS antenna with additional
beam scanning in the H-plane (from [15]) . . . . . . . . . . . . . . . 265
Fig. 10.11 Schematic depicting diode modified patch element for typical
reconfigurable patch based PRS antenna (from [15]) . . . . . . . 266
Fig. 10.12 a Schematic of 1-D reconfigurable leaky-wave antenna
b detail of HIS cell (from [17]) . . . . . . . . . . . . . . . . . . . . . . . 268
Fig. 10.13 Schematic representation of the Brillouin diagram for a
parallel plate waveguide loaded with transverse periodic
surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
Fig. 10.14 Computed beam scan angles for 15 GHz LWA’s design
to provide 10° beamwidths at scanned angles of 5° (blue long
dashed trace), 20° (green solid trace), 40° (brown dashed
trace) and 60° (red short dashed trace) . . . . . . . . . . . . . . . . . . 270
Fig. 10.15 Top and side views of a schematic representation of a planar
leaky-wave array antenna formed from a tunable textured
surface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271
Fig. 10.16 Plane wave scattering at a passive metallic screen supporting
a periodic array of apertures separated by distance d . . . . . . . 273
xxx List of Figures

Fig. 10.17 Typical FSS formats for CRA’s with (a) and (b) forming
reflecting surfaces while (c) and (d) present transparent
surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
Fig. 10.18 Elemental scatterers for periodic surfaces described as
metamaterials fabricated on copper coated planar dielectric
substrates. a Centre connected, b loop types, c patch types,
d combinations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
Fig. 10.19 Plane wave refraction at a smooth flat interface between
materials with differing refractive index. a Regular materials,
b regular material in contact with metamaterial . . . . . . . . . . . 277
Fig. 10.20 Surface reflection and refraction at (a) a conventional
interface and (b) an interface with a metamaterial. In both
qffiffiffiffiffiffiffi
cases n1=2 ¼  ll2 ee21 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
278
1

Fig. 10.21 Conventional optical lens compared with ‘flat’ lens. For
a glass lens in air (lr = 1, er = 1) and for convex lens (lr = 1,
er > 1) and for b the flat lens in air it has lr < 0, er < 0 . . . . 280
Fig. 11.1 Phase conjugation through mixing . . . . . . . . . . . . . . . . . . . . . 287
Fig. 11.2 Phase conjugation circuit employing phase comparison . . . . . 288
Fig. 11.3 Phase conjugation using PLL chip . . . . . . . . . . . . . . . . . . . . . 289
Fig. 11.4 Phase conjugation using I and Q processing . . . . . . . . . . . . . . 290
Fig. 11.5 Retro-directive transmit/receive array with orthogonally
polarised patch radiators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292
Fig. 11.6 Schematic of dual-mode dual-polarised patch antenna
parasitically fed from a non-resonant X-slot by means of
microstrip lines above and below the slot . . . . . . . . . . . . . . . . 294
Fig. 11.7 Schematic of a dual-mode dual-polarised patch antenna
parasitically fed from a non-resonant X-slot by means of
balanced microstrip lines above and below the slot to improve
bandwidth and axial ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
Fig. 11.8 Schematic of dual-mode dual-polarised circular aperture array
antenna parasitically fed from balanced microstrip lines
above and below the four elements . . . . . . . . . . . . . . . . . . . . . 296
Fig. 11.9 Schematic of dual-mode dual-polarised patch antenna.
a Substrate 1—copper patch L1  L1, thickness t1, relative
permittivity er1, b substrate 2—copper patch L2  L2,
thickness t2, relative permittivity er2, c two dumbbell
coupling slots in top face, 50X feed lines in lower face (see
Gao [26] for detail), substrate thickness t3, relative
permittivity er3, d edge view showing air/polystyrene layer
between substrate 1 and substrate 2, thickness t4, relative
permittivity er4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 298
Fig. 11.10 Circularly polarised dual-mode square patch antenna with slot
coupling and microstrip hybrid feed arrangement . . . . . . . . . . 299
List of Figures xxxi

Fig. 11.11 Dual-mode dual-polarised microstrip patch antenna


implemented using circular patches for enhanced axial ratio
bandwidth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 300
Fig. 11.12 Phase conjugating active antenna element incorporating a
balanced quasi-optical mixer . . . . . . . . . . . . . . . . . . . . . . . . . . 301
Chapter 1
Evolution of Compact Slot Antennas

1.1 Free-Space Waves and Transmission Lines

The aim of this chapter is to introduce to readers the idea that even quite complex
geometrical arrangements of slots or patches in compact array antennas can be
evaluated and understood, at least initially, by empirical or heuristic methods, based
on a good understanding of properties of the most basic electromagnetic wave
solution governed by the Maxwell equations. The form of this solution is generally
referred to as the plane wave or the transverse electromagnetic (TEM) wave. In fact
this understanding need not extend beyond the underlying physics of plane waves
in order to permit satisfactory comprehension of radiation sources and the influence
of boundaries and antenna geometries on radiation patterns. This empirical exercise
actually represents a very useful preliminary stage in the antenna design process,
particularly since today novel designs are generally procured through the imple-
mentation of simulations on full-wave software analysis tools such as HFSS,
COMSOL, or CONCERTO for example. Of course the efficient application of these
tools, is greatly enhanced if the user has acquired a thorough appreciation of the
mathematical procedures which are incorporated into such simulations. This rec-
ommendation is addressed in much more detail in Chaps. 4 and 5.
Readers of this text are assumed to have some background in electromagnetic
theory, at least to the level of familiarity with the formulation of electromagnetic
wave solutions from the Maxwell’s equations for passive environments such as
free-space, simple transmission lines and regular waveguides. For those requiring a
memory boost and who have limited access to alternative texts (see Bibliography),
Appendices A and B contain summaries of the requisite material.

© Springer Nature Switzerland AG 2019 1


A. J. Sangster, Compact Slot Array Antennas for Wireless Communications, Signals
and Communication Technology, https://doi.org/10.1007/978-3-030-01753-8_1
2 1 Evolution of Compact Slot Antennas

1.2 Plane Electromagnetic Wave in Free-Space

The plane electromagnetic wave is, in words, a radio-wave, or even a wave at light
frequencies, in open space where reflecting and refracting boundaries are absent.
The essence of the wave’s field structure is depicted in Fig. 1.1.
In Fig. 1.1, E and H represent the electric and magnetic field intensities in the
region of interest. We can also define D = eE where D is the electric flux density
and B = lH where B is the magnetic flux density. Note that in vacuum, or air,
permittivity e = e0 and permeability l = l0, while the phase coefficient
pffiffiffiffiffiffiffiffiffi
k ¼ k0 ¼ x l0 e0 ¼ x=c. Fuller parameter definitions are provided in Appendix A.

Fig. 1.1 Representation of a simple sinusoidal plane electromagnetic wave showing the
fundamental orthogonal relationships between the vector electric (E) and magnetic (H) fields
and the direction of propagation (k)
1.2 Plane Electromagnetic Wave in Free-Space 3

The mathematical expression for the general free-space wave (see Appendix A) is:
Ex ¼ Af ðz  ctÞ þ Bf ðz þ ctÞ ð1:1Þ

where f is an unspecified function for a wave travelling in the z-direction which is


aligned with k (Fig. 1.1). For a sinusoidal wave, Eq. (1.1) is replaced by:
Ex ¼ A sinðz  ctÞ þ B sinðz þ ctÞ ð1:2Þ

The first term in Eq. (1.1) represents a wave travelling in the +z direction while
the second allows for a reflected wave, if such exists. In this case a standing wave is
formed on the mismatched line. Given that velocity is the rate of change of z with
respect to time, it is evident that c represents velocity (actually phase velocity) of
the electromagnetic wave in ‘free space’. For vacuum it is equal to 3  108 m/sec.
The application of Maxwell’s equations also gives Hz = 0 and:
A B
Hy ¼ f ðz  ctÞ þ f ðz þ ctÞ ð1:3Þ
g0 g0

Also
rffiffiffiffiffi
Ex l0
¼ ¼ g0 ð1:4Þ
Hy e0

η0 is termed the free space wave impedance, which for air or vacuum has the
value 120pX. The resultant solution is a plane electromagnetic wave, also termed a
transverse electromagnetic (TEM) wave, for which E and H are transverse to the
direction of propagation and orthogonal to each other. E and H are also in time
phase, as Eqs. (1.1) and (1.3) attest (see Fig. 1.1). If time phase between E and H
prevails, which it does in lossless free-space, the power flow density along the
direction k is given by:
jE jjH j
p¼ Watts/m2 ð1:5Þ
2

The plane wave fronts in Fig. 1.1b lie transverse to k and, in theory, cover the
entire, theoretically infinite, transverse plane. On any given plane, E and H are of
constant magnitude, represented by equally spaced vectors. Further, the magnitudes
of E and H vary in accordance with Eqs. (1.2) and (1.3) respectively, as we move
from one plane to the next.
While this plane wave solution to the Maxwell equations evidently provides a
useful guide to the nature of electromagnetic waves in free-space, the importance of
this wave structure, from an antenna perspective, is that it also describes electro-
magnetic wave propagation in a variety of transmission systems which are
important antenna feed systems. The obvious reason for this beneficial position lies
in the fact that guided electromagnetic waves can be procured by simply trapping
the TEM wave [see Bibliography] between a pair of flat conducting surfaces that lie
4 1 Evolution of Compact Slot Antennas

normal to the electric field vectors. Surface charge on the conductors instanta-
neously redistributes itself, thus ensuring that the boundary conditions on both
normal D and tangential H are satisfied. That is the charge adjustment ensures that
(using unit vector notation):

^D¼q
n and ^H¼J
n ð1:6Þ

where q and J are respectively the surface charge density (Coulomb/m2) and the
surface current density (A/m) on the metal surface. Following convention, the unit
vector is directed normal to the boundary surface.

1.3 TEM Transmission Lines

The most common TEM transmission line alternatives are shown in Fig. 1.2.
Parallel plate waveguide (Fig. 1.2a) provides only limited guidance in directions
other than normal to the plates, but clearly shows how the insertion of smooth
conducting planes normal to the E-field has negligible effect on the propagation
conditions for the TEM mode. The resultant surface currents, dictated by the
boundary conditions, are represented by blue dashed lines. Full trapping is provided
by coaxial line (Fig. 1.2b) but at the expense of phase velocity reduction associated
with the unavoidable dielectric ‘separator’ between the inner and outer conductors.
The potential for power loss in this dielectric medium is a drawback particularly in
flexible coaxial cables. The axial current flow on the conductor surfaces are again
represented by blue dashed lines. Stripline (Fig. 1.2c) is essentially ‘flattened’
coaxial line and has the advantage of ease of fabrication of intricate transmission
networks using printed circuit board (pcb) techniques. In coaxial line and in stripline
the dielectric separator usually displays a relative permittivity (er) of between 2 and
3. By increasing this to between 6 and 10 in microstrip (Fig. 1.2d), it becomes
possible to dispense with the upper ground plane and create an open structure into
which microwave components can relatively easily be inserted. The primary dis-
advantage, as the schematic suggests, is high dielectric losses and radiation losses.

1.3.1 TEM Transmission Line Fed Radiating Elements

The location of surface currents in the TEM transmission lines of Fig. 1.2 has been
highlighted for antenna purposes. In order to create or enhance radiation from such
lines it is generally necessary to disrupt the current flow directions. This is illus-
trated perfectly in Fig. 1.3 for microstrip line. This is the version from which
radiation is most easily extracted.
The slot is located in the ground plane symmetrically and directly above (or
below) the microstrip line and orthogonal to it, although this need not be the case.
1.3 TEM Transmission Lines 5

Fig. 1.2 Trapped TEM mode


transmission lines systems

For illustrative purposes, the axial TEM mode currents are assumed to be
approximately constant over a length of line which can be presumed to be less than
k/4 long, where k is the TEM wavelength. The currents on the ground plane are
strongest directly above the strip and consequently the slot exercises maximum
6 1 Evolution of Compact Slot Antennas

Fig. 1.3 Slot radiator located


in the ground plane of
microstripline

disruptive effect. Near the centre of the slot the current is totally impeded and
induces a build-up of charge on the slot rims, shown by the positive and negative
symbols. This is not unlike current flow into a capacitor, and as with a capacitor,
electric field is formed in the gap. On the other hand, near the slot extremities,
current paths can continue to flow by ‘bending’ around the slot ends. This increase
in current path length, as in a coil, introduces inductance into the equation. Hence
the slot can be modelled electrically as a two wire transmission line shunted by a
parallel combination of a capacitor and an inductor. If the slot length is much less
than half the free-space wavelength (<ko/2) the inductance (L) is negligible and the
slot is largely capacitive (C) in nature. However, as the slot length increases toward
ko/2 the inductance begins to match the capacitance and resonance occurs. In
addition at this point the slot radiates strongly with the E and H vectors in the slot,
at any given instant in time, aligned to produce an outwardly directed Poynting
vector (Appendix A—Eq. A.19). In modelling terms, a resistive component (Rrad)
may be included along with the L and the C for long slots.
An alternative scheme for inducing radiation from a microstrip line, which is,
perhaps, more in sympathy with its basic electrical characteristics, is the microstrip
patch antenna. In its simplest and original form the usually 50X characteristic
impedance of the feed line is significantly reduced by abruptly widening the line in
a step transition (see Fig. 1.4a). If the wide portion is then terminated in an
open-circuit, a mechanism for resonance by reflection is created between the
open-circuit and the abrupt transition. The transition, viewed from within the very
low impedance patch, also presents a high impedance reflective discontinuity.
Consequently, if the distance between the open-circuit termination and the step
junction is set to k/2, where k is the TEM wavelength within the microstrip, the first
standing wave resonance is trapped in the patch as shown in Fig. 1.4b, c. The patch
is usually chosen to be approximately ko/2 in width for optimum radiation from the
patch edges. The two ‘edges’ can be viewed as ‘patch slots’. In directions normal to
the patch face the slot pair radiate in phase developing maximum power density as
shown in Fig. 1.4b. Note that the feed line can be matched to the resonant patch by
means of rectangular inserts as shown in Fig. 1.4a. The technique is not difficult for
a radiating patch. In this case, in the absence of a reflection, the line current or
voltage amplitude is unvarying in the feed line as shown. Finally relative to the
1.3 TEM Transmission Lines 7

Fig. 1.4 Fundamental mode of operation for patch antenna, a top view of patch antenna, b side
view showing trapped E-field—note edge radiation can be viewed as radiation from a two element
slot array, c voltage and current variation in patch and line

propagation direct (x in this case) the trapped mode under the patch displays a one
half-cycle variation in x, zero in y and zero in z, so the mode is termed TE100. Note
that if the x-direction were chosen to be normal to the patch surface then the same
mode would be termed TM100.

1.3.2 Radiating Element with Polarisation Agility

Hopefully, it has become clear from the above discussion, that it is possible to
evaluate the radiation mechanism associated with TEM transmission line fed
compact antennas, by performing a heuristic examination of the disruption caused
to current flows and to E and H field directions in the TEM line. Examples of
applying this kind of cogitative process are becoming quite common in the liter-
ature assisted by visual field presentations constructed from electromagnetic
8 1 Evolution of Compact Slot Antennas

simulations [1–3]. In reference [4] a wide, but by no means exclusive, range of


possible compact antenna geometries based on microstrip line technology has been
developed. It has arguably been, and continues to be, a promising source of stimuli
for novel compact antenna designs.
To emphasise the potential usefulness of the technique described here we apply
it to a geometrically quite complex slot radiator fed from microstrip line, which is
envisaged as providing polarisation agility in a compact inexpensive radiating
element for application in an array antenna environment. It comprises a polarisation
agile, T-shaped slot powered from microstripline. By insightful application of
boundary conditions, and with particular attention to the slot field and surface
current density distributions, the evolution of the antenna geometry is fully
explained. The arrangement, which is depicted in Fig. 1.5, aims to achieve polar-
isation agility by independently exciting a pair of orthogonally aligned slots from
microstrip. Thus the slot pair can radiate axial or transverse linear polarisations if
excited separately (see Fig. 1.6), or if excited simultaneously and in phase

Fig. 1.5 T-slot radiator fed


from balanced microstrip line
1.3 TEM Transmission Lines 9

Fig. 1.6 Mechanism for delivering polarisation agility based on microstrip

quadrature circularly polarised radiation is produced [5]. The T-shaped slot is


machined into the metal back-plate depicted in the Fig. 1.5. This plate is located in
close proximity to back-to-back ‘balanced’ microstrip lines so that the slots are
immersed in the fringing edge magnetic fields of these lines.
The fringing fields from the balanced microstrip lines induce z-directed current
flows in the back-plate and hence an electric field in the transverse leg of the T-slot
which in turn initiates the mechanism for radiation, as illustrated in Fig. 1.6. The
balanced microstrip line can be operated in two modes, namely an even mode with
the TEM mode E-fields in phase across the ground plane and an odd mode when
they are in antiphase. The resultant fields in the slot are depicted in Fig. 1.6 which
shows that the odd mode sets up transversely (y-directed) polarised radiation while
the even mode generates axial (z-directed) polarisation. It should be note that the
incident TEM mode, whether odd or even, directly excites only the ‘vertical’
y-directed, or transverse, component of the T-slot essentially through the fringing
magnetic field components and the resultant z-directed current which is interrupted
only by this component of the T-slot. The axial portion of the T-slot cannot be
excited directly be a TEM mode.
In the case of even-mode excitation, while the upper half of the transverse slot
component is, at the instant shown, immersed in a downward (y-directed) magnetic
field, the opposite is the case for the lower half of the slot. The E-fields in the upper and
10 1 Evolution of Compact Slot Antennas

lower portions of the transverse slot are in antiphase so that the slot as a whole is
predicted to radiate very weakly. On the other hand, these antiphase slot fields in the
transverse slot can combine at the junction with the axial slot, producing a strong
parasitic excitation of this slot, if it is long enough. Radiation which displays y-directed
polarisation is the result. The septum slot, or clearance slot, in the microstrip ground
plane is introduced to minimise E-field suppression at the junction by proximity of the
ground plane edge. For odd-mode excitation of the balanced microstrip line the E-fields
in the upper and lower portions of the transverse slot are predicted to be in phase
(Fig. 1.6) with negligible parasitic coupling into the axial slot. Radiation is therefore
almost entirely from the transverse slot generating z-directed polarisation. In principle,
the above radiator is capable of generating circularly polarised radiations, in addition to
the linear possibilities already indicated. To achieve this the exciting TEM mode
components will require to be in quadrature phase.
Based on a moment method study [5] of the proposed T-slot geometry the
potential of this novel and compact slot to generate effective polarisation agility has
been established quantitatively, by modelling and measuring its radiation and
polarisation properties. Electromagnetic modelling tools based on the moment
method and focused on the scattering properties of slot radiators are comprehen-
sively developed in Chaps. 4 and 5.

1.4 Waveguide Feed System

It is perhaps not too unreasonable to suggest from a cursory scanning of the


literature that compact antennas based on slots radiators are much more commonly
fed from TE mode waveguide than from TEM transmission lines. So, following the
same empirical route pursued in our examination of TEM line excited radiators, we
shall now develop, from a plane wave perspective, the modal field and current
patterns for rectangular waveguides, which are essentially hollow metallic pipes.
These patterns are subsequently studied with the aim of conducting empirical
explorations of a range of typical waveguide excited slot radiators.
Snell’s laws can be used to explore the processes behind electromagnetic wave
trapping, or guidance, concepts which are needed in later chapters. While at low
frequencies, as we have seen, TEM waves can be guided by a pair of conductors,
such as in power lines, in parallel wire telephone lines, or in coaxial lines, high
frequency wave trapping in hollow conducting pipes is not so easy to comprehend.
Such waveguides are increasingly being used in many of the compact antenna
configurations employed in mobile communications. This method of guidance is
very efficient, and is especially applicable to high power transmission at microwave
frequencies. It relies on the nature of plane wave interference patterns and can,
perhaps, best be explained by consideration of Figs. 1.7 and 1.8.
1.4 Waveguide Feed System 11

Fig. 1.7 Plane wave interference pattern

Fig. 1.8 Interference patterns of plane electromagnetic waves reflecting off a flat perfectly
conducting surface
12 1 Evolution of Compact Slot Antennas

1.4.1 Trapped TE/TM Waveguide Modes

Figure 1.7 depicts (in two dimensions for simplicity) a pair of plane electromag-
netic waves (TEM waves) of equal magnitude travelling in different directions A
and B. The waves are represented by their wavefronts, with the wave peaks in each
case denoted by solid transverse lines (planes in 3-D) and wave troughs by dashed
lines. The distance between a wave peak and wave trough is, of course, half of the
free-space wavelength (ko/2). The waves are travelling at the velocity of light [c] in
the directions of the large arrows. On examination of this wave pattern it is not too
difficult to observe that along the horizontal chain-dotted line (or in three dimen-
sions—the y-z plane) peaks of wave A coincide with those of wave B, and troughs
coincide with troughs—and this is independent of the movement of the waves. This
line (or plane) represents a stationary (in the x-direction) field maximum ‘inde-
pendent of time’ while the waves continue to exist.
In contrast, along the green dashed line, peaks of wave A coincide with troughs
of wave B, and vice versa, resulting in a stationary field null at these positions.
Consequently, if a perfectly conducting sheet of infinite extent, orientated normal to
the x-axis, is located at the stationary null position (Z-Z) the field pattern remains
unchanged (see Fig. 1.8). For a sheet at the upper dashed line, the red direction
arrow then represents an incident wave and the blue arrow a reflected wave, which
according to Snell’s laws at a perfect mirror, reflects with a magnitude equal to the
incident wave and at an angle such that hr = hi, as is required to retain the pattern.
For perpendicularly polarised plane waves with the E-field confined to the
y-direction, the E-field pattern forms a cosine distribution between the null planes.
This pattern can be trapped or guided by introducing a second conducting sheet at
the lower null locus in Fig. 1.8 (green dashed line). The trapped pattern travels in
the z-direction with a phase velocity:

vp ¼ c= sin h ð1:7Þ

and a wavelength:

kp ¼ ko = sin h ð1:8Þ

The phase coefficient for this trapped wave bp ¼ 2p=kp and hence

bp ¼ 2p sin h=ko ¼ ko sin h ð1:9Þ

c is the speed of light, ko is the TEM wavelength in free-space, ko is the phase


coefficient for the constituent plane wave and h ¼ hi ¼ hr . The magnetic field
distribution can easily be deduced by applying trigonometrical rules and the total
E-H pattern is termed a TE (Transverse Electric) guided wave. Alternative TM
(Transverse Magnetic) guided wave can be formed by commencing with parallel
polarised TEM components (For mathematical details see Appendix B).
1.4 Waveguide Feed System 13

Fig. 1.9 TE10 mode field


patterns for rectangular
waveguide: width a, height b,
a E and H-fields in the y-z
plane, b E-field in the x-y
plane, c E and H-fields in the
x-z plane (E-field = green
solid lines: H-field = blue
dashed lines: current density
J = orange chain-dotted lines)

A TE wave between parallel conducting planes separated by a distance a is


illustrated in Fig. 1.9. The sinusoidal field variations in x are clearly shown in cases
(b) and (c). The relationship between plane separation a and wavelength ko can
again be deduced from trigonometry and yields:

mko
cos h ¼ ð1:10Þ
2a

m is the number of half-sinusoids of field pattern between the null planes. In


Fig. 1.8 it is not necessary to choose the nearest null planes to create a trapped
pattern. Equation (1.10) only has meaning for mko \2a, so that for m = 1 the case
14 1 Evolution of Compact Slot Antennas

depicted in Fig. 1.9, the free-space wavelength must be less than 2a for propagation
to occur. The corollary is that the frequency of the wave f (=1/ko) must be greater
than a certain critical value or cut-off value corresponding to the cut-off wavelength
kc ¼ 2a. Furthermore if a\ko \2a the m = 2, 3, 4, solutions all yield the
impossible requirement that cos h [ 1. This means that in the prescribed frequency
range only the m = 1 solution is possible. The solution is termed the dominant
mode for the parallel plane waveguide of separation a, and is defined as the TE10
mode—with one E-field variation in x and zero variation in y. This mode is shown
in Fig. 1.9a–c. Perfectly conducting ‘lids’ can be introduced at y = 0 and y = b to
form a rectangular waveguide, without altering the pattern, because the E-field is
normal to these walls, and the H-field induces surface current density flows
(J) which ensure magnetic field boundary conditions are satisfied. These current
flows are shown in Fig. 1.9b, c. The b-dimension is usually chosen to be approx-
imately half the a-dimension to maximise bandwidth.
Equations (1.7), (1.8) and (1.9) embody the essential features of the frequency
characteristics of the dominant TE10 mode. Equation (1.9) implies that at cut-off
(h = 0) for the TE10 mode bp = b10 = 0, while at very high frequencies well above
cut-off where h = 90° then b10 ! ko . At the same time from Eq. (1.7) we observe
that vp10 ! c. This behaviour is summarised in the Brillouin diagram presented in
Fig. 1.10, which for a waveguide of width a shows the TE10 mode cutting-on at
koa = p. Thereafter it grows in an parabolic manner (see Appendix A.2) becoming
increasingly asymptotic to the vp = c line. Close to cut-off where b10 approaches

Fig. 1.10 Brillouin diagram showing graphically the frequency characteristics of the TE10 mode
in rectangular waveguide (orange line denotes b10 = ko and thus vp = c for all frequencies: blue
solid line = b10 for a rectangular waveguide of width a)
1.4 Waveguide Feed System 15

zero the guide wavelength kg10 ! 1 making wavelength-dependent component


design impossible. Consequently this region is avoided in engineering practice. The
practical frequency range is generally assumed to run from koa = 1.25p
(f = 1.25fc10) to koa = 2p where the TE20 mode starts to propagate. Note that in this
region the TE10 mode plot is relatively linear thus minimising the frequency dis-
persion which can be an awkward feature, in communication terms, of hollow metal
waveguide.
Apart from the TE10 mode, in the dominant mode frequency range, all TEmn
modes for which ð1 \ m \ 1; 0 \ n \ 1Þ and TMmn modes for which
ð0 \ m \ 1; 0 \ n \ 1; m ¼ n 6¼ 0Þ are cut-off and decay exponentially
from the electromagnetic wave source. These modes are referred to as evanescent
modes and carry no power. They account for stored energy which accumulates
around discontinuities in the waveguide in order that the total field structure sat-
isfies the boundary conditions there. Furthermore, this doubly infinite set of
orthogonal TE and TM modes is mathematically ‘complete’ and enables full wave
modelling of waveguide discontinuities such as slot radiators, as we shall see in
Chaps. 4 and 5. There we develop a moment method model of the slot discontinuity
problem. Texts 4 and 9 in the Bibliography, provide full and comprehensive
treatments of EM modelling techniques.

1.5 Waveguide Slot Radiators

Arguably, waveguide fed slot radiators have provided a very significant source of
ideas for recent developments in compact antenna technology. However, early
embodiments, such as those depicted in Fig. 1.11, were impelled by large antenna
arrays for military radars [6–8]. Rapid developments in the 1940s were driven by

Fig. 1.11 A hollow


rectangular TE10 mode
waveguide showing some
possible slot locations
required to secure radiation.
One guide wavelength (kg10)
of the modal current pattern is
depicted
16 1 Evolution of Compact Slot Antennas

the demands of World War II for long range radio detection systems. The fre-
quencies used were generally below or at the low end of the microwave range, so
the waveguide and the resultant antennas were inevitably large. Nevertheless,
war-time advances on slotted waveguide array antennas laid important and reliable
foundations on the electromagnetic issues associated with waveguide fed slots and
waveguide slot arrays [8]. This early work continues to inform current develop-
ments, as hinted at in Chaps. 4, 5, 6 and 7.

1.5.1 Conventional Waveguide Slot Radiators

Traditional slot array antennas generally employ either longitudinally directed offset
‘shunt’ slots, or transverse centrally located ‘series’ slots, in the broadwall of the
rectangular waveguide feed. Inclined slots located in the sidewall of waveguide are
also common (see Fig. 1.11). Once the TE10 mode wall current flows have been
established, as shown in the Fig. 1.9, the slot excitation mechanism is not too
different from that of the TEM mode examples. Current flow interruption by the slot
induces charge build up on its long edges resulting in an instantaneous E-field
formation across the narrow dimension of the slot. This slot field changes in
sympathy with the modal pattern as it passes through the waveguide at the phase
velocity given by Eq. (1.7) (or Eq. B.15 in Appendix B). At a given incident power
level in the propagating mode, the slot radiation is maximised by operating it near
its resonant length, namely L = ko/2.
Shunt slot in waveguide
Figure 1.11 also reveals that the shunt slot is weakly excited close the waveguide
centre line where the current flow is largely axially directed, and strong near the
side wall where the current flow is almost entirely transverse. This behaviour is
used to control radiation strength. The scattering nature of waveguide fed slot
radiators in explored in detail in Chap. 5.
For the shunt slot, which is z-directed in Fig. 1.11, the induced electric field is
clearly aligned with x axis. Therefore, in the direction of the pattern maximum,
this orientation of the slot dictates that the radiation from it is polarised in a
direction transverse to the axis of the waveguide—usually referred to as trans-
versely polarised. This also true of a slot at position (B) which is separated axially
from the slot at (A) by kg10/2 and on the opposite side of the centre line. In phase
terms the slots are separated by 360° which can be interpreted as 0°. This means
that the slots radiate in phase in the y-direction. The importance of this feature is
outlined in Chap. 7.
The radiation pattern, when the slot is part of a planar array for example, is
predominantly directed into the forward half-space (y > b) since the array surface
presents a large conducting ground plane to the slot. It can be viewed as a
‘half-dough-ring’ in shape with nulls in the axial direction (± z-direction), and
1.5 Waveguide Slot Radiators 17

Fig. 1.12 Computed radiation pattern for a six element shunt slot waveguide fed array (with
acknowledgement to www.remcom.com)

displaying omni-directional power density distribution in the radial direction in the


x-y plane. These features can be discerned from Fig. 1.12 which displays a com-
puted radiation pattern for a linear array of six shunt slots.
The structure of a typical shunt slot array, in planar form, is illustrated in
Fig. 1.13. If fabricated on a programmed lathe, the top plate and base plate are
generally created separately. The sizes, shapes and locations of the slots forming the
radiating slot array, are determined from simulation on an electromagnetic solver,
and these dimensional details are programmed for automatic machining into a
metallic top plate of predetermined thickness. The base plate, on the other hand,
carries the network of feed waveguides, again with dimensions predetermined by
simulation, enabling computer controlled machining on a thick metallic (usually
aluminium) work piece. In the example shown the eight parallel grooves forming
the waveguide feed structures are shorted at each end forming resonant cavities.
This arrangement is narrow band (see Chap. 7) but very cost efficient. Power is
leaked into the resonant cavities from the transverse waveguide by narrow apertures
in their end walls. The coupling apertures are separated by kg10/2 in the transverse
member which means that the feed waveguides are excited in antiphase as
Fig. 1.13b indicates. To ensure that the radiating slots are all radiating in-phase at
the operating frequency the wall current directions are exploited to ensure that the
slot staggering takes account of this. The transverse member is, in turn, powered
through a T-junction formed by introducing an input waveguide from behind the
array feed block. The ‘choke’, which is included in Fig. 1.13, comprises a quarter
wavelength deep groove machined into the base plate approximately a quarter
wavelength from the waveguide recess. It ensures that radio frequency leakage
between the base plate and the slotted lid is negligible.
18 1 Evolution of Compact Slot Antennas

Fig. 1.13 Planar shunt slot


resonant waveguide array,
a isometric view, b end
sectional view

Such an array can be surprisingly compact in high frequency sensor applications.


For example, at mm-wave frequencies (circa 100 GHz) and employing (say) WG27
(2.4 mm  1.2 mm) a sensor array could exhibit dimensions typically in the range
2.8 cm wide, 3.8 cm long and 2.5–3 mm thick. The radiating slots would be of
length *1.5mm and *0.2 mm wide. Conventional slot arrays of the type illus-
trated in Fig. 1.13 are examined fully in Chap. 6.
Series slot in waveguide
The transverse slot in Fig. 1.11 is noticeably less versatile, and hence less practical,
than the shunt slot. Since the slot is now excited by the longitudinal currents flows
in the broadwall of the waveguide, which replicates the axial current flow along the
wires of a two wire transmission line, the slot exhibits the behaviour of a lumped
electrical component inserted in series with the line. For a slot, which is most
1.5 Waveguide Slot Radiators 19

commonly operated close to resonance (L = ko/2), transverse positional movements


in order to control radiation strength are severely limited (L is not much smaller
than a). Furthermore, from an array perspective broadside radiation requires the z-
polarised transverse slots to be separated by the guide wavelength kg10, to procure
in-phase excitation. If the waveguide is air filled this magnitude of slot spacing has
significant and undesirable radiation pattern consequences. These are thoroughly
examined in Chaps. 6 and 7. The solution is usually to fill the waveguide with a
dielectric, or load it with a periodic surface, so that kg10 is reduced in length toward
ko/2. In a closely related application, as outlined in Chap. 8, series slots are ideal
elements for the implementation of waveguide fed leaky-wave array antennas. This
is because transverse broadwall slots, when tightly spaced axially, create periodic
ladder structures which readily support surface waves and leaky waves.
The radiation pattern for a single slot is omnidirectional in the y-z plane while in
the x-z plane nulls occur in the x direction. Apart from polarisation direction, the
pattern differs from that of the shunt slot only in radiation distribution at
wide-angles. This is attributable to divergences in the electromagnetic wave scat-
tering into the backward half-space.
Edge slot in waveguide
Planar slotted-waveguide array antennas which employ inclined edge slots have
been quite frequently reported upon in the early literature with a focus on antennas
for radar systems operating at microwave frequencies [6–8]. On the sidewall of
rectangular waveguide the TE10 mode surface current flows are directed trans-
versely to the waveguide axis, that is y-directed in Fig. 1.11. Such currents could
obviously be interrupted by z-directed slots. Strong excitation would result, but
such slots offer little control of the radiation strength if half wavelength long slots
are preferred. The usual practice therefore is to incline the slots which introduces
radiation strength control using the tilt angle (the greater the tilt the stronger is the
coupling to the TE10 mode).
Needless to say the tilted slots exhibit disadvantages and advantages in slot array
applications. Firstly, because of the tilt alone the electric field induced in the slot is
neither aligned with, nor transverse to, the waveguide axis. Consequently in order
to avoid the linearly polarised radiation being in a ‘squint’ relationship to the
waveguide axis adjacent slots in a linear array are inclined in opposite directions. It
is not difficult to deduce from Fig. 1.11 that the resultant polarisation for such an
array is axial. While the cross-polar (y-directed) components of the slot radiations
cancel in the broadside direction, strong cross-polar sidelobes of the array appear at
wide radiating angles. This represents a major disadvantage of inclined edge slot
arrays. A schematic illustration of such an array is provided in Fig. 1.14 showing
clearly the oppositely inclined slot implementation scheme. For this array, if the
waveguides are parallel to the ground, the antenna provides a horizontally and
linearly polarised radiating beam in the forward broadside direction.
20 1 Evolution of Compact Slot Antennas

Fig. 1.14 Horizontally polarised resonant waveguide planar array antenna formed from edge-slots

In principle, vertical stacking of an edge slot array should be simpler than for
broadwall slots since the waveguide height (b) is generally considerably smaller that
ko/2, the critical distance for the avoidance of grating lobes (see Chaps. 6 and 7).
However this advantage is somewhat negated by the encroachment of resonant
length edge slots into the broadwall (see Fig. 1.11), so that electromagnetic wave
leakage into the inter-guide gaps occurs. The result is radiation pattern unpre-
dictability and additional cross-polar deficiencies. With each of the slotted waveg-
uides in Fig. 1.14 depicted as terminating in a short circuit usually kg10/4 from the
last slot and the slots separated by exactly kg10/2, the sketch is representative of a
‘resonant’ array (see Chap. 7). It produces a precise high gain principal beam in the
broadside direction at the design frequency. However, high gain delivery is available
over only a very narrow bandwidth.

1.5.2 Unconventional Waveguide Slot Radiators

Omni-directional slot doublet


The heuristic approach to evaluating, or comprehending, the radiation mechanisms
developed by slots in transmission systems, can quite successfully be applied to
more complex geometries. Below we investigate several examples which are ger-
mane to later chapters in this text.
Perhaps, the simplest unconventional slot pattern in waveguide is the slot dou-
blet which has applications in medicine when embedded in a surgical implement or
1.5 Waveguide Slot Radiators 21

Fig. 1.15 Omni-directional


slot-doublet in rectangular
waveguide

as a compact, robust search and rescue transponder [9, 10]. In much the same way
as is expounded in the previous section the mode of operation of the slot doublet in
waveguide (Fig. 1.15) can be explained by knowledge of the TE10 mode wall
currents and the disruption caused to these currents by the presence of the slots.
Instantaneous charge build along the long edges of the slot is the source of electric
field in the slot which always traverses the shortest distance within it (see
Fig. 1.15). For the TE10 mode, at the instant shown where the top wall currents flow
outward to the sidewalls they flow inwards on the bottom wall. This means that the
electric field in the top wall slot is polarised in the opposite direction to the bottom
slot, and this remains true for all instances in time as the mode propagates through
the waveguide. Electromagnetic wave scattering by the upper slot produces –x po-
larised radiation into the upper half-space while the lower slot radiates into the
lower half-space with +x polarised radiation. A comprehensive simulation of the
slot doublet geometry has been formulated [9, 10] using the moment method
(Chaps. 4 and 5). It exhibits good omni-directional radiation in the transverse plane
of the waveguide (Fig. 1.16) for a low height waveguide (25% less than conven-
tional waveguide). For the pattern in the figure the ratio of the maximum to min-
imum radiation intensity (at constant radius) is of the order of 1 dB. Perhaps not
surprising, for a full height waveguide it is not nearly so good, at approximately
10 dB. Unfortunately lowering the waveguide height severely restricts power
handling capability.
Polarisation agile waveguide slots [11]
An example of a significantly more complex arrangement of radiating slots in
waveguide has been proposed for a polarisation agile 3-D antenna for a long range
22 1 Evolution of Compact Slot Antennas

Fig. 1.16 Simulated


omni-directional radiation
pattern in x-y plane for
slot-doublet radiator

ground radar [11, 12]. A typical system of this kind usually comprises a parabolic
reflector providing azimuth gain and a front fed linear array providing elevation
gain and frequency scanning (see Chap. 8). To bring about polarisation agility it is
necessary to create a means of altering the current flows in the walls of the
waveguide feed structure in order to switch TE10 mode excitation from transversely
polarised slots to axially polarised slots. One way of doing this is to employ
bifurcated waveguide as suggested in Fig. 1.17.
This particular bifurcated waveguide format can be excited from a conventional
four port ‘magic-T’ (see Bibliography: D.M. Pozar). The two output ports dovetail
with the bifurcated waveguide feed while of the two input ports one sets up the even
mode of propagation, the other the odd mode alternative. Switching power between
the two input ports represents routine microwave technology. Consequently,
polarisation switching requires a broadwall slot pattern with orthogonally directed
slots which are independently excited by the even and odd propagation modes. The
slot pattern is dictated by the TE10 mode surface current flows as illustrated in
Fig. 1.17.
Given that for effective switching slot radiators must be immersed in the even
mode as a whole, or the odd mode as a whole, there are two possibilities—a
transverse slot straddling the central wall, or bifurcation, and a longitudinal slot
running along the bifurcation. The former is excited by axial wall current flows as
we have already seen. In this case only the even mode is effective in exciting the
transverse slot with an uni-directional axially polarised electric field (see
Fig. 1.17a). For the odd mode the induced electric field in the transverse slot is
weak and in anti-phase across the centre line. The resultant radiation from this slot
1.5 Waveguide Slot Radiators 23

Fig. 1.17 Dual-mode bifurcated rectangular waveguide structure showing even mode to odd
mode polarisation switching mechanism

is relatively insignificant. Note the inclusion of a clearance slot, or septum slot, in


the bifurcation in order to minimise suppression of the slot field by the metal wall.
A longitudinally directed slot located above the common wall of the dual mode
waveguide can be excited only by the odd mode. This mode generates wall current
flows which cross the centre line and can be intercepted by an axially directed slot
(Fig. 1.17b). However, an axial slot positioned over the common wall is impractical
because of the damping effect, on the slot E-field, of the metal wall beneath it. To
overcome this problem, the simple answer is to split the axial slot into two elements
which straddle the bifurcation, as shown in the figure.
If a bifurcated waveguide array antenna, formed from slots of this description,
were mounted vertically, relative to the ground, even-mode excitation would
generate vertical polarisation, while the odd-mode would produce a horizontally
24 1 Evolution of Compact Slot Antennas

polarised radiation pattern. If the inputs to the ‘magic-T’ were fed in quadrature
phase through a 3 dB hybrid, for example, the radiation pattern would be circularly
polarised.

1.5.3 Waveguide Cavity Backed Slot Radiators

A not unrelated, but subtly different, route toward a polarisation agile slot pattern in
waveguide involves using orthogonal modes which are formed in a square waveg-
uide cavity [13, 14], rather than the even and odd mode scheme outlined above.
Rectangular waveguide is readily convertible to a high-Q cavity resonator by ter-
minating a section of waveguide with a short circuit. Transmission line theory then
informs us that this zero impedance load is replicated at every half-wavelength (kg/2)
as we withdraw from the load toward the source (see Bibliography: J.D. Kraus). If
the propagating mode in the original waveguide is the TE10 mode then by locating a
second short circuit at the nearest E-field null to the load traps a TE101 cavity mode,
while the next one traps a TE102 cavity mode—and so on. For a square cavity of
dimensions a  a  b the resonant frequency for the cavity mode can be deduced
pffiffiffi pffiffiffi
from equation (B.15). It is fixed and given by f0 ¼ c= 2a for TE101 and f0 ¼ 2c=a
for TE102. For example for WG16 for which a = 22.86 mm a TE101 square cavity
resonates at 9.3 GHz, while a TE102 cavity resonates at 18.6 GHz.
Appropriate broadwall slot locations, commensurate with strong radiation, are
indicated in Fig. 1.18. The positions shown are determined, as with the waveguide
slots discussed earlier, by the surface current density flows (J) in the upper (or
lower) inner square faces of the cavity. For the TE101 mode depicted in the left hand
diagram of Fig. 1.18, electromagnetic coupling to the cavity fields occurs equally
strongly at four possible slot locations as shown. Four slots in a balanced or

Fig. 1.18 Polarisation agile waveguide cavity backed slot antenna


1.5 Waveguide Slot Radiators 25

symmetrical arrangement are chosen to extract orthogonal linear polarisations.


However, in the broadside direction (normal to the square surface) this slot pattern
actually produces negligible radiation for TE101 mode excitation. The instantaneous
current flow directions lead to slot electric fields, in the horizontal slot pair and in
the vertical slot pair, which are opposed in phase resulting in E-field cancellation,
and hence power suppression, in the desired radiation direction. The figure suggests
that the TE101 mode can be excited from stripline tabs located beneath one or more
of the top-wall slots [14], but other mechanisms are equally possible.
While the TE101 cavity mode does not lead to a slot based antenna with
meaningful polarisation agility, this is not true for a cavity tuned to the TE102 mode,
as is illustrated in the right hand diagram of Fig. 1.18. The same balanced and
symmetrical slot pattern is employed and it is assumed that the higher frequency
TE102 mode can be excited by similar stripline tabs to those described above for the
TE101 mode. If we again follow the empirical process of examining the relationship
between the modal wall current flows and the disruptively placed slots it is not
difficult to appreciate that the slot electric field alignments are as shown in the
figure. These current flows for the TE102 mode dictate the vertical slot pair do not
radiate, since the E-field reverses within the length of each slot. On the other hand
the horizontal pair radiate strongly and in phase producing a vertically polarised
radiation pattern with a maximum at broadside. This mode and this polarisation
would be produced by a stripline feed entering the cavity either below the upper slot
or the lower slot, or both. Feeding the cavity from the side, under either the left
hand or the right hand slot sets up the TE201 mode, for which the magnetic field and
current density patterns are rotated by 90° relative to the TE102 mode depicted in
Fig. 1.18. In this case the horizontal slots are weakly excited while the vertical slots
radiate strongly, and in phase, generating a horizontally polarised broadside beam.
An antenna of this description can therefore be switched between orthogonal
linear polarised emissions by electronically switching the input signal between the
two stripline ports. Furthermore, if the ports are fed simultaneously, and in
quadrature phase, circularly polarised radiations can be generated. In microwave
circuits this is generally achieved by incorporating a 3 dB hybrid coupler between
the antenna and the source. A photograph of an experimental prototype [14] of an
antenna with this functionality is presented in Fig. 1.19. It is fabricated from two
copper clad substrates which have been combined to create a triplate sandwich. In
the cavity region the inner copper cladding is cleared leaving only the stripline feed
tabs. The outer cladding forms the top and bottom walls of the square cavity which
is completed by the metal plated vias clearly shown in the photograph. The area
between the cavity and the coaxial connectors supports a stripline feed network
which comprises a 3 dB hybrid coupler to generate circularly polarised radiation
from the slot pattern. The presence of the dielectric within the cavity lowers its
frequency and its Q-level, as does the via constructed sidewalls. However, mea-
surements demonstrate levels of Q which are more than high enough to maintain
modal purity [14].
26 1 Evolution of Compact Slot Antennas

Fig. 1.19 Dual-polarised


cavity resonator antenna for
RFID scanning applications at
2.45 GHz (*4 cm wide by
7 cm long)

1.6 Planar Reflect-Arrays for Compact Antennas

Pressure from communication systems developers for access to compact alterna-


tives to traditional antenna types, such as the mainstay parabolic reflector antenna,
has increasingly been driven by the inevitable drift toward higher and higher fre-
quencies by expanding demand for broadband systems and from growing
requirements for mobile platforms which can provide direct access to satellites. At
below microwave frequencies this role has traditionally been filled by versatile
reflector antennas which provide high efficiency, high gain, high bandwidth and can
readily deliver tracking using well established motion control techniques.
Nevertheless, the conventional parabolic reflector antenna presents a primary dis-
advantage in applications such as mobile platforms capable of tracking low earth
orbit (LEO) satellites, namely in its relatively bulky shape and in its weight.
Incorporating such an antenna, while preserving low wind resistance, into a fast
moving vehicle, which is itself streamlined, is difficult and costly. Furthermore,
tracking a satellite from a fast moving vehicle severely stretches the capability of
mechanical control systems. Solutions based on simply ‘flattening’ the reflector,
while preserving the focusing property of the original antenna, are being explored.
The electromagnetic processes which require to be followed to ‘flatten’ a
parabolic reflector antenna, without sacrificing ‘focusing’, are developed in detail in
Chap. 9. However, it is not too difficult to illustrate the process with a few simple
2-D sketches as shown in Fig. 1.20. The ‘flattening’ argument applies equally to
3-D developments. In Fig. 1.20a the parabolic reflector surface (shown as a brown
1.6 Planar Reflect-Arrays for Compact Antennas 27

Fig. 1.20 Evolution of ‘flat’ focused reflectors for incorporation into compact reflectarray
antennas with applications in communication systems

dashed line) is presumed to be replicable by a stepped, or pixelated, surface which


is evidently possible if the pixellation steps (p) are infinitesimally small. In practice
p is set at about 10% of the operating wavelength. Ideally at the reflector aperture
(blue dashed line) the normal reflected wave from the centre of each step should
match that of the reflection from a similar point on the original reflector, to procure
‘equivalence’ between the parabolic surface and the pixelated replacement.
To ‘flatten’ the reflector as suggested in Fig. 1.20b, its upper surface has to
replicate the reflection phase of the original reflector at the aperture plane (blue
dashed line) in Fig. 1.20a. This is achieved by arranging that at the pixelated step
locations a ‘shorted’ groove is provided which reflects the incident electromagnetic
wave so that at the ‘flat’ surface the reflected wave exhibits the phase distribution of
the original reflector at its aperture plane. In three dimensions the grooves can be
viewed as shorted waveguides of a depth which generates the requisite phase. An
alternative approach is depicted in Fig. 1.20c where thin walled ‘waveguides’ are
raised from a flat backing plate. The upper surface in this case is not exactly flat but
the reflector is lighter than the case (b) alternative. In Chap. 9 more sophisticated
solutions are addressed. These are based on the use of patch, or slot, arrays etched
into flat copper clad substrates.
28 1 Evolution of Compact Slot Antennas

1.7 Compact Cavity Resonator Antennas

As we have observed in earlier sections of this chapter the manipulation of elec-


tromagnetic wave radiation patterns, in the context of compact antennas, can be
procured and controlled by the judicious disruption of surface current flows in
suitable feed systems such as striplines and waveguides. Furthermore, if bandwidth
is not an issue, it is suggested that additional forms of pattern control can be devised
by cleverly amalgamating slot patterns and orientations with carefully selected
cavity modes. But if antenna compactness, ease of fabrication and cost effectiveness
are primary developmental goals, then the implementation of cavity antennas in
conventional copper clad substrate form, is generally preferred and becomes a high
priority. Unfortunately, as Fig. 1.19 shows, stripline or microstrip technology does
not easily lend itself to high-Q cavity formation without resorting to the use of
non-standard techniques such as multiple metallised vias to realise sidewalls. This
represents a distinct disadvantage for cavity based antenna initiatives where fab-
rication costs can be critical.
Needless to say solutions have been, and are being, sought by for example
employing resonance conditions in parallel plate waveguide, and in particular the
Fabry-Perot resonance to circumvent stripline fabrication issues. The Fabry-Perot
scheme is shown diagrammatically in Fig. 1.21a. It can be viewed as a rectangular
waveguide (see Fig. 1.9), with width a replaced by separation distance d, lying on
its side with the top and bottom walls removed. The Fabry-Perot resonance equates
to cut-off for the TE10 mode. Note that this mode is not influenced by the absence of
walls at y = −b/2 and y = +b/2 (see Fig. 1.9b). The surface current flows in the
walls at z = 0 and z = d will be y-directed and unvarying for all y, so the
Fabry-Perot resonator can act as an antenna for x-directed slots in the z = d wall,
assuming excitation of the cavity is through the z = 0 wall. For two dimensions the
nature of this radiation, for an array of slots forming a partially reflective surface
(PRS) at z = d, is not difficult to explain. The radiating elements of the PRS are
excited in phase and consequently a high directivity principal beam is formed at
broadside. It is analysed and discussed fully in Chap. 10.
In three dimensions, provided the resonator is excited by a non-directional feed
such as a z-aligned coaxial probe, orthogonal TE10 modes (with x-directed or y-
directed E-field) are possible, both of which behave as described above. The
resultant broadside radiated pencil beam will be polarised at 45° to x or y. However,
in the planar Fabry-Perot cavity antenna, each mode exhibits additional features
which are much less straightforward. When the structure is viewed on the x-z plane,
as shown in Fig. 1.21b, a more complicated field pattern for the same TE10 mode is
revealed. The surface wall current flows remain y-directed and continue to excite x-
directed slots. But now, for slots distributed in the x directions, in-phase excitation
is only guaranteed at cut-off where kg10 ) 1. In this case a high gain ‘pencil’
beam is generated by the 2-D Fabry-Perot cavity resonator antenna for this mode
1.7 Compact Cavity Resonator Antennas 29

Fig. 1.21 Schematic representation of a Fabry-Perot resonator antenna, a end view of TE mode
structure, b top view of TE mode structure

and the orthogonal mode. However, less helpfully, at frequencies above cut-off,
waves begin to propagate in the x directions (or y) within the parallel plate
structure leading to frequency sensitive and scanning leaky-wave beams. This
operating condition is examined in detail in Chap. 10.
30 1 Evolution of Compact Slot Antennas

1.8 Retro-Directive Array Antennas

The principle behind the retro-directive array, which is addressed in Chap. 11, has
been known about since the 1960s. Retro-direction, whereby an incoming radar
signal, on striking a target reflector is returned back in the direction from which it
came, usefully boosts the ‘visibility’ of the original scatterer. This outcome has
historically been achieved by employing ‘corner reflectors’. These are often seen on
the masts of yachts [15]. The device increases the radar cross-section of the target
thereby enhancing its detectability against severe electromagnetic scattering ema-
nating from the sea. A schematic drawing illustrating the principle of operation of
the corner reflector is presented in Fig. 1.22. Generally it comprises three flat
conducting metal surfaces positioned orthogonally to each other—for example the
inner faces of a cube as suggested in Fig. 1.22a. In practice the corner reflector is
made less bulky and hence lighter than the ‘open cube’ by refashioning it into a
trihedral shape as shown in Fig. 1.22b.
The retrodirective nature of the corner reflector can be illustrated relatively easily
by expressing the ray directions in Fig. 1.22a in vectorial terms. For example the
incident portion of ray 1 can be defined as travelling in direction

r 1 ¼ ax x 1 þ ay y 1 þ az z 1 ð1:11Þ

where ax, ay, az are the unit vectors in the x, y and z directions respectively. If this
ray represents a plane electromagnetic wave at frequency x propagating towards
the corner reflector, then (see Appendix A) the electric field component can be
expressed as:

E1 ¼ a? E1 exp jðxt  k0 r1 Þ ð1:12Þ

where a? defines a unit vector normal to r1, and k0 is the free-space propagation
coefficient. On the first reflection on the bottom x-y surface of the corner reflector,
the components of r1 in the x and y directions are unchanged while the z-component

Fig. 1.22 Traditional corner reflectors showing retrodirective operational mode


1.8 Retro-Directive Array Antennas 31

is reversed by Snell’s law, to −azz1. The resultant reflected ray direction can be
expressed vectorially as:

r01 ¼ ax x1 þ ay y1  az z1 ð1:13Þ

At the next reflection on the rear x-z plane (Fig. 1.22a), the x and z components
of r1 are unchanged while it is the y component that is now reversed in phase.
Hence

r001 ¼ ax x1  ay y1  az z1 ð1:14Þ

Finally, at the third reflection at the y-z surface, only the x component is reversed
in phase, ending up with a ray travelling in the following direction:

r000
1 ¼ ax x1  ay y1  az z1 ð1:15Þ

This is a precise reversal of the incident wave direction (it is said to be


retro-directed) resulting in an E-field representation of the retro-wave which can be
expressed mathematically as

E1 ¼ a? E1 exp jðxt þ k0 r1 Þ ð1:16Þ

For perfect conductors Eqs. (1.12) and (1.16) are essentially identical except for
the sign of the second term within the exponential argument. The spatial component
of the reflected wave is phase conjugated relative to the incident wave. Note that
this is generally true of all incident plane waves which illuminate the corner
reflector.
In mobile communication terms, and particularly in communications with
satellites, the ability to retain a communications link by retro-directing the received
signal back to its source has considerable potential, if it can be achieved with a less
limited and bulky antenna which the corner reflector represents. The key feature of
retro-directive action is the complex conjugate relationship between the received
and transmitted wave phases. It is shown in Chap. 11 that retro-directivity can be
procured using elegant antenna array formats with each element backed by a phase
conjugating circuit. Not surprisingly, retro-directive antennas, in array format and
incorporating varying levels of technical sophistication, have been evolving since
pioneering work in the 1960s [16, 17] and arguably the planar retro-directive array
offers the most cost effective option for satellite tracking. Recently, a planar
retro-directive array design has been reported for potential use in terminals
accessing the Inmarsat BGAN system [18].
32 1 Evolution of Compact Slot Antennas

References

1. M. Nosrati, N. Tavassolian, Miniaturised circularly polarised square slot antenna with


enhanced axial ratio bandwidth using an anti-podal Y-strip. IEEE Antennas Wirel. Propag.
Lett. 16, 817–820 (2016)
2. Y. Seo, H. Lee, Y. Lim, Design of a compact dual-meander antenna for WLAN operation.
Microwave Opt. Technol. Lett. 54(7), 1595–1599 (2012)
3. C.-Z. Li, C.-M. Tong, W. Shaoqing, L.-H. Qi, Compact slot Yagi-Uda like antenna design
with directional pattern, in IEEE International Conference on Communications Software &
Networks, Chengdu, China, 6–7 June 2015. https://doi.org/10.1109/iccsn.2015.7296143
4. K.R. Carver, J.W. Mink, Microstrip antenna technology. IEEE Trans. Antennas Propag. 29
(1), 2–24 (1981)
5. A.J. Sangster, P. Smith, A polarisation agile T-shaped slot antenna fed from balanced
microstripline, in 26th European Microwave Conference Proceedings, vol. 1, pp. 288–292,
Prague, 9–12 Sept 1996. ISBN 1 899919 08 2
6. S. Silver, Microwave Antenna Theory & Design, vol. 12 (Radiation Laboratory Series, 1949)
7. R.S. Elliott, Antenna Theory and Design (Prentice-Hall Ltd., Englewood Cliffs, 2006)
8. R.C. Johnson, Antenna Engineering Handbook, 3rd edn. (McGraw-Hill Inc., New York,
1993)
9. H.Y. Wang, S.Q. Zhang, A.J. Sangster, Numerical modelling of omni-directional slotted
waveguide antennas. Microwave Opt. Tech. Lett. 31(5), 374–377 (2001)
10. A.J. Sangster, H. Wang, Omni-directional blade antenna for microwave hyperthermia
treatment. J. Electromagn. Waves Appl. 19(14), 1935–1948 (2005)
11. K. Milne, The combination of pulse compression with frequency scanning for three
dimensional radars, in Proceeding of Symposium on Signal Processing in Radar and Sonar
Direction Systems, Birmingham, 6–9 July 1964
12. A.W. Rudge, K. Milne, A.D. Olver, P. Knight (eds.) The Handbook of Antenna Design:
Volumes I and II (Peter Peregrinus Ltd., London, 1986)
13. A.J. Sangster, H. Wang, P. Smith, R. Jacobs, Moment method solutions for polarisation agile
slots in bifurcated waveguide. Electromagnetics 19(1/2), 91–107 (1999)
14. A.J. Sangster, P. Smith, E. McErlean, K. Sinclair, R. Jacobs, A dual-polarised stripline fed
slot antenna incorporating signal cancellation. IEE Proc. (MAP) 148(6), 357–362 (2001)
15. L.G. Van Atta, Electromagnetic Reflector. US Patent 2,908,002, 6 Oct 1959
16. C.Y. Pon, Retrodirective array using heterodyne technique. IEEE Trans. Antennas Propag.
AP-12, 176–180 (1966)
17. R.N. Ghose, Electronically adaptive antenna systems. IEEE Trans. Antennas Propag. AP-12
(2), 161–169 (1964)
18. N. Buchanan, V.F. Fusco, M. Van Der Vorst, N. Williams, C. Winter, New Retro-directive
antenna for mobile terminal applications, in 32nd Antenna Workshop, ESA/ESTEC,
Noordwijk, Netherlands, 5–8 Oct 2010

Bibliography

1. R. Ferrari, An Introduction to Electromagnetic Fields (Van Nostrand Reinhold Co., Ltd., New
York, 1975)
2. C.T.A. Johnk, Engineering Electromagnetic Fields and Waves (Wiley, New York, 1988)
3. R.P. Feynman, R.B. Leighton, M. Sands, Lectures in Physics, vol. II (Addison-Wesley
Publishing Co., London, 1972)
4. P.M. Morse, H. Feshbach, Methods of Theoretical Physics (McGraw-Hill Book Co., New
York, 1953)
5. J.D. Kraus, Electromagnetics (McGraw-Hill Book Co., London, 1984)
Bibliography 33

6. P. Lorrain, D. Corson, Electromagnetic Field and Waves (W.H. Freeman & Co., San
Francisco, 1962)
7. A.J. Baden-Fuller, Engineering Electromagnetism (Wiley, New York, 1993)
8. P. Hammond, Applied Electromagnetism (Pergamon Press Ltd., Oxford, 1971)
9. J.A. Stratton, Electromagnetic Theory (Wiley, New Jersey, 2007)
10. D.M. Pozar, Microwave Engineering, 4th edn. (Wiley, USA, 2011). ISBN: 0470631554
Chapter 2
Fundamentals of Electromagnetic
Radiation

2.1 Theoretical Formulation of Radiation from a Slot

Today, our deep understanding of the nature of antennas, and in particular slot
antennas, originated with the study of what is now a classical problem in electro-
magnetics—namely that of the aperture in a plane conducting screen of infinite
extent illuminated by a plane electromagnetic wave onto one face (see Fig. 2.1).
The earliest significant examination of this, at the time intriguing, problem of how
to calculate the electromagnetic fields, at arbitrary location P(x, y, z), in the empty
half-space to the right of the screen, is attributed to Lord Rayleigh in 1897 [1].
The essence of a solution lies in creating a credible representation of the induced
electric field or magnetic field within the aperture which becomes the source of the
scattered fields. This is not a trivial problem and called for some considerable
ingenuity on the part of early investigators at a time when the available mathe-
matical tools were strictly analogue in character and while, contemporaneously,
only closed form solutions were practicable or workable.
Antenna radiation (or reception) is classified as a field theory problem of the
boundary value type not too unlike the electromagnetic propagation problems
outlined in Appendix A. If the field region is remote from the source, field theory,
as we have seen, identifies transverse electromagnetic waves (TEM waves) in
free-space, or trapped electromagnetic waves in confined or bounded spaces. In this
case the boundary value problem is generally termed ‘source-free’. For such
problems Maxwell’s equations, in the source free differential form presented below,
are detailed in Appendix A.
r  E ¼ jxlH ð2:1Þ

r  H ¼ jxeE ð2:2Þ

rD¼0 ð2:3Þ

© Springer Nature Switzerland AG 2019 35


A. J. Sangster, Compact Slot Array Antennas for Wireless Communications, Signals
and Communication Technology, https://doi.org/10.1007/978-3-030-01753-8_2
36 2 Fundamentals of Electromagnetic Radiation

Fig. 2.1 Original problem of electromagnetic scattering by an aperture in a conductive plane


screen of infinite extent

rB¼0 ð2:4Þ

In radiation problems on the other hand, the source in no longer remote and becomes
an integral part of the field problem; radiation from a slot in a conducting surface is just
such a problem. It should, perhaps, be noted that the normal convention, in formulating
the mathematical equations associated with radiation, is to presume that the waves
emanate from the ‘source’. However, the transmit/receive process is entirely reciprocal
as indicated in Fig. 2.2, so we could equally evolve the equations on the presumption
that the focus of the waves is a ‘sink’ and the mathematics would not change.
In these source/sink present electromagnetic boundary value problems the most
general form of Maxwell’s equations are required. These are, in phasor form:

r  E ¼ jxlH ð2:5Þ

r  H ¼ jxeE þ J ð2:6Þ

rD¼q ð2:7Þ

rB¼0 ð2:8Þ

where J and q are the impressed ‘electric’ current density and ‘electric’ charge
density respectively, which form the source terms for the differential equations.
In radiation problems where the electromagnetic fields emanate from an aperture
(e.g. horn and slot radiators) the basic source of the radiation is by analogy a
2.1 Theoretical Formulation of Radiation from a Slot 37

Fig. 2.2 a Radiation source


and b radiation sink

‘magnetic’ current (electric field) rather than conventional ‘electric’ current (mag-
netic field), in which case it is more appropriate to use the following ‘dual’ forms
for Maxwell’s equations:

r  E ¼ jxlH  Jm ð2:9Þ

r  H ¼ jxeE ð2:10Þ

rD¼ 0 ð2:11Þ

r  B ¼ qm ð2:12Þ

where Jm and qm are the impressed ‘magnetic’ current density and ‘magnetic’
charge density respectively representing ‘aperture-source’ terms for the differential
equations. These forms are pursued in Chap. 3.
It is important, particularly in the ‘electric’ case to distinguish between the
impressed currents and charges, and currents and charges which arise because of the
presence of the fields in a medium having finite conductivity. Conduction current
density J is proportional to the electric field and is given by rE, where r is the
conductivity of the material. This current component can readily be accounted for,
in the Maxwell equations, by replacing the permittivity e by the complex form
e(1 − jtand) = e(1 − jr/xe). The density of free charge, apart from that associated
with impressed currents, may be presumed to be zero. Thus, in subsequent
discussion, the charge density q and the current density J appearing in the field
equations can be taken as comprising purely impressed charges and currents.
38 2 Fundamentals of Electromagnetic Radiation

Any other currents, if they arise, will be accounted for by the complex electric
permittivity which will be written simply as e for convenience.

2.2 Reciprocity for Macroscopic Antennas

The reciprocity hinted at in Fig. 2.2 can be established as follows. The two antennas
one in transmit mode (a) and one in receive mode (b) exhibit, respectively, input
impedances ZT and ZR. It is assumed that the antennas exist is the same linear,
isotropic space, which is not necessarily homogeneous, and that the electromagnetic
power radiated from antenna (a) propagates toward (b). The equivalent circuit
formations representing this antenna juxtaposition are presented in Fig. 2.3.
In the transmitter circuit (Fig. 2.3a) the generator internal impedance is represented
by Zg and to procure maximum power transfer to space it is set equal to the complex
conjugate of the antenna impedance ZT Zg ¼ ZT . In the receiver circuit, on the other
hand, optimum power transfer to the load (ZL) requires that ZL is set equal to the
complex conjugate of the receive antenna ZR; that is ZL ¼ ZR . These impedance
settings are convenient for the following reciprocity calculation, but are not essential.
The calculation proceeds by computing, firstly, the power (P1) delivered to
antenna (a) from the generator. This power is given by:
    2
1 1 Vg ZT Vg Vg 
P1 ¼ ReðVT IT Þ ¼ Re ¼ ð2:13Þ
2 2 ZT þ Zg ðZT þ Zg Þ 8RT

This power radiates into the surrounding space, and some of it is collected by the
receive antenna and becomes absorbed by the load. This power P2 can by computed
if we ascribe a transfer admittance Y21 to the transmit/receive network which
incorporates the generator impedance, the impedances of the antennas and the load
impedance. Hence

1
1  2
P2 ¼ Re ZL ðVg Y21 ÞðVg Y21 Þ ¼ RL Vg  jY21 j2 ð2:14Þ
2 2

Fig. 2.3 Equivalent circuit representation of transmit/receive antenna combination


2.2 Reciprocity for Macroscopic Antennas 39

The division of Eq. (2.14) by Eq. (2.13) yields an expression for the
receive-to-transmit power ratio for the antenna combination—namely:

P2
¼ 4RL RT jY21 j2 ð2:15Þ
P1

If the transmit antenna and the receive antenna are interchanged with (b) trans-
mitting power P1 and (a) receiving power P2 in its load, the transfer admittance is
now Y12 and the power ratio becomes:

P2
¼ 4RL RT jY12 j2 ð2:16Þ
P1

For linear circuits it is clear that Y21 ¼ Y12 and consequently the antenna
transmit/receive combination displays the property of reciprocity.

2.2.1 Reciprocity Theorems for Electromagnetic Fields

The property of duality in electromagnetic theory as it evidently exists between


electric current and magnetic current sources, is augmented by the concept of
reciprocity which essentially dictates that sources can be sinks and vice versa as we
have seen for macroscopic antennas occupying the same volume of space. Actually,
familiarity with electrical circuit theory in general makes the idea of reciprocity not
too surprising since it is quite natural to accept that in a linear circuit driven by a
current source in one part of the circuit with a voltmeter monitoring the voltage in
another, interchanging the current source and the voltmeter returns the same meter
reading. The same is true of a voltage source and an ammeter in a linear circuit.
Nevertheless, reciprocity is particularly appropriate to the development of antenna
theory as we have illustrated above where it evidently enables transmitting antenna
behaviour to be reinterpreted for the illumination of, or the better understanding of,
receiving antenna characteristics. The reverse is also true.
To establish the reciprocity theorem at the microscopic field level we commence
by assuming that electric current densities J1, J2, and magnetic current densities
Jm1, Jm2 form two sets of sources in a linear isotropic medium which may, or may
not, be homogeneous. We further assume that the sources radiate singly or together
into the same medium at the same frequency and generate fields E1, H1 and E2, H2
respectively. Such sources and fields, which are governed by the Maxwell equa-
tions, are as a consequence forced to satisfy the Lorentz reciprocity theorem [1, 2],
which in differential form yields:

r  ðE1  H2  E2  H1 Þ ¼ E1  J2 þ H2  Jm1  E2  J1  H1  Jm2 ð2:17Þ


40 2 Fundamentals of Electromagnetic Radiation

Applying a volume integration over V to both sides of this equation, and sub-
sequently enforcing the divergence theorem to the right hand side, leads to the
arguably more useful integral form of the Lorentz reciprocity theorem—namely:
ZZ ZZZ
 ðE1  H2  E2  H1 Þ  dS ¼ ðE1  J2 þ H2  Jm1  E2  J1  H1  Jm2 ÞdV
S V
ð2:18Þ

Equation (2.18) leads to some useful relationships in certain limited, but rele-
vant, conditions. In a source free region (J1 = J2 = Jm1 = Jm2 = 0) Eq. (2.18)
evidently reduces to:
ZZ
ðE1  H2  E2  H1 Þ  dS ¼ 0 ð2:19Þ
S

This equation is a special case of the Lorentz reciprocity theorem and must be
satisfied on the enclosing surface S of a source free region containing the fields E1,
H1 and E2, H2 simultaneously.
From an antenna perspective it is often the case that the sources and sinks
(transmit and receive antennas) can be considered to reside in a sphere (free space)
of infinite radius. At the surface of this sphere (S) at infinity, the fields generated by
the finite sources within it (J1, J2, Jm1, Jm2) will have reduced to zero, which means
that the left hand side of Eq. (2.18) is zero. Hence:
ZZZ
ðE1  J2 þ H2  Jm1  E2  J1  H1  Jm2 ÞdV ¼ 0 ð2:20Þ
V

which can obviously be expressed in the form:


ZZZ ZZZ
ðE1  J2 þ H2  Jm1 ÞdV ¼ ðE2  J1  H1  Jm2 ÞdV ð2:21Þ
V V

It is pertinent to note here that Eqs. (2.19) and (2.21) are not power or energy
expressions because complex conjugation is absent. What these equations do rep-
resent is field/source coupling or reaction resulting in further field components [1].
In antenna terms, and in particular array antennas, the process gives rise to mutual
coupling (see Chap. 6) which can present a major source of error, if handled
incorrectly, when precision pattern control is sought in such arrays. In addition, the
above reciprocity relations are particularly relevant to the electromagnetic mod-
elling of elemental antennas, such as dipoles and slots embedded within compact
antenna geometries, as we shall see in Chap. 5.
2.2 Reciprocity for Macroscopic Antennas 41

2.2.2 Inhomogeneous Vector Helmholtz Equation

The impressed currents and charges, expressed explicitly in the source-present form
of Maxwell’s equations (Eqs. 2.5–2.8, and 2.9–2.12), must be related to each other
through the equation of continuity for current and charge. That is:

@q
rJ¼  ð2:22Þ
@t

The dual relationship for ‘magnetic current’ and ‘magnetic charge’ is

@qm
r  Jm ¼  ð2:23Þ
@t

Furthermore, to solve the Maxwell equations for E-field or H-field behaviour in a


bounded region it is first necessary to form an equation for either E or H alone
which is facilitated by employing Eq. (2.22) or (2.23) as appropriate. The standard
procedure for achieving this conversion is to perform a curl operation on either the
curl equation for E or the corresponding equation for H (see Appendix A). This
process leads to second order differential equations in E and H. In the electric
current case we obtain:

r  r  E  k 2 E ¼ jxlJ
ð2:24Þ
r  r  H  k2 H ¼ r  J

where q is electric charge density. Equation (2.24) are inhomogeneous vector


Helmholtz equations in which the term on the right hand side of the equals sign
forms a source function. By employing the equation of continuity for charge and
current, the source function can be expressed in terms of the more familiar dif-
ferential operators on the left of the equation, but at the expense of more complex
source terms on the right. That is

rr  J
r2 E þ k 2 E ¼ jxlJ  ð2:25Þ
jxe

By an entirely analogous procedure, a corresponding inhomogeneous vector


wave equation for H can be derived, which has the form:

r2 H þ k2 H ¼ r  J ð2:26Þ

In deriving the above equations it has been presumed that l and e are invariant
with time. In free space or vacuum, l ¼ l0 ¼ 4p  107 H=m and
e ¼ e0 ¼ 36p
1
 109 F=m.
42 2 Fundamentals of Electromagnetic Radiation

In a source free region, for which the impressed current J is non-existent,


Eqs. (2.25) and (2.26) reduce, as they should, to the well known homogeneous
vector Helmholtz equations, namely:

r2 E þ k 2 E ¼ 0
ð2:27Þ
and r2 H þ k 2 H ¼ 0

These are the phasor forms of the wave equations.


When a radiating source of the magnetic current type (Jm—e.g. slot radiator)
exists within the region of interest, a similar derivation when applied to Eqs. (2.25)
and (2.26) leads to the following inhomogeneous vector differential equation

r  r  H  k 2 H ¼ þ jxeJm
ð2:28Þ
r  r  E  k2 E ¼ r  Jm

or equivalently:

r 2 E þ k 2 E ¼ r  Jm
rr  Jm ð2:29Þ
and r2 H þ k 2 H ¼ jxeJm þ
jxl

These magnetic current sources, and the antenna forms which they generate, are
treated in depth in Chap. 3.
It is clear on examining their forms that the impressed current density vectors
J and Jm enter into the inhomogeneous differential Eqs. (2.25), (2.26) and (2.29) in
a particularly complicated manner, making their direct solution extremely difficult,
although with modern computer based electromagnetic solvers employing finite
difference or finite element techniques—not impossible. Nevertheless, this difficulty
is normally circumvented by resorting to the use of auxiliary potential functions,
which generate more easily solvable wave equation forms. These auxiliary potential
functions may, or may not, represent clearly definable physical entities (especially
is this so in the absence of sources), and so it is usual to adopt the view point that
these potentials are merely useful functions from which the electromagnetic fields
can be derived. The two most common potential functions which arise in radiation
problems are discussed in the next section.

2.3 Auxiliary Potential Functions

In formulating auxiliary potential functions, with the aim of facilitating the solution
of the source present second order differential equations as represented by
Eqs. (2.25), (2.26), (2.28) and (2.29), it is usually advisable to choose forms that
2.3 Auxiliary Potential Functions 43

satisfy at least one or more of the basic Maxwell equations. For the ‘electric’ current
Eqs. (2.25), (2.26) such a choice is:

B¼rA ð2:30Þ

which immediately satisfies Eq. (2.8), since by definition r  r  S ¼ 0 for any


vector S. In addition the choice:

@A
E¼ þ r/ ð2:31Þ
@t

satisfies Maxwell Eq. (2.5) since r  ru ¼ 0 for any scalar function. The auxil-
iary vector A is usually termed the magnetic vector potential, which has relevance
in magnetostatics, while auxiliary scalar / is termed the electric potential (some-
times denoted by V) and has its origins in electrostatics. Consequently, if we can
develop solvable source present second order differential equations for A and /,
then the associated electric and magnetic field forms can be deduced subsequently
from the above two equations.
To develop differential forms in A and / we again enrol the Maxwell equations
as expressed in Eqs. (2.5)–(2.8). On substituting D (= e0E for vacuum) in Eq. (2.7)
with the expression for E given in Eq. (2.31) we obtain:

@ q
r2 /  ðr  AÞ ¼ ð2:32Þ
@t e0

which hardly seems an advance, in simplification terms, on Eqs. (2.25) and (2.26).
However, Eqs. (2.30) and (2.31) only partially define A and /. Equation (2.30)
delineates only a rotational component of A (Ar say) while Eq. (2.31) defines a part
of / which has finite gradient in the space domain. Therefore, we can add to A any
component which has a zero curl (termed a lamellar component) without altering
Eq. (2.30), and we can add to / any function whose gradient is zero, leaving
Eq. (2.31) unchanged. Provided the Maxwell relationships are not disrupted, these
arbitrary introductions to A and / can be chosen in a manner which simplifies
Eq. (2.32). Such a choice is the Lorentz condition whereby:

1 @/
rA¼ ð2:33Þ
c2 @t

Actually this choice is not totally arbitrary. The Lorentz condition can be
identified with the Kirchhoff law which relates conduction current to rate of change
of charge (Eq. 2.22). It also brings the resultant radiation equations into line with
the requirements of special relativity [3].
On substituting Eq. (2.33) into Eq. (2.32) a second order differential in / alone
is formed, namely:
44 2 Fundamentals of Electromagnetic Radiation

1 @2/ q
r2 /  ¼ ð2:34Þ
c @t
2 2 e0

or in phasor form:
q
r2 / þ k02 / ¼  ð2:35Þ
e0

where
x pffiffiffiffiffiffiffiffiffi
k0 ¼ ¼ x l0 e0 ð2:36Þ
c

Recalling that H ¼ B=l0 and e = e0 in free space then substituting Eqs. (2.30)
and (2.31), together with the Lorentz condition, into Eq. (2.6), yields an equivalent
second order differential equation for A. In phasor form, this equation is termed the
vector Helmholtz equation:

r2 A þ k02 A ¼ l0 J ð2:37Þ

Equations (2.35) and (2.37) are clearly wave equations but with source functions
on the right of the equals sign which are much less convoluted than those in
Eqs. (2.25) and (2.26). These source functions have now been simplified to the
impressed charge density (divided by a constant eo) in the region of interest in
Eq. (2.35), and impressed current density (multiplied by a constant lo) in
Eq. (2.37). If q or J are known in space and time, on an antenna located in a volume
of space (say), then in principle the above equations can be solved directly for / or
A respectively. In practice, only one of the two wave equations need be solved
since / and A are related through the Lorentz condition. However, rather than
demonstrate the direct route to a solution, it is actually more instructive to construct
a solution by employing electrostatic and magnetostatic analogies.

2.4 Electrostatics Analogy

If Eq. (2.7) is applied to a small spherical volume of stationary and isolated free
charge (q), located in a magnetic field free region, then symmetry and the circu-
lation law for E (Eq. 2.5) dictate that the electric vector D will be everywhere radial
and unvarying over an enveloping spherical surface (S) of radius r (Fig. 2.4),
enclosing a volume V. Under these circumstances volume integration on both sides
of Eq. (2.7) yields:
2.4 Electrostatics Analogy 45

Fig. 2.4 Electric field flux


through a spherical surface
containing a point charge q

ZZZ ZZ
r  DdV ¼  D  dA ¼ 4pr 2 Dr ¼ q ð2:38Þ
V S

Clearly Maxwell’s first equation reduces to the inverse square law, or Gauss’ law
which is usually written:
q
Er ¼ ð2:39Þ
4pr 2 e0

provided that, for air, or vacuum:

D ¼ e0 E ð2:40Þ

In the m.k.s. system, D has the dimension C/m2, while the constant of propor-
tionality e0 is termed the absolute permittivity of free space. Measurement reveals
that it has the value 8.84  10−12 F/m, where in the m.k.s. system a Farad in a
Coulomb/volt. More generally, D = ere0E, where er is termed the relative permit-
tivity of the material concerned. It need not be a constant, and could be a complex
tensor to represent a general non-linear, anisotropic, inhomogeneous medium.
In conjunction with the Lorentz force equation:

F ¼ qE Newtons

Equation (2.39) yields the Coulomb force law which can be expressed as:

qQ
F¼ Newtons ð2:41Þ
4pr 2 e0

for isolated point charges q and Q (in coulombs) distance r metres apart in a
vacuum.
46 2 Fundamentals of Electromagnetic Radiation

Equation (2.39) really only applies if q is concentrated at a point, and so the


situation in Fig. 2.5 where a distributed charge of density q is assumed to surround
the origin of the Cartesian coordinate system (shaded region). To determine E in
this case we must first dice the shaded volume up into elemental volumes such a
dV 0 ð¼ dx0 dy0 dz0 Þ at r 0 . For such a volume the charge enclosed is q ¼ qðx; y; zÞdV 0
coulombs and if it can be considered to reside at a point in space, then given that
potential / ¼ Edr where dr ¼ jr  r 0 j, the potential at P due to this element is:

qðx; y; zÞdV 0
d/ ¼ ð2:42Þ
4pe0 jr  r 0 j

The potential at P due to the entire charge volume (V) at O, is the summation of
all contributions qðx; y; zÞdV 0 at all possible positions O′ throughout the charge
volume. In the limit this summation becomes an integral, giving:
ZZZ
1 qdV 0
/¼ ð2:43Þ
4pe0 jr  r 0 j
V

While this equation exhibits the correct spatial form for electric potential at
distance from an arbitrary charge distribution, it is applicable only to time invariant
charge concentrations. The discussion in relation to Fig. 3.1 suggests that if the
charge alters with time then Eq. (2.43) could justifiably be applied to fields in close
proximity to the charge, but at more distant locations as a time retardation mecha-
nism is required. So can we build time delay into Eq. (2.43)? It turns out that we can.
If the charge distribution at O′ in Fig. 2.5 alters with time, the information that
this change has occurred can be felt at point P only after a period of time jr  r 0 j=c,
which is the time it takes for a TEM wave to travel distance jr  r 0 j. Hence, it is not

Fig. 2.5 Geometry for


construction of retarded
potentials
2.4 Electrostatics Analogy 47

implausible to suggest that at time t and at position r the contribution to /(r, t) at P


can be calculated using Eq. (2.43) if we insert the charge distribution at time
t  jr  r 0 j=c. This is true for all volume elements dV 0 within V. Consequently, the
desired time varying solution to Eq. (2.35) can be formulated by simply incorpo-
rating the time retardation into Eq. (2.43), to give:
ZZZ
1 qðr0 ; t  jr  r0 j=cÞdV 0
/ðrÞ ¼ ð2:44Þ
4pe0 j r  r0 j
V

That Eq. (2.44) is a solution to Eq. (2.35) can be established by substitution. The
solution to the phasor form of Eq. (2.35) simplifies to:
ZZZ
1 expðjko jr  r0 jqðr0 ÞdV 0
/ðrÞ ¼ ð2:45Þ
4pe0 j r  r0 j
V

with time expressed through the common factor exp(jxt) which can be suppressed.

2.5 Magnetostatics Analogy

The fundamental measurement based laws of magnetostatics, as deduced by the


early pioneers in electrical science can be constructed in similar manner from
Eqs. (2.10) and (2.12). Equation (2.12), which is usually referred to as the flux law
for magnetic field, simply states that however generated—by a permanent magnet,
by the Earth’s core, by current on a wire, by a current carrying coil—magnet flux
density forms closed loops. At the macroscopic level there are no magnetic field
sources performing the same role as charge in electrostatics. Hence, if we apply
Eq. (2.10) to a steady (DC) current (I) on a long straight wire, we can confidently
say that the magnetic field which the law predicts must form closed circular loops
centred on the wire—assuming the wire has a cylindrical cross-section. We
therefore obtain:
I
H  d‘ ¼ 2prHt ¼ I
C
I ð2:46Þ
Ht ¼
2pr
l0 I
Bt ¼
2pr

Equation (2.46) is Ampere’s law, and since I is expressed in amps (A) in m.k.s.
system, the dimension of magnetic field strength H is A/m. With H in A/m and B in
Tesla consistency of the magnetic field laws requires that for vacuum:
48 2 Fundamentals of Electromagnetic Radiation

B ¼ l0 H ð2:47Þ

where the constant l0 is termed the absolute permeability of free space and has the
magnitude 4p  10−7 H/m (Henry = T  m2/A). As with electrostatics we can say
that more generally, B = lrl0H, where lr is termed the relative permeability of the
material concerned. It need not be a constant, being a function of H in a ferrite
material, and it could be a complex tensor to represent a general, anisotropic,
inhomogeneous medium. Application of the circuital law for H to a current element
Id‘ rather than to a current on a long straight wire, leads to the Biot and Savart law
(see Balanis in Bibliography). Referring to Fig. 2.6 where a current element Idl0 is
located at O0 a distance r 0 from the origin at O. In determining the magnetic flux
density at P distance r from O, we can assume from Ampere’s law that the flux
forms circular loops centred on the current I. The contribution dB at P must be
proportional to I in magnitude and inversely proportional to the square of the
distance between O0 and P, namely jr  r0 j. Directionally the vector dB must be
orthogonal to both Idl0 and the vector ðr  r0 Þ. So mathematical reasoning suggests
that we can express dB in the following form:

l0 I dl0  ðr  r0 Þ
dB ¼ ð2:48Þ
4p jr  r0 j3

Fig. 2.6 Magnetic field at P


due to current element Idl0
2.5 Magnetostatics Analogy 49

This the Biot-Savart law which is most commonly expressed in the form:

l0 Id‘
dB ¼ ð2:49Þ
4pr 2

for a current element at the centre of the magnetic field loop. If the straight portion
of wire carrying I is actually part of a complete electrical circuit s, then on summing
all the Idl0 contributions around s leads, in the limit of infinitesimally small dl0 , to an
integration giving:
I
l0 I dl0  ðr  r0 Þ
BðrÞ ¼ ð2:50Þ
4p j r  r0 j 3
s

Remembering that we have chosen to define the vector magnetic potential


through Eq. (2.30) and noting that [3]:

dl0 ð r  r0 Þ
r ¼   dl0
j r  r0 j jr  r0 j3

thence, combining Eqs. (2.30) and (2.50) yields the following:


I
l I dl0
AðrÞ ¼ 0 ð2:51Þ
4p j r  r0 j
s

The time retarded form for A can be deduced in much the same way as we have
already done for / with the additional observation that current and current density
are related through:
Idl0 ¼ JdV 0 ð2:52Þ

Hence we obtain:
ZZZ
l0 Jðr0 ; t  jr  r0 j=cÞ 0
Aðr; tÞ ¼ dV ð2:53Þ
4p j r  r0 j
V

or in phasor form:
ZZZ
l0 expðjko jr  r0 jÞJðr0 Þ 0
AðrÞ ¼ dV ð2:54Þ
4p j r  r0 j
V

The retarded potentials A and / represent powerful tools in the search for
solutions to complex electromagnetic radiation and reception problems, and are
widely used in various simulation packages by providing the ‘building blocks’,
such as the ‘short current element’, to construct field patterns for distributed current
50 2 Fundamentals of Electromagnetic Radiation

sources, which are known in space and time. This is particularly true of those
packages employing moment methods. The short current element, or Hertzian
dipole, is examined in the next section.
Needless to say, equivalent forms to A and / also exist, which provide expedient
routes to the solutions to Eqs. (2.28) and (2.29). These are usually defined by the
parameters Am—an auxiliary potential function associated with ‘magnetic current’
Jm, and /m—an auxiliary potential linked to ‘magnetic charge’ qm. These forms
will be considered in more detail in Chap. 3.

2.6 Hertzian Dipole

The Hertzian dipole and the ‘short current element’ are essentially synonymous.
The concept is presented schematically in Fig. 2.7. The short current element is
viewed as a filament of current Idl which exhibits constant magnitude along its
length dl. As such it is physically impossible to realise, but remains a helpful
mathematical construct. It functions as an exceptionally useful building block in
antenna modelling which enables the field patterns of practical antenna structures to
be determined by integration. Consequently it is instructive to examine the field
behaviour of this current filament, and this is implemented by applying the
appropriate potential function to the example shown in Fig. 2.7.
The current element is assumed to be located at the origin of a Cartesian
coordinate system and that it takes the form a constant z-directed current varying in
time at frequency x rad/sec—i.e. the current I ¼ I0 cos xt. If time retardation is
incorporated into the definition of current then this becomes:

Fig. 2.7 Short current


element at origin of Cartesian
coordinate system
2.6 Hertzian Dipole 51

r
I ¼ I0 cos xðt  Þ ¼ I0 cosðxt  k0 r Þ ð2:55Þ
c

In complex notation this can be written:

I ¼ Re½I0 exp ðjðxt  k0 rÞÞ


pffiffiffiffiffiffiffiffiffi
In these two equations the phase coefficient k0 ¼ x=c ¼ x l0 e0 for free space.
So, for this one dimensional element of length dl in the z direction Eq. (2.54)
becomes:

l0 I0 dl
A¼ expðjðxt  k0 rÞÞ ð2:56Þ
4pr

 If we adopt the unit vector notation, for the Cartesian system, namely
^ax ; ^ay ; ^az then:

l0 I0 dl^az
A ¼ Az ^az ¼ expðjðxt  k0 rÞÞ ð2:57Þ
4pr

It is evident from Eq. (2.57) that at distant point P (Fig. 2.7) the magnetic vector
potential is z-directed. However, we know from our earlier discussion on the
radiation mechanism that the important field components are orthogonal to the
direction vector r and are therefore best represented by spherical components.
Consequently we let:

A ¼ ^ar Ar þ ^ah Ah þ ^a/ A/ ð2:58Þ

where Ar ¼ Az cos h, Ah ¼ Az sin h and A/ ¼ 0. Hence the magnetic field com-
ponents of the radiated wave at P can be derived using B ¼ l0 H ¼ r  A resulting
in [6]:
 
I0 dl jk0 1
H/ ¼ þ 2 sin h expðjk0 rÞ ð2:59Þ
4p r r

Hh ¼ Hr ¼ 0 ð2:60Þ

The electric field components of the same radiated wave can be constructed by
applying the Maxwell Eq. (2.2) to generate:
rffiffiffiffiffi  
l0 I0 dl jk0 1 j
Eh ¼ þ 2  3 sin h expðjk0 rÞ ð2:61Þ
e0 4p r r k0 r
rffiffiffiffiffi  
l0 I0 dl 1 j
Er ¼  cos h exp ðjk0 rÞ ð2:62Þ
e0 2p r 2 k0 r 3
52 2 Fundamentals of Electromagnetic Radiation

E/ ¼ 0 ð2:63Þ

From the discussion in Sect. 2.1 the fields radiating away from the short current
element are expected to be normal to the direction vector r, and to each other. They
are also predicted to diminish in magnitude at a rate which is inversely proportion to
distance r from the source. The only field components which meet this requirement
are the first term of H/ in Eq. (2.59) and the first term of Eh (Eq. 2.61). So in the far
field, generally defined as occupying the region given by k0r > 1:

jk0 I0 dl
H/ ¼ sin h expðjk0 rÞ ð2:64Þ
4pr
rffiffiffiffiffiffiffiffiffiffiffiffi
l0
Eh ¼ H / ¼ g0 H / ð2:65Þ
e0

A pictorial representation of the Eh field distribution is shown in Fig. 2.8. Note


that the pattern is circularly symmetric in the ‘horizontal’ plane (x-y plane in
Fig. 2.7) which slices through the current element, but it is not ‘isotropic’—radi-
ating equally in all directions. The ‘solid’ pattern is ‘dough-ring’ shaped with
electric (and magnetic) field nulls along the z-axis (h = 0°, 180°) of the source
current. The H/ field is everywhere normal to the E-field, and normal to the plane
of the page containing Fig. 2.8.
So, what is the significance of the other field components, which diminish with
distance as 1/r2 and 1/r3 in Eqs. (2.61) and (2.62)? The 1/r3 term is reminiscent of
the electrostatic field behaviour for an electric dipole [6]. It arises because the
current I must, in accordance with Kirchoff’s law, be associated with positive and
 
negative charge build up I ¼  dq dt at its extremities, so forming, at any given

Fig. 2.8 Electric field


contours for radiation from a
short current element at a
specific instant in time (red =
peaks; blue = troughs in the
wave)
2.6 Hertzian Dipole 53

instant in time, an electric dipole. The field exists only in the ‘near’ field region
(kor < 1) of the source, being negligible in the far field. It supports electrostatic or
capacitive energy storage.
The 1/r2 which exists in both the electric and magnetic field formulations
(Eqs. 2.59, 2.61 and 2.62) exhibits the characteristics of the stored inductive
magnetic field in the vicinity of a current carrying wire. In magnetostatics it is
derived using the Biot-Savart law. The electric field components arise because the
antenna current, and hence this inductive magnetic field, is time varying. The fields
are derivable using the Faraday law (Eq. 2.5). Again these fields diminish rapidly
with distance from the source and are essentially zero in the far field.
The nature of these field components can be further illuminated by examining
complex power flow in the near field of the source. Complex power flow density in
W/m2 is generally expressed in the form:

1
p ¼ E  H W/m2 ð2:66Þ
2

where p is a vector directed normal to E and H. Following convention, H is


conjugated to suppress the exponential term, in much the same way as in electrical
circuit theory, complex power is expressed as P ¼ 12VI  . Hence expressing E and
H in their r, h and / components we obtain:

1
p ¼ ð^a/ Er Hh  ^ah Er H/  ^a/ Eh Hr þ ^ar Eh H/ þ ^
ah E/ Hr  ^
ar E/ Hh Þ ð2:67Þ
2

On introducing Eqs. (2.59) to (2.63) into Eq. (2.67) the following components to
the complex power density p for the short current element are derived:
rffiffiffiffiffi 
jk0 I02 dl2 l0 1 1
p ¼ ^ah þ sin h cos h
16p2 e0 r 3 k02 r 5
rffiffiffiffiffi  ð2:68Þ
k0 I 2 dl2 l0 k0 j
þ ^ar 0 2  sin2 h
32p e0 r 2 k02 r 5

The real power flow density radiating in the r-direction away from the source is
given by pr ¼ Re p ¼ 12 ReðE  H Þ and has the form:
rffiffiffiffiffi 2
k02 I02 dl2 l0 sin h
pr ¼ ð2:69Þ
32p2 e0 r 2

The remaining components of Eq. (2.68) are imaginary in character and repre-
sent stored energy in the near field on the source. These components can be
extracted by expressing pi ¼ Im p. This leads to an r-directed component:
54 2 Fundamentals of Electromagnetic Radiation

rffiffiffiffiffi 2
jI02 dl2 l0 sin h
pi j r ¼ ð2:70Þ
32k0 p2 e0 r 5

and a h-directed component:


rffiffiffiffiffi 
jk0 I02 dl2 l0 1 1
pi j h ¼ þ sin h cos h ð2:71Þ
16p2 e0 r 3 k02 r 5

The real power flow density (Eq. 2.69) diminishes, as expected, in proportion to
the inverse radius squared (inverse square law), while the sine squared dependence
of the equation denotes that the radiating lobe maximises in the x-y plane (at
h = 90°) for all values of /. The stored energy density term pi jr (Eq. 2.70), which
decays in inverse proportion to radius to the power of five, also exists mainly in the
x-y plane. This electric dipole energy essentially oscillates at the frequency of the
source, between electric and magnetic stored forms, with the instantaneous power
flow density moving in a radially direction. On the other hand, the stored energy
density represented by Eq. (2.71), which displays h-directed instantaneous power
flow, is concentrate in ‘energy lobes’ located at h ¼ 45 , for all values of / (see
Fig. 2.9). Both electric dipole (1/r3) and inductive (1/r5) energy form are represented.
To complete the picture of the radiation behaviour of the short current element it
is instructive to examine to total power delivered by the element for a given current.
To determine the total power it is necessary to integrate the power density over a
spherical surface centred on the element and enclosing it. That is:

ZZ Z2p Zp rffiffiffiffiffi 2
k02 I02 dl2 l0 sin h 2
Prad ¼ pr ^ar  dS ¼ r sin h dh d/
32p2 e0 r 2
S 0 0
ð2:72Þ
rffiffiffiffiffi Zp
k02 I02 dl2 l0
¼ 2p sin3 h dh
32p2 e0
0

Fig. 2.9 Power and energy


flows in near and far field
regions for short current
element
2.6 Hertzian Dipole 55

Rp
But sin3 h dh ¼ 4=3 and k0 ¼ 2p
k0 where k0 is the free space wavelength. Hence:
0

 
pg0 I02 dl 2
Prad ¼ ð2:73Þ
3 k0

This can be further reduced to a now traditional form:


 2
dl
Prad ¼ 40p 2
I02 Watts ð2:74Þ
k0

by using the relation η0 = 120p. Not unexpectedly, the power radiated is propor-
tional to the square of the current applied, but more importantly it is proportional to
the square of the element length normalised to the free space wavelength. This ratio
occurs repeatedly in antenna theory and underpins the unmistakeable trend that
antennas largely diminish in size as frequency rises.
In electrical circuits it is useful to express power in terms of a resistance, and so
it is convenient to make the relationship:

Prad ¼ Irms
2
Rrad ð2:75Þ

where Irms is the root mean squared current in the source, to yield:

1
Prad ¼ I02 Rrad ð2:76Þ
2

Rrad is termed the radiation resistance of the source. On combining Eqs. (2.76)
and (2.74), the following classical equation is obtained:
 2
dl
Rrad ¼ 80p2 Ohms ð2:77Þ
k0

If an antenna is viewed as a device which provides a transitioning mechanism


from a guided wave in a feed line supplied from a transmitter to a free space
(TEM) wave, then radiation resistance represents the coupling between the antenna
and free space through its radiation pattern. It should be understood that Rrad has
nothing to do with ohmic loss in the antenna structure. For optimum coupling to
free space the output impedance of the transmitter, the characteristic impedance of
the feed line and the radiation resistance should be matched. For the short current
element, however, this is not possible. The assumptions incorporated into the
development of the above equations require dl=k0 \0:1, which gives Rrad < 8 X.
An antenna feed line typically exhibits a characteristic impedance between 50 and
100 X—obviously very different from Rrad. Clearly techniques are required to
56 2 Fundamentals of Electromagnetic Radiation

embed and match practical antennas into real transmission systems and these will
be explored in Chaps. 6, 7 and 8.

2.7 Chapter Summary

The role of the Maxwell equations in the formulation of a comprehensive theory of


electromagnetic radiation is explored. It is shown that in the presence of an active
source of radiation the solution of the Maxwell’s equations leads to the formulation
of second order vector differential equations with decidedly complex source
functions. Substitution techniques are required to simplify the equations and thus
effect a solution. It is demonstrated that the most common and effective substitution
entails the introduction of a magnetic vector potential A and a scalar electric
potential /. The functional forms of these intermediary terms are constructed from
the physical nature of both electric and magnetic fields at the static level. The full
time varying forms for / and A are deduced from the fundamental nature of
electromagnetic wave propagation.
While the mathematical niceties of the Hertzian dipole are not too relevant to
antennas evolved from slots (Chap. 3), the well established representations of
radiated field distributions, of power density distributions, of radiation pattern
fundamentals, of radiated power and stored energy, of radiation resistance and
impedance relations, provide an invaluable insight into the nature of radiation and
antennas.

Bibliography

1. C.A. Balanis, Antenna Theory (Wiley, New York, 1982)


2. R.E. Collin, F.J. Zucker, Antenna Theory (McGraw-Hill Book Co., New York, 1969)
3. I.S. Grant, W.R. Phillips, Electromagnetism (Wiley, Chichester, 1975)
4. F.H. Read, Electromagnetic Radiation (Wiley, Chichester, 1980)
5. F.T. Ulaby, Fundamentals of Applied Electromagnetics (Prentice Hall Inc., New Jersey, 1997)
6. J.D. Kraus, Electromagnetics (McGraw-Hill Book Co., Singapore, 1984)
Chapter 3
Compact Aperture Antennas

3.1 Aperture Antenna Models

The auxiliary potential functions introduced in Chap. 2, help to facilitate the


analysis and modelling of wire, or current carrying, antennas, including simple
reflectors as used in a wide range of applications. Their successful adoption is
dependent on knowing the surface current density everywhere on the metallic
structure. This is very often not possible as with lens antennas and dielectric rod
antennas for which there is no identifiable conduction current flow, while it is very
difficult to arrange, for antennas such as horns. These antenna types (see Fig. 3.1)
are generally referred to as aperture antennas. Those represented by sketches a, b
and c in Fig. 3.1 are horn types in which an empty rectangular waveguide opens out
into a rectangular or square aperture depending on the application. In sketches d and
e the slots cut into the broad and side walls of the waveguide form the aperture
radiators, while f represents a leaky-wave dielectric rod radiator, where the aperture
is the surface of the dielectric rod. The classification also includes reflector antennas
(g), although these are also amenable to the techniques described in Chap. 2.
To determine the fields radiated by an aperture antenna it is usually necessary to
employ a ‘field equivalence principle’ [1–3]. By adopting such a procedure, the
antenna structure of interest, which would represent a largely intractable mathe-
matical problem in its original manifestation, is typically reconfigured to a much
simpler more amenable geometrical form, while remaining in essence electro-
magnetically equivalent to the original. While this modified form may no longer
lead to an exact solution on analysis, it can be modelled more readily, and if
constructed insightfully and intelligently, generally yields more than acceptable
results. The basic concept is illustrated in Fig. 3.2a. In the left hand diagram input
electric current and magnetic current densities J and Jm (e.g. on and inside a horn
antenna) sets up a radiation field E, H in a large area of space (V2). Within this area
an arbitrary surface S of volume V1 (say) is inserted, which is invisible to the field
distribution E, H. The equivalence theorem states that if on the surface S surface

© Springer Nature Switzerland AG 2019 57


A. J. Sangster, Compact Slot Array Antennas for Wireless Communications, Signals
and Communication Technology, https://doi.org/10.1007/978-3-030-01753-8_3
58 3 Compact Aperture Antennas

Fig. 3.1 Examples of aperture antennas: a H-plane horn, b E-plane horn, c sectoral horn,
d shunt-slot in waveguide, e edge-slot in waveguide, f dielectric rod, g horn fed parabolic antenna

currents Js and Jms are imposed such that the external field E, H is unaltered, then
the introduced currents are equivalent sources for the region V2. Therefore on S,
Js ¼ n^  ðH  Hi Þ and Jms ¼ ^ n  ðE  Ei Þ where the interior of S is presumed
to contain a field distribution Ei, Hi which is specified only on S by means of the
3.1 Aperture Antenna Models 59

above boundary relations. Since the interior volume V1 is not of interest in setting
up the equivalent sources, the interior fields Ei, Hi can be arbitrarily chosen. The
most common choice is Ei = Hi = 0 which is termed Love’s equivalence principle
^  H and Jms ¼ ^
[3]. In this case, on S, Js ¼ n n  E. Furthermore this means that
the medium within S can be chosen arbitrarily. For example, a conductor can be
judiciously placed there with no electromagnetic effect on the arrangement.
In Fig. 3.2a it should be noted that both electric current density Js and magnetic
current density Jms have been inserted onto the surface S. However, according to
the uniqueness theorem [3] we need only specify Js or Jms on S, not both, to ensure
that the resultant field solution is unique. For example, consider the conducting
body S with an aperture A as shown in Fig. 3.2b for which the field in the aperture
(Ea  n ^) is known. The radiated field in the exterior region is unaffected if the
aperture in the conducting body is closed by a conducting wall as depicted in
Fig. 3.2b (right hand diagram) and replaced by the magnetic current Jms ¼ Ea  n ^
placed closely adjacent to the position of the original aperture—essentially on the
surface of the aperture wall.

Fig. 3.2 Equivalence principles a for arbitrary surface S b for conducting body S with aperture A
60 3 Compact Aperture Antennas

Let us consider applying this equivalence technique to the horn antenna


(Fig. 3.3). Firstly, the horn—the electromagnetic wave source—is enclosed
(Fig. 3.3b) by an arbitrary surface which both accommodates the ‘radiating aper-
ture’ of the source and completely encases it. It is commonly referred to as the
Huygen’s surface. A field free region is introduced within the surface S if surface
currents are introduced to form fields external to S equivalent to those generated by
the original source. This entails placing on S a distribution of electric and magnetic
surface current densities Js and Jms. If selected appropriately these current
replacements generate the original external field. Given that on a defined surface the
electromagnetic boundary conditions dictate that Js ¼ n ^  H and Jms ¼ ^ n  E,
where n ^ denotes the unit normal to the surface, the application of the equivalence
principle requires precise knowledge of both E and H over the surface S. These
field definitions must obviously be dictated by the original source geometry.
Actually, because of the relationship between E and H, built into the Maxwell
equations, it is not necessary to define both field quantities over S. It is usually
enough to define either E or H alone over the entire surface, or more commonly
E alone over part of S and H alone over the rest. The resultant external field
solution is unique, provided that on no part of S is surface current density
unspecified. This means that S can be viewed as a perfect conductor, with Jms
specified over an area adjacent to the position of the horn aperture where Ea is finite
(Fig. 3.3b). In principle this is a simpler problem to solve using auxiliary potential
functions, particularly as we are relatively free in the way we choose the shape and
form of S [3]. The E-field is imaged in the conducting surface to satisfy the
boundary conditions so that Jms ¼ 2^ n  Ea . For example, if the portion of the
surface which intercepts the aperture is presumed to be a plane perfectly conducting
surface of infinite extent, then the equivalent model now comprises two half-spaces
separated by an infinitesimally thin conducting plane, with the horn aperture field
replaced by a magnetic current sheet adjacent to the conducting plane as suggested
in Fig. 3.3c. This is clearly a much simpler radiation problem to solve than the raw
original. To solve the problem all we now need is a knowledge of the E-field, and
hence Jms, in the horn aperture, which is generally available from waveguide
theory. Hence using an appropriate auxiliary function the radiated field in the right
hand half space can be determined (see Sect. 3.2). The disadvantage of the tech-
nique is that only field patterns in the forward half space of the antenna are gen-
erated—back radiation is ignored. However, this is seldom a serious limitation
since it is the forward pattern that is of primary concern in the majority of antenna
applications whether operated in transmission or reception mode.

3.2 Auxiliary Potentials

In very much the same way as was expounded in Chap. 2, if we resort to an


auxiliary potential function to simplify the electromagnetic boundary value problem
created by a radiation source, the choice should satisfy at least one of the
3.2 Auxiliary Potentials 61

Fig. 3.3 Application of equivalence principle a original horn structure b enclosing surface S
c equivalent aperture in infinite screen

source-free Maxwell equations in an electric charge-free space. Here we are con-


cerned with magnetic current (electric field) sources and consequently an apt choice
is to set (see [3]):

1
E¼ r  Am ð3:1Þ
e0

where Am is the auxiliary potential function. Unlike the magnetic potential function
A in Chap. 2, Am has little electrical significance, other than as a useful mathe-
matical generator of electric field, and hence magnetic field.
By duality, the Maxwell equations which are applicable to the problem of
radiation from a magnetic current source of density Jm and magnetic charge density
qm are:

r  E ¼ jxlH  Jm ð3:2Þ

r  H ¼ jxeE ð3:3Þ
62 3 Compact Aperture Antennas

rD¼0 ð3:4Þ

r  B ¼ qm ð3:5Þ

The magnetic current density Jm and magnetic charge density qm are linked
through the equation:
r  Jm ¼ jxqm ð3:6Þ

which is suggestive of the Kirchoff relationship between electric current and charge
(see Chap. 2).
Following a derivation paralleling Chap. 2, these lead to two second order vector
differential equations for electric and magnetic fields, namely:

r 2 E þ k 2 E ¼ r  Jm ð3:7Þ

and

rr  Jm
r2 H þ k2 H ¼ jxeJm  ð3:8Þ
jxl

Clearly the right hand side source functions in these equations are unappealingly
complex. However the auxiliary potential function choice delineated above, when
substituted into the derivation process as outline in Chap. 2, generates a much more
amenable second order vector differential equation, as follows:

r2 Am þ k02 Am ¼ e0 Jm ð3:9Þ


pffiffiffiffiffiffiffiffiffi
where k ¼ k0 ¼ x l0 e0 for radiation into free space or vacuum. By borrowing an
obvious parallel with the construction of the solution to the analogous equation in
Chap. 2 [Eqs. (2.28) and (2.45)], it is not difficult to assert that the solution to
Eq. (3.9) must have the phasor form.
ZZZ
e0 expðjk jr  r0 jÞJm ðr0 Þ 0
Am ðrÞ ¼ dV ð3:10Þ
4p j r  r0 j
V

That Eq. (3.10) is a solution to Eq. (3.9) can be established by substitution.


Although the functional forms of Eqs. (2.45) and (3.10) are essentially identical the
volumes V are not the same. While the current element J exists in isolation in a
‘whole’ space (V an infinite volume), Jm is presumed to reside on an infinite
conducting screen (Fig. 3.3) and radiates into a half-space (V).
The evident mathematical duality that exists between radiation solutions ema-
nating from surface conduction current densities (J  Js ) on ‘wire’ structures, and
magnetic current densities (Jm  Jms ) representing field sources in apertures, is
highlighted in Table 3.1.
3.3 Radiation from a Linear Aperture 63

Table 3.1 Duality of equations associated with electric and magnetic current sources
Electric sources (J 6¼ 0, Jm = 0) Magnetic sources (Jm 6¼ 0, J = 0)
r  E ¼ jxl0 H r  H ¼ jxe0 E
r  H ¼ J þ jxe0 E r  E ¼ Jm þ jxl0 H
H ¼ l1 r  A E ¼  e10 r  Am
0

E ¼ jxA  j xl1 e0 rðr  AÞ H ¼ jxAm  j xl1 e0 rðr  Am Þ


0 0

r2 A þ k02 A ¼ l0 J r2 Am þ k02 Am ¼ e0 Jm


l0 RRR Jðr0 ;tÞ 0 e0
RRR Jm ðr0 ;tÞ 0
Aðr; tÞ ¼ 4p jrr0 j dV Am ðr; tÞ ¼ 4p jrr0 j dV
V V

3.3 Radiation from a Linear Aperture

Radiation mechanisms associated with antennas which fall into the category of
‘aperture antennas’ are attributable to diffraction from the electric field distribution
formed across the aperture at any given instant in time. The field distribution from
an aperture can be modelled in many ways (as we shall see in Chap. 4) but the
radiation mechanism can be understood most clearly and directly by examining the
electromagnetic penetration of a narrow one-dimensional slit as depicted in
Fig. 3.4. The arrangement comprises a perfectly flat conducting screen of infinite
extent in the x-y plane, containing a slit of width a in the x-direction and extending
to infinity at y. A plane electromagnetic wave in the left hand half-space is
incident on the screen and hence the slit and sets up, in sympathy with the
x-polarised plane wave, an x-directed electric field E0 in the aperture, which we can
presume to be constant across the narrow aperture.

3.3.1 Huygen’s Principle and Equivalent Sources

At every point on the E-field in the aperture it can be presumed, referring to


Fig. 3.4, that there exists a source of spreading waves into the half-space to the right
of the aperture. In the left hand half-space the back scattered waves from the slot, if
it is small are negligible by comparison with waves reflected by the screen and can
therefore be ignored at this stage. This approximate technique is termed the
Huygen’s principle. In two dimensions these waves spread with cylindrical
pffiffi
wave-fronts into the half-space with 1= r dependence on distance from the source.
That this is so, is evident from the nature of Eq. (3.10). In three dimensions they
would spread out spherically with 1/r dependence for the fields.
So let us consider an element of the slit of width dx at x in the x-direction
(Fig. 3.4). We can assume that the fields everywhere are invariant in the y-direction
for a slit of constant width for all y. In accordance with Huygen’s principle the field
within the element can be considered to be a source of cylindrical waves in the
right-hand medium (z > 0). The strength of this ‘secondary’ source will obviously
64 3 Compact Aperture Antennas

Fig. 3.4 Electromagnetic wave diffraction by a parallel sided slit in a perfectly conducting
membrane (Huygen’s mechanism)

be proportional to E0 and also to the width dx of the element. Hence the field at the
remote point P in Fig. 3.4, which is located at distance r′ from the source at (x, 0, 0)
and distance r from the origin at O, can logically be represent by a function of the
following form:

AE0 dx
dEP ¼ pffiffiffiffi expðjk0 r 0 Þ ð3:11Þ
r0

where A is an amplitude constant, and k0 is the propagation coefficient for free space
electromagnetic waves in the right-hand region. If P is very remote from the
aperture such that r  a, then we can reasonably presume that r 0  r in the
denominator of Eq. (3.11). However, for the much more sensitive phase shift we
use r 0 ¼ r  x sin h and so Eq. (3.11) becomes:

AE0 dx
dEP ¼ pffiffi expðjk0 ðr  x sin hÞÞ ð3:12Þ
r

where h is the angle between OP and the z-axis in Fig. 3.4.


3.3 Radiation from a Linear Aperture 65

The total diffracted field at P due to the field E0 in the aperture can now be
determined by summing all of the dx contributions at P for all values of x between
−a/2 and +a/2. This results in the following integral equation:

Z a=2
þ
AE0
EP ðr; hÞ ¼ pffiffi expðjk0 rÞ expðjk0 x sin hÞdx ð3:13Þ
r
a=2

which, on performing the integration, reduces to:

AE0 a sinðpka0 sin hÞ


EP ðr; hÞ ¼ pffiffi expðjk0 rÞ ð3:14Þ
r pka0 sin h

Significantly, the form of this equation reveals that the diffracted wave at P from
the superposition of cylindrical waves, emanating from all parts of the aperture
containing the field E0, combine to form an ‘interference pattern’ which in this case
has a characteristic sin X/X distribution. Distributions of this nature are ubiquitous
in aperture antenna theory, and they will be examined in more detail in the fol-
lowing section.

3.3.2 Plane Wave Spectrum

In several branches of electromagnetism, the wave solutions to the electromagnetic


boundary value problem are too complex to recruit directly into device develop-
ment. Fortunately however, such solutions can usually be resolved into constituent
plane waves. A good example [2] is rectangular, or circular cylindrical, waveguide
where the behaviours of the TE or TM modes can best be implemented by
expressing them in plane wave terms (Appendix B). This approach can also be used
to advantage in certain radiation problems, particularly those of the aperture type.
Mathematically, as we shall see, the plane wave spectrum technique, as it is termed,
is not difficult to implement for antennas with known aperture field shapes, and can
lead to quick and reliable predictions of an antenna’s radiation characteristics.
The plane wave spectrum of a radiating aperture is a fundamentally more rig-
orous means of estimating the diffraction pattern than the Huygen’s method. For
example, a point on the cylindrical wave front depicted in Fig. 3.4 can locally be
viewed as comprising a ‘plane’ wave within an elemental sector dh, particularly for
large r values. This idea is illustrated in Fig. 3.5. The combination of plane waves
representing all sectors dh between h = −p/2 and h = +p/2 mathematically
reconstructs the original cylindrical wave front. This combination of plane waves is
usually termed the angular spectrum, or spectrum function (F(h)), and for the case
of a very narrow slit it has a constant magnitude. Actually this function is con-
ventionally expressed as F(sin h) rather than F(h).
66 3 Compact Aperture Antennas

Fig. 3.5 Plane wave diffraction at a narrow slit in an infinite perfectly conducting screen—
illustrating the diffraction pattern is a spectrum of plane waves

On the other hand for a radiating slit of finite width the spectrum function is by
no means constant. In this more general case, it can be evaluated be reference to
Fig. 3.5, where the slit has a width a in the x-direction. As in the Huygens example
the slit is irradiated from the left by a single plane wave of E-field amplitude E0.
The figure shows a typical plane wave component of the radiated power in the right
hand half-space. It propagates in the direction k0 which is at an angle h to the z-axis.
pffiffiffiffiffiffiffiffiffi
If the right hand region is free-space the magnitude of k0 is given by k0 ¼ x l0 e0 ,
and the direction is given by that of the constituent plane wave. The point P(x, z)
(Fig. 3.6) is expressed in the cylindrical coordinate frame by the vector r. The plane
wave propagating in the direction h subtends angle dh at the origin, and logically
the wave magnitude will be proportional to both F(sin h) and dsin h, on using sin h
rather than h to denote angle. Hence the E-field at P due to this representative plane
wave can be expressed as:

dEP ðx; zÞ ¼ Fðsin hÞdðsin hÞ expðjk0  rÞ


ð3:15Þ
¼ Fðsin hÞdðsin hÞ expðjk0 ðx sin h þ z cos hÞÞ

and on collecting and adding all possible contributions from all typical plane waves
the following integral function is created:
3.3 Radiation from a Linear Aperture 67

Fig. 3.6 Cylindrical coordinates (r, w) of a field point P

Z1
EP ðx; zÞ ¼ Fðsin hÞ expðjk0 ðx sin h þ z cos hÞÞdðsin hÞ ð3:16Þ
1

The integration limits are taken from 1 to þ 1 to ensure that all possible
contributions to the diffracted field are included—both real and imaginary.
If the field in the aperture is expressed functionally as:

f ðxÞ ¼ EP ðx; 0Þ ð3:17Þ

then Eq. (3.16) becomes:

Z1
f ðxÞ ¼ Fðsin hÞ expðjk0 x sin hÞdðsin hÞ ð3:18Þ
1

In electrical circuit theory Fourier transforms are regularly used to relate fre-
quency to time and vice versa through the equations:

Z1
1
f ðtÞ ¼ FðxÞ expðjxtÞdx ð3:19Þ
2p
1
68 3 Compact Aperture Antennas

Z1
FðxÞ ¼ f ðtÞ expðjxtÞdt ð3:20Þ
1

So by analogy with this time/frequency transform pair, it is reasonable to suggest


that Eq. (3.18) has the transform partner:

Z1
1
Fðsin hÞ ¼ f ðxÞ expðjk0 x sin hÞdx ð3:21Þ
k0
1

This transform relationship between the angular spectrum and aperture field,
admittedly expressed in a limited two dimensional form here, in Eqs. (3.18) and
(3.21), is a very important one in antenna theory and will be considered in more
detail below.
In the elementary diffraction example presented in Fig. 3.5, the electric field, and
hence f(x), in the range of x from 1 to þ 1, is non-zero only in the slit:
therefore,

E0 for j xj a=2
f ðxÞ ¼ ð3:22Þ
0 for j xj [ a=2

On substituting this form for f(x) into Eq. (3.21) we obtain:

Z1
E0
Fðsin hÞ ¼ expðjk0 x sin hÞdx
k0
1

This integral is well known and exhibits the following solution:


 
pa sin h
E0 a sin k0
Fðsin hÞ ¼ pa sin h
ð3:23Þ
k0 k 0

It has the same angular dependence in the far-field as the radiation pattern
represented by Eq. (3.14) which was deduced from the Huygen’s principle. That
the spectrum function bears a direct relationship with the far-field pattern of an
aperture antenna is more fully developed in many texts on antenna theory [3]. This
powerful relationship remains true for all aperture antennas, not just the slit example
examined here.
3.4 Spectrum Functions 69

3.4 Spectrum Functions

The Fourier transform pair [Eqs. (3.18) and (3.21)] which expresses the relationship
between spectrum function (far-field) and an aperture field function (antenna source
field) provides an excellent base from which to examine the fundamental radiation
characteristics of aperture antennas. They are of course limited to furnishing pattern
information for the principal plane of the antenna only—that is the plane containing
the major axis of the aperture and the normal to the aperture (the x-z plane in
Fig. 3.5). Notwithstanding this limitation all the essential features of the antenna
pattern can be demonstrated. To assess antenna behaviour in planes other than the
principal plane the three dimensional forms of Eqs. (3.18) and (3.21) must be used.
These are developed in Sect. 3.5.
In Fig. 3.7 spectrum functions for four commonly used antenna aperture field
distributions are displayed in two-dimensional rectangular rather than polar form.
The functions themselves are listed in Table 3.2. The rectangular form is more
traditional but with the growing availability of sophisticated drafting tools within
EM solvers the polar representation is becoming increasingly common place. Of
course, for ‘real’ antennas radiation patterns are three dimensional, and modern
drafting software can generate 3-D images. However 3-D representations on a 2-D
page can lose subtle details of a given pattern, and so the 2-D depictions continue to

Fig. 3.7 Far-field patterns for apertures with the following field distributions—uniform, cosine,
cosine squared, cosine squared on a 10 dB pedestal
70 3 Compact Aperture Antennas

Table 3.2 Commonly used spectrum functions


Aperture distribution normalised to k0/a Spectrum function
f(x) F(sin h)
 
1 Uniform f(x) = k0/a sinðpauÞ
u ¼ sink0 h
pau
  p2 cosðpauÞ
2 Cosine f ðxÞ ¼ ka0 cos px
a 2 ðp=2Þ2 ðpauÞ2
k0
 
3 Cosine squared f ðxÞ ¼ a cos2 px
a
sinðpauÞ 1
pau 1ðauÞ2
h  2 i h i
4 Parabolic f ðxÞ ¼ ka0 1  2x
a
3
ðpauÞ2
sinðpauÞ
pau  cosðpauÞ
h   2 i2 h h i i
5 k0 15 3 sinðpauÞ
 cosðpauÞ  sinðpauÞ
Parabolic squared f ðxÞ ¼ a 1  2x
a ðpauÞ2 ðpauÞ2 pau pau
h i
6 Cosine squared on 10 dB pedestal sinðpauÞ 10:481ðauÞ2
   pau 1ðauÞ2
f ðxÞ ¼ ka0 0:3162 þ 0:6838 cos2 px
a

be used to ensure that the pattern of interest is fully recorded. Normally this means
that at least two planar patterns are required representing the major planes (or ‘cuts’)
of the antenna. These are usually chosen to be orthogonal to each other. The
traditional planes employed are predominantly the vertical and horizontal planes for
ground mounted antennas. These are also termed the elevation and azimuth planes
particularly in radar applications. The reference direction au = 0 (i.e. h = 0°) is
generally assumed to be aligned with the normal to the surface of the aperture.
These directions coincide in practice if the antenna is symmetrically distributed
relative to h = 0° axis.

3.4.1 Pattern Nulls and Sidelobes

The primary function of any electromagnetic wave transmitting or receiving device


is to produce a predominant ‘look’ direction which provides a distinctly optimum
level of transmission or reception, with signals in all other directions being largely
suppressed. This is reflected in the 2-D radiation patterns depicted in Fig. 3.7.
Normally, any given pattern contains a single main lobe and several subsidiary
sidelobes in real space—usually defined as −90° < h < 90°. The sidelobes repre-
sent wasted power in transmission, and non-optimum suppression on reception.
They are caused by unavoidable diffraction processes in a finite size aperture, and
their distribution and magnitude are a function of the ratio a/k0. The predominant
requirement of the antenna design engineer is to ensure that the sidelobes are kept
as low as possible with respect to the mainlobe, and avenues for achieving this are
suggested in the above figures. These show that illumination functions (f(x)) which
are shaped to produce low edge fields in the aperture produce low sidelobes.
However, at a fixed a/k0 this is achieved at the expense of widening the angle
subtended by the mainlobe. On the other hand, the main beam width, which
determines the directivity of the antenna, can be reduced by increasing a/k0. This
3.4 Spectrum Functions 71

stratagem introduces more nulls into the pattern, and hence moves the first null
(nearest the mainlobe peak) toward au = 0, that is h = 0, thus ‘squeezing down’ the
mainlobe width.
For example if a = 0.3 m and k0 = 3 cm the first null occurs at 5.7° for a
uniform aperture illumination (Fig. 3.7). If a is now increased to 0.6 m the first null
shifts to 2.85° thus narrowing the primary lobe. The magnitude of the first sidelobe
relative to the peak of the mainlobe, as can be seen from Fig. 3.7, has the magnitude
0.22 = −13.1 dB. For most applications this would normally be viewed as too high
and aperture field shaping would be introduced to lower it.

3.4.2 Mainlobe Beamwidth

The angular width of the mainlobe of a directional antenna is arguably the


parameter of primary concern in relation to achieving a design which meets a
specific electromagnetic wave transmission or reception requirement. In two
dimensions, the definition of this lobe-width or beamwidth could simply be the
angular distance between first nulls as the spectrum patterns suggest. In three
dimensions, as we shall see in Sect. 3.5, beamwidths in at least two orthogonal
planes are usually required to define the lobe.
For the example considered above, of a 0.3 m wide uniformly illuminated slit
radiator, the distance between first nulls can be deduced from Fig. 3.7 (solid red
curve) and gives a null beamwidth of 11.4°. While in theory this is a perfectly
acceptable definition, in practice this is not so, for three very sound reasons:
1. There may not be a null—a feature of some limited pattern types.
2. If the null exists, its precise location may be difficult to determine accurately.
3. The minimum detectable level of an EM signal in any practical measurement is
well above zero field, as a result of noise and other signal perturbations.
As a consequence of these difficulties the convention, which has sensibly
evolved, defines beamwidth as the angular separation in a given plane, between the
points on either side of the mainlobe, which are 3 dB in power density below the
density at the beam maximum. This is termed the half-power beamwidth (HPBW)
and is identified on the spectrum pattern in Fig. 3.7. Note that the pattern is nor-
malised to give a peak value of unity. Since the spectrum actually represents a field
pffiffiffi
pattern, the 3 dB positions are located where the field is 1= 2 below the peak
value. The HPBW is determined by equating the spectrum function (f(x)) to 0.707.
In the case of uniform illumination this gives: sin pau ¼ 0:707pau with h ¼ h3 dB .
For example, for a 0.3 m wide uniformly illuminated slit at 10 GHz (k0 = 3.0 cm)
the beamwidth from Fig. 3.7 is approximately 5.1°, whereas for a cosine illumi-
nation it is 6.8°. Thus the improved sidelobe performance associated with the cosine
illumination is obtained at the expense of some deterioration in beamwidth. It is
assumed that a/k0 is the same in both cases.
72 3 Compact Aperture Antennas

3.4.3 Pattern Gain

A property of a directional antenna that is arguably of primary interest to users,


whether in reception or transmission roles, is its gain. In any physically realisable
antenna there will generally be one direction in which the radiated or collected
power density is greatest. To give a quantitative value to this optimum operating
condition, the power density at this pattern location is compared with some gen-
erally accepted reference level. This level is provided by estimating the power
density in the same distance and direction produced by an isotropic source or sink
delivering (or receiving) the same total power as the original antenna. As we have
already noted, an isotropic antenna notionally produces an omni-directional or
spherical radiation pattern. So, for a lossless antenna, or for a low loss antenna for
which resistive and other losses can be ignored, the ratio of the peak power density
delivered or received by the test antenna in the direction of the mainbeam to that
produced by an equivalent isotropic antenna is termed the directive gain or direc-
tivity (D). On the other hand, when the antenna losses are not ignored, the resultant
gain (G) is inevitably lower, and it is usually termed the power gain. Directivity and
gain are related through the relation:

G ¼ gD ð3:24Þ

where η (the efficiency) is a measure of the total power transmitted by the given
antenna divided by the total power delivered to it, or in reception terms the total
power delivered to the receiver divided by the total power received by the antenna.
Directivity and gain will be examined further in Sect. 3.5.

3.5 Rectangular Aperture

While the examination of linear apertures has the advantage of simplifying the basic
analysis process associated with the evolution of radiation patterns and with
identifying the nature and properties of these patterns, we still need tools to
investigate practical aperture antennas. These real antennas exhibit, of course, two
dimensional apertures, and hence they generate three dimensional radiation pat-
terns. Here it will be demonstrated that two dimensional apertures can be modelled
by revisiting Eq. (3.10) and extending its applicability to two dimensions.
In the computational modelling of two dimensional aperture problems the main
difficulty lies in performing the integrations embedded in Eq. (3.10). However, if
interest is limited to the far-field, which is the normal scenario, the complexity of
the integrations can be significantly reduced. It is helpful at this stage to make some
simplifying definitions. Firstly in the far-field the following approximations can be
made with little impact on accuracy:
3.5 Rectangular Aperture 73

R  jrj ¼ r ðfor magnitude changesÞ ð3:25Þ

R ¼ r þ r 0 cos n ðfor phase changesÞ ð3:26Þ

Here n denotes the angle between the vectors r and r′ which are depicted in
Fig. 3.8.
For far-field computations n  0 in which case R and r are essentially parallel.
Hence applying Eq. (3.10) to a radiating aperture which can be represented by a
surface S, we obtain:
ZZ
e0 expðjk0 RÞ 0
Am ¼ Jm dS ð3:27Þ
4p R
S

If we now incorporate Eqs. (3.25) and (3.26) into Eq. (3.27), Am can be written
as:
ZZ
e0 expðjk0 rÞ
Am  Jm expð þ jk0 r 0 cos nÞdS0 ð3:28Þ
4pr
S

In the polar (r, h, /) coordinate system the vector Am has two components in the
far-field, namely:

Am ¼ ah Amh þ a/ Am/ ð3:29Þ

Fig. 3.8 Geometry for two-dimensional aperture antenna analysis


74 3 Compact Aperture Antennas

where ah and a/ are unit vectors. There is no Amr since in the far-field the radiation
is transverse electromagnetic in character. At the position of the aperture surface at
r′ (x′, y′, z′) the vector magnetic current can be expressed as:

Jm ¼ ax Jmx þ ay Jmy þ az Jmz ð3:30Þ

Hence combining Eqs. (3.28), (3.29), and (3.30) yields the following accessible
forms:
ZZ
e0 expðjk0 rÞ 
Amh ¼ Jmx cos h cos / þ Jmy cos h sin /  Jmz sin h expð þ jk0 r 0 cos nÞ dS0
4p
S

ð3:31Þ
ZZ
e0 expðjk0 rÞ 
Amu ¼  Jmx sin / þ Jmy cos / expð þ jk0 r 0 cos nÞ dS0 ð3:32Þ
4p
S

The electric and magnetic fields in the far-field can be generated using the
appropriate equations in Table 3.1—noting that the divergence of Am is zero in the
far-field.
Hence we obtain, finally, the far-field solution for the general two dimensional
aperture antenna:

Er  0 ð3:33Þ

jk0 expðjk0 rÞ
Eh   Amu ð3:34Þ
4pr
jk0 expðjk0 rÞ
E/  þ Amh ð3:35Þ
4pr
Hr  0 ð3:36Þ


jk0 expðjk0 rÞ Amh
Hh   ð3:37Þ
4pr g0


jk0 expðjk0 rÞ Am/
H/   ð3:38Þ
4pr g0

3.5.1 Uniformly Illuminated Rectangular Aperture

It is useful to establish the effectiveness of these equations by examining a standard


example which can be found in Balanis [3], Chap. 12. It comprises a rectangular
aperture of width a and height b located within an infinite perfectly conducting
3.5 Rectangular Aperture 75

Fig. 3.9 Rectangular aperture antenna embedded in a conducting ground plane of infinite extent

ground plane (see Fig. 3.9). The electric field in the aperture is assumed to be
constant in both the x- and y-directions and to be y-polarised. This is not a practical
field distribution but it serves the purpose of the demonstration with minimal
complexity. This field distribution can be expressed as:

Ea ¼ ay E0 ð3:39Þ

existing in the range a=2 x0 a=2 and  b=2 y0 b=2. E0 is a constant in this
case.
The equivalent solvable geometry comprises a ‘closed’ half-space bounded by a
flat conducting screen extending over the entire x-y plane at z = 0. Hence applying
the boundary conditions arising from Fig. 3.3, we obtain:
76 3 Compact Aperture Antennas


a/2 x0 a=2
n  ^ay Ea ¼ 2^ax E0
Jms ¼ 2^ ð3:40aÞ
b/2 y0 b=2

Jms ¼ 0 feverywhere outside aperture ð3:40bÞ

The far-fields radiated by this magnetic current distribution located in a rect-


angular aperture can be found by inserting definitions (3.40a) and (3.40b) into
Eqs. (3.31) and (3.32) to evolve the potential function components, and hence into
Eqs. (3.33) to (3.38) to generate the radiated fields. The result is:

Er ¼ 0 ð3:41Þ
 
jabk0 E0 expðjk0 rÞ sin X sin Y
Eh ¼ sin / ð3:42Þ
2pr X Y
 
jabk0 E0 expðjk0 rÞ sin X sin Y
E/ ¼ cos h cos / ð3:43Þ
2pr X Y

Hr ¼ 0 ð3:44Þ

E/
Hh ¼  ð3:45Þ
g0

Eh
H/ ¼  ð3:46Þ
g0

X and Y in these equations are given by:

k0 a k0 b
X¼ sin h cos / and Y¼ sin h sin /
2 2

The normalised E-field distributions typical of the above equations are plotted
using standard mathematical software in Fig. 3.10 for a rectangular aperture with
a = 3ko and b = 2ko, and in Fig. 3.11 for a square aperture with a = b = 3ko. Note
that in Fig. 3.10 the E-field is y-polarized (across the narrow dimension) in
Fig. 3.11 it is x-polarized.
Figures 3.10 and 3.11 provide a good pictorial representation of the penetration
of the aperture radiation pattern into the forward half space of the antenna, provided
it is fully realised that the depicted lobe surfaces are actually constant magnitude
power density contours. They could be viewed as ‘iso-density’ surfaces. Lower
power density contours would push much further out into the half-space, while
higher power densities produce surfaces closer to the antenna. Actually, experi-
mentally this form of presentation is seldom used because the volume of mea-
surement and the consequent time involved in generating such images is much too
high to be cost effective. So, as we have intimated earlier the convention is to
3.5 Rectangular Aperture 77

Fig. 3.10 Three-dimensional E-field pattern for a rectangular aperture (3ko  2ko) in a large
conducting plane—constant Ey-directed case

Fig. 3.11 Three-dimensional E-field pattern for a square aperture (3ko  3ko) in a large
conducting plane—constant Ex-directed case

present antenna patterns as two primary ‘cuts’—generally the E-plane and the
H-plane (see Fig. 3.10). As the figure demonstrates the E-plane for the aperture is
the y-z plane or the / = p/2 plane. The E-field distribution on this plane is
therefore:

Er ¼ E/ ¼ 0 ð3:47Þ
"  #
jabk0 E0 expðjk0 rÞ sin k20 b sin h
Eh ¼ ð3:48Þ
2 sin h
2pr k0 b
78 3 Compact Aperture Antennas

On the H-plane (x-z plane or / = 0 plane) the E-field has the form:

Er ¼ Eh ¼ 0 ð3:49Þ
"  #
jabk0 E0 expðjk0 rÞ sin k20 a sin h
Eu ¼ cos h ð3:50Þ
2 sin h
2pr k0 a

Interestingly, the term inside the square brackets in Eqs. (3.48) and (3.50) has
reduced to the spectrum functions (see Table 3.2) of the uniformly illuminated
linear apertures of length b and of length a respectively. The important conclusion
here is that the radiation patterns of two dimensional apertures can usually be
adequately represented on the two major planes, or ‘cuts’, by recruiting the spec-
trum functions. These functions generally provide good first approximations to
mainlobe beamwidths, the locations of the sidelobe zeros and maxima, as well as
the magnitudes of the sidelobes.

3.5.2 Directivity

As discussed in Sect. 3.4.2 the directivity of an antenna gives a measure of the


enhanced power density at the peak of the mainlobe of an antenna of finite size
relative to that of an isotropic antenna delivering the same power at the same
far-field radius. If the antenna power density at the mainbeam peak at radius r is pa
(W/m2) and the total power radiated is Prad (Watts) the directivity of the antenna is
defined as:

pa 4pr 2 pa
D¼ ¼ ð3:51Þ
Prad =4pr 2 Prad

For an aperture antenna, as described above, if the electric and magnetic fields in
the far-field have already been determined, as detailed in the previous section, then
the radial power density flow at any angle in the forward half-space is given by:

1    

Uðh; /Þ ¼ Re ^ah Eh þ ^a/ E/  ^ ah H h þ ^
a/ H /
2
1   2  ð3:52Þ
¼ jEh j2 þ E/ 
2g0

If at the ‘nose’ of the mainlobe Eh ¼ Eh0 and E/ ¼ E/0 then the power flow
density there we can computed from:
3.5 Rectangular Aperture 79


1  0 2  0 2
pa ¼ Eh þ E/  W/m2 ð3:53Þ
2g0

The total power Prad is more difficult to evaluate. It involves a surface integration
of the general power density term U(h, /) over a hemisphere in the forward
half-space of the radiating aperture (see for example Balanis [3]).
However, since the chosen aperture model is embedded in a plane conducting
surface of infinite extent a simpler procedure exists to compute the total radiated
power. If the electric field ðE0 ðx0 ; y0 ÞÞ in the aperture is available, which it usually is
for known excitation methods, the power density in the aperture can be calculated.
Thence the total power entering the forward half-space namely Prad, can be com-
puted by integrating the power density over the aperture area. That is:
ZZ
1
jE0 ðx0 ; y0 Þj dx0 dy0
2
Prad ¼ ð3:54Þ
2g0
S

where S is the aperture surface. For the uniformly illuminated rectangular aperture
in Fig. 3.9 this gives:

jE0 j2
Prad ¼ ab W ð3:55Þ
2g0

Also, for this aperture example, Eq. (3.53) in combination with Eqs. (3.48) and
(3.50), evaluated at h = 0, yields:
2
ab jE0 j2
pa ¼ W/m2 ð3:56Þ
k 2g0 r 2

Thus, for this particular case we obtain:

4ppa 4pab 4pAeff


D¼ ¼ 2 ¼ ð3:57Þ
Prad k0 k20

where Aeff is the effective area of the aperture. It is worth stressing that, Eq. (3.57) is
actually a very useful relationship for gauging the directivity (and hence gain) of
any aperture antenna, not just for the cited example.

3.6 Chapter Summary

By taking advantage of well-established field equivalence theorems, it is demon-


strated here that a duality exists between (a) time varying electric surface current
density (Js) on a metal interface in air, which is the source electromagnetic wave
80 3 Compact Aperture Antennas

radiation from ‘wire’ antennas, and (b) a magnetic surface current density (Jms) on a
non-conducting interface, which is the source of radiation from an aperture.
Consequently, it becomes possible to take advantage of a suite of analytical pro-
cedures, which have been evolved to explain and simulate wire antenna charac-
teristics (see Chap. 2), and apply them to the related but separate area of
investigation presented by aperture antennas, including slots.
From the perspective of electromagnetic theory the duality between electric
current sources and magnetic current sources is emphasised in a table comparing
and contrasting the fundamental equations. Furthermore, by employing Huygen’s
principle and plane wave spectrum techniques, the precise and uncompromising
relationships between aperture size and shape, and radiation pattern characteristics,
are exposed and explored. As we shall see these seemingly unyielding physical
constraints on radiation pattern manipulation have major significance for compact
slot based antennas.

Bibliography

1. R.E. Collin, F.J. Zucker, Antenna Theory (McGraw-Hill, New York, 1969)
2. R.F. Harrington, Time Harmonic Electromagnetic Fields (McGraw-Hill, New York, 1961)
3. C.A. Balanis, Antenna Theory (Wiley, New York, 1997)
Chapter 4
Computational Modelling Techniques
for Slot Antennas

4.1 Introduction—Inhomogeneous Helmholtz Equation

In so far as the source present Maxwell equations (Eqs. 2.5–2.12) provide a rig-
orous theoretical model of electromagnetic radiation in a properly specified source
present space, so the second order differential equations, repeated below, namely
Eqs. (4.1) and (4.2) are rigorous representations expressed in terms of the vector
potential functions A and Am defined in Chap. 2.

r2 A þ k02 A ¼ l0 J ð4:1Þ

r2 Am þ k02 Am ¼ e0 Jm ð4:2Þ

The heuristically deduced solutions (see Chaps. 2 and 3) are:


ZZZ
l Jðr0 ; tÞ 0
Aðr; tÞ ¼ 0 dV ð4:3Þ
4p j r  r0 j
V
ZZZ
e0 Jm ðr0 ; tÞ 0
Am ðr; tÞ ¼ dV ð4:4Þ
4p jr  r0 j
V

Although these solutions to the second order differential equations have been
constructed by analogy with statics it should be stressed that the differential
equations themselves remain rigorous provided that the parameters and the solution
space are properly defined, as we have already emphasised.
However, for incorporation into computer simulation software, where wide
applicability is a prime goal, a general approach to the solution of second order
differential equations generated by electromagnetism (and other disciplines) is

© Springer Nature Switzerland AG 2019 81


A. J. Sangster, Compact Slot Array Antennas for Wireless Communications, Signals
and Communication Technology, https://doi.org/10.1007/978-3-030-01753-8_4
82 4 Computational Modelling Techniques for Slot Antennas

required. The required versatility can be established by noting that Eqs. (4.1) and
(4.2) can be rewritten in the forms:

½r2 þ k02 A ¼ l0 J ð4:5Þ

½r2 þ k02 Am ¼ e0 Jm ð4:6Þ

Equations of this type belong, generally, within the mathematics of linear spaces
and operators (see Harrington [1]), where they take the form:

LðfÞ ¼ g ð4:7Þ
 
Here L denotes the scalar operator ( r2 þ k02 in the above equations) while
vector function f is the operated upon function and vector g represents the range of
the operator domain [the source term in the inhomogeneous Eqs. (4.5) and (4.6)].
Equation (4.7) can be written in the form:

Lð^ ^g
uf Þ ¼ u ð4:8Þ

where u^ denotes the unit vector, in the direction of g, which applies on both sides of
the equation for a linear operator.
In Eq. (4.8), if f and g are functions of a continuous variable x, then evidently:

Lðf ðxÞÞ ¼ gðxÞ ð4:9Þ

Also, if L is a linear operator, as we have already supposed, then it must [1] have
an inverse L−1 which leads to an inverse form of Eq. (4.9) namely:

f ðxÞ ¼ L1 gðxÞ ð4:10Þ

For an operator L as defined by Eq. (4.9) it is possible to show that a function


‘ðx; x0 Þ termed the kernel function [2, 3] must exist (see Appendix C), such that:
Z
gðxÞ ¼ ‘ðx; x0 Þf ðx; x0 Þdx0 ð4:11Þ

Of particular interest to electromagnetic modelling is the specific kernel dðx; x0 Þ


which exhibits the property that for any continuous function e(x), the following
very useful formulation becomes available:
Z
eðxÞ ¼ dðx; x0 Þeðx0 Þdx0 ð4:12Þ

dðx; x0 Þ is termed the Dirac delta function.


More importantly, this property of the Dirac function leads to one of its most
significant applications, namely in Green’s functions, which form a very powerful
4.1 Introduction—Inhomogeneous Helmholtz Equation 83

theoretical treatment of second order inhomogeneous differential equations (see


Bibliography).
In general mathematical terms, Eqs. (4.9) and (4.10) respectively represent an
inhomogeneous differential equation with operator L and its integral solution.
Furthermore g(x) is a continuous function of x, and x can be viewed as a vector with
n components x1, x2, x3 …, xn. The d-function can now be employed to express the
right hand side of Eq. (4.9) in the form:
Z
gðxÞ ¼ dðx; x0 Þgðx0 Þdx0 ð4:13Þ

and recalling that L operates only on the non-primed x values we obtain in place of
Eq. (4.10):
Z
f ðxÞ ¼ gðx0 ÞL1 dðx0  xÞdx0 ð4:14Þ

Consequently, we can deduce that the solution to Eq. (4.9) takes the form:
Z
f ðxÞ ¼  gðx0 ÞGðx; x0 Þdx0 ð4:15Þ

We can make this change provided the introduced term G(x, x′) is a solution of

Gðx; x0 Þ ¼ L1 dðx0  xÞ ð4:16Þ

G(x, x′) is defined as the Green’s function and is a solution to the second order
inhomogeneous differential equation for which the source term is the d-function.
This may or may not be the whole solution. The right hand side of Eq. (4.16) could
contain a non-singular term, for example Go(x, x′), which is a solution only of the
homogeneous equation, not unlike the ‘constant’ in conventional integration:

LðGo Þ ¼ 0 ð4:17Þ

Consequently Eq. (4.16) should more correctly be written:

Gðx; x0 Þ ¼ L1 dðx0  xÞ þ Go ð4:18Þ

Hence, we have:

L Gðx; x0 Þ ¼ dðx0  xÞ þ L Go ðx; x0 Þ ¼ dðx0  xÞ ð4:19Þ

The additional term Go(x, x′) ensures that all the relevant boundary conditions,
and initial conditions are satisfied, leading to a unique solution. In electromagnetic
radiation terms this formulation can reasonably be viewed as the modelling of
radiation from a ‘point’ emitter. For a defined space containing the point source of
84 4 Computational Modelling Techniques for Slot Antennas

radio waves, G(x, x′) is the unique electromagnetic field solution which satisfies the
boundary conditions.
In physical terms, Eq. (4.9) can be interpreted as expressing the relationship
between the given source g and the field distribution which it generates. The
Green’s function expresses that part of the field at a distant point x (at some later
instant t) which is generated by a unit intensity source located at the point x′ (and
earlier time t′) within the source volume. The total field f can then be formed by
integrating (see Eq. 4.10) over all source locations x′ and over all time intervals
over which it acts.

4.2 Common Green’s Functions in Electromagnetic


Radiation

In Chap. 2 it has been shown that in a bounded volume of space containing a source
of radiation, the solution to Maxwell’s equations leads to the formulation of an
inhomogeneous scalar Helmholtz equation in the magnetic vector potentials A, and
a corresponding one Am, with source terms representing electric current density J or
magnetic current density Jm, respectively. It should be noted that the Helmholtz
equation is not the only second order differential equation which emanates from
Maxwell’s equations. However, these other forms are encapsulated within the
vector potential formulation. So we need examine only the Green’s functions which
are applicable to the Helmholtz equation. There is actually a range of such Green’s
functions for the solution of antenna related problems. These arise to accommodate
changes in boundary conditions.
For the Helmholtz equation (Eqs. 4.5 and 4.6) we have demonstrated in the
preceding section (see Eq. 4.19) that an appropriate Green’s function is a solution
of:

½r2 þ k 2 Gðr=r0 Þ ¼ dðr0  rÞ ð4:20Þ

where r denotes the point (x, y, z) while r0 denotes the point ðx0 ; y0 ; z0 Þ as shown in
Fig. 4.1.
In free space where the origin is naturally placed at zero (i.e. r0 ¼ 0) the Green’s
form of the scalar Helmholtz equation becomes:

½r2 þ k 2 GðrÞ ¼ dðrÞ ð4:21Þ

This equation can be solved rigorously [4, 5] if the Sommerfeld radiation con-
dition (phasor form) at infinity is applied. It is:
4.2 Common Green’s Functions in Electromagnetic Radiation 85

Fig. 4.1 Field and source coordinate system

jrGj [K 
lim r @G ð4:22Þ
r!1 @r þ jkG ¼ 0

The parameter K is an arbitrary constant. Actually the upper inequality is


redundant since it must apply if the lower equation is enforced. The solution to
Eq. (4.21) is then:

expðjkrÞ
GðrÞ ¼ ð4:23Þ
4pr

If the source is restored to the point r0 in Fig. 4.1 then the scalar Green’s function
can be expressed more generally as:

expðjk jr  r0 jÞ
Gðr=r0 Þ ¼ ð4:24Þ
4pjr  r0 j

In Eqs. (4.5) and (4.6) the right hand source term is −loJ, or −eoJm, respec-
tively, rather than the delta function source as in Eq. (4.21). The solutions to
Eqs. (4.5) and (4.6) can therefore be established by employing Eq. (4.15) to yield:
86 4 Computational Modelling Techniques for Slot Antennas

ZZZ
expðjk jr  r0 jÞ 0
AðrÞ ¼ l0 Jðr ÞdV 0 ð4:25Þ
4pjr  r0 j
V0

or
ZZZ
expðjk jr  r0 jÞ
Am ðrÞ ¼ e0 Jm ðr0 ÞdV 0 ð4:26Þ
4pjr  r0 j
V0

where V′ is the source volume. Not surprisingly Eqs. (4.25) and (4.26) replicate the
heuristically deduced forms in Chaps. 2 and 3 (see Eqs. 2.45 and 3.10). It should be
emphasised here that the advantage of the Green’s function route toward a solution
to a given second order differential equation is that it offers a very general method
for a wide range of problems. A successful outcome depends on the nature of the
integral and whether or not it can be performed over the source volume V′.
There are many situations, for example where the boundary conditions are more
easily formulated for electric field E than for the equivalent magnetic vector po-
tential. In this case it can be more convenient to proceed using Eq. (2.15) or (2.19).
Equation (2.15) can be written in the form:

½r  r  k 2 EðrÞ ¼ jxlJðr0 Þ ð4:27Þ

To solve this equation using the Green’s function strategy we need an electric
dyadic Green’s function which is a solution of:

½r  r  k2 Ge ðr=r0 Þ ¼ Idðr  r0 Þ ð4:28Þ

It is also chosen to satisfy the radiation condition. The quantity I is the unit
dyadic

I ¼ ^ax ^ax þ ^ay ^ay þ ^az ^


az ð4:29Þ

and ^ax ; ^ay ; ^az ; are the unit vectors in the x, y and z directions respectively.
Some standard manipulation within the rules of vector algebra [5] demonstrates
that Ge is a solution of the vector Helmholtz equation:
 
rr
½r2 þ k2 Ge ðr=r0 Þ ¼  I þ 2 dðr  r0 Þ ð4:30Þ
k

In other words, the electric dyadic Green’s function Ge can be deduced from the
scalar Green’s function G, which satisfies Eq. (4.20) by operating on G with dyadic
operator I þ rr
k 2 . Consequently, given that the scalar Green’s function is already
known (Eq. 4.24) then we must have:
4.2 Common Green’s Functions in Electromagnetic Radiation 87

   
0  rr expðjk jr  r0 jÞ  rr
Ge ðr=r Þ ¼ I þ 2 ¼ Iþ 2 G ð4:31Þ
k 4pjr  r0 j k

With the Green’s function now known for the vector Helmholtz equation we can
deduce, from Eq. (4.15), that the solution for E has the form:
ZZZ
EðrÞ ¼ jxl Jðr0 Þ  Ge ðr0 =rÞdV 0 ð4:32Þ
V0

provided r is not within V 0 .


The inhomogeneous vector Helmholtz equation for H in Eq. (2.15) is similarly
solvable if the requisite Green’s function is available. It is a solution of:

½r  r  k 2 Gm ðr=r0 Þ ¼ r  ½Idðr  r0 Þ ð4:33Þ

where Gm ðr=r0 Þ is termed the magnetic dyadic Green’s function. Since Ge ðr=r0 Þ
and Gm ðr=r0 Þ must generate electric and magnetic fields, respectively, which obey
the Maxwell equation, it follows that

r  Ge ¼ Gm ð4:34Þ

Hence, as with the electric case, we can express the free-space magnetic dyadic
Greens function in terms of the scalar Green’s function by combining Eqs. (4.31)
and (4.34) to yield:

Gm ¼ ½rGðr=r0   I ð4:35Þ

and, analogously to Eq. (4.32), H(r) can be determined from the equation:
ZZZ
HðrÞ ¼ Jðr0 Þ  Gm ðr0 =rÞdV 0 ð4:36Þ
V0

The duality between E and H in the Maxwell equations dictates that similar
definitions exist for the magnetic current source Jm. In this case the Green’s
functions Ge and Gm are solutions of the equations:
 
rr
½r2 þ k2 Gm ðr=r0 Þ ¼  I þ 2 dðr  r0 Þ ð4:37Þ
k

½r  r  k2 Ge ðr=r0 Þ ¼ r  ½Idðr  r0 Þ ð4:38Þ


88 4 Computational Modelling Techniques for Slot Antennas

and E and H can be deduced from:


ZZZ
HðrÞ ¼ jxe Jm ðr0 Þ  Gm ðr0 =rÞdV 0 ð4:39Þ
V0
ZZZ
EðrÞ ¼ Jm ðr0 Þ  Ge ðr0 =rÞdV 0 ð4:40Þ
V0
Ge and Gm can be expressed in terms of the scalar free space Green’s function G,
in much the same way as Ge and Gm in Eqs. (4.31) and (4.35). The results are:
   
rr expðjk jr  r0 jÞ I þ rr G
Gm ðr=r0 Þ ¼ I þ 2 ¼ ð4:41Þ
k 4pjr  r0 j k2

Ge ¼ ½rGðr=r0   I ð4:42Þ

Table 4.1 demonstrates that for a source of electromagnetic radiation in free


space Green’s functions exist for solving the major inhomogeneous wave equations
shown in the left hand column. For the scalar inhomogeneous Helmholtz equation
the relevant Green’s function is expressed in full in the top two entries in column 3.
For dyadic equations listed in entries 4, 5, 6, and 7 of column 1 the appropriate
Green’s functions are formulated in terms of the scalar solution.

4.2.1 Green’s Function Solution for Hertzian Dipole

As we have seen in Chap. 2, the essence of dipole radiation can be deduced by


examining the radiation from a fictitious current element I of finite length dl
(Fig. 4.2). The element is usually located at the origin of the Cartesian system and
z-directed. If the current density (J A/m2) is uniformly distributed through the
element over a cross-sectional area A, then we can write:

Idl^az ¼ Jz dV^az ð4:43Þ

where dV = Adl.
The relevant Helmholtz equations for this set up are presented in rows 1, 3 or 4 of
Table 4.1. The most common approach is to first solve for the vector potential A.
So applying Eq. (4.15) we obtain:
ZZZ
expðjk jr  r0 jÞ 0
AðrÞ ¼ l0 Jðr ÞdV 0
4pjr  r0 j
V0
Table 4.1 Established Electromagnetic Green’s Functions
Electromagnetic Helmholtz Equation Green’s Equation Green’s function
2 2 2 0 0 0
r A þ k02 A ¼ l0 J ½r þ k Gðr=r Þ ¼ dðr  rÞ jrr jÞ
Gðr=r0 Þ ¼ expðjk
4pjrr0 j
0
r2 Am þ k02 Am ¼ e0 Jm ½r2 þ k 2 Gðr/r0 Þ ¼ dðr0  rÞ jrr jÞ
Gðr=r0 Þ ¼ expðjk
4pjrr0 j
  0
 
½r  r  k 2 EðrÞ ¼ jxl Jðr0 Þ ½r2 þ k 2 Ge (r=r0 Þ ¼  I þ rr
k 2 dðr  r Þ Ge ðr=r0 Þ ¼ I þ rr k2 G
½r  r  k 2 HðrÞ ¼ r  Jðr0 Þ ½r  r  k 2 Gm ðr=r0 Þ ¼ r  ½Idðr  r0 Þ Gm ¼ ½rGðr=r0   I
  0
 
½r  r  k 2 HðrÞ ¼ þ jxeJm ðr0 Þ ½r2 þ k 2 Gm ðr=r0 Þ ¼  I þ rrk 2 dðr  r Þ Gm ðr=r0 Þ ¼ I þ rr k2 G
½r  r  k 2 EðrÞ ¼ r  Jm ðr0 Þ
4.2 Common Green’s Functions in Electromagnetic Radiation

½r  r  k 2 Ge ðr=r0 Þ ¼ r  ½Idðr  r0 Þ Ge ¼ ½rGðr=r0   I


89
90 4 Computational Modelling Techniques for Slot Antennas

Fig. 4.2 Short current


element Idl at origin of
Cartesian system

For the short current element, with V 0  dV 0 and r0 ¼ 0, this becomes:

expðjkrÞ
AðrÞ ¼ l0 Idl ^
az ð4:44Þ
4pr

which compares, as it should, with the equivalent result in Chap. 2 (Eq. 2.48).
Further detailed analysis of the distributions of electric and magnetic fields in the
radiation pattern, including power and energy flows, would precisely parallel the
discussion in Sect. 2.5 of Chap. 2, and need not be repeated here.

4.2.2 Modified Green’s Functions for Electromagnetic


Radiation

Clearly, an a priori knowledge of the current density distribution on the antenna of


interest, as in the case of the Hertzian dipole above, (and other wire antennas
radiating into free-space—see [5]), makes the Green’s function method for the
computation of radiated fields relatively straightforward. Needless to say, however,
for real antennas in complex radiation environments solutions to the relevant
inhomogeneous Helmholtz equation are generally more difficult to construct. In
practice, radiation occurs in environments where conductive reflectors and scat-
terers can be present giving rise to imaged electric current distributions which in
themselves are secondary sources of radiation. Real spaces can also contain
dielectric interfaces on which secondary magnetic current densities can appear
forming further sources of radiation.
Only in symmetrical geometries of largely simple layout is it possible to generate
modified Green’s functions capable of yielding direct solutions to the
4.2 Common Green’s Functions in Electromagnetic Radiation 91

electromagnetic boundary value problem in a manner paralleling the Hertzian


dipole example. Such modified Green’s functions have been studied comprehen-
sively in reference [4]. It encompasses, for example, the half-space, the parallel
plate waveguide, the rectangular waveguide, the cylindrical waveguide, the circular
cylinder in free-space, the elliptical cylinder in free-space, and the conducting
sphere in free-space. Here we replicate the Green’s function development for the
half-space, which is arguably the most common radiation environment encountered
in the electromagnetic simulation of antennas.
As we have seen above, the scalar inhomogeneous Helmholtz equation can be
solved, for a free-space environment, using the Green’s function G which has the
form:

expðjk jr  r0 jÞ
Gðr=r0 Þ ¼ :
4pjr  r0 j

In terms of G the electric dyadic Green’s function can be written in the form (see
Eq. (4.31)):
 
rr
Ge ðr=r0 Þ ¼ I þ 2 Gðr=r0 Þ ð4:45Þ
k

where, referring to the upper half of Fig. 4.3,


h i
2 1=2
jr  r0 j ¼ ðx  x0 Þ þ ðy  y0 Þ þ ðz  z0 Þ
2 2
ð4:46Þ

Equation (4.45) applies to the upper half of Fig. 4.3, for a point source in V′, as
if the conducting sheet were non-existent. To accommodate its presence imaging of
the source in the ‘mirror’, which the sheet represents, is the commonly applied
technique [4]. This results in the imaged source region Vi′ in the imaged Cartesian
system x′, y′, z′. This imaged source also exhibits a Green’s function contribution at
the field point P, namely:
  
1 @ @ @
Ge i ðr=r0i Þ ¼ ð^ax ^ax  ^ay ^ay þ ^az ^az Þ þ ^
r ^ax  ay ^
þ az Gðr=r0i Þ
k2 @x @y @z
ð4:47Þ

where

expðjk
r  r0i
Þ
Gðr=r0i Þ ¼ ð4:48Þ
4pjr  r0i j
92 4 Computational Modelling Techniques for Slot Antennas

Fig. 4.3 Field and source regions imaged in an infinite and perfectly conducting ground plane

and



h      i1=2

r  r0
¼ x  x0 2 þ y  y0 2 þ z  z 0 2 ð4:49Þ
i i i i

Not unexpectedly, although it appears very different, Eq. (4.47) mirrors


Eq. (4.45). The first term (in the round brackets) in Eq. (4.47) replaces the unit
dyadic. Note that the green arrows in the upper half-space indicate the positive
E-field directions due to the ‘real’ source, while the blue arrows in the lower
half-space show the corresponding field directions associated with the image.
4.2 Common Green’s Functions in Electromagnetic Radiation 93

So, the elements in Eq. (4.47) have to be shown explicitly because of the sign
differences associated with the imaged E-fields. In the second term of Eq. (4.47), a
similar situation arises with the del operation on the field coordinates of the
image-source.
Hence the Green’s function for the half-space Ge hs ðr=r0 Þ is merely the combi-
nation of Ge ðr=r0 Þ and Ge i ðr=r0 Þ. That is:

Ge hs ðr=r0 Þ ¼ Ge ðr=r0 Þ þ Gei ðr=r0 Þ ð4:50Þ

Finally, by resorting to the Maxwell equations, we can generate the magnetic


dyadic Green’s function for the half-space from:

Gm hs ðr=r0 Þ ¼ r  Ge hs ðr=r0 Þ ð4:51Þ

The fields in the upper half-space of Fig. 4.3 can now be generated by applying
Eq. (4.15), which for the electric field leads to:
ZZZ
EðrÞ ¼ jxl Jðr0 Þ  Gehs ðr0 =rÞdV 0 ð4:52Þ
V0

where Jðr0 Þ includes the imaged current density and V 0 incorporates Vi0 .
Consequently, on adopting Eq. (4.50) we obtain:
ZZZ ZZZ
0 0 0
   
EðrÞ ¼ jxl Jðr Þ  Ge ðr =rÞdV  jxl J r0i  Gei r0i =r dV 0 ð4:53Þ
V0 Vi0

4.2.3 Hertzian Dipole Adjacent to Conducting Ground


Plane

If J were chosen to represent the current density for a z-directed short current
element (current Io and length l), located distance d in front of the perfectly con-
ducting sheet, application of Eq. (4.53) leads, after some mathematical manipulation
[4, 5], to the following scalar equation for the h-component of the electric field at P:

jxl I0 l ejkr1 ejkr2
Eh ¼ sin h1 þ sin h2 ð4:54Þ
4p r1 r2

where r1 ; r2 ; h1 and h2 are defined in Fig. 4.4. Similar scalar equations can be
generated for Er and E/, using:
94 4 Computational Modelling Techniques for Slot Antennas

Fig. 4.4 Current element above a perfectly conducting ground plane and perpendicular to it
a Converging rays at point P b rays converging at infinity

Er ¼ sin h cos uEx


Eh ¼ sin h sin uEy ð4:55Þ
Eu ¼ cos hEz

Hence, on applying geometrical relationships, we observe from Fig. 4.4a, that in


general:
4.2 Common Green’s Functions in Electromagnetic Radiation 95

 1=2
r1 ¼ r 2 þ d 2  2rd cos h1
 1=2 ð4:56Þ
r2 ¼ r 2 þ d 2  2rd cosðp  h2 Þ

For far field radiation simulations (r > > d, Fig. 4.4b), h1 = h = h2 and
Eqs. (4.56) reduce, using binomial series, to:

r1  r  d cos h
ð4:57Þ
r2  r þ d cos h

In the far field the dcos h term is negligible by comparison to r which implies
that r1 and r2 are essentially parallel vectors as depicted in Fig. 4.4b. This, in turn,
means that from an amplitude variations perspective we can, with little error,
assume that:
r1  r2  r ð4:58Þ

Consequently, Eq. (4.54) can expressed more simply as:

jxl I0 l ejkr
Eh ¼ sin h½2 cos ðkd cos hÞ ð4:59Þ
4p r

for z > 0.
The term preceding the square brackets in Eq. (4.59) represents the Eh field
distribution for a single z-directed current element located at the origin of the
coordinate system in Fig. 4.4, which is modified by an array factor (term inside
square brackets—see Chap. 5) accounting for the fact that the source is at distance
d from the perfectly conducting ground plane. By employing the modified free
space Green’s function which accommodates the presence of an infinite perfectly
conducting ground plane—the half-space Green’s function—the array factor in
Eq. (4.59) results. Equation (4.59), normalised to a 0 dB maximum field strength,
is plotted in Fig. 4.5 for four d values, namely d = 0, k/4, k/2, k. The strong
influence of the ground plane on the radiation pattern is clearly illustrated.
In summary, where a modified Green’s function exists for a space containing a
radiating source then a solution for the electromagnetic fields radiated into the
designated volume can be constructed using Eq. (4.15). Unfortunately the range of
antenna configurations which can be solved in this way is limited by the not
insignificant difficulty of generating modified Green’s functions for other than
relatively symmetrical or regular spaces. For example Tai [4] has generated dyadic
Green’s functions for the following regular spaces and for volumes containing
simple geometrical obstacles:
(1) parallel plate waveguide,
(2) rectangular waveguide,
(3) cylindrical waveguide,
96 4 Computational Modelling Techniques for Slot Antennas

Fig. 4.5 Elevation plane field patterns in dB for a vertically orientated current element located at
different heights above the perfectly conducting ground plane

(4) circular cylindrical obstacle in free space,


(5) conducting elliptical cylinder in free space,
(6) perfectly conducting wedge,
(7) the half-sheet,
(8) spheres and conducting cones.
While these examples can lead to radiation solutions which provide genuine
insight into some fundamental antenna issues they are nevertheless limiting in
scope. In reality, an electromagnetic antenna can comprise two or more linked
regions, each of which may be represented by a Green’s function, but no region
separately represents a solution to the antenna problem. For example a compact slot
radiator fed from waveguide can be represented by two interlinked regions namely
the waveguide, which can be rectangular, cylindrical or elliptical, the exterior
radiation space (usually a half-space). They are electromagnetically linked through
a slot ‘cavity’ if the waveguide wall has finite thickness. If the coupled regions
possess Green’s function representations, the problem can be solved by a technique
termed the moment method [1], which involves discretisation of the slot field and
iterating toward a stable solution. It is summarised in Sect. 4.3.
We have chosen here to emphasise the moment method because generally the
technique combines accuracy with computational speed which is very important in
antenna array contexts where calculations for individual radiators may have to be
repeated many times over as the simulation as a whole advances toward a stable
solution. An alternate simulation technique termed the variational method [3] is
arguably computationally swifter but it lacks the level of accuracy required by
modern antenna design practice. On the other hand high modelling accuracy is
4.2 Common Green’s Functions in Electromagnetic Radiation 97

provided by the finite element method (see bibliography [C]) but at the expense of
computational speed arising from a very direct computational method which simply
discretizes the space to be solved and then performs perhaps millions of field
calculations over the finite element grid until stability is secured. This approach
represents a distinct disadvantage in the context of array modelling.

4.3 Concise Formulation of the Moment Method

For linear differential equations, represented in general terms by Eq. (4.9), the
method of moments (MoM) solution to this type of problem can be enunciated
briefly as follows. For a linear operator L in an equation (e.g. equation 4.7) for
which g is known, the determination of the unknown f can readily be computed as
we have seen if the region supporting f possesses a Green’s function (see Eq. 4.15).
However if the region for which f is the solution is (say) perturbed away from a
regular shape and does not exhibit a Green’s function, then a solution is still
possible by resorting to a discretisation of f [6]. That is we expand f in a series of
functions f1, f2, f3, …. defined in the domain of L. Therefore we express f as
follows:
X
f ¼ an f n ð4:60Þ
n

where an are multipliers and are constants to be determined. The fn terms are
generally referred to as basis functions or expansion functions. For an exact solution
to the relevant differential equation, the summation in Eq. (4.60) would normally be
infinite and the fn form a complete set (Appendix C) of basis functions. Otherwise
the summation is truncated, within defined mathematical rules, to minimise com-
putational effort.
For a linear operator, substitution of Eq. (4.45) into Eq. (4.9) yields:
X
an Lðfn Þ ¼ g ð4:61Þ
n

For a linear L, the relationship between f and g must also be linear in which case
an inner product (denoted by hf ; g i) must exist. For example a common, but not
unique, formulation is:

Z1
hf ; gi ¼ f ðxÞgðxÞdx ð4:62Þ
0
98 4 Computational Modelling Techniques for Slot Antennas

The moment method is constructed by defining a set of testing, or weighting


functions, w1, w2, w3, … in the domain of L. So if we form inner products of each
testing function wm (m = 1, 2, 3…) with both sides of Eq. (4.55) we obtain:
X
an hwm ; Lf n i ¼ hwm ; gi ð4:63Þ
n

Equation (4.63) actually represents a set of equations which when expressed in


matrix form encapsulates the essence of the moment method. That is:
 
Im;n ½an  ¼ ½gm  ð4:64Þ

where
2 3
hw1 ; Lf1 i hw1 ; Lf2 i     
  6 hw2 ; Lf1 i hw2 ; Lf2 i      7
Im;n ¼ 6
4              5
7 ð4:65Þ
            
2 3
a1
6 a2 7
6 7
6 a3 7
½ an  ¼ 6
6 : 7
7 ð4:66Þ
6 7
4 : 5
:
2 3
hw1 ; gi
6 hw2 ; gi 7
6 7
6 hw3 ; gi 7
½ gm  ¼ 6
6 : 7
7 ð4:67Þ
6 7
4 : 5
:

Generally the matrix (4.59) will be non-singular (i.e. have a non-zero determi-
nant), and hence will be invertible. We can therefore set up an equation for the
unknown an constants which takes the form:
 
½an  ¼ I1
nm ½gm  ð4:68Þ

Once the an constants have been determined the solution for f is given by
Eq. (4.45). The result may be exact, or approximate, depending on the original choice
of fn and wn. It is relevant to note here that the common choice of equating fn to wn
ðfn ¼ wn Þ is referred to as the Galerkin method [1]. Commercial software packages
4.3 Concise Formulation of the Moment Method 99

which have been constructed around the moment method are listed in references [7–
11]. An antenna format which is particularly amenable to the MoM, and is used as a
textbook example, is the dipole antenna. For this antenna, the method yields rapid and
accurate solutions by employing well suited sinusoidal basis functions [1].

4.4 Chapter Summary

The moment method, which has been comprehensively evaluated in this Chapter, is
shown to be an elegant mathematical technique and that it is particularly suited to
the solution of a class of electromagnetic boundary value problem created by
antennas and radiation. Importantly, in computational terms, the method provides a
source of ‘high speed’ electromagnetic solvers, by offering a problem-specific
approach which closely reflects the underlying physics. Furthermore, the rapid
growth in the application of array antennas to new roles in mobile and satellite
communications, has created design requirements for the electromagnetic mod-
elling of arrays, and this calls for the embedding of boundary value solvers within
array simulations. For these embedded MoM solvers computational speed is
essential, since they may need to be accessed many thousands of times during the
course of a design calculation.
Expressed in words, it can be averred that the MoM provides a solution to the
problem of determining the electromagnetic fields generated by a known source
located in a volume bounded by accurately defined surfaces. The mathematical pro-
cesses required to achieve this outcome are summarised in this chapter. It is
demonstrated that, in essence, the method relies on knowledge of a simpler electro-
magnetic wave solution for a comparable bounded space when excited by a unit
intensity point source within it. It is shown that mathematically such a source is
represented by a delta function. The solution produced is termed the Green’s function
for the specified region and Green’s functions for a range of bounded spaces, which
are relevant to antenna problems, are tabulated for easy referencing. The total field
solution for a defined source distribution can then be formed by integrating the point
source solution over all source locations and over all time intervals over which it acts.
Further, it is shown that the linearity of the resultant integral equations permits dis-
cretisation of the relevant functions to form matrix representations which yield readily
to computer processing. This procedure is embellished further in Chap. 5.

References

1. R.F. Harrington, Field Computation by Moment Methods (MacMillan, New York, 1968)
2. P.H. Morse, H. Feshbach, Methods of Theoretical Physics (McGraw-Hill Book Co., New
York, 1953)
100 4 Computational Modelling Techniques for Slot Antennas

3. L. Cairo, T. Kahan, Variational Techniques in Electromagnetism (Blackie & Sons Ltd.,


London, 1965)
4. Chen-To Tai, Dyadic Green Functions in Electromagnetic Theory (IEEE Press, New Jersey,
1994)
5. R.E. Collin, F.J. Zucker, Antenna Theory, Part 1 (McGraw-Hill Book Co., New York, 1969)
6. R.F. Harrington, Matrix methods for field problems. Proc. IEEE 55(2), 136–149 (1967)
7. Wave 3D, CEM Works, Winnipeg, Manitoba, Canada
8. HyperLynx 3D EM, Mentor Graphics, Fremont, California, USA
9. Antenna VLab, EM Cos, Tbilisi, Georgia
10. MMANA-GAL, DL1PBD & DL2KQ Partnership, Sebaststrasse 8, 53115 Bonn, Germany
11. R.C. Hansen (ed.), Moment Methods in Antennas and Scattering (Artech House Inc., London,
1990)

Bibliography

1. R.E. Collin, Field Theory of Guided Waves (McGraw-Hill Book Co., New York, 1960)
2. R.F. Harrington, Time-Harmonic Electromagnetic Fields (McGraw-Hill Book Co., New
York, 1961)
3. J.-M. Jin, Finite Element Methods in Electromagnetics (Wiley, New York, 2014)
Chapter 5
Moment Method Models of Compact
Slot Antennas

5.1 Introduction

In Chap. 4 it has already been noted that for open, or closed, spaces which support
time varying electromagnetic fields, excited by sources which can readily be
expressed mathematically, then the electromagnetic boundary value problem pre-
sented can be solved without undue difficulty if the Green’s function for the space in
question is available. Unfortunately, this is seldom the reality, particularly in antenna
developments. The typical problem which is encountered tends to comprise two or
more interconnected regions each capable of supporting electromagnetic waves.
A solution to such a problem requires that the fields in each region satisfy the local
boundary conditions and simultaneously the boundary conditions at interfaces. An
example of a problem of this type is illustrated in Fig. 5.1a. It comprises a narrow
and inclined slot located in the upper broadwall of the rectangular waveguide (in-
ternal dimension; width a; height b and wall thickness T). The waveguide walls are
generally assumed to be perfectly conducting. Figure 5.1b shows the slot detail
(length L; width W) and the local slot coordinate system (n, η, f).
Electromagnetically the antenna structure represented by Fig. 5.1a comprises
two regions; the free-space exterior to the waveguide (above the slot and usually
modelled as a half-space), and the waveguide region (usually modelled as infinitely
long in the z-direction). These regions are coupled electromagnetically through the
slot. The computationally most efficient, and accurate, procedure for solving this
problem, which is not untypical of a wide range of compact radiator geometries, is
the moment method. Other such geometries include slots fed from transmission
lines such as stripline, microstrip line, coplanar waveguide, but also slot fed pat-
ches, slot modified patches, cavity backed slots and combinations of all of these.
Many will be addressed in later chapters.
Other powerful electromagnetic solvers based on the finite element and finite
difference techniques can undoubtedly also be used to simulate compact slot
radiators. But these techniques, while they are in principle unlimited in their

© Springer Nature Switzerland AG 2019 101


A. J. Sangster, Compact Slot Array Antennas for Wireless Communications, Signals
and Communication Technology, https://doi.org/10.1007/978-3-030-01753-8_5
102 5 Moment Method Models of Compact Slot Antennas

Fig. 5.1 a Waveguide fed compact slot radiator and b slot detail

applications, exhibit a consequential ‘downside’ of very large memory require-


ments and relatively ‘slow’ computational speeds. While very accurate simulations
are possible, these methods, unlike the moment method, take no a priori account of
model geometry and are as a result much more computationally demanding. This
can represent a major disadvantage in the context of, for example, slot antenna array
development [1]. In synthesising such an array to achieve target radiation pattern
specifications each slot in the array, and there may be over one thousand, has to be
5.1 Introduction 103

accessed and adjusted many times. This means a slot simulation has to be entered
repeatedly and clearly the computational speed offered by the moment method
generates design advantages in this scenario. Recent developments in the use of
artificial neural networks to facilitate the optimisation of compact slot antennas [2]
involves ‘training’ and ‘testing’ of the neural system by repeated reference to the
compact slot-antenna model. So again we encounter a design situation where we
require efficient electromagnetic simulations to avoid excessive, or time consuming,
computational effort.

5.2 Moment Method Solution

The essence of the moment method is summarised mathematically in Chap. 4,


Sect. 4.3. Here we will apply the technique outlined there to the slot geometry
presented in Fig. 5.1a. It is clear from a cursory examination of Fig. 5.1a that the
primary coupled regions, namely the waveguide (region a) and the external
half-space (region c) are regularly shaped spaces thus possessing Green’s function
solutions if excited from a point source. However, in the geometry shown, the
behaviour of the fields in the slot (region b) has to be determined before the
scattered fields in the waveguide and the radiated fields in the external space can be
computed. Note that in the absence of the slot (i.e. if the slot were closed by a
perfectly conduction insert) then the waveguide would support only an incident
wave travelling from a remote matched source at one or other of its distant ter-
minations. This wave can generally be presumed to be represented by the funda-
mental TE10 mode. On the other hand, if the slot is open it becomes a source of
fields scattered into the waveguide and fields radiated into the half-space. In theory
these scattered and radiated fields could be computed using a form of Eq. (4.15) if
the instantaneous field distributions on the slot surfaces S and S′ were fully known.
While approximate estimates of the slot field distribution for a very narrow slot in a
‘zero thickness’ waveguide wall have been reported in some early papers [3, 4], the
field magnitude still remains to be computed for the wide range of possible slot
locations, orientations, or shapes, which differ from those in the original papers. To
secure a solution in the general case of a waveguide slot radiator, it is necessary to
resort to discretizing the slot fields on both surfaces. Whereupon, on application of
an iterative technique such as the moment method, a search for slot field values is
progressed until a steady state solution to the boundary value problem is estab-
lished. If the problem is properly posed, electromagnetically, then this iterative
process should converge on the unique solution.
In formulating a moment method solution for any given slot radiator geometry, a
number of conditions must be incorporated into the model, if a valid solution is to
be generated. Firstly, and obviously the simulated fields must satisfy the Maxwell
equations. Secondly, the external radiated fields must satisfy the radiation condition
104 5 Moment Method Models of Compact Slot Antennas

at infinite distance from the radiator. Thirdly, it is essential that at the inner and
outer surfaces of the slot (S and S′) that field continuity conditions are applied. The
first and second of these three conditions are immediately enforced by directing the
analysis toward a solution of the relevant inhomogeneous Helmholtz equations for
E and H (Eqs. 2.19 and 2.20) and by forming the relevant Green’s functions for the
coupled regions (a) and (c). Any solution constructed around the Green’s functions
ensures that the eventual field solutions emanating from the analysis satisfy both the
electromagnetic boundary conditions on all conducting surfaces, and the radiation
condition at infinity in region (c). Consequently, the key to formulating a set of
equations which correctly represent the radiation problem depicted in Fig. 5.1a lies
in proper establishment of the field boundary conditions at those slot surfaces which
connect regions (a), (b) and (c).
The analysis begins by identifying the electromagnetic fields which must exist in
each region in Fig. 5.1a. We have already noted that if the slot is ‘closed’ by a
fictitious perfectly conducting insert a single incident wave (TE10 mode) continues
to propagate through the waveguide. The fields of this wave are defined as Ei, Hi.
When the slot is ‘open’, and capable of interfering with the wall currents of the
TE10 mode, it becomes a discontinuity in the waveguide scattering fields back into
it. These fields we will term Ea, Ha. So, in region (a) the total fields are Ei + Ea,
Hi + Ha. The discontinuity fields on the surface S also scatters fields Eb1, Hb1 into
the slot cavity thus generating further components Eb2, Hb2 on surface S′. These are
the source of radiation into region (c) expressed through Ec, Hc. In order to
complete the electromagnetic modelling process on the structure of Fig. 5.1a it is
necessary to ensure that the above fields satisfy the boundary conditions on both
surfaces S and S′. For aperture problems (see Chap. 3) the discontinuity fields on
these surfaces are the electric field components tangential to the relevant surface.
Consequently, the boundary conditions can be formulated as:

^  Ea ¼ ^
n n  Eb1 on S ð5:1Þ

^  Eb2 ¼ ^
n n  Ec on S0 ð5:2Þ

^ points into the waveguide


Note that on the left side of Eq. (5.1) the unit vector n
whereas on the right it is directed into the slot cavity. In Eq. (5.2) the opposite is
true. Also Ei exhibits a zero tangential component on the wall containing the slot
and so does not appear in the field continuity equations.
To simplify the process of representing this problem mathematically, it is
advantageous here to employ the field equivalence principles outlined in Chap. 3 (see
Figs. 3.2 and 3.3). This principle, firstly, allows us to replace the open slot surfaces
S and S′ by an ‘electric current wall’ (i.e. a perfectly conducting membrane shown
as chain-dotted line in Fig. 5.2), which means that the regions (a), (b) and (c) be-
come respectively a classical closed rectangular waveguide, a perfect rectangular
cavity and a classical half-space. All three regions now possess Green’s function
5.2 Moment Method Solution 105

Fig. 5.2 Field equivalence representation of slot radiator in thick walled waveguide

representations as required by the moment method. Secondly, the representation is


completed with each region being excited by magnetic current densities Jm, Jm’ on
S and on S′. These replace the aperture fields E and E’ on S and S′ in the original
problem. Thus we have:

Jm ¼ n^  E on S in region a
Jm ¼ ^ n  E on S in region b
J m ¼ n  E0 on S0 in region b
0
^0

J0 m ¼ ^
n0  E0 on S0 in region c

This theoretical model is illustrated in Fig. 5.2. Each region is now separately
excited by a magnetic current and can be represented mathematically by the inho-
mogeneous Helmholtz equation rewritten below [see Eq. (2.19) and Table 4.1].

½r  r  k2 HðrÞ ¼ þ jxeJm ðr0 Þ ð5:3Þ

with zero tangential electric field on the conducting surfaces, which can be
expressed through the equation:

^  r  HðrÞ ¼ 0
n ð5:4Þ

on the metal surfaces. The appropriate Green’s equation for each region (see
Table 4.1) is:
106 5 Moment Method Models of Compact Slot Antennas

 m ðr=r0 Þ ¼ Idðr  r0 Þ
½r  r  k2 G ð5:5Þ

which can also be written in the form


 
 m ðr=r0 Þ ¼  I þ rr dðr  r0 Þ
½r2 þ k2 G ð5:6Þ
k2

 m mirrors Eq. (5.4); namely


The boundary condition on G

n  m ðr=r0 Þ ¼ 0
^rG ð5:7Þ

Once the as yet unknown Jm (or J0 m ) is evaluated the magnetic field distribution
in each region is calculable using an equation of the form of Eq. (5.8) (see Chap. 4)
which is expressed in terms of an arbitrary surface S0 and location vector r0:
ZZ
HðrÞ ¼ j xe  m ðr=r0 Þ  Jm ðr0 ÞdS0
G ð5:8Þ
So

In order to determine the variables Jm and J0 m we need to formulate a pair of


simultaneous equations for these two unknowns. These can be constructed by
applying the boundary conditions for the magnetic fields H at the slot surfaces
supporting Jm and J0 m , namely S and S′.
The formulation leads to the following:
On surface S, which supports the magnetic current density Jm, magnetic field
continuity yields:
ZZ
 
^  Hi ðrÞ ¼ jxe0 n
n ^  ma ðr=r0 Þ þ G
G  mb ðr=r0 Þ  Jm ðr0 ÞdS0
S
ZZ ð5:9Þ
 
^
þ jxe0 n G mb ðr=r0 Þ  J0 m ðr0 ÞdS0 for r on S
S

While, on surface S′ which supports the magnetic current density J0 m the con-
tinuity condition produces:
ZZ
 
^
0 ¼ jxe0 n G mb ðr=r0 Þ þ G
 mc ðr=r0 Þ  J0 m ðr0 ÞdS0
S0
ZZ ð5:10Þ
 
^
þ jxe0 n G mb ðr=r0 Þ  Jm ðr0 ÞdS0 for r on S0
S0

In Eqs. (5.9) and (5.10) G  ma ; G


 mb ; and G
 mc represent the dyadic Green’s functions
for regions (a), (b) and (c) respectively. Similar simultaneous equations are discernible
5.2 Moment Method Solution 107

in reference [5]. The relative complexity of Eqs. (5.9) and (5.10) make direct solution
unfeasible. Consequently, a numerical method designed for the task has to be adopted;
such a method is the moment method developed in the previous chapter [6, 7].
Electromagnetic radiation from a narrow rectangular slot fed from waveguide is,
as we have already intimated, not a new problem. It has been studied widely for use
as a compact radiating element in linear and planar array antennas. In today’s world
of burgeoning mobile communications, ‘compactness’ has become more critical
than array applications, and the waveguide feed arrangement has inevitably been
superseded by spatially constrained alternatives. These will be addressed in a later
Chapter. Nevertheless, the waveguide fed slot provides a cogent example of how
modelling by means of the moment method is developed in slot radiation problems.
Some of the original theoretic work on slot antennas, most notably by Oliner [3]
and Das [4], has demonstrated that the electric field distribution along the slot
length is well modelled by a single trigonometric function. Examples, generally
evolved through the application of variational techniques, are shown in Fig. 5.3.
The indicated differences are largely attributed to the focus on near resonant slots by
Oliner, and on marginally non-resonant slots by Das. This precedent suggests that
there could be potentially significant advantage, from a computational speed per-
spective, in adopting basis functions for the discretized slot field with trigonometric
properties. Mathematically this leads to the following constructions (see Eq. 4.45):

X
N
^E¼n
n ^ as fs ^ag on S ð5:11Þ
s¼1

Fig. 5.3 Slot field approximations for a narrow (w/L < 0.1), near resonant slot (L/k = 0.5) in the
broadwall (T = 0) of rectangular waveguide, as adopted in early radiation computations [Solid line
with diamond symbols = Oliner: Dashed line with square symbols = Das]
108 5 Moment Method Models of Compact Slot Antennas

X
N
n0  E0 ¼ ^
^ n0  a0s fs ^ag on S0 ð5:12Þ
s¼1

where, referring to Fig. 5.1b:

spðn þ L=2Þ
fs ðnÞ ¼ sin s ¼ 1; 2; 3; . . .; N ð5:13Þ
L

Following the procedures outlined in Chap. 4, Sect. 4.3, which are developed in
more detail in [6], we choose testing functions [Eq. (5.14)] which mirror the basis
functions, namely the Galerkin method

fi ðnÞ^ag i ¼ 1; 2; 3; . . .. . .; N ð5:14Þ

In order to generate a matrix representation of the simultaneous equation. This


has the form:
 " # " #
½ A ½C ½ a ½ h
¼ ð5:15Þ
½D ½B ½ a0  ½ 0

where:
ZZ ZZ
 
Ais ¼ jxe0 fi ^ag  f^
n G ma ðr=r0 Þ þ G
 mb ðr=r0 Þ  fs ^
an dS0 gdS ð5:16aÞ
S S
ZZ ZZ
 
Bis ¼ jxe0 fi0 ^ag  f^
n0  G  mc ðr=r0 Þ  f 0 ^
 mb ðr=r0 Þ þ G
s an dS0 gdS ð5:16bÞ
S0 S0
ZZ ZZ
 
Cis ¼ jxe0 fi0 ^ag  f^
n  mb ðr=r0 Þ  fs ^
G an dS0 gdS ð5:16cÞ
S S0
ZZ ZZ
 
Dis ¼ jxe0 n0 
fi ^ag  f^  mb ðr=r0 Þ  f 0 ^
G s an dS0 gdS ð5:16dÞ
S0 S
ZZ
hi ¼  fi ^ag  ½n
^  Hi ðrÞdS ð5:16eÞ
S

The matrix entries Ais, Bis, Cis, Dis given mathematical expression in Eqs. (5.16a,
5.16b, 5.16c, 5.16d, 5.16e) are essentially electromagnetic coupling terms repre-
senting to the mutual coupling which occurs between the assumed current distri-
butions (fi,s) on the interior and external surfaces of the radiating aperture. In the
5.2 Moment Method Solution 109

particular case of the conventional slot (in Fig. 5.1) they can be ascribed the fol-
lowing specific meanings:
Ais represents self-coupling between magnetic current elements on the inner slot
surface S.
Bis represents self-coupling between magnetic current elements on the outer slot
surface S′.
Cis represents mutual coupling to the inner slot surface (S) due to magnetic
current elements on the outer surface (S′).
Dis represents mutual coupling to the outer slot surface (S′) due to magnetic
current elements on the inner surface (S).
Once the integrals in Eqs. (5.16a, 5.16b, 5.16c, 5.16d, 5.16e) have been eval-
uated, which is mathematically routine but tedious [8, 9], the matrix Eq. (5.15) is
amenable to solution by a range of matrix manipulation tools. These can be found
in standard mathematical textbooks [10, 11] and in convenient to use computer
software packages [12]. The outcome of the process is plausible values for s, fs, a′s
and as. These can then be used to compute the scattering parameters for the compact
radiator as outlined below.

5.3 Scattering for Longitudinal Slot in Waveguide

The general broadwall inclined slot radiator depicted in Fig. 5.1a can be resolved
into a longitudinal shunt slot when h = 0°, and into a transverse series slot when
h = 90°. In fact mathematically it is more sensible to do this than attempt to solve
the inclined slot example directly [9]. In practice the inclined broadwall slot has
very few applications and most studies have been directed at modelling the series
slot and the shunt slot. Below we revisit some validating results for the shunt slot,
which is by far the most commonly used in array applications.
A cursory examination of the matrix entries in Eq. (5.16a, 5.16b, 5.16c, 5.16d,
5.16e) quickly establishes that the controlling matrix in Eq. (5.15) grows in size as
the square of the number of basis functions s, i. The more basis functions selected,
therefore, the slower will be the progress toward a computed solution. For con-
ventional slots, at least, the advantage of the use of trigonometric basis functions is
that accurate solutions are available for no more than three basis functions. This is
demonstrated in Fig. 5.4 where the slot field distribution for a longitudinal slot in
rectangular waveguide as computed by the moment method employing three basis
functions (solid curve—round symbols), is compared with the predictions of Oliner
(diamond symbols) and Das (square symbols) for the same geometry and discussed
earlier. The moment method solution converges on a slot field distribution, which is
approximately cosinusoidal as predicted by Oliner, but with a rather broader, less
peaky, shape than the pure cosine. This modified shape is also predicted by Vu
Khac (solid line—triangular symbols) employing a moment method constructed
around thirty ‘pulse’ type basis functions [13]. The Das distribution with its sharp
discontinuous peak can no longer be justified.
110 5 Moment Method Models of Compact Slot Antennas

Fig. 5.4 E-field distributions for a shunt slot radiator in the broadwall of WG16 waveguide as
predicted by the moment method (freq = 9.375 GHz, L = 0.361 k, x1 = 6.43 mm, w = 1.58 mm,
T = 1.27 mm)

The key to determining slot scattering lies with Eq. (5.8) which provides the link to
the fields generated by the slot discontinuity once Jm and J0 m are known. These become
calculable [Eqs. (5.11) and (5.12)] once as and as′ are evaluated through the moment
method computation. Given that in employing the equivalence theorem, Jm and J0 m
are positioned at the slot locations, but on ‘unbroken’ waveguide walls, the required
waveguide Greens function for calculating internal scattering is the ‘sum of modes’
formulation [9] (see Appendix C). However, real power in the transmission and
reflection signals within the waveguide is carried by the TE10 mode. Consequently we
can write for the transmitted magnetic field component of the TE10 mode H10t :
ZZ
1X N
þ
H10t ðrÞ ¼ H10i  as H10 ðrÞ H
10 ðrÞ  fs ^
an dS0 ð5:17Þ
2 s¼1
S

where H 10 are the forward and backward propagating components of the magnetic
field for the TE10 mode in the rectangular waveguide, while H10i denotes the
incident mode in the waveguide. The reflected field component H10r can be cal-
culated from:
ZZ
1X N
 þ
H10r ðrÞ ¼  as H10 ðrÞ H10 ðrÞ  fs ^
an dS0 ð5:18Þ
2 s¼1
S
5.3 Scattering for Longitudinal Slot in Waveguide 111

The derivation of closed form, computer programming friendly, algorithms from


Eqs. (5.17) and (5.18) is presented in Appendix C. It results in equations for the
parameters of primary interest for compact antennas with ‘known’ radiation pat-
terns, namely the reflection coefficient within the waveguide S11, the transmission
coefficient within the waveguide S12 and the relative power radiated into the
half-space pr/pin. Typical results are graphed and discussed in the next section.

5.3.1 Computed and Measured Results

The moment method program formulated using trigonometric basis functions has
been tested exhaustively [8, 9] to assess its sensitivity to the number of basis
functions employed. For the computation of slot radiation, and slot scattering
parameters, the sensitivity is demonstrated to be low, and it is safe to say that one
basis function is generally satisfactory for modelling conventional waveguide slot
radiators. This is demonstrated in Fig. 5.5 where the relative power radiated, as a
function of slot length, for a conventional longitudinal slot in WG16 waveguide, as
predicted by the moment method, is presented. Computations were performed for
up to five basis functions. On the scale adopted in the graph there was no dis-
cernible difference between three or more basis functions, while the single basis

Fig. 5.5 Relative power radiated by a broadwall waveguide slot as a function of slot length L,
with the number of basis functions as parameter (Freq = 10 GHz, a = 22.86 mm, b = 10.16 mm,
x1 = 2.00 mm, w = 1.5 mm, T = 0 mm)
112 5 Moment Method Models of Compact Slot Antennas

function prediction is marginally lower than the three-basis case only for below
resonance slot lengths. This is because close to resonance the slot internal field
distribution is well represented by a simple cosine, and hence one basis function is
all that is needed to generate accurate results. Needless to say this has had major
significance for the speed and efficiency of array antenna simulation incorporating
the method of moments. For example, in a multi-slot array simulation where the
MoM is embedded into an iterative program designed to perform pattern synthesis
[1], the MoM for each slot has to be accessed repeatedly, and computational speed
becomes critical.
To emphasise the point further, the trigonometric formulation with three basis
functions has been compared with the Vu Khac moment method [13] employing 30
pulse basis functions. The comparison was carried out on a longitudinal slot in
WG16 waveguide at a frequency of 9.375 GHz. The WG16 had a wall thickness of
1.27 mm and the shunt slot, with w = 1.58 mm, was off-set from the sidewall of the
waveguide by 1.43 mm. The results are presented in Figs. 5.6 and 5.7 which
display reflection coefficient magnitude |S11| and transmission coefficient magnitude
|S12| as a function of slot length L.
It is clear from Figs. 5.6 and 5.7 that the trigonometric moment method
employing three basis functions computes values for |S11| and |S12| which are in
very good agreement—largely within the computational errors—with the pulse
basis moment method. Deviations outwith the computational error level only occur

Fig. 5.6 Reflection coefficient magnitude (|s11|) as a function of slot length as predicted by
trigonometric basis (Lyon) and by pulse basis (Vu Khac) moment methods (WG16,
Freq = 9.375 GHz, T = 1.27 mm, x1 = 1.43 mm, w = 1.58 mm)
5.3 Scattering for Longitudinal Slot in Waveguide 113

Fig. 5.7 Transmission coefficient magnitude (|s12|) as a function of slot length as predicted by
trigonometric basis (Lyon) and by pulse basis (Vu Khac) moment methods (WG16,
Freq = 9.375 GHz, T = 1.27 mm, x1 = 1.43 mm, w = 1.58 mm)

well above the resonant slot length (L * 15.5 mm). At such slot lengths, reso-
nances of higher order than the half-cosine shape depicted in Fig. 5.4 begin to
influence the calculations and it becomes necessary to include more than three basis
functions used to generate Figs. 5.6 and 5.7. Phase computations, namely/S11 and
/S12, have also been compared [8], and yield largely similar levels of agreement.
To complete the validation exercise on the trigonometric moment method,
computed results have also been compared with measured characteristics for a
typical longitudinal slot radiator in WG16 waveguide. Typical results for S12
magnitude and phase are presented in Fig. 5.8a, b. The measurements were pro-
cured by the application of a cavity technique [8]. This involves resonating the
waveguide in a section containing the slot, which then becomes a detuning dis-
continuity, then measuring changes in resonant frequency and quality factor (Q) as
a function of slot length. While the level of agreement is generally good, and
certainly within the measurement errors, the displacement of the measured resonant
frequency to a longer slot length than predicted by theory, was initially puzzling.
But this was found to be almost wholly attributable to the fact that measurements
were performed on round-ended slots for practical fabrication reasons. The theory
assumed square-ended slots.
114 5 Moment Method Models of Compact Slot Antennas

Fig. 5.8 a Transmission coefficient magnitude for a rectangular waveguide containing a


longitudinal slot of changing length (Freq = 10.48 GHz, T = 1.32 mm, x1 = 2.00 mm,
w = 1.5 mm) b Transmission coefficient phase for a rectangular waveguide containing a
longitudinal slot of changing length (Freq = 10.48 GHz, T = 1.32 mm, x1 = 2.00 mm,
w = 1.5 mm)
5.4 T-Shaped Slot in Bifurcated Waveguide 115

5.4 T-Shaped Slot in Bifurcated Waveguide

A more complex slot example, comprising a polarisation agile, T-shaped, slot fed
from bifurcated waveguide has been included here to show that with insightful
application of boundary conditions, and with particular attention to the slot field
distribution and the basis function choice, the trigonometric moment method is of
sufficient flexibility to establish accurate modelling tools for a wide range of
compact slot geometries. The arrangement, which is depicted in Fig. 5.9, was
devised originally to secure polarisation agility by independently exciting a pair of
orthogonally aligned slots (see Chap. 1, Fig. 1.22) in dual-mode waveguide. Thus
the slot pair is arranged to radiate axial or transverse linear polarisations if excited
separately using mode switching, or if excited simultaneously and in phase
quadrature, circularly polarised radiation is produced [14]. While the longitudinal,
or shunt, slot portion of the orthogonal slot pair has been thoroughly examined in
the literature, as outlined in Sect. 5.3 above, a transverse slot straddling the central

Fig. 5.9 a Transverse T-slot radiator in dual-mode rectangular waveguide. b Transverse element
of T-slot showing coordinate system
116 5 Moment Method Models of Compact Slot Antennas

wall of a bifurcated wall is relatively novel [15] and requires advanced electro-
magnetic modelling to establish its radiation characteristics. We revisit this case
here to demonstrate the potential that exists in the moment method for addressing
unconventional applications such as those encountered in evolving novel compact
slot antenna geometries examined in ensuing chapters.
The matrix entries Ais, Bis, Cis, Dis defined in Eqs. (5.16a, 5.16b, 5.16c, 5.16d,
5.16e), have been evolved for a moment method formulation of the general case of
electromagnetic power transfer from a transmission system (waveguide) through a
slot cavity to a half-space where radiation occurs (Fig. 5.2). In the previous section
(Sect. 5.3) it has been applied comprehensively to the well established example of a
shunt slot in waveguide. We will now apply these Eqs. (5.16a, 5.16b, 5.16c, 5.16d,
5.16e) to the switchable T-slot radiator on dual mode waveguide, as depicted in
Fig. 5.9.
From a mathematical perspective the fundamental difference between the sym-
metrically located T-shaped slot in bifurcated waveguide (Fig. 5.9a) and the con-
ventional slot in regular waveguide shown in Fig. 5.1a, lies with the clearance slot
which attracts additional mutual coupling between the inner surface S of the
transverse slot and the septum slot surfaces Sv. The computation of these additional
coupling terms requires that a suitable trial magnetic field distribution can be for-
mulated for the T-slot which encompasses all three surfaces S, S’ and Sv. Once this
is accomplished the preliminary formulation of the moment method solution for the
T-slot proceeds much as outlined in Sect. 5.3.
Consider, firstly, the tangential electric field distribution which the waveguide
modal fields induce on the inner surface S of the transverse component of the T-slot
(Fig. 5.9b). The functional form of this field will obviously be influenced by the
septum slot and its depth L3. For example if L3  0 then the top of the waveguide
bifurcation will suppress any induced electric field (Eη) near the centre of the inner
surface S of the transverse slot, resulting in a field distribution not unlike that
depicted in Fig. 5.10a. That is, it is plausibly presumed that when the septum slot
approaches zero length the Eη field distribution on the inner surface S of the
transverse slot is essentially the same as that of two slots, of length L/2, positioned
end-to-end. Note that this is not necessarily true for the outer slot surface S’ par-
ticularly if the waveguide wall thickness is not infinitesimally small. On the other
hand, when the septum slot depth is finite, it becomes less easy to predict possible
field distributions on the surfaces S and Sv. However, from qualitative considera-
tions it seems evident that the clearance slot will be parasitically excited by the
radiating slot fields, and thus the Eη fields at the centre of the transverse slot will no
longer be suppressed on the inner surface S. So, it seems reasonable to suggest that
the Eη field on S will exhibit a functional form not too dissimilar to Fig. 5.10b.
With the above observations in mind we can proceed to the construction of
suitable basis functions for the fields on S and on Sv by noting the analogy between
the E-field forms in the T-slot to the voltage profiles on an equivalent T-network of
transmission lines [16]. Suppose that the transmission line depicted in Fig. 5.11a
represents a conventional slot positioned wholly within the broadwall of a regular
rectangular waveguide feed. The input impedance at the right hand extremity of the
5.4 T-Shaped Slot in Bifurcated Waveguide 117

Fig. 5.10 Suggested electric field distributions in the transverse component of the T-slot a L3 = 0,
b 0 < L3 < L/2

normalised line, for which Z0 = 1, can, from transmission line theory, be expressed
as:

Zin ¼ j tan bL ð5:17Þ

where b ¼ 2p=k and k is the line wavelength at the frequency of operation.


However, with a short located at x = L conditions must exist enforcing Zin = 0. This
in turn dictates that

tan bL ¼ 0 ð5:18Þ
118 5 Moment Method Models of Compact Slot Antennas

Fig. 5.11 Transmission line based equivalent circuits for waveguide slot radiators a conventional
slot, b T-slot in bifurcated waveguide

leading to the result that


np
b¼ ðn ¼ 1; 2; 3. . .Þ ð5:19Þ
L

The implication of this is that the line voltage (or current) forms a standing wave
with the functionality
npx
V ¼ V0 sin ð5:20Þ
L

By analogy the slot voltage ðV ¼ E  wÞ will have this same functionality, as


will the E-field in the slot. This is precisely the trigonometric basis function form
which was used successfully in Sect. 5.3.
In Fig. 5.11b we extend this idea to the T-slot of Fig. 5.9 with the aim of
constructing a plausible basis function for this more complex slot. The transmission
line model comprises two transmission lines of length L/2 representing the
5.4 T-Shaped Slot in Bifurcated Waveguide 119

transverse slot portions on either side of the central bifurcation and a line, L3 in
length, representing the clearance notch. The three lines are connected in parallel at
A, and all three are terminated in short circuits. From transmission line theory the
input impedance at A, for the paralleled L/2 lines, is given by:

j bL
ZA ¼ tan ð5:21Þ
2 2

for normalised line impedances (Z0 = 1). Hence, with ZA as the load, the impedance
at B can be expressed as:
1
tanðbL=2Þ þ j tanðbL3 Þ
ZB ¼ 2
ð5:22Þ
1  12 tanðbL=2Þ tanðbL3 Þ

But ZB must be zero for the shorted line, which implies:

tanðbL=2Þ ¼ 2 tanðbL3 Þ ð5:23Þ

This transcendental equation can be solved to give n discrete values for b, in


sympathy with the conventional case. However, given that we have already
observed in Sect. 5.3 that the s = 1 basis function on its own can procure good
results, up to the first resonance, for the conventional slot radiator, it seems sensible
to follow the same route with the T-slot. Consequently, only the first root of
Eq. (5.23) needs to be retained in the following moment method solution [15]. This
root (b1) can be determined quite straightforwardly by incorporating a regula falsi
[10] procedure within the moment method analysis. It follows immediately that
suitable basis functions on the inner surface S, which divides into two S/2 com-
ponents, of the transverse slot are:

L L
fs ¼ sin b1 ð  j xjÞ ¼ sin b1 ð  jnjÞ ð5:24Þ
2 2

and preserving continuity with the conventional slot analysis we choose to employ
testing functions (fi) for the S/2 surfaces such that ðfi ¼ fs Þ. On the outer surface of
the transverse slot (S’), where the waveguide bifurcating wall is less influential with
T > 0, the basis and testing ðfs0 ¼ fi0 Þ functions are unchanged from those used in
Sect. 5.3. That is:

spðn  L2Þ
fs0 ¼ sin ð5:25Þ
L

with s = 1 for the fundamental term.


From the transmission line model in Fig. 5.11b it is quite clear that voltage
across the two transverse arms will have the mathematical form:
120 5 Moment Method Models of Compact Slot Antennas

 
L
V ¼ sin b1  j xj ð5:26Þ
2

mirroring the trial field or basis function on S [Eq. (5.24)], while on the clearance or
septum slot surfaces Sv Fig. 5.11b suggests that here the basis function form will be:

V ¼ V 0 sin b1 ½L3  ðb  yÞ ð5:27Þ

Since these voltages (or equivalent E-fields in slot terms) must equate at the
junction A then we can write:

b1 L
sin ¼ V 0 sin b1 L3 ð5:28Þ
2

Hence, by analogy, an appropriate basis function for the field forms on the Sv
surfaces is:
sinðb21 LÞ
fsv ¼ sin b1 ðy  b þ L3 Þ ð5:29Þ
sinðb1 L3 Þ

With the trial field expressions now defined through Eqs. (5.24), (5.25) and
(5.29) for the relevant surfaces of the T-slot, the analysis can proceed in a manner
not too different from the development in Sect. 5.2. Commencing with Eqs. (5.9)
and (5.10), which are generally applicable to slot radiators, a matrix equation for the
T-slot can be formed with the following entries [see Eq. (5.15)]:
ZZ ZZ
 
Ais ¼ 2jxe0 fi ^ag  f^
n G ma ðr=r0 Þ  fs ^
an dS0 gdS
S=2 S=2
ZZ ZZ
 
 jxe0 fi ^ag  f^
n  mb ðr=r0 Þ  fs ^
G an dS0 gdS ð5:30aÞ
S
ZZ ZZ
S
 
 2jxe0 fi ^ag  f^
n  ma ðr=r0 Þ  fsv ^
G an dS0 gdS
S=2 Sv
ZZ ZZ
 
Bis ¼ jxe0 fi0 ^ag  f^0
n  G  mc ðr=r0 Þ  f 0 ^
 mb ðr=r0 Þ þ G
i an dS0 gdS ð5:30bÞ
S0 S0
ZZ ZZ
 
Cis ¼ jxe0 fi0 ^ag  f^
n  mb ðr=r0 Þ  fs ^
G an dS0 gdS ð5:30cÞ
S S0
ZZ ZZ
 
Dis ¼ jxe0 fi ^ag  f^0
n   mb ðr=r0 Þ  f 0 ^
G s an dS0 gdS ð5:30dÞ
S0 S
5.4 T-Shaped Slot in Bifurcated Waveguide 121

ZZ
hi ¼  fi ^ag  ½n
^  Hi ðrÞdS ð5:30eÞ
S

Except for the Ais term, the above integral expressions are formally identical to
the corresponding conventional slot versions. They differ only in the assumed
functional form for the basis function fi and the testing function fs. On the other hand,
Ais contains an additional term, relative to the conventional analysis, to accommo-
date the mutual coupling which can occur between the inner surface of the transverse
slot and the septum slot surfaces. The Green’s functions G  ma ; G
 mb ; and G
 mc are
essentially the same as for the conventional slot analysis, in Sect. 5.2, except that for
G ma which must be treated in two parts with in-phase dominant modes on either side
of the bifurcation. This is done by coordinate system adjustments. The previously
defined rectangular slot Green’s function G  mb now applies to the transverse com-
ponent of the T-slot. It is important to note that in the computation of Bis, Cis, and Dis
higher order slot cavity modes will require to be included, more so than in the
conventional slot solution, to accommodate basis functions such as fs in Eq. (5.24)
which will tend to generate such modes. The integral expressions delineated in
Eqs. (5.16a, 5.16b, 5.16c, 5.16d, 5.16e) have been fully evaluated in references [8]
and [9] and some typical results have been presented in Sect. 5.3. Needless to say the
evaluation of Eqs. (5.30a, 5.30b, 5.30c, 5.30d, 5.30e) are not too different and are
detailed in [8]. It should be noted here that the moment method computations have
generally been performed for the s = 1 basis function on the inner slot surface S, and
for the first basis function on the outer surface S’ except where Lv is short, in which
case the s = 3 basis function is introduced.
Below we present a selection of representative results to demonstrate the
effectiveness of the moment method in modelling more complex and compact
antenna designs. As with the longitudinal slot example in Sect. 5.2, once the matrix
equation, with the entries detailed in Eq. (5.30a, 5.30b, 5.30c, 5.30d, 5.30e), has
been solved by moment method interaction, plausible values for s; as ; fs ; fs0 ; and fv
should become available. In a process little different to that adopted for calculating
the longitudinal slot scattering values, those for the transverse T-slot can be simi-
larly calculated. The scattering formulae are detailed in Appendix C.

5.4.1 Computed and Measured Results

The proof that a new engineering analysis, or that an analysis of a novel engineering
advance, is mathematically and technically sound lies either by showing compar-
ison with the results of other researchers, or by comparing the theory with mea-
surements on carefully prepared models, as outlined in Sect. 5.3. Here we follow
this measurement route, since at the time when the T-slot in bifurcated waveguide
was proposed [15], no earlier results by other investigators were available.
122 5 Moment Method Models of Compact Slot Antennas

Measurements were performed on a dual-mode waveguide fabricated from two


lengths of standard WG16 (22.86 mm  10.16 mm) brazed together along care-
fully machined sidewalls such that the common wall, or bifurcation, was 1.5 mm in
thickness. The T-slot radiators were formed by machining dual-mode waveguide
samples. Slots ranging in length from a minimum case with L = 10 mm and
L3 = 0.0 mm, and increased incrementally in approximately 1 mm length steps up
to slots with L = 20 mm and L3 = 5.5 mm.
Normally, slot discontinuities in waveguide can be measured and characterised
using network analysis procedures. These can be very accurate if ‘de-embedding’
[17], of the unavoidable coaxial line to waveguide junctions, is undertaken care-
fully. However, for the dual-mode waveguide feed (Fig. 5.9) in addition to any
coax-to-waveguide transitions which may be necessary, transitions from conven-
tional waveguide to dual-mode waveguide are also required, resulting in an overall
transition arrangement which is too difficult to de-embed. This was certainly true
when the T-slot radiator was under development. A solution which was found to be
accurate for below resonance slots involved resonating the device under test,
including its conventional-to-dual waveguide transition, and performing measure-
ments of the resonant frequency and the quality factor (Q) of the system as a whole
[17, 18]. The arrangement is depicted schematically in Fig. 5.12.
In the figure the device under test (DUT)—namely the dual-mode waveguide
feed supporting a T-slot radiator—was sandwiched between a pair of folded
‘magic-tee’ waveguide couplers. These components provide wideband even-mode
(direct port) and odd-mode (orthogonal port) excitation of the bifurcated waveg-
uide. Only the even-mode port was used for T-slot measurements. The odd-mode
port can be, and has been, used for the characterisation of parallel shunt slot
radiators, such as those illustrated in Fig. 1.22. At the even-mode port of each of the
‘magic tees’ a precision machined iris was attached thus forming a waveguide
resonator. The irises were designed to be as identical as possible and each was
accurately calibrated for its transmission coefficient as a function of frequency. The
system was driven by a crystal controlled microwave source through a calibrated
precision attenuator and a precision multi-hole waveguide sampling coupler (10 dB
coupling). A similar coupler was also introduced into the output port of the system
with the transmitted power subsequently dissipated in a well matched waveguide
load. The signals emanating from the sampling ports of the two couplers were fed
via diode detectors to a standard microwave display, which was controlled from the
source. The measurement of cavity resonant frequencies and the cavity Q’s for a
broad range of T-slot geometries is relatively routine and has been implemented.
Furthermore, the results have been converted into reflection coefficient (magnitude
and phase) and transmission coefficient (magnitude and phase) by standard pro-
cedures [19]. The results are summarised in Figs. 5.13, 5.14, 5.15 and 5.16.
In Fig. 5.13 the power radiated by a 16 mm long T-slot with a 5.48 mm
clearance slot is plotted as a function of frequency for both the moment method
simulation and measured results. Equation (C.25) has been used to convert the
reflection and transmission results into power radiated. The figure demonstrates,
5.4 T-Shaped Slot in Bifurcated Waveguide 123

Fig. 5.12 Resonant cavity set-up for T-slot scattering in dual mode waveguide

rather as expected from perusal of the conventional slot case, that radiation is weak
at low frequencies which are well below the T-slot resonance, and then grows
steadily in strength as the source frequency approaches this resonance—predicted
124 5 Moment Method Models of Compact Slot Antennas

Fig. 5.13 Relative power radiated (Prad/Pinc) versus frequency for a transverse T-slot in bifurcated
waveguide in even-mode (WG16 wall thickness T = 1.32 mm; slot width w = 1.5 mm; slot length
L = 16 mm; septum slot L3 = 5.48 mm)

Fig. 5.14 T-slot transmission coefficient magnitude a and phase b as a function of slot length L
(Even mode excitation; WG16; T = 1.32 mm; L3 = 5.48 mm; bifurcation thickness = 1.5 mm;
frequency = 9.89 GHz)

by the moment method to be at 11.1 GHz for the given dimensions. It is noticeable
that agreement between theory and measurement is good, and well within the
estimates of experimental error at frequencies below *10.5 GHz where the T-slot
is distinctly non-resonant. Close to the slot resonance, around 11 GHz, the mea-
sured results begin to deviate significantly from the moment method predictions.
This can be attributed to a basic limitation in the cavity based measurement tech-
nique. For weakly radiating non-resonant slots the amount of power extracted from
5.4 T-Shaped Slot in Bifurcated Waveguide 125

Fig. 5.15 T-slot reflection coefficient magnitude a and phase b as a function of slot length L
(Even mode excitation; WG16; T = 1.32 mm; L3 = 5.48 mm; bifurcation thickness = 1.5 mm;
frequency = 9.89 GHz)

Fig. 5.16 Relative power radiated in dB as a function of clearance slot depth (L3) for a
symmetrically located T-slot—even mode excitation (WG16; T = 1.32 mm; L3 = 5.48 mm;
bifurcation thickness = 1.5 mm; frequency = 9.89 GHz)

the cavity is insufficient to ‘dampen’ the fundamental dual-mode waveguide reso-


nance, which means that sharp, peaky, Q-curves with well defined, easily measured,
characteristics continue to be observed. However, as the drive frequency approa-
ches the T-slot resonant value, and the slot begins to radiate strongly, the dual-mode
waveguide cavity becomes increasingly heavily loaded, thus ‘flattening’ the
Q-curves. This then makes it very difficult to establish the precise cavity resonant
frequency and to determine a value for the very low Q-level. The measurement
error becomes large, as is reflected in the deviation between theory and experiment
in the vicinity of the slot resonant frequency.
126 5 Moment Method Models of Compact Slot Antennas

The above comments are reinforced in Figs. 5.14 and 5.15, which present,
respectively, plots of transmission coefficient amplitude and phase and reflection
coefficient amplitude and phase as functions of slot length. The plots are typical of a
series resonant slot in waveguide. While the computed transmission coefficient
magnitude falls from approximately unity to almost 0.94 at resonance, the reflection
coefficient magnitude mirrors this, rising from almost zero for non-resonant lengths
to almost 0.06 at resonance. The phase shift predictions establish that the slot, in
transmission line terms, could be modelled by an RLC circuit in series with the line,
with L  C for short slots. At resonance L = C and the slot behaves as a series
resistance in the line.
The comparisons between theory and measurement provided in Figs. 5.14 and
5.15 fully confirm the observation made with respect to Fig. 5.13 with ‘good’ levels
of agreement for non-resonant slots (within experimental error). It is perhaps per-
tinent to note that while the measured points in Figs. 5.14a and 5.15a offer poor
confirmation of the predicted resonant behaviour of the slot, this is less true of the
phase measurements. These indicate an experimental resonant frequency of *18,
0.5 GHz below the theoretical value, and well within the measurement error
estimate.
Arguably, a feature of primary importance for potential users of the T-slot
radiator is the degree to which its radiation level is influenced by the clearance slot
length (L3). A typical result, depicting relative power radiated as a function of
septum slot length, is presented in Fig. 5.16. Not unexpectedly for L3 approaching
zero length the radiation is very weak at about −36 dB. It is then predicted to rise
rapidly as L3 is increased, to almost −20 dB when L3 = 7 mm. The measured results
confirm the theoretical prediction to within the experimental error. The experimental
errors arise from several sources one of which is cavity damping for near resonant
slots. Others are attributed to the square ended slots used in the theory. These were
difficult to replicate experimentally, and the slot used were generally round-ended.
Roughness of the machined surfaces can have a deleterious effect on measurements
as can imprecise positioning of the slots within the dual-mode waveguide test piece.
Generally the degree of agreement between theory and experiment is sufficient to
assert that the moment method is correctly simulating the T-slot radiator located in
dual-mode rectangular waveguide. Needless to say, today, we can establish the
operational integrity of any new computational analyses based on the moment
method by performing a ‘one-off’ simulation of the geometry of interest using a
modern commercially available finite element package such as Concerto [20], and
comparing results. This has been done in Fig. 5.17 for an ‘extreme’ T-slot radiator
with a very short radiating element (L = 2 mm) and a relatively long clearance
notch (L3 = 13, 14 and 15 mm) in a deep dual-mode waveguide structure. The plots
show strong resonant behaviour with the resonant frequency shifting to lower
values as the overall slot length increases—in accordance with Eq. (5.23). More
importantly the degree of agreement between the moment method (solid lines) and
the finite element technique (dashed lines) is excellent and well within modelling
errors, particularly given the limited number of basis functions adopted in the
moment method.
5.5 Chapter Summary 127

Fig. 5.17 Moment method solution (solid lines) for T-slot compared with finite element software
(dashed lines) for deep dual-mode waveguide (L = 2 mm, L3 = 13, 14, 15 mm, w = 1.5 mm,
t = 1.27 mm)

5.5 Chapter Summary

Computer models for simulating electromagnetic scattering by elemental radiators


have become key components for modern antenna design techniques, and in this
chapter we have aimed at establishing the important role of the moment method in
this trend. To do this two illustrative examples are explored in detail. While the
examples focus on waveguide fed slots, it is important to recognise that the ana-
lytical procedures described are applicable more widely to other compact radiators
such as dipoles and patches, driven from other transmission line forms, such as
stripline and microstrip. Any limitations which may exist revolves around the
availability of appropriate Green’s functions.
Firstly, a conventional well established example of electromagnetic wave scat-
tering by an inclined slot in the broadwall of waveguide is investigated. This entails
the elaboration of a full solution for the conventional ‘shunt’ slot in rectangular
waveguide. The relevant Green’s functions for the coupled regions created by this
slot geometry are formed. They comprise the Green’s functions for rectangular
waveguide, a half-space and a cavity. Once these are established it becomes pos-
sible to construct the integral equations describing the boundary value problem.
Subsequently, the essence of the solution of these equations is the thoughtful choice
of efficient trial functions or basis functions for the unknown aperture field. This is
illustrated by describing and detailing the selection and formulation of a trigono-
metric option. It is shown that the trigonometric choice optimises calculation speed
without sacrificing the accuracy of the scattering predictions for a range of shunt
slot geometries. Secondly, the T-slot radiator in bifurcated waveguide is also
128 5 Moment Method Models of Compact Slot Antennas

subjected to the MoM to emphasise the flexibility of the technique. Without sig-
nificant amendments to the computational steps pursued in modelling the con-
ventional shunt slot in waveguide, the key to the evolution of a MoM for this quite
different more complex scattering geometry is the careful construction of an
appropriate basis function.
It is pertinent to note here, that the geometries of many compact radiators, which
are amenable to solution by the method of moments, are undoubtedly solvable
using finite element methods (FEM) such as HFSS and COMSOL. However, while
these solvers permit model construction with minimal knowledge of electromag-
netic theory, essentially by exchanging user modelling choices for computational
intensity, they exhibit the significant downside of being relatively slow computa-
tionally in reaching a solution. As has already been intimated this rules out the use
of FEM in embedded modules within array simulation packages.
Finally, it is arguably true to say that accurate computational design of aperture
antennas, and in particular array antennas, made possible using the method of
moments, has virtually obviated the traditional need for laborious back-up mea-
surement procedures. In the not too distant past, 25–30 years ago, it was necessary
to provide precise design data on waveguide slots, ranging over a multiplicity of
sizes and orientations, in order to facilitate array synthesis calculations. At that time
measurements were the only source of such data.

References

1. A.J. Sangster, A.H.I. McCormick, Theoretical design/synthesis of slotted waveguide arrays.


Proc. IEE 136H, 39–46 (1989)
2. T. Khan, A. De, M. Uddin, Prediction of slot size and inserted air gap for improving the
performance of rectangular microstrip antennas using artificial neural networks. IEEE
Antennas Wirel. Propag. Lett. 12(1), 1367–1371 (2013)
3. A.A. Oliner, The impedance properties of narrow radiating slots in the broad face of
rectangular waveguide. IRE Trans. AP-5, 4–20 (1957)
4. B.N. Das, G.S. Sanyal, Network parameters of a waveguide broad wall slot radiator. Proc.
IEE 117(1), 41–44 (1970)
5. T. Vu Khac, C.T. Carson, Coupling by slots in rectangular waveguide with arbitrary wall
thickness. Electron. Lett. 8, 456–458 (1972)
6. R.F. Harrington, Field Computation by Moment Methods (MacMillan, New York, 1968)
7. R.F. Harrington, Matrix methods for field problems. Proc. IEEE 55(2), 136–149 (1967)
8. R.W. Lyon, An investigation of apertures in dual mode rectangular waveguide for variable
polarisation arrays, Ph.D. thesis, Heriot-Watt University, 1979
9. R.W. Lyon, A.J. Sangster, Efficient moment method analysis of radiating slots in thick walled
rectangular waveguide. IEE Proc. 128(4), 197–205 (1981)
10. M.G. Salvadori, M.L. Baron, Numerical Methods in Engineering (Prentice-Hall International,
London, 1961)
11. T.J. Akai, Numerical Methods (Wiley, Canada, 1993)
12. R.S. Esfandiari, Numerical Methods for Engineers and Scientists using MATLAB (C.R.C.
Press, Boca Raton, USA, 2013)
13. T. Vu Khac, Solution of some waveguide discontinuities by the method of moments. IEEE
Trans. MTT-20, 416–417 (1972)
References 129

14. J.S. Ajioka, D.M. Tang, N.W. Wong, Arbitrarily polarised slot radiators in bifurcated
waveguide. IEEE Trans. AP-22, 196–200 (1974)
15. A.J. Sangster, R.W. Lyon, Moment method analysis of a T-shaped slot radiator in bifurcated
waveguide. IEE Proc. 129(6), 299–306 (1982)
16. R.J. Chignell, J. Roberts, Compact resonant slot for waveguide arrays. IEE Proc. 125(11),
1213–1216 (1978)
17. A.E. Bailey, Microwave Measurements (Peter Peregrinus, 1989)
18. N.F. Wong, R. Tang, G.A. Okubo, Investigation of Characteristics and Practical
Implementation of Arbitrary Polarised Radiators (Hughes Aircraft Co., Ground Systems
Group, Fullerton, California, 1971)
19. D.M. Pozar, Microwave Engineering, 4th edn. (Wiley, 2011)
20. CONCERTO, Vector Fields Ltd. http://metamaterials-modelling.net/software-concerto.php
Chapter 6
Resonant and Non-resonant Linear
Slot Arrays

6.1 Introduction

At microwave frequencies, array antennas generally comprise a linear or planar


arrangement of periodically spaced aperture antennas. These may be of the slot
type, open-ended waveguide type, or horn type, as described in Chap. 3. To sim-
ulate the electromagnetic wave nature of such periodic distributions of radiators it is
usual, firstly, to model an equivalent periodic array of isotropic radiators, and
secondly, to employ the principle of pattern multiplication to determine the radi-
ation pattern of the original array. As the name implies, this principle suggests that
if the far-field pattern of the isotropic array is multiplied with that of an individual,
or typical, aperture antenna which forms an element of the actual array, the resultant
pattern will be that of the original array. However, care must be exercised in
following this procedure. Unless the element pattern has been obtained in the array
environment this process ignores element-to-element interaction (mutual coupling)
and the resultant pattern becomes an approximation to the required pattern.
Nevertheless, the technique is a powerful one and is widely used to assess the
properties of potential array designs.
Array antennas will also be examined using the now conventional z-transform
method, since it gives access to the study of elementary array synthesis procedures.
Although the z-transform method again neglects the effects of mutual coupling
between closely spaced neighbouring radiating elements of the target array, for
many compact microwave antennas of the array type, the method remains tolerably
effective because mutual coupling is often too small to cause significant deviation
from the modelled result.

© Springer Nature Switzerland AG 2019 131


A. J. Sangster, Compact Slot Array Antennas for Wireless Communications, Signals
and Communication Technology, https://doi.org/10.1007/978-3-030-01753-8_6
132 6 Resonant and Non-resonant Linear Slot Arrays

6.2 Uniform Linear Array of Isotropic Elements

A linear array is one that is formed from a single line of multiple radiating sources.
When these individual sources radiate equally in all directions the array is said to
comprise isotropic elements. In addition the array is described as uniform when the
elements are equally spaced and fed with equal signal levels. Such an array is
depicted in Fig. 6.1. Although planar array (an array of linear arrays) antennas are
more commonly used in mobile communications systems, it is advantageous to
study the radiation patterns of linear arrays which are mathematically less complex.
Nevertheless, the knowledge gained from linear array studies remains entirely
relevant to, and representative of, planar arrays.
At a distant point from the antenna the total E-field, at radius r due to n elements
can, by superposition, be expressed as:

ET ¼ E0 ð1 þ ejw þ e2jw þ    þ ejðn1Þw Þ ð6:1Þ

where E0 is the field magnitude at the distant point (r) due to a single element of the
array, and very importantly:

w ¼ k0 d sin / þ a ð6:2Þ

Here d is the element-to-element spacing, k0 is the free-space phase coefficient, /


is the real angle from broadside, and a is the progressive element-to-element phase
shift impressed on the array by the feed structure. Note that the factor (expðjk
r
0 rÞ
),
which is common in all radiation field representations, has been suppressed for
convenience. If we normalise Eq. (6.1) to the optimum field E0 we obtain:

Fig. 6.1 Geometrical relationships for linear array with element spacing d
6.2 Uniform Linear Array of Isotropic Elements 133

ET
¼ 1 þ ejw þ e2jw þ    þ ejðn1Þw ð6:3Þ
E0

By replacing the exponential terms by the variable z (= ejw) the far-field pattern
of the linear array can be expressed mathematically as a polynomial, as follows:

1ðzÞ ¼ 1 þ z þ z2 þ z3 þ    þ zn1 ð6:4Þ

This polynomial form has been shown [1] to be a powerful mathematical tool in
the analysis and synthesis of array antennas. However, before advancing to the
treatment of arrays in general, it is perhaps instructive to consider some special
cases.
When f(z) is multiplied by z in Eq. (6.4) we obtain:

z1ðzÞ ¼ z þ z2 þ z3 þ    þ zn ð6:5Þ

Hence subtracting Eq. (6.4) from Eq. (6.5) gives:

ðz  1Þ1ðzÞ ¼ zn  1

giving

zn  1
1ðzÞ ¼ ð6:6Þ
z1

With a little manipulation, Eq. (6.6) can be written in the form:

zn=2 ðzn=2  zn=2 Þ


1ðzÞ ¼ ð6:7Þ
z1=2 ðz1=2  z1=2 Þ

If this equation is recast in its exponential form, with z = ejw, while recognising
the identity:

ejh  ejh
¼ sin h
2

then we discover that:

E0 sin nw2
ET ¼ ð6:8Þ
sin w2

Element phase is referred to the array mid-point to secure this simple form.
Functionally, Eq. (6.8) is not too dissimilar to Eq. (3.23), the spectrum function for
a uniformly illuminated linear aperture. The primary difference is the oscillatory
nature of the denominator of Eq. (6.8) which strongly influences the pattern
134 6 Resonant and Non-resonant Linear Slot Arrays

distribution. The result is that the linear array pattern and the linear aperture pattern
are comparable only for a limited range of w values, which implies a limited range
of element spacings d, as illustrated in Sect. 6.2.1.

6.2.1 Radiation Patterns

Equation (6.8) represents the functional form of the far-field radiation pattern, or
polar diagram, of the uniformly excited linear array of isotropic elements. Since the
radiation in the forward half-plane −p/2 < / < p/2 is clearly identical to the
radiation into the reverse half-plane p/2 < / < 3p/2, we need consider only the
former in the following discussion.
The ratio of sines in Eq. (6.8) implies that the pattern must be multi-lobed. The
primary or main lobe occurs, where simultaneously w ! 0 and sin ðnw=2Þ !
nw=2. For this lobe ET =E0 ¼ n; that is at the peak of the primary lobe the E-field at a
given radius for the array is n times the E-field at the same radius for a single
isotropic element. However, note that if the element spacing d is equal to or greater
than ko there may be more than one principal maximum—these secondary maxima
are termed grating lobes and appear where w = p, 2p …. It is apparent from
Eq. (6.2) that the principal maximum at w = 0 occurs where:
a
sin / ¼  ð6:9Þ
k0 d

When this principal maximum appears in a direction normal to the line of the
array, i.e. at / = 0°, the array is termed a broadside array (see Fig. 6.2). Clearly, for
this case a = 0°, which means that the element-to-element phase shift is set to zero
degrees. In the diagram, /1 denotes the angle to the first null measured from the
array normal, while d1 provides essentially the same information, but measured
from the array base line - d1 ¼ 90  /1 .
When / = 90° the principal maximum lies along the line of the array (Fig. 6.3)
and the array, in this case, is termed end-fire. The angle from the lobe maximum to
the first null is denoted by d2 in the figure. From Eq. (6.9) it is clear that for an
end-fire array we require a = −kod.

6.2.2 Broadside Array

For a broadside array of equally space and equally excited radiators (uniform array)
the nulls of the radiation pattern occur where the sine ratio in Eq. (6.8) exhibits zero
values, i.e. where:
6.2 Uniform Linear Array of Isotropic Elements 135

Fig. 6.2 Principal lobe—broadside radiation

Fig. 6.3 Principal lobe—


end-fire array

nw
¼ mp ð6:10Þ
2

Hence substituting for w using Eq. (6.2) and noting that for a broadside array
a = 0,

nk0 d sin /
¼ mp ð6:11Þ
2
136 6 Resonant and Non-resonant Linear Slot Arrays

which implies that:

2mp
sin / ¼  ð6:12Þ
nk0 d

Therefore, the nulls in the pattern appear at angles given by:


 
mk0
1
/m ¼ sin ð6:13Þ
nd

If the array is large, such that nd  mk0 then /m is small for the first few nulls
nearest to the main lobe, and is given by:

mk0 m
/m ¼  ð6:14Þ
nd L=k0

In this relationship LðndÞ is the total length of the array. Strictly, L = (n − 1)


d but the error is small for long arrays. The first nulls which define the principal
lobe occur where m = 1, and for this broadside case the beamwidth between first
nulls (BWFN) is given by:

2 114:6
BWFN ¼ 2/1  radians ¼ ð6:15Þ
L=k0 L=k0

For reasons intimated earlier it is much more usual for antenna engineers to
employ half-power beamwidth (HPBW) in defining the radiated main beam and for
the broadside uniform array it is not difficult to show that

HPBW ¼ 0:44  BWFN ð6:16Þ

So, for this array:

2  0:44 50:4
HPBW  radians ¼ ð6:17Þ
L=k0 L=k0

For this uniform broadside array of isotropic elements with d < ko/2 the radiated
power density in the principal lobe can be approximated by:

PT
parray  W=m2 ð6:18Þ
2pr 2 h3dB

where PT denotes power transmitted and r is a representative far-field distance from


the array. For a single isotropic radiator delivering the same power PT, the power
density at r is:
6.2 Uniform Linear Array of Isotropic Elements 137

PT
piso ¼ W=m2 ð6:19Þ
4pr 2

Hence the directivity (D) of the array has the simple form:

parray 2
D¼  ð6:20Þ
piso h3dB

In the initial process of assessing array requirements in any given application,


simple formulae such as Eqs. (6.17) and (6.20) are very useful for acquiring a quick
appreciation of the trade off between the beamwidth desired and antenna size.

6.2.3 End-Fire Array

The only difference between a uniform end-fire array and a uniform broadside array
is in the value of the progressive phase shift a, so the form of Eq. (6.8) is unaltered
by the change. Consequently, the condition for pattern nulls remains as in
Eq. (6.10). However, in the end-fire case, with a = −k0d, the equation relating w
and / becomes:

w ¼ k0 d sin /  k0 d ð6:21Þ

This can be written in a more convenient form if we note that d = 90° − /,


namely

w ¼ k0 d cos d  k0 d ð6:22Þ

Hence combining Eq. (6.22) with Eq. (6.10) we get:


n
k0 dðcos d  1Þ ¼ mp
2

or

2mp mk0
cos d  1 ¼  ¼ ð6:23Þ
nk0 d nd

Thus in terms of d the null angles are located at:


rffiffiffiffiffiffiffiffi!
dm mk0
¼ sin1  ð6:24Þ
2 2nd
138 6 Resonant and Non-resonant Linear Slot Arrays

Once again, directing our attention to a long array with nd  mko , the nearest
pattern nulls occur at angles given by:
rffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffiffi
dm mk0 m
 ¼
2 2nd 2L=k0

i.e.
sffiffiffiffiffiffiffiffiffiffi
2m
dm  ð6:23Þ
L=k0

So, finally we obtain the useful relations:


sffiffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffi
2 2
BWFN ¼ 2 radians ¼ 114:6 degrees ð6:24Þ
L=k0 L=k0

For the end-fire array it is not difficult to show that HPBW = 0.66  BWFN,
and therefore we can write:
sffiffiffiffiffiffiffiffiffiffi
2
HPBW  76:3 degrees ð6:25Þ
L=k0

Note that the half-power beamwidth for the end-fire array is very much larger
than for the corresponding broadside array. For example, when the ratio L/ko = 50
the end-fire array beamwidth is approximately 15 times greater than that of the
equivalent broadside format. Some improvements in end-fire gain can be achieved
without increasing the number of radiating elements by judicious adjustment of the
element spacings d. The optimum spacing is termed the Hansen-Woodyard con-
dition, which will be examined below.

6.2.4 Scanned Array

The above examination and discussion of broadside and end-fire arrays underlines a
major property of array antennas, namely their ability to provide electronic beam
scanning. The two cases represent the scanning extremes of zero degrees (broad-
side) and ninety degrees (end-fire) when the progressive phase shift a is set to,
respectively, zero and −kod. At the / = 0° (d = 90°) scan condition the principal
beam exhibits minimum beamwidth and optimum directivity. As the beam is
scanned away from broadside towards the end-fire condition both the beamwidth
and the directivity deteriorate, slowly at first, and then increasingly rapidly as
end-fire is approached. This behaviour is illustrated graphically in Fig. 6.4 which
6.2 Uniform Linear Array of Isotropic Elements 139

Fig. 6.4 Beamwidth variation with scanning (adapted from [2])

presents 3 dB beamwidth as a function of array length for a uniform linear array.


These beamwidths were generated by forming the following relationship:

sin nw2
¼ 0:707 ð6:26Þ
sin w2

for uniform linear arrays with d = ko/2. The curves in Fig. 6.4 represent varying
scan angles (/) and the scan limit defines the angle to which the principal beam can
be scanned unmodified by the proximity of end-fire. For example, a very long array
can be scanned much closer to end-fire than a short one. This figure makes it very
clear that array length is the major determinant of the principal lobe beamwidth and
hence of the array gain.
It is pertinent to ask here whether or not, for any given number of elements n,
there is an element spacing d and an array length L = (n − 1)d which procures
optimum array gain and minimum beamwidth. For a broadside array calculation
based on Eq. (6.8) it can be demonstrated that minimum beamwidth and maximum
gain (allowing for power losses in sidelobes and grating lobes) occur where the
element spacing is just less than one wavelength. This is shown clearly in Fig. 6.5
where gain is presented as a function of d/ko for a range of array options with
values of n ranging from 2 to 24. The pronounced dip in the value of the gain at
140 6 Resonant and Non-resonant Linear Slot Arrays

Fig. 6.5 Gain as a function of d/k0 for a uniform broadside array (adapted from [3])

d=k0  0:95 particularly where n is large, is produced by the appearance of the first
pair of grating lobes in the end fire directions.

6.3 Array Design Using Theory of Polynomials

In this short section the intention is to introduce the reader to antenna synthesis,
rather than analysis, where we have been directing our attention up to this juncture.
The method which will be outlined applies strictly to uniform arrays. Nevertheless,
the procedures developed here provide a good general introduction to antenna
synthesis concepts without embarking on an overly complex mathematical journey.
Actually, in practice, antenna design engineers generally use very sophisticated
commercial software to perform the complex operations involved in synthesis, but
to do this efficiently some awareness of the basic process is essential.
It has already been shown (Eq. 6.4) that the far-field pattern of an n-element
uniform linear array can be expressed in the polynomial form:
6.3 Array Design Using Theory of Polynomials 141

1ðzÞ ¼ 1 þ z þ z2 þ z3 þ    þ zn1

where z ¼ ejw and w is the difference in phase of waves radiated from any pair of
adjacent elements. By a fundamental theory of algebra a polynomial of order
(n − 1) has (n − 1) zeros (some of which may be multiple zeros) and can be
factored into (n − 1) binomials [1]. Thus

1ðzÞ ¼ ðz  r1 Þðz  r2 Þðz  r3 Þ. . .ðz  rn1 Þ ð6:27Þ

The quantities r1, r2, …, rn−1 are termed the roots of the (n − 1)th order poly-
nomial. Since z ¼ ejw has magnitude unity, these roots can be represented graph-
ically as located on a unit circle in the complex z-plane. The magnitude of f
(essentially the far-field pattern) can then be determined by multiplication of the
vector magnitudes (z − r) as illustrated in Fig. 6.6, which shows z ¼ ejw with w
measured clockwise from the real axis denoted by Re. The imaginary axis is
denoted by Im. Complex vector r1 ¼ ejw1 is also represented on the diagram, and
the vector (z − r1) is just the vector difference between z and r1.
For arrays formed from isotropic elements, radiation will be symmetrically
distributed around the axis of the array, and consequently only radiation angles / in
the range −p/2 to p/2 need be considered. From Eq. (6.2), this means that w, the
argument of z, increases from −kod + a to kod + a as / varies from −p/2 to p/2.
Thus the range of w is 2kod. For example, for a broadside array for which a = 0, if
the element spacing is set at half the free-space wavelength (d = k0/2) then w varies
from −p to +p, as shown in Fig. 6.7a. Note that z is presumed to travel in the

Fig. 6.6 Unit circle diagram


142 6 Resonant and Non-resonant Linear Slot Arrays

Fig. 6.7 Range of w for broadside array (a = 0)

clockwise direction on the unit circle. For d = ko/4 (Fig. 6.7b) the range of w is
from −p/2 to +p/2, while the range is from +p/2 to −p/2 when d = 3ko/4, with
z traversing the unit circle 1.5 times (Fig. 6.7c). Given that mainbeams occur when
w = 0°, 360°, …, 2 mp, arrays with large element spacings ðd k0 Þ generate a
grating lobe each time the locus of z passes zero on the circle diagram.
For a uniform linear array antenna the roots of its polynomial representation
(Eq. 6.27) are not restricted as to where on the unit circle they lie, and embedded in
this observation resides the basis of array synthesis. The roots rn can be located at
any angle w except w = 0 where the primary beam is always located. For the
uniform array case the roots, plus the pole at the w = 0 point (black dot), divide the
circle into n equal arcs. This is illustrated in Fig. 6.8a for a four element array.
The far-field pattern magnitude is obtained by forming the product of the vectors
(z − ri). This can be done graphically by multiplying the lengths of the blue vectors
in Fig. 6.8b for all positions of z (red dot) as it moves around the circle. It is not
difficult to observe that the principal maximum occurs at w = 0 where all three
vectors exhibit their maximum length. Subsidiary, less pronounced maxima, also
occur mid-way between the zeros, where the pattern nulls occur. These features can
be located in ‘real space’ (denoted by /) by replacing w with /, using Eq. (6.2) For
example, the nulls of the four element array occur at the w values −p/2, −p, −3p/2,
or more generally at wm = −2 mp/n with m = 1, 2, 3. Therefore the nulls in the
‘real space’ radiation pattern occur at angles /m given by:
6.3 Array Design Using Theory of Polynomials 143

Fig. 6.8 Location of roots on circle diagram for four element uniform array

a 2mp
sin /m ¼   ð6:28Þ
k0 d nk0 d

The ‘real space’ far-field pattern of the four element array can now be con-
structed either graphically, or by plotting the function f(z), which has the form:

fðzÞ ¼ ðz  ejp=2 Þðz  ejp Þðz  ej3p=2 Þ


ð6:29Þ
¼ z3 þ z2 þ z þ 1

For a = 0 and d = k0/2 an approximation to the resultant pattern as a function of


/ is shown in Fig. 6.9. This technique for array pattern synthesis is generally
attributed to Schelkunoff [1]. The range of w, in this case, is 2p.
For an end-fire array with four elements the root diagram remains unchanged,
but with a = −kod, this repositions the nulls in ‘real space’. If in addition we reduce
d to k0/4 the range of w ranges from w = 0 to w = p around the lower half of the
circle diagram as shown in Fig. 6.10. The root at 3p/2 is now no longer contributing
to the shape of the pattern in real space.
The resultant pattern for this case is shown in Fig. 6.12 (curve A). It has a very
wide primary lobe and high side lobes, because of the non-contributing zero.
However if we move all of the roots into the range of w as suggested in Fig. 6.11, a
much improved pattern emerges for the same number of elements. This arrange-
ment of roots on the circle diagram is generated by a polynomial of the form:

1ðzÞ ¼ ðz  ejp=3 Þðz  ej2p=3 Þðz  ejp Þ


ð6:30Þ
¼ 1 þ 2zejp=3 þ 2z2 ej2p=3 þ z3 ejp
144 6 Resonant and Non-resonant Linear Slot Arrays

Fig. 6.9 Far-field pattern for


a four element uniform
broadside array with d = k0/2

By making the relation z1 ¼ zejp=3 Eq. (6.30) can be written in the more
convenient form:

1ðz1 Þ ¼ 1 þ 2z1 þ 2z21 þ z31 ð6:31Þ

This represents an array having radiating elements distributed in amplitude as 1:


2: 2: 1 and with a progressive phase shift of

p p 5p
a¼  ¼ ð6:32Þ
2 3 6

This modified end-fire array pattern is sketched in Fig. 6.12 (curve B). Note the
improvement in directivity.
If the overall length of the array is fixed but the number of elements it contains is
allowed to increase, it becomes possible to improve the directivity still further if the
nulls are optimally spaced within the range of w on the circle diagram. Curve C in
Fig. 6.12 shows the pattern that results when the number of elements is increased to
6.3 Array Design Using Theory of Polynomials 145

Fig. 6.10 Root location for four element end-fire array

Fig. 6.11 Root locations for four element end-fire array with improved far field pattern
146 6 Resonant and Non-resonant Linear Slot Arrays

Fig. 6.12 End-fire array pattern improvements by root adjustments

seven by reducing the inter-element spacing to one eighth of a wavelength, so that


the overall array length is still 3ko/4. To obtain this result the nulls are equispaced in
the range w = 2kod = p/2. Curve D demonstrates the pattern which results when the
number of elements is increased even further (to 13) by reducing the inter-element
spacing to ko/16. Again the nulls are equispaced in the range of w. However, these
improvements in directivity as will be shown later tends to be achieved at the
expense of other important factors of antenna performance—a typical engineering
‘trade-off’.
For the uniform array, as we have seen, the maximum directivity and gain are
directly related to the array length. In contrast, the above null adjustment proce-
dures, which impinge on the element excitement levels and phase relations, appear
to offer the possibility of arbitrarily improving directivity with an array of fixed
length by simply using a sufficiently large number of elements. However, as always
in technology, there is a balancing consequence. Closely spaced elements, phased
and level controlled to achieve high directivity, present very low impedances to the
input feed lines. This means very high currents in the feed lines—or in microstrip
terms the feed lines become unfeasibly wide. In other words, with ‘real’ antennas
the practical requirements of efficient delivery of power to each element imposes a
limit to the directivity improvement which can be extracted from a fixed length
array.
6.3 Array Design Using Theory of Polynomials 147

6.3.1 Optimum Element Spacing

Having discovered that the location of the roots, on the circle diagram for the array
polynomial, has a major influence on the radiated pattern, it becomes possible to use
this graphical technique to determine the optimum element spacing for a uniform
array. For the broadside case the optimum pattern results from including the
maximum number of nulls in the range of w. Therefore, for a large number of
elements n, the maximum number of nulls are active when the range of w is 4p, as
shown in Fig. 6.13a. Since the range of w for a broadside array is 2kod this implies
that 4pd/ko = 4p and hence the optimum element spacing is d = ko.
In the case of the end-fire array the maximum number of nulls become active
when the range of w approaches 2p, as suggested in Fig. 6.13b, which means that in
this case the optimum element spacing d is almost ko/2. Note that in the limit where
d = ko/2 there are two opposed end-fire mainbeams, and the array is bidirectional.
Also, for the end fire array, if it is uniform, the roots are dispersed in an equally
spaced format around the unit circle, and so their spacing is 2p/n, except on either
side of the pole at z = 1. Actually to further improve directivity the width of the
principal lobe can be reduced further, than suggested in Fig. 6.13b, by simply
adjusting the range of w so that for d = 0 the pole lies midway between w = 0 and
the first zero of the polynomial. This is termed the Hansen-Woodyard condition [4].
To achieve this adjustment an extra inter-element phase shift −p/n has to be
introduced, or equivalently a phase shift of almost −p distributed along the length
of the array.

Fig. 6.13 Diagrams showing the range of w for optimum element spacing. a Uniform broadside
array, b uniform endfire array. (The dotted arrows imply that the zeros occupy the entire circle)
148 6 Resonant and Non-resonant Linear Slot Arrays

6.3.2 Binomial Array

With fully adjustable arrays, in both element phase and in element power level,
detailed radiation pattern control, encompassing even far-out sidelobes, becomes
feasible. To emphasise this capability, here we will consider the formation of
radiation patterns with no sidelobes.
If the roots of the polynomial f(z) are all equal and can be co-located at the p
position on the unit circle, as suggest in Fig. 6.14, then the resultant pattern has
only one null or zero at w = p and consequently should possess no sidelobes. So
does this root location scheme lead to a sensible polynomial which can be trans-
formed into a realisable array? For the endfire array with the unit circle represen-
tation shown in Fig. 6.14, the polynomial must have the form:

1ðzÞ ¼ ðz þ 1Þn1 ð6:33Þ

The levels to which the individual elements of the array must be driven to secure
this solution can be deduced by expanding Eq. (6.33) as a binomial series. For
example for a four element array the relative magnitudes are in the ratios 1: 2: 2: 1.
This element excitation scheme will result in no sidelobes for the four element array
provided the element spacing d and the progressive phase shift a are also properly
imposed.

Fig. 6.14 Roots locations on unit circle for binomial array


6.4 Chapter Summary 149

6.4 Chapter Summary

Compact antennas for modern mobile communication systems generally require


some degree of beam formation and steering in order to optimise signal strength in
the direction of the satellite transmitter or relay antenna. This is particularly true in
communication systems based on interaction with low earth orbit (LEO) satellites
which traverse the sky in fifteen or so minutes. This can be done mechanically on
vehicles, such as ships, which can support gyro-stabilised antennas. For more
compact systems on much smaller platforms the ideal solution is electronic beam
control using arrays, a topic which attracts significant research interest.
The chapter aims to demonstrate that the essence of array pattern control,
associated with both uniform linear, and planar, array forms can be neatly sum-
marised using polynomial representations of their basic periodic forms. This now
classical analysis technique is employed throughout the chapter to examine the
relationship between array structure, in particular element numbers, element spac-
ing and element weightings, and how it impinges fundamentally on radiation pat-
tern attributes such as directivity, gain, beamwidth, sidelobes and grating lobes. It is
also emphasised that this powerful polynomial formulation presents an elegant
route, not just to array analysis, but also to array synthesis. Several cases are
examined to illustrate the technique.

References

1. S.A. Schelkunoff, W.T. Friis, Antennas, Theory and Practice (Wiley, New York, 1952)
2. R.S. Elliott, Beamwidth and directivity of large scanning arrays. Microwave J. (1963)
3. The Microwave Engineers Handbook, Page T-136 (1963)
4. M.M. Dawoud et al., Realization of superdirectivity from active and passive array antennas, in
IEE Conference Publication No. 169 (1978)

Bibliography

1. R.E. Collin, F.J. Zucker, Antenna Theory (McGraw-Hill Book Co., New York, 1969)
2. R.F. Harrington, Time Harmonic Electromagnetic Fields (McGraw-Hill Book Co., New
York, 1961)
3. C.A. Balanis, Antenna Theory (Wiley, New York, 1997)
4. R.S. Elliott, Antenna Theory and Design (Prentice-Hall Ltd., Englewood Cliffs, N.J., 1981)
5. R.C. Hansen, Microwave Scanning Antennas (Academic Press Ltd., New York, 1964)
6. E.C. Jordan, Electromagnetic Wave and Radiating Systems (Prentice-Hall Ltd., Englewood
Cliffs, N.J., 1968)
7. A.W. Rudge, K. Milne, A.D. Olver, P. Knight (eds.), The Handbook of Antenna Design:
Volumes I and II (Peter Peregrinus Ltd., London, 1986)
8. B.A. Munk, Metamaterials: Critique and Alternatives (Wiley, New York, 2009)
Chapter 7
Conventional Waveguide-Fed
Travelling-Wave Slot Arrays

7.1 Introduction

In preceding chapters we have examined in detail the nature of radiation (Chap. 2)


and how radiation is generated by individual aperture type radiators, including slots
(Chap. 3). Antennas of this type can be modelled very accurately and efficiently
using the moment method as we have demonstrated in Chaps. 4 and 5. Once
accurate models of individual radiating elements have been established we have
demonstrated in Chap. 6 how radiation can be focused and manipulated to procure
high gain and radiation pattern control by embedding the elements within a periodic
array. Unfortunately, the array techniques outline in Chap. 6 assume that rather
complex methods of feeding electromagnetic power to the individual elements are
available—namely by placing a stable source and phase shifter behind each ele-
mental radiator, or by adopting a complex transmission line distribution network to
channel power from a single source to the phase shifter controlled elements. Neither
scheme is particularly appropriate for realising compact antenna structures which
we are concerned with here.
Compact, efficient and inexpensive arrays are, arguably, most commonly
secured by applying travelling-wave principles to a linear array of radiating ele-
ments. The technique has been explored in the literature for over sixty years [1–4],
particularly in relation to slot arrays implemented in waveguide. Travelling-wave
arrays employing radiators other than slots, and transmission systems other than
waveguide [5] are also well established. Furthermore, planar array forms can be
realised by combining together similar linear arrays in appropriately aligned parallel
geometries. Examples of both linear and planar travelling-wave array antennas are
depicted in Chap. 1, Figs. 1.13 and 1.14. It is well known that the rectangular
waveguide provides an efficient low-loss method of transmitting electromagnetic
waves over finite distance. But, in addition, if the waveguide contains more than
one slot radiator, such as depicted in Fig. 1.12 all slots are excited by the same
travelling-wave, potentially forming a simple and efficient antenna array. As we

© Springer Nature Switzerland AG 2019 151


A. J. Sangster, Compact Slot Array Antennas for Wireless Communications, Signals
and Communication Technology, https://doi.org/10.1007/978-3-030-01753-8_7
152 7 Conventional Waveguide-Fed Travelling-Wave …

have seen in Chap. 6, desirably focused and directed radiation patterns depend on
the phase and radiation strength of the exciting electromagnetic wave at the position
of each slot. In the case of excitement by a travelling wave, this means slots
equi-spaced in the axial direction of the waveguide, generating *360° progressive
phase increases to secure an approximately broadside main radiation lobe. Lobe
location is a key issue with travelling wave arrays as we shall see.
Radiation from an aperture or slot machined or etched into a conducting wall of
a rectangular waveguide, or any other shape of waveguide for that matter, is
maximised by orientating the slot into alignment with the TE10 mode magnetic field
(H10) adjacent to the inner surface, or equivalently normal to the accompanying
surface current density (Js10). This behaviour is illustrated in Chap. 1 and
demonstrated theoretically in Chap. 5. For consistency, given that the longitudinal
shunt slot is examined in detail in Chap. 5, we will direct our attention here toward
the longitudinal slot array. Generally, for such an array resonant slots are employed
to maximise array bandwidth. In Chap. 5, Figs. 5.6, 5.7 and 5.8 inform us that at
resonance a longitudinal broadwall slot presents a real impedance to the waveguide
—so in transmission line terms it can be modelled with a shunt resistance or
conductance as shown in Fig. 7.1. A shunt, rather than a series, connection is
dictated by the phase behaviour of S11 and S12 for the longitudinal slot.
The travelling wave array antenna formed from resonant longitudinal slots
inserted periodically into a broadwall of waveguide has been examined widely in
the literature. This interest has been encouraged by the early observation [2] that the
off-set resonant slot, as depicted in Fig. 7.1a, can be modelled in transmission line
terms by a simple shunt conductance (Fig. 7.1b). In fact Stevenson [2] has gen-
erated an elegant closed form equation for the normalised shunt conductance g of a
narrow slot radiating into a half-space, namely
   
Ys kg a 2 pk0 pd
g¼ ¼ 2:09 cos sin2 ð7:1Þ
Y0 k0 b 2kg a

where d = a/2 − X1, Ys is the slot admittance in the waveguide (=G when real), Y0
is the waveguide characteristic admittance, while kg = kg10 is the TE10 mode
wavelength in the waveguide at the operating frequency. The validity of Eq. (7.1)
as a ‘trial’ value in any travelling wave array synthesis procedure is confirmed by
the moment method computations presented in Chap. 5. If the power entering the
waveguide in the TE10 mode is known then the guide voltage and hence the
equivalent transmission line value V (Volts) is available. Hence the power radiated
by the slot, normalised to the input power, for the scenario in Fig. 7.1b is simply:

V 2g
pr ¼ ð7:2Þ
2

In words, the slot excitation level is proportional to g, which in turn [Eq. (7.1)] is
 
proportional to sin2 pda : The primary mechanism for controlling the radiation
pattern for a shunt slot travelling wave array thus suggests itself. Namely, from
7.1 Introduction 153

Fig. 7.1 Longitudinal (shunt) slot in TE10 mode rectangular waveguide: a waveguide geometry,
b equivalent circuit for resonant slot with normalised g = Ys/Yo

estimated elemental power requirements, consistent with establishing a specified


radiation pattern, we determine the conductances of the slots forming the array, and
hence we can evaluate the slot off-sets d. The procedure is elaborated upon in the
next section.
154 7 Conventional Waveguide-Fed Travelling-Wave …

7.2 Travelling-Wave Array—Design Principles

The essence of establishing a slotted waveguide travelling-wave array design can


probably be explained most elegantly by focusing on the resonant slot example.
This approach is competently developed in Johnson [6], borrowing from earlier,
fuller treatments, in Silver [7], Collin and Zucker [8], and Elliott [9]. The method is
summarised here for completeness and ease of reference.
The basic geometry of the shunt slot array antenna, as implemented in rectan-
gular waveguide, is illustrated in Fig. 7.2. The longitudinal slots (z-directed) are
off-set from the waveguide side wall by distances xn for the n slots, and each
resonant slot is assumed to present a real conductance to the waveguide g1, g2, …..
gN, as indicated in Fig. 7.2b. Adjacent slots are on opposite sides of the centre line
to avoid grating lobes (see Chap. 6, Sect. 6.3.1) in the y-z plane. If the axial
separation distance between slots d is approximately equal to kg/2 then adjacent
slots differ in phase by almost 360° thus placing the main beam in the y-z plane
close to, but not precisely at, broadside (/ = 0°). It should be noted that d = kg/2 is
not possible for a slot array terminated in a matched load, as shown in Fig. 7.2. This
is because the resonant slots reflect in phase, resulting in unacceptable levels of

Fig. 7.2 Resonant slot travelling-wave array antenna (a) and equivalent transmission line model (b)
7.2 Travelling-Wave Array—Design Principles 155

combined power being returned to the array input, and hence the microwave source.
Operation of a travelling wave array exactly at broadside will be examined as a
special case at the end of this section.
The main lobe of the travelling wave array radiation pattern is located in the
angular range p=2  /  p=2 in the y-z plane. In the transverse x-y plane the
pattern is essentially omni-directional in the forward half-space (+y-direction).
Furthermore, the antenna is generally assumed to be terminated in a perfectly
matched waveguide load represented by admittance YL in the equivalent trans-
mission line.
For radiation pattern calculations in the ensuing section, we note here, referring
to Fig. 7.2, that geometrically the centre of any given slot (n) can be defined as
located at rn where

^y
rn ¼ nd^az þ xn ^ax þ b ð7:3Þ

For a fixed height waveguide, b is common to all slots in the array, and hence it
has no influence on pattern calculations and can therefore be omitted. Thus

rn ¼ nd^az þ xn ^ax ð7:4Þ

7.2.1 Staggered Shunt Slot Array

In Chap. 6 it has been noted that the radiation pattern for an array antenna can, in
the absence of mutual coupling, be elegantly formed from the product of the array
factor and the element factor. The former applies to the radiation pattern of the
equivalent array comprising isotropic radiators, while the latter alludes the radiation
pattern of the isolated element. For a uniform array of isotropic elements it is shown
that the radiation pattern can be simply represented mathematically [see Chap. 6,
Eq. (6.3)] as:

E
¼ 1 þ ejw þ e2jw þ    þ ejðn1Þw ð7:5Þ
E0

where Eo is the field at a given radius for a single element, while E is the radiated
field strength for the array at the same radius in the plane containing the array (y-z
in Fig. 7.2). Also, for unstaggered elements, w the propagation phase difference
between adjacent elements is given by (see Eq. 6.2) namely:

w ¼ k0 d sin / þ a ð7:6Þ

However, for the travelling wave array, the slots are seldom equally excited and
are seldom unstaggered, so Eqs. (7.5) and (7.6) have to be adjusted slightly to
represent a non-uniform structure. Hence we obtain:
156 7 Conventional Waveguide-Fed Travelling-Wave …

E
¼ a1 ejk0^rr1 þ jpjb10 d þ a2 ejk0^rr2 þ 2jp2jb10 d þ    þ an ejk0^rrn þ jnpjnb10 d ð7:7Þ
E0

where d is the element separation in the z-direction. It is also the distance from the
origin to the first element. an is the field excitation amplitude at the nth slot while
b10 is the propagation phase for the TE10 mode in the waveguide. The unit vector ^r
is given by:

^r ¼ ^ax sin D cos h þ ^ay sin D sin h þ ^


az cos D ð7:8Þ

with D ¼ p2  / in the y-z plane, and h the angular rotation from the x-axis in the
x-y plane. Consequently

w ¼ k0 d sin / þ p  b10 d þ k0 xn cos / cos h ð7:9Þ

Generally we are interested in the radiation pattern in the y-z plane (h = p/2
plane) in which case the most relevant and useful form of Eq. (7.9) is:

w ¼ k0 d sin / þ p  b10 d ð7:10Þ

Evidently, and perhaps not surprisingly, Eq. (7.10) closely matches Eq. (7.6)
except that the a phase term is replaced by the slot-to-slot phase delay (−b10d) for
the TE10 mode in the waveguide. The additional p phase shift term accounts for slot
staggering. It disappears for an array with shunts slot located only on one side of the
centre line.
Referring back to Chap. 6, we have observed there that the main radiating beam
in the plane of the array (y-z plane) occurs where w = 0, while secondary beams or
grating lobes occur at w = 2 mp (m = 1, 2, … M). That is, for primary lobes we
must have:

k0 d sin / þ p  b10 d ¼ 2mp ð7:11Þ

Hence primary lobes occur at values of the slot spacing d given by:

ð2m  1Þk0 kg10


d¼ ð7:12Þ
2ðkg10 sin /  k0 Þ

Equation (7.12) is plotted in Fig. 7.3, for low values m, for a resonant shunt slot
travelling-wave array located in 22.86 mm  10.16 mm (WG16) rectangular
waveguide at an operating frequency of 10.0 GHz. Equation (7.12) reveals that the
curves in Fig. 7.3 intercept the / = −90° axis at

ð1  2mÞk0 kg10
d¼ ð7:13Þ
2ðk0 þ kg10 Þ
7.2 Travelling-Wave Array—Design Principles 157

Fig. 7.3 Influence of slot spacing d on the principal beam locations (/ degrees) for a staggered
shunt slot travelling wave array in WG16 waveguide at 10 GHz

and on the / = +90° axis at

ð2m  1Þk0 kg10


d¼ ð7:14Þ
2ðk0 þ kg10 Þ

while the infinity occurs where

k0
sin / ¼ ð7:15Þ
kg10

The figure thus provides a clear and rapid indication of the number of principal
beams generated by the array for a particular value of d. At a given value of
d (60 mm say), if a horizontal line (blue dashed line) is inserted onto the figure at
this value then the possible main beams, or grating lobes, in /-space will be
predicted to occur at approximately −20°, 7°, 33° and 75°. To secure a single main
beam in the y-z plane the slot spacing d must be located between the m = 0 curve
and the m = −1 curve at / = −90°. That is, we must have:

k0 kg10 3k0 kg10


d ð7:16Þ
2ðk0 þ kg10 Þ 2ðk0 þ kg10 Þ

For WG16 operating at 10 GHz this implies that d should be in the range 8.55–
25.65 mm. The m = 1 solution between / = 75° and / = 90° is not significant in
158 7 Conventional Waveguide-Fed Travelling-Wave …

the above range of d if kg < 2k0. This condition lies in the normal operating range
for standard rectangular waveguides.

7.2.2 Unstaggered Shunt Slot Array

Although seldom adopted because of grating lobe issues a brief examination of the
unstaggered shunt slot array provides further insight into the nature of the
travelling-wave array antenna. As we have already noted, the fundamental phase
relationship governing primary beam locations, namely Eq. (7.11), is modified by
the absence of the p term when applied to the unstaggered case. Hence the
equivalent d-equation [to Eq. (7.12)] locating main beams becomes

mk0 kg10
d¼ ð7:17Þ
ðkg10 sin /  k0 Þ

This equation is presented in graphical form in Fig. 7.4.


Figure 7.4 is superficially the same as Fig. 7.3, but careful examination reveals
two main differences. The m = 0 solution in the staggered array case has become
the m = −1 solution. However this does not mean that an m = 0 solution does not
exist. From Eq. (7.17) it is clear that d is non-zero for m = 0 only where the
denominator is also zero, namely at sin / ¼ kkg10
0
: This condition is represented by

Fig. 7.4 Influence of slot spacing d on the principal beam locations (/ degrees) for a unstaggered
shunt slot travelling wave array in WG16 waveguide at 10 GHz
7.2 Travelling-Wave Array—Design Principles 159

the blue solid line in Fig. 7.4. It is solid because it now denotes a real solution,
usually termed the d-independent solution. Physically this operating condition
occurs when the phase velocity of the TE10 mode in the waveguide vp10 = x/bg10
matches the z-component of the velocity of the radiated beam along the outer
surface of the waveguide. For this solution to be a single beam operating mode, the
element spacing d must be less than for the m = −1 solution at w = −90°, namely

k0 kg10
d¼ ð7:18Þ
ðkg10 þ k0 Þ

or d < 17.1 mm in WG16 at 10 GHz.


This situation can be improved marginally by changing the frequency of oper-
ation to a lower value. This moves the TE10 mode toward its cut-off value
(6.56 GHz) such that kg10 increases more rapidly than k0. The result, as Fig. 7.5a
shows, is that the blue vertical line, at the ‘infinity’, moves leftwards toward a lower
value of /, increasing the separation between the curves representing m = −1, −2,
−3, and hence improving the d-range for single beam operation. The downside of
this change is a more dispersive TE10 mode. The opposite trend is evident when the
system frequency is raised as shown in Fig. 7.5b with the d-independent beam
shifting to a higher value of /.

7.2.3 Travelling-Wave Array Antenna Design

Long before the advent of electromagnetic wave simulation software, based on


either moment methods or finite element methods, and before the appearance of
desk top computers with both sufficient memory capacity, and computational speed,
to accommodate complex antenna investigations, successful techniques were being
developed to solve the specific problem presented by travelling-wave arrays of the
shunt slot variety [9, 10]. For dependable and quick assessments of preliminary
array designs, or for design engineers lacking access to modern simulation software
and the means to run it, the above problem specific solutions continue to be helpful.
Their relative simplicity mathematically, and their impressive accuracy, largely lies
with the fact that shunt (end-to-end) slots in waveguide incur negligible mutual
coupling, whereby the local fields from one slot, influence the fields in near
neighbours (see Chap. 5).
The design method summarised here is attributable to Kaminow and Stegen [4],
and is largely formulated by applying power conservation principles and trans-
mission line theory, to a relatively complete two-wire transmission circuit model for
the shunt slot array antenna, which is illustrated in Fig. 7.6. In the model the slots
are represented by normalised admittances (yn) rather than conductances (gn) to
accommodate the fact that longitudinal slots of finite length, even if resonant
(L = k0/2), introduce phase shift into the transmission (sn) and reflection coeffi-
cients (qn). Hence we define the complex admittance (normalised) of the nth slot as:
160 7 Conventional Waveguide-Fed Travelling-Wave …

Fig. 7.5 Influence of slot spacing d on the principal beam locations (/ degrees) for a unstaggered
shunt slot travelling wave array in WG16 waveguide at a 8 GHz and b 12 GHz

yn ¼ gn þ jbn ð7:19Þ

The normalised admittance ‘seen’ by the nth slot ‘looking’ toward the termi-
nation is also defined as:
7.2 Travelling-Wave Array—Design Principles 161

Fig. 7.6 Shunt slot travelling wave array antenna—design example a array configuration,
b general transmission line model

yTn ¼ gTn þ jbTn ð7:20Þ

Consequently, ‘looking’ toward the nth slot from the (n − 1)th slot, the com-
bined admittance becomes:

yTn1 ¼ yn þ yTn ¼ gTn þ jbTn þ gn þ jbn ð7:21Þ

Hence, using transmission line theory, the reflection coefficient seen by the nth
slot can be expressed as:

1  yTn
qn ¼ ð7:22Þ
1 þ yTn

For generality, the waveguide is assumed to be imperfect exhibiting a small level


of wall loss. This is accommodated in the model by defining a complex propagation
coefficient:
162 7 Conventional Waveguide-Fed Travelling-Wave …

c10 ¼ a10 þ jb10 ð7:23Þ

Now the full range of power contributions at the nth slot is delineated in
Fig. 7.6b, and hence applying power conservation there, we obtain:

Pn ¼ Pfn  Pqn þ Pbn  Psn ð7:24Þ

where, Pn is the power radiated by slot n, Pfn is the power in the waveguide
reaching slot n, Psn is the forward power in the waveguide on the load side of the
nth slot, Pqn is the backward power on the source side of the nth slot and Pbn is the
‘backward’ flowing power in the waveguide on the load side of the nth slot.
Equation (7.24) can be simplified into (Fig. 7.6b):

Pn ¼ Pn jin Pn jout ð7:25Þ

where Pn jin and Pn jout represent the net powers in the waveguide flowing into and
out of the nth slot. If the voltage on the transmission line representation of the slot
array (Fig. 7.6b) at the nth slot is Vn then the following definitions apply:

1
Pn ¼ jVn j2 gn ð7:26Þ
2

and

1
Pfn ¼ jVn j2 gTn ð7:27Þ
2

Also at the termination:

1
PT ¼ jVT j2 gT ð7:28Þ
2

The travelling-wave array design commences with the observation, as we have


already noted, that the required radiation pattern for the array is dictated by the slot
excitation levels an (see Chaps. 3 and 6), such that Pn / a2n . We also note that in a
lossless waveguide Vn = VT = V, for all n. Generally the travelling-wave array is
created in low loss waveguide, in which case a trial starting point for a
travelling-wave array design is to assume that the waveguide loss is sufficiently
close to zero to let Vn = V = constant. In this case, from Eq. (7.26), we can state
that:

Pn ¼ Cg a2n ð7:29Þ

The constant Cg can be estimated by introducing a trial value for that proportion
of the source power reaching the load after passing through the array (r say). Thus
7.2 Travelling-Wave Array—Design Principles 163

PT
r¼ ¼ PT ð7:30Þ
Pin

if power from the source Pin is normalised to unity. With these simplifications we
can define [9] a continuous conductance per unit length g(z) such that:
gn
gðzÞ ¼ gðndÞ ¼ ð7:31Þ
d

In this approximated ‘array’, now represented by a continuous conductance, we


can also replace Pn by PðzÞ and Pfn by Pf ðzÞ: Consequently for a loss free
waveguide and a perfectly matched load, if the input power is normalised to unity
then at any position z:

PðzÞ
gðzÞ ¼ ð7:32Þ
Pf ðzÞ

Zz Zz
Pf ðzÞ ¼ 1  PðzÞdz ¼ 1  Cg ðaðzÞÞ2 dz ð7:33Þ
0 0

with a(z) = an/d and using the continuous version of Eq. (7.29). Furthermore at the
load end of the array (z = L) we must have, from Eqs. (7.30) and (7.33):

ZL
r ¼1 Cg ðaðzÞÞ2 dz ð7:34Þ
0

On combining Eqs. (7.32), (7.33) and (7.34) an elegant expression for the
continuous conductance is generated—namely:

ðað1ÞÞ2
Lgð1Þ ¼ ð7:35Þ
RL RL
1
1r ðað1ÞÞ2 d1  ðað1ÞÞ2 d1
0 0

where f = z/L. Equation (7.35) is plotted in Fig. 7.7 for a(f) distributions with
Taylor functionality in order to produce Chebyschev radiation patterns [10]—one
which produces −20 dB flat sidelobes in the array radiation pattern, while the
second case develops −25 dB sidelobes.
Four curves are presented in Fig. 7.7, two associated with −20 dB sidelobe
designs (blue traces with circle markers) and two associated with −25 dB sidelobe
designs (red traces with square markers). In both cases the curves are differentiated
by the power (r), as a percentage of the source power, reaching the load. It is
noticeable that for the low power cases (r = 5%) g(f) as a function f is much more
sensitive to increasing f than for the cases where r = 35%. The reason for this
164 7 Conventional Waveguide-Fed Travelling-Wave …

Fig. 7.7 Continuous conductance profile for lossless, perfectly matched, shunt-slot
travelling-wave array antennas (with acknowledgement to [9])

difference in behaviour lies with the fact that with only 5% of the input power
reaching the load the last few slots in the array are exposed to a very weak level of
pffiffiffiffiffiffiffiffiffi
TE10 mode excitation. However, given that að1Þ / Pð1Þ , it becomes clear from
Eq. (7.31) that a low power level (Pf(f)) requires a large g(f) value to secure the
desired slot excitations a(f). This difficulty is exacerbated for the −20 dB sidelobe
array which requires higher a(f) levels at the slots near the array extremities to
procure the higher sidelobes. Note that this option equates to higher gain. In fact the
way in which the magnitude of g(f) escalates as f approaches unity highlights a
significant design limit for the shunt slot travelling-wave array antenna. Eventually,
as r is reduced, the level of g(f) cannot be realised by a shunt slot in waveguide
where the maximum g(f) is restricted by the optimum off-set (d = a/2) of a slot
from the centre line of the waveguide [see Fig. 7.1 and Eq. (7.1)].
A compromise between the magnitude of r, array efficiency and the desired gain
becomes inevitable. High efficiency (low r) is easier to procure with low sidelobe
levels which equate to less than optimum gain. In general this balancing process
becomes less pronounced for long arrays with large N. With r at the 35% level in
Fig. 7.7 it is interesting to note that the g(f) distribution along the array follows
closely the Taylor distribution applied to a(f). This is simply because Pf(f) remains
relatively constant along the array in these cases. Finally, low sidelobe designs
require care in implementing a terminating load which must be well matched to the
feed waveguide forming the array. An imperfect match results in a reflected TE10
mode which becomes a source of a secondary radiation pattern with a ‘main’ beam
pointing in a direction equal and opposite, relative to the array normal, to the
7.2 Travelling-Wave Array—Design Principles 165

designed beam. This ‘reflection’ beam together with the original sidelobes in the
same direction should ideally combine to form a lobe which is higher than the
designed sidelobe level. Calculations performed by Dion [10] suggest that at the
−30 dB sidelobe level in an array designed on the basis of ‘dumping’ 5% of the
source power in the termination, a load with a generous VSWR of *1.34 would be
satisfactory. On the other hand, at r = 35% a high quality load with a VSWR  1.1
is required to maintain the sidelobe specification. At the −20 dB sidelobe level the
load requirements are, not unexpectedly, considerably more relaxed with a load
with a VSWR  1.4 being acceptable, even at r = 35%.
For a slot array of predetermined length L, the Dion continuous conductance
technique permits the allocation of trial values for the discrete conductances gn in an
ensuing practical design. For such an N element array design with L = Nd we
choose gn as defined in Eq. (7.31) with z = nd, on the assumption that the first slot
is located a distance d from the array input. An estimate for the parameter r is also
furnished by the Dion analysis and hence the power PT transmitted into the load is
calculable.
For an array more accurately modelled by admittances (Fig. 7.6b) embedded in a
waveguide, which is not loss free, Kaminow and Stegen [4] provide an iterative
procedure to procure a design. This commences with the trial conductances from
the above procedure and the known conditions at the load, then progresses slot by
slot back toward the array input, by employing power conservation principles and
transmission line theory. The power conservation requirement is defined in
Eq. (7.24), with power definitions in Eqs. (7.26), (7.27) and (7.28). Since it is
unlikely that the termination will be other than well matched we can reasonably set
pffiffiffiffiffiffi
gT = 1, in which case VT ¼ PT .
At slot N, which we assume is distance d from the load, the power conservation
requires:

PN jout ¼ expða10 dÞPT ð7:36Þ

and for a matched termination Eq. (7.32) applies, giving:

PN PN
gN ¼  ð7:37Þ
PfN PN jin

Now, for a lossy line, transmission line theory gives the voltage at slot N as:

VN ¼ VL ðcoshðc10 dÞ þ yT sinhðc10 dÞÞ ð7:38Þ

while at slot N − 1 we obtain:

VN1 ¼ VN ðcoshðc10 dÞ þ yTN sinhðc10 dÞÞ ð7:39Þ


166 7 Conventional Waveguide-Fed Travelling-Wave …

using Eq. (7.21). At the Nth slot transmission line theory also yields:

yT þ tanhðc10 dÞ
ysN ¼ ð7:40Þ
1 þ yT tanhðc10 dÞ

Furthermore from Eq. (7.21)

yTN ¼ yN þ ysN ¼ gN þ jbN þ gT ð7:41Þ

with bN computed from:

jVn yn
\ ¼ \ expðjb10 dÞ ð7:42Þ
jVn þ 1 yn þ 1

This equation ensures that the slot to slot excitation phase shifts in accordance
with array beam formation in the specified design direction. It leads to the following
iterative relationship for phase, namely

bn bn þ 1 Vn þ 1
tan1 ¼ tan1 þ b10 d þ \ ð7:43Þ
gn gn þ 1 Vn

For the Nth slot, ‘looking’ into a matched termination, this reduces to:

bN
tan1 ¼ b10 d ð7:43bÞ
gN

With PN, gN and bN at the final slot now established the iteration becomes
focused on the (N − 1)th slot. Firstly the power levels at slot N can be transferred,
in accordance with transmission line theory and power conservation, back to slot
N − 1 by applying power conservation at a reference plane on the load side of slot
N − 1 (see Fig. 7.6). This leads to

PN1 jout ¼ Ps;N1  Pb;N1 ¼ expð2a10 dÞPf ;N  jqN j2 expð2a10 dÞPf ;N


h i
¼ 1jqN j2 expð2a10 dÞPf ;N þ jqN j2 expð2a10 dÞPf ;N  jqN j2 expð2a10 dÞPf ;N
h i
¼ 1jqN j2 expð2a10 dÞPf ;N þ 2jqN j2 sinhð2a10 dÞPf ;N
ð7:44Þ

And
h i
PN jin ¼ 1  jqN j2 Pf ;N ð7:45Þ

The following useful equation [10] relating the N − 1th slot to the Nth slot then
results:
7.2 Travelling-Wave Array—Design Principles 167

( )
2 j qN j 2
PN1 jout ¼ PN jin expð2a10 dÞ þ sinhð2a10 dÞ ð7:46Þ
1  jqN j2

where qn is defined in Eq. (7.22). Hence with PN known at the Nth slot and PN jin
available from Eq. (7.37) the power leaving the N − 1th slot can be calculated
using Eq. (7.46). The power on the source side of slot N − 1 is computed using
Eq. (7.25) with gn−1 derived from the transmission line Eq. (7.40). VN−1 is deter-
mined from transmission line theory [Eq. (7.39)] which then permits the evaluation
of bN−1. The process is then repeated for slot N − 2, N − 3, etc. back to slot 1.
Eventually, with all of the gn’s and bn’s evaluated the slot offsets dn, the slot
spacings dn and the slot lengths Ln can be deduced from Eqs. (7.1) and (7.12) with
the slot resonant lengths ‘tweeked’ to accommodate the susceptance values.
Actually, the sophisticated antenna simulation software packages which are avail-
able to today’s designers, allows the electrical data generated above for the
travelling-wave array antenna to be inserted into a waveguide slot analysis program
(see Chap. 5) to ascertain accurate dimensional design parameters including slot
widths and waveguide wall thicknesses.

7.2.4 Resonant Array Design

A disadvantage of the travelling-wave array antenna described in the preceding


sections is that the operating beam, or main beam, is not located at broadside
(/ = 0), and furthermore it scans with changes in frequency, as is made clear in
Figs. 7.3, 7.4 and 7.5. However, a broadside shunt slot array can be realised by
‘resonating’ the waveguide if bandwidth is not an issue. A typical application where
a very narrow bandwidth is acceptable is in low cost antennas for search and rescue
radar beacons. The concept is illustrated in Fig. 7.8. The array is ‘shorted’ exactly
kg10/4 (or 3kg10/4) beyond the last slot, while the staggered shunt slots forming the
array are spaced exactly kg10/2 apart. The result is a standing-wave excited slot
array rather than a travelling-wave excited array. Note that while the slots are also
operated at their resonant length (k0/2), the array format is termed ‘resonant’
because of the standing-wave excitation mode. The essence of this mode is depicted
in Fig. 7.9. The TE10 mode which is presumed to enter the waveguide from the left
(Fig. 7.8a) is totally reflected at the shorting plate terminating the waveguide on the
right. In the absence of slots, and in a lossless waveguide, a standing-wave with
perfect nulls at distances pkg10/2 (p = 1, 2, 3 …) in front of the shorting wall will be
formed.
For resonant length slots separated by precisely kg10/2 transmission line theory
dictates that the effective conductance at the input to the array (ge), as depicted in
Fig. 7.9, is given by:
168 7 Conventional Waveguide-Fed Travelling-Wave …

Fig. 7.8 Resonant form of the waveguide shunt slot array a schematic representation, b circuit
representation

Fig. 7.9 Pictorial representation of the slot excitation mechanism in the resonant array
7.2 Travelling-Wave Array—Design Principles 169

X
N
ge ¼ gn ð7:47Þ
n¼1

where N is the number of slots and for a good array matched to the source, normally
we would choose ge = 1. Note that the kg10/4 short circuited line presents an open
circuit across the final slot.
The matching condition alone is insufficient to determine all of the gn values, but
we have seen from Chap. 3, and from Chap. 6, that the field distribution across the
aperture of the antenna, or the element excitation level in the case of an array, can
be shaped to procure high directivity or low side lobe radiation patterns (see
Fig. 3.7). Consequently, for the resonant array let the excitation level, or coefficient,
for the nth slot be an. For a uniform array (see Chap. 6) the an’s have a fixed level,
but may be set to a cosine pattern (for example) if lower sidelobes are desired. Since
the slot conductance is proportional to the radiated power [see Eq. (7.2)] we can
therefore write:

gn ¼ Cg a2n ð7:48Þ

where Cg is a constant yet to be defined. Hence combining Eqs. (7.47) and (7.48)
yields:

X
N
Cg a2n ¼ 1 ð7:49Þ
n¼1

for a matched array. If the an’s are established to meet pattern requirements then the
constant Cg becomes calculable. Once Cg is known the gn’s can be determined and
hence the remaining unknowns required to complete the array design, namely the
slot off-sets d, can be computed by resorting to Eq. (7.1).

7.3 Some Slot Array Applications

With their relative simplicity, from a geometrical perspective, their robustness, their
reliability and efficiency, slotted waveguide array antennas have found particular
favour in airborne radar, radiometry and satellite communications roles.
Furthermore, in both resonant and travelling-wave forms, control of slot excitation
levels and phase relationships can be sufficiently precise, as we have seen, to permit
array designers to procure antenna radiation patterns with tightly specified gains,
and sidelobe levels. This leads to high operating efficiencies. It is fair to say that
modern precision manufacturing techniques now permit relatively faithful repli-
cation of computer simulations, and this has resulted in array designs displaying
sidelobe levels well below −20 dB. A typical ‘flat plate’ waveguide slot array is
170 7 Conventional Waveguide-Fed Travelling-Wave …

shown in Fig. 1.13. The flat plate diameter, in the case of a circular embodiment,
can range from well over a metre at S and C band frequencies to little more than a
few centimetres at mm-wave frequencies (>60 GHz). Applications for such
antennas range widely from airborne tracking and seeking roles to ground based
mobile phone systems which predominantly employ linear array versions of the
slotted waveguide antenna (see Johnson [6] and Rudge [11]).
Following pioneering effort in the 1940s and 1950s [1–3], slot array antennas in
radar, and in communications applications flourished, arguably most fruitfully and
determinedly during the ‘cold war’. However by the 1990s these developments had
seemed to stall. The advent of ever more precise manufacturing methods, such as
laser machining, electron beam technology, photographic etching and 3D printing,
have recently begun to push applications into the mm-wave frequency range (circa
60 GHz). The low-loss advantage of waveguide, over alternative feed mechanisms,
remains a major attraction for antenna designers, and unquestioningly helps to
maintain interest in slotted waveguide arrays. At mm-wave frequencies relatively
small arrays can procure the tightly controlled radiation patterns, which are
instrumental in the realisation of atmospheric sensors, automotive radar sensors and
gigabit/second wireless interfaces [12]. In the ‘wireless’ category, a number of
applications are being planned such as high definition multimedia interfaces
(HDMI) enabling (1) uncompressed high definition (HD) video, streaming wire-
lessly rather than by inconvenient cable technology, (2) computer mobility in
distributed environments, (3) wireless docking stations, (4) wireless gigabit
Ethernet operation, (5) fast transfers of bulky files, (6) wireless gaming [13]. Other
applications will undoubtedly emerge in the near future.
Wireless systems at 60 GHz have materialised as one of the most promising
candidates for multi-gigabit indoor communication links. Evidently, a deciding
factor that encouraged this recent interest, lies with the huge unlicensed bandwidth
(up to 7 GHz) available worldwide. While this is comparable to the unlicensed
bandwidth allocated for ultra-wideband (UWB) systems, at 60 GHz the avail-
able bandwidth is continuous and power limits are much less restrictive. This
difference is governed by the fact that an UWB system must operate in concert with
other systems, and is thus subject to very much more demanding regulations [5].
In communication terms, the bandwidth available at 60 GHz band is one of the
largest unlicensed bandwidths which has ever been allocated. The huge 7 GHz
bandwidth has the potential to enable systems demanding very high levels of
channel capacity and flexibility thus making 60 GHz technology particularly
attractive for gigabit wireless applications. Furthermore, 60 GHz regulation allows
much higher transmit power compared to other existing wireless local area net-
works (WLANs) and wireless personal area networks (WPANs) operating at
microwave frequencies, so accommodating and counteracting the higher transmit
power levels, which are necessary to overcome the inevitably higher path losses
encountered at 60 GHz.
At the time of writing, it is true to say that a large number, and a wide range, of
planar antennas have been studied for millimeter-wave radio and radar applications.
The most obvious alternative to the waveguide slot array is the substrate mounted
7.3 Some Slot Array Applications 171

patch array [6]. While high gain operations have been demonstrated with the latter,
microstrip substrates suffer from serious loss mechanisms at mm-wave frequencies.
Consequently, the efficiency diminishes as the gain and/or frequency is pushed
upwards. Despite the maturity of patch array technology these limitations are
unavoidable. It has been estimated that the efficiency of microstrip arrays in the
60 GHz band, with gains of the order of 35 dBi, could be 20% lower than an
equivalent waveguide slot array [16]. On-chip antennas have also been mooted for
mm-wave wireless roles but these exhibit their own particular drawbacks. On
conductive high-permittivity silicon their radiation efficiency is inevitably low. As a
result, in a WPAN environment, where wireless communication distances of the
order of 10 m need to be accomplished, multi-element arrays are necessary to
achieve the 15–20 dBi gain levels required for reliable transmission. If imple-
mented in MMIC chip technology such systems attract high cost [16].
It is, therefore, difficult not to conclude that, for wireless applications, waveguide
slot antenna arrays represent the most attractive candidates where high-gain planar
antennas, with high efficiency, are sought. Nevertheless waveguide fed slot arrays
present their own difficulties where low cost is important. Primarily, the compli-
cated three dimensional feed arrangements in early designs tended to limit their use
to cost-insensitive military or professional applications. An array configuration with
manufacturing costs potentially comparable to that of microstrip counterparts was
required. The solution lay with single-layer waveguide slot array designs which
lend themselves to mass production [13–18].
The impetus for this development was to note that, in principle, planar
waveguide fed slot arrays can be reduced to two parts. These are the slotted top
plate and the grooved bottom plate containing the waveguide channels (Fig. 7.10).
Several versions of these single-layer waveguide arrays, intended for high efficiency
over 12 and 20 GHz bands, were initially developed [13–17]. Subsequently,
manufacturing methods improved sufficiently to permit the extension of single layer
waveguide arrays to higher frequencies up to and including 60 GHz [18–20].
Slotted waveguide, single-layer, planar array antennas, aimed at satisfying
mm-wave developments in communications, evolved rapidly after 1995, under the
promptings of research scientists at the Tokyo Institute of Technology in particular.
Initially these revolved around the conceptual arrangements depicted in Fig. 7.10
and in particular the alternate phase arrangement in Fig. 7.10b [13]. This anti-phase
relationship between adjacent waveguides is counteracted by the ‘mirrored’ slot
staggering, which is clearly seen by comparing Fig. 7.10a, b. So while the phase
reversals have no influence on the beam forming, it has significant implications for
the wave propagation in the waveguide grooves. With no net current flow in the
internal walls the degree of electrical contact between the slot plate and the
channelled plate could be greatly relaxed. This meant that more cost effective array
manufacturing techniques could be explored [15, 16].
In Ref. [15] it was shown that while anti-phase operation of a waveguide fed
planar slot array eliminated the need for ‘good’ electrical contact between the
common internal waveguide walls, and the slotted top plate, the same did not apply
to the peripheral walls. The solution was to introduce traditional ‘choke’ technology
172 7 Conventional Waveguide-Fed Travelling-Wave …

Fig. 7.10 Single layer planar slotted waveguide array antennas with a co-phase corporate feed
b anti-phase transverse linear waveguide feed
7.3 Some Slot Array Applications 173

(see Fig. 1.13), long used to suppress leakage between minimal contact waveguide
flanges. With the addition of choke suppression, Ref. [16] demonstrated that mass
production of the waveguide antenna at low cost was possible. Die-casting was
adopted for the corrugated base plate of the antenna to circumvent high precision
metal cutting which was considered to be too time-consuming and expensive. The
dimensional tolerance associated with die-casting is typically only 50lm and
normally some further machining is necessary in the 26 GHz band. To avoid this, a
high precision casting technique was developed (see Ref. [16]).
To confirm the fabrication process, a travelling-wave array antenna with a
specified gain of over 30 dBi was designed, for operation at approximately
25 GHz. The aperture area occupied by slots was set to be 16 cm  15 cm. It
comprised 16 parallel waveguides, with a = 8 mm and b = 3 mm, each feeding 19
staggered shunt slots. For the die-cast antenna the gain exceeded 31 dBi over the
frequency range and the corresponding efficiency was more than 65%. The return
loss was a creditable −20 dB. At that juncture in the evolution of arrays for
mm-wave systems (circa 2005), the planar slotted waveguide array antenna could
justifiably be claimed to surpass competitive offerings in any given frequency and
gain range.
More recent developments [17] acknowledged the ‘beam-squint’ problem
encountered with travelling wave arrays (see Figs. 7.4, 7.5 and 7.6) which led to the
evolution of matched shunt slots incorporated into a planar array of centre fed
slotted waveguides. By feeding each linear sub-array, not from their ends as in
Fig. 7.10, but from their mid-points, the long line effect was halved with con-
comitant bandwidth enhancement. But more significantly, this symmetrical feed
scheme enforced the main beam of the array as a whole to remain on the boresight.
Furthermore, the necessity for the sub-arrays to be designed for non-boresight
operation to minimise return loss was greatly mitigated by employing matched
slots. Reflection cancellation can be procured in a number of ways, usually by
introducing an insert in the vicinity of the slot, which exhibits a reflection phase
counter to that of the slot. The inductive post was found to be the most effective
mechanism for suppressing the cumulative slot reflection at the input port, thus
allowing the beam tilting technique to be discarded [17]. The resultant antenna
displayed wide bandwidth and a main beam directed steadily at boresight. However
sidelobes were higher than equivalent end-fed examples of the same antenna due to
the blockage incurred by the centred feed geometry.
Later developments in the slotted waveguide array scenario were very much
driven by poor transmission compatibility with microwave integrated circuits. This
disadvantage does not arise to the same extent with patch arrays fed from mi-
crostripline or coplanar waveguide. Needless to say the realisation of single-layer
waveguide arrays offering well designed well matched transition structures for
efficient integration with other planar circuits became a high priority [18]. This so
called substrate integrated waveguide (SIW), also called post-wall waveguide or
laminated waveguide, was realised in a metal-clad dielectric substrate by intro-
ducing two rows of metallised via-holes by a routine printed circuit-board fabri-
cation technique (see Fig. 7.11). Importantly this could be accomplished at low
174 7 Conventional Waveguide-Fed Travelling-Wave …

Fig. 7.11 Millimetre-wave slotted waveguide array antenna in substrate based post-wall or SIW
waveguide [18]

cost. It has been demonstrated [18] that the SIW slot array antenna could easily be
integrated with other circuits, resulting in cost-effective subsystems. A range SIW
slot antenna arrays and beam forming networks have been developed [19]. In Ref.
[19] in addition to the extension of the SIW concept from 60 to 79 GHz the
designed antenna was formed on sufficiently thin substrate sheets (flexible PC foil)
to limit waveguide leakage problems in the dielectric and at the post-wall, which
can arise at mm-wave frequencies. In the polyimide foil the post wall is replaced by
a closely packed ‘forest’ of nickel wires formed by a process termed ion track
technology. Physical flexibility was a useful bonus.
The loss of gain and efficiency which is unavoidable with substrate based slot
array solutions, led some researchers to persist with hollow waveguide options in
order to retain its intrinsic low loss advantages [20–23]. To procure such a hollow
waveguide antenna design, at a cost level compatible with substrate based alter-
natives, development gravitated toward a structure described as a double-layer,
corporate-feed, waveguide slot array [20]. The typical antenna consisted of a cor-
porate waveguide feed (see Fig. 7.10a) arranged in the lower layer, and 2  2 small
subarray units in the upper layer. The 4 elements in the subarray are almost equally
spaced from the coupling aperture between layers at the end of each branch of the
corporate feed so that the array is substantially fed by a full corporate feed
(Fig. 7.12). With constant element spacing of less than the free-space wavelength,
grating lobes were effectively controlled.
7.3 Some Slot Array Applications 175

Fig. 7.12 Corporate feed and laminated waveguide slot array antenna

However, fabricating the low-loss hollow waveguide feed structure embedded


within a solid metal plate was found to be verging on impossible by the available
manufacturing techniques, such as machining or die-casting. The answer was to
construct the structure from very thin metallic plates etched with carefully posi-
tioned apertures forming elements of the feed waveguide. When laminated, regis-
tered and secured by diffusion bonding, the layered structure incorporated the
required hollow waveguides, coupling slots and radiating slots (Fig. 7.12) of the
designed waveguide slot array antenna. This method was demonstrated to provide
the necessary high precision at potentially low cost.
176 7 Conventional Waveguide-Fed Travelling-Wave …

While in the 12 GHz band, a cavity-backed planar slot array antenna with
16  16-elements had demonstrated a gain of 33.7 dBi with 85% efficiency and
12% bandwidth (VSWR < 2) using conventional fabrication techniques, it was
difficult to repeat these characteristics in the millimetre wave band. Greater fabri-
cation accuracy was needed. This was further illustrated [24] in a 40 GHz band,
two-layer, slotted-waveguide antenna array with 24  24-elements. While it dis-
played a creditable gain of 33.8 dBi, efficiency was appreciably lower at 51%, and a
narrower bandwidth of 5.5% was measured, even when the array was carefully
assembled and secured with screws. In contrast, later attempts [20, 21] which have
employed the laminated fabrication method, achieved superior efficiency and
bandwidth outcomes from a corporate feed hollow waveguide array. The figures
compared very favourably with lower frequency examples, even in the high fre-
quency band of 60 GHz. An illustrative photograph of the 60 GHz corporate feed
hollow waveguide array antenna (16  16 slots), showing the laminated con-
struction is presented in Fig. 7.13. The robust yet compact form for of the manu-
factured array is clearly seen particularly when compared with a coin similar in size
to a UK sterling 10p piece. The inclination of the slots by 45o relative to the main
axis of the almost square antenna, represents a simple yet effective scheme to
minimise grating lobes [21].

7.4 Chapter Summary

The slotted-waveguide travelling wave array, which forms a primary source of


robust, efficient, high gain and cost effective antennas, represents an ideal candidate
for many communications roles. It has been examined in detail in this chapter. In
particular, it is shown how the distinctive and basic dimensional choices presented
by the slot array exert a strong influence the radiation pattern trade-offs of gain,
sidelobe level and absence of grating lobes. Furthermore, they are demonstrated to
have significant implications for frequency sensitive beam scanning, which inevi-
tably dictates available bandwidth.
Iterative procedures, relying both on established transmission line models of the
slotted waveguide and on power conservation requirements at every element of the
array, are revisited to illustrate the fundamental design method. Typical results are
evaluated.
The chapter is completed with a review of recent developments aimed at
achieving the importation of slotted waveguide array techniques into antenna
applications which have been propelled by the opening up of millimetre-wave
frequency bands for exploitation by digital communication systems, atmospheric
sensing and wireless networks.
References 177

Fig. 7.13 Photograph of laminated hollow waveguide slot array a top view, b side view [21]

References

1. H.A. Bethe, Theory of diffraction by small holes. Phys. Rev. 66(7–8), 168–183 (1941)
2. A.F. Stevenson, Theory of slots in rectangular waveguide. J. Appl. Phys. 19, 24–38 (1948)
3. A.A. Oliner, The impedance properties of narrow radiating slots in the broad face of
rectangular waveguide. IRE Trans. AP-5, 4–20 (1957)
4. I.P. Kaminow, R.J. Stegen, Waveguide Slot Array Design (Hughes Aircraft Co. Technical
Memorandum, 348, Culver City, California, 1954)
5. A. Derneryd, Linear polarised microstrip antennas. IEEE Trans. Antennas Propag. AP-24,
846–850 (1976)
6. R.C. Johnson, Antenna Engineering Handbook (McGraw-Hill Inc., New York, 1993)
7. S. Silver, Microwave Antenna Theory and Design (McGraw-Hill Book Co., New York, 1949)
8. R.E. Collin, F.J. Zucker, Antenna Theory (McGraw-Hill Book Co., New York, 1969)
178 7 Conventional Waveguide-Fed Travelling-Wave …

9. R.S. Elliott, Antenna Theory and Design (Prentice-Hall Ltd., Englewood Cliffs, N.J., 1981)
10. A. Dion, Non-resonant slotted arrays. IRE Trans. Antennas Propag. AP-6, 360–365 (1958)
11. A.W. Rudge, K. Milne, A.D. Olver, P. Knight (eds.), The Handbook of Antenna Design:
Volumes I and II (Peter Peregrinus Ltd., London, 1986)
12. S.K. Yong, C.-C. Chong, An overview of multigigabit wireless through millimeter wave
technology: potentials and technical challenges. EURASIP J. Wirel. Commun. Netw. 2007,
1–10 (2007)
13. K. Sakakibara, Y. Kimura, A. Akiyama, J. Hirokawa, M. Ando, N. Goto, Alternating
phase-fed waveguide slot arrays with a singlelayer multiple-way power divider. Proc. Inst.
Electr. Eng. Microwaves Antennas Propag. 144, 425–430 (1997)
14. M. Ando, J. Hirokawa, in High Gain and High Efficiency Single Layer Slotted Waveguide
Array in 60 GHz Band. Proceedings of the 11th International Conference on Antennas and
Propagation, Publication No. 436, vol. 1 (Edinburgh, UK, 1997), pp 464–468
15. Y. Kimura, T. Hirano, J. Hirokawa, M. Ando, Alternating-phase fed single-layer slotted
waveguide arrays with chokes dispensing with narrow wall contacts. Proc. Inst. Electr. Eng.
Microwaves Antennas Propag. 148, 295–301 (2001)
16. Y. Kimura, Y. Miura, T. Shirosaki, T. Taniguchi, Y. Kazama, J. Hirokawa, M. Ando, A
low-cost and very compact wireless terminal integrated on the back of a waveguide planar
array for 26 GHz band fixed wireless access (FWA) systems. IEEE Trans. Antennas Propag.
53(8), 2456–2463 (2005)
17. S. Park, Y. Tsunemitsu, J. Hirokawa, M. Ando, Centre feed single layer slotted waveguide
array. IEEE Trans. Antennas Propag. 54(5), 1474–1480 (2006)
18. X.-P. Chen, K. Wu, L. Han, F. He, Low cost high gain planar array antennas for 60 GHz band
applications. IEEE Trans. Antennas Propag. 58(6), 2126–2129 (2010)
19. S. Cheng, H. Yousef, H. Kratz, 79 GHz slot antennas based on substrate integrated
waveguides (SIW) in a flexible printed circuit board. IEEE Trans. Antennas Propag. 57(1),
64–71 (2009)
20. Y. Miura, J. Hirokawa, M. Ando, Y. Shibuya, G. Yoshida, Double-layer full-corporate-feed
hollow-waveguide slot array antenna in the 60 GHz-band. IEEE Trans. Antennas Propag. 59
(8), 2844–2851 (2011)
21. T. Tomura, Y. Miura, M. Zhang, J. Hirokawa, M. Ando, A 45o linearly polarised hollow
waveguide corporate-feed slot array antenna in the 60 GHz band. IEEE Trans. Antennas
Propag. 60(8), 3640–3646 (2012)
22. M. Zhang, J. Hirokawa, M. Ando, Fabrication of a slotted waveguide array at 94 GHz by
diffusion bonding of laminated thin plates. IEICE Trans. Commun. E93-B(10), 2538–2544
(Oct 2010)
23. D. Kim, M. Zhang, J. Hirokawa, M. Ando, in Design of a Dual-Polarisation Waveguide Slot
Array Antenna Using Diffusion Bonding of Laminated Thin Plates for 60 GHz Band. IEEE
Antennas and Propagation Society International Symposium, 2012. [https://doi.org/10.1109/
aps.2012.6348506]
24. S.S. Oh, J.W. Lee, M.S. Song, Y.S. Kim, Two-layer slotted waveguide antenna array with
broad reflection/gain bandwidth at millimetre-wave frequencies. IEE Proc. Microwave
Antennas Propag. 51(5), 393–398 (2004)
Chapter 8
Frequency Scanned and Leaky-Wave
Linear Slot Arrays

8.1 Introduction

The aim of this chapter is to provide the reader with a thorough introduction to the
nature of the transition from ‘simple’ slot arrays in conventional waveguides to
often quite complex configurations, generated using advanced simulation tech-
niques. Such antennas cannot readily be analysed using the methods outlined in
Chaps. 5, 6, and 7. Additionally it is intended that the following sections will
provide a ‘bridge’ to the exploration in Chap. 9 of open periodic structures in so far
as they are employed to realise compact array forms, and thereby radiation pattern
optimisation. As we shall see, these antenna forms can be categorized as frequency
selective surfaces and are sometimes referred to as meta-structures.
In the previous Chapter (Sects. 7.2.1 and 7.2.2) we have already discovered that
for linear slot arrays implemented in conventional rectangular waveguide a beam
squinting propensity is exhibited with changing frequency. This is usually con-
sidered to be problematic, particularly in bandwidth terms, and certainly not
helpful. Nevertheless, a linear slot array which offers main beam positioning or
scanning that is highly sensitive to the antenna operating frequency can have very
definite and positive applications.

8.2 Frequency Scanned Slotted Waveguide Array

The evolution of the frequency scanned antenna was arguably propelled by the
desire to achieve range, height and azimuth information on a target from a single
radar platform. Unlike the active phased array antenna which can achieve this role
readily, the frequency scanned antenna option, by being wholly passive, is very
reliable and cost effective [1]. It is well known that slotted waveguide linear arrays
are troubled by beam squinting if operated over a significant bandwidth.

© Springer Nature Switzerland AG 2019 179


A. J. Sangster, Compact Slot Array Antennas for Wireless Communications, Signals
and Communication Technology, https://doi.org/10.1007/978-3-030-01753-8_8
180 8 Frequency Scanned and Leaky-Wave Linear Slot Arrays

The essence of the problem is illustrated in Figs. 7.4 and 7.5. Normally radar
engineers would wish to minimise this difficulty, but in the frequency scanned
antenna the beam squinting is purposefully enhanced. This can be done in several
ways, given that scanning of the primary beam for a slot array in air filled
waveguide is governed by Eq. (7.11). That is:

b10 d  2mp
sin / ¼ ð8:1Þ
k0 d

Note that staggering (see Chap. 7) becomes irrelevant in waveguides modified


for scanning.

8.2.1 Dielectric Loaded Waveguide

Enhanced scanning is most readily and obviously procured by introducing a low


loss dielectric of relative permittivity er into the waveguide to stretch the electrical
length of the modal path between slots. Equation (8.1) becomes:
pffiffiffiffi
er b10 d  2mp
sin /  ð8:2Þ
k0 d

for operation well above cut-off. Obviously the larger we make the relative per-
mittivity of the waveguide filling, the greater is the influence of b10d in Eq. (8.2)
thus augmenting the influence of frequency on /. Unfortunately, at typical fre-
quencies for long range radar operating in the L, S and C bands, this solution results
in feed structures which are unacceptably heavy and inefficient, when conventional
dielectric materials are employed.

8.2.2 Sinuous Waveguide Feed

The logical alternative to raising the relative permittivity of the guide filling to boost
the phase shift term in Eq. (8.1), is to simply increase the interior distance between
radiating apertures to lg (>d) so that Eq. (9.1) becomes:

b10 lg  2mp
sin / ¼ ð8:3Þ
k0 d

To achieve this operating mode it becomes necessary to physically fold the


waveguide into a sinuous or serpentine shape [1, 2]. The basic arrangement is
shown schematically in Fig. 8.1.
8.2 Frequency Scanned Slotted Waveguide Array 181

Fig. 8.1 Schematic demonstrating frequency scanning using a sinuous waveguide feed

The sinuous structure, sometimes referred to as a serpentine feed, is usually


formed from conventional air filled rectangular waveguide. Fabrication can be
made simpler if low height waveguide is employed but this means sacrificing power
handling. The feed is excited at one end by a frequency agile source and is ter-
minated in a wideband matched load. The radiating slots are located in the
waveguide wall forming the outer curved surfaces, and are separated by distance
d externally, and by distance lg internally. The main beam direction is denoted as
angle / from broadside. For a broadside beam (/ = 0) the internal slot spacing lg
must be equal to mkg10 where m is an integer (m = 1, 2,…, M). Consequently b10lg
= 2 mp and the right hand side of Eq. (8.3) becomes zero as required. At the
frequencies where this condition holds we replace kg10 by kg10m.
Consequently, for a primary beam (m = 1), the angle / in Eq. (8.3) can be
written:
 
lg k k
sin / ¼  ð8:4Þ
d kg10 kg10m

which is generally termed the frequency scan equation. To scan the primary beam
over all of ‘real’ space from −90° to +90° then we must have −1 < sin / < 1.
Clearly this range can be achieved with low frequency deviation if lg  d. Note
that d and lg can be adjusted independently thus permitting separate optimisation of
scan sensitivity and of grating lobe suppression.
For grating lobes (see circle diagrams in Chap. 6) at angles /n and /n+1
(say) Eq. (8.3) can be manipulated into the following grating lobe equation:

k0
sin /n  sin /n þ 1 ¼ ð8:5Þ
d
182 8 Frequency Scanned and Leaky-Wave Linear Slot Arrays

Fig. 8.2 Double curvature


parabolic reflector antenna
illuminated by a sinuous feed.
Courtesy of NRPL Group:
www.radartutorial.com

Clearly this equation is independent of lg which implies that the grating lobes of
the linear sinuous array antenna simply obey the strictures for linear arrays as
presented in Chap. 6.
In practice, frequency scanned array antennas based on the sinuous feed seldom
exist as stand-alone linear arrays. It is much more common to see the sinuous feed
combined with a parabolic reflector as shown in Fig. 8.2, or as an end feed to a
planar array of parallel slotted waveguides (see Chap. 7) as presented in Fig. 8.3.
These antennas provide 3-D scanning capability with azimuth information devel-
oped from the rotation of the structure as a whole while height information is
generated from the frequency scanned feed. Early versions were by no means
compact.

8.2.3 Periodically Loaded Waveguide Feed

As we have seen in the preceding sections the key to enhanced frequency scanning
in a waveguide feed is the b10d term in Eq. (8.1) which needs to be much larger
than the kod term. Enhancement of b10 is possible by dielectric loading  as
demonstrated in Sect. 8.2.1, while by differentiating the internal distance d ! lg
between apertures from the external spacing d (see Sect. 8.2.2) it is possible with
the resultant sinuous structure to improve scanning sensitivity by making lg  d.
Unfortunately dielectric loading presents considerable antenna weight and poor
electrical efficiency, while serpentine feeds are bulky, incur aperture blockage
difficulties in reflector antennas (see Fig. 8.2), and the long sinuous waveguide
structure attracts power loss issues.
8.2 Frequency Scanned Slotted Waveguide Array 183

Fig. 8.3 Planar array of


slotted waveguide radiators
end-excited by a sinuous feed
(on the left of the planar array
structure) (see [1])

A third possible solution is to enhance b10 relative to ko by adopting periodically


loaded waveguide, which has applications in many other areas of microwave
technology. For example, iris loaded circular waveguide is widely used in linear
accelerators [3–5], and in power travelling wave tubes [3, 4], while rectangular
waveguide loaded with capacitive irises has traditionally formed the basis of high
power, high frequency, bandpass filters [6–8]. More recently periodic loading of
flared waveguide has become a common feature of low noise feeds for satellite
systems [9]. The advantages of the ‘slow-wave’ feed waveguide, over the ser-
pentine feed, for frequency scanning radar are mainly mechanical. The former is
considerably smaller, and hence much less heavy for equivalent electrical perfor-
mance. The role of frequency scanning feeds in 3-D radar are examined in detail in
references [10, 11].

8.3 Transition from Uniform Waveguide


to Slow-Waveguide Feed

The feed structure for a slotted waveguide linear array possessing enhanced scan-
ning ability usually comprises a fin loaded rectangular waveguide which ‘slows’ the
propagating mode relative to empty waveguide. Hence the term ‘slow-wave’ feed.
184 8 Frequency Scanned and Leaky-Wave Linear Slot Arrays

So, in order to progress, we need to be able to assess the influence of the period-
ically spaced metal fins or irises on the waveguide propagation characteristics. We
will attempt to do this in a manner which is physically intuitive. The literature on
the topic of slow-wave structures is very extensive and the early contributions, in
the 1950s and 1960s, are highly mathematical. So the subject can be somewhat
impenetrable without deep study.

8.3.1 Fast Waveguide Frequency Characteristics

For the original empty, perfectly conducting, rectangular waveguide of width a, and
height b, we discovered in Chap. 1 that the waveguide supports a dominant TE10
mode, which cuts off at a frequency given by koa = p, that is fc = c/2a Hz. This is
depicted in Fig. 7.14a (generally referred to as the x–b diagram) where the modal
frequency characteristics are presented diagrammatically, with the parabolic shape of
the curve representing the characteristic equation for the TE10 mode (see Appendix B):
p2
ðb10 aÞ2 ¼ ðk0 aÞ2  ð8:6Þ
a

Actually, in general the x–b diagram (Brillouin diagram) for empty rectangular
waveguide can exhibit many higher order solutions, with cut-offs reaching
increasingly higher values on the koa axis, representing all of the possible TEmn and
TMmn modes as encapsulated by Eq. (A2.13) in Appendix B.
Interestingly, when the rectangular waveguide is modified by the insertion of a
periodic array of metal fins (period = p, height = h) as suggested in Fig. 8.4, the
structure can still support an electromagnetic wave [6]. Not surprisingly its prop-
agation characteristics change significantly to accommodate the influence of the
fins. In short any valid modal solution for the periodic waveguide must satisfy the
electromagnetic boundary conditions on E and H over the fin surfaces as well as
over the waveguide inner surfaces, as encountered in the empty waveguide case.
Note, that if the TE10 mode waveguide contained only one metal iris or fin,
reflection and transmission of this mode at the iris would occur. However, while
these modal constituents would continue to satisfy the electromagnetic boundary
conditions on the waveguide walls they can only partially satisfy the boundary
conditions in the vicinity of the fin. Complete satisfaction of the boundary condi-
tions for the structure as a whole requires a suitable combination of higher order TE
and TM modes [6, 7]. These modes will be evanescent (see Chap. 1) and contribute
to capacitive stored energy fields in the immediate vicinity of the iris [4]. Now,
when a multiplicity of fins in a periodic array are present in the waveguide, as
hinted at in Chap. 7, the boundary requirements at each fin does not change and
multiple TE and TM waveguide modes will continue to be necessary to meet them.
However, the fact that the fins are actually closely spaced, in a periodic manner, sets
a concomitant requirement of periodicity on those modal fields which are in close
proximity to the surface (see Fig. 8.5) of the fin-loaded structure.
8.3 Transition from Uniform Waveguide to Slow-Waveguide Feed 185

Fig. 8.4 Frequency/phase diagrams (x–b diagrams) for a uniform rectangular waveguide—TE10
mode = red curve: TE20 mode = mauve curve

Fig. 8.5 Fin or iris loaded rectangular waveguide

8.3.2 Floquet Harmonics

Mathematically the condition of periodicity is expressed through Floquet’s


Theorem. This theorem can be succinctly stated as: A steady-state modal solution to
the electromagnetic boundary value problem, which accommodates a periodic
186 8 Frequency Scanned and Leaky-Wave Linear Slot Arrays

structure, has the property that the fields in adjacent cells, distance p apart, are
related by a complex constant. For a lossless structure, it can be expressed in the
form:

E0 ðx; y; z þ pÞ ¼ E0 ðx; y; zÞ expðjb0 pÞ ð8:7Þ

E0(x, y, z) represents a possible electromagnetic wave solution in the zero’th


mode. In an infinitely long, closed and lossless, structure the solution is inevitably
multi-modal. It is also a periodic function of z, the propagation direction in the
waveguide, while bo is the phase constant for the fundamental space harmonic of
the zero’th mode. That E0(x, y, z) is a solution of the Maxwell equations which are
applicable to the slow waveguide structure, we firstly represent it as follows:

E0 ðx; y; zÞ ¼ f ðx; y; zÞ expðjb0 zÞ ð8:8Þ

where f(x, y, z) is also a periodic function of z with period p. It follows therefore,


referring to Fig. 8.5, that at location z + p we must have:

E0 ðx; y; z þ pÞ ¼ f ðx; y; z þ pÞ expðjb0 ðz þ pÞÞ ð8:9Þ

But, since f ðx; y; z þ pÞ  f ðx; y; zÞ for a periodic function, then Eq. (8.9)
becomes:

E0 ðx; y; z þ pÞ ¼ f ðx; y; zÞ expðjb0 zÞ expðjb0 pÞ ð8:10Þ

Hence, on combining Eqs. (8.10) and (8.8) we observe that the Floquet rela-
tionship [Eq. (8.7)] is confirmed for the electromagnetic solution of a closed
periodic structure.
That the Floquet formulation is consistent with the Maxwell equations can be
deduced by considering a parallel problem in signal processing theory, namely the
representation of square or digital waveforms in the time domain by electrically
more convenient sinusoidal waveforms of varying frequencies. The mechanism for
generating this relationship is provided by Fourier series. The theory simply states
that any periodic function that is finite, single-valued and continuous can be rep-
resented by a series of trigonometric functions. Good pictorial representations of the
process are to be found on the internet [12]. Consequently, the periodic field
function E0(x, y, z) can by expanded into a Fourier series in the space domain
z (rather than time t as in signal processing) such that:

X
1  
j2prz
E0 ðx; y; zÞ ¼ Er ðx; yÞ exp  expðjb0 zÞ
r¼1
p
ð8:11Þ
X1
¼ En ðx; yÞ expðjbr zÞ
r¼1
8.3 Transition from Uniform Waveguide to Slow-Waveguide Feed 187

Fourier theory dictates that the amplitudes of the r space harmonics Er(x, y) of
the zero’th mode are given by,

Zp  
1 j2prz
Er ðx; yÞ ¼ E0 ðx; y; zÞ exp dz ð8:12Þ
p p
0

and

2pr
br ¼ b0 þ ð8:13Þ
p

is the phase coefficient for the rth space harmonic ð1\r\1Þ. Note that in the
context of a fin loaded structure based on rectangular waveguide, as depicted in
Fig. 8.5, the zero’th mode is essentially an evolution of the TE10 mode of the empty
waveguide.
For higher order modes Emn(x, y) (m = 1, 2, 3,…, n = 0, 1, 2,…) the corre-
sponding possible solution forms are

X
1  
j2prz
Em:n ðx; y; zÞ ¼ Er ðx; yÞ exp  expðjb0 zÞ
r¼1
p
ð8:14Þ
X1
¼ Er ðx; yÞ expðjbr zÞ
r¼1

where b0 becomes the fundamental phase coefficient for the mnth mode, and the Er
represent the relevant harmonic magnitudes for this mode.
The question then remains as to whether or not the field forms represented by
Eqs. (8.13) and (8.14) satisfy the electromagnetic wave equation as they must? For
the electric field in z-directed waveguide this has the form (see Appendix B):

r 2 E  c2 E ¼ 0 ð8:15Þ

Consequently the postulated periodic structure solution E0(x, y, z) must satisfy:

r2 E0 ðx; y; zÞ  c2 E0 ðx; y; zÞ ¼ 0 ð8:16Þ

A similar requirement applies to Em,n (x, y, z). Hence combining Eq. (8.16) with
Eq. (8.11) we obtain:

X
1 X
1
r2 Er ðx; yÞ expðjbr zÞ  c2 Er ðx; yÞ expðjbr zÞ ¼ 0 ð8:17Þ
r¼1 r¼1

which, given the linearity of the wave equation, leads to:


188 8 Frequency Scanned and Leaky-Wave Linear Slot Arrays

X
1

r2 Er ðx; yÞ expðjbr zÞ  c2 Er ðx; yÞ expðjbr zÞ ¼ 0 ð8:18Þ


r¼1

Therefore, the postulated field solution E0(x, y, z) for the periodic structure will
be a solution to the wave equation, and hence Maxwell’s equations, if each space
harmonic is also a solution. This will normally be the case since the space har-
monics are generated by solving Maxwell’s equations for individual periodic cells.
The cells are generally represented by resonant structures exhibiting much simpler
boundary conditions.
From an engineering perspective, the seemingly intractable complexity of
potential solutions to the electromagnetic boundary value problem, as presented by
a periodic structure, can be ameliorated significantly by noting that the structure is
composed of a linear assembly of identical cells (Fig. 8.5). Actually, the resonant
frequencies of these cells can be determined with little difficulty on an electro-
magnetic wave simulator such as HFSS [13] or COMSOL [14], and once known
the formation of the phase (x–b) diagram for the structure as a whole becomes
possible. Furthermore, when the cells are at resonance, only two logical possibilities
exist for neighbouring units. The fields either assume a symmetrical mode distri-
bution with zero phase shift (bp = 0) from cell-to-cell, or they assume an
anti-symmetrical form with p-phase shift (bp = p) from cell-to-cell. This behaviour
can be observed in low frequency LCR filter circuits [5, 15]. In filter terms the first
option represents a zero phase shift frequency cut-off condition while the second
equates to a cut-off condition exhibiting p-phase shift.

8.3.3 Slow Waveguide Frequency Characteristics

For the fin-loaded rectangular waveguide the lowest possible operating frequency is
determined by the fundamental (TE10) mode cut-off in the empty waveguide, given
by Eq. (8.6) with b10a = 0. At this frequency kg10 ) 1 which means that the gaps
between the fins have effectively zero depth electrically and the modal fields are
restricted to the empty region above the fins. That is, the periodic structure behaves
electromagnetically like an empty rectangular waveguide of height b and replicates
its frequency behaviour at cut-off (see Figs. 8.4 and 8.6) at bp = 0 and ba = 0
respectively). In cell modelling terms this resonance is generated when magnetic
walls (H-walls) are located above neighbouring irises, as suggested in Fig. 8.7. The
H-walls reflect the cell-to-cell zero phase shift requirement at this cut-off. The finite
element simulation predicts, not unexpectedly, 6.55 GHz (koa = p) in accordance
with Fig. 8.4 and with Eq. (8.6).
When an EM simulation is performed on the cell model with E-walls (Fig. 8.7b)
the principal bp = p frequency is predicted to be 9.87 GHz (koa = 4.726) with a full
half cycle of TE10 mode field pattern formed between the base of the fin-gap and the
top wall of the periodic structure. It is plotted on Fig. 8.6 as the first cut-off
8.3 Transition from Uniform Waveguide to Slow-Waveguide Feed 189

Fig. 8.6 Frequency/phase diagram (x–b diagram) for rectangular waveguide periodically loaded
with metal fins—showing fundamental plus first higher space harmonic (NB slow-wave solutions
to right of v = c line: fast-wave solutions occur in region to the left of v = c line—note in an open
structure (see Chap. 10) waves in the fast region become leaky)

maximum. Needless to say given the nature of the Floquet theorem, the bp = 0
frequency is repeated at bp ¼ 2p; 4p; . . .1, while the bp = p frequency is
repeated at bp ¼ 3p; 5p; . . .1.
The precise shape of the curve in Fig. 8.6 between the locations of key cut-off
frequencies, which have just been identified, requires the solution of the Maxwell
equations for the periodic structure as a whole. An investigator possessing elec-
tromagnetic simulation software such as HFSS [13] or COMSOL [14] could
attempt a direct application of a finite element, or finite difference technique,
although this route would demand considerable computational power to accom-
modate a model of the complexity presented by a periodically loaded waveguide.
The computationally more efficient alternative would to set up a problem specific
solution based on the moment method as outlined in Chap. 5, or perhaps on a
variational method as advocated by Cairo and Kahan [16] or an alternative MoM
[17]. Software such as MATLAB [18] can greatly facilitate the evolution of a
suitable algorithm to achieve this, particularly in the case of the moment method.
190 8 Frequency Scanned and Leaky-Wave Linear Slot Arrays

Fig. 8.7 Periodic cell models showing E-wall and H-wall inserts. a H-wall, b E-wall

8.3.4 Approximate Solution for Phase Coefficient

As we observed in Chap. 5 the moment method furnishes us with a mathematical


procedure for solving electromagnetic boundary value problems generally com-
prising two coupled regions which are themselves fully described electromagneti-
cally. In the slot radiator example the ‘known’ regions are empty rectangular
waveguide and a half-space coupled through the ‘uncharacterised’ radiating aper-
ture. Here, the coupled regions are the gaps in the periodic fin structure which
behave electromagnetically like shorted low height waveguides, and the rectangular
waveguide space located above the fins (see Fig. 8.5). The two regions are coupled
through the ‘apertures’ representing the interface between the air gaps, between y =
b and y = b + h (linked mathematically by means of the Floquet theorem), and the
empty waveguide region. The unknown in the analysis is the axial propagation
coefficient imposed by the periodicity. Needless to say, the literature presents a
multitude of variational method, or moment method, solutions for this regular
structure, and for a wide range of related structures [3, 6, 7, 11].
Alternatively, a sense of the nature of the solution to electromagnetic propaga-
tion in a periodically loaded waveguide can be achieved by following a heuristic
methodology, as displayed in Chap. 2. There the approach aimed at formulating
potential functions relevant to electromagnetic radiation [Eqs. (2.45), and (2.54)].
8.3 Transition from Uniform Waveguide to Slow-Waveguide Feed 191

Here we wish to generate potential functions which are solutions of the electro-
magnetic wave equations within a fin loaded periodic waveguide (see Fig. 8.5). It is
evident that Fig. 8.5 represents a rectangular waveguide, as discussed in Chap. 1,
except that the lower broadwall presents an impedance other than zero to the
electromagnetic waves. Consequently it as appropriate to evaluate waveguide
propagation in this modified set-up.
In Chap. 1, it is demonstrated that the fundamental frequency characteristics of
rectangular waveguide are contained in the propagation equation [Eq. (1.68)],
which is also presented graphically in Fig. 1.6. For air filled waveguide with per-
fectly conducting walls this equation is:
mp2 np2
b2wmn ¼ k02  kcmn
2
¼ k02   ð8:19Þ
a b

for both TE and TM modes. If these are propagating solutions it is necessary that
the propagation coefficient for the mnth mode in the waveguide, bwmn is real, in
which case:
k0 [ kcmn and bmn \k0

Since bwmn ¼ x=vpmn and k0 ¼ x=c then clearly vpmn [ c. That is, the
solutions are ‘fast’ with the phase velocity of the mnth mode travelling at a speed
greater than that of light axially down the waveguide. So what conditions would be
required for the propagating modes to be ‘slow’ with vpm \c? Mathematically it
would require:
 2 mp2
b2m ¼ k02 þ kcm
2
¼ k02  kcy  ð8:20Þ
a

‘Slow’ solutions seem intuitively more comprehensible in a world of physics


which is used to the idea that nothing travels faster than light. The question,
therefore, appears to be a not unnatural one. Note that since the x-directed geometry
of the waveguide is unchanged by the fins (Fig. 8.5), the third term on the right of
the equals sign in Eq. (8.20) must also remain unchanged. In the y-direction this is
certainly not true, and therefore we introduce a kcy parameter (to be determined) to
reflect this. Hence we can write for the mth mode:
mp2
2
kcy ¼ k02  b2m  ð8:21Þ
a

For slow-wave solutions we have already stipulated that bm > k0 and given that
all the terms on the right side of Eq. (8.21) are real, kcy must be imaginary (say jam).
This implies that in the fin loaded waveguide, in the space (0 < y < b) above the
fins, valid fields solutions must display exponential decay in the −y direction.
Embracing the evident analogy between the fin loaded rectangular waveguide
and its empty counterpart (Chap. 1) the following potential functions in phasor
form can be constructed for TE and TM mode solutions. For the mth such mode:
192 8 Frequency Scanned and Leaky-Wave Linear Slot Arrays

rX
¼1
1 mp
zm ¼
ATE 2
½Ar sinhðamr yÞ þ Br coshðamr yÞ sin exp jbmr z
k
r¼1 cu
a
rX
¼1
ð8:22Þ
1 mp
ATM
zm ¼ 2
½Cr sinhðamr yÞ þ Dr coshðamr yÞ cos exp jbmr z
k
r¼1 cu
a

summed over an infinite number of Floquet space harmonics. Note that for higher
 2 2
order space harmonics a2cmr ¼ b2mr þ mp a k0 . Also kcu2
¼ k02 þ a2cmr and
bmr ¼ bm0 þ 2rp p , while p = G + t.
On applying the Maxwell equations the following field expressions are gener-
ated for the mth mode in the rectangular space (a  b) above the fins (denoted by
subscript ‘u’):

X1
1 n mp 
Eux ¼  j b A mr þ jxl a cmr Dmr sinh acmr y
k2
r¼1 cu
a mr 0
 mp  o mpx
þ j bmr Bmr þ jxl0 acmr Cmr cosh acmr y  cos exp jbmr z
a a
ð8:23Þ

X1
1 n mp 
Euy ¼  ja cmr b mr A mr þ jxl 0 Dmr cosh acmr y
k2
r¼1 cu
a ð8:24Þ
 mp  o mpx
þ jacmr bmr Bmr  jxl0 Cmr sinh acmr y  sin exp jbmr z
a a
X
1
mpx
Euz ¼ ðAmr sinh acmr y þ Bmr cosh acmr yÞ sin exp jbmr z ð8:25Þ
r¼1
a

X1
1 n mp 
Hux ¼  jxe 0 a cmr Amr þ j b Dmr cosh acnr y
k2
r¼1 cu
a mr ð8:26Þ
 mp  o mpx
þ jxe0 acmr Bmr þ j b Cmr sinh acmr y  sin exp jbmr z
a mr a
X1
1 n mp 
Huy ¼  2
jxe0 Amr þ jbmr acmr Dmr sinh acmr y
k
r¼1 cu
a
 mp  o mpx
þ jxe0 Bmr þ jbmr acmr Cmr cosh acmr y  cos exp jbmr z
a a
ð8:27Þ

X
1
mpx
Huz ¼ ðCmr sinh acmr y þ Dmr cosh acmr yÞ cos exp jbmr z ð8:28Þ
r¼1
a
8.3 Transition from Uniform Waveguide to Slow-Waveguide Feed 193

In the fin loaded region (b < y < h + b) electromagnetic TE and TM modes also


exist in the gaps between the fins. This region will be denoted by subscript ‘f’
representing the ‘fin’ region. These gaps can sensibly be treated as short lengths
(h) of low height rectangular waveguide of cross-sectional area G  a, terminated
in electrical short circuits. For a periodic structure with i fins and gaps (0 < i < I),
the z-position of the ith gap zi is given by zi = z − (i − 1)(G + t). Given that the
mth mode in any prescribed gap displays the same x-functionality as the modes in
the space above the fins, then for the mth mode we can write:

X1
1 mp np  mpx npzi
Efx ¼  2
K mn þ jxl 0 bgmn L mn cos sin sin bgmn ðy  b  hÞ
k
n¼0 cf
a G a G
ð8:29Þ

where it is presumed that the Floquet periodicity is built into the constants Kmn and
Lmn, for the mnth gap mode. This ensures that the gap modes are phase shifted from
one gap to the next in sympathy with the eventual mnth modal solution (bm0) for the
periodic structure as a whole. Thus Kmn (or Lmn) at the ith slot (say) is related to Kmn
in the first slot by,

ðiÞ
Kmn ¼ Kmn expðjibm0 ðG þ tÞÞ:

Also

X1
1 np mp  mpx npzi
Efy ¼  2
b gmn K mn þ jxl0 Lmn sin sin cos bgmn ðy  b  hÞ
k
n¼0 cf
G a a G
ð8:30Þ

X
1
mpx npzi
Efz ¼  Kmn sin cos sin bgmn ðy  b  hÞ ð8:31Þ
n¼0
a G

X1
1  mp np  mpx npzi
Hfx ¼ 2
jxe b K
0 gmn mn  Lmn sin cos cos bgmn ðy  b  hÞ
k
n¼0 cf
a G a G
ð8:32Þ

X1
1  mp np  mpx npzi
Hfy ¼ 2
jxe 0 Kmn þ b gmn Lmn cos cos sin bgmn ðy  b  hÞ
k
n¼0 cf
a G a G
ð8:33Þ

X
1
mpx npzi
Hfz ¼ Lmn cos sin cos bgmn ðy  b  hÞ ð8:34Þ
n¼0
a G
194 8 Frequency Scanned and Leaky-Wave Linear Slot Arrays

The phase coefficient bgmn applies to the low height gap-waveguide and is given
by the following equation (see Chap. 1):
mp2 np2
b2gmn ¼ k02   ð8:35Þ
a G

and
np 2
kcf2 ¼ k02  ð Þ ð8:36Þ
G

Note that for higher order modes in the gaps, the right-hand side of Eq. (8.35)
can become negative, in which case the bgmn terms in Eqs. (8.29) to (8.34) are
replaced by jagmn, when appropriate in the ensuing formulation. Also for n = 0,
bgm0 = bwm0.

8.3.5 Boundary Condition at Periodic Interface

The above partial field solutions, have been constructed by employing conventional
waveguide analogies in the upper space (u) and the fin gap (f), by applying the
electromagnetic field boundary conditions on the side walls at x = 0 and x = a, and
on the fin surfaces within the gaps, and at the bottom of the gaps (y = b + h). Thus
all are automatically satisfied. We are therefore left to ensure the boundary con-
dition Etan = 0 is satisfied at y = 0, that is:
(1) Ex = 0 and Ez = 0 at y = 0
Furthermore the boundary conditions at the periodic surface at y = b must be
observed. This implies that:
(2) Etan and Htan are continuous across the y = b interface. It is sufficient to take
care of the tangential field components only—since the Maxwell equations
ensure that the normal components fall into line.
The application of condition (1) immediately requires that:

Bmr ¼ Cmr ¼ 0 ð8:37Þ

Condition (2) requires that Eux ¼ Efx and Euz ¼ Efz at y = b in the range
0 < z < G + t. This is automatically the case at the fin tips (G < z < G + t) where
Eux ¼ Efx ¼ Euz ¼ Efz ¼ 0.
Continuity of Ez is enforced mathematically by performing an averaging process
over one period of the periodic surface. That is:
8.3 Transition from Uniform Waveguide to Slow-Waveguide Feed 195

Z
Gþt ZG
Euz expðjbmr zÞ dz ¼ Efz expðjbmn zÞ dz ð8:38Þ
0 0

On substituting Eqs. (8.25) and (8.31) into Eq. (8.38) and setting y = b yields:

ZG X
1
npz
ðG þ tÞAm0 sinh acm0 b ¼  Kmn sin bgmn h cos expðjbgmn zÞ dz ð8:39Þ
n¼0
G
0

Note that on the left of the above equation the space harmonic summation
disappears since the higher harmonics integrate to zero over the period p of the
corrugated surface. Am0 represents the fundamental term in the space harmonic
series for the mth slow-waveguide mode. This mode is neither TE nor TM. It is
generally referred to as the HE11 hybrid mode. Equation (8.39) can be reconfigured
to give:

X1  
G jbm0 G
Am0 ¼ Kmn bm0 sinðbgmn hÞ exp Uðbm0 nÞ ð8:40Þ
ðG þ tÞ sinh acmr b n¼0 2

Similarly enforcing continuity of Ex at the y = b boundary gives:

ðG þ tÞ h mp i
2
jbm0 Am0 þ jxl0 acm0 Dm0 sinh acm0 b
kcu G a
X1
np=Ghmp np i jb G
¼ 2
Kmn þ jxl0 bgmn Lmn sinðbgmn hÞ expð m0 ÞUðbm0 nÞ
n¼0
k cf a G 2
ð8:41Þ

Continuity of Hz at y = b generates:

X1 np  
jbmr G n Lmn
j Dmr coshðacmr bÞ exp  Vðbmr nÞ ¼ n cosðbgmn hÞ ð8:42Þ
n¼1
G 2 2
196 8 Frequency Scanned and Leaky-Wave Linear Slot Arrays

Finally, continuity of Hx at y = b gives:

X1
bmr h mp i b G
2
jxe 0 a cmr Amr þ j bmr D mr coshðacmr bÞ expðj mr ÞVðbmr nÞ
k
r¼1 cu
a 2
n h mp np i
¼ n2 jxe0 bgmn Kmn  Lmn cosðbgmn hÞ
2kcf a G
ð8:43Þ

The following definitions apply to the above:


8
bmr < sinww m even
Uðbmr nÞ ¼  
b2mr  ðnpG Þ2 : þ w
jcosw ð8:44Þ
m odd

8
bmr < sinww m even
Vðbmr nÞ ¼  
bmr  ðnpG Þ2 :  w
2 jcosw ð8:45Þ
m odd


2 ðn ¼ 0Þ
where wr ¼ bmr2 G and nn ¼ .
1 ðn [ 0Þ

8.3.6 Determinantal Equation

Relatively routine algebraic manipulation of Eqs. (8.40), (8.41), 8.42) and (8.43),
leads to a pair of simultaneous equations for Kn and Ln (for the mth mode), which
can be summarised as outlined below. Note that in terms involving n, which occur
outside of the summations over n, it is convenient to replace the integer n with
integer s.

X
1
fWnrs Kns þ Xnrs Lns g ¼ 0
r¼1
ð8:46Þ
X1
fYnrs Kns þ Znrs Lns g ¼ 0
r¼1
8.3 Transition from Uniform Waveguide to Slow-Waveguide Feed 197

where
" #
coth acmr b X1 mpnp np2 k 2
Wnrs ¼ Vðbmr sÞ a G
b sin bgmn h 
2 cu
Uðbmr nÞ
acmr n¼0
jxl0 mr G kcf2
  1 
Gþt cos bgms h coth acmr b X np2 k2
Xnrs ¼ dns ns þ Vðbmr sÞ bgmn cu Uðbmr nÞ
G 2 acmr n¼0
G kcf2
 
Gþt cos bgms h
Ynrs ¼ dns ns jxe0 bgms
G 2kcf2
" #
X1
ðmpa Þ2 ðnpG Þ2
2 coth acmr b jxe0 2
þ bmr Vðbmr sÞ bw10 sin bgmn h  Uðbmr nÞ
acmr n¼0
k02 kcf2
 
Gþt mp np cos bgms h
Znrs ¼ dns ns
G a G 2kcf2
coth acmr b X1 mpnp
j a G bmr bgmn
þ b2m0 Vðbmr sÞ Uðbmr nÞ
acmr n¼0
kcf2
ð8:47Þ

The Kronecker delta dns is introduced to ensure that modal orthogonality is


maintained (dns ¼ 1; when n ¼ s; otherwise it equates to zero).
There is a non-trivial solution to Eqs. (8.46) only if the determinant of the matrix
coefficients equates to zero. That is, if,

Wnrs Xnrs
¼0 ð8:48Þ
Ynrs Znrs

The accuracy of any solution for the propagation behaviour of periodic waveg-
uide, based on Eq. (8.48), is dependent on the number of terms in the infinite series
that is retained in computations. Obviously there is a limit. Infinite series of similar
character are reported in Ref. [19], which provides useful pointers to possible sum-
mation techniques based on the calculus of residues. However, in the present context
we can test the validity of the heuristically deduced solution by noting that only the
first passband for the periodic structure with a single field variation in the x-direction
(m = 1) is significant. In this scenario drastic approximation can legitimately be made.
In Eq. (8.48) the only term of any significance is Ynrs, since this is the only term which
does not equate to zero when n = 0. Consequently, Eqs. (8.48) can justifiably be
approximated by setting YnrsKns = 0, to produce the following identity. It is pertinent
to note that the W and Z terms in Eq. (8.48) emanate from the equations for Ex and Hz
respectively. Thus equating them to zero to form (8.49) implies that Ex = 0 and Hz = 0
at y = b. In conjunction with the boundary conditions for E the implication is that Ex =
0 everywhere inside the fin-loaded guide. While this is approximately in accord with
reality for the fundamental mode, it is unlikely to remain valid for most high order
modes in higher passbands.
198 8 Frequency Scanned and Leaky-Wave Linear Slot Arrays

8 bgms cos bgms h


9
1 > >
Gþt
X < dns G ns 2kcf2
Kms =
0¼ 
> þ b2 coth acmr b Vðb sÞ Km0 b2 w10 sinðb hÞUðb 0Þ þ P
1 K b2
Uðbmr nÞ >
2
r¼1 : ;
mn gmn
m0 acmr mr k w10 mr k2
cf n¼0 cf

ð8:49Þ

Equation (8.49) has been programmed for three space harmonics above the
corrugated surface and for three waveguide modes in the inter-fin gaps below the
corrugated surface. The results for the phase coefficient (b11) of the fundamental
mode HE11 in the guide, as a function of frequency, are presented in Fig. 8.8. Two
fin-loaded waveguide examples are examined, and when compared with mea-
surements, performed carefully on precisely machined prototypes, the level of
agreement between calculation and experiment is clearly extremely good. For the
fundamental mode it is fair to conclude that the heuristically deduced solution,
despite the evident approximations involved, is capable of providing a reliable
source of design information for fin loaded slow-waveguide structures. Of course
today, structures of this type can be comprehensively and accurately modelled by
resorting to full-wave electromagnetic solvers [12–14]. Nevertheless such solvers
have to be set up properly—a process which is open to errors—and the availability
of accessible equations such as Eq. (8.49) can provide a very useful check on the
modelling process.

Fig. 8.8 Predicted and measured phase coefficients, as a function of frequency, for a fin-loaded
rectangular waveguide a a = 72.14 mm, b = 12.24 mm, h = 21.8 mm, G = 4.76 mm,
t = 1.58 mm, b a = 64.5 mm, b = 12.24 mm, h = 26.5 mm, G = 11.11 mm, t = 1.58 mm
(curves = theory, points = measured)
8.4 Slot Radiators in a Periodic Waveguide Feed 199

8.4 Slot Radiators in a Periodic Waveguide Feed

The design of a scanned linear array antenna based on fin-loaded waveguide fed slot
radiators requires knowledge of the radiation properties of such slots, which may be
transversely or longitudinally directed, or inclined to the guide axis. Here we will
limit our attention to the transverse slot radiator, as shown schematically in Fig. 8.9,
which is fully representative of the computational process for slots in general. The
application of a moment method to this problem leads, as we have seen for the
conventional waveguide fed slot in Chap. 5, to a matrix equation of the form:
  
½ A ½C ½ a ½ h
¼ ð8:50Þ
½D ½B ½ a0  ½ 0

The same equation applies to the periodic waveguide example in Fig. 8.9 but the
matrix entries on the left of the equation require slight modification to reflect the
waveguide change. The entries are:
8 9
ZZ < ZZ =
 
Ais ¼ jxe0 fi ^ag n^ Gma ðr=r0 Þ þ Gmb ðr=r0 Þ fs ^
an dS0 dS ð8:51aÞ
: ;
S S
8 9
ZZ < ZZ =
  0
Bis ¼ jxe0 fi0 ^ag ^0 
n Gmb ðr=r0 Þ þ Gmc ðr=r0 Þ fs ^
an dS0 dS
: ;
S0 S0
ð8:51bÞ
8 9
ZZ < ZZ =
 
Cis ¼ jxe0 fi0 ^ag ^
n Gmb ðr=r0 Þ fs ^
an dS0 dS ð8:51cÞ
: ;
S S0

Fig. 8.9 Fin-loaded feed waveguide for transverse slot radiator in upper wall
200 8 Frequency Scanned and Leaky-Wave Linear Slot Arrays

8 9
ZZ < ZZ =
  0
Dis ¼ jxe0 fi ^ag ^0 
n Gmb ðr=r0 Þ fs ^
an dS0 dS ð8:51dÞ
: ;
S0 S
ZZ
hi ¼  fi ^ag ½n
^  Hi ðrÞ dS ð8:51eÞ
S

The matrix entries Ais, Bis, Cis, Dis given mathematical expression in
Eqs. (8.51a) are essentially electromagnetic coupling terms giving mathematical
expression to the mutual coupling which occurs between the assumed current
distributions (fi,s) on the interior and external surfaces of the radiating aperture. For
the ‘slot-cavity’ and for the exterior half-space the magnetic dyadic Green’s
functions Gmb and Gmc respectively, remain unchanged from the conventional
waveguide moment method derivation in Chap. 5. On the other hand, the magnetic
dyadic Green’s function Gma for the periodic feed waveguide clearly differs from
Gma . Nevertheless it can readily be deduced from the plain waveguide form, as
elaborated in Appendix C. For the transverse slot shown in Fig. 8.9 the matrix
entries have the same meanings as for the slot radiator in conventional waveguide.
These are:
Ais represents self-coupling between magnetic current elements on the inner slot
surface S.
Bis represents self-coupling between magnetic current elements on the outer slot
surface S′.
Cis represents mutual coupling to the inner slot surface (S) due to magnetic current
elements on the outer surface (S′).
Dis represents mutual coupling to the outer slot surface (S′) due to magnetic current
elements on the inner surface (S).
Once the integrals in Eqs. (8.51a) have been evaluated, which is mathematically
routine but tedious [18, 20], the matrix Eq. (8.50) is amenable to solution by a
range of matrix manipulation tools. These can be found in standard mathematical
textbooks [21, 22] and in convenient to use computer software packages [12–14,
23]. The outcome of the process is plausible values for s, fs, a′s and as. These can
then be used to compute the scattering parameters for the compact radiator much as
outlined in Chap. 5.
Moment method computations have been performed on a fin-loaded waveguide
fabricated in WG10 (a = 72.14 mm, b = 34.04 mm) designed to provide a feed for
a frequency scanned linear array of transverse slots. These are summarised in
Fig. 8.10, where the results are compared with an equivalent set-up in conventional
waveguide (also WG10). In the figure the normalised power radiated by a centred
transverse slot is presented as a function of frequency, for four different slot lengths
of 30, 40, 42 and 46 mm, all at constant width (w = 4.5 mm). The solid (green)
curves denote the conventional waveguide case, while the dashed (blue) traces
8.4 Slot Radiators in a Periodic Waveguide Feed 201

Fig. 8.10 Relative power radiated as a function of frequency for a centred transverse slot in a
waveguide feed (solid curves) and a periodic waveguide feed (dashed curves) with slot length l as
parameter (l = 30, 40, 42, 46 mm: w = 4.5 mm) Waveguide WG10: a = 72.14 mm,
b = 34.04 mm Slow-waveguide: a = 72.14 mm, b = 12.24 mm, h = 21.8 mm, G = 4.7625 mm,
t = 1.5875 mm

represent corresponding slow-waveguide calculations. At low frequencies it is clear


that for the slow waveguide, with 21.24 mm deep fins separated by 4.76 mm, the
normalised radiated power exceeds that of the equivalent slot in unmodified
waveguide. This is largely because the height b (=12.24 mm) in the slow waveg-
uide is made low enough to counteract the decay in fundamental mode field
strength in the y-direction in the space above the periodic surface, thus enforcing
good field coupling at the slot. This is emphasised by the curves depicting radiation
from slots resonant at all frequencies (chain dotted) with the normalised power
radiated deteriorating rapidly for the slow-wave case (red chain dotted) as the
frequency extends beyond 3.15 GHz. As the frequency increases the decay phe-
nomenon is enhanced [see Eq. (8.21)] so that between 3.1 and 3.2 GHz the plain
waveguide begins to generate stronger radiation for the same feed conditions for all
slot examples. For both feed types it is noticeable that for a 46 mm slot resonant at
*3.13 GHz, approximately 50% of the incident power is radiated and the oper-
ating bandwidths are comparable. These results confirm that the slow waveguide
can provide an effective feed for a scanned linear array antenna.
202 8 Frequency Scanned and Leaky-Wave Linear Slot Arrays

8.4.1 Frequency Scanned Linear Slot Array Employing


Periodic Waveguide Feed

The fin-loaded periodic waveguide has been shown to exhibit phase shift charac-
teristics, as a function of frequency (see Fig. 8.8), which are comparable with the
phase sensitivities manifested by dielectric loaded waveguide or sinuous waveg-
uide. Furthermore slots located in the top wall opposite to the periodic surface
(Fig. 8.9) can be excited to a level not unlike slots in a conventional waveguide
feed (Fig. 8.10). Consequently, as a feed for a frequency scanned linear array
antenna [10, 11, 24] the fin-loaded waveguide structure offers the advantages of
much less weight than the dielectric loaded option and much less bulk than the
serpentine waveguide alternative. Early attempts at forming fin-loaded waveguide
feeds for a frequency scanning radar were derailed by heavy propagation losses
[24]. However, it was eventually discovered that by adopting the highest accuracy
and control of the manufacturing process the periodic feeder loss can be made
comparable with a sinuous waveguide exhibiting similar scanning capability.
A slow-waveguide feed with the dimensions detailed in Fig. 8.8 [curve (a)] was
used to create a scanned linear array comprising six slots on the broadface. The
slots were of equal size, and with relatively low radiation per slot the field distri-
bution across the aperture was close to uniform. The measured beam width and
sidelobe levels for the array were consistent with this excitation set-up. The devi-
ation of the peak of the main radiated beam has also been investigated. The pre-
dicted squint based of the theoretical phase coefficient change with frequency
shown in Fig. 8.8 [curve (a)] was obtained using Eq. (8.1). It is plotted as a solid
blue curve in Fig. 8.11, and compared with measured results (red diamonds). The
degree of agreement further confirms the reliability of the developed formula
[Eq. (8.49)] for the phase coefficient of the slow-wave feed.

8.5 Uniform Leaky-Wave Antennas

Careful and perceptive readers will have noted by now that electromagnetic
waveguides display a degree of commonality in their frequency characteristics. This
can largely be ascribed to the propagation equation which for conventional rect-
angular waveguide is given by Eq. (8.6). Standard waveguide text books [6, 7]
confirm this by demonstrating that the same equation applies to empty waveguides
of other cross-sections, such as circular, coaxial and elliptical. The only difference is
that the transverse resonance terms, (mp/a) and (np/b) in the rectangular case, must
be replaced by Bessel function roots or elliptic function roots. Consequently, it is
convenient to express Eq. (8.6) in a more general form, namely:
8.5 Uniform Leaky-Wave Antennas 203

Fig. 8.11 Beam scan as a function of frequency for a slow-wave fin-loaded waveguide feed
comprising six slots of the same size, equi-spaced by 65.4 mm (solid curves = theory, diamond
markers = measurement)

k02 ¼ kx2 þ ky2 þ kz2 ð8:52Þ

where, for the rectangular waveguide: kx ¼ mp a ; ky ¼ b ; kz ¼ bmn .


np

For slow rectangular waveguide we have observed that it is necessary for kz [ k0


and this is constructed by making ky2 negative in which case ky must be imaginary
and equal to jam (say). That such solutions are possible is demonstrated in Sect. 8.3
for fin-loaded waveguide. Partial dielectric loading also generates solutions of this
form.
Equation (8.52) points not only to trapped waves of the fast and slow variety,
but also to leaky waves in an open structure. If such a wave propagates in the axial
direction z, then we must have:

kz ¼ bm  jam ð8:53Þ

for the mth mode for which kx ¼ mp a . bm denotes the axial phase shift per unit
distance in rad/m while am denotes axial attenuation rate in nepers/m. In this case
Eq. (8.52) requires that:
204 8 Frequency Scanned and Leaky-Wave Linear Slot Arrays

Fig. 8.12 Radiation from a leaky wave structure

ky ¼ bmy þ jamy ð8:54Þ

if the space into which the electromagnetic waves leak is defined as y > 0 in
Fig. 8.12. In this direction the leaky wave field strength has the functional form:

expðamy  jbmy Þy ð8:55Þ

The amy term if positive implies, rather curiously that in the y-direction the leaky
wave fields grow without limit. The consequent singular behaviour at infinity,
means that mathematically, while it is a solution of the homogeneous field equa-
tions, the leaky wave cannot be a proper mode of the source free Maxwell equa-
tions. Nevertheless, despite this ‘improperness’ leaky waves can still be valid
representations of the fields generated by an open structure in certain restricted
regions provided a source is present [25].
The situation can perhaps be clarified by reference to Fig. 8.12 which depicts a
leaky waveguide containing a long axially directed slit (commonly in the side wall
[25]) fed from the left from a closed waveguide and an intervening matching
transition. To the right, in the z-direction, the structure is assumed to be infinite. The
slit is located at y = 0 and extends to infinity in the z-direction along with the
waveguide. From Fig. 8.12 it is readily deduced that if the wave solution within the
waveguide exhibits a dominant ‘mode’ phase coefficient b0 then the wave com-
ponent parallel to the waveguide and travelling along the outer surface must match
it. This implies that:
8.5 Uniform Leaky-Wave Antennas 205

b0 k0
/  cos1 ¼ cos1 ð8:56Þ
k0 kg0

The 3 dB beamwidth D/ can be deduced from array theory in Chap. 6


[Eq. (6.17)]. Thus for uniform excitation of the slit:

0:88k0
D/  ð8:57Þ
L sin /

In Eq. (8.56) kg0 is the wavelength of the dominant field solution in the guide.
Note that it demands that b0 \k0 . Consequently, leaky behaviour is linked to ‘fast’
waves and the radiation is launched into the forward quadrant. If the attenuation
coefficient for this solution is a0, then the field magnitude in the waveguide must
diminish as z increases, which in turn implies that the power available for radiation
decays. The resultant higher power density in the radiation fields at the input to the
slit is represented in the diagram by closely spaced rays, with the spacing growing
with increasing z. Consequently, it is not difficult to appreciate that if leaky radi-
ation occurs the radiation intensity, at any position z′, grows in the y-direction up to
the point y ¼ z0 tan /. This growth is in accordance with Eq. (8.55). In practical
terms the leaky wave radiation is confined to the region y\z tan /.
The key to long-slot leaky-wave antenna design in which the rectangular
waveguide feed is modified by a long axial slit, through which the modal energy in
the waveguide dissipates, lies in determining the propagation and attenuation
coefficients for the adapted waveguide. For long leaky slits which can be presumed,
with little error mathematically, to be of infinite extent, an electromagnetic
boundary problem is formed. The resultant boundary value problem is readily
amenable to solution by an electromagnetic solver either based on the moment
method, or on the finite element method [13, 14]. As we have noted previously,
while such solvers can be very accurate, if set up properly, it is generally good
engineering practice to anticipate their use by developing simple analogues which
are open to approximate field solutions. In this way a thorough insight into the
antenna characteristics becomes available, while at the same time furnishing a
method of checking the predictions of an electromagnetic solver if adopted.
For the long slot leaky-wave antenna with the aperture located in the sidewall of
the rectangular feed waveguide, an elegant and surprisingly accurate field solution,
using approximate methods, is provided by Goldstone and Oliner [25]. The analysis
employs a solidly established transverse resonance technique which recognises that
the non-axial propagation coefficients (kx, ky) in Eq. (8.52) represent the transverse
resonances of all possible modes. Transmission line techniques are used to model
the resultant field interactions in the transverse directions, including the slit dis-
continuity, represented as a complex impedance, between the transmission line
analogue and a half-space. For this class of leaky-wave antenna it is recognised that
the leaky-wave solution is essentially a perturbation of the electromagnetic mode
existing in the waveguide in the absence of the slit. Consequently, perturbation
206 8 Frequency Scanned and Leaky-Wave Linear Slot Arrays

theory is used to establish ‘closed’ form equations to represent the real and
imaginary parts of the propagation coefficient in the leaky-wave structure.
For a rectangular waveguide perturbed by a slit running along the length of a
sidewall Eq. (8.52) can be expressed as:

k02 ¼ ðkx þ DjÞ2 þ ky2 þ kz2 ð8:58Þ

where Dj is the perturbation on the closed waveguide value for the x-directed mode
number, and is assumed to be complex, which means that kz is complex and given by:

kz ¼ b  ja ð8:59Þ

Hence, some routine algebra leads to:


!
b k0 k0 kx k2go
¼  1 Djr
k0 kg kgo 4p2
ð8:60Þ
k0 kgo
ak0  kx Djj
2p

where kgo is the guide wavelength in the unperturbed waveguide, while


Djr and Djj are the real and imaginary parts of Dj. Transmission line analysis of
the transverse field relations in the perturbed waveguide yields the following
approximate equations [25]:
 
j YT0 ðj0 Þ  j cot j0 a
Dj  ð8:61Þ
a csc2 j0 a

Here YT0 denotes the normalised terminating admittance of the transmission line
analogue. Expressions for the terminating impedance or admittance, where the slit
as ‘seen’ from the transverse propagation perspective can be viewed as a capacitive
iris, are available in many standard texts on electromagnetism [6, 7, 26]. Not
surprisingly they confirm that this admittance is strongly dependent on slit width
d. Reference [25] gives:
" #
GT p2 bð1  ðdbÞ2 Þ2 2p
G0T ¼  0:285  
Y0 16d cos2 pd2b
jd
" # ð8:62Þ
BT p2 bð1  ðdbÞ2 Þ2 2p
B0T ¼  0:156  
Y0 16d cos2 pd 2b
jd
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
with j ¼ k02  ðpaÞ2  b2 .
The phase shift constant b expressed as k0/kg, and the attenuation coefficient aa
are plotted in Fig. 8.13 as functions of the slit width d. The plots show clearly the
8.5 Uniform Leaky-Wave Antennas 207

Fig. 8.13 Guide wavelength and attenuation coefficient as a function of slit width d in a
rectangular waveguide with a = 22.86 mm and b = 10.16 mm at a frequency of 9.4 GHz. (Dashed
curves = approximate solution: Solid curves = exact solution)

dependence of a and b on the slit width, with the influence on the attenuation rate
more evident. However, more importantly the curves show minimal deviation from
the approximate closed form solutions (Eq. (8.54); dashed curves) from
super-positions which are nominally ‘exact’. From an engineering view-point it
cannot be too often stated that the availability of easily understood and easily
implemented solutions for a given class of antenna is immensely helpful to the
design process.
Once a and b have been ascertained to a requisite degree of accuracy, it becomes
a relatively straightforward matter to determine the radiation characteristics of the
leaky-wave antenna. Such characteristics include beam direction, beamwidth,
radiation efficiency and scan sensitivity to frequency. Since, in this section scanning
antennas are the focus of our attention we will limit ourselves to the scan sensitivity
issue. For long-slot antennas which are largely uniform in cross-section, and
unchanging geometrically in the axial direction, the scan behaviour usually depends
on whether or not the feed waveguide is air filled or partially dielectric filled.
Generally, the air filled option provides a superior scanning antenna. This is
explicable by directing one’s attention toward transverse wave numbers (kx, ky)
which in the air filled case are independent of frequency [see Eq. (8.52)]. The
implication is that the beamwidth of the radiation pattern of the leaky-wave antenna
does not change as the beam is scanned with frequency. This scan behaviour is
summarised in Eq. (8.56), which dictates that the beam scan angle / progresses
from broadside ðb=k0 ) 0:/ ¼ 90
Þ near cut-off, to endfire at high frequencies in
208 8 Frequency Scanned and Leaky-Wave Linear Slot Arrays

the direction of propagation within the feed (+z-direction: b=k0 ) 1). Furthermore,
the application of elementary trigonometry to (8.56) yields:

b2
sin2 / ¼ 1  cos2 / ¼ 1 
k02
 b 2  k 2
But b2 ¼ k02  kc2 , consequently for an air filled feed 1  k0 ¼ c
k0 , and
therefore:

kc
/ ¼ sin1 ð8:63Þ
k0

Substituting into (8.57) then reveals that:

5:53
D/  ð8:64Þ
kc L

Note that (8.64) confirms that the beamwidth is independent of frequency for the
air filled feed, as indicated above.
In the partially dielectric loaded feed increasing the frequency causes the field
distribution of the internal mode to become more and more confined to the
dielectric, weakening the slit excitation level. This process significantly degrades
the beamwidth which is inevitably considered to be a disadvantage. On the other
hand, it is well known that the phase shift coefficient in a dielectric loaded
waveguide exhibits an increasing magnitude with frequency, which is significantly
larger than for an equivalent empty waveguide. The consequence is, that with
dielectric loading, the uniform leaky waveguide antenna can be designed to scan
over a larger range of angles than the air filled alternative, for a prescribed fre-
quency range.

8.6 Periodic Leaky-Wave Antenna Arrays

While leaky-wave antennas evidently manifest themselves in a wide variety of


geometries, which are uniform rather than periodic in nature [1], nevertheless we
will direct our attention to the antenna array with periodically and closely spaced
slot radiators, as depicted in Fig. 8.14b. This is consistent with the slot antenna
emphasis of the book, and in this section, it is the evolution of linear slot arrays into
structures incorporating overt slow-wave characteristics [1, 11, 25], which is of
primary interest. This means that in a rectangular waveguide fed travelling-wave
array, formed from transverse slots in the broadwall, if the slots become so closely
spaced (S ko; see Fig. 8.14b) that mutual coupling makes a significant contri-
bution to the array behaviour, then the slot array as a whole is considered to
represent periodic loading of the waveguide. The periodically loaded waveguide
8.6 Periodic Leaky-Wave Antenna Arrays 209

Fig. 8.14 Waveguide linear array antennas employing transverse (series) slots. a Travelling-wave
array, b leaky wave array

structure must then be treated as a single electromagnetic boundary value problem,


the solutions for which are rather different from those of the original empty
waveguide as we have seen. Electromagnetically, the transitioning from uniform
rectangular waveguide to a periodically loaded version has been addressed in
Chap. 7, Sect. 7.3 and this material provides a helpful route into the analysis of the
closely spaced transverse slot array.
The reader will be aware at this juncture that the travelling-wave array antenna in
a rectangular waveguide feed is largely realised using longitudinally directed
(shunt) slots. Nevertheless alternative slot arrangements for travelling-wave arrays
210 8 Frequency Scanned and Leaky-Wave Linear Slot Arrays

Fig. 8.15 Pictorial representation of the mutual coupling mechanism for a shunt slots and
b transverse or series slot in a linear waveguide array antenna

have been tried, most commonly in the form of inclined slots and transverse (series)
slots. The transverse slot in rectangular waveguide as shown in Fig. 8.14 most
clearly illustrates the difficulty of adopting slot forms other than the shunt variety,
so the following exposition is focused upon it.
The overwhelming reason for the low adoption of non-longitudinal slots in
waveguide arrays of this type is mutual coupling between neighbouring slots, which
is strongly influenced by slot orientation and positioning. The electromagnetic field
mechanism underpinning divergent levels of mutual coupling between notionally
similar slots in waveguide is readily explained by reference to Fig. 8.15. The figure
depicts a y-z plane cut of a rectangular waveguide feeding a pair of longitudinal
slots in sketch (a) and a pair of transverse slots in sketch (b). For narrow slots it is
now well established that the electric field induced in the slot by the exciting TE10
mode is almost wholly aligned in a direction normal to the major axis of the slot.
So, for the longitudinal slots in Fig. 8.15a the E-field in each slot is x-directed. This
means that in the y-z plane containing the slot the radiated E-field is also x-directed,
exhibiting the typical (dipole) pattern shape with nulls at the conducting surface of
the waveguide where the x-directed electric field is ‘shorted’. This pattern dictates
that at the neighbouring slot located d * ko/2 away the E-field is weak and mutual
coupling is very low. Similar observations apply to the internally scattered fields.
As a consequence shunt slots in waveguide represent almost ideal array elements
and shunt slot arrays can be investigated easily, as we have seen, using the con-
ventional array theory presented in Chap. 6.
On the other hand, the electromagnetic interaction between a neighbouring pair
of transverse (series) slots in the broadwall of a rectangular waveguide is by no
means insignificant, as Fig. 8.15b purports to demonstrate. In this case, with the
major axis of each slot aligned in the x-direction the induced E-field in both is
z-directed. Thus, in any y-z plane intercepting the slots the radiated E-field forms
omni-directional, semi-circular, equi-field contours as shown. This results in the
radiation power density propagating over the guide surface (z-direction) matching
the normally directed radiation (y-directed). Furthermore, the E-field for this
8.6 Periodic Leaky-Wave Antenna Arrays 211

z-directed surface radiation is normal to the waveguide wall. This behaviour means
that the electromagnetic coupling between slots is no longer weak unless the slots
are separated by more than the free space wavelength. Actually to form a broadside
mainbeam with an array of series slots, with no possibility of a p-phase shift
adjustment, as with shunt slots, slot separation is required to be *kg10. For an air
filled waveguide feed this equates to a slot interval d > ko. While this is advanta-
geous in lowering mutual coupling, it has the effect of permitting the formation of
undesirable grating lobes (see Chap. 6). It can be solved by dielectric loading of the
waveguide which has the effect of shrinking kg10. However, such a change simply
negates the attempt to reduce mutual coupling. Actually, dielectric loading is sel-
dom an acceptable solution due to weight and power loss penalties. It is perhaps
pertinent to note that conventional array theory (Chap. 6) can continue to be applied
to transverse slots afflicted by mutual coupling provided the radiation pattern of
each slot element in the array is not modelled as an isolated slot as in Chap. 5.
Rather the presence of coupled neighbouring slots is modelled by assuming that the
active slot is embedded in a ground plane of passive slots arranged in a periodic
grid representative of the array antenna design. Note that in planar slot arrays
transverse slot mutual coupling is largely unavoidable. Not surprisingly, for
inclined slots mutual coupling strength is a function of the inclination angle [1].
The consequence of the above is that transverse slot arrays can be realised only if
the mutual coupling between slots is viewed as instrumental to the array operation
rather than as disruptive to it. This means that the feed is no longer just a support
structure whereby the TE10 mode illuminates a sequence of slot radiators
(Fig. 8.14a), but forms an integral part of a slow-wave structure created by the
periodic array of slots. In addition to the TE/TM modes of rectangular waveguide
the ladder structure (Fig. 8.14b) can also support surface wave and leaky-wave
modes, which significantly contribute to the radiation characteristics of the antenna.
For the waveguide periodically loaded with slot radiators (Fig. 8.14b) the leaky
phenomenon, as we have seen, results in complex propagation coefficients for the
fundamental space harmonic of each mode—a and b (say) for the dominant mode
[27]. For a typical structure with b/a = 0.466, L/a = 1.0 theoretical estimates of
these parameters, normalised to a/ko and b/ko, are presented in Figs. 8.16 and 8.17.
In a manner reminiscent of the uniform leaky-wave antenna, radiation beamwidth
and radiation efficiency are influenced by the magnitude of a, so that operation near
the waveguide cut-off, where a is large, results in a wide primary beam, which
narrows as the frequency is raised (see Fig. 8.16). Above cut-off a rises rapidly to a
peak once finite power is able to enter the guide. It then falls away with increasing
frequency as the slot radiators, with L = a, become increasingly non-resonant. The
relationship between attenuation constant and frequency can obviously be tailored
by dimensional adjustments as is hinted at in Fig. 8.16.
Generally the ‘leaky’ slotted ladder in the broadwall of waveguide, as suggested
in Fig. 8.14, does not impose heavy loading on the waveguide mode. Consequently
the phase coefficient for the modified guide deviates little from the closed
waveguide prediction, as indicated in Fig. 8.17. The noteworthy observation, which
can be extracted from Fig. 8.17, is that the closed waveguide b-curve represents the
212 8 Frequency Scanned and Leaky-Wave Linear Slot Arrays

Fig. 8.16 Attenuation coefficient as a function of frequency for the dominant mode of a leaky
ladder of slot radiators (with L = a)

Fig. 8.17 Brillouin diagram for the dominant mode of a leaky ladder of slot radiators:
w/a = 0.0016, S/w = 5
8.6 Periodic Leaky-Wave Antenna Arrays 213

border between fast wave and slow wave solutions. Where the leaky wave curve
falls below this line, at approximately b = 0.6ko, the solution ceases to be leaky and
reverts to a trapped surface wave. Therefore, for the specified antenna geometry,
leaky wave performance is available only between cut-off at koa = p and koa = 4.
Finally, it is pertinent to note here that an important difference between the
uniform and the periodic leaky-wave antenna is that the dominant mode in the
former is a fast wave so that the structure radiates whenever it is coupled to
free-space. On the other hand, the dominant mode for a periodic leaky-wave
antenna is a slow-wave that is intrinsically non-radiating. However, ‘fast’ space
harmonics of a higher order mode can exist, and one of these can be a source of
leaky-waves in an unclosed structure such as Fig. 8.14b. Note that if a harmonic is
‘leaky’ the mode as a whole is ‘leaky’. Since the engineering aim is to create an
antenna that forms a single radiating beam the structure must be designed to favour
the mode which is potentially leaky. Sophisticated design of a periodic leaky-wave
antenna, combining a slow-wave structure with a leaky ladder of slots, can achieve
very large scan range, potentially over the angular range −90° < / < 90° by
operating over a frequency range which transitions from a backward space har-
monic to a forward space harmonic. This is not possible with the uniform slit based
alternative.

8.7 Chapter Summary

Linear slotted array antennas implemented in regular rectangular waveguide are


generally considered to be restricted in their use because of ‘squinting’ which links
the beam angle to frequency (see Chap. 7). However, as is often the case in
engineering developments, disadvantages of a technology in one context can be a
boon in another.
Array antennas, which are designed to enhance beam scanning for 3-D radar
applications, have been examined in this chapter. Several different forms are con-
sidered including the sinuous feed, the dielectric loaded feed, and the periodic
waveguide feed. The latter relies on the enhanced frequency dependency of the
phase shift constant of the fundamental mode in the periodically loaded waveguide.
The key to the development of scanned antenna applications is precise knowledge
of the modal characteristics of the waveguide. This is particularly so for periodic
waveguide. The required knowledge is provided in this chapter, while relevant
theoretical methods for estimating frequency scanning capability are described.
When the periodicity is created by the slot array itself, it is demonstrated that this
leads to leaky-wave behaviour. Leaky-wave antennas exist in two basic forms,
namely uniform versions and periodic versions. These are compared in this chapter.
It is shown that scan range is optimal in the leaky-wave antenna array format,
provided full cognizance of the complex modal structure is exercised in the design
process.
214 8 Frequency Scanned and Leaky-Wave Linear Slot Arrays

References

1. R. Johnson, Antenna Engineering Handbook (McGraw-Hill, New York, 1993)


2. R.C. Hansen, Microwave Scanning Antennas, Vol III, (Academic Press Inc., New York,
1966), N.A. Begovich, Chap. 2
3. J.C. Slater, Microwave Electronics (Van Nostrand, 1950)
4. R.M. Bevensee, Slow-wave Structures (Wiley, New York, 1964)
5. W. Walkinshaw, Theoretical design of linear accelerators for electrons. Proc. Phys. Soc. 61,
246–254 (1948)
6. R.E. Collin, Field Theory of Guided Waves (McGraw-Hill, New York, 1960)
7. J.A. Stratton, Electromagnetic Theory (Wiley, New Jersey, 2007)
8. S.B. Cohn, Analysis of a wideband waveguide filter. Proc. IRE 37, 651–656 (1949)
9. P.J.B. Clarricoats, P.K. Saha, Propagation and radiation behaviour of corrugated feeds. Proc.
IEE 118(9), 1167–1186 (1971)
10. K. Milne, The combination of pulse compression with frequency scanning for three
dimensional radars. Radio Electron. Eng. 26(2), 89–97 (1964)
11. A.W. Rudge, K. Milne, A.D. Olver, P. Knight, The Handbook of Antenna Design, vol.
2 (Peter Peregrinus Ltd., London, 1983)
12. http://mathworld.wolfram.com/FourierSeriesSquareWave.html
13. HFSS, High Frequency Structure Simulator, Ansoft Corp
14. COMSOL, https://www.comsol.com/comsol-multiphysics
15. R.F. Harrington, Moment Methods
16. MATLAB, Mathworks Ltd., https://uk.mathworks.com/products/matlab.html
17. A.J. Sangster, B. Maher, Higher order evanescent modes on slow wave structures. Electron.
Lett. 8, 608–609 (1972)
18. R.W. Lyon, An Investigation of Apertures in Dual Mode Rectangular Waveguide for Variable
Polarisation Arrays, Ph.D. Thesis, Heriot-Watt University, 1979
19. R.A. Hurd, The propagation of an electromagnetic wave along an infinite corrugated surface.
Can. J. Phys. 32(12), 727–734 (1954)
20. R.W. Lyon, A.J. Sangster, Efficient moment method analysis of radiating slots in thick walled
rectangular waveguide. IEE Proc. 128(4), 197–205 (1981)
21. M.G. Salvadori, M.L. Baron, Numerical Methods in Engineering (Prentice-Hall International,
London, 1961)
22. T.J. Akai, Numerical Methods (Wiley, Canada, 1993)
23. R.S. Esfandiari, Numerical Methods for Engineers and Scientists using MATLAB (CRC Press,
Boca Raton, USA, 2013)
24. A.J. Sangster, D.C. Hawkins, in IEEE Conference Publication. A slotted waveguide circularly
polarised linear array for a 3D antenna system, No. 105, pp. 94–97, Oct 1973
25. L.O. Goldstone, A.A. Oliner, Leaky-wave antennas – part 1: rectangular waveguides. IRE
Trans. Antennas Propag. AP-7, 307–319 (1959)
26. N. Marcuvitz, Waveguide Handbook (M.I.T. Radiation Laboratory Series, McGraw-Hill Book
Co., Inc., New York, 1951)
27. R.F. Hyneman, Closely spaced transverse slots in rectangular waveguide. IRE Trans.
Antennas Propag. AP-7, 335–342 (1959)
Chapter 9
Compact Planar Reflectarrays

9.1 Introduction

It is appropriate at this juncture in our examination of slot antennas to apply what


we have learned so far, on slots and radiation, to the examination of planar arrays,
and in particular planar slot arrays. Chapters 2–7 have furnished the reader with the
electromagnetic ‘tools’ required to evaluate radiation from antennas of the slot type,
particularly when excited by the most commonly employed electromagnetic feed
structure, namely rectangular waveguide. This propagation format is almost
invariably adopted wherever significant power delivery and/or low loss are desir-
able. However, to be an effective antenna in most practical scenarios the slot
radiator cannot act on its own. It inevitably needs to form an element in an array.
The ‘tools’ needed to evaluate the radiation behaviour of a range of possible linear
array configurations are provided in Chaps. 6 and 7. It is pertinent to note that ‘full
blown’ planar phased array systems, where each array element is backed by a
self-contained transmit/receive device, remain very expensive. Consequently, they
are generally limited to military installations such as long range radar, and to ‘big
science’ applications such as astronomy. Needless to say, a ‘half-way house’ planar
array antenna can readily be formed from the paralleling of linear arrays fed from
branching networks of couplers. This is particularly true of waveguide systems
where the linear slotted waveguide arrays are excited by end-fed, or centre fed,
waveguide distribution systems. However, such antenna developments tend to be
equally large being aimed at providing long range 3-D radar installations in the
lower microwave frequency bands (L-band to X-band). They remain costly despite
the technological simplifications which can be adopted by comparison with max-
imally versatile phased arrays. An illustrative example is shown in Fig. 8.3.
At the time of writing this text, it is probably fair to say that commercial
applications of planar arrays have tended to concentrate on formats which offer
specified functions designed to deliver highly specialised aims at affordable cost
levels. Such systems predominantly provide compactness, conformability and

© Springer Nature Switzerland AG 2019 215


A. J. Sangster, Compact Slot Array Antennas for Wireless Communications, Signals
and Communication Technology, https://doi.org/10.1007/978-3-030-01753-8_9
216 9 Compact Planar Reflectarrays

lightness based on the adoption of advanced fabrication methods, such as 3-D


printing. They find applications in mobile and satellite communications systems and
satellite sensing and radiometry, generally in operating bands at the mm-wave and
low infrared sections of the electromagnetic spectrum, competing with slotted
waveguide array antennas as indicated in Chap. 7, Sect. 7.2.5. To be consistent with
the thrust of this book, it is appropriate here to direct our attention toward these
planar slot array antennas, which fall into the compact category. The availability of
increasingly accurate fabrication methods enables the realisation of the high pre-
cision periodic surfaces which are in evidence in planar arrays operating at
mm-wave and low infra-red frequencies. Typical examples which we will explore
in this Chapter are the reflectarrays. Antennas in the Fabry-Perot cavity resonator
(CRA) class, which offer similar sensing solutions, are discussed in Chap. 10.
In essence the reflectarray [1, 2] represents an attempt by antenna designers, at
least at the outset, to replace ubiquitous and very traditional parabolic reflector
antennas in satellite communications roles, by planar slot/patch array technology
which offers potential improvements such as lower weight, less bulk, increased ease
of assembly and installation (in space), and a more cost effective alternative to the
parabolic dish. The latter is particularly true at mm-wave frequencies.

9.2 Passive Arrays

While both the reflectarray and CRA antennas presents an electromagnetic


boundary value problem which is straightforward enough to permit direct analysis
by commercially available full wave finite element based solvers, such as HFSS or
COMSOL, significant computing power remains necessary to simulate, with
meaningful detail and accuracy, a high gain structure. On the other hand, by
generating an approximate analysis constructed from established field equations and
antenna analysis techniques, basic scattering relationships can be deduced while at
the same time procuring valuable insight into the fundamental nature of the passive
array antenna. Such an analysis can usefully double up as a check on simulation
predictions where these are pursued.
It is perhaps not too surprising that the passive array can be analysed using
antenna array theory (Chap. 6) together with aperture theory (Chap. 3) and plane
wave spectrum theory (Chap. 3). It can comprise many elemental scatterers (patches
[3] or grooves [4] or dielectric protrusions [5]), and to some extent the most
appropriate analysis technique is dependent of the array size and hence the number
of scatterers. For small arrays an element-to-element or mutual impedance approach
is preferred [6] whereby the interactions between every element in the array is
tabulated leading to a potentially large set of simultaneous equations. While the
method is intrinsically simple the computation of the mutual interactions between
realistically modelled elements is not. Consequently even for relatively small arrays
computations can be restrictively time consuming.
9.2 Passive Arrays 217

The alternative approach, which is more relevant to multi-element arrays,


embraces the argument that in a large array the vast majority of elements, remote
from the array periphery, are embedded in an environment which is largely that of
an infinite array. The adoption of an infinite array analogy introduces the significant
mathematical advantage of permitting the array to be treated as a periodic surface.
This facilitates [7] the application of plane wave expansion or spectral techniques to
the passive reflectarray geometry.
In Chap. 3 the essence of antenna analysis is demonstrated to reside in the
integral equations (see Table 3.1):
ZZZ
l0 Jðr0 ; tÞ 0
Aðr; tÞ ¼ dV ð9:1Þ
4p jr  r0 j
V
ZZZ
e0 Jm ðr0 ; tÞ 0
Am ðr; tÞ ¼ dV ð9:2Þ
4p j r  r0 j
V

As we have observed in earlier chapters for wire, dipole and patch antenna
elements on which an electric current density J exists (Chap. 2), the starting point
for a radiation calculation is Eq. (9.1), while for antenna elements of the aperture
type (slots, holes, slits and patch/slots) where radiation emanates from an E-field
usually referred to, by analogy, as the magnetic current density (Jm), Eq. (9.2)
represents the appropriate commencement of any calculation. It is, hopefully, also
clear from Chaps. 2, 3 and 4 that these equations lead to the powerful observation
that for any wire antenna or aperture antenna, if the current densities on a specified
antenna geometry are known and integrable, and if the dyadic Green’s functions
(see Table 4.1) for the coupled spaces are available, then near and far field radiation
patterns can be computed.
The introduction of passive arrays, such as reflectarrays and frequency selective
surfaces, into the analysis mix requires the application of Eqs. (9.1) and (9.2) to a
space, or spaces, which are not accommodated by the traditional antenna analysis
tools. A procedure for accommodating radiators, which form elements of a periodic
surface, is required [7]. We commence the process by replacing Eq. (9.1) by its
arguably more familiar usage as it applies to the Hertzian current element (see
Chap. 2). It has the form:

l0 I0 dl
A¼ expðjðxt  k0 rÞÞ ð9:3Þ
4pr

We could have commenced with Eq. (9.2) and proceeded in a similar manner by
applying it to an isolated magnetic current element (i.e. an elemental slot).
However, we shall persevere with Eq. (9.3) since periodic surfaces used in antenna
pattern control are more commonly fabricated from multiple scatterers of the ‘wire’
based variety (see Fig. 9.1). Note that in Eq. (9.3) the term
218 9 Compact Planar Reflectarrays

Fig. 9.1 Plane wave illumination of passive periodic array of current elements

1
expðjðxt  k0 rÞÞ
4pr

represents in essence the scalar Green’s function for free-space in phasor form. The
vector form is given in Table 4.1.
Generally speaking, the incorporation of frequency selective surfaces or periodic
surfaces into antenna technology aims to influence radiation characteristics by
making use of the controlled and adjustable scattering behaviour of the surface,
when illuminated by a plane electromagnetic wave. There are few antenna situa-
tions where plane wave illumination cannot be applied and therefore we concentrate
on this aspect. In Fig. 9.1 a plane electromagnetic wave is presumed to propagate
from the left, in direction kð¼ ^ak k0 Þ, where ^ak is the unit vector and k0 = x/c, and
illuminates an inclined flat surface supporting a regular lattice of dipole like scat-
terers. The periodic surface lies in the x-y plane of the x, y, z Cartesian system, with
x-directed metallic dipoles spaced Dx apart in the x-direction and Dy apart in the
y-direction. A more general system would have the dipoles directed at angle w to
the x-axis, but in such a case the dipole can be resolved into x-directed and
y-directed portions, so it is not too restrictive to limit consideration to the less
complex x-directed case. In the x-direction the elements are numbered from
1 \ p \ 1 while in the y-direction they are numbered from 1 \ q \ 1,
so we can identify the current on a typical element as ^ ax Ipq .
9.2 Passive Arrays 219

At the reference element Ioo the plane wave arrives in the direction.

k ¼ ^ax kx þ ^ay ky þ ^az kz ð9:4Þ

Thus, if the incident plane wave k induces a current Ioo at the reference element,
then at a typical element located at p, q, where p is the element count in the
x-direction and q is the count in the y-direction, we must have, on applying the
Floquet theorem:

Ipq ¼ I00 expðjpkx Dx Þ expðjqky Dy Þ ð9:5Þ

The exponential factors express the plane wave phase shift between element 0, 0
and element p, q. Note that the currents Ipq are also secondary sources of radiation,
and if the periodic surface is immersed in an incident plane wave of infinite extent,
the elements will radiate equally strongly.
To assess the scattering from the passive array as a whole, we first need to
evaluate the radiation at an observation point r(x, y, z) exterior to the array surface
from each current element p, q. Relative to this observation point the distance to the
element p, q is given by (see Fig. 9.1):

rpq ¼ z2 þ ðpDx  xÞ2 þ ðqDy  yÞ2 ð9:6Þ

Consequently, by analogy with Eq. (9.3) it is possible to expect that the mag-
netic vector potential at r, due to Ipq, is Apq and that it can be expressed as:

l0 I0 dl
Apq ¼ expðjk0 rpq Þ ð9:7Þ
4prpq

Given that all of the current elements are aligned in the direction dlð¼ ^
ax dlÞ then
the magnetic vector potential Ap for the entire linear array at p, with 1\q\1
can be expressed as:

l0 I0 dl X
1
Ipq expðjk0 rpq Þ
Ap ¼ ð9:8Þ
4p q¼1 rpq

This can be formulated in terms of Ioo by substituting Eq. (9.5) into Eq. (9.8), to
give:

l0 I00 dl X1
expðjqky Dy Þ expðjk0 rpq Þ
Ap ¼ expðjpkx Dx Þ ð9:9Þ
4p q¼1
rpq

Generally the infinite sum in the equation is convergent, except where kyDy is a
multiple integer of p—a grating lobe condition. Unfortunately, the summing con-
vergence rate is very slow [6]. Relatively routine mathematical manipulation of the
220 9 Compact Planar Reflectarrays

summation in Eq. (9.9) can generate a more amenable version [6] which can be
expressed as:

X1   
l0 I00 dl ky nk0 ð2Þ
Ap ¼ expðjpkx Dx Þ exp jk0 y þ H0 ðk0 asn Þ ð9:10Þ
4jDy n¼1
k o D y

where
a2 ¼ z2 þ ðpDx  xÞ2 ð9:11Þ

and
"  2 #1=2
ky nk0
sn ¼ 1  þ ð9:12Þ
k0 Dy

ð2Þ
In (9.10), H0 is the Hankel function of order zero and of the second kind. While
Eq. (9.10) is useful, here it represents a step toward a solution for the infinite planar
array, since it is the adoption of planar periodic surfaces, rather than passive linear
arrays, that provide routes toward developing compact, electronically controllable
antennas. For the infinite planar array formed from an array of y-directed linear
arrays (q), it is clear that:
X
1
A¼ Aq ð9:13Þ
p¼1

On substituting Eq. (9.10) into Eq. (9.13) we obtain:


  1
l0 I00 dl X
1
ky nk0 X ð2Þ
Ap ¼ expðjk0 y þ Þ expðjpkx Dx ÞH0 ðk0 asn Þ ð9:14Þ
4jDy n¼1 ko Dy p¼1

To improve the rate of convergence the p-summation can again be manipulated,


in much the same way as for the q summation. This is expounded in detail in [7].
The result is:

P
1  h i P
1  h i
A ¼ l4jD
0 I00 dl k
y
exp jk0 y koy þ nk Dy
0 2p
Dx exp jk 0 x k0 þ
kx ik0
Dx
n¼1
 i¼1

2 1=2 ð9:15Þ
expðjk0 z 1ðkk0x þ ikD0x Þ ðkk0y þ nkDy0 Þ
2
Þ
  1=2
1ðk0 þ Dx Þ ðk0 þ Dy Þ
kx ik0 2 ky nk0 2

Equation (9.15) can be written more compactly as:

l0 I00 dl X 1 X1
expðjk0 r  ^
ar Þ
A¼ ð9:16Þ
2jk0 Dx Dy n¼1 i¼1 sz
9.2 Passive Arrays 221

where r(x, y, z) is the vector denoting the reference position (Fig. 9.1) while the unit
vector is given by:
   
kx ik0 ky nk0 kz
^ar ¼ ^ax þ þ ^ay þ ^
az ð9:17Þ
k0 Dx ko Dy k0

Also
" 2 #1=2
  
kz kx ik0 2 ky nk0
¼ 1 þ  þ ð9:18Þ
k0 k0 Dx k0 Dy

By analogy with Eq. (9.3) it is evident that the Eq. (9.16) represents, in effect,
the free-space Green’s function for a space bisected by a plane periodic surface.
Furthermore, relatively straightforward applications of the Maxwell equations (see
Chap. 4) to Eq. (9.16) generates the corresponding expressions for electric field and
magnetic field, namely:

g0 I00 dl  ^ar  ^ar X1 X1


expðjk0 r  ^
ar Þ
E¼ ð9:19Þ
2Dx Dy n¼1 i¼1
s z

I00 dl  ^ar X
1 X1
expðjk0 r  ^
ar Þ
H¼ ð9:20Þ
2Dx Dy n¼1 i¼1 sz
pffiffiffiffiffiffiffiffiffiffiffi
where g0 ¼ l0 =e0 . Note that Eqs. (9.16), (9.19) and (9.20) apply to both the
forward and backward scattering into the neighbouring half-spaces. A thorough
assessment of the infinite array Green’s function issue is presented in Ref. [7].
The nature of the solutions contained in Eqs. (9.16), (9.19) and (9.20) is depicted
schematically in Fig. 9.2. The plane passive periodic surface created by an array of
current elements (dipoles, patches) is supported by a dielectric sheet (yellow) which
lies in the x-y plane of the Cartesian coordinate system. The sheet is illuminated by
an incident plane electromagnetic wave propagating in the direction ^ ak illustrated
by the large blue arrow.
The aligned scatterers in the uniform passive array are equally excited by the
incoming plane wave and they adopt its phase profile over the array surface. The
initial response electromagnetically is specular scattering of a single reflected wave
into the incidence half-space (z < 0), and a single transmitted wave into the reverse
space (z > 0). This behaviour matches that of plane wave incidence on a smooth flat
semi-transparent screen. The reflected wave (red arrow) is formed in the direction
^ar , while the transmitted wave (green arrow) is directed along ^ ar þ . For a plane
screen this is the only scattering which occurs, whereas the periodic surface
generates additional outcomes depending on the magnitudes of the ratios Dx/ko and
Dy/ko.
222 9 Compact Planar Reflectarrays

Fig. 9.2 Electromagnetic wave scattering for FSS illuminated by a plane wave

The key equations are (9.17) and (9.18). When i = n = 0 in these equations, the
fundamental non-periodic scattering is being implemented. The equations reduce
to:

kx ky kz
^ar ¼ ^ax þ ^ay  ^
az ð9:21Þ
k0 k0 k0

indicating, as anticipated, that two plane waves are scatterered in the directions
^ar , namely the transmitted wave ð^ar þ k0 Þ and the reflected wave ð^
ar k0 Þ. Note that
these directions are independent of the nature of the surface periodicity. On the
other hand, for large enough values for Dx and Dy it is possible for other values of
i and n to generate more real solutions for ^ar in which case additional scattered
waves can appear. These are the grating lobes (see Chap. 6) as shown schematically
by the orange coloured arrows in Fig. 9.2, approximating the particular case of
i = 0 and n = −1. Furthermore, it is clear from Eq. (9.18) that for high enough
values of i and n the propagation ratio kz/k0 can be imaginary (= jkz/k0), whereas kx
and ky remain real. This implies that above (+z) and below (−z) the periodic surface,
waves can exist which propagate freely in the x and y directions, but decay
exponentially in the z-directions. These solutions are termed surface waves, slow
9.2 Passive Arrays 223

waves or trapped waves. They are depicted in Fig. 9.2 as blue bands above and
below the array which fade monotonically with distance jzj. While in radiation
terms these surface waves are insignificant in magnitude a finite distance away from
the array, they represent vital stored energy in the array and are very influential to
the radiation control which the passive array surface provides.

9.3 Reflectarrays

The reflectarray provides a good example of a flat periodic surface being used to
control the scattering of incident electromagnetic waves. There are three basic
options (1) all metallic, (2) patch, (3) printed. The design principles underpinning
the reflectarray concept are fully explored in Refs. [1, 2] and are well rehearsed in
[3–6, 8–11]. Obviously, the essence of focusing electromagnetic waves with a
reflector is well understood and can readily be explained by reference to Fig. 9.3a.
In antenna terms, the requirement of the reflector is to reflect the spreading (un-
focused) rays from the source (usually a horn) so that waves emanating from the
source arrive at a plane surface ahead of the reflector with the same phase, i.e.
focused. This surface is termed the equiphase surface (Chap. 3). It is usually chosen
to coincide with the extremities of the reflector and is then viewed as the radiating
aperture of the antenna as a whole. For curved reflector systems the route to
forming an equiphase surface is to ensure that all rays from the horn or feed to this
‘aperture’ surface are equal in electrical length. This in turn dictates that the
reflector is parabolic in shape.
For the reflector in Fig. 9.3a, at the equiphase front, the total phase delay from
the horn phase centre is

k0 ðr0 þ r1 Þ ¼ constant ð9:22Þ

for all possible ro and r1 values, where ko = 2p/ko is the free space propagation
coefficient. So provided the reflector has a parabolic profile a focused beam is
formed in the direction normal to the reflector (z-direction).
The reflectarray has been proposed as a means of producing focused reflection
using flat, low cost microstrip or stripline substrates. More recently reflectarrays
formed using all metallic grooved plates or 3-D dielectric printed protrusions on a
metal base have been advocated. In the microstrip option, it usually comprises a
periodic array of square resonant patches or narrow resonant slots, as suggested in
Fig. 9.3b. This patch/slot dual interpretation is possible since the radiation mech-
anism for a resonant patch is associated with the fields in the ‘open-circuited’ edges
(edge slots) of the patch (see Fig. 1.4). However if the patches are all the same each
patch reflects in mirror fashion so that the array as a whole acts like a flat mirror,
reflecting the radiated wave from the horn back towards it without modifying the
spreading unfocused pattern as Eqs. (9.17) and (9.21) predict.
224 9 Compact Planar Reflectarrays

Fig. 9.3 a Focused reflector antenna, b unfocused slot/patch array, c reflectarray focusing

To procure focusing with a flat reflectarray an adjustment mechanism not unlike


that distilled in Eq. (9.22) is required where a phase altering procedure can be
incorporated, not by path length, but by reflection phase modifications at the
individual patches [1, 2]. If this patch phase adjustment for the mnth patch is /mn,
then for the reflectarray in Fig. 9.3c to be focused, we require:

/mn ¼ k0 ðr0mn þ r1mn Þ  ð2N þ 1Þp ð9:23Þ

where /mn counteracts the physical length differences for the rays from the horn to
the equiphase surface. Note that the 2Np factor incorporates the arbitrary and
electrically meaningless full cycle phase differences between well separated rays.
These are irrelevant to the design calculation since electrically 2p ¼ 360  0 .
The additional −p accommodates the reflection phase shift for a resonant patch
which is 180°.
To get a better appreciation of the dimensional constraints it is generally quite
helpful to consider a real example such as a 12 GHz design [3]. In this design the
9.3 Reflectarrays 225

Fig. 9.4 Basic dimensions for 10 cm  10 cm reflectarray

flat array is required to subtend an angle of 72° at the horn phase centre to achieve a
compromise between excessive aperture blockage (high sidelobes) and excessive
distance (roo) which results in an unwieldy array and lowered bandwidth. The
square array dimensions are set at 10 cm  10 cm. The angle choice immediately
leads to the dimensions for roo and rnn as shown in Fig. 9.4. For an array with a
period of 1.22 cm the reflectarray in this case will comprise a 9  9 grid of patches.
When Eq. (9.2) is applied to this geometry we obtain, at 12.25 GHz, the fol-
lowing reflection phase requirements for a central patch and an edge patch. The
free-space wavelength at 12.25 GHz is ko = 2.45 cm. Therefore, for the 6.7 cm ray
there are 6.7/2.45 = 2.75 wavelengths between the horn phase centre and the central
patch of the array. The 2 full wavelengths can be ignored (2N term). Thus for the
central patch a reflection phase of /oo = 0.75ko = 0.75  360° = 270°. From this
we subtract 180° to allow for the reflection phase of a resonant patch. For the 8.2 cm
long ray to an edge-patch a similar calculation yields /mn = 0.347ko − 180° =
−55.1°. The establishment of the reflection phase settings for all patches, consistent
with forming a focused beam, is relatively routine (see Fig. 9.5) requiring only a
simple algorithm to step through all ray possibilities. This procedure is particularly
required for much larger arrays.
The phasing over the surface of a 30  30 element front fed reflectarray of
square patches, as computed by a commercial mathematics package, is depicted in
Fig. 9.5. The square elements of the array are colour coded to display the reflection
phases required by every element in order to produce a focused beam at broadside
to the reflectarray surface. The right hand scale demonstrates the link between the
colour range and reflection phase, with dark blue representing 0° and dark red/
brown representing 360°. A typical rippling colour pattern is depicted as the
226 9 Compact Planar Reflectarrays

Fig. 9.5 Array phasing to form a broadside beam for a 30  30 square reflectarray (0° = dark
blue, 360° = brown)

element phases compensate for increasing ray path length between the centre and
the periphery of the reflectarray surface. Note that the proximity of deep red and
dark blue patches does not indicate a large phase jump, since 360° equates to 0°.

9.3.1 Reflection Phase Computation

Once the element phase requirements are known it becomes necessary to implement
these phases within the hardware. In the case of a patch array the patch size is used
with a reflection phase of 180° at resonance. In all metal arrays the groove depth is
tailored to the phase requirement, while in 3D printed arrays the protrusion heights
provide phase control. Note that the computation of the reflection phase shifts of
elements in an array cannot be determined from calculations on an isolated element,
because the mutual interactions from neighbouring elements in the array contribute
significantly to the result. However, for arrays of the order of 10  10 elements, or
more, the finite array can generally be replaced by an infinite array with seriously
influencing the behaviour of individual elements. This, in turn, means that periodic
array techniques can be adopted to determine reflection phase. The most efficient
route to a solution is by means of the moment method [6, 7] applied to an integral
equation for the unknown currents in the patch element, or the unknown fields at
the groove surface in the case of all metal arrays (see Chap. 5). Of course the
9.3 Reflectarrays 227

Fig. 9.6 Typical reflection phase versus patch size for 28 GHz reflectarray [3]

external region must be represented by the periodic surface Green’s function


(Sect. 9.1) which consists of a double summation of Floquet modes. In Ref. [3],
Pozar used a simulation package based on the moment method to determine the
reflection phases of patches within a 28 GHz reflectarray fabricated on a 0.02 in.
thick substrate with er = 2.95. A typical result from the Pozar analysis is presented
in Fig. 9.6.
The square patch used in the derivation of Fig. 9.6 had a resonant length
Lres = 0.2756 cm, and the curve shows the change in reflection phase from *345°
well below resonance to *30° well above it. This is a sufficiently large phase swing
to enable an effective reflectarray to be formed. Note that the unavoidable ‘mirror’
reflection phase shift of p at the surface of each reflectarray patch is incorporated
within the total reflection phase of the detuned patch. Hence the resonant patch at
DL/L = 0 in Fig. 9.6 has a phase shift of 180°. DL is the incremental size change
from the resonant case.
In simulation terms, a patch element, a metal reflectarray groove, or printed array
protrusion, forming a segment of an infinite array, can be modelled quite readily, by
any modern finite element based simulation package such as COMSOL or HFSS.
This is achieved by embedding the element of interest in a ‘square waveguide’
whose ‘walls’ are periodic boundaries, as suggested in Fig. 9.7. Periodic boundaries
are standard features in most packages and replicate the role of the Green’s function
for the periodic surface as required in moment method calculations. The results
obtained from these calculations are similar to those displayed in Fig. 9.6.
Once the element phasings have been established, the radiation characteristics of
the reflectarray antenna as a whole can be investigated. Generally, this takes the
form of an iterative process towards satisfying a predetermined radiation pattern
228 9 Compact Planar Reflectarrays

Fig. 9.7 COMSOL model to determine the patch reflection phase when embedded in an array

specification. Given that the relative position of the feed horn and the reflect array
are defined in order to compute the element phases, the precise illumination of the
array elements by the chosen feed, entails a determination of its radiation pattern.
The techniques for doing this are provided in Chap. 3. With this information,
backed by geometric optics, it becomes possible to compute the magnitude and
phase of the incident radiation at every element. Each element, of the now finite
reflectarray, can be viewed as a radiator with known excitation phase and signal
level. Hence, the radiation pattern of the reflectarray antenna as a whole can be
constructed from the array theory developed in Chap. 6. If the pattern fails to meet
the desired specification on an initial pass then further iterations will be required.
This computational procedure will generally lead to acceptable design [3].
Computed E-plane and H-plane patterns of just such an exercise are presented in
Fig. 9.8. These were calculated, as described above, for a 77 GHz reflectarray,
which is defined in Ref. [3]. The differences between the traces can be attributed to,
firstly, pattern differences in the E and H planes for the feed and, secondly, to the
patch elements which radiate differently in the E and H planes near endfire. For the
front mounted feed, blockage can differ between the two principal planes and may
be reflected, in addition, in the overall patterns. Comparisons with measurements,
which are generally excellent, are provided in the original paper.

9.4 Reflectarray—State of the Art

At the time of compiling this text, in 2017, with the primary aim of collecting, in a
single volume, a compendium of design techniques which are generally now
available for creating antennas employing passive arrays of slots, the reflectarray
was featuring strongly in the literature. Some versions of the reflectarray naturally
emphasise the use of slots, while others based on patches, for example, can be
viewed as slot dependent insofar as patch radiation can be attributed to fields in the
edge ‘slots’ (see Chap. 1, Fig. 1.4). Consequently, it seems apposite at this point to
9.4 Reflectarray—State of the Art 229

Fig. 9.8 Typical H-plane a and E-plane b radiation patterns for a 77 GHz, 6 in. square
reflectarray, formed from 0.005 in. thick Duroid substrate with 5776 linearly polarized patch
elements

review recent publications on this topic while placing these contributions in the
context of earlier work on slot array antennas described in previous chapters.

9.4.1 Microstrip Reflectarrays

The design principles underpinning the reflectarray concept are fully explored in
Ref. [3] and are well rehearsed in Refs. [8–11], which are summarized in Sect. 9.2.
230 9 Compact Planar Reflectarrays

The Pozar paper [3] provides three examples of antennas designed for use as high
frequency space-based sensors. These would in the past have been implemented
using parabolic dish technology, but have instead been designed around planar
reflectarrays, formed from microstrip patches. Such arrays are potentially more
convenient to transport and to install. Simple square patches are used in all three
designs with the required reflection phase adjustment at every patch being achieved
by dimensional modifications. The operational conditions for individual patches
may thus move from below resonance to above resonance, thus providing the
required reflection phase ranging from −180° to +180°. The examples presented in
[3] are of millimetre-wave designs—three operating at 28 GHz and two at 77 GHz.
Of the 28 GHz designs two are directed at similar 6 in. square reflectarrays but
fabricated onto different substrates—0.01 in. thick Taconic, and 0.02 in. thick RT
Duroid. The former required 528, 0.609 cm  0.667 cm, off-square patches, while
the latter was formed from 784, 0.544  0.544 square patches. In reported antenna
modelling outcomes, and in comparisons with measurement, the reflectarrays are
illuminated from the boresight direction from suitably designed circular corrugated
horns. The differences in performance between the two antennas are shown to be
small, and measurements are demonstrated to be in very good agreement with
theory, confirming the accuracy of the analysis method outlined in Sect. 9.2.
Comparisons are reported to have been made with an equivalent parabolic
reflector antenna, and generally indicate that any differences in performance are
slight, except for bandwidth, which is at best 5% for the reflectarray, much less than
for a parabolic reflector antenna which is limited mainly by the feed. The source of
this limited bandwidth is the resonant patches themselves which display impedance
bandwidths in the 5% range. Nevertheless, for antennas operating in the
millimetre-wave bands, a 5% bandwidth still provides considerable gross frequency
range, in GHz terms, and for many applications, it seems to be not too restrictive.
On the other hand, in the 12–14 GHz frequency band, for antennas in satellite
communications roles 5% is likely to represent a distinct disincentive to the
adoption of reflectarray technology.
The 77 GHz reflectarray [3] is presented both in prime focus, and in Cassegrain,
formats. The microstrip reflectarray itself is 6 in. square, fabricated on 5 thou-
sandth’s of an inch thick RT Duroid substrate, and required 5776 patches. Both
antenna types are shown to exhibit high gain, but the Cassegrain version is less
efficient largely because of higher ‘blockage’ generated by the sub-reflector. The
antennas are reported to exhibit −1 dB gain bandwidths of 2.5%. A typical example
of a prototype microstrip reflectarray in a Cassegrain geometry, and aimed at
operation in W-band (94 GHz), is shown in Fig. 9.9.
The bandwidth limitation inherent to the single layer microstrip patch reflec-
tarray is tackled in three ways in the literature. Firstly, non-linear changes in
reflection phase when presented as a function of patch size (see Sect. 9.2) is con-
sidered to be a major factor [3] in narrowing the bandwidth in patch based
reflectarrays, and it has been suggested that much more linear phase change could
be secured by using line length [8]. The proposed X-band array in this submission
[8] employs equi-size square patches and each patch is coupled through ground
9.4 Reflectarray—State of the Art 231

Fig. 9.9 Square patch


reflectarray fabricated in
microstrip and illuminated by
a Cassegrain sub-reflector
(see Yoon [13])

plane slots to a microstrip circuit on the reverse of the ground plane. This circuit
contains couplers and junctions linked to tailored line-lengths which establish the
required reflection phases at each patch in accordance with beam focusing
requirements. Needless to say, with the introduction of a microstrip circuit to
control reflection phases it is a small step to consider introducing active elements
[9] to enhance antenna performance. Such solutions are inevitably complex. From
the perspective of securing low costs in satellite communications and sensing
systems, there seems, at least in 2017, little advantage in taking this route.
Secondly, multiple substrate layers have been employed for bandwidth
enhancement in other patch related applications with adjacent patches on different
layers tuned to slightly different frequencies. The approach is advocated by Encinar
and Zornoza [11] who describe two-layer and three-layer printed reflectarrays. In
Ref. [10] two 40 cm circular reflect arrays with slightly different patch size patterns
are described. The paper observes that for a single layer reflectarray a reflection
phase range >300° can be secured by adopting as thin a substrate as possible, but
that this means that the phase, as a function of patch size, becomes highly
non-linear, thus limiting bandwidth. Modelled and measured results for patch
reflectarrays fabricated on two layers, show that with the outer patch of the
co-located patch pair is set smaller than the inner patch by 30%. It is observed that
reflection phase as a function of patch size can be linearised without degrading
phase range. A designed 40 cm reflectarray, front fed from an 11.95 GHz corru-
gated horn used in TV receivers, displays a creditable −1.5 dB gain bandwidth of
16.7%. However, when the two layer technology is applied to a 150 cm off-circular
reflectarray the bandwidth decreases significantly to *4% [9]. This is largely
attributed to insufficient reflection phase variation in order to accommodate the
much larger range of ray lengths from horn to patch. By introducing a third layer [9,
10] the reflection phase range can be increased from 400° to 800° [9]. With the
introduction of a three layer patch array the 1.5 m diameter reflectarray antenna is
shown to exhibit a credible 10% operating bandwidth [10] but obviously at the
expense of fabrication complexity.
232 9 Compact Planar Reflectarrays

An alternative scheme to achieve enhanced gain-bandwidth in a single-layer


microstrip reflectarray is reported by Yoon [12, 13]. The proposal comprises an
array formed from patch elements of differing geometries and differing frequency
characteristics. The recommended elements are the standard patch (SP) the hollow
patch or square ring (SR), and the ring loaded patch (RLP), which is a combination
of the SP and SR. Curves of reflection phase as a function of element size are
presented to demonstrate how the elements can be adjusted to accommodate dif-
ferent phase ranges between −300° and zero. A 15 GHz reflectarray on a 2.4 mm
substrate and with a diameter of 25.5 cm is described. It is claimed to possess a
−1 dB gain bandwidth of 11.7%. In Ref. [12] the Yoon multi-element concept is
developed to a further level with the range of element geometries increased to nine.
These geometries are the square, round, oval and triangular patch, and the square,
circular, elliptical and triangular metallised track or ring. The ninth form is a square
track with a T-shaped insert. To realise the reflectarray these elements are etched
onto a 1 mm thick RT Duroid 5580 substrate, to a predetermined software gen-
erated pattern. Interestingly, the prototype 10 GHz antenna incorporating the
reflectarray is illuminated by a fixed beam feed which is itself an array, comprising
sixteen patch elements, and having a 17 dB directivity. The overall outcome is a
prototype reflectarray antenna exhibiting an impressive −3 dB measured gain/
bandwidth of 24.73%.
The above scheme has also been applied to a 95 GHz Cassegrain antenna in
which both the main reflector and the subreflector are implemented by reflectarrays
of this multi-element type [13]. A Cassegrain antenna for remote sensing of the
atmosphere (radiometry) at 95 GHz is also reported in Ref. [4]. It uses a conven-
tional parabolic main reflector but a reflectarray for the subreflector. A 3% band-
width is claimed.

9.4.2 All-Metal Reflectarrays

Conceptually, the all-metal reflectarray is arguably a less complex entity than the
patch versions described in the previous section [14–19]. As with the patch based
reflectarray, formed from a flat microstrip substrate, the all-metal alternative is
configured from a flat metal sheet. In order to replicate the beam forming charac-
teristics of a parabolic reflector it must adhere to the reflection phase requirements
at the elements as represented by Eq. (9.23). Inevitably these adjustments at the
element level are achieved rather differently in the all metal alternative. There are
actually three reflection phase altering mechanisms which have been thoroughly
explored in the literature. These are (a) a flat metal plate with (usually) square
grooves of varying depth, (b) a flat conducting sheet with raised metallic
‘pill-boxes’ hollowed to form elemental shorted waveguides, and (c) a pixelated or
stepped structure without the hollowing. These alternatives are illustrated
schematically in Fig. 9.10.
9.4 Reflectarray—State of the Art 233

Fig. 9.10 2-D schematics of possible all-metal reflectarray antennas a grooved metal plate,
b hollowed ‘pill-boxes’ raised from surface of metal backing plate, c stepped or pixelated metal
plate

Case (a) The periodic surface in this option is strictly flat and comprises an array
of small rectangular recesses of varying depth. Radiation from the individual ele-
ments of the array emanates from a magnetic current density Jm at the outer surface
of the groove created by the incident illumination reflecting off its base. That is, the
periodic surface can be viewed as a conducting plane supporting a lattice of
magnetic current radiators. The arrangement is in electromagnetic terms the dual of
the patch reflectarray studied in Sect. 9.2, where the dielectric surface of the sub-
strate supports a lattice of electric current radiators. A solution paralleling the
procedure outlined for the microstrip reflectarray in Sect. 9.2 is evidently available.
A Green’s function which is in effect the dual of Eq. (9.15) has been formulated
elsewhere for the grooved flat metal plate (see Eqs. (9.9) and (9.10) in Ref. [20]).
This Green’s function is defined as applicable to the ‘semi-infinite’ array which
means an infinite array curtailed in size to the required reflectarray area but without
accommodating edge effect. This is not unreasonable for large arrays where edge
elements represent a very insignificant contribution to the overall array behaviour.
The availability of a suitable Green’s function means that it is not too difficult to
formulate an integral equation for the unknown aperture magnetic current, and the
moment method can, as described in Chap. 5 and in Sect. 9.2, be employed to
234 9 Compact Planar Reflectarrays

develop depth profiles for the recess elements, which acknowledges that they are
embedded within a periodic surface.
Actually, much of the literature directed at the all metal reflectarray, in the case
(a) configuration, is noticeably exercised by the limitations of the semi-infinite array
model [15–17]. The alternative solution accommodates edge effect by adopting a
modelling technique (T-block Method) which is largely borrowed from a decon-
struction procedure widely used in dispersion calculations for ridged metal
waveguide [5]. It is claimed to offer more rapid computations than earlier solutions.
The introduction of the T-block concept into the reflectarray analysis is depicted
schematically in Fig. 9.11. In the central region, elemental recesses in the metal
ground plane can be viewed as residing in a semi-infinite periodic surface, for
which the half-space Green’s function exists. Array scattering computations for this
portion of the array can proceed as outlined above. For the edge portions of the
periodic array an alternative procedure is required. In the T-block method each
groove on this part of the array is viewed as a separate element in a finite con-
ducting ground plane. Scattering solutions can be formulated for this geometry by
using diffraction theory for an aperture in a finite ground plane [21]. This is per-
formed for all edge grooves with their relative positions in the periodic lattice are
guaranteed by the T-block process [15–17]. T-block superposition is used subse-
quently to determine the scattering for the periodic surface as a whole. In some
contributions, this problem specific analysis is backed up by a full wave electro-
magnetic simulation based on the finite element technique, mainly as a check on
accuracy.
Beam focusing by means of the plane all-metal reflectarray antenna requires
reflection phase adjustment from individual grooves paralleling the microstrip
based reflectarray summarised in Eq. (9.23). The desired phase condition is illus-
trated in Fig. 9.12 where the reflection phase from each groove is adjusted using its
depth (d(n)) to achieve a condition which matches that of the equivalent parabolic
reflector antenna. Relative to the reflectarray coordinate system at (0, 0) the focus of

Fig. 9.11 Multiple groove flat metal-only reflectarray—solution options


9.4 Reflectarray—State of the Art 235

Fig. 9.12 Phase matching between all-metal reflect array and equivalent parabolic reflector

the parabolic curve is at (0, fo − do) where do is the distance from (0, 0) to the
lowest point on the parabola represented mathematically by the relation:

y ¼ a0 x 2  d0 ð9:24Þ

Hence, using Fig. 9.12 phase equality requires that:


h i1=2 
2b10 d ðnÞ þ k0 x2 þ ðf0  d0 Þ2 ¼ k 0 f 0 þ a0 x 2 þ j y j ð9:25Þ

where b10 is the phase coefficient for the lowest mode in the groove waveguide, k0
is the free-space phase coefficient, d(n) is the depth of the nth groove, while f0 and d0
are defined on the diagram. The nth groove, in T-block notation, is defined as being
located distance T(n) from the origin. Consequently, if we take a0 = 1/(4f0), which
is not untypical of reflectors for very high frequency sensor systems, then the
groove depth requirement for a focused reflectarray is given by [5, 15]:
 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
kg10
d ðnÞ ¼ d0 þ f0  ðT ðnÞ Þ2 þ ðf0  d0 Þ2 ð9:26Þ
2k0

The reflection phase as a function of groove depth, normalised to the deepest


case, is presented as a chain-dotted orange curve in Fig. 9.13. There it is compared
with the equivalent patch array curve normalised to the patch dimension at reso-
nance. It is evident that the grooved reflectarray provides a much more linear
relationship between reflection phase and element size. This is responsible for much
improved bandwidth for the antenna. That the curve for the grooved all-metal
reflectarray is not more linear than is depicted in Fig. 9.13 can be attributed both to
mutual coupling between grooves and edge effect.
236 9 Compact Planar Reflectarrays

Fig. 9.13 Reflection phase computations for three all-metal reflectarray options

Fig. 9.14 Prototype all-metal


reflectarray formed from 5961
rectangular grooves machined
into a 30 cm diameter base
plate (from [5])

A prototype of a front illuminated grooved all-metal reflectarray [5], fabricated


from a 1 cm thick, 30 cm diameter, flat metal plate, is presented in Fig. 9.14. It
comprises 5961 rectangular grooves and is fed from a pyramidal horn designed to
operate at 78.5 GHz. At this frequency a gain of 42 dBi is reported for the antenna
as a whole.
Case (b) An alternative to the tuned recesses of Ref. [5] is to introduce raised
metallic inserts above a flat conducting plate (see Fig. 9.10). Such a scheme is
9.4 Reflectarray—State of the Art 237

described by Chou et al. [14] where the variable height metal protrusions essentially
enclose shorted cylindrical waveguides, which operate not unlike the recesses in
[5]. The claimed advantage of this approach is that the reflectarray can be manu-
factured using a low cost die-cast process. However, this is off-set by the fact that
the array surface is no longer planar and consequently the semi-infinite array
approach to reflectarray analysis, employed earlier, is less valid.
A 48 cm  48 cm square reflectarray operating at 12.45 GHz and fed from an
off-set circular corrugated horn is described in [14]. A measured gain of 33 dB
together with a −3 dB gain/bandwidth of *3% is demonstrated.
Case (c) The elements for an all-metal reflectarray described by Lee et al. [18,
19] take the form of non-resonant square cross-section metal posts or cells rising
above the flat metal base plate of the reflector. The lattice of protrusions of varying
height results in a quasi-curved reflecting surface which generates a high gain beam
when illuminated from a horn feed. In effect what is created here is a parabolic
reflector with a ‘pixellated surface’ (Fig. 9.10). It can be designed using geometric
optics. It is not really a reflectarray in the conventional ‘flat plate’ sense. The
design, fabrication and measurement of a precision machined 7.2 cm  7.2 cm
square reflector operating at 94 GHz is described in the published papers.
Reflection phase calculations emanating from the surface of each step are simple.
For a step of height d above a reference ground level it is given by:

4pd
\q ¼  þp ð9:27Þ
k0

Clearly the reflection phase is a linear function of height d, and this is shown in
Fig. 9.13 (green dashed trace). However, it is also a linear function of frequency,
which inevitably curtails the available bandwidth, when compared with the
equivalent parabolic reflector.
A gain of *35 dB for the front fed antenna is reported in the literature [18]. The
−3 dB bandwidth is not specified but it is likely to closely match that of an
equivalent parabolic dish version of this mm-wave antenna. In fact, a Cassegrain
version of this antenna [19], which displays a slightly higher gain of 36 dB, has
been fabricated and measured and shown to have a creditable −1 dB gain/
bandwidth of 10.7%. A photograph of this reflectarray is reproduced in Fig. 9.15.

9.4.3 3D-Printed Reflectarrays

The growing interest in reflectarrays at the beginning of this millennium, which is


revealed by the antenna literature, shows a predominant emphasis in mm-wave
systems largely for radiometry. Around the same time polymer jetting 3D printing
technology was moving from a novel laboratory curiosity to a serious contender for
the manufacturing of complex devices with sub-millimetre dimensions. Inevitably,
at the frequencies involved, from 100 GHz to 1 THz, the very small array
238 9 Compact Planar Reflectarrays

Fig. 9.15 Cassegrain illuminated all-metal reflectarray with pixelated or stepped periodic surface
(from [19])

Fig. 9.16 Plane wave illumination of 3D-printed reflectarray in off-set receiver geometry

dimensions and element sizes of reflectarray solutions, was deemed to favour this
evolving technology [6]. The paper describes the printing of array elements in the
form of variable height dielectric blocks onto a flat conducting surface. With the
blocks occupying the whole periodic cell—an option rather than a necessity—the
process results in a stepped surface formed from dielectric blocks of square
cross-section and varying height. Each step in the non-flat periodic surface acts as a
dielectric resonator if tuned to the array operating frequency with height used to
procure reflection phase adjustments, as suggested in Fig. 9.16. The reference plane
for phase determination is adjusted to coincide with the tallest block, and all blocks
are then adjusted by the same height in order to ensure that the reflection phase from
the tallest element at the centre of the array is 360°.
9.4 Reflectarray—State of the Art 239

Fig. 9.17 Phase and magnitude of reflection coefficient (q) for a dielectric block in 3D-printed
reflectarray

A drawback for the printed reflectarray is that unless the dielectric used is very
dense, and therefore heavy, only partial reflection occurs at the top surface of each
block. This is illustrated in Fig. 9.17. The transmitted component reflects off the
conducting base plate creating a resonant condition within the bloc. The
re-transmitted portion then interferes with the direct reflection from the block
resulting in the typical oscillatory reflection pattern (see Fig. 9.17) for the reflection
magnitude (green curve, triangles) and phase (blue curve, circles) as block height
changes. Interestingly, to correct this difficulty, the paper suggests that in the future
dielectric constant variations, rather than cell height, could represent a better
method of phase correction.
Three 100 GHz reflectarrays are reported to have been designed and fabricated
using different criteria for determining the dielectric resonator patterns. These arrays
are horn fed from a 30° off-set position. Gains of *25 dB and encouragingly high
−1 dB gain/bandwidth levels ranging from 14 to 20% are claimed.

9.5 Chapter Summary

A comprehensive examination of the reflectarray is furnished in this chapter and it


is shown to provide an instructive insight into the design of planar passive periodic
structures aimed at specialist antenna applications. It is clear from the literature that
the recent surge in the development of reflectarrays has been driven by growing
240 9 Compact Planar Reflectarrays

interest in atmospheric monitoring from space at mm-wave frequencies and this has
inevitably influenced developments.
The relevant technologies which have evolved to date are as follows.
1. Double layer microstrip array as described in [22].
2. Variably shaped reflecting patch elements in single layer microstrip reflectarray
[4].
3. Die-cast metal reflectarray as described in [6].
The analysis of arrays of this description is dependent of the formulation of a
general mathematical representation of radiation in the presence of a passive
periodic surface. The basic steps towards establishing such a formulation are pre-
sented in Sect. 9.1. By commencing with the Hertzian potential function for radi-
ation from a current element, derived in Chap. 2, and by assuming, not
unreasonably, that the non-edge elements of a finite multi-element passive array are
effectively embedded in an infinite array, it is demonstrated that a Green’s function
for an element of the periodic array radiating into the surrounding space can be
formed. It then becomes possible to synthesise radiation patterns for typical passive
arrays such as the reflect-array. The phase compensation calculations required to
procure parabolic reflector type focusing from a ‘flat’ array are also studied.
The chapter is completed with a resume of the state-of-the-art in reflect-arrays.

References

1. D.G. Berry, R.G. Malech, W.A. Kennedy, The reflectarray antenna. IEEE Trans. Antennas
Propag. 11(10), 645–651 (1963)
2. J. Huang, J.A. Encinar, Reflectarray Antennas (Wiley-IEEE, Hoboken, NJ, 2008)
3. D.M. Pozar, S.D. Targonski, H.D. Syrigos, Design of millimetre wave microstrip
reflectarrays. IEEE Trans. Antennas Propag. 45(2), 287–296 (1997)
4. Hu Wenfei et al., 94 GHz dual-reflector antenna with reflectarray subreflector. IEEE Trans.
Antennas Propag. 57(10), 3043–3050 (2009)
5. Y.H. Cho, W.J. Byun, M.S. Song, High gain metal-only reflectarray antenna composed of
multiple rectangular grooves. IEEE Trans. Antennas Propag. 59(12), 4559–4568 (2011)
6. P. Nayeri et al., 3D printed dielectric reflectarrays: low cost high-gain antennas at
sub-millimetre waves. IEEE Trans. Antennas Propag. 62(4), 2000–2008 (2014)
7. B.A. Munk, Frequency Selective Surfaces (Wiley, New York, 2000)
8. A.W. Robinson, M.E. Bialkowski, H.J. Song, A passive reflectarray with dual-feed microstrip
patch elements. Microw. Opt. Tech. Lett. 23(7), 295–299 (1999)
9. A.W. Robinson, M.E. Bialkowski, H.J. Song, A 137-element active reflectarray concept.
Microw. Opt. Tech. Lett. 26(3), 147–151 (2000)
10. J.A. Encinar, Design of two layer printed reflectarrays using patches of variable size. IEEE
Trans. Antennas Propag. 49(10), 1403–1410 (2001)
11. J.A. Encinar, J.A. Zornoza, Broadband design of three-layer printed reflectarrays. IEEE Trans.
Antennas Propag. 51(7), 1662–1664 (2003)
12. J.H. Yoon et al., Single-layer reflectarray with combination of element types. Electron. Lett.
50(8), 574–576 (2014)
13. J.H. Yoon et al., Axially symmetric dual-reflectarray antennas. Electron. Lett. 50(13), 908–
910 (2014)
References 241

14. H.-T. Chou, C.-Y. Lin, M.-H. Wu, A high efficiency reflectarray antenna consisted of periodic
all-metallic elements for the Ku-band DTV applications. IEEE Antennas Wirel. Propag. Lett.
14, 1542–1545 (2015)
15. Y.H. Cho, W.J. Byun, M.S. Song, Metallic-rectangular-grooves based 2-D reflectarray
antenna excited by an open-ended parallel plate waveguide. IEEE Trans. Antennas Propag. 58
(5), 1788–1792 (2010)
16. Y.H. Cho, Transverse magnetic plane-wave scattering equations for infinite and semi-infinite
rectangular grooves in a conducting plane. IET Microw. Antennas Propag. 2(7), 704–710
(2008)
17. Y.H. Cho, TM plane-wave scattering from finite rectangular grooves in a conducting plane
using overlapping T-block method. IEEE Trans. Antennas Propag. 54(2), 746–749 (2006)
18. W. Lee et al., Non-resonant conductor reflectarray element for linear reflection phase.
Electron. Lett. 51(9), 669–671 (2015)
19. W. Lee, Y.J. Yoon, A broadband dual-metallic reflectarray antenna for millimetre wave
applications. IEEE Antennas Wirel. Propag. Lett. 15, 1–4 (2016)
20. J.P. Skinner, P.J. Collins, A one-sided version of the Poisson sum formula for semi-infinite
array Green’s function. IEEE Trans. Antennas Propag. 45(4), 601–607 (1997)
21. R.E. Collin, F.J. Zucker, Antenna Theory (McGraw-Hill Book Co., New York, 1969)
22. A.P. Feresidis, J.C. Vardaxoglou, High gain planar antenna using optimised partially
reflective surfaces. IEE Proceedings-Microwave, Antennas Propag. 148(6), 345–350 (Dec.
2001)
Chapter 10
Compact Planar Resonator Arrays

10.1 Introduction

Arguably the search for low cost antenna solutions, which embody aspects of
phased array technology, has underpinned recent interest in compact array antennas
(CRA) excited by the modal fields in a planar resonant cavity. This technology also
offers the intriguing possibility of securing what has been termed ‘configurability’,
which simply implies an ability to achieve some level of electronic positioning or
adjustment of the radiated beam. Electronic beam control is a technology which has
traditionally been restricted to full phased arrays. Intrinsically high gain, and effi-
ciency, represent a further attraction for CRA’s, as does the possibility of low
fabrication costs for antennas in the millimetre-wave range of the electromagnetic
spectrum.
In its original form, as described by Trentini [1], the planar resonator array was
referred to as a partially reflective surface (PRS) antenna and is founded on an
essentially rather simple concept. It comprises a Fabry-Pérot cavity resonator cre-
ated by the sandwich formed by the electrically conducting (perfect electric con-
ductor—PEC) antenna ground plane and an electromagnetically (‘porous’) periodic
surface (PRS) located approximately half of a wavelength above it. In principle the
structure, which can realistically be viewed as a parallel plate waveguide, may be
unconfined in the transverse directions; that is, parallel to the reflecting planes (x-
y plane in Fig. 10.1). An elementary feed antenna, which may be a microstrip patch
antenna or a ground plane slot, excites the system [1–4]. A weakly resonant
electromagnetic mode (usually TE10) is set up in the cavity due to the multiple
reflections between the reflecting layers. Curiously, antennas of this type have been
simultaneously and independently studied from the viewpoint of the leaky-wave
antenna (LWA), rather as if an alternative technology was being pursued [5–8].
However, this is strictly not true since the antennas in question are also based on the
Fabry-Perot resonator. They differ from the PRS class by trapping the electro-
magnetic resonance between a high impedance surface (HIS) or an artificial

© Springer Nature Switzerland AG 2019 243


A. J. Sangster, Compact Slot Array Antennas for Wireless Communications, Signals
and Communication Technology, https://doi.org/10.1007/978-3-030-01753-8_10
244 10 Compact Planar Resonator Arrays

Fig. 10.1 Radiation mechanism for a partially reflective surface on an infinite ground plane
a E-plane (y-z plane); b H-plane (x-z plane)

magnetic conductor (AMC), replacing the PEC in the PRS version. Electrically the
operating principles are interchangeable so that the PRS antenna is actually a
special case of the LWA. Consequently simplicity is served by viewing all antennas
of this type as cavity resonance antennas (CRA).

10.2 Cavity Resonator Arrays

In order to realise a CRA it is essential that the PRS is only partly reflective to the
cavity mode fields, resulting in electromagnetic coupling to the external space
through periodically spaced apertures, or radiating elements, embedded in the PRS.
Depending in the rate of power leakage the finite volume containing meaningful
10.2 Cavity Resonator Arrays 245

stored energy in the Fabry-Perot resonator will rise or shrink to fulfil the dictates of
power conservation. This is depicted by the shaded areas in Fig. 10.1a, b. High
leakage will result in minimal or weak mode formation and hence low directivity
with only a small area of the PRS illuminated by the minimally established mode.
On the other hand a highly reflective surface, offering a low level of PRS trans-
mission, permits high energy accumulation before the power radiated from a now
larger area of the PRS balances the input power. This condition equates to high
directivity for the antenna as a whole [1, 3].

10.3 PRS Antenna

The original parallel plate cavity resonator, on which the PRS antenna is based, is
essentially a 2-D resonator trapping a TE10 mode (one half-cycle E-field variation in
the z-direction, invariant in the y-direction). If the waveguide is cut-off at the
resonant frequency the fields decay monotonically and rapidly in the x-direction
and contribute little to the resonance condition (c.f. Figure 10.1a). For the PRS
therefore, it is clear from Fig. 10.1 that the essentially two dimensional Fabry-Perot
cavity will support quasi-TEm0 modes which could be either y-polarised as shown,
or x-polarised. With a preselected reflector spacing of d * k/2 the quasi-TE10
mode, as depicted, will be formed. Note that while the quasi-TE10 mode produces a
largely monotonic field variation (in y) due to power loss through the PRS, as
suggested in Fig. 10.1a, for an infinite structure the x-directed propagation must
display a travelling wave behaviour [1] as shown in Fig. 10.1b. The same obser-
vation is true for the orthogonal TE10 mode if excited by an omni-directional
source. However, for d * k/2 the mode is very close to cut-off and the guide
wavelength for the fundamental mode, ky = kg10, will be very large. Thus from the
slowly varying excitation field, exhibiting a very long wavelength, a broadside
pencil beam is formed by the PRS for both modes. Clearly, in the Fig. 10.1a
representation, coupling apertures on the PRS will be excited in phase, resulting in
an intrinsically broadside mode of operation for the antenna in both planes. One the
other hand, the aperture field distribution forming the radiation pattern is not a
uniform one, since the mode, as explained above is concentrated in the central
region of the resonator. This is also true in the H-plane (Fig. 10.1b), if kg10 is long
enough, in which case essentially only one cycle is effective.
For a Fabry-Perot resonator operating in the n = 1 mode (quasi-TE10 mode) as
depicted in Fig. 10.2 the plate separation d dictates the internal reflection angle h in
accordance with the relationship:
pffiffiffiffi
2pd 2p er
¼ d cos h ¼ p ð10:1Þ
kz k0
246 10 Compact Planar Resonator Arrays

Fig. 10.2 Ray optics representation of PRS cavity antenna

If the cavity is dielectric filled, with a relative permittivity of er , the radiation


angle hr is given by Snell’s laws and hence for the dielectric filled cavity:

0:5k0
d ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð10:2Þ
er  sin2 hr

Clearly for a broadside beam we require d  k=2. The dimensions of the cavity
(d) and the relative permittivity of the dielectric spacer (er) require to be adjusted to
ensure higher order mode radiation lobes (secondary lobes) are suppressed partic-
ularly where scanning of the main beam is anticipated. Generally the scanning
requirement is that the main (n = 1) beam can be located anywhere between hr = 0
and hr = p/2 without the n = 2 mode forming a lobe in real space. The n = 1 mode
can be located at endfire when:

0:5k0
d ¼ pffiffiffiffiffiffiffiffiffiffiffiffi ð10:3Þ
er  1

and the n = 2 mode will be present, with a lobe located at broadside, if:

k0
d ¼ pffiffiffiffi ð10:4Þ
er

Equating (10.3) and (10.4) leads to the restriction that for beam scanning without
secondary lobes the relative permittivity must be greater than 1.333. The maximum
n = 1 beam scan angle, which can be secured without degrading the radiation
pattern with a higher order mode lobe, is then given by:
10.3 PRS Antenna 247

pffiffiffiffiffiffi
1 3er
hr ¼ sin ð10:5Þ
2

This condition should not be confused with the occurrence of grating lobes (see
Chap. 6) which is dictated by the element separation in the periodic surface of the
PRS antenna.
Scanning in the z-x plane is possible [5, 6] for a PRS antenna by frequency
adjustment. For a periodic element separation, or pitch, in the x-direction of
magnitude p, array theory in Chap. 6 informs us the when the main (n = 1) beam is
scanned to endfire in the forward direction a grating lobe at backward endfire is
avoided if p/k0 < 0.5. This generally represents the normal mode of operation of a
PRS.

10.3.1 Radiation Pattern Estimation

The excitation of the PRS array is mainly shaped by the modal concentration, and
significantly it is only very slightly influenced by the feed. A radiation pattern with
a tendency toward low sidelobes is the result [1]. This can be illustrated very
effectively by performing a simple ray analysis of the parallel plate geometry. It is a
generally accepted principle in electromagnetic wave systems, as we have already
observed, that any legitimate field pattern can be resolved into a spectrum of plane
waves (see Chap. 3). In a rectangular box, such as a conventional waveguide, it is
shown in Chap. 1 that the waveguide mode patterns can be broken down into plane
wave components. For the parallel plate waveguide, which is just a rectangular
waveguide minus the top and bottom walls, the TE10 mode (see Fig. 10.1) is
formed from a pair of y-polarised plane waves repeatedly reflecting on the side
walls at z = 0 and z = d. However, for the PRS antenna the wall at z = d is only
partially reflective with a reflection coefficient qðhÞ\/ðhÞ which is a function of the
angle of incidence h of the constituent plane waves. This is illustrated in Fig. 10.2.
Transmission through the wall at z = d can therefore be expressed as:
 1=2
s o ¼ 1  q2 ð10:6Þ

at the location of ray zero, for example.


For ray 1, the incident field amplitude at the PRS has, ignoring secondary effects,
the magnitude q\ð/  p  4pd ko cos hÞ for an air filled structure. Therefore, at the
position of ray 1:
 1=2 4pd
s 1 ¼ q 1  q2 \ð/  p  cos hÞ ð10:7Þ
ko

and so on toward the more remote of the n rays.


248 10 Compact Planar Resonator Arrays

Hence, the radiated field pattern for the PRS antenna can be crudely expressed
as:

X
1
Erad ¼ E0 gðhÞqn ð1  q2 Þ1=2 expðjUn Þ ð10:8Þ
n¼0

where the angles Un represent the range of possible reflection angles at the PRS for
the n possible rays. In this equation g(h) is the element pattern for the source at (0,
0, 0) in Fig. 10.2, and
 
4pd
Un ¼ nH ¼ n /  p  cos h ð10:9Þ
ko

Since the reflection coefficient magnitude for the PRS is by definition less than
unity it follows that:

X
1
1
ðq exp jUn Þn ¼ ð10:10Þ
n¼0
1  q exp jH

On inserting Eq. (10.10) into Eq. (10.8) an equation for the radiated power
density (P(h)) emerges [1]:

ð1  q2 ðhÞÞ
PðhÞ ¼ g2 ðhÞ ð10:11Þ
1 þ q2 ðhÞ  2qðhÞ cos½/ðhÞ  p  4pd
k0 cos h

Equation (10.11) emphasises the earlier qualitative observation that the larger
the magnitude of q the sharper is the primary beam resulting in high directivity.
The radiation mechanism could be viewed as electromagnetic seepage from a
standing wave (in the z-direction) within the parallel plate housing. For a patch
based PRS this mechanism has been illustrated in reference [4] by modelling the
near field 2 mm above the PRS (see Fig. 10.3). It is important to note that with a
dual polarised feed (square waveguide or square patch the modal patterns in
Fig. 10.1 will occur with Fig. 10.1a representing the y-z plane, and Fig. 10.1b
representing the x-z plane (TE01 mode replacing TE10). If these orthogonal modes
are excited simultaneously and in phase then a high gain, linearly polarised and
symmetrical pencil beam is formed at broadside. The leaky energy dissipation rate
results in an aperture field shape which is virtually free of sidelobes (see Chap. 6,
Sect. 6.3.2). If the feed modes are in quadrature phase the radiation will be cir-
cularly polarised for a PRS with dual-polarisation capability—for example a peri-
odic surface of square patches [4] or square apertures [3] or cross-slots [6].
10.4 Leaky-Wave Fabry-Perot Antenna 249

Fig. 10.3 PRS antenna near-field 2 mm above periodic surface. Note the concentration of energy
towards the mid-region (gridded) for a surface modelled as unrestricted in x and y (from [4])

10.4 Leaky-Wave Fabry-Perot Antenna

In parallel with innovative developments on the PRS antenna described above,


research has simultaneously been proceeding strongly on the rather similar planar
leaky-wave array antenna which has also evolved from the Fabry-Perot cavity. It is
probably fair to say the PRS version represents a special case, or particular
embodiment, of the planar leaky-wave antenna. The latter is discussed in references
[5] through [9].
The primary features which distinguish the PRS antenna from the leaky wave
alternative are depicted in Fig. 10.4. Both antennas are derived from planar
Fabry-Perot cavities but, while in the PRS antenna the cavity is formed by a
periodic surface (PRS) in parallel with, and separated by k/2, from a perfectly
conducting ground plane (PEC), in the LWA the ground plane is replaced by a high
impedance surface (HIS), or perfect magnetic conductor (PMC), and the separation
distance reduces to *k/4, or less. In principle the LWA makes possible a much
slimmer package.

10.4.1 Reciprocal Periodic Moment Method for CRA

It is apparent from just a cursory look at Fig. 10.4b that any serious theoretical
treatment of the cavity resonator antenna has to accommodate the presence of one
or more periodic surfaces. As we have noted in earlier chapters, while it is possible
with today’s ‘full-wave electromagnetic solvers’ based on either finite element or
finite difference techniques to ‘crunch out’ a solution to any given CRA geometry,
250 10 Compact Planar Resonator Arrays

Fig. 10.4 Sketch depicting the fundamental difference between a PRS antenna and a leaky-wave
antenna a PRS antenna with Fabry-Perot cavity formed by a PRS in parallel with a perfect
electrical conductor (PEC) or ground plane and separated by a distance of k/2 b LWA employing a
Fabry-Perot cavity formed from a PRS above a high impedance surface (HIS) and separated by a
distance < k/2 (h = k/4 if the HIS functions as a perfect magnetic conductor (PMC))

insight and understanding of the electromagnetic wave interactions within the CRA
usually entails modelling, typically based on the moment method. In Chap. 9 the
issue of electromagnetic scattering, and the formulation of a moment method model
in the presence of a periodic surface, also arose in connection with planar reflec-
tarray antennas. However, from a periodic surface perspective, while superficially
similarities exist between a CRA and a reflectarray, electromagnetically they call
for rather different solutions. With the passive reflectarray, scattering at the periodic
surface is made accessible to the moment method, by adopting the notion of the
infinite array. This then permits a degree of mathematical simplification by enabling
the imposition of Floquet’s theorem to the electromagnetic field representations and
hence the formulation of a Green’s function for an exterior space bounded by a
periodic surface. For relatively large (in wavelength terms) reflectarray surface
areas this is actually a rather insignificant approximation, as we have observed.
In contrast to the reflectarray, the CRA is an active antenna, for which the
Fabry-Perot resonator acts as an electromagnetic field transition between the ele-
mentary source, whether slot, patch, dipole or probe, and the antenna aperture
formed by the PRS (see Fig. 10.4a). Consequently, modelling for the CRA can
follow more conventional antenna array and moment method techniques as outlined
in Chaps. 3, 5 and 6. The essence of such a moment method analysis is detailed in
Chap. 5 Sect. 5.2. In simplistic terms it entails the computation of the coupling
between two electromagnetic systems such as a feed system (waveguide in Chap. 5
) and free space usually modelled as a half-space. The unknown quantity, which
needs to be established, is the field distribution in the coupling mechanism,
10.4 Leaky-Wave Fabry-Perot Antenna 251

generally a radiating aperture or slot. This is accomplished by applying field


boundary conditions at the interface. The resultant coupled equations are then
solved by means of the moment method. Once the aperture fields are established it
becomes possible following conventional antenna theory to determine the scattering
parameters and the antenna radiation pattern through the plane wave spectrum
technique applied to the already computed aperture field distribution.
However, with the CRA, the open microstrip substrate, bounded as it is on one
side by a PEC and on the other by a PRS (Fig. 10.5a), presents a complication to
the direct application of a moment method solution both because of its ‘open’
geometry and also because of the periodicity of the PRS. The solution to this
dilemma is presented in references [5, 6]. Well established aperture antenna theory
(Chap. 3) dictates that the radiation pattern for such an antenna can be computed by
representing the radiation as a spectrum of plane waves. Consequently, by invoking
the reciprocity, which exists between an antenna in transmission and an antenna in
receive mode, and provided that the moment method calculations are performed in
the spectral domain, it is feasible to calculate an antenna radiation pattern by
illuminating it with plane waves from a testing current element in the radiation
space. For the CRA the far-field pattern computation then reduces to the calculation
of the field in the interior of the substrate due to incident plane waves (Fig. 10.5a).
The CRA analytical model for a reciprocity computation of the antenna far-field
pattern is depicted in Fig. 10.5. The remote testing dipole is directed along ^ ah ,
although it could equally be aligned with ^a/ . The computation endeavours to equate
the testing field, emanating from the remote dipole, to the field Ex at the location of
the CRA source (at the PEC) as generated by the illuminating plane wave from the
test dipole. In short, the radiation pattern computation for the CRA reduces to the
substrate pattern due to an incident plane wave. As with conventional aperture
antennas such as slot radiators (Chap. 5) it is appropriate to adopt the equivalence
principle which permits the CRA slots to be replaced by magnetic currents Jmx.
The direction vector and the field vectors for plane wave incidence on a surface
in the x-y plane are shown in Fig. 10.6. For excitation of a substrate mode it is the
plane wave component tangential to the surface which is relevant. In the figure the
x-direction is highlighted, but the y-direction could equally have been selected. The
direction of the incoming wave is denoted by ko while the tangentially directed
wave component lies along kx (or ky). Consequently for propagation in the x-z
plane, we can write:

kx ¼ k0 sin hr ð10:12Þ

and k02 ¼ kx2 þ kz2 ð10:13Þ

The incident electric field and magnetic field components associated with the ko
directed plane wave are Einc and Hinc. In the left hand diagram Hinc is entirely
transverse to kx and is labelled as TM incidence, while in the right hand diagram
Einc is entirely transverse to kx generating TE incidence.
252 10 Compact Planar Resonator Arrays

Fig. 10.5 Fabry-Perot resonator antenna a illuminated by a plane wave from a distant test dipole
at r, h, / b PRS comprising a periodic array of slots

For a closed PRS, that is a surface with no slots, the plane wave reflects totally
on the outer surface and hence for TE incidence (see Fig. 10.6):

Hxpþ ¼ 2Hxinc ð10:14Þ


10.4 Leaky-Wave Fabry-Perot Antenna 253

Fig. 10.6 Electromagnetic wave field components at CRA surface when illuminated by a plane
wave and transmission-line equivalent circuits

where Hxpþ denotes the x-directed magnetic field tangential to the outer surface of
the plate. Hxinc is the x-directed magnetic field component of the incident plane
wave. Furthermore for this surface wave directed along kx, E and H are related
through:

Eyinc ¼ Z0TE Hxinc ð10:15Þ

A similar relationship exists for the TM case. Given that the plane wave at the
CRA emanates from a test dipole at distance r from the antenna, the field Einc can
be expressed as:
254 10 Compact Planar Resonator Arrays

jxl0
Einc ¼ a^h  expðjk0 rÞ expðjk0  ^
ar rÞ ð10:16Þ
4pr

The term inside the square brackets represents radiation from a point source, and
the second term ensures that, of all possible plane waves, the plane wave directed
along ^ar is mathematically selected. A similar equation can be formed for the TM
case with Einc directed along ^a/ . It is also helpful here to define Z0TE and Z0TM :
Hence:

Einc Einc xl0


Z0TE ¼ ¼ ¼ ð10:17Þ
Hzinc Hinc cos h kz0

Ezinc Einc cos h kz0


Z0TM ¼ ¼ ¼ ð10:18Þ
Hinc Hinc xe0

For narrow x-directed slots as suggested in Fig. 10.5b the Hxinc field component,
generated by the TE illumination of the PRS, can in turn be associated with trial
magnetic current densities (unknown at this stage) in the antenna slots. These are
denoted by Jmx s
ðxp ; yq Þ on the slot array surfaces at z = 0 in real space, and by
~Jmx
s
ðkxp ; kyq Þ in the plane wave spectral domain. The integers p and q define slot
locations. For mathematical convenience, rather as outlined in Chap. 9, Sect. 9.1,
p and q are allowed to range in value from 1 to þ 1 since those slots outside
the limited number of active examples at the heart of the array contribute nothing to
the results.
The adoption of the infinite array concept in Sect. 9.1 yielded the analytical
benefit of making it possible to model the reflectarray with minimal complexity and
tedious repetition. When the array is of infinite extent its periodicity can be
expressed mathematically through the Floquet Theorem. Furthermore, considerable
mathematical simplification accrues to the CRA analysis presented below by per-
forming the computation in the spectral domain. When all relevant functions are
Fourier transformed into the spectral domain the array can be modelled quite
straightforwardly as summations over identically excited elements. The Floquet
modes are accommodated automatically within the Fourier transforming process. It
is well established in signal theory that an infinite sequence of identical ‘pulses’
transforms into a frequency spectrum of sinusoidal waves. By analogy the periodic
array of identical aperture fields is transformed into Floquet modes, which satisfy
the periodic boundary conditions at the substrate surface.
The x-directed magnetic current density Jmx s
ðxp ; yq Þ in the slots of the PRS
þ
generate a magnetic field Hxs adjacent to the outer surface (z > 0) of the PRS, and

simultaneously a magnetic field Hxs adjacent to the inner surface (z < 0). If the
spectral domain magnetic Green’s function ðG ~  ðkxp ; kyq ÞÞ for the periodic surface is
m
known, then we can write:
10.4 Leaky-Wave Fabry-Perot Antenna 255

 1 X 1 X1
~  ðkxp ; kyq Þ~J s ðkxp ; kyq Þ exp jðkxp x þ kyq yÞ
Hxs ¼ G ð10:19Þ
p1 p2 p¼1 q¼1 m mx

where p1 and p2 are the element to element pitches in respectively the x and
y directions.
In the spectral domain, the above Green’s function can be deduced relatively
simply by employing a transmission line equivalent network (TEN) [10]. The
analogous circuits show downward wave propagation in the negative z-direction in
the radiation space z > 0, and transmission into the substrate z < 0 (Fig. 10.6). The
figure presents TEN’s for both the TM orientated incident plane wave and the TM
alternative. The more general case of oblique incidence is presented in reference
[10]. The transmission lines are presumed to extend to infinity in the +z direction
where they are terminated in Rrad the radiation resistance. In the air region the
transmission line characteristic impedances are identified as Z0TM and Z0TE , while in
the substrate we use Z1TM and Z1TE . At z = 0 on the PRS the scatterers, in our case
slots, are excited by the Hn (TM) and Hη (TE) of the obliquely incident plane wave,
where n is rotated angle d from the x axis. Consequently they can induce magnetic
current densities ~Jmx
TM
and ~Jmx
TE
in the x-aligned slots. Operation in the spectral do-
main is presumed. For slots ~Jmy is usually negligible so we can concentrate on ~ Jmx .
For z-directed propagation, the TEN relates the fields in the interior and the exterior
of the substrate and across the interface at z = 0 [10]. In this representation the
current density excitations at z = 0 can be replaced by the currents ~Imx TM
and ~Imx
TE
,
where ~Imx ¼ ~Jmx L and ~Imx ¼ ~Jmx L. Hence, on applying network theory to the
TM TM TE TE

equivalent circuits in Fig. 10.7, the voltage at the junction at z = 0 is given by:

~inTM ¼ ~Imx
V TM TM
Zin
ð10:20Þ
~in ¼ ~Imx ZinTE
V TE TE

where ZinTE;TM are the parallel impedances of the lines above and below z = 0. These
are respectively Z0TE;TM (see Eqs. (10.17), (10.18)) and
TE;TM
Zsub ¼ jZ1TE;TM tan b1 d ð10:21Þ

and for the substrate region

b1 xlr l0 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Z1TM ¼ and Z1TE ¼ and b1 ¼ x er e0 lr l0 ð10:22Þ
xer e0 b1

On combining Eqs. (10.20), (10.21) and (10.22), in the spectral domain, this
yields the relation:
256 10 Compact Planar Resonator Arrays

Fig. 10.7 Scanned E-plane (a), and H-plane (b) radiation patterns for a slot based CRA with a
dielectric filled resonator (er = 2.2) for which L = 0.6 cm, w = 0.05 cm, p1 = 1 cm and
p2 = 0.3 cm. The CRA is resonant at 11 GHz for d = 0.85 mm. Over the frequency range
depicted the height (d) varies from d  k=2 to d  k=3 (from [6])
10.4 Leaky-Wave Fabry-Perot Antenna 257

h i
~ xs ðkx ; ky Þ ¼  1 kx2 IxTE þ ky2 IxTM ~
H J ðk ; k Þ ð10:23Þ
0 kt2 0 0 ms x y

which is evaluated at z = 0. Referring to the transmission line analogy, at z = 0, we


have:

Ix j0 ¼ Yin VinTE;TM ð10:24Þ

where Vin is the voltage at the surface (Fig. 10.6). To form a Green’s function in the
spectral domain we choose Vin = 1 to represent the ‘point’ source. Hence

1
Ix j0 ¼ Yin ¼ Y0TE;TM ¼ ð10:25Þ
Z0TE;TM

The field due to Ix just above the PRS is therefore


h i
~ þ ðkx ; ky Þ ¼  1 k2 Y TE þ k 2 Y TM
G ð10:26Þ
m x 0 y 0
kt2

Z0TE and Z0TM and hence Y0TE and Y0TM are defined in Eqs. (10.17) and (10.18).
Also
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
kz0 ¼ ðk02  kt2 Þ ¼ ðk02  kx2  ky2 Þ
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð10:27Þ
kz1 ¼ ðk12  kt2 Þ ¼ ðk12  kx2  ky2 Þ
pffiffiffiffi
where k1 ¼ b1 ¼ er k0 .
Similarly, for the magnetic field just below the PRS and within the substrate
(region 1 with relative permittivity = er) the appropriate Green’s function is:
h i
~  ðkx ; ky Þ ¼ 1 jk 2 Y TE cotðkz1 dÞ þ jk 2 Y TM cotðkz1 dÞ
G ð10:28Þ
m
kt2 x 1 y 1

with
xlr l0
Z0TE ¼ ð10:29Þ
kz0

kz0
Z0TM ¼ ð10:30Þ
xer e0

The magnetic fields above and below the PRS can now be expressed in terms of
the unknown slot magnetic current densities. When the boundary conditions are
applied at a reference slot (usually the p = 0, q = 0 slot), that is:
258 10 Compact Planar Resonator Arrays

Hxpþ þ Hxsþ ¼ Hxs



ð10:31Þ

where Hxpþ is the tangential magnetic field on the ‘closed’ metal plate, Hxsþ is the

slot magnetic current induced field just above the slotted surface and Hxs is the
equivalent field just below the surface. At this stage we choose to follow the
Galerkin approach (see Chap. 5) to the implementation of the moment method and
employ entire domain basis functions in the trial form of the ‘unknown’ slot
magnetic current density Jms. Hence

X
N
Jms ðx; yÞ ¼ An Fn ðx; yÞ ð10:32Þ
n¼1

A suitable choice for Fn is [6]:


  
L x2
sin np L
Fn ðx; yÞ ¼ rhffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
i ð10:33Þ
 w 2
p 2 y 2

In the spectral domain the slot magnetic current density can therefore be written
as:

X
N X
N
~Jmx ¼ ~n ðkx ; ky Þ ¼
An F An~fn ðkx Þ~
gn ðky Þ ð10:34Þ
n¼1 n¼1

where:

~fn ðkx Þ ¼ exp jð 2 Þ½npL þ expðjkx LÞðnpL cos npÞ


kx L
ð10:35Þ
ðkx LÞ2 ðnpÞ2

ky w
~gn ðky Þ ¼ J0 ð10:36Þ
2

Hence on applying the Galerkin method the following matrix equation for the
unknown An’s is generated:

½Zmn ½An  ¼ ½Rm  ð10:37Þ

The entries in the right hand column matrix are:

2 ~m ðkx0 ; ky0 Þ
Rm ¼  Eyinc F ð10:38Þ
Z0TE;TM
10.4 Leaky-Wave Fabry-Perot Antenna 259

while the entries in the Zmn matrix are:

1 X 1 X1
~ þ ðkxp ; kyq ÞF
~n ðkxp ; kyq ÞF
~m ðkxp ; kyq Þ
Zmn ¼ G
p1 p2 p¼1 q¼1 m
ð10:39Þ
1 X 1 X1
~  ðkxp ; kyq ÞF
~n ðkxp ; kyq ÞF
~m ðkxp ; kyq Þ
þ G
p1 p2 p¼1 q¼1 m

Once the unknown An’s are determined by solving Eq. (10.37) the field Ey inside
the substrate at the position of the source can be calculated, and by reciprocity,
which involves sampling the cavity field, the far field becomes calculable.

10.4.2 CRA Radiation Characteristics

Typical E-plane and H-plane far-field computations for slot based CRA’s at fre-
quencies near resonance (f = 11 GHz, d = 0.85 mm) have been generated and are
depicted in Fig. 10.7. The calculations apply to a substrate with er = 2.2 and for a PRS
for which L = 0.6 cm, w = 0.05 cm, p1 = 1 cm and p2 = 0.3 cm. For an infinite
parallel plate waveguide with a dielectric filling (er = 2.2), and a plate separation of
0.85 mm, propagation for the TE10 mode cuts-on at 11.89 GHz in the absence of slots.
The slot array lowers the resonance to 11.0 GHz, represented by the blue solid line in
Fig. 10.7a, b. At this frequency the radiation peak is precisely at broadside and is
sidelobe free for both the E-plane and the H-plane, as predicted by Trentini [1]. The
symmetry of Eq. (10.36) suggests that the E-plane and H-plane patterns for the CRA
should be the same. That they are not is due to the adoption of narrow slots to form the
PRS. As we have seen in earlier chapters the E-plane (y-z plane in Fig. 10.5) pattern
for a narrow slot in a half-space is semi-circular, while in the H-lane (x-z plane in
Fig. 10.5) it is a half ‘doughring’ with nulls along the major axis of the slot.
Consequently, in Fig. 10.7 the E-plane pattern exhibits (Fig. 10.7a) much stronger
radiation close to +90° and −90° than occurs in the H-plane (Fig. 10.7b).
More importantly Fig. 10.7 highlights the issue of beam splitting with a CRA, or
leaky wave antenna, and how the split beam maxima are influenced by antenna
thickness (d) and operating frequency. It is evident that the major influence on beam
split angles is the ratio of separation height (d) between the PRS and the ground
plane to the wavelength (k) within the substrate (i.e. d/k) and that in the H-plane it
is possible to locate a maximum beam anywhere between broadside and 75°. In the
E-plane it is predicted that the peaks can scan almost to end-fire, as indicated in
Fig. 10.7a. What is unexpected is that for 2-D antennas of this type so called
‘scanning’ is fundamentally associated with the formation of a split beam, partic-
ularly at scan angles well away from broadside. The mechanism for this behaviour
is suggested in Fig. 10.1. In the H-plane (Fig. 10.1b) a single beam, at broadside, is
formed only at resonance since this ensures that the fundamental parallel plate mode
is cut-off. This in turn implies that the TE mode in the positive and the negative
260 10 Compact Planar Resonator Arrays

x-directions in Fig. 10.1b exhibits zero phase shift. The radiating elements of the
PRS are thus excited in phase so forming a single broadside beam in both the x-
z plane and the y-z plane. At higher frequencies the parallel plate TE mode is no
longer cut-off in the +x and −x directions, for mid-point excitation as shown in
Fig. 10.1. Leaky-wave beams are thus possible in these directions resulting in the
2-D split beam patterns shown in Fig. 10.7. Note that since d and k are interlinked,
and essentially interchangeable in the basic mode of operation of the Fabry-Perot
resonator, beam splitting behaviour similar to Fig. 10.7 can be observed when
thickness adjustments (d) replace frequency [5, 6]. The beam-splitting is particu-
larly associated with the mid-point excitation commonly adopted with the CRA.
One directional scanning occurs if the array can be end fed.
This split beam characteristic represents an awkward feature of the leaky-wave
CRA, which is highlighted in references [5, 6], namely that the desired broadside
pencil-shaped radiating beam degrades rapidly with small changes to cavity height
from the resonant condition. Given that in theory the CRA periodic surface is
presumed to be of infinite extent the PRS could take the form of a circular plate
with radiating elements spreading radially outwards from the centre. Propagation
from a centrally located feed will also spread radially outwards at above resonance
frequencies. In three dimensions, at above cut-off, the split beam maxima of
Fig. 10.7 will occur along radii at all possible azimuth angles thus forming a hollow
conical beam with a circular rim. While some form of frequency scanning seems
possible, the obvious practical implication of conical beam formation, is that the
basic centre fed CRA antenna provides a broadside pencil beam radiation pattern
only over a relatively narrow bandwidth.

10.5 CRA Thinning

A leaky-wave CRA geometry in which the perfectly conducting ground plane is


replaced by a second periodic array, with the characteristics of a high impedance
surface (HIS—see Fig. 10.4b), offers some support that a wider band CRA is
achievable [8]. The introduction of the HIS primarily makes possible a significant
level of antenna thinning. In theory, a leaky-wave CRA can be reduced in thickness
to d  k=7 from the usual k/2 by the adoption of an HIS. The mechanism is made
clear by considering the simplest possible manifestation of this adjustment,
whereby the HIS presents an open-circuit wall to the internal fields the CRA. It is
not too difficult to appreciate that in this case the thickness, for no change in
resonance, is immediately reduced by 50%. By carefully tailoring the surface
impedance it is possible to reduce the thickness even further. Improved bandwidth
accrues from the fact that in the thinned CRA the beam scanning becomes much
less sensitive to d variations and hence to changes in frequency.
A similar scheme, reported in reference [2], describes a planar Fabry-Perot
leaky-wave cavity antenna employing a PRS above an artificial magnetic conductor
(AMC—see Fig. 10.8). The AMC is simply an HIS by another name. For an
10.5 CRA Thinning 261

Fig. 10.8 Schematic of a 2-D


leaky-wave Fabry-Perot
resonator antenna fed by
square patch at centre of
cavity (from [8])

antenna operating at 14 GHz the AMC is produced by printing a closely packed


array of almost square patches (4.1 mm  4.4 mm) on a 1.13 mm thick grounded
substrate (er = 2.2). A highly reflective PRS is adopted to procure high gain. The
upper layer PRS is constructed from almost square patches (10 mm  11 mm) on
an ungrounded substrate similar to that used for the AMC. The antenna is fed from
an off-square patch embedded at the centre of the HIS. The general arrangement is
shown schematically in Fig. 10.8. A full-wave 3-D electromagnetic simulation of
the proposed design revealed an antenna thickness of close to k/4 for resonance at
14 GHz. Modelling and measurements indicate that an optimum gain of 20 dBi at
broadside is potentially available for the linearly polarised beam. The bandwidth
was estimated at 2%. A CRA with similar performance but exhibiting a much
thinner (d = k/10) profile has also been reported. The key to thinning is in the
nature of the HIS which can in principle provide a surface impedance which lies
anywhere between a PEC (d = k/2) and a PMC (d = k/4).

10.6 Bandwidth Enhancement

The narrowness of the bandwidth of the basic CRA is studied in reference [9] where
design guidelines are presented to tailor the CRA’s directivity bandwidth. This is
predicated on a thorough understanding of the role of higher order modes in the
leaky-wave CRA which was achieved by applying reciprocity, and the full-wave
spectral domain periodic analysis, as outlined above, to a 2-D geometry. Improved
bandwidth is possible, not by replacing the conducting ground plane by an HIS as
in Fig. 10.4b, but by introducing a second PRS inserted between the original PRS
and the ground plane forming in effect a pair of coupled Fabry-Perot cavities (c.f.
Fig. 10.8). The technique is not unlike patch antenna bandwidth enhancement
262 10 Compact Planar Resonator Arrays

which involves the stacking of two or perhaps three patch radiators in close
proximity to form a coupled resonator system. Bandwidth improvement follows
separate adjustments of the resonant frequencies of each patch. In the case of the
CRA, bandwidth enhancements of six times that of the single layer LWA, at the
same directivity level, can be accomplished.
High directivity, high gain, beam formation in a two dimensional leak-wave
antenna (2D-LWA) of the Fabry-Perot type is described in references [5–9]. The
basic format of the studied LWA is shown schematically in Fig. 10.8 with a PRS of
square patches on the underside of the upper substrate, separated by a distance
d from an HIS formed from a square array which has twice as many much smaller
square patches distance hd2 above a perfectly conducting ground plane. Actually,
the periodic surfaces representing the PRS and the HIS need not be square patches
but could be crossed dipoles, or crossed slots, or round holes, in a conducting
screen. In [5] the PRS is formed from printed dipoles, in [6] it is constructed from
slots etched into a plane conducting screen, while the antennas in [8, 9] are
developments based on square patches. Bandwidth enhancement, as indicated
above, is linked to the reduced thickness (d) which the HIS enforces.
High gain with improved bandwidth potential can be procured in theory by
omni-directional excitation of a circular plate shaped CRA. In Sect. 10.3, where the
basic mechanism of operation of the PRS antenna is examined (c.f. Fig. 10.1), the
explanation given there for leaky-wave radiation, presumes that the parallel plate
waveguide forming the antenna as a rectangular structure, which is best represented
by the Cartesian system. However, given that the Fabry-Perot cavity is theoretically
infinite in extent [1] this need not be the case. Cylindrical symmetry could equally
be assumed as is done in [5]. In this case the antenna is much more obviously a
travelling-wave device with the field pattern of Fig. 10.1b travelling radially out-
wards from the source. If the source is omni-directional (e.g. a coaxial probe) then
the dominant TM mode propagates radially outwards with equal magnitude in all
directions exciting the PRS as it does so. This operational model is tentatively
hinted at in Fig. 10.3. For an air dielectric the fundamental mode, as in the PRS
antenna case, cuts-on at d * ko/2, and a pencil beam at broadside is formed.

10.7 Reconfigurable PRS for Fixed Frequency Scanning

CRA modifications, discussed in preceding sections of this chapter, were aimed


both at controlling frequency sensitive beam splitting to improve bandwidth, and to
take advantage of it to procure practical beam scanning with a CRA. The discussion
has, thus far, circumvented the idea of beam scanning at a fixed frequency, or in
effect to introduce quasi-phased-array behaviour into the CRA. It is perhaps hardly
surprising, nevertheless, that no sooner had the breakthrough to practical
Fabry-Perot cavity based antennas been achieved [1–9] largely by employing
microwave substrates, that efforts were being made to introduce beam steerability
10.7 Reconfigurable PRS for Fixed Frequency Scanning 263

into the mix. The route to this goal has largely been through the introduction of
‘reconfigurable’ array elements into the PRS [11–16] and/or the HIS [17].
The methodology of electronic control in a CRA for beam scanning is hinted at
in the fundamental Eq. (10.11). However, this equation should be treated with
caution insofar as it is two dimensional and has been formulated using a simple ray
theory and multiple reflections. With q(h) denoting the internal PRS reflection
coefficient at incident angle h (from x) in Fig. 10.1a, it is clear that control of q is
key to pattern modification [7]. Note that the g(h) term is a magnitude factor only—
influencing energy distribution in cavity. It has no influence on phase at the radi-
ating surface. However, in practice control of q has been implemented by the
introduction of a second periodic surface (HIS) either replacing the ground plane or
inserted between the ground plane and the PRS as indicated in Fig. 10.9a.
Reconfigurability has generally been advanced by introducing phase agile cells into
this additional periodic surface rather than the PRS.
Arguably, the introduction of reconfigurable components into PRS antennas
appeared first as a frequency compensation technique [12], whereby PRS directivity
could be maintained over an enhanced tunable frequency range. An antenna
comprising a PRS formed from an array of square patches is described, and fre-
quency adjustment is achieved by introducing a phase-agile surface on a substrate
positioned close to, and just above the ground plane. The geometry is shown
schematically in Fig. 10.9, where the lower periodic surface is represented as being
‘phase agile’. This is achieved by employing split patches, termed reflection cells,
with the gap bridged by one or more varactor diodes (c.f. Fig. 10.9b). The
reflectivity of the reflection cells, and hence the phase, is controlled by the suitably
located varactor diodes. It is reported [12] that CRA gain can be maintained con-
sistently above 10 dBi over a tuned frequency range of 13.5% from 5.2 to
5.95 GHz. The bandwidth is six time larger than for the standard PRS antenna. By
further exploiting the principle enunciated in [7] and encapsulated in Eq. (10.11), it
has been suggested that dynamic beamwidth reconfiguration in PRS antennas is
feasible and a range of techniques are explored in references [13–15].
While the reconfigurable PRS has presented the primary route toward controlling
and optimising beamwidth and directivity, it has also been proposed [15] that PRS
beam positioning or scanning can, in principle, be procured by phase control of a
two element feed antenna embedded in the Fabry-Perot cavity as shown in
Fig. 10.10.
Referring back to Eq. (10.11), it has already been noted that, in the E-plane of
the Fabry-Perot resonator antenna the g(h) term, representing the feed, only indi-
rectly influences the radiation pattern P(h) and consequently, in this plane, phase
changes at the two element feed cannot modify P(h) and the antenna’s broadside
radiation characteristic. Consequently beam scanning in this case must be restricted
to the H-plane with the PRS excited from a linearly polarised feed. While the
simulated beam scanning predictions are presented for the H-plane, that scanning is
restricted to this plane is not made clear in references [14, 15]. It is fairly evident
from Fig. 10.1b that excitement of the PRS array elements is provided by the axial
(z-directed) magnetic field of the close-to-cut-off quasi-TE10 mode in the parallel
264 10 Compact Planar Resonator Arrays

Fig. 10.9 CRA with a phase agile HIS and b typical phase agile cell incorporating varactor diode
switches
10.7 Reconfigurable PRS for Fixed Frequency Scanning 265

Fig. 10.10 Schematic of a high directivity PRS antenna with additional beam scanning in the
H-plane (from [15])

plate waveguide. If the PRS is reconfigurable as suggested in Fig. 10.11, this allows
the surface reflection coefficient to be adjusted by controlling the capacitances of
the varactor diodes. The introduction of a two-element feed can be made to favour
the propagation of this mode in the forward (+z) or backward (−z) directions, thus
providing a scanning mechanism which results in three beam positions in the
H-plane given by the basic leaky-wave equation.
266 10 Compact Planar Resonator Arrays

Fig. 10.11 Schematic depicting diode modified patch element for typical reconfigurable patch
based PRS antenna (from [15])

k
H ¼ cos1 ð10:40Þ
kg

This is confirmed by performing an elementary calculation on the fully


dimensioned PRS antenna in reference [15]. The Fabry-Perot cavity can, simplis-
tically, be viewed as an air-filled parallel plate waveguide with d = 78.4 mm.
Hence for the quasi-TE10 mode we estimate that kc = 156.8 mm, while at 2 GHz,
the free-space wavelength ko = 150 mm. Consequently kg10 = 515.0 mm and from
Eq. (10.12) we have three possibilities for the mainbeam offset from broadside,
namely −17°, 0°, +17°. The paper provides simulated H-plane results showing a
−15°, 0°, +15° beam swing for a low directivity example, and −20°, 0°, +20° for a
high directivity design. Given the assumptions inherent in the use of Eq. (10.11),
this is good agreement. It is perhaps pertinent to note that attempts to achieve
significant beam scanning in the H-plane tend to be incompatible with good PRS
performance in the E-plane. A not unfamiliar ‘trade-off’ which is typical of choices
regularly encountered in electrical engineering.
10.8 Beam Scanning in a Leaky-Wave CRA 267

10.8 Beam Scanning in a Leaky-Wave CRA

Electronic beam scanning in a CRA represents active research at the time of writing
in 2017 but solutions are proving to be elusive. While a range of modifications,
which have been introduced into the original Fabry-Perot resonator, such as addi-
tional periodic layers variously referred to as an HIS, AMC, FSS or metamaterial,
have achieved improved bandwidths, gain and efficiency by controlling
beam-splitting, frequency independent scanning has seemingly remained out of
reach. Nevertheless, comprehensive modelling of the CRA points to the leaky-wave
mechanism as the most promising route to effective non-phased-array beam scan-
ning at a fixed frequency.
Leaky-wave scanning performance of a cavity antenna, based on the application
of reconfiguration to a PRS layer, is partly addressed in [16]. Careful inspection of
Fig. 10.1b reveals that for a non-directional feed, an in-phase, close-to-cut-off,
quasi-TE10 mode field propagates in the PRS antenna to the left and right of the
source; that is in the −z and +z directions. The PRS thus splits into two ‘half-arrays’
whose in-phase radiations add constructively into the free-space region contributing
in the H-plane to the predicted [1] broadside pencil beam. In reference [16] this split
operational mode is emphasised by introducing slightly different array geometries
in the left and right regions of the PRS antenna. In summary, the arrangement has
arguably produced a limited 1-D leaky-wave scan mechanism in the H-plane of the
cavity resonator antenna which was originally presented as a PRS antenna. The
investigation outlined in [16] suggests that more significant and more practical
levels of scanning in cavity resonator antennas require more emphasis on the
leaky-wave mechanism. This is addressed in the next section.

10.8.1 Linear Leaky-Wave Array

The leaky-wave antenna was addressed in several contributions to the literature


after 2010. For example, a 1-D Fabry-Pérot resonator, for which the leaky-wave
beam scanning requirements are clearly expressed, is reported in [17].
Beam-scanning of up to 40° is demonstrated at a single design frequency. The
essence of the method is in the implementation (see Fig. 10.12) of a high impe-
dance surface (HIS) in place of the electrically conducting ground plane, which is
normally encountered in the Fabry-Perot cavity antenna (see Sect. 10.4). This
leaky-wave structure is analysed using the spectral domain moment method and
reciprocity [8] as described in Sect. 10.4.1.
The operational mode of the 1-D leaky-wave antenna can relatively easily be
deduced from further consideration of Fig. 10.1a, b. With reference to Fig. 10.1,
the conducting side walls in Fig. 10.12 are aligned parallel to the x-z plane, on
either side of the feed slot in Fig. 10.1a and above and below the feed in Fig. 10.1b,
with no effect on the internal fields. Consequently, if the HIS in Fig. 10.12 were
268 10 Compact Planar Resonator Arrays

Fig. 10.12 a Schematic of 1-D reconfigurable leaky-wave antenna b detail of HIS cell (from [17])

replaced by a perfect conductor, then the TE10 mode as depicted in Fig. 10.1 would
be set up in the 1-D leaky-wave antenna. However, the HIS in its passive state will
likely exhibit a high impedance approaching that of an open circuit or a perfect
magnetic wall, in which case the TE10 mode becomes a TE1/2,0 at resonance, with
the tangential electric field progressing from a maximum at the HIS to approxi-
mately zero at the PRS.
The situation is represented schematically in the accompanying Brillouin dia-
gram (Fig. 10.13). The lowest modes in the parallel plate waveguide, namely the
TEM mode and the TE10 mode, are required to meet the electromagnetic boundary
conditions presented by the periodic inserts which introduces space harmonics and
slow-wave modes to the right of the ‘light’ line (dashed green trace) and
leaky-wave modes to the left. The original TEM mode can generally be identified
with a slow TM mode (yellow dashed traces in Fig. 10.13), while the empty
waveguide TE10 mode transmutes into the lowest TE slow mode, retaining the
predictable cut-off frequencies associated with the conventional waveguide. So, a
leaky wave antenna array supporting a TE mode with a low cut-off frequency at x1
will radiate at broadside at this frequency, and at endfire just above x4, as dictated
by Eq. (10.40). Thus frequency dependent beam scanning proceeds away from
broadside toward endfire. Any further increase in frequency moves the operational
mode into the slow-wave region where the propagation mechanism becomes a
surface wave, and radiation is suppressed. Actually, in a reconfigurable geometry,
leaky wave radiation is recoverable at higher frequencies by capacitance changes in
10.8 Beam Scanning in a Leaky-Wave CRA 269

the varactor diodes shown in Fig. 10.12. These modify the HIS impedance such
that the operational mode is shifted to a higher resonance or cut-off frequency.
But, more importantly, the Brillouin diagram shows how frequency independent
beam scanning can be realised with a reconfigurable HIS. At the frequency xs for
example, if the varactor diode adjustment is capable of shifting the TE mode
frequency characteristics to the positions shown, then the propagation coefficient
for the leaky wave mode is moved from bp1 to bp2 to bp3. The beam scans from
endfire to broadside. In this 1-D format the antenna is largely omni-directional in
the y-z plane (E-plane) above the parallel plates. The main-beam, which is scanned
in the H-plane (x-z plane), is therefore ‘fan’ shaped. In an LWA designed to operate
at 5.6 GHz, a beam scan range in the H-plane of 8°–30° from broadside is
demonstrated without loss of gain (Fig. 10.14). The influence of HIS changes on
the radiation characteristics of similar leaky-wave CRA’s are also reported in [11].
To avoid beam splitting (see Fig. 10.9) the 1-D structure is end fed from rectangular
waveguide. This represents a complete departure from the normally centre-fed CRA
geometry.
That the 1D leaky wave antenna depicted in Fig. 10.12 possesses the range of
adjustments required to generate effective H-plane beam scanning is demonstrated
in reference [18]. While the evaluation did not include diode switching within the
studied device, it is shown that predictable and reliable positioning of the primary
beam can be produced at a fixed frequency by simple adjustments in the length of
the dipoles forming the PRS (LPRS), and the length of those forming the HIS (LHIS).
It is anticipated that these length modifications could be implemented electronically
by embedded varactor diodes as illustrated in Fig. 10.13. Moment method

Fig. 10.13 Schematic representation of the Brillouin diagram for a parallel plate waveguide
loaded with transverse periodic surfaces
270 10 Compact Planar Resonator Arrays

Fig. 10.14 Computed beam scan angles for 15 GHz LWA’s design to provide 10° beamwidths at
scanned angles of 5° (blue long dashed trace), 20° (green solid trace), 40° (brown dashed trace)
and 60° (red short dashed trace)

computations, which have been performed on several versions of the LWA, are
summarised in Fig. 10.14. The analysis has been confirmed by measurements [18].

10.8.2 Planar Leaky-Wave Array

Largely in parallel with the linear LWA developments described in the previous
section, the introduction of a reconfigurable HIS into a planar antenna geometry has
been reported [19–22] with the aim of forming a reflecting surface for use as a
reflectarray, or as a compact conformal scanned LWA. However, while competing
developments in CRA technology have tended to separate the PRS from the HIS,
within the parallel plate structure, and apply reconfigurability by means of the HIS,
in the cited articles varactor diode control is incorporated into a periodic structure
which notionally combines the two roles. The arrangement is described variously as
a tunable textured ground plane, a tunable impedance surface and an artificial
magnetic conductor. It is depicted schematically in Fig. 10.15.
The periodic surface comprises a regular array of square patches above a con-
ducting ground plane, not unlike CRA’s described earlier in this chapter. However,
the inclusion of shorting posts between the patches and the ground plane makes the
notion of a parallel plate operating mechanism less sustainable. Nevertheless, ref-
erences [21, 22] suggest that the Brillouin diagram for this periodic surface exhibits
10.8 Beam Scanning in a Leaky-Wave CRA 271

Fig. 10.15 Top and side


views of a schematic
representation of a planar
leaky-wave array antenna
formed from a tunable
textured surface

similar characteristics to those depicted in Fig. 10.13. The frequency characteristics


of this surface (Fig. 10.15) can be constructed by viewing it in LC terms.
Capacitance per unit area (C) is furnished by the patch-to-patch gaps while in-
ductance per unit area (L) is created by the metal posts. Consequently, the surface
impedance can be equated to that of a matrix of parallel resonant LC circuits. In
simplistic terms it can be expressed as [21]:

jxL
Zs ¼ ð10:41Þ
ð1  x2 LC Þ

Close to resonance, the surface impedance Zs tends towards infinity and the
periodic structure behaves as an HIS, or artificial magnetic conductor, with a zero
reflection phase shift. The term textured surface has been introduced to reflect the
non-uniformity of the Zs for this structural format.
As with the linear leaky-wave antenna described in the previous section, scan-
ning can be either frequency dependent, or stepped at fixed frequency by varactor
272 10 Compact Planar Resonator Arrays

diode adjustment. In the former case, increasing frequency from the dominant TE
mode cut-off results in growing bp (Fig. 10.13) until the chosen curve crosses the
‘light’ line (vp = c). In this frequency range the TE mode is leaky and according to
Eq. (10.40) the primary radiating beam scans from broadside to end-fire. However,
the inclusion of varactor diodes as suggested in Fig. 10.15, confers reconfigura-
bility on the periodic surface, by changing patch-to-patch capacitance and hence
resonance (see Eq. (10.41)). This manifests itself as repositioning of TE mode
cut-off frequencies from x1 to x4 in Fig. 10.13. So, in much the same manner as for
the linear leaky-wave structure, at a fixed frequency xs (say) diode adjustment can
move the phase shift coefficient bp from bp1 to bp2 to bp3, thus scanning the
leaky-wave beam. Scanning in the forward direction of the incident mode is clearly
demonstrated in reference [21]. Furthermore, it is also suggested there that the
textured surface of Fig. 10.15 also supports a backward TE mode which enables
scanning into the backward sector from 0° to −50°.
In practical examples of this textured surface leaky-wave antenna, excitation of
the surface relies on either an off-set flat notch shaped feed located just above it
(by *ko/50) [21], or on a flat bow-tie dipole antenna [22]. The scanning is 1-D, in
either the x-z or y-z plane depending on the location of the feed. A successful
graduation to 2-D scanning had not been reported by the end of 2017.

10.9 The Role of Metamaterials in Compact Slot Arrays

Arguably, without being too disingenuous, there was a period over the past ten to
twenty years when the route to virtually guaranteed paper publication on any
antenna topic was to include the word ‘metamaterial’ in the title. Actually many
such papers [23–27] did not include metamaterials as strictly defined—namely a
material, or medium, displaying both negative permittivity and negative perme-
ability, sometimes referred to as a negative refractive index material (see below).
Most of these contributions describe the employment of periodic surfaces in novel
antenna forms, using the unique features of such surfaces, in particular frequency
selectivity and reconfigurability, as we shall see in the following sections.

10.9.1 Arrays as Passive Periodic Structures

Before we consider the metamaterial role it is important that we fully appreciate the
EM characteristics of a regular passive periodic surface from an antenna perspec-
tive. As has already been observed, the primary difference, electromagnetically,
between an array antenna and passive periodic surface used as a scatterer, such as a
frequency selective surface (FSS/metamaterial), is in the nature of the energy
source. For the array antenna, as we have seen, each element of the array is an
individual source of EM radiation (Chap. 6, Fig. 6.1), whereas the passive FSS is
10.9 The Role of Metamaterials in Compact Slot Arrays 273

swept by an incoming plane wave (Fig. 10.16) which excites simultaneously the
periodically spaced elements. This is, of course, closely analogous to the leaky
wave array discussed in the preceding section of this chapter. Hence, as elements of
an array, the FSS apertures combine following antenna array rules, thus generating
scattered waves in the forward (transmitted) and backward (reflected) directions. In
practice the periodic array would normally be planar. However, since the planar
array is merely a combination of parallel linear arrays, we can deduce the essence of
the behaviour of a passive periodic surface by limiting our attention to a linear
geometry, as we have done for the array antenna analysis in Chap. 6.
With reference to Fig. 10.16, it is useful initially to consider the limiting cases of
EM wave incidence on the conducting screen containing a periodic array of
apertures of arbitrary size, usually ranging from zero to half-wavelength long in
their major dimension. For infinitesimal slots the screen is perfectly conducting
everywhere, and by Snell’s laws leads to a single reflected wave equal in magnitude
to the incident wave and reflected in the direction −/ when the incident angle, to
the surface normal, is +/. The opposite limit, which is rather meaningless practi-
cally, involves apertures which are large enough to totally replace the screen. In this
case, there is a single transmitted wave at +/ and no reflected wave.
The more practical large iris limit is the half-wavelength option, in which case a
transmitted wave and a reflected wave is formed with essentially equal magnitude.
At resonance each iris radiates equally into the forward and backward half-spaces.
If we limit our attention to a two dimensional scenario, these re-radiated directions
can be deduced from the linear array theory presented in Chap. 6. For the incident

Fig. 10.16 Plane wave scattering at a passive metallic screen supporting a periodic array of
apertures separated by distance d
274 10 Compact Planar Resonator Arrays

wavefront direction depicted in Fig. 10.16 (dark blue solid trace below the array),
the element to element phase shift w (Eq. 6.2) is now dictated by the direction of
the incoming phase front. For a passive surface there is no independent phase delay
or advance and so a ¼ 0 . Equation (6.2), therefore, takes the following form for a
uniform periodic array excited by an incident plane wave at incident angle /:

w ¼ k0 d sin / ð10:42Þ

For the n-th element we therefore have:

nw ¼ nk0 d sin / ð10:43Þ

where n is the element number and d is the element spacing.


For scattering into the forward half-space (z-positive) the periodic surface acts
like a uniform array for which (in two dimensions) the forward scattered radiation
can be expressed by Eq. (6.8) with w given by Eq. (10.42). If d k0 =2 the pattern
exhibits a single forward scattered beam in the direction /. Perhaps, unsurprisingly,
this beam is just the continuation directionally of the incoming plane wave, but
attenuated by the apertures as explained above. If d [ k0 =2 scattering conditions
become significantly more complicated due to the onset of grating lobes (see
Chap. 6).
Scattering into the backward half-space can be treated in a similar manner, but
now scattering is in the –z direction. From a cursory examination of Fig. 10.16, it is
clear that the progressive delay nw along the array is quite independent of whether
or not the scattering is into the forward or backward half-spaces. Consequently, if
d k0 =2 then a single reflected beam is formed, as dictated by Eq. (6.8), in the
backward space, propagating in the −/ direction (green wavefront). The power
density in this reflected beam will be equal to, or less, than that of the incident beam
depending of the size of the major dimension of the apertures forming the array.
Directionally the system is in accord with Snell’s laws.

10.9.2 FSS as a Metamaterial

In preceding sections of this chapter it has generally been assumed that the cavity
resonator antennas, which have been chosen for study, have incorporated periodic
surfaces formed from scattering elements of relatively regular shape (Fig. 10.17).
For passive arrays of such patches or apertures at microwave and millimetre-wave
frequencies it is evident, as demonstrated above, that conventional antenna array
theory, including Snell’s laws comprehensively define the surface characteristics.
However the passive surface need not necessarily be composed of regularly shaped
scatterers and an indication of the range of possible shapes which could be used in
appropriate circumstances are illustrated in Fig. 10.18.
10.9 The Role of Metamaterials in Compact Slot Arrays 275

Fig. 10.17 Typical FSS formats for CRA’s with (a) and (b) forming reflecting surfaces while
(c) and (d) present transparent surfaces

Periodic surfaces incorporating these more complex scatterers sometimes get


classified as metamaterials. The accepted definition for such a material (usually
artificial periodic structures in the form of 2-D surfaces or 3-D volumes) is that
macroscopically it displays permittivity and permeability values which are simul-
taneously negative. The non-simultaneous case is not possible since for any
homogeneous isotropic medium we must have from Maxwell’s equations:
rffiffiffiffiffiffiffiffiffi
lr l0 c
g¼ and v ¼ pffiffiffiffiffiffiffiffi ð10:44Þ
er e0 l r er

where η is the characteristic impedance and v is the EM velocity in the medium.


Such a choice would lead to imaginary values for the characteristic impedance and
the EM wave velocity in the medium – a wholly impractical outcome. Interestingly
no double negative materials have been discovered in nature. Confusion arises with
these materials in the engineering literature because complex periodic structures are
often erroneously termed metamaterials even when they do not possess the above
property. For example, a relatively simple capacitive/inductive wire-grid combi-
nation [23] is classed as a metamaterial, although it is not, while the term is also
linked to electromagnetic band gap (EBG) structures, metallodielectric EBG for-
mations [24], metasurfaces and metastructures [25], artificial magnetic conductors
(AMC) [2, 26], all of which have roles in compact antenna development [27], but
not as metamaterials.
276 10 Compact Planar Resonator Arrays

Fig. 10.18 Elemental scatterers for periodic surfaces described as metamaterials fabricated on
copper coated planar dielectric substrates. a Centre connected, b loop types, c patch types,
d combinations

10.9.3 Negative Refractive Index

Unfortunately, considerable scientific ‘hype’ has been attached to the metamaterial


because of its potentially huge engineering significance. From an antenna per-
spective it has now impacted on the microwave/millimetre-wave scene as unwanted
‘baggage’ [28–31]. Claims for properties such as
• Negative refractive index n
• Propagation phase of EM wave advances rather than delays
10.9 The Role of Metamaterials in Compact Slot Arrays 277

• Evanescent waves can grow rather than attenuate


• For an electromagnetic wave in a metamaterial the electric field vector, the
magnetic field vector and the propagation direction vector form a left-handed
triplet
all of which seemingly contravene the Snell’s Laws and the Maxwell equations,
are perplexing. Fortunately, these claims have recently been comprehensively
unpicked [32].
The original source of the negative refractive index ‘conundrum’ is attributed to
a 1968 paper [33]. In it the author, V.J. Veselago, explored theoretically the notion
of an artificial material possessing both negative permittivity and negative perme-
ability. Referring to Fig. 10.19, we know that if the input medium is air (say), then
for a regular dielectric (case (a)):
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
vair c 1 lr l0 er e0 pffiffiffiffiffiffiffiffi
nsolid ¼ ¼ ¼ pffiffiffiffiffiffiffiffiffi
¼ lr er ð10:45Þ
vsolid vsolid l0 e0 1

Fig. 10.19 Plane wave


refraction at a smooth flat
interface between materials
with differing refractive
index. a Regular materials,
b regular material in contact
with metamaterial
278 10 Compact Planar Resonator Arrays

For case (b), which replaces the regular solid with a metamaterial we obtain:
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
vair c 1 ðlrm l0 Þðerm e0 Þ pffiffiffiffiffiffiffiffiffiffiffiffiffi
nmeta ¼ ¼ ¼ pffiffiffiffiffiffiffiffiffi
¼  lrm erm ð10:46Þ
vmeta vmeta l0 e 0 1

if the negative root is chosen as in reference [33]. For the metamaterial the relative
permittivity and permeability are respectively −erm and −lrm. The positive root
would imply that Eqs. (10.45) and (10.46) are identical and that EM refraction at
the metamaterial surface is no different to the regular material, which is hardly an
advance. If substantiated, the electrical ramifications of the negative choice in
Eq. (10.46) are not insignificant, and hence the Veselago paper caused quite a ‘stir’.
For plane wave refraction at a plane interface between two regular materials, the
angle of refraction (ht), which is intrinsic to the design of optical devices, is given
by Snell’s law, namely:

sin hi v1 c pffiffiffiffiffiffiffiffi
¼ ¼ ¼ lr er ¼ nsolid ð10:47Þ
sin ht v2 vsolid

For case (b) we obtain:

sin hi v1 c pffiffiffiffiffiffiffiffiffiffiffiffiffi
¼ ¼ ¼  lrm erm ¼ nmeta ð10:48Þ
sin ht v2 vmeta

In other words, for a metamaterial interface, the refraction angle (ht) is the
opposite sign to the corresponding angle in a regular set up (see Fig. 10.19).
Since the Snell’s laws arise by enforcing the electric and magnetic field
boundary conditions for the plane waves at the regular interface (Fig. 10.20a) then

Fig. 10.20 Surface reflection and refraction at (a) a conventional interface and (b) an interface
qffiffiffiffiffiffiffi
with a metamaterial. In both cases n1=2 ¼  ll2 ee21
1
10.9 The Role of Metamaterials in Compact Slot Arrays 279

the same must be true for the metamaterial interface. The situation is illustrated in
Fig. 10.20.
Figure 10.20a depicts a mirror flat interface between two conventional materials
(1) and (2) for which the permittivities are respectively e1 and e2, and the perme-
abilities are respectively l1 and l2. Satisfying the electromagnetic field boundary
conditions at the interface at z = 0 leads to Snell’s laws (Appendix A), which are
represented in vector form in the figure. The directional vectors ki1, kr1 and kt2 must
obey the specified angular relationships while the field vectors Ei, Hi, Er, Hr, Et and
Ht automatically satisfy the boundary conditions. Power flow requirements dictate
that for the incident, reflected and transmitted plane waves E, H and k obey the
clockwise rule (Appendix A) with the Poynting vector, and hence power flow,
directed along the relevant k vector. Note that the boundary conditions must be met
everywhere over the z = 0 interface at any given instant in time. This is only
possible if

v1 tan ¼ v2 tan ð10:49Þ

It is not difficult to show that adherence to the Snell’s laws ensures that this is the
case.
So, does case (b) in Fig. 10.20 also meet all of these fundamental physical
constraints? The negative refractive index for the metamaterial requires that the
refractive wave is directed along kt2 at an angle –ht below the z-axis. However, in
addition a negative refraction index implies that, assuming for simplicity that region
(1) is air or vacuum:

x pffiffiffiffiffiffiffiffiffiffiffiffiffi xn1=2
kt2 ¼ lmr emr ¼  ð10:50Þ
c c

The significance of Eq. (10.50) is that kt2 is directed towards the interface. So
the field boundary conditions are in accord with case (a) with the tangentially
directed wave components on each side of the interface meeting the dictates of
Eq. (10.50). However, while the boundary conditions are seemingly satisfied for the
negative index metamaterial, power flow directions are puzzling. The directions of
Et and Ht in case (b) dictate that the Poynting vector S ¼ E  H and hence power
flow is in the direction opposite to kt2. While for power flow normal to the interface
energy conservation is satisfied, for tangential power flow in the z = 0 plane the
power flow in the regular medium (1) opposes the flow in the metamaterial (2),
which suggests that the negative index material is not physically possible. This
embodiment of the refracted wave has been compared to a backward wave which is
a possible solution to the Maxwell equations in periodic structures. But in backward
wave modes, such as those that occur in backward wave microwave oscillators, for
example, power flow everywhere within the modal field structure propagates with
the same group velocity, which is in opposition to the phase velocity direction –
hence the description ‘backward wave’. The electromagnetic field solution in case
(b) of Fig. 10.20 contravenes this behaviour.
280 10 Compact Planar Resonator Arrays

10.9.4 The Flat Lens

The ‘coup de gras’ for artificial metamaterials displaying negative refractive index
is arguably the ‘flat lens’ proposed by Veselago [33], although many other practical
anomalies have been explored and discarded [32]. These intensive studies suggest
that most metamaterial claims are identifiable with established properties of peri-
odic structures.
The Veselago flat lens is viewed in conjunction with a conventional lens per-
forming the same focusing role in Fig. 10.21. For the conventional lens the
focusing mechanism is easily explained. A source at point (A) in Fig. 10.21a is
imaged at point (B) by ensuring the electrical path lengths between (A) and (B) are
the same, balancing electrical path lengths in the air with path lengths in the lens of
lower phase velocity. In addition the lens surface is parabolically shaped to ensure
that the glass/air refractive exit angles match the entry angles. The conventional
lens is predominantly used at optical frequencies and consequently it is usually
made of hard transparent material such as glass with lr = 1. For the flat lens this
restriction does not necessarily apply. It is generally imagined to operate at below
light frequencies as a 3-D periodic structure for which lr and er are negative and
greater than unity. The operational principles of the flat lens are not too different to
those of the conventional lens. The electrical path lengths from the source at (A) to
the image at (B) must be the same for all rays, while the negative refraction angles
must be arranged to secure a focused image on the opposite side of the lens as
suggested in Fig. 10.21b. If the path between (A) and the secondary focus at (C) is
examined for both an edge ray, and a straight through ray, then to arrive at point

Fig. 10.21 Conventional


optical lens compared with
‘flat’ lens. For a glass lens in
air (lr = 1, er = 1) and for
convex lens (lr = 1, er > 1)
and for b the flat lens in air it
has lr < 0, er < 0
10.9 The Role of Metamaterials in Compact Slot Arrays 281

(C) in phase requires that the additional phase delay in the air for the edge ray must
be negated by an extra phase advance in the metamaterial. A similar requirement
applies to all other rays. Equations (10.48) and (10.50) suggest that this is possible
for phase. The difficulty with this argument is that it omits the time element. Phase
coherence at (C), and hence at (B), must occur at the same instant of time, at all
times. This implies that the additional time delays in the air for the longer paths
must be balanced by time advances in the metamaterial, and this is synonymous
with negative time or time reversal! In real physical structures and systems, at least
in this corner of the universe, cause and effect is sacrosanct, and hence the flat
metamaterial lens is not possible.

10.10 Chapter Summary

The Fabry-Perot resonator antenna or cavity resonator antenna in its most basic
form, is shown to be a leaky parallel plate waveguide operated close to the cut-off
of the fundamental TE mode. It is examined initially from a plane wave ray per-
spective and to advance the analysis the periodic waveguide wall is modelled as a
uniform and partially reflective surface (PRS). This leads to simple but actually
rather useful estimates of directivity and in particular how it relates to the reflec-
tivity of the periodic surface. This type of antenna is usually realised as a copper
clad microwave substrate with one surface made periodic by the presence of a
suitably designed array of radiating patches or radiating slots.
More generally, the CRA is shown to represent a class of compact planar
antenna which includes the partially reflective surface (PRS) antenna, the
Fabry-Perot cavity antenna, and the leaky wave CRA. All of these forms can be
modelled by the reciprocal periodic moment method which is fully developed in
Sect. 10.4. Supported by evidence from the literature, modelling demonstrates that
high gain and/or directivity is available from a PRS antenna, but with very restricted
bandwidth. Inconvenient beam splitting and scanning occur as the operating fre-
quency is raised above cut-off. Nevertheless bandwidth enhancement is an active
pursuit.
Splitting of the primary radiated beam also occurs with centre fed leaky-wave
antennas if the applied frequency is above the fundamental mode cut-off value.
Consequently, as with the PRS antenna, beam scanning using frequency is prob-
lematic. However, beam scanning remains a goal of CRA development and it is
shown how this can be achieved, in 1-D embodiments, by using end-fed arrays
incorporating one or more additional reconfigurable periodic layers in the form of
high impedance surfaces (HIS). The planar leaky-wave antenna incorporating a
textured surface has not yet demonstrated 2-D scanning although it possesses the
potential to do so. The conundrum then remains of applying leaky-wave methods to
secure 2-D beam scanning from a compact planar CRA. To do so will require novel
feed arrangements which probably means abandonment of conventional centre fed
geometries associated with the CRA. With 1-D scanning the complex diode
282 10 Compact Planar Resonator Arrays

switching algorithm, which is required to control the multi-diode network, becomes


much more multi-layered and difficult where 2-D scanning in arbitrary directions is
demanded.
Complex periodic surfaces and structures, for which element sizes and spacings
are much less than the wavelength of any penetrating EM waves, are increasingly
being introduced into the literature. They are usually correctly referred to as arti-
ficial dielectrics. Unfortunately, in some contributions they have simultaneously
been described as metamaterials which is also the generic term for so called neg-
ative index materials. As is indicated in this chapter the latter materials have never
been realised in practice. It is to be hoped that as long as new uses of periodic
structures are clearly linked to their periodic behaviour, the metamaterial descrip-
tion will gradually lose its power to confuse.

References

1. G. von Trentini, Partially reflective sheet arrays. IRE Trans. Antennas Propag. 4(4), 666–671
(1956)
2. A.P. Feresidis, G. Goussetis, S. Wang, J.C. Vardaxoglou, Artificial magnetic conductor
surfaces and their application to low-profile high-gain planar antennas. IEEE Trans. Antennas
Propag. 53(1), 209–215 (2005)
3. N. Guerin, S. Enoch, G. Tayeb, P. Sabouroux, P. Vincent, H. Legay, A metallic Fabry-Perot
directive antenna. IEEE Trans. Antennas Propag. 54(1), 220–224 (2006)
4. A. Foroozesh, L. Shafai, Investigation into the effects of the patch-type FSS superstrate on the
high-gain cavity resonance antenna design. IEEE Trans. Antennas Propag. 58(2), 258–270
(2010)
5. T. Zhao, D.R. Jackson, J.T. Williams, H.-Y.D. Yang, A.A. Oliner, 2-D periodic leaky-wave
antennas-part I: metal patch design. IEEE Trans. Antennas Propag. 53(11), 3505–3514 (2005)
6. T. Zhao, D.R. Jackson, J.T. Williams, 2-D periodic leaky-wave antennas-part II: slot design.
IEEE Trans. Antennas Propag. 53(11), 3515–3524 (2005)
7. A.P. Feresidis, J.C. Vardaxoglou, High gain planar antenna using optimised partially
reflective surfaces. IEEE Proc. Microwaves Antennas Propag. 148(6), 345–350 (2001)
8. C. Mateo-Segura, G. Goussetis, A.P. Feresidis, Sub-wavelength profile 2-D leaky-wave
antennas with two periodic layers. IEEE Trans. Antennas Propag. 59(2), 416–424 (2011)
9. C. Mateo-Segura, A.P. Feresidis, G. Goussetis, Bandwidth enhancement of 2-D leaky-wave
antennas using double-layer periodic surfaces. IEEE Trans. Antennas Propag. 62(2), 586–593
(2014)
10. T. Itoh, W. Menzel, A full-wave analysis method for open microstrip structures. IEEE Trans.
Antennas Propag. AP-29(1), 63–68 (1981)
11. L. Zhou, H. Li, Y. Qin, Z. Wei, C.T. Chan, Directive emissions from subwavelength
metamaterial-based cavities. Appl. Phys. Lett. 86, 101101-1–101101-3 (2005)
12. A.R. Weily, T.S. Bird, Y.J. Guo, A reconfigurable high–gain partially reflecting surface
antenna. IEEE Trans. Antennas Propag. 56(11), 3382–3390 (2008)
13. T. Debogovic, J. Perruisseau-Carrier, J. Bartolic, Partially reflective surface antenna with
dynamic beamwidth control. IEEE Antennas Wirel. Propag. Lett. 9, 1157–1160 (2010)
14. T. Debogovic, J. Bartolic, J. Perruisseau-Carrier, Array-fed partially reflective surface antenna
with dynamic beamwidth control and beam-steering. Paper presented at the 6th European
Conference on Antennas and Propagation (EuCAP 2012, Prague, 2012)
References 283

15. T. Debogovic, J. Perruisseau-Carrier, Array-fed partially reflective surface antenna with


independent scanning and beamwidth dynamic control. IEEE Trans. Antennas Propag. 62(1),
446–449 (2015)
16. L.-Y. Ji, Y. Jay Guo, P.-Y. Qin, S.-X. Gong, R. Mittra, A reconfigurable partially reflective
surface (PRS) antenna for beam steering. IEEE Trans. Antennas Propag. 63(6), 2387–2395
(2015)
17. R. Guzman-Quirós, J.L. Gómez-Tornero, A.R. Weily, Y.J. Guo, Electronically steerable 1-D
Fabry-Perot leaky-wave antenna employing a tunable high impedance surface. IEEE Trans.
Antennas Propag. 60(11), 5046–5055 (2012)
18. M. Garcia-Vigueras, J.L. Gomez-Tornero, G. Goussetis, A.R. Weily, Y.J. Guo, 1D-leaky
wave antenna employing parallel-plate waveguide loaded with a PRS and HIS. IEEE Trans.
Antennas Propag. 59(10), 3687–3694 (2011)
19. D. Sievenpiper, J. Schaffner, J.J. Lee, S. Livingston, A steerable leaky wave antenna using a
tunable impedance ground plane. IEEE Antennas Wirel. Propag. Lett. 1, 179–182 (2002)
20. D. Sievenpiper, J. Schaffner, H. Jae Song, R.Y. Loo, G. Tangonan, Two-dimensional beam
steering using an electrically tunable impedance surface. IEEE Trans. Antennas Propag. 51
(10), 2713–2722 (2003)
21. D. Sievenpiper, Forward and backward leaky wave radiation with large effective aperture
from an electronically tunable textured surface. IEEE Trans. Antennas Propag. 53(1), 236–
247 (2005)
22. F. Costa, A. Monorchio, S. Talarico, F.M. Valeri, An active high-impedance surface for
low-profile tunable and steerable antennas. IEEE Antennas Wirel. Propag. Lett. 7, 676–680
(2008)
23. A. Ourir, S.N. Burokur, A. de Lustrac, Phase-varying metamaterial for compact steerable
directive antennas. Electron. Lett. 43(9), 493–494 (2007)
24. Y.R. Lee, A. Charaya, D.S. Lockyer, J.C. Vardaxoglou, Dipole and tripole metallodielectric
photonic bandgap (MPGB) structures for microwave filter and antenna applications. Proc.
IEEE Optoelectron. 127(6), 395–400 (2000)
25. H.L. Zhu, S.W. Cheung, X.H. Liu, T.I. Yuk, Design of polarisation reconfigurable antenna
using metasurface. IEEE Trans. Antennas Propag. 62(6), 2891–2898 (2014)
26. Y. Zhang, J. von Hagen, M. Younis, C. Fischer, W. Wiesbeck, Planar artificial magnetic
conductors and patch antennas. IEEE Trans. Antennas Propag. 51(10), 2704–2712 (2003).
(Special Issue on Metamaterials)
27. S. Enoch, G. Tayeb, P. Sabouroux, N. Guerin, P. Vincent, A metamaterial for directive
emissions. Phys. Rev. Lett. 89(21), 2139–2142 (2002)
28. J.B. Pendry, Negative refraction makes perfect lens. Phys. Rev. Lett. 85(18), 3966–3969
(2000)
29. J.B. Pendry, A.J. Holden, W.J. Stewart, I. Young, Extremely low frequency plasmons in
metallic mesostructure. Phys. Rev. Lett. 76(20), 4773–4776 (1996)
30. J.B. Pendry, A.J. Holden, D.J. Robbins, W.J. Stewart, Magnetism from conductors and
enhanced non-linear phenomena. IEEE Trans. Microwave Theor. Tech. 47(11), 2075–2084
(1999)
31. R.A. Shelby, D.R. Smith, S. Schultz, Experimental verification of a negative index of
refraction. Science 292, 77–79 (2001)
32. B.A. Munk, Metamaterials: Critique and Alternatives (Wiley, New York, 2009)
33. V.G. Veselago, The electrodynamics of substance with simultaneously negative values of e
and l. Sov. Phys. Usp. 10(4), 509–524 (1968)
Chapter 11
Retro-directive Compact Array
Antennas

11.1 Retro-directive Tracking—Basic Concept

There are essentially three routes to tracking a moving signal source, such as a
satellite, with array technology. Firstly, in a full blown phased array antenna with a
transmit/receive unit behind each element of the array, feeding into a high speed
digital signal processor. In this case the antenna radiation pattern can be configured
almost instantaneously into any form required. This is the expensive option. An
undoubtedly cheaper alternative is to use a form of monopulse tracking commonly
seen in radar and in parabolic reflector tracking systems [1, 2]. But this is also
technically challenging for commercial applications aimed at satellite/mobile/
wireless communications.
A third form of tracking, which has been given impetus by recent reconfigurable
developments in array antennas, is provided by the ‘retro-directive’ array [3, 4]. The
principle is not new. Retro-direction, whereby an incoming radar signal, on striking
a target is returned back in the direction from which it came, is achieved (see
Chap. 1) by ‘corner reflectors’, often seen on the masts of yachts [5]. The device
increases the radar cross-section of the target thereby enhancing its detectability
against the background of the sea. Arguably the retro-direction in an array offers the
most cost effective option for emerging wireless systems.
The retro-directive array topology can be further sub-divided into two basic
classifications, depending on whether or not the front-end transmit/receive devices
are ‘mixer’ based, or ‘phase-lock-loop (PPL)’ based. The essence of the two
alternatives is outlined below.

© Springer Nature Switzerland AG 2019 285


A. J. Sangster, Compact Slot Array Antennas for Wireless Communications, Signals
and Communication Technology, https://doi.org/10.1007/978-3-030-01753-8_11
286 11 Retro-directive Compact Array Antennas

11.2 Retrodirective Implementation Through Mixing

The existence of retro-directive antenna action lies in the nature of travelling waves
in free space and on how this behaviour is expressed mathematically. An electro-
magnetic wave travelling away from a source in the direction r varies with time and
with distance from the source. At a single frequency x the variation will be
sinusoidal and this can be expressed as:

E ¼ A expðjxtÞ expðjk0 rÞ ð1Þ

where A is an arbitrary amplitude parameter representing the source signal mag-


nitude and k0 = x/c, where c is the speed of light. The above equation can be
re-written in the form:
r
E ¼ A exp jxðt  Þ ð2Þ
c

which has the functional form of a travelling wave moving in the direction r at
velocity c. Now for a wave of the same frequency x, travelling in the opposite
direction (−r) with amplitude A0 the electric field can be expressed as:
r
E 0 ¼ A0 expðjxtÞ expðjk0 rÞ ¼ A0 exp jxðt þ Þ ð3Þ
c

and E 0 is the retro-directive version of E. The fundamental difference between E 0


and E lies in the exp(jk0r) factor, which is the complex conjugate of exp(−jk0r). In
communication terms this implies that in order to transmit an electromagnetic wave
back toward the source of the received wave, the phase of the incoming wave must
be conjugated before retransmission.
Interestingly it is a property of signal mixing that when a received signal is mixed
with the output of a reference local oscillator at exactly twice the frequency of the
incoming signal (or the carrier frequency if modulated) the resultant difference signal
is the complex conjugate of the original [3–10]. This is illustrated in Fig. 11.1 where
a conjugating mixer circuit is attached to one element of a supposed array via a
microwave circulator (directional filter). All other elements would be backed by
exactly similar circuits all fed from the common local oscillator. It is easy to see that
the transmitted (difference) signal is the conjugate of the incoming wave except for
the fixed phase /0 introduced by the local oscillator. This phase is not relevant to the
operation of the circuit. The phase / occurs when the incoming wave direction is at
an angle h to the array normal. It is basically the phase delay of the received signal
for two adjacent elements separated by a distance d, and is given by:

/ ¼ k0 d sin h ð4Þ

So phase conjugation occurs automatically for any angle of arrival of the


received wave at the retrodirective array. A significant disadvantage of this circuit
11.2 Retrodirective Implementation Through Mixing 287

Fig. 11.1 Phase conjugation through mixing

arrangement, apart from the requirement for a local oscillator at double the oper-
ating frequency, is the presence of the circulator. This has to be a high quality
device, to minimise tracking errors, and at microwave frequencies such a device is
very expensive. Furthermore, this cost is multiplied by the number of elements in
the array.
One way around the circulator difficulty is to provide separate transmit and
receive arrays, as advocated by Myamoto [9, 10] which allows the circulator to be
replaced by a ring hybrid coupler, or rat-race. However, the requirement to
duplicate the transmitter and receiver array hardware is clearly a disadvantage. The
Myamoto solution requires up and down links at different frequencies which, in
turn, demands frequency doubling and frequency shifting circuitry in the device
being tracked, such as a satellite. This could present unwelcome hurdles in certain
applications. Such difficulties have led to the development of phase locked loop
(PLL) based conjugating circuits originally proposed by Ghose [4], and later pur-
sued in Refs. [11–16]. These are discussed in the next section.
288 11 Retro-directive Compact Array Antennas

11.3 Conjugation Employing PLL’s

The earliest example of a conjugating circuit incorporating a phase comparison


technique is attributable to Chernoff [6], with the basic circuit shown in Fig. 11.2.
The voltage controlled oscillator (VCO), which is adjusted by a voltage signal
emanating from the phase comparator (at B) and passed through a loop filter and a
voltage amplifier. The VCO supplies the transmit antenna with a signal of mag-
nitude F1 and frequency x1, and a presumed phase shift of /. Part of the VCO
output is mixed with a signal from the receive antenna (F2 with phase shift U) and
the summed output component is split at point A. From there a component (divided
by n) is fed back to the phase comparator while the other component is demodu-
lated to extract data. That this circuit produces phase conjugation is easily
demonstrated as follows. At position A, the up-converted RF signal can be
expresses as

V ¼ VA expðx1 t þ / þ x2 t þ UÞ ð5Þ

where VA is an arbitrary amplitude factor, and / and U are transmit and receive
phases associated with a main beam at an angle relative to the array normal. If we
assume (Fig. 11.2) that for transmit and receive magnitudes F1  F2 then when the
loop is phase locked we must have U = −/, i.e. phase conjugation. Note that this
phase condition for phase conjugation requires that phase locking is performed on
the sum signal from the mixer at A. Physically, this means that at phase lock, the
signal transmitted by the array is forced to return in the direction of the incoming
received signal—or expressed another way the primary beam of the array is forced
to be pointing at the source. The second mixer at the lower left of the diagram

Fig. 11.2 Phase conjugation circuit employing phase comparison


11.3 Conjugation Employing PLL’s 289

provides heterodyning of the received signal at A with a local oscillator signal at


frequency x0, to produce an intermediate frequency (IF) signal modulated by the
required system data.
A marginally more efficient circuit is presented in Fig. 11.3 [11–14]. In the
version described in Ref. [12], the circuit is implemented around a 200 kHz phase
detector shown on the left of the circuit. The input from a 10 MHz crystal stabilised
oscillator feeds one port of the phase detector once divided by 50, while at the other
port a signal emanating from a receive array element is applied. The received signal
at 1050 MHz is mixed with the transmit signal from the VCO and the sum output at
1990 MHz is then divided by 9950. The voltage at the output of the detector at B,
essentially at DC, is proportional to the detected phase offset. It is added to an
externally applied DC offset voltage to provide a pre-setting voltage for the voltage
controlled oscillator (VCO) at F1 = 0.94 GHz. The ‘error’ voltage from the phase
detector provides continual adjustment to the VCO in a direction which ensures that
the 200 kHz signal entering the phase detector exhibits zero phase offset. Part of the
VCO output passes through a buffer amplifier to the transmit radiator, while part
enters a following mixer. As indicated above, there it is mixed with the received
signal, and the resultant sum signal is directed through a bandpass filter back to the
phase locked loop (PLL). Note that the difference signal from the mixer provides
(see [12]) a constant phase IF signal. These signals from all other receive elements
of the array antenna are then added before demodulation. This obviates the need for
a second mixer and local oscillator as in the Chernoff circuit.

Fig. 11.3 Phase conjugation using PLL chip


290 11 Retro-directive Compact Array Antennas

11.4 Phase Conjugation Using IQ Processing

The phase conjugation scheme represented by Fig. 11.3 assumes that a double array
exists with the receive array separate from the transmit array. This configuration, as
already observed, could obviously be a disadvantage in many situations, as could
the employment of differing transmit and receive frequencies. Furthermore at
microwave frequencies the very high sum frequency fed back to the PLL requires a
costly BPF and a high degree of division. An ‘in-phase/quadrature phase’
(IQ) modification has been proposed [15, 16] to circumvent these difficulties and is
shown diagrammatically in Fig. 11.4. The IQ modulator on the right of the circuit is
pivotal to the approach. The circuit as a whole (one to each array element) con-
jugates the signal (xRF) received at the left port, from the appropriate array element,
before dispatching it at the right (xout) to the transmit radiating element. As will be
demonstrated below, the transmit and receive elements need not be separate radi-
ators, as Fig. 11.4 seems to suggest. The IQ modulator, in analogue form, primarily
comprises a pair of up-converting mixers, a 90° phase shift and a summing
junction.
Phase conjugation is procured as follows. A received signal (xRF) at frequency
2400.5 MHz (say), and phase delay /, is mixed with a 2400 MHz signal from the
frequency controlled local oscillator (VCO1) to form a 0.5 MHz IF signal at A, i.e.

V ¼ VA cosðxRF t  xLO1 t þ /Þ ð11:6Þ

Fig. 11.4 Phase conjugation using I and Q processing


11.4 Phase Conjugation Using IQ Processing 291

where VA is an arbitrary complex magnitude. At A this signal is split equally into


two separate paths. In one of these paths a −90° phase shift is applied before the
signals enter the IQ modulator. There they are mixed with a second local oscillator
(xLO2) signal at 2410 MHz say so that at point B we have:

VB ¼ V1 cosðxRF t  xLO1 t þ xLO2 t þ /  90 Þ


ð7Þ
þ V1 cosðxRF t þ xLO1 t þ xLO2 t  / þ 90 Þ

and at C:

VC ¼ V1 cosðxRF t  xLO1 t þ xLO2 t þ / þ 90 Þ


ð8Þ
þ V1 cosðxRF t þ xLO1 t þ xLO2 t  / þ 90 Þ

When summed, the resultant signal at xout has the form:

Vout ¼ V2 cosðxLO1 t þ xLO2 t  xRF t  / þ 90 Þ ð11:9Þ

which is phase conjugated relative to the input as required. The output frequency is
equal to the input frequency if xLO1 þ xLO2 ¼ 2xRF .

11.5 Retro-directive Array Implementation

It is clear from the literature that the performance of retro-directive array antennas is
measurably susceptible to direct leakage between the transmission and receiver
sections of the system, particularly if transmit and receive operations are performed
at the same frequency. The main source of this leakage is mutual coupling between
closely spaced array elements, and to avoid this, the transmit and receive functions
are often totally separated, by adopting two distinct arrays [6, 12, 16]. Leakage
between the transmission and reception circuits both reduces the sensitivity of the
system and diminishes tracking accuracy.
It is not difficult to appreciate that doubling up on array hardware to counteract
leakage seems like technological ‘over kill’. It certainly adds significantly to build
costs. The usual way around this is to employ a duplexer behind each array element
which allows the transmit and receive channels to use the same antenna element.
This is suggested in Fig. 11.1 where the duplexer takes the form of a circulator.
Unfortunately, as indicated previously, duplexers, which display high enough
isolation between transmit and receive ports to permit reliable and accurate
retro-directive behaviour in an array, tend to be very expensive.
Given advances in planar microwave circuit fabrication techniques, a potentially
much more cost effective solution is to employ array radiator elements which have
dual-mode capability [17]. An example is depicted in Fig. 11.5, where the
retrodirective array is formed from square patches capable of receiving/transmitting
292 11 Retro-directive Compact Array Antennas

vertically or horizontally polarised electromagnetic waves. For this particular


dual-mode antenna array the transmit circuit port in each phase conjugating circuit
(PCC) is linked to the vertical polarisation connector at the patch, while the receive
port is linked to the horizontal polarisation connector. Isolation between the vertical
polarisation and horizontal polarisation ports is required to be of the order of 30 dB
[16, 17]. Radiators which can potentially provide high isolation between orthogonal
modes of operation are discussed below.
For some communications applications, for example between a ground station,
or a moving vehicle on the surface of the earth, and a potentially gyrating satellite, it
is almost inevitable that severe signal fading would occur if this dual-mode linearly
polarised technique were adopted in a retrodirective tracking array. Generally the
answer to the problem of sustaining a channel of communication between plat-
forms, which could be rotating and moving, involves the introduction of circularly
polarised transmission and reception channels. However, circularly polarised patch
antennas, which are also dual mode, are less well established than linear alterna-
tives, as the patent situation [18–21] and the literature [22–30] attests. Thus, the
adoption of retrodirective technology, if it is to be commercially viable, calls for the
use of dual-polarised array elements to eliminate the need for separate transmit and
receive arrays while maintaining signal fidelity in a mobile communications system.
Unfortunately, in microstrip technology, this in turn can lead to rather complex feed
arrangements, particularly when the radiators are incorporated into arrays of four or
more elements, and particularly when beam-switching is involved [3, 4]. As has
been noted earlier, it is clear from the literature that circularly polarised slot or patch
antennas, which are also dual-mode, are less well established than their linear
alternatives. Nevertheless, microstrip transmission line technology is the preferred
platform for mobile communications. This limits possible antenna array options to
three distinct alternatives, as is outlined below.

Fig. 11.5 Retro-directive


transmit/receive array with
orthogonally polarised patch
radiators
11.6 Cross-slot Fed Microstrip Patch Antenna 293

11.6 Cross-slot Fed Microstrip Patch Antenna

Arguably, the most commonly employed mechanism for securing a compact


dual-polarised array element based on microstrip technology is the square con-
ducting patch excited parasitically from a circularly symmetric cross-slot [25, 28].
Edimo et al. describe arguably the simplest arrangement, which is illustrated
schematically in Fig. 11.6. The non-resonant X-slot is located centrally below the
square patch radiator which means, referencing Fig. 1.4, that the orthogonally
directed slots of the X-slot separately excite the degenerate orthogonal TE101 modes
of the patch, thus securing dual-polarised radiation. Each of the two microstrip feed
lines, in principle, excites only the slot element whose major dimension is trans-
verse to it. However, the geometry enforces the feed lines to be on opposite sides of
the slot plane and thus occupy separate substrates, which adds undesirable com-
plexity in fabrication terms.
The patch excitation arrangement depicted in Fig. 11.6 displays a degree of
imbalance between the patch and the feed lines—the fields of one line are shielded
from the patch, while for the other they are not. This leads to issues of differential
gain between the desired polarisations, cross-polarisation differences and poor axial
ratio outcomes. Nevertheless the simple arrangement depicted in Fig. 11.6 has been
demonstrated [25] to operate over an *5% bandwidth, for a 5 GHz design
example. The cross-coupling between the orthogonal ports is modest, and predicted
to be of the order of −20 dB.
The troublesome cross-coupling generated by the slot excitation scheme
depicted in Fig. 11.6 can be significantly reduced [28] by simply incorporating
microstrip T-junctions into each feed line (see Fig. 11.7) so that it is no longer in
close proximity to the orthogonal uncoupled slot. The arrangement introduces a
modicum of additional complexity to the layout and to the fabrication of the
antenna, but not without considerable performance gains. For an antenna designed
to operate at 1.5 GHz, Sanford and Tengs report a 20% bandwidth capability
(defined by S11 < −10 dB) with port isolation (S21) approaching −35 dB over the
band.

11.7 Dual-Polarised Radiation from Apertures

While retro-directive array antennas for mobile applications are, on the basis of
current technology, likely to be optimally realised using patch radiating elements,
needless to say, non-patch alternatives, providing circularly polarised and/or
dual-mode capability, are also scattered through the literature [29]. Generally, these
competing possibilities comprise either microstrip fed cross-slot radiators, or
microstrip fed circular apertures. An example which gives a flavour of non-patch
possibilities is depicted in Fig. 11.8. The four holes etched into the metallic ground
plane of the middle substrate (b) are sandwiched between substrates (a) and
294 11 Retro-directive Compact Array Antennas

Fig. 11.6 Schematic of dual-mode dual-polarised patch antenna parasitically fed from a
non-resonant X-slot by means of microstrip lines above and below the slot
11.7 Dual-Polarised Radiation from Apertures 295

Fig. 11.7 Schematic of a dual-mode dual-polarised patch antenna parasitically fed from a
non-resonant X-slot by means of balanced microstrip lines above and below the slot to improve
bandwidth and axial ratio
296 11 Retro-directive Compact Array Antennas

Fig. 11.8 Schematic of dual-mode dual-polarised circular aperture array antenna parasitically fed
from balanced microstrip lines above and below the four elements

(c) which carry branching microstrip lines. The elliptic shapes, incorporated into the
branch-lines, should register with the circular apertures, and are designed to
improve the matching process. The upper substrate er1 excites vertically polarised
radiations from the four apertures while the lower substrate er3 generates
11.7 Dual-Polarised Radiation from Apertures 297

transversely polarised waves. If the combination is fed from a quadrature hybrid


then this four hole antenna element provides circular polarised radiation. The
metallic layer on the back surface of substrate er1 ensures one-directional radiation
—to the right on the end view in Fig. 11.8.
While this concept of circular apertures, fed from back-to-back microstrip feed
structures, offers some advantages in fabrication terms, cited performance figures
tend to be modest. The −10 dB bandwidth is of the order of 5% while
cross-polarisation levels tend to be high at approximately −20 dB. In addition, it
should be noted that, like slot radiators, holes are prone to high mutual coupling,
which is best avoided in retro-directive array applications.

11.8 Parasitically Fed Microstrip Patch Antenna

In order to eliminate the disadvantage of requiring two separate feed substrates to


procure dual polarisation in a single parasitically coupled radiating element, it is
necessary to feed the patch through two orthogonally directed non-contacting
apertures or slots. The general arrangement takes the basic form shown in Fig. 11.9,
where the square patch (or patches) is excited by a pair of orthogonally directed
slots which in turn set up spatially orthogonal TE101 modes under the patch. This
geometry permits slot excitation by a single microstrip substrate with feed lines on
one face. Such dual-mode aperture coupled patch arrangements are described in
papers by Adrian and Schaubert [22], Gosalia and Lazz [23] and Gao et al. [26]. In
[22], the antenna is designed to operate at *2.215 GHz, and is fabricated on a
substrate with er = 2.55. It displays a modest 3 dB gain bandwidth of 3.5% with an
axial ratio of 2 dB across this band. This equates to a bandwidth of approximately
8 MHz. However this can be improved by introducing a second patch as described
by in reference [26]. Designed for 2.6 GHz their dual patch design with dumbbell
coupling slots, as shown in Fig. 11.9, is reported to provide an S11 = −10 dB over a
bandwidth of *21% with cross-polarisation levels of better than minus 22 dB. To
secure dual polarisation and circular polarisation simultaneously with the dual mode
patch antennas described above, it is necessary to feed the antenna through a 3 dB
hybrid coupler [30] (see Fig. 11.10). This scheme generates the required quadrature
phase relationship at the antenna ports. However, the combination can be space
demanding, which can be an impediment to the formation of a planar array with
grating-lobe free element spacing ði:e:  k0 =2Þ. The problem is partially solved in a
paper by Sangford and Tengs [28] where the hybrid is merged, with the slot feed
lines and the coupling slots, below the patch. However, the outcome is very
complex in fabrication terms, making it potentially unattractive in any cost con-
scious development.
While good impedance bandwidth of the order of *10% is most often quoted in
assessing antenna suitability to communication systems, some applications also
demand comparable circular polarisation ellipticity bandwidths [31–34]. Typically
ellipticity bandwidths can be 30% less than the impedance bandwidth of dual
298 11 Retro-directive Compact Array Antennas

Fig. 11.9 Schematic of dual-mode dual-polarised patch antenna. a Substrate 1—copper patch
L1  L1, thickness t1, relative permittivity er1, b substrate 2—copper patch L2  L2, thickness t2,
relative permittivity er2, c two dumbbell coupling slots in top face, 50X feed lines in lower face
(see Gao [26] for detail), substrate thickness t3, relative permittivity er3, d edge view showing air/
polystyrene layer between substrate 1 and substrate 2, thickness t4, relative permittivity er4

polarised patch antennas. Dual-mode dual polarised microstrip antennas have


generally employed square patches, but a desire for wider ellipticity bandwidth has
generated interest in circular patches, as is shown by Karmakar and Bialkowski
11.8 Parasitically Fed Microstrip Patch Antenna 299

Fig. 11.10 Circularly polarised dual-mode square patch antenna with slot coupling and microstrip
hybrid feed arrangement

[31]. A typical manifestation of this approach is depicted in Fig. 11.11. The circular
symmetry of the round patch confers sufficient marginal improvement in polari-
sation circularity, at wide angles in the radiation pattern, to acquire enhanced
ellipticity bandwidth for the circular radiator, when compared with square patch
alternatives.

11.9 Phase Conjugated Active Antenna Element

Currently, the most attractive method of securing phase conjugation in a


retro-directive array antenna is furnished by heterodyne mixing. However, as we
know from the literature this essentially requires that the RF (receive) and the IF
(transmit) frequency be the same, or very nearly the same. This means that RF
leakage reflected or radiated back toward the source can significantly interfere with
the retro-directive signal. Getting rid of the unwanted RF leakage is not possible
using conventional filtering techniques because the IF is at the same frequency. The
active antenna element represents a possible solution to this problem [34, 35].
The signal (RF at 6 GHz) received by the active phase conjugating antenna (the
red patch—Fig. 11.12), is split into two components, to the right of the figure.
These are then forced into an antiphase relationship by using line length. The local
300 11 Retro-directive Compact Array Antennas

Fig. 11.11 Dual-mode


dual-polarised microstrip
patch antenna implemented
using circular patches for
enhanced axial ratio
bandwidth

oscillator signal at twice the RF frequency is introduced at the left of the circuit. It is
also split into two equal components with a differential 180° phase shift added by
line length. The advantage of the RF frequency signals on the microstrip lines, in
the upper and lower portions of the circuit, being in antiphase is that leakage or
cross-coupling across the channels, which can degrade the retro-directive operation,
is minimised. The FET gate mixers provide both phase conjugation and gain, with
11.9 Phase Conjugated Active Antenna Element 301

Fig. 11.12 Phase conjugating active antenna element incorporating a balanced quasi-optical
mixer

the difference frequency IF signals adding at the T-junction before passing to the
antenna for transmission. Note that at the patch the transmit and receive signals are
isolated from each other by employing the orthogonal linear polarisations of the
square patch. For retro-directive communication between a mobile platform and a
satellite, which depends on circularly polarised radiations, the Fig. 11.12 circuit is
inappropriate. While the concept is promising, clearly further research is needed to
achieve both a dual-polarised and a circularly polarised active antenna element for a
mobile communications system with fully operational retro-directive capability.

11.10 Chapter Summary

A review is furnished here of state-of-the-art technology (relevant up to 2017),


which aims to provide compact antenna designs for wireless applications where
relative motion between the receive antenna and the source of a desired commu-
nications signal exists. The development involves the introduction of retrodirective
capability into compact transmit/receive arrays on, for example, mobile platforms
and similar applications. It is shown that the technology follows three distinct paths
which are presented in detail.
Several antenna formats, exhibiting potential compatibility with the retrodirec-
tive mode of operation, are studied, and it is concluded that compactness and low
cost are best served with antennas fabricated in microstrip technology. Furthermore,
current developments indicate that the antenna forms which are generally preferred
are those using one or more square patches excited through separated parasitic
coupling slots, in order to secure highly desirable circularly polarised radiation.
302 11 Retro-directive Compact Array Antennas

It is tentatively suggested that the future may reside with active antenna ele-
ments, but this remains some distance away, until dual-polarised, circularly
polarised capability can be confidently demonstrated in commercial products.

References

1. R.C. Johnson, Antenna Engineering Handbook, 3rd edn. (McGraw-Hill Inc., New York,
1993)
2. A.W. Rudge, K. Milne, A.D. Olver, P. Knight (eds.), The Handbook of Antenna Design:
volumes I and II (Peter Peregrinus Ltd., London, 1986)
3. C.Y. Pon, Retrodirective array using heterodyne technique. IEEE Trans. Antennas Propag.
AP-12, 176–180 (1966)
4. R.N. Ghose, Electronically adaptive antenna systems. IEEE Trans. Antennas Propag. AP-12
(2), 161–169 (1964)
5. L.G. Van Atta, Electromagnetic Reflector. US Patent 2,908,002, 6 Oct 1959
6. R.C. Chernoff, Large active retrodirective arrays for space applications. IEEE Trans.
Antennas Propag. AP-27(4), 489–496 (1979)
7. K. Leong, Y. Wang, T. Itoh, A full duplex capable retrodirective array system for high speed
beam tracking and pointing applications. IEEE Trans. Microwave Theory Tech. MTT-52(5),
1479–1489 (2004)
8. G.S. Shiroma, R.Y. Miyamoto, W.A. Shiroma, A full-duplex dual-frequency self-steering
array using phase detection and phase shifting. IEEE Trans. Microwave Theory Tech.
MTT-54(1), 128–134 (2006)
9. R.Y. Myamoto, Y. Qian, T. Itoh, An active integrated retrodirective transponder for remote
information retrieval. IEEE MTT-S, Int. Microwave Symp. 3, 1431–1434 (2000, June)
10. R.Y. Myamoto et al., Digital wireless sensor server using an adaptive smart-antenna/
retrodirective array. IEEE Trans. Vehicular Technol. 52(5), 1181–1189 (2003)
11. S.L. Karode, V.F. Fusco, Frequency offset retrodirective antenna array. Electron. Lett. 33(16),
1350–1351 (1997)
12. S.L. Karode, V.F. Fusco, Multiple target tracking using retrodirective antenna arrays. Nat.
Conf. Antennas Propag. Publication No. 461, 178–181 (1999)
13. V.F. Fusco, C.B. Soo, N. Buchanan, Constant phase IF self-tracking receiver analysis and
characterisation. IEEE Proc. (MAP) 151(6), 507–513 (2004)
14. N.B. Buchanan, V.F. Fusco, Triple mode PLL antenna array. IEEE MTT-S Int. Microwave
Symp. Digest, 1691–1694 (2004)
15. V.F. Fusco, C.B. Soo, N. Buchanan, Analysis and characterisation of PLL-based retrodirec-
tive array. IEEE Trans. Microwave Theory Tech. MTT-53(2), 730–738 (2005)
16. V.F. Fusco, N. Buchanan, High performance IQ modulator based phase conjugator for
modular retrodirective antenna array implementation. IEEE Trans. Microwave Theory Tech.
MTT-57(10), 2301–2306 (2009)
17. V.F. Fusco, N. Buchanan, Dual mode retrodirective phased array. Electron. Lett. 45(3), 139–
141 (2009)
18. T.D. Huynh, Dual polarised aperture coupled microstrip patch antenna system. US Patent
No. 5,896,107, 1999
19. L. Desclos et al., Single and dual mode patch antennas. US Patent No. 6181281, 30 Jan 2001
20. N. McDonald, Dual polarised patch antenna. European Patent No. EP 0901185, 14 Nov 2001
References 303

21. S.M. Sanzgiri et al., Dual polarised microstrip patch antenna array for PCS base stations. US
Patent No. 5,923,296, 13 July 1999
22. A. Adrian, D.H. Schaubert, Dual aperture-coupled microstrip antenna for dual or circular
polarisation. Electron. Lett. 23(23), 1226–1228 (1987)
23. K. Gosalia, G. Lazz, Reduced size dual-polarised microstrip patch antenna for wireless
communications. IEEE Trans. AP-51(9), 2182–2186 (2003)
24. S.L. Karode, V.F. Fusco, Dual-polarisation microstrip patch antenna using feedforward
isolation enhancement for simultaneous transmit/receive applications. Nat. Conf. Antennas
Propag. Conference Publication No. 461, 49–50 (1999)
25. M. Edimo, A. Sharaiha, C. Terret, Optimised feeding of a dual-polarised broadband
aperture-coupled printed antenna. Electron. Lett. 28(19), 1785–1787 (1992)
26. S.C. Gao, L.W. Li, P. Gardner, P.S. Hall, Wideband dual-polarised microstrip patch antenna.
Electron. Lett. 37(20), 1213–1214 (2001)
27. A. Uz Zaman, L. Manholm, D. Derneryd, Dual-polarised microstrip patch antenna with high
port isolation. Electron. Lett. 43(10), 551–552 (2007)
28. J.R. Sandford, A. Tengs, A two substrate dual-polarised aperture-coupled patch. IEEE
Antennas Propag. Soc. Int. Symp. AP-S Digest 3, 1544–1547 (1996)
29. E. Rammos, A new wideband high-gain stripline planar array antenna for 12 GHz TV.
Electron. Lett. 18(6), 252–253 (1982)
30. A.J. Sangster et al., A dual-polarised stripline fed slot antenna incorporating signal
cancellation. IEE Proc. (MAP) 148(6), 357–362 (2001)
31. N.C. Karmakar, M.E. Bialkowski, Circularly polarised aperture coupled circular microstrip
patch antennas for L-band applications. IEEE Trans. AP-47(5), 933–940 (1999)
32. N.C. Karmakar, M.E. Bialkowski, A beam forming network for a circular switched-beam
phased array antenna. IEEE Microwave Wireless Lett. 11(1), 1–9 (2001)
33. M.E. Bialkowski, N.C. Karmakar, A two ring circular phased array antenna for satellite
communications. IEEE Antennas Propag. Mag. 41(3), 14–23 (1999)
34. K. Chang, R.A. York, P.S. Hall, T. Itoh, Active integrated antennas. IEEE Trans. MTT-50(3),
937–944 (2002)
35. R.Y. Miyamoto, Y. Qian, T. Itoh, Active retrodirective array for remote tagging and wireless
sensor applications. IEEE MTT-S Int. Microwave Symp. Dig., Boston, MA, 1431–1434
(2000, June)
Appendix A
Free-Space Electromagnetic Waves

A.1 Plane Wave Solution to the Maxwell Equations

All materials contain electric charges bound loosely or otherwise within atoms and
molecules. If these materials exist in an environment which naturally or artificially
causes agitation of the charge, and hence changes in the associated electric and
magnetic fields, then electromagnetic waves are unavoidable. These waves can
appear in quite complex forms such as trapped wave, fast waves, slow waves,
surface waves, leaky waves, evanescent fields and radiating waves. To describe
these various and quite diverse formats a set of equations is required which rep-
resent the field components of the wave at a point in space. The point forms
(differential forms) of Maxwell’s equations are developed in a host of text books on
the topic (see Bibliography) and essentially entails the recruitment of well known
vector-differential theorems such as the divergence theorem and Stokes’ law to
accomplish the transitions.
Many electromagnetic wave guiding and manipulating problems are of the
source-free variety, which implies that the source is so far distant that the waves of
interest are freely propagating and in an unbounded region (free-space) are plane
waves. These waves, also termed transverse electromagnetic (TEM) waves, are
described as ‘plane’ because the radius of curvature of the wave-front is very large,
and thus the natural rate of curvature of the front, can be deemed mathematically
insignificant allowing it to be fully described by means of Cartesian coordinates. In
this scenario, the EM problem reduces to a boundary value problem, for which
Maxwell’s equations, in differential form, become:

rD¼0 ðA:1Þ

rB¼0 ðA:2Þ

© Springer Nature Switzerland AG 2019 305


A. J. Sangster, Compact Slot Array Antennas for Wireless Communications, Signals
and Communication Technology, https://doi.org/10.1007/978-3-030-01753-8
306 Appendix A: Free-Space Electromagnetic Waves

@B
rE¼ ðA:3Þ
@t
@D
rH¼ ðA:4Þ
@t

where E and H represent the electric and magnetic field intensities in the region
of interest. As always, D = eE is the electric flux density while B = lH is the
magnetic flux density. The ‘del’ operator ðrÞ expresses directional derivatives in
the three space directions. It is a vector, which in the Cartesian system (for example)
has the form:

@ @ @
r ¼ ^ax þ ^ay þ ^az ðA:5Þ
@x @y @z

where ^ax ; ^ay and ^az are unit vectors directed along x, y, and z respectively.
When the del operator is multiplied to a scalar (/(x, y, z) say) the result is a vector
which expresses the gradient or slope of the function / in all three space directions.
i.e.:

@/ @/ @/
r/ ¼ ^ax þ ^ay þ^
az ðA:6Þ
@x @y @z

Cross multiplication of del with a vector produces the operation of ‘curl’, while
dot multiplication produces the operation of ‘divergence’ (‘div’). Crudely, curl is
circulation at a point, while divergence is flux at a point.

A.2 Second Order Differential Equation

To solve the Maxwell equations for E-field or H-field behaviour in a bounded


region it is first necessary to form an equation either E or H alone. The standard
procedure for achieving this conversion is to perform a curl operation on either the
curl equation for E or the corresponding equation for H. This gives, for example,
using Eq. (A.3)

@
rrE¼ lr  H
@t
@ @
¼  ½ le E ðA:7Þ
@t @t
@2E
¼ le 2
@t
Appendix A: Free-Space Electromagnetic Waves 307

Hence, on using a convenient vector identity, which states that for any vector F

r  r  F ¼ rr  F  r2 F ðA:8Þ

Equation (2.39) can be re-expressed as follows:

@2E
rr  E  r2 E ¼ le ðA:9Þ
@t2

But, from Eq. (A.1), r  E ¼ 0, for a linear, homogeneous medium for which
permeability l and permittivity are constants. Therefore

@2E
r2 E ¼ le ðA:10Þ
@t2

and by analogy:

@2H
r2 H ¼ le ðA:11Þ
@t2

Equations (A.10) and (A.11) are wave equations. Equations of this nature, with
appropriate variables, appear in most branches of science and engineering and their
solutions have been studied widely. Solutions depend very much on the boundary
conditions—namely the conditions imposed on the variables at the periphery or
containing surface of the solution region. They can fix the magnitude of the variable
(Dirichlet condition) or the rate of change of the variable (Newman condition) or a
mixture of both. A unique solution depends on the conditions being neither under
specified or over specified.
For example, let us consider formulating a solution to Eq. (A.10), and inevitably
Eq. (A.11) because of the Maxwell linkages, for a region of free space (l = l0:
e = e0) which is large enough to presume that all boundaries are effectively at
infinity. In this case we can choose to represent the region mathematically using
Cartesian coordinates, and furthermore since we anticipate that the solution is a
waveform we can arbitrarily determine that the waves travel in the z-direction. This
implies that the rates of change of the E-field in x and y are zero, and using (A.1) it
follows that Ez = 0. The equation to be solved, therefore, is:

@2E @2E
¼ l0 e0 2 ðA:12Þ
@z 2 @t

where, in general, E ¼ a^x Ex þ a^y Ey . However, if we choose to align the coor-


dinate system so that E lies along the x-axis (x polarised solution), then Ey = 0 and
the wave equation reduces to the scalar form:
308 Appendix A: Free-Space Electromagnetic Waves

@ 2 Ex 1 @ 2 Ex
¼ ðA:13Þ
@z2 c2 @t2

With ‘speed of light’ c ¼ pffiffiffiffiffiffi


1 ffi
l e0 .
0

A.3 General Solution

Equation (A.13) has a wave solution of the general form:

Ex ¼ Af ðz  ctÞ þ Bf ðz þ ctÞ ðA:14Þ

This is easily demonstrated by substitution back into the equation. The first term
represents a wave travelling in the +z direction while the second allows for a
reflected wave, if such exists. Given that velocity is the rate of change of z with
respect to time, it is evident that c represents velocity (actually phase velocity) of
the electromagnetic wave in ‘free space’. For vacuum it is equal to 3  108 m/sec.
The application of Maxwell’s equations also gives Hz = 0 and:

A B
Hy ¼ f ðz  ctÞ þ f ðz þ ctÞ ðA:15Þ
g g

Also
rffiffiffiffiffi
Ex l0
¼ ¼ g ðA:16Þ
Hy e0

η is termed the free space wave impedance which for air or vacuum has the value
120p X. The resultant solution is a plane electromagnetic wave, also termed a
transverse electromagnetic (TEM) wave, for which E and H are transverse to the
direction of propagation and orthogonal to each other. E and H are also in time
phase, as Eq. (A.16) attests (see Fig. A.1).
Electrical engineers are generally very familiar with the relationship between
power (P), voltage (V) and current (I) in the form:

1
P ¼ VI Watts ðA:17Þ
2

where V and I are defined in peak, rather than in the more common r.m.s.,
format. But, voltage is simply integrated electric field E (V/m), and from Ampere,
current is integrated magnetic field intensity H (A/m), so by analogy we can suggest
that for the plane wave:
Appendix A: Free-Space Electromagnetic Waves 309

Fig. A.1 TEM wave field and directional relationships

1 1
p ¼ EH ¼ ce0 E2 Watts=m2 ðA:18Þ
2 2

This means that p is the real power flow density in the TEM wave. In general,
complex power flow density in an electromagnetic wave is given by the Poynting
vector S, where:

1
S ¼ E  H Watts=m2 ðA:19Þ
2

In electrical engineering it is much more usual to examine wave solutions at a


single frequency (x rad/sec)—namely sinusoidal solutions. This actually incurs
little loss of generality, since any arbitrary time variation carried on a radio-wave
can be resolved into a spectrum of single frequency components. The adoption of a
single frequency, or a spectral frequency, in carrying through time varying com-
putations has the distinct advantage that the time variable can be omitted. The
calculations are then progressed in phasor notation. In trigometric form Eq. (A.14)
becomes:

Ex ¼ A exp jðxt  bzÞ þ B exp jðxt þ bzÞ ðA:20Þ

where A and B are complex constants. The phasor form is:

Ex ¼ j Aj expðjbz þ uÞ þ jBj expðjbz þ hÞ ðA:21Þ

with u and h representing the phases respectively of A and B.


310 Appendix A: Free-Space Electromagnetic Waves

A.4 Snell’s Laws

When a plane electromagnetic wave at the frequency of light, or in fact any radio
frequency, is incident upon a smooth interface (by ‘smooth’ it is meant that any
surface undulations or protuberances are in size very much less than the wavelength
of the impinging waves) between two extended propagating media, part of the wave
is reflected back into the incident medium while part is transmitted or refracted into
the second medium, usually with a change of direction.
Analytically, the relationships between the incident and reflected waves can be
developed by considering a plane electromagnetic wave, incident at a physically
real angle hi to the normal, at the interface between two semi-infinite regions of
space, as suggested in Fig. A.2. Each region is presumed to comprise linear
homogeneous medium with a different index of refraction (n). The index of
refraction is defined as:
c
n¼ ðA:22Þ
v

where c is the speed of light in vacuum, or free space, while v is its speed within
the specified medium. Also, with reference to Fig. A.2, the following definitions
apply:

Fig. A.2 Reflection and refraction at a dielectric interface—perpendicularly polarised case


Appendix A: Free-Space Electromagnetic Waves 311

1
c ¼ pffiffiffiffiffiffiffiffiffi ðA:23Þ
l0 e 0

and

1
v1 ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðA:24Þ
l0 e0 er1

1
v2 ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðA:25Þ
l0 e0 er2

Here, e0 and l0 are the free space permittivity and permeability respectively,
while er1 is the relative permittivity of medium 1 and er2 is the relative permittivity
of medium 2. Both media are assumed to be loss-less and non-magnetic in which
case l1 = l2 = l0. The indices of refraction for the two media then become:
pffiffiffiffiffiffi pffiffiffiffiffiffi
n1 ¼ er1 ; n2 ¼ er2 ðA:26Þ

Maxwell’s equations in the semi-infinite regions remote from the interface are,
as we have seen above, fully satisfied by TEM plane waves. It remains then to
satisfy the Maxwell boundary conditions at the interface. If this can be done the
resultant solutions represent complete EM solutions for the specified boundary
value problem. For an incident TEM electromagnetic wave, as depicted in Fig. A.2,
the E-field vector and the H-field vector must be mutually orthogonal to each other
^ directed in
and to the direction of propagation, usually defined by a unit vector k,
the direction of the relevant ray. In this case we can write:

1 ^
H ¼ ðk  EÞ ðA:27Þ
g

where η is the wave impedance for the medium containing the wave. Hence for
regions 1 and 2 respectively:
rffiffiffiffiffiffiffiffiffi
l0
g1 ¼ ðA:28Þ
er1 e0
rffiffiffiffiffiffiffiffiffi
l0
g2 ¼ ðA:29Þ
er2 e0

However, this condition does not fully establish the polarisation direction, which
must also be specified. There are two basic choices from which any other polari-
sation possibilities can be deduced. We can choose the E-field vector of the incident
wave to be either normal to the yz-plane, or parallel to it. The yz-plane in Fig. A.2 is
generally termed the plane of incidence for the incoming wave, being the plane that
312 Appendix A: Free-Space Electromagnetic Waves

contains both the direction vector k ^ and the unit normal to the interface ð^ nÞ. When
the electric field in the incident TEM wave is normal to the plane of incidence, the
wave is said to be perpendicularly polarised, and when it is parallel to this plane it is
described as parallel polarised. Note that in relation to the surface of the earth, while
parallel polarisation equates to horizontal polarisation, perpendicular polarisation
can be termed vertical polarisation only if hi approaches 90°. Perpendicular polar-
isation is often termed transverse electric (TE) propagation while parallel polarised
waves get the complementary description of transverse magnetic (TM) waves.
Now that we know the electromagnetic field forms (TEM waves) remote from
the interface between regions 1 and 2 in Fig. A.2, we can examine the field con-
ditions (boundary conditions) precisely at the interface. For the diffraction set up
depicted in Fig. A.2 with a perpendicularly polarised TEM wave incident at hi the
field directions at a given instant in time can be represented vectorially as shown.
Just at the interface a typical ray of the incident TEM wave is both reflected off the
surface and transmitted through it. Also, for a ‘smooth’ surface ‘common sense’
suggests that it is safe to presume that the reflected and transmitted waves retain the
polarisation of the incident wave. Furthermore, there will be a single reflected ray
and a single transmitted ray. Actually, this latter assumption is not strictly necessary
as we will show presently.
When the TEM wave direction (or ray) lies in paths other than along the
coordinate axes, it is usual to define the ray direction by the vector k which is
chosen to be equal in magnitude to the wave coefficient k. That is k ¼ kk. ^ Hence
we can express mathematically the wave component in any other direction (r say).
For the case shown in Fig. A.3, where the electric field is x-directed, the expression
has the form:

Fig. A.3 Representation of TEM wave with E, H, and k in mutually orthogonal directions
Appendix A: Free-Space Electromagnetic Waves 313

Ei ¼ ^ax Ei expðjk  rÞ ðA:30Þ

Consequently, if r and k lie in the yz-plane as suggested in Fig. A.3, then


clearly:

k ¼ ^ay ky þ ^az kz ðA:31Þ

r ¼ ^ay y þ ^az z ðA:32Þ

Also

k2 ¼ ky2 þ kz2 ðA:33Þ

so we can conveniently write:

ky ¼ k sin h ðA:34Þ

and

kz ¼ k cos h ðA:35Þ

Hence, employing these relationships, Eq. (A.30) can be expanded into the
non-vectorial form:

Exi ¼ Ei expðjðxt  k1 z cos hi  k1 y sin hi Þ ðA:36Þ

where
x x
k1 ¼ ¼ n1 ðA:37Þ
v1 c

For a transverse electromagnetic wave the electric and magnetic fields are related
through Eq. (A.27). Hence on combining Eqs. (A.36) and (A.27), and observing the
field directions in Fig. A.2 we obtain for magnetic fields:

Hyi ¼ Hi cos hi exp½jðxt  k1 z cos hi  k1 y sin hi Þ ðA:38Þ

Hzi ¼ Hi sin hi exp½jðxt  k1 z cos hi  k1 y sin hi Þ ðA:39Þ

Also, we note that if these field components represent a TEM wave, then we
must have:

Ei g
¼ g1 ¼ 0 ðA:40Þ
Hi n1
314 Appendix A: Free-Space Electromagnetic Waves

Similar constructions lead to the following equations for the reflected and
transmitted field components:

Exr ¼ Er expð jðxt þ k1 z cos hi  k1 y sin hi Þ ðA:41Þ

Hyr ¼ Hr cos hi exp½ jðxt þ k1 z cos hi  k1 y sin hi Þ ðA:42Þ

Hzr ¼ Hr sin hi exp½ jðxt þ k1 z cos hi  k1 y sin hi Þ ðA:43Þ

with

Er
¼ g1 ðA:44Þ
Hr

and

Ext ¼ Et expð jðxt  k2 z cos ht  k2 y sin ht Þ ðA:45Þ

Hyt ¼ Ht cos ht exp½ jðxt  k2 z cos ht  k2 y sin ht Þ ðA:46Þ

Hzt ¼ Ht sin ht exp½ jðxt  k2 z cos ht  k2 y sin ht Þ ðA:47Þ

where

Et
¼ g2 ðA:48Þ
Ht

and
x
k2 ¼ n2 ðA:49Þ
c

The above field expressions for the incident and reflected waves in region 1 and
the transmitted waves in region 2 each separately satisfy Maxwell’s equations in
these regions. A solution that satisfies Maxwell’s equations for the entire volume
including the interface is achieved by enforcing the electromagnetic field boundary
conditions, given in equations at the interface. That is, at z = 0, we require that
across the divide between regions 1 and 2:

Ex is continuous ðA:50Þ

Hy is continuous ðA:51Þ

Bz is continuous ðA:52Þ
Appendix A: Free-Space Electromagnetic Waves 315

On combining Eq. (A.50) with the field expressions (A.36), (A.41) and (A.45)
we obtain with little difficulty:

Exi þ Exr ¼ Ext ðA:53Þ

on the z = 0 plane. The implication is that:

Ei expðjk1 y sin hi Þ þ Er expðjk1 y sin hr Þ ¼ Et expðjk2 y sin ht Þ ðA:54Þ

This equation must remain true over the entire z = 0 boundary, from
1  y  þ 1. This is only possible if:

k1 sin hi ¼ k1 sin hr ¼ k2 sin ht ðA:55Þ

It is pertinent to note here that if at the commencement of this derivation we had,


without pre-knowledge of refraction rules, chosen to presume that several reflected
waves at angles hr1, hr2, hr3 …, and several transmitted waves at angles ht1, ht2, ht3 …,
were possible, then the equivalent form of Eq. (A.54) would lead to:

k1 sin hr1 ¼ k1 sin hr2 ¼ k1 sin hr3 ¼ . . .. . . ðA:56Þ

and

k2 sin ht1 ¼ k2 sin ht2 ¼ k2 sin ht3 ¼ . . .. . . ðA:57Þ

These equations clearly dictate that hr1 = hr2 = hr3 = …, and ht1 = ht2 = ht3 =
… In other words, an ‘optically smooth’ surface produces only one reflected wave
and one transmitted wave.
Equation (A.55) is the source of Snell’s Laws which state that at an optically
smooth interface between two lossless media:

hr ¼ h i ðA:58Þ

sin ht k1 n1
¼ ¼ ðA:59Þ
sin hi k2 n2

However, these laws govern only the reflection and refraction angles. We also
need to have knowledge of the relative magnitudes of the reflected and transmitted
wave, and how these are influenced by material properties.
When Eqs. (A.50), (A.51) and (A.52) are applied to the TEM field components
at the boundary, while also applying Snell’s laws, the following relations are
generated:
316 Appendix A: Free-Space Electromagnetic Waves

Ei þ Er ¼ Et ðA:60Þ

ðHi  Hr Þ cos hi ¼ Ht cos ht ðA:61Þ

ðBi  Br Þ sin hi ¼ Bt sin ht ðA:62Þ

Equation (A.60) can be converted to magnetic field form by employing Eqs.


(A.44) and (A.48) leading to:

g1 ðHi þ Hr Þ ¼ g2 Ht ðA:63Þ

Consequently, if we choose to define reflection coefficient for this perpendicu-


larly polarised example (TE case) as:

Hr
qTE ¼ ðA:64Þ
Hi

then making use of Eqs. (A.61) and (A.63) the following useful relationship is
deduced:

g2 cos hi  g1 cos ht
qTE ¼ ðA:65Þ
g2 cos hi þ g1 cos ht

This can also be expressed in a slightly more familiar form, which explicitly
incorporates the indices of refraction, namely:

n1 cos hi  n2 cos ht
qTE ¼ ðA:66Þ
n1 cos hi þ n2 cos ht

Similarly, if we choose to define the transmission coefficient as:

Ht
sTE ¼ ðA:67Þ
Hi

then:

2n2 cos hi
sTE ¼ ðA:68Þ
n1 cos hi þ n2 cos ht

It is not difficult to demonstrate that:

Hr Er
qTE ¼ ¼ ðA:69Þ
Hi Ei
Appendix A: Free-Space Electromagnetic Waves 317

and

Et n1
¼ sTE ðA:70Þ
E i n2

An analogous derivation can also be followed through for the parallel polari-
sation case (TM case). If this is done we obtain:
 
Er Hr n2 cos hi  n1 cos ht 
jqTM j ¼ ¼ ¼ ðA:71Þ
Ei Hi  n2 cos hi þ n1 cos ht 

and

2n1 cos hi
sTM ¼ ðA:72Þ
n2 cos hi þ n1 cos ht

The reflection coefficient, as a function of incident angle for both the TE and TM
cases, is plotted in Fig. A.4. Clearly, for an interface between lossless dielectrics of
differing refractive indices the reflection behaviours are distinct. While for the TE
case it increases monotonically from a magnitude of 0.33 (n1 = 1, n2 = 2) at hi = 0,
to unity at hi = 90°, it drops to zero close to 60° in the TM case. At the zero

Fig. A.4 Reflection coefficient at an air/dielectric interface for TE and TM incidence


318 Appendix A: Free-Space Electromagnetic Waves

reflection angle the two surfaces are said to be ‘matched’ for surface normal wave
components. It is termed the Brewster angle, a physical property which underpins
the design of light polarizers.

Bibliography

1. R. Ferrari, An Introduction to Electromagnetic Fields (Van Nostrand Reinhold Co. Ltd.,


New York, 1975)
2. C.T.A. Johnk, Engineering Electromagnetic Fields and Waves (Wiley, New York, 1988)
3. R.P. Feynman, R.B. Leighton, M. Sands, Lectures in Physics, Vol. II (Addison-Wesley
Publishing Co., London, 1972)
4. P.M. Morse, H. Feshbach, Methods of Theoretical Physics, (McGraw-Hill Book Co.,
New York, 1953)
5. J.D. Kraus, Electromagnetics (McGraw-Hill Book Co., London, 1984)
6. P. Lorrain, D. Corson, Electromagnetic Field and Waves (W.H. Freeman & Co., San
Francisco, 1962)
7. A.J. Baden-Fuller, Engineering Electromagnetism, (Wiley, New York, 1993)
8. P. Hammond, Applied Electromagnetism (Pergamon Press Ltd., Oxford, 1971)
9. R.M. Bevensee, Slow-wave Structures (Wiley, New York, 1964)
10. J.A. Stratton, Electromagnetic Theory (Wiley, New Jersey, 2007)
Appendix B
Basic Waveguide Theory

B.1 Introduction

In Chap. 1 it is demonstrated, in an empirical manner, that the hollow rectangular


waveguide provides an effective transmission medium for electromagnetic waves
by trapping a free-space TEM wave within its confines. Actually it is not only
hollow pipes of rectangular cross-section which do this. Circular and elliptical
cross-sections are also possible. Given that the hollow pipe exhibits an unbroken
metallic structure to a source of electromagnetic waves, it is clear that at DC, and at
low enough frequencies, the waveguide will present a short circuit to the source. At
what frequency does the waveguide change from a non-propagating to a propa-
gating structure? This question, together with the multi-mode nature of waveguide,
will be addressed in this Appendix.
It is pertinent to note, at the outset, that the solutions to Maxwell’s equations can
be of four distinct types. In the Cartesian frame the categories are related to the
existence or otherwise of field components pointing in the propagation direction of
the wave. If we take this to be the z-direction, as in Fig. B.1, then we can make the
following statements:

Ez ¼ Hz ¼ 0 defines a TEM wave


Ez 6¼ 0; Hz ¼ 0 transverse magnetic ðTMÞ
Ez ¼ 0; Hz 6¼ 0 transverse electric ðTEÞ
Ez 6¼ 0; Hz 6¼ 0 hybrid mode ðEH or HEÞ

As we have seen the TEM solution occurs in unbounded free space. It also forms
in two wire transmission systems such as coaxial line. In hollow pipes, solutions are
of the TE or TM type. A preliminary insight into the nature of propagation in a
hollow rectangular waveguide is provided in Chap. 1. It is evident that the solutions
to Maxwell’s equations in this type of structure divide into two classes, namely
transverse electric (TE) modes and transverse magnetic (TM) modes. The modes

© Springer Nature Switzerland AG 2019 319


A. J. Sangster, Compact Slot Array Antennas for Wireless Communications, Signals
and Communication Technology, https://doi.org/10.1007/978-3-030-01753-8
320 Appendix B: Basic Waveguide Theory

Fig. B.1 Rectangular waveguide coordinate system

are distinguished by the presence of a field component in the direction of propa-


gation (z-direction in Fig. B.1)—a magnetic field (Hz) for TE modes and an electric
field (Ez) for TM modes as indicated above. These modes in waveguide can be
generated from potential functions (see Collin in Bibliography), with mathematical
efficiency recommending the use of a magnetic-type Hertzian potential for TE
modes and an electric type for TM modes. Employing the Collin terminology these
potentials can be expressed as:

Ph ¼ ^az Ph ðB:1Þ

and

Pe ¼ ^az Pe ðB:2Þ

Other potential function choices are possible depending on the dictates of the
problem to be solved. For example in antenna analyses the functions A (vector
magnetic potential function) and / (electric potential function) are solutions
commonly used (see Chap. 2). Here we concentrate on the TE mode solutions to
illustrate the typical waveguide analysis procedure. In terms of Ph the electric and
magnetic fields in the waveguide are given by:
Appendix B: Basic Waveguide Theory 321

E ¼ jxl0 r  Ph ðB:3Þ

H ¼ k02 Ph þ rr  Ph ðB:4Þ

On applying the Maxwell equations it follows that Ph is a solution of the wave


equation. That is, in phasor form:

r2 Ph þ k02 Ph ¼ 0 ðB:5Þ
pffiffiffiffiffiffiffiffiffi
For an empty waveguide k0 ¼ x l0 e0 ¼ 2p=k0 where k0 is the free-space
wavelength. Since the solution to Eq. (B.5) must represent z-directed propagating
waves in the waveguide, then we can assume that:

Ph ¼ ^az wh ðx; yÞ exp cz ðB:6Þ

where c ¼ a  jb is the complex propagation coefficient, while wh is a solution


of:

r2t wh þ kc2 wh ¼ 0 ðB:7Þ

rt operates on x and y only, while

kc2 ¼ k02 þ c2 ðB:8Þ

In the Cartesian coordinate frame (Fig. B.1), a wave equation such as Eq. (B.7)
can be solved by the method of separation of variables, which means that wh can be
expressed as Hz = X(x)Y(y). Consequently, on applying the boundary conditions
on E and H, and hence wh, at the perfectly conducting walls of the waveguide (see
Fig. B.1), the solution to Eq. (B.7) is easily found to be:
mpx npy
wh ¼ Cðcos cos Þ exp jbmn z ðB:9Þ
a b

where cmn ¼ jbmn for lossless waveguide.

B.2 TE Solutions for Rectangular Waveguide

With wh now known it is not difficult to establish the fields within the waveguide.
From Eq. (B.4) we obtain:

Hz ¼ r2t wh ¼ k02 wh ðB:10Þ


322 Appendix B: Basic Waveguide Theory

Consequently for the mnth mode we can write:


mpx npy
Hzmn ¼ C cos cos exp jbmn z ðB:11Þ
a b

In this equation m and n are integers such that m = 0, 1, 2 … 1, and n = 0, 1, 2


… 1 (n = m 6¼ 0). This solution also requires that:
mp2 np2
kc2 ¼ kcmn
2
¼ þ ðB:12Þ
a b

Hence from Eq. (B.8) we must have


mp2 np2
c2mn ¼ þ k02 ¼ kcmn
2
 k02 ðB:13Þ
a b

where k0 = x/c. In this equation k0 is the free-space propagation coefficient, kcmn


is the cut-off coefficient, while c is the waveguide propagation coefficient for the
mnth solution or mode. The other field components of the mode are derived from
Hz using the following relations derived from the Maxwell equations:

cmn @Hzmn
Hxmn ¼  2
kcmn @x
cmn @Hzmn
Hymn ¼ 2
kcmn @y
ðB:14Þ
jxl @Hzmn
Exmn ¼ 2 0
kcmn @y
jxl @Hzmn
Eymn ¼ 2 0
kcmn @x

It is important to observe, that these modal field expressions form (in mathe-
matical terms) a complete set. This means that any arbitrary field distribution in the
waveguide, due for example to the presence of a discontinuity such as a post or an
iris, can always be represented mathematically as a combination or summation of
the known modes of the empty waveguide by a process not unlike the use of
Fourier series to represent arbitrary periodic waveforms with sinusoids.
Appendix B: Basic Waveguide Theory 323

B.3 Frequency Characteristics of Waveguide Modes

For fast wave solutions in uniform metallic waveguides the propagation equation
has the general form:

c2mn ¼ kcmn
2
 k02 ðB:15Þ

where kcmn is a constant whose magnitude is governed by the size and shape of
the waveguide, and k=k0 is the plane wave phase constant for the interior medium.
For rectangular waveguide, we have already seen that:
mp2 np2
2
kcmn ¼ þ ðB:16Þ
a b

It has other dimensionally dependant values for other waveguides, such as cir-
cular waveguide and elliptical waveguide. However, for all such waveguides the
propagation equation has the same form, as expressed in Eq. (B.16).
In a lossless waveguide for which kcmn is real and constant, and
pffiffiffiffiffiffiffiffiffi
k0 ¼ x l0 e0 ðB:17Þ

is real and linearly related to frequency, two types of modal solution exist in
waveguide.
These are:
1. Those for which ko > kcmn giving

 1=2
c ¼ jbmn where bmn ¼ k02  kcmn
2
ðB:18Þ

bmn is termed the phase coefficient in rad/m for the mnth mode. The exponential
multiplier in the field expression now has the form:

expðjðbmn z  xtÞÞ

which implies that the mode in question propagates with no attenuation in a


lossless waveguide.
2. Those for which ko < kcmn giving

 2 1=2
cmn ¼ amn ¼ kcmn  k02 ðB:19Þ

amn is termed the attenuation coefficient in nepers/m. The exponential multiplier


in the field expressions now has the form
324 Appendix B: Basic Waveguide Theory

expðamn zÞ expðjxtÞ

and this implies that the mode produces a field which attenuates in z. It does not
propagate. Such a mode is said to be evanescent.
The frequency at which ko = kcmn is termed the cut-off frequency. It is the
frequency at which, in a loss-less waveguide, a given mode changes abruptly from
propagating to non-propagating. For any waveguide there is in general a separate
and distinct cut-off frequency for every TE mode, and every TM mode, in the mode
set. On the other hand it is not uncommon for a TE mode and a TM mode to have
the same cut-off frequency. Such modes are said to be degenerate. For a TE mode
and a TM mode in rectangular waveguide the cut-off frequency is evaluated as
follows:
Let xc denote the frequency at which ko = kcmn. Then:
r ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffiffi mp2 np2
xc l0 e0 ¼ þ
a b
r ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 
mp 2 np2
i:e: xc ¼ c þ
a b
s ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðB:20Þ
c m 2 n2
Hence fcmn ¼ þ 2
2 a2 b
c pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi
¼ ðm2 b2 þ n2 a2 Þ
2ab

Also, since c = fo, then

2ab
kcmn ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðB:21Þ
ð m b2 þ n2 a2 Þ
2

B.4 TE10 Mode

The mode with the lowest cut-off frequency, which means that it can exist alone in a
smooth empty waveguide, is the TE10 mode. The field expressions for this mode (in
magnitude), using Eqs. (B.11) and (B.14), are:
Appendix B: Basic Waveguide Theory 325

px
Hz ¼ cos
a
Hy ¼ 0
b a px
Hx ¼ j 10 sin ðB:22Þ
p a
Ex ¼ 0
ka px
Ey ¼ j g0 sin
p a

Each component in Eq. (B.22) is of course multiplied by the wave term

expðjðb10 z  xtÞÞ

The functional forms in Eq. (B.22) are sufficiently simple to permit the con-
struction of field distributions in pictorial form, as shown in Fig. B.2. This is also
true of the next dozen or so TE and TM modes as many texts on electromagnetic
theory demonstrate [see Bibliography].
For this TE10 mode the cut-off frequency is given by:
c
fc10 ¼ and
2a
kc10 ¼ 2a

In a typical X-band waveguide (8.0–12.4 GHz) with a broad dimension of 2 cm


this simple relation sets cut-off at 7.5 GHz (k0a = p in Fig. B.3).
A number of waveguide properties are encapsulated within the propagation
equation, which are very effectively illuminated by constructing a k0a − ba, or
Brillouin diagram, for the waveguide. The beauty of this form of representation is
that the slope of any line on the curve is representative of velocity, and Eq. (2.65)
plots as a simple circle when c (a) is real, and as a parabola, when c (b) is
imaginary. This is shown in Fig. B.3. The figure gives a quick and comprehensive
picture of the most important waveguide characteristics, such as cut-off frequencies
(p for TE10, 2p for TE20 etc.), bandwidth (distance between cut-off points on k0a
axis), phase velocities (slope of line from origin to the frequency of interest on the
TE10 mode propagation solution), group velocities (slope of the TE10 mode prop-
agation solution at frequency of interest), higher order mode rates of attenuation
(distance from k0a axis to intercept on aa solution at frequency of interest), and
frequency stratification of the modes. Usually only the first few TE modes are
shown (TE10, TE20, TE11, TE21 etc.) but in principle the complete TEmn mode
structure could be displayed. A similar diagram exists for TM modes. Note that the
parabolic curves depicting propagating solutions lie above the v = c line. Such
326 Appendix B: Basic Waveguide Theory

Fig. B.2 TE10 mode field patterns, a E and H-fields in the y-z plane, b E-field in the x-y plane, c E
and H-fields in the x-z plane

solutions for uniform waveguide are termed ‘fast’ waves. Below cut-off, k0a = p
for the TE10 mode, the solutions are evanescent as we have already seen and these
cut-off solutions are represented by the quadrant of a circle (blue dotted curve for
TE10 mode—it is not quite circular because of the slightly different axis scales). The
attenuation coefficient a in nep/m can be deduced from the horizontal distance from
the frequency of interest on the k0a axis to the dotted curve of the mode of interest.
Appendix B: Basic Waveguide Theory 327

Fig. B.3 Brillouin diagram for TE10 mode showing frequency (k0a) as a function of phase (b1oa)
for rectangular waveguide (dark blue curve: a = 2 cm, b = 1 cm), and as a function of a10a (green
dotted curve). Orange line represents the speed of light line v = c

Bibliography

1. R.E. Collin, Field Theory of Guided Waves (McGraw-Hill Book Co., New York, 1960)
2. R. Ferrari, An Introduction to Electromagnetic Fields (Van Nostrand Reinhold Co. Ltd., New
York, 1975)
3. C.T.A. Johnk, Engineering Electromagnetic Fields and Waves (Wiley, New York, 1988)
4. R.P. Feynman, R.B. Leighton, M. Sands, Lectures in Physics, Vol. II (Addison-Wesley
Publishing Co., London, 1972)
5. P.M. Morse, H. Feshbach, Methods of Theoretical Physics (McGraw-Hill Book Co., New
York, 1953)
6. J.D. Kraus, Electromagnetics (McGraw-Hill Book Co., London, 1984)
Appendix C
Waveguide Green’s Function
and Scattering Parameters

C.1 Green’s Function for Empty Rectangular Waveguide

For the moment method solution of broadwall slot radiators in empty rectangular
waveguide, the following are the relevant magnetic dyadic Green’s functions for the
geometry of Fig. 5.1.
(1) The exterior region usually treated as a half-space is (see Morse and Feshbach
and Collin in the Bibliography):

 mc ðr=r0 Þ ¼ ðI þ rrÞ exp jk0 jr  r0 j


G ðC:1Þ
k02 2pjr  r0 j

(2) The appropriate Green’s function for an arbitrary slot radiator in waveguide
[see Chap. 5, Eq. (5.5)] is the magnetic dyadic Green’s function. However the
problem can be simplified by noting that an inclined slot radiator in the
broadwall of waveguide as depicted in Fig. 5.1 can generally be resolved into
an x-directed (transverse) element and a z-directed (longitudinal) element.
Actually in practice broadwall waveguide slot radiators are either purely
transverse or wholly longitudinal. Inclined slots are hardly ever seen. This
means that the magnetic dyadic Green’s function for the feed waveguide can be
separated into a part that is relevant to the transverse slot computation and a part
that applies to the longitudinal slot.
As has already been hinted at above the magnetic dyadic Green’s function for
waveguide, see Collin in Bibliography, can be constructed by a summation of the
complete set of modes for the rectangular box forming the waveguide. That is:

© Springer Nature Switzerland AG 2019 329


A. J. Sangster, Compact Slot Array Antennas for Wireless Communications, Signals
and Communication Technology, https://doi.org/10.1007/978-3-030-01753-8
330 Appendix C: Waveguide Green’s Function and Scattering Parameters

8
> P
1
> 1
< 2jxe0 Hnþ ðrÞH
n ðrÞ z [ z0
 ma ¼
G n¼1
ðC:2Þ
> P1
>
: 2jxe
1
H þ
n ðrÞHn ðrÞ z\z0
0
n¼1

where

H
n ¼ ðhn þ hzn Þ expð Cmn zÞ ðC:3Þ

E
n ¼ ðen  ezn Þ expð Cmn zÞ ðC:4Þ

These vector fields, representing the normal mode field solutions for the
waveguide, are normalised to be in accord with:
ZZ
½en  hm   ^az dS ¼ dnm ðC:5Þ
Sx

while
mp2 np2
Cmn ¼ þ k02 ¼ kc2  k02 ðC:6Þ
a b

Computationally friendly forms for E 


n and Hn can be constructed by employ-
ing the potential function generators as outlined in Collin. Hence for TE modes:

H ¼ kc2 wh expð Cmn zÞ^az Cmn expð Cmn zÞrt wh


ðC:7Þ
E ¼ jxl0 ^az  rt wh expð Cmn zÞ

and for TM modes:

E ¼ kc2 we expð Cmn zÞ^az þ Cmn expð Cmn zÞrt we


ðC:8Þ
H ¼ jxe0 ^az  rt we expð Cmn zÞ

Collin provides the following forms for the potential functions:


mpx npy
wh ¼ cos cos
a   b
mpx ðC:9Þ
npy
we ¼ sin sin
a b

By combining Eq. (C.2) through (C.9) the Green’s function component relevant
to transverse slot modelling is as follows
Appendix C: Waveguide Green’s Function and Scattering Parameters 331

8 1 1 9
> P P npy >
>
> ðxe Þ2 np 2 2
ð Þ j  jÞ mpx0 npy >
b >
mpx
> 0 B expðC mn z z0 sin sin cos cos >
>
< m¼1 n¼1
b mn a a b >
=
 1
xx
Gma ðr=r0 Þ ¼ P1 P1
mp 2 2
2jxe0 >
>  ð a Þ Cmn Amn expðCmn jz  z0 jÞ sin a sin a cos b cos b >
2 mpx mpx0 npy npy
>
>
> >
>
>
:
m¼0 n¼0 >
;
n 6¼ m ¼ 0
ðC:10Þ

The factors Amn and Bmn arise from the normalisation process represented by
Eq. (C.10) and are given by:
eom eon
A2mn ¼ ðC:11Þ
jxl0 Cmn kc2 ab

4
B2mn ¼ ðC:12Þ
jxe0 Cmn kc2 ab

where eom and eon are equal to zero if m ¼ 0 or n ¼ 0 otherwise they are equal to
unity.
For the longitudinal slot radiator in rectangular waveguide it is shown in Collin
that the sum of modes formulation produces a function for which non-convergence
issues arise. However a direct solution to Eq. (5.5) in Chap. 5 together with the
appropriate waveguide boundary conditions has been determined by Rahmat-Samii.
The result is the following:

X1 X 1
eom eon mpx mpx0 npy npy0
 zz ðr=r0 Þ ¼
G expðCmn jz  z0 jÞ cos cos cos cos
ma
m¼0 n¼0
2abC mn a a b b
ðC:13Þ

(3) For a waveguide with finite wall thickness (T) the slot is treated as a rectangular
cavity of dimensions L  W  T, with a local coordinate system n; g; 1 as
indicated in Fig. 5.1. For this cavity the relevant magnetic dyadic Green’s
function (consistent with n-directed magnetic currents at open ends) can be
formulated heuristically following the summation of modes approach in Collin
(see Bibliography). The slot is viewed as a very short (length = T) of rectan-
gular waveguide (a replaced by L: b replaced by W). This yields:
332 Appendix C: Waveguide Green’s Function and Scattering Parameters

X
1 X
1    
 mb ¼ el0 en0 lpðg þ W2 lpðg0 þ W2 Þ npðn þ L2Þ npðn0 þ L2Þ
G cos cos sin sin
l¼0 n¼0
WL W W L L
 " #
1ðkn0pLÞ2 cosðkg 1Þ cosðkg ð10 þ TÞÞ
 1 6¼ 10
kg sinðkg TÞ cosðkg ð1 þ TÞ cosðkg 10 Þ
ðC:14Þ

where

np 2 lp
kg2 ¼ k02  ð Þ  ð Þ2 ðC:15Þ
L W

C.2 Longitudinal Slot in Conventional Waveguide

For a slot in waveguide, or in any metallic surface, excitation of the slot is asso-
ciated with magnetic field components aligned with its major axis, and hence
current density flows normal to this axis (see Chap. 1). Consequently for the
longitudinal slot in waveguide as shown in Fig. 5.1, with h = 0°, the component of
H10 which is effective in exciting the slot is the z-component H10z. From
Appendix B, this has the form:
p2 px

H10z ¼ A10 cos expð C10 zÞ ðC:16Þ
a a

with

A10 ¼ ðk0 Z0 b10 abÞ1=2

for fields normalised to unity power flow. Also



1=2
p2
C10 ¼ jb10 ¼ 2  k02
a

Hence substituting into Eq. (5.17) yields:

x1Zþ w=2 Z L=2


þ
p2 px 
1X N
þ spðz þ L=2Þ
H10r jz ¼  as H10z A10 cos expðC10 zÞ sin dzdx
2 s¼1 a a L
x1 w=2 L=2

ðC:17Þ

On combining Eqs. (C.16) and (C.17) the reflection coefficient (S11) for back-
ward scattering into the TE10 mode in the waveguide we obtain:
Appendix C: Waveguide Green’s Function and Scattering Parameters 333


N " #
A10 p pðx1 þ w=2Þ pðx1  w=2Þ X sp
S11 ¼ pffiffiffi sin  sin as L

j 2a a a s¼1 ðspL Þ2 b210


( ðC:18Þ
2 cos b102 L s odd

j2 sin b102 L s even

The transmission coefficient (S12) directly from the TE10 mode magnetic field, as
above, or from the relation:

S12 ¼ 1  ð1Þs S11 ðC:19Þ

By conservation of energy the power radiated (prad) by the longitudinal slot can
be deduced to be:

prad ¼ pinc  ðpt þ pr Þ ðC:20Þ

This can be expressed more conveniently as:


prad  
¼ 1  S211 þ S212 ðC:21Þ
pinc

C.3 T-Slot in Bifurcated Waveguide

As far as the calculation of the waveguide scattering and power radiated is con-
cerned the T-slot radiator differs little from the conventional longitudinal slot
presented above but using moment method results presented in Chap. 5, Section 5.4
. However the derivations are slightly complicated by the fact that the T-slot scatters
into both halves of the bifurcated waveguide. The internal scattering coefficients
þ
have therefore been considered to comprise two components S11 and S11 repre-
senting backward scattering into the left hand (x-positive) and right hand
þ
(x-negative) waveguides respectively, while S12 and S
12 are the corresponding
forward scattered components. With these definitions established it is not difficult to
show that:
 
b10 w L p
S
11 ¼ 2jA10 sin a1 r ; ðC:22Þ
2 2 a

for s = 1 since, as we have noted in Chap. 5, the first basis function is enough to
secure good results, and
334 Appendix C: Waveguide Green’s Function and Scattering Parameters

( q sin bpb sin qp


q2 b2
ðq 6¼ bÞ
rðp; qÞ ¼ 1 ðsin qp þ sin bpÞ
ðC:23Þ
2 qþb  pcosbp ðq ¼ bÞ

where b is the phase coefficient for the slot transmission line. In this case the
transmission coefficient can be calculated from:

S 
12 ¼ 1  S11 ðC:24Þ

As in the case of the longitudinal slot power conservation can be applied to


determine the power radiated by the T-slot, which generates the relation:

 2
prad þ 2 þ
¼ 1  S11 Þ þ ðS12 ðC:25Þ
pinc

This equation accommodates the double waveguide feed.

Bibliography

1. P.M. Morse, H. Feshbach, Methods of Theoretical Physics (McGraw-Hill Book Co., 1953)
2. R.E. Collin, Field Theory of Guided Waves (McGraw-Hill Book Co., 1960)
3. R.E. Collin, On the incompleteness of E and H modes in waveguides. Can. J. Phys. 51, 1135–
1140 (1973)
4. Y. Rahmatt-Samii, On the question of computation of Dyadic Green’s function at the source
region in waveguides and cavities. IEEE Trans. Microwave Theory Tech. 23, 762–765 (1975)
Appendix D
Green’s Function for Fin Loaded
Waveguide

The Green’s function for slow-waveguide periodically loaded with metallic fins can
be developed heuristically from the well-established empty rectangular waveguide
formulation (see Appendix C). The rectangular waveguide Green’s function in
Appendix C can be separated into three elements.
(1) The real part of the dominant mode term.
(2) The imaginary part of the dominant mode term.
(3) The imaginary part formed from the higher order all of which can generally be
assumed to be evanescent modes.
By analogy it is not unreasonable to suggest that a Green’s function for the
fin-loaded rectangular waveguide will display similar components. In relation to
item (3), the effect of periodic loading on the form of a potential Green’s functions
for a slow-wave structure has been studied elsewhere and clearly indicates that in
the space a  b above the corrugated surface the evanescent modes are essentially
the same as for the equivalent empty waveguide. Thus referring to Eq. (C.10) in
Appendix C we can deduce that the evanescent mode contribution to the require
Green’s function is:
8P 1 P 1  2 2 9
>
> ðxe0 Þ2 np npy >
>
b Bmn expðCmn jz  z0 jÞ sin a sin a cos b cos b >
mpx mpx0 npy
>
> >
>
> m¼1 n¼1 >
>
>
< 1  
>
=
1 P1 P

G
xx
ðr=r Þ ¼  mp 2 2
C A2
expðC mn j z  z 0 jÞ sin mpx
sin mpx0
cos npy
cos npy
=ma 0 mn mn
2jxe0 >
> m¼0 n¼0
a a a b b
>
>
>
> >
>
>
> n 6¼ m ¼ 0 >
>
>
: >
;
m ¼ 1 : n 6¼ 0
ðD:1Þ

Furthermore, with reference to item (2) above, in a waveguide feeding slot


radiators, we already know that the imaginary part of the dominant mode term has
no influence on a scattering calculation for a centred transverse slot and can be
equated to zero. This assumption applies equally to the centred transverse slot in the

© Springer Nature Switzerland AG 2019 335


A. J. Sangster, Compact Slot Array Antennas for Wireless Communications, Signals
and Communication Technology, https://doi.org/10.1007/978-3-030-01753-8
336 Appendix D: Green’s Function for Fin Loaded Waveguide

upper broadwall of periodic waveguide. In addition to the above comparisons


between conventional waveguide and periodic waveguide, the latter case has within
every modal solution embedded space harmonics, in order to satisfy the periodic
surface boundary condition. Fortunately, when it is operated as a feed waveguide
the presence of the space harmonics can largely be ignored, since the space har-
monic field components exist only in very close proximity to the fin tips and thus
exhibit negligible influence on slot excitation at the radiating wall, provided b is not
too small. This is unlikely, since generally b has to be large enough to permit useful
power flow through the waveguide.
For the formulation of a serviceable Green’s function for the periodic waveguide
feed we can therefore concentrate on the contribution of the real part of the dom-
inant HE11 mode namely G  xx ðr=r0 Þ. As with the conventional waveguide for-
<ma
mulation the xx component of the Green’s function can be deduced from the
x-directed magnetic potential:

ph ¼ ^ax wh ðD:2Þ

By analogy with the periodic waveguide field solutions presented in


Chap. 8 [Eqs. (8.17)–(8.21)] it is reasonable to suggest that:
mpx
ph ¼ ^ax cos cosh ac11 ðy  bÞ expðjb11 zÞ ðD:3Þ
a

For the coordinate system adopted in Fig. 8.9


Following the derivation steps detailed in Appendix C we eventually obtain:
(  1
 xx xe0 b210 1 sinh 2ac11 b px pxo
G <ma ðr=r0 Þ ¼ ^ax ^ax cos cos
b11 k02 ab 2 4ac11 b a a

xe0 b11 p2 1 sinh 2ac11 b 1 px pxo
þ ^az ^az sin sin
b10 k02 a3 b 2 4ac11 b a a
 )
jxe0 p 1 sinh 2ac11 b 1  px pxo px pxo 
 ^ax ^az 2 2 sin cos  cos sin expðjb11 jz  zo jÞ
k0 a b 2 4ac11 b a a a a

ðD:4Þ

As noted above the third term in Eq. (D.4) does not contribute to radiation from
a transverse slot in the y = b wall and can be ignored. Finally

 xx ðr=r0 Þ ¼ G
G  xx ðr=r0 Þ þ G
 xx ðr=r0 Þ ðD:5Þ
ma <ma =ma
Index

A novel, 272
Admittance reciprocity, 261
transfer, 39 reflector, 57
Ammeter, 39 retro-directive, 31, 286, 299
Ampere’s law, 47, 48 simulation, 167
Amplifier synthesis, 133
buffer, 289 thinning, 260
voltage, 288 Aperture
Antenna antenna, 57, 69, 78
3-D, 21 blockage, 182, 225
1-D format, 269 circular, 293
active, 250, 267, 302 dual-mode, 298
aperture, 57, 75 effective area, 79
broadside, 248 elliptic, 296
cavity resonator, 216, 249, 267, 274 field distribution, 251
compact, 8, 28, 40, 151, 243 linear, 72
configurable, 263 radiating, 251
conformal, 270 rectangular, 72, 76, 79
dielectric rod, 57 source, 21, 35, 37, 62, 69, 80
directional, 71 theory, 216
directivity, 78 two dimensional, 72, 74, 78
efficiency, 72 Array
Fabry-Perot, 262 analysis, 140
flat plate, 169 antenna, 140
gain, 79 apeture, 293
horn, 60 beam squint, 173, 179
impedance, 38 bidirectional, 147
infrared, 216 binomial, 148
integral equation, 226 boresight, 173
leaky-wave, 19, 57, 205, 208, 213, 215, broadside, 134, 140
243, 260, 267, 272, 281 cavity resonator, 249
lens, 57 colour coded, 225
long-slot, 207 compact, 179, 220, 270, 285
macroscopic, 38, 39 control, 223
mm-wave, 216, 237 design, 154

© Springer Nature Switzerland AG 2019 337


A. J. Sangster, Compact Slot Array Antennas for Wireless Communications, Signals
and Communication Technology, https://doi.org/10.1007/978-3-030-01753-8
338 Index

Array (cont.) synthesis, 133


design iteration, 166 theory, 228
directivity, 144 tracking, 292
efficient, 151 transverse slot, 209–211
element current, 152 travelling-wave, 151, 154, 156, 158, 159,
element excitation, 169 209
element number, 149 uniform, 132
endfire, 137, 147 uniform linear, 139
factor, 95, 155 unstaggered, 155, 158, 160
flat, 223 Attenuator
flat plate, 170 calibrated, 122
frequency scanned, 182 Auxillary, 43
gain, 140, 151 potential, 60
high directivity, 169 scalar, 43
inexpensive, 151 Azimuth plane, 70
infinite, 221, 262
infinite analogy, 217 B
isotropic, 136 Balanis, 79
leaky-wave, 213, 281 Bandwidth, 1
length, 142, 147, 159 ellipticity
linear, 134, 220 enhancement, 281
long, 139 improvement, 262
low sidelobe, 164 Base
main lobe, 136 corrugated, 173
matched, 169, 173 Basis
model, 176 function, 107, 109, 111, 120, 121, 333
multi-element, 171 pulse, 109, 112, 113
multi-lobed, 134 sinusoidal, 99
no sidelobes, 148 trigonometric, 109, 118
passive, 223, 228 Beam
patch, 171 backward endfire, 247
pattern, 134 conical, 260
periodic, 131, 184 control, 267
phase coefficient, 132 electronic control, 263
phase shift, 132, 148 endfire, 269
phased, 179, 215 focused, 223, 225
phasing, 226 forward scattered, 274
planar, 132, 149, 182, 183, 215 offset, 266
polynomial, 133, 140 power density, 71
principal lobe, 139 primary, 269
quasi-phased, 262 radial scan, 260
reconfigurable, 263 scanning, 246, 247, 260, 262
reference element, 219 secondary, 156
resonant, 20, 167–169 split, 259
retro-directive, 30, 291 split maxima, 259
rules, 273 squint, 179
scanned, 138 switching, 292
semi-infinite, 237 Beamwidth, 71, 136
shorted, 167 half-power, 136, 138
shunt slot, 152 HPBW, 71, 136
sidelobe, 139 minimum, 139
simulation, 159 optimisation, 263
slotted waveguide, 154, 169, 179, 216 Bessel
square patch, 270 function, 202
Index 339

Binomial, 141 Complex, 53


series, 95, 148 z-plane, 141
Boundary Computational
conditions, 84, 86 accuracy, 96
E-wall, 188 speed, 96
field interface, 251 Computer
H-Wall, 188, 190 simulation, 159
periodic, 227 Concerto, 126
Boundary conditions, 4, 15 Conductance
Brillouin, 14 continuous, 163
diagram, 184 Conductor
Broadside artificial, 243, 260
beam, 259 artificial magnetic, 244
perfect, 268
C perfect electric, 243, 250
Cancellation, 25 perfect magnetic, 249
Capacitive Conjugation, 290
stored energy, 184 complex, 40
Capacitor, 8 Continuity
Cassegrain, 230, 232, 237 equation, 41
Casting Contour
high precision, 173 power density, 136
Cavity, 125 Convergence
high-Q, 24, 128 rate, 219
mode, 24 Copper
planar, 243 clad, 27
resonator, 28 Corner
square,, 25 reflector, 30
stored energy, 245 Coulomb law, 45
Cells Coupled
periodic, 185, 186 regions, 127
Charge Coupler
density, 37 precision, 122
free, 61 Coupling
magnetic, 41 electromagnetic, 24
Chebyschev pattern, 163 mechanism, 250
Circle mutual, 131, 235
diagram, 143, 144 self, 109, 200
unit, 142 Cross
Circuit coupling, 109, 200
conjugating, 292 polarisation, 293, 297
equivalent, 118 Cross-polar, 20
resonant LC, 271 sidelobes, 19
theory, 39 Cross-slot, 248, 293
Circuital law, 48 Curl
Circular, 25 operation, 41
polarisation, 8 Current
Circulator, 286, 287 coil, 47
Clockwise, 142 conduction, 37
Communications, 10 DC, 47
mobile, 132 density, 37, 93, 217
Compact element, 88, 217
antenna, 128 imaged, 93
array antenna, 151 impressed, 37
340 Index

Current (cont.) current, 37


lattice, 233 dipole, 52
magnetic, 37, 41, 233 potential, 43
short element, 50, 52 source, 63
wire, 48 Electric field, 16
Current element Electromagnetic
short, 93 3-D simulation, 261
testing, 251 boundary, 184
Cylinder, 91 classical, 35
coupling, 10, 200
D field transition, 250
Das, 107, 109 full-wave solver, 249
De-embed, 122 leakage, 244
Demodulation, 289 light, 2
Determinant, 139, 196, 197 model, 10, 82, 104
Dielectric, 19 modelling, 10, 81
density, 248 plane wave, 1
loss, 5 power, 116, 151
low loss, 180 radiation, 49
relative permittivity, 246 radio-wave, 2
resonator, 256 refraction, 278
Differential simulation, 8
equation, 43 solver, 17, 42, 198
operator, 41 Transverse electromagnetic (TEM), 1
Diffraction, 63–65 TEM wave, 35
theory, 234 wave, 35, 234
Diffusion Electron beam, 170
bonding, 175 Electrostatic, 44
Diode analogy, 44
capacitance, 265 Elevation plane, 70
varactor, 264 Elliptic
Dion, 165 function, 202
Dipole Ellipticity
lattice, 218 bandwidth, 297–299
radiation, 88 Endfire
scatterer, 218 beam, 269
test, 252 Energy, 54
Dirac, 82 capacitive, 53
Directivity, 78, 137 conservation, 279
Improvement, 146 inductive, 54
optimum, 138 leakage rate, 244
Discontinuity lobes, 54
fields, 104 storage, 53
Discretisation, 97, 99 Equation
Divergence, 40 characteristic, 184
Domain convergent, 219
entire, 258 determinantal, 196
spectral, 313, 316, 318, 332 differential, 35, 41, 81, 84
Duplexer, 291 homogeneous, 83
Dyadic, 86 integral, 217, 226
unit, 86 leaky-wave, 265
second order, 35, 41, 81
E simultaneous, 196
Electric transcendental, 119
Index 341

Equi-field polymide, 174


contours, 210 Fourier
Equivalence transform, 67, 69
principle, 57, 59 transformation, 254
Etching transform pair, 57
photographic, 170 Fourier series, 186
Ethernet, 170 Free-space
Exponential environment, 91
argument, 31 Frequency
behaviour, 188
F carrier, 286
Fabrication characteristic, 184
3-D printing, 216 compensation, 263
accuracy, 176 difference, 301
laminated, 176 dispersion, 234
Fabry-Perot, 216 independent, 267
cavity, 243, 245, 249, 250 intermediate, 289
couple cavities, 261 resonant, 125, 126
resonator, 252, 260 scan, 181
Faraday’s law, 53 scan equation, 181
Far-field, 68, 78 scanned, 207
Feed shift, 291
corporate, 172 sum, 290
serpentine, 181 sum signal, 289
sinuous, 182, 183 up-converted, 288
slow-wave, 183 Function
Ferrite, 48 basis, 128
Field testing, 121
continuity, 104
contours, 52 G
cosinusoidal, 109 Gain, 22
coupling, 40 Galerkin method, 108
equivalence, 79 Gauss law, 45
exponential decay, 191 Generator, 38
monotonic, 245 Grating
reflected, 110 lobe, 139, 142
scattered, 210 Green’s
Field theory, 35 equation, 105
Filter Green’s function, 82–84, 93, 200, 217, 227,
bandpass, 183, 289 233, 240
directional, 286 electromagnetic, 89
LCR circuit, 188 method, 90
loop, 289 modified, 91, 95
Finite difference, 189 Ground
Finite element, 42, 97, 188 conducting plane, 249
Finite element method, 159 Ground plane, 95
Floquet perfect, 95
modes, 254 textured, 270
periodicity, 193 Guide
relationship, 233 wavelength, 19
theorem, 189, 190, 219, 254
Flux H
magnetic, 48, 306 Half-space, 35, 205, 221
Foil backward, 333
342 Index

Hankel K
function, 220 Kernel, 82
Hansen-Woodyard, 138 Kirchoff, 62
Helmholtz, 41, 84 Kirchoff law, 43
equation, 87 Kronecker delta, 197
vector equation, 41, 44
Helmholtz equation, 105 L
Hemisphere, 79 Laser
Hertzian machining, 170
dipole, 88, 90, 93 Law
element, 217 inverse square, 45, 54
Hertzian dipole, 50 Leaky
Heterodyne, 299 ladder, 212
Heuristic, 7, 81 Leaky-wave
Horizontal plane, 70 1-D scan, 269
Horn antenna pattern, 249
circular, 230 array antenna, 249
corrugated, 230 attenuation, 207, 212
phase centre, 223 linear array, 209, 267
pyramidal, 236 long slot antenna, 207
Huygen phase shift, 272
method, 89 phenomenon, 211
principle, 63 radiation, 262
Hybrid slit, 217
coupler, 25 slit width, 217
quadrature, 297 Lens
flat, 280
I optical, 280
Illumination, 71 parabolic, 280
uniform, 71 Line
Image current, 146
focused, 280 parallel, 151
Impedance, 38 shorted, 119
characteristic, 55 voltage, 118
complex, 205 Linear
input, 116, 119 operator, 82
load, 119 polarisation, 8, 115, 301
mutual, 216 space, 82
Inductor, 8 Linear accelerator, 183
Inhomogeneous, 41, 84 Load
Inmarsat, 31 matched, 154
Interference Specification, 165
pattern, 65 Lobe
Iris grating, 176
capacitive, 183, 206 grating condition, 219
Isotropic primary, 156
elements, 155 Longitudinal slot, 210
Iteration, 166 Lord Rayleigh, 35
Lorentz
J condition, 43
Junction reciprocity, 39
summing, 290 Loves principle, 59
Index 343

M T-junction, 301
Macroscopic Microstripline, 173
values, 275 Microwave, 4
Magic-tee, 122 circuit, 25
Magnet frequency band, 215
permanent, 47 integrated circuit, 173
Magnetic Mirror, 12
charge, 37 flat, 223
charge density, 61 Mixer, 289, 301
conductor, 271 conjugating, 286
current, 37, 200 gate, 300
current density, 59 heterodyne, 299
current element, 200 quasi-optical, 301
flux density, 47, 48 up-converting, 290
source, 63 Mobile, 10
vector potential, 43, 51 communications, 301
Magnetic field, 14 platform, 301
Magnetostatic Mode
analogy, 44, 47 anti-symmetrical, 188
Magnetostatics, 43 backward TE, 272
Mainbeam backward wave, 279
Mainlobe cut-off, 188
nose, 78 degenerate, 293, 324
width, 71 dominant, 204
Material evanescent, 15
double negative, 275 even, 9, 22
MATLAB, 189 fast, 213
Matrix, 200 hybrid, 195
entries, 199 improper, 204
equation, 258 leaky-wave, 268
impedance, 271 odd, 9, 22
manipulation, 200 orthogonal, 15
Matrix form orthogonality, 197
Maximum parallel plate, 259, 262, 265, 268
principal, 134 perturbation, 206
secondary, 134 quasi-TE10, 245
Maxwell, 35, 84 receive, 38
equations, 81 slow-wave, 195
Measurement sum of, 331
error, 113, 126 surface wave, 253
Medium symmetrical, 188
anisotropic, 45 TE, 10
homogeneous, 45 TE10, 13
inhomogeneous, 45 TEM, 9
isotropic, 39 transmit, 38
non-linear, 45 transverse resonance, 202, 205
Metallodielectric, 275 Model
Metamaterial, 267, 272 ComsolL, 216
frequency selective, 272 T-block, 235
Meta-structures, 179 Modulator
Metasurface, 275 IQ, 290, 291
Microstrip, 28 Moment method, 10, 21, 96, 99, 189, 190, 199,
feed, 293, 297 200, 205, 226, 227, 233
patch, 300 concise, 97
344 Index

Moment method (cont.) square, 230


periodic, 249 standard, 232
reciprocal, 281 trapped mode, 7
Mutual Pattern
coupling, 200 beam splitting, 260
interactions, 216, 226 cosine, 169
doughring, 259
N fan shaped, 269
Near field, 53 far-field, 143
Network first null, 134
distribution, 151 H-plane scan, 259
equivalent, 147, 152 low sidelobes, 247
theory, 255 major planes, 78
Neural network, 103, 128 maximum, 134
Nulls modification, 224
equispaced, 146 multiplication, 131, 136, 137, 142
reflected beam, 274
O sidelobe free, 259
Obstacle split maxima, 260
regular, 95 synthesis, 143
Oliner, 107, 109 Periodic
Omni-directional, 20 cells, 188
Operator function, 186, 193
linear, 82 interface, 194
Optical loading, 183
device, 278 passive surface, 274
frequency, 280 semi-infinite surface, 234
ray, 312 structure, 179
Orbit Permeability, 48, 275
satellite, 26 negative, 277
Orthogonal, 48, 51 tensor, 45, 48
Ports, 293 Permittivity, 37
Oscillator complex, 37
backward wave, 279 negative, 277
crystal, 289 relative, 180
local, 286 Perturbation
voltage controlled, 289 theory, 206
Phase, 25
P adjustment, 238
Parabolic, 26 advance, 263
equivalent, 253 agile, 263
profile, 223 agile cells, 264
Parasitic, 10 approximation, 198
Patch coefficient, 187
Antenna, 7 coherence, 281
dual-mode, 300 colour coded, 225
dual-polarised, 294, 300 comparison, 288
equi-size, 230 computation, 113, 226
geometrical range, 232 condition, 245
mode, 297 conjugation, 40, 286, 287, 290, 299
resonant, 223 conjugation circuit, 288
ring loaded, 232 constant, 289
round, 232 delay, 290
split, 263 detector, 289
Index 345

groove depth, 226 real, 53


line-length adjustment, 231 relative, 78
locked, 289 transfer, 38
locked-loop, 287, 289 Poynting
matching, 280 vector, 6
profile, 221 Poynting vector, 279
quadrature, 10, 25, 115 Processing
receive, 289 IQ, 290
shifter, 151 Processor
transmit, 291 Signal, 285
Phasor, 36, 218 Product
form, 36, 44, 47, 49 inner, 97
Pixellation, 27 Propagation
Plane wave, 31 coefficient, 64, 190, 191
incident, 219 complex, 218
oblique incidence, 255 equation, 191
refraction, 277
scattering, 251 Q
Polar Quadrature, 9
Form, 69 Quality factor, 113, 122
Polarisation, 8
agile, 24 R
agility, 7 Radar, 15
circular, 25, 297, 298, 299, 300 airborne, 169
linear, 8 beacon, 167
orthogonal, 25 cross-section, 30, 285
perpendicular, 12 frequency scanned, 202
slot, 16 monopulse, 285
switching, 22 three dimensional, 229
transverse, 16 Radiation
Polynomial 3D pattern, 69
Pole, 147 broadside, 260
root diagram, 143 circular polarised, 297
roots, 141, 142 condition, 72
Port dipole, 88
even-mode, 22 directivity, 28
odd-mode, 23 dough-ring, 52
sampling, 122 dual-polarised, 292, 293
Potential far-field, 68
auxiliary, 42 isotropic, 52, 134
function, 42, 190 loss, 72
magnetic vector, 219 mechanisms, 20
Power near field, 248
backward, 162 omni-directional, 17, 21
complex, 53 pattern, 1, 90, 95, 134
computing, 216 pattern cuts, 77
conservation, 159, 162, 165 pattern null, 70
density flow, 78 pattern sidelobes, 70
density radiated, 248 pencil beam, 28
forward, 162 point source, 254
maximum, 38 polar pattern, 134
normalised, 152, 163, 200, 201 principal beam, 20, 28
radiated, 38, 111, 122, 201 resistance, 56
ratio, 39 slit, 66
346 Index

Radiation (cont.) Reflector


slot, 8, 35, 223 antenna, 285
sources, 90 corner, 285
Radiator flat, 27
T-slot, 116 focused, 223
Radiometry, 169, 216 parabolic, 22, 27, 182, 216, 230
Ray, 31 pixellated, 237
Ray optics, 246, 280 Refraction, 277
Reaction, 40 negative index, 279
Reciprocity, 38, 39, 251, 259 Refractive index, 272
Reconfigurable negative, 276
component, 263 Regions
PRS, 263 coupled, 190
Rectangular, 10 Regula falsi, 119
form, 69 Relativity
Reflectarray, 26, 223, 228, 230 special, 43
3-D printed, 223, 226, 237 Residues
all metal, 234 calculus of, 197
bandwidth, 230, 232 Resonator, 24
die-cast, 239 coupled, 251
dielectric block, 239 Fabry-Perot, 28
edge effect, 281 parallel plate, 245
element phase, 228 Retardation, 47
finite, 228 potential, 49
focused, 235 Retro-direction, 285
gain/bandwidth, 239, 243, 267 array, 30
groove, 227, 235
groove depth, 235 S
iteration, 228 Satellite, 26
metal protrusion, 227 communications, 169
microstrip, 234 gyrating, 292
mm-wave, 237 tracking, 31
multi-layer Scalar
multiple groove, 234 operator, 82
parabolic equivalent, 230 Scan
pattern, 231 range, 213
phase comparison, 288 sensitivity, 181, 207
phase linearity, 248 Scanning
pill-box, 232 beam, 29
pixellated, 237 leaky-wave, 29
scattering, 250 Scatterer
simulation, 227 metamaterial, 272
state-of-the-art, 240 multiple, 217
Reflection, 317 specular, 221
angle, 245 wire, 217
coefficient, 111, 112, 122, 125, 161 Scattering
linear phase, 230 Conditions, 274
multiple, 243 parameters, 200
oscillatory, 239 T-slot, 121
partial, 247 Schelkunoff, 143
phase, 224, 226 Screen
phase correction, 239 conducting, 62
phase linearity, 231 infinite, 35
phase range, 231 semi-transparent, 221
Index 347

Self-coupling, 200 magnetic current, 251


Septum, 10 off-set, 153
Septum slot, 23, 116 orthogonal, 8, 115
Series parasitic, 10, 301
infinite, 197 power radiated, 152
Sheet radiator, 6, 10, 212
conducting, 93 reference, 257
Short resonant, 167
current element, 93 resonant length, 16
Sidelobe round-ended, 113, 126
design level, 165 scattering, 10
specification, 165 septum, 10
Signal series, 18, 210
anti-phase, 171 shunt, 17, 210
fidelity, 292 spacing, 156, 160
isolation, 292 square-ended, 113
mixing, 286 staggered array, 155, 157
modulated, 289 staggering, 17
receive, 289 susceptance, 167
retro-directive, 299 T-shaped, 8
RF leakage, 299 tilt, 19
tracking, 292 transverse, 210, 211
Silicon, 171 travelling-wave fed, 151
Simulation, 8 unstaggered array, 158
COMSOL, 1, 188 X-slot, 293, 294
CONCERTO, 126 Slow-wave, 183
full-wave, 261 Slow-waveguide, 195, 202
HFSS, 188 cut-off, 213
Simultaneous, 106 feed, 202
Simulator fields, 267
electromagnetic wave, 188 mode, 243
Sink, 37 phase diagram, 185
Sinusoidal, 2 Snell;s law, 12, 246, 278
Slot contravention, 277
admittance, 152 Software
alignment, 152 analysis, 1
array, 16, 213, 215 full-wave, 261
array applications, 169 Solution
broadwall, 210 moment method, 103
circuit model, 159 valid, 104
clearance, 10 Solver
compact, 8, 10, 12, 15, 18, 21, 96, 101, 103, finite element, 216
115, 216, 220, 272 full-wave, 198
conductance, 153, 169 Sommerfeld, 84
coupling, 299 Source, 35, 81, 230
discontinuity, 15 aperture, 21, 37
doublet, 22 crystal controlled, 122
dumbbell, 297 equivalent, 63
edge, 19 field, 35
electric field, 16, 19 free, 42
inclined, 211 frequency agile, 181
ladder, 213 function, 41
linear array, 202 image, 91
longitudinal, 22, 210 isotropic, 39
348 Index

Source (cont.) isodensity, 76


point, 83 partially reflective, 28
secondary, 63, 219 passive, 221
Source free, 35 periodic, 194
Space periodic spacing, 274
bounded, 35 Protrusion, 227
domain, 186 quasi-curved, 237
harmonic, 186–189, 198 reconfigurable, 272, 281
homogeneous, 38 refraction, 278
isotropic, 38 roughness, 126
real, 70 scatterer, 217, 221
regular, 95 textured, 271
symmetrical, 95 tunable impedance, 270
Spacing wave, 19, 213, 218
inter-element, 146, 147 Surface wave, 222
optimum, 138 Switch
Spectral varactor diode, 263
domain, 251, 254, 255
Spectrum, 69, 78 T
angular, 65, 68 Target, 30
electromagnetic, 216 reflector, 30
frequency, 254 Taylor function, 163
function, 65 Technique
plane wave, 65, 216 cavity, 28
Sphere, 40 iterative, 103
Stevenson Tensor, 45, 48
slot, 152 Testing function, 98
Stripline, 4, 28 Theorem
feed, 25 divergence, 40
Structure Theory
3-D periodic, 280 polynomial, 141
band-gap, 275 transmission line, 166
complex periodic, 275 Time
fin-loaded, 184 negative, 281
periodic, 186 retardation, 47, 50
Substrate, 25 reversal, 281
Duroid, 232 Tracking
microstrip, 251 retro-directive, 285
Taconic, 230 Transcendental, 119
thinning, 261 Transform
Superposition, 65, 132 pair, 68
Surface Transistor
additional, 263 Transmission
capacitance, 268 coaxial line
corrugated, 195 coefficient
current, 19 line
current density, 57 line model
electromagnetically porous, 243 microstrip
equiphase, 223, 224 stripline
frequency selective, 179, 218 Transmission line, 151
high impedance, 243, 249, 250, 260 analogue, 206
Huygen, 60 equation, 167
impedance, 243 lossy, 165
inductance, 271 model, 154, 161
Index 349

Transmitter, 55 microwave, 155


impedance, 55 milli-metre, 151
Trapped wave, 223 plane, 1
Travelling-wave pattern, 10
array, 151 polarised, 12
tube, 183 reflected, 221
Trentini, 243 scattered, 273
Trigonometric sinusoidal, 3
function, 186 slow, 183
MoM, 189 spherical, 72
Trihedral, 30 standing, 118
Triplate, 25 TE, 12
TM, 12
U trapped, 12, 203
Unique travelling, 286
Solution, 83 trough, 12
Uniqueness Wavefront, 12
Theorem, 59 Incident, 274
Waveguide, 91
V anti-phase, 171
Varactor diode, 265, 269 attenuation, 205
embedded, 269 bifurcated, 22, 23
Variational method, 190 cavity, 24
Vector characteristic admittance, 152
equation, 41 choke, 17
lamellar, 43 circular, 183
potential, 84, 86, 88 common wall, 23
product, 142 corrugated, 198
rotational, 43 cylindrical, 262
Velocity dielectric, 202
group, 279 dielectric loaded, 202
phase, 3 double-layer, 174
Vertical plane, 70 dual-mode, 23
Veselago, 277 empty, 187
Vias equation, 187
metallised, 28 fast, 184
Voltmeter, 39 feed, 17, 199
Vu Khac, 112 fin loaded, 183, 187, 191, 198
grooves, 17
W hollow, 174
Wave hybrid, 24
cylindrical, 63, 65 iris, 184
diffracted, 65 iris loaded, 185
equation, 42, 187 irises, 184
evanescent, 277 laminated, 173, 175, 177
fast, 213 load, 155
focusing, 223 magic-T, 24
guided, 12 matched load, 163
impedance, 116 multi-modal, 186
leaky, 213 parallel plane, 14
350 Index

parallel plate, 91 voltage, 152


periodic, 184, 191, 197 wavelength, 206
periodically loaded, 183, 213 Waveguide dielectric, 19
perturbed, 206 Wavelength, 15, 19
planar array, 173 free space, 55
post-wall, 173 TE10 mode, 152
power loss, 182 Wireless
rectangular, 10, 210 interfaces, 170
resonator, 24 mm-wave, 171
ridged, 234 networks, 176
scanned array, 138 streaming, 170
serpentine, 202
sinuous, 181, 202 Y
slot, 16 Yacht, 30, 285
slot array, 167, 169
slotted, 183 Z
slow, 186 Z-transform, 131, 78, 139, 230, 281
square, 24
substrate integrated, 173

You might also like