Approximate Antenna Analysis for CAD
Approximate Antenna Analysis for CAD
Hubregt J. Visser
© 2009 John Wiley & Sons, Ltd. ISBN: 9780470512937
Approximate Antenna Analysis for CAD Hubregt J. Visser Antenna Engineer, The Netherlands
A John Wiley and Sons, Ltd, Publication
This edition first published 2009 © 2009 John Wiley & Sons Ltd Registered office John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester, West Sussex, PO19 8SQ, United Kingdom For details of our global editorial offices, for customer services and for information about how to apply for permission to reuse the copyright material in this book please see our website at www.wiley.com. The right of the author to be identified as the author of this work has been asserted in accordance with the Copyright, Designs and Patents Act 1988. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, except as permitted by the UK Copyright, Designs and Patents Act 1988, without the prior permission of the publisher. Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic books. Designations used by companies to distinguish their products are often claimed as trademarks. All brand names and product names used in this book are trade names, service marks, trademarks or registered trademarks of their respective owners. The publisher is not associated with any product or vendor mentioned in this book. This publication is designed to provide accurate and authoritative information in regard to the subject matter covered. It is sold on the understanding that the publisher is not engaged in rendering professional services. If professional advice or other expert assistance is required, the services of a competent professional should be sought. Library of Congress CataloginginPublication Data Visser, Hubregt J. Approximate antenna analysis for CAD / Hubregt J. Visser. p. cm. Originally presented as author’s thesis–Ph. D. Includes bibliographical references and index. ISBN 9780470512937 (cloth) 1. Antennas (Electronics)–Computeraided design. 2. Electromagnetic fields–Computer simulation. I. Title. TK7871.6.V569 2009 621.382’4–dc22 2008041825 A catalogue record for this book is available from the British Library. ISBN 9780470697160 (H/B) Set in 10/12pt Times by Sunrise Setting Ltd, Torquay, UK. Printed in Great Britain by Antony Rowe.
Contents
Preface Acknowledgments
xi xiii
Acronyms
xv
1
Introduction 1.1 The history of Antennas and Antenna Analysis 1.2 Antenna Synthesis 1.3 Approximate Antenna Modeling 1.4 Organization of the Book 1.5 Summary References
1 1 5 7 9 12 13
2
Intravascular MR Antennas: Loops and Solenoids 2.1 Introduction 2.2 MRI 2.2.1 Magnetic Properties of Atomic Nuclei 2.2.2 Signal Detection 2.3 Intravascular MR Antennas 2.3.1 Antenna Designs for Tracking
19 20 22 22 24 27 28
vi
CONTENTS
2.3.2 Antenna Designs for Imaging MR Antenna Model 2.4.1 Admittance of a Loop 2.4.2 Sensitivity 2.4.3 Biot–Savart Law 2.4.4 Model Veriﬁcation Antenna Evaluation 2.5.1 Antennas for Active Tracking 2.5.2 Antennas for Intravascular Imaging 2.5.3 Antenna Rotation In Vitro Testing 2.6.1 Sensitivity Pattern 2.6.2 Tracking Antenna Synthesis 2.7.1 GeneticAlgorithm Optimization Safety Aspects 2.8.1 Static Magnetic Fields and Spatial Gradients 2.8.2 Pulsed Gradient Magnetic Fields 2.8.3 Pulsed RF Fields and Heating Conclusions Appendix 2.A. Biot–Savart Law for QuasiStatic Situation References
30 30 34 40 41 43 58 59 65 71 75 75 77 80 80 86 87 88 88 89
PCB Antennas: Printed Monopoles 3.1 Introduction 3.2 Printed UWB Antennas 3.2.1 Ultrawideband Antennas 3.2.2 TwoPenny Dipole Antenna 3.2.3 PCB UWB Antenna Design 3.2.4 BandStop Filter 3.3 Printed Strip Monopole Antennas 3.3.1 Model of an Imperfectly Conducting Dipole Antenna 3.3.2 Dipole Antenna with Magnetic Coating 3.3.3 Generalization of the Concept of Equivalent Radius 3.3.4 Equivalent Dipole with Magnetic Coating 3.3.5 Validation
97 97 99 99 100 100 109 117
2.4
2.5
2.6
2.7 2.8
2.9
3
90 92
118 121 122 125 125
vii
CONTENTS
3.3.6 3.4
4
5
MicrostripExcited Planar Strip Monopole Antenna Conclusions References
127 135 136
RFID Antennas: Folded Dipoles 4.1 Introduction 4.2 Wire FoldedDipole Antennas 4.2.1 Symmetric FoldedDipole Antenna 4.2.2 Asymmetric FoldedDipole Antenna 4.3 Impedance Control 4.3.1 Power Waves 4.3.2 Short Circuits 4.3.3 Parasitic Elements 4.4 Asymmetric CoplanarStrip FoldedDipole Antenna on a Dielectric Slab 4.4.1 Lampe Model 4.4.2 Asymmetric CoplanarStrip Transmission Line 4.4.3 Dipole Mode Analysis 4.5 FoldedDipole Array Antennas 4.5.1 Reentrant FoldedDipole Antenna 4.5.2 SeriesFed Linear Array of Folded Dipoles 4.5.3 Model Veriﬁcation 4.5.4 Inclusion of Eﬀects of Mutual Coupling 4.5.5 Veriﬁcation of Modeling of Mutual Coupling 4.6 Conclusions References
139 139 142 142 144 146 147 150 152 153 155 157 166 169 170 171 172 174 176 178 179
Rectennas: Microstrip Patch Antennas 5.1 Introduction 5.2 Rectenna Design Improvements 5.3 Analytical Models 5.3.1 Model of Rectangular Microstrip Patch Antenna 5.3.2 Model of Rectifying Circuit 5.4 Model Veriﬁcation 5.5 Wireless Battery 5.5.1 Single Rectenna 5.5.2 Characterization of Rectenna 5.5.3 Cascaded Rectennas
183 183 185 187 187 193 198 200 202 203 204
viii
CONTENTS
5.6 5.7
5.8
6
Power and Data Transfer RF Energy Scavenging 5.7.1 GSM and WLAN Power Density Levels 5.7.2 GSM Mobile Phone as RF Source Conclusions References
Large Array Antennas: OpenEnded RectangularWaveguide Radiators 6.1 Introduction 6.1.1 Mode Matching and Generalized Scattering Matrices 6.2 Waveguide Fields 6.2.1 TE Modes 6.2.2 TM Modes 6.2.3 Transverse Field Components 6.3 Unit Cell Fields 6.3.1 TE Modes 6.3.2 TM Modes 6.3.3 Transverse Field Components 6.4 CrossSectional Step in a Rectangular Waveguide 6.4.1 Boundary Conditions Across the Interface 6.4.2 Creation of a Finite System of Linear Equations 6.4.3 Matrix Formulation and GSM Derivation 6.5 Junction Between a Rectangular Waveguide and a Unit Cell 6.5.1 GSM Derivation 6.6 Dielectric Step in a Unit Cell 6.6.1 GSM Derivation 6.7 FiniteLength Transmission Line 6.7.1 GSM Derivation 6.8 Overall GSM of a Cascaded RectangularWaveguide Structure 6.9 Validation 6.9.1 Initial Choice of Modes 6.9.2 Relative Convergence and Choice of Modes 6.9.3 Filter Structures 6.9.4 Array Antenna Structures 6.10 Conclusions
204 211 211 215 216 217
221 222 222 224 227 228 229 231 232 234 234 236 237 239 243 245 246 248 249 251 252 254 256 256 258 262 265 272
ix
CONTENTS
Appendix 6.A. Waveguide Mode Orthogonality and Normalization Functions Appendix 6.B. ModeCoupling Integrals for WaveguidetoWaveguide Junction Appendix 6.C. Unit Cell Mode Orthogonality and Normalization Functions Appendix 6.D. ModeCoupling Integrals for RectangularWaveguidetoUnitCell Junction References 7
Summary and Conclusions 7.1 FullWave and Approximate Antenna Analysis 7.2 Intravascular MR Antennas: Loops and Solenoids 7.3 PCB Antennas: Printed Monopoles 7.4 RFID Antennas: Folded Dipoles 7.5 Rectennas: Microstrip Patch Antennas 7.6 Large Array Antennas: OpenEnded RectangularWaveguide Radiators References
Index
273 277 281 282 288 293 293 295 297 297 298 299 299 301
Preface
In der Beschränkung zeigt sich erst der Meister,1 wrote Johann Wolfgang von Goethe on 26 June 1802. It is a quote much used in PhD theses to accentuate and justify the compactness of a thesis. For this book, which also serves the purpose of a PhD thesis, this quote is completely unjustified. I have tried to be as elaborate as possible in explaining the approximate antenna models developed. This book is the result of more than 15 years of work in the field of antenna modeling. After working for a number of years on the fullwave modeling of large phased array antennas, I found that, for a customer, it is very hard to wait till a fullwave computer code has been developed. Therefore I started developing socalled ‘engineering’ or approximate models in parallel with the fullwave models. These engineering models, which can be produced much faster, but at the cost of reduced accuracy, can give the customer a preview of what will be possible, and may be used to create ‘predesigns’ to be finetuned by applying the fullwave model. Nowadays I focus completely on developing approximate models. Most of the topics encountered in this book were developed over the last few years, but some date back almost 15 years. The reason for being ‘as elaborate as possible’ in explaining the approximate models is twofold. First, as a young engineer fresh from university, I found it hard, when starting on a new assignment, to work backwards from a relevant paper and understand all the steps taken in the development of a model. In those days, I would have wanted a book that would have taken me by the hand and explained to me all the necessary steps taken in the development of
1 ‘Constraint is where you show you are a master’.
xii
PREFACE
a model. With this book, I have tried to accommodate this wish. Second, I have always been in the privileged situation of having literature search facilities and a large technical library at my immediate disposal. For those not in this privileged situation, it may be very hard to get access to the necessary references. Therefore, rather than just referring to the sources, I have also written down all of the equations needed for implementing the model into software. This may have the effect that the book will become a bit dreary for experienced antenna engineers. For the inexperienced antenna engineer, I hope that, referring again to Goethe, the following quote will be appropriate after reading the book: Das also war des Pudels Kern2 [1].
REFERENCE 1. J.W. von Goethe, Faust: Der Tragödie erster und zweiter Teil. Urfaust, Beck Verlag, Munich, Germany, 2006.
Hubregt J. Visser Veldhoven, The Netherlands
2 ‘So this, then, was the kernel of the brute’.
Acknowledgments
This book could not have been written without the help of many individuals whom I would like to thank for their contributions. Chapter 2 is the result of a cooperation between the Electromagnetics Department of the Faculty of Electrical Engineering of Eindhoven University of Technology (TU/e) and the Image Science Institute of the University Medical Center Utrecht (UMC Utrecht), both in The Netherlands. From UMC Utrecht, I would like especially to thank Chris Bakker, JanHenry Seppenwoolde and Wilbert Bartels. I would also like to thank my MSc students Nicole Op den Kamp and Marjan Aben for contributing to that chapter. I would like to thank my MSc students Iwan Akkermans and Jeroen Theeuwes for their contributions to Chapter 5. Frank van den Boogaard, from TNO Defence and Safety, is thanked for his kindness in permitting me to use material on waveguide array antenna modeling for Chapter 6. K.K. Chan from Chan Technologies, Inc., Canada, is thanked for his many helpful suggestions and support in developing the model. A word of special thanks is reserved for Anton Tijhuis from TU/e for being my promoter and pushing me forward to finish this work. Also, a word of special thanks is reserved for Guy Vandenbosch from the Catholic University of Leuven, Belgium, for also being my promoter and for keeping faith in me for more than ten years. Ad Reniers is thanked for preparing the many antenna prototypes and performing part of the measurements. Sarah Hinton, Sarah Tilley and Tiina Ruonamaa from Wiley are thanked for their incredible patience and support. Finally, I would like to thank my wife Dianne and daughter Noa for accepting, again, a long period of bookrelated neglect. H.J.V.
Acronyms
AC
Alternating Current
BBC
British Broadcasting Corporation
CAT
Computed Axial Tomography
COTS
Commercial OfftheShelf
CPS
Coplanar Strip
CPW
Coplanar Waveguide
CT
Computed Tomography
DC
Direct Current
FE
Finite Element
FFT
Fast Fourier Transform
FID
Free Induction Decay
FIT
Finite Integration Technique
FR
Flame Retardant
GA
Genetic Algorithm
GPS
Global Positioning System
GSM
Global System for Mobile Communications; Generalized Scattering Matrix
iMRI
Interventional Magnetic Resonance Imaging
xvi
ACRONYMS
MEN
Multimode Equivalent Network
MIT
Massachusetts Institute of Technology
MoM
Method of Moments
MRI
Magnetic Resonance Imaging
NEC
Numerical Electromagnetic Code
NMI
Nuclear Medicine Imaging
NMRI
Nuclear Magnetic Resonance Imaging
OFDM
Orthogonal Frequency Division Multiplexing
PCB
Printed Circuit Board
PEC
Perfect Electric Conductor
PET
Positron Emission Tomography
RC
Relative Convergence
RF
Radio Frequency
RFID
Radio Frequency Identification
RK
Runge–Kutta Method
SAR
Specific Absorption Rate
SMA
Subminiature Version A
SNR
SignaltoNoise Ratio
SPECT
SinglePhotonEmission Computed Tomography
TE
Transverse Electric
TEM
Transverse Electromagnetic
TL
Transmission Line
TLM
Transmission Line Matrix
TM
Transverse Magnetic
UWB
Ultrawideband
WAIM
WideAngle Impedance Match
WLAN
Wireless Local Area Network
1 Introduction
From the moment that Heinrich Rudolf Hertz experimentally proved the correctness of the Maxwell equations in 1886, antennas have been in use. The fact that Guglielmo Marconi’s success depended on the ‘finding’ of the right antenna in 1895 indicates the importance of antennas and thus of antenna analysis. It was, however, common practice up until the middle of the 1920s to design antennas empirically and produce a theoretical explanation after the successful development of a working antenna. It took a world war to evolve antenna analysis and design into a distinct technical discipline. The end of the war was also the starting point of the development of electronic computers that eventually resulted in the commercial distribution of numerical electromagnetic analysis programs. Notwithstanding the progress in numerical electromagnetic analysis, a need still exists for approximate antenna models. They are needed both in their own right and as part of a synthesis process that also involves fullwave models.
1.1
THE HISTORY OF ANTENNAS AND ANTENNA ANALYSIS
The history of antennas dates back almost entirely to the understanding of electromagnetism and the formulation of the electromagneticfield equations. In the 1860s, James Clerk Maxwell saw the connection between Ampère’s, Faraday’s and Gauss’s laws. By extending Ampère’s law with what he called a displacement current term, he united electricity and magnetism into electromagnetism [1]. His monumental work of 1873, A Treatise on Electricity and Magnetism, is still in print [2]. With light now described as and proven to be an electromagnetic phenomenon, Maxwell had already predicted the existence of electromagnetic waves at radio frequencies, i.e. at much lower frequencies than light.
Approximate Antenna Analysis for CAD
Hubregt J. Visser
© 2009 John Wiley & Sons, Ltd. ISBN: 9780470512937
2
INTRODUCTION
Switch
Adjustable capacitor sphere Interrupter
Battery Spark gap Spark gap
Pri
Sec
Oneturn coil
Core Induction coil Transmitter
Receiver
Figure 1.1 Hertz’s open resonance system. With the receiving oneturn loop, small sparks could be observed when the transmitter discharged. From [4].
It was not until 1886 that he was proven right by Heinrich Rudolf Hertz, who constructed an open resonance system as shown in Figure 1.1 [3, 4]. A spark gap was connected to the secondary windings of an induction coil. A pair of straight wires was connected to this spark gap. These straight wires were equipped with electrically conducting spheres that could slide over the wire segments. By moving the spheres, the capacitance of the circuit could be adjusted for resonance. When the breakdown voltage of air was reached and a spark created over the small airfilled spark gap, the current oscillated at the resonance frequency in the circuit and emitted radio waves at that frequency (Hertz used frequencies of around 50 MHz). A singleturn square or circular loop with a small gap was used as a receiver. Without being fully aware of it, Hertz had created the first radio system, consisting of a transmitter and a receiver. Guglielmo Marconi grasped the potential of Hertz’s equipment and started experimenting with wireless telegraphy. His first experiments – covering the length of the attic of his father’s house – were conducted at a frequency of 1.2 GHz, for which he used, like Hertz before him, cylindrical parabolic reflectors, fed at the focal point by halfwave dipole antennas. In 1895, however, he made an important change to his system that suddenly allowed him to transmit and receive over distances that progressively increased up to and beyond 1.5 km [5–7]. In his own words, at the reception for the Nobel Prize for physics in Stockholm in 1909 [7]: In August 1895 I hit upon a new arrangement which not only greatly increased the distance over which I could communicate but also seemed to make the transmission independent from the effects of intervening obstacles. This arrangement [Figure 1.2(a)] consisted in connecting one terminal of the Hertzian oscillator or spark producer to earth and the other terminal to a wire or capacity area placed at a height above the ground and in also connecting at the receiver end [Figure 1.2(b)] one terminal of the coherer to earth and the other to an elevated conductor.
3
THE HISTORY OF ANTENNAS AND ANTENNA ANALYSIS
(a)
(b)
Figure 1.2 Marconi’s antennas of 1895. (a) Scheme of the transmitter used by Marconi at Villa Griﬀone. (b) Scheme of the receiver used by Marconi at Villa Griﬀone. From [4]. Reproduced, with permission, from Oﬁr Glazer, BioMedical Engineering Department, TelAviv University, Israel. Part of M.Sc. ﬁnal project, tutored by Dr. Hayit Greenspan.
Marconi had enlarged the antenna. His monopole antenna was resonant at a wavelength much larger than any that had been studied before, and it was this creation of longwavelength electromagnetic waves that turned out to be the key to his success. It was also Marconi who, in 1909, introduced the term antenna for the device that was formerly referred to as an aerial or elevated wire [7, 8]. The concept of a monopole antenna, forming a dipole antenna together with its image in the ground, was not known by Marconi at the time of his invention. In 1899, the relation between the antenna length and the operational wavelength of the radio system was explained to him by Professor Ascoli, who had calculated that the ‘length of the wave radiated [was] four times the length of the vertical conductor’ [9]. Up to the middle of the 1920s it was common practice to design antennas empirically and produce a theoretical explanation after the successful development of a working antenna [10]. It was in 1906 that Ambrose Fleming, a professor at University College, London, and consultant to the Marconi Wireless Telegraphy Company, produced a mathematical explanation of a monopolelike antenna1 based on image theory. This may be considered the first ever antenna design that was accomplished both experimentally and theoretically [10]. The first theoretical description of an antenna may be attributed to H.C. Pocklington, who, in 1897, first formulated the frequency domain integral equation for the total current flowing along a straight, thin wire antenna [11].
1 This antenna was a suspended long wire antenna, nowadays also called an inverted L antenna or ILA, and used for transatlantic transmissions.
4
INTRODUCTION
The invention of the thermionic valve, or diode, by Fleming in 1905 and of the audion, or triode, by Lee de Forest in 1907 paved the way for the reliable detection, reception and amplification of radio signals. From 1910 onwards, broadcasting experiments were conducted that resulted, in Europe, in the formation in 1922 of the British Broadcasting Corporation (BBC) [12]. The early antennas in the broadcasting business were makeshift antennas, derived from the designs used in pointtopoint communication. Later, Tconfigured antennas were used for transmitters [13], and eventually vertical radiators became standard, owing to their circularly symmetrical coverage (directivity) characteristic [13, 14]. The receiver antennas used by the public were backyard Lstructures and Tstructures [4]. In the 1930s, a return of interest in the higher end of the radio spectrum took place. This interest intensified with the outbreak of World War II. The need for compact communication equipment as well as compact (airborne) and highresolution radar made it absolutely necessary to have access to compact, reliable, highpower, highfrequency sources. In early 1940, John Randall and Henry Boot were able to demonstrate the first cavity magnetron, creating 500 kW at 3 GHz and 100 kW at 10 GHz. In that same year, the British Prime Minister, Sir Winston Churchill, sent a technical mission to the United States of America to exchange wartime secrets for production capacity. As a result of this Tizard Mission, named after its leader Sir Henry Tizard, the cavity magnetron was brought to the USA and the MIT Rad Lab (Massachusetts Institute of Technology Radiation Laboratory) was established. At the Rad Lab, scientists were brought together to work on microwave electronics, radar and radio, to aid in the war effort. The Rad Lab closed on 31 December 1945, but many of the staff members remained for another six months or more to work on the publication of the results of five years of microwave research and development. This resulted in the famous 28 volumes of the Rad Lab series, many of which are still in print today [15–42]. In relation to antenna analysis, we have to mention the volume Microwave Antenna Theory and Design by Samuel Silver [26], which may be regarded as one of the first ‘classic’ antenna theory textbooks. Soon, it was followed by several other, now ‘classic’ antenna theory textbooks, amongst others Antennas by John Kraus in 1950 [43], Antennas, Theory and Practice by S.A. Schelkunoff in 1952 [44], Theory of Linear Antennas by Ronold W.P. King in 1956 [45], Antenna Theory and Design by Robert S. Elliott in 1981 [46] and Antenna Theory, Analysis and Design by Constantine A. Balanis in 1982 [47]. Specifically for phased array antennas, we have to mention Microwave Scanning Antennas by Robert C. Hansen [48] (1964), Theory and Analysis of Phased Array Antennas by N. Amitay, V. Galindo and C.P. Wu [49] (1972), and Phased Array Antenna Handbook by Robert J. Mailloux [50] (1980).2 At the end of World War II, antenna theory was mature to a level that made the analysis possible of, amongst others, freestanding dipole, horn and reflector antennas, monopole antennas, slots in waveguides and arrays thereof. The end of the war was also the beginning of the development of electronic computers. Roger Harrington saw the potential of electronic computers in electromagnetics [51] and in the 1960s introduced the method of moments (MoM) in electromagnetism [52]. The origin of the MoM dates back to the work of
2 For the ‘classic’ antenna theory textbooks mentioned here, we refer to the first editions. Many of these books have
by now been reprinted in second or even third editions.
ANTENNA SYNTHESIS
5
Galerkin in 1915 [53]. The introduction of the IBM PC3 in 1981 helped considerably in the development of numerical electromagnetic analysis software. The 1980s may be seen as the decade of the development of numerical microwave circuit and planar antenna theory. In this period, the Numerical Electromagnetics Code (NEC) for the analysis of wire antennas was commercially distributed. The 1990s, however, may be seen as the decade of numerical electromagneticbased design of microwave circuits and (planar, integrated) antennas. In 1989 the distribution of Sonnet started, followed, in 1990, by the HP (now Agilent) High Frequency Structure Simulator (HFSS)4 [51]. These two numerical electromagnetic analysis tools were followed by Zeland’s IE3D, Remcom’s XFdtd, Agilent’s Momentum, CST’s Microwave Studio, FEKO from EM Software & Systems, and others. Today, we have evolved from the situation in the early 1990s when the general opinion appeared to be ‘that numerical electromagnetic analysis cannot be trusted’ to a state wherein numerical electromagnetic analysis is considered to be the ultimate truth [51]. The last assumption, however, is as untrue as the first one. Although numerical electromagnetic analysis software has come a long way, incompetent use can easily throw us back a hundred years in history. One only has to browse through some recent volumes of peerreviewed antenna periodicals to encounter numerous examples of bizarrelooking antenna structures designed by iterative use of commercially offtheshelf (COTS) numerical electromagnetic analysis software. These reported examples of the modern variant of trial and error, although meeting the design specifications, are often presented without even a hint of a tolerance analysis, let alone a physical explanation of the operation of the antenna. The advice that James Rautio, founder of Sonnet Software, gave in the beginning of 2003 [51], No single EM tool can solve all problems; an informed designer must select the appropriate tool for the appropriate problem,
is still valid today, as a benchmarking of COTS analysis programs showed at the end of 2007 [54, 55]. Apart from the advice to choose the right analysis technique for the right structure to be analyzed, these recent studies also indicate the importance of being careful in the choice of the feeding model and the mesh for the design to be analyzed. So, notwithstanding the evolution of numerical electromagnetic analysis software, it still takes an experienced antenna engineer, preferably one having a PhD in electromagnetism or RF technology, to operate the software in a justifiable manner and to interpret the outcomes of the analyses. Having said this, we may now proceed with a discussion of how to use fullwave analysis software for antenna synthesis.
1.2
ANTENNA SYNTHESIS
Antenna synthesis should make use of a manual or automated iterative use of analysis steps. The analysis techniques occupy a broad time consumption ‘spectrum’ from quick physical
3 4.77 MHz, 16 kB RAM, no hard drive. 4 Currently Ansoft HFSS.
6
INTRODUCTION
Figure 1.3 Analysis techniques ordered according to calculation time involved.
Figure 1.4 Stochastic optimization based on iteration of fullwave analysis is a (too) timeconsuming process.
reasoning (‘the length of a monopolelike antenna should be about a quarter of the operational wavelength’) to lengthy (in general) fullwave numerical electromagnetic analysis. The ‘spectrum’ of analysis techniques is shown in Figure 1.3, where the hourglasses indicate symbolically the time involved in applying the various analysis techniques. For an automated synthesis, starting with mechanical and electromagnetical constraints and possibly an initial guess,5 we have to rely on stochastic optimization. Since stochastic optimization needs a (very) large number of function evaluations or analysis steps, such an optimization scheme based on fullwave analysis (Figure 1.4) is not a good idea. Therefore, we propose a twostage approach [56], where, first, a stochastic optimization is used in combination with an approximate analysis and, second, line search techniques are combined with fullwave modeling (Figure 1.5). Since one of the key features of the approximate analysis model needs to be that its implementation in software is fast while still sufficiently accurate, we may employ many approximate analysis iterations and therefore use a stochastic optimization to get a predesign. This predesign may then be finetuned using a limited number of iterations using linesearch techniques. Owing to the limited number of iterations, we may now – in the final synthesis stage – employ a fullwave analysis model. Using an approximate but still sufficiently accurate model, the automated design – using stochastic optimization – may be sped up considerably. The output at this stage of the synthesis process is a preliminary design. Depending on the accuracy of this design and
5 An initial guess may be created by randomly choosing the design variables.
APPROXIMATE ANTENNA MODELING
7
Figure 1.5 Antenna synthesis based on stochastic optimization in combination with an approximate model and line search with a fullwave model.
the design constraints, it is very well possible that the design process could end here; see for example [56]. If a higher accuracy is required or if the design requirements are not fully reached, this preliminary design could be used as an input for a line search optimization in combination with a fullwave model. For the complete synthesis process using both approximate and fullwave models (Figure 1.5), the time consumption will drop with respect to a synthesis process involving only a fullwave model. The reason is that the most timeconsuming part of the process, i.e. when the solution space is randomly sampled, is now conducted with a fast, approximate, reducedaccuracy model. The question that remains is what may be considered to be ‘sufficiently accurate’.
1.3
APPROXIMATE ANTENNA MODELING
From the point of view of synthesis, approximate antenna models are a necessity. They need to be combined with a fullwave analysis program, but if – depending on the application – the accuracy of the approximation is sufficient, the approximate model alone will suffice. In [51, 54], the use of (at least) two fullwave simulators is advised, but not many companies or universities can afford to purchase or lease multiple fullwave analysis programs. For many companies that do not specialize in antenna design, even the purchase or lease of one fullwave analysis program may be a budgetary burden. Therefore the availability of approximate, sufficiently accurate antenna models is required not only for the full synthesis process. It is
8
INTRODUCTION
also valuable for anyone needing an antenna not yet covered in the standard antenna textbooks who does not have access to a fullwave analysis program. The purpose of the approximate and fullwave models is to replace the realization and characterization of prototypes, thus speeding up the design process. This does not mean, however, that prototypes should not be realized at all. At least one prototype should be realized to verify the (pre)design. A range of slightly different prototypes could be produced as a replacement for the finetuning that employs line search techniques in combination with fullwave modeling. A question that still remains with respect to the approximate modeling is what may be considered ‘sufficiently accurate’. This question cannot be answered unambiguously. It depends on the application; the requirements for civil and medical communication antennas, for example, are much less stringent than those for military radar antennas. If we look at a communication antenna to be matched to a standard 50 transmission line, we should not look at the antenna input impedance but rather at the reflection level. In general, any reflection level below −10 dB over the frequency range of interest is considered to be satisfactory. This means that, if we assume the input impedance to be realvalued, we may tolerate a relative error in the input impedance of up to 100%. For lowpower, integrated solutions, working with a 50 standard for interconnects may not be the best solution. A conjugate matching may be more efficient. If we are looking at antennas to be conjugately matched to a complex transmitter or receiver frontend impedance, however, we cannot tolerate the aforementioned large impedance errors. In general, we may say that we consider an approximate antenna model sufficiently accurate if it predicts a parameter of interest to within a few percent relative to the measured value or the (verified) fullwave analysis result. Such an accuracy also prevents the answer drifting away during the stochastic optimization. Another question is when to develop an approximate model. The answer to this question is dictated both by the resources available and a company’s longterm strategy. If neither a fullwave analysis program for the problem at hand nor an existing approximate model is available, then one can resort to trial and error or develop an approximate model or a combination of both, where the outputs of experiments dictate the path of the development of the model. If a fullwave analysis program is available and the antenna to be designed is a oneofakind antenna or time is really critical, one can resort to an educated software variant of design by trial and error, meaning that the task should be performed by an antenna expert. When the antenna to be designed can be considered to belong to a class of antennas, meaning that similar designs are foreseen for the future, but for different materials and other frequency bands or for use in other environments, it is beneficial to develop a dedicated approximate model. The additional effort put into the development of the model for the first design will be compensated for in the subsequent antenna designs. An antenna design may also be created by generating a database of substructure analyses, employing a fullwave analysis model. Then, a smart combining of these preanalyzed substructures results in the desired design. The generation of the database will be very timeconsuming but once this task has been accomplished, the remainder of the design process will be very timeefficient. The last question is how to develop an approximate model. First of all, the approximate model should be tailored to the antenna class at hand. To achieve that, the antenna structure should be broken down into components for which analytical equations have been derived in the past, in the precomputer era, or for which analytical equations may be derived.
ORGANIZATION OF THE BOOK
9
By distinguishing between main and secondary effects, approximations may be applied with different degrees of accuracy, thus speeding up computation time. It appears that much of the work performed in the 1950s, 1960s and 1970s that seems to have been forgotten is extremely useful for this task. In this book, we have followed this approach for a few classes of antennas. For each class of antennas, we have taken a generic antenna structure and decomposed it into substructures, such as sections of transmission line, dipoles and equivalent electrical circuits. For these substructures and for the combined substructures, approximate analysis methods have been selected or developed. The main constraints in developing approximate antenna models were the desired accuracy in the antenna parameter to be evaluated (the amplitude of the input reflection coefficient or the value of the complex input impedance) and the computation time for the software implementation of the model. Examples of the development of approximate models will be given in the following chapters.
1.4
ORGANIZATION OF THE BOOK
In Chapter 2, we start with the development of an approximate model for intravascular antennas, i.e. loops and solenoids embedded in blood (Figure 1.6). A reason for undertaking this development was the unavailability of a fullwave analysis program fit for the task at the time of development. But even if such a program had been available, it would have taken too much time to be of practical value in designing intravascular antennas. The antennas were meant as receiving antennas in a magnetic resonance imaging (MRI) system, either for visualizing catheter tips during interventional MRI or for obtaining detailed information about the inside of the artery wall. The figure shows that the quasistatic model developed here may be used in a stochastic optimization process. The optimization times were of the order of minutes. In Chapter 4, we describe an example of the use of a fullwave analysis program for designing a printed ultrawideband (UWB) monopole antenna, the reason being that this antenna was a ‘oneofakind’ design. We begin with physical reasoning about how the proposed antenna operates. In the design process, it becomes clear that it may be beneficial to use or develop approximate models for parts of the structure, such as filtering structures in the feeding line. Next, an approximate model is developed for a nonUWB printed monopole antenna (Figure 1.7) that is considered to belong to a class of antennas. The model is based on an equivalentradius dipole antenna with a magnetic covering. Then, in Chapter 5, we discuss foldeddipole antennas and some means to control the input impedance of these antennas. The envisaged application is in the field of radio frequency identification (RFID), where the antenna needs to be conjugately matched to the RFID chip impedance, which will, in general, be some complex value different from 50 . An approximate model based on dipole antenna analysis and transmission line analysis is applied to both thinwire foldeddipole structures and foldeddipole structures consisting of strips on a dielectric slab. Also, arrays of reentrant folded dipoles will be analyzed, as shown in Figure 1.8. Pursuing the modeling of ‘non50 ’ antennas, in Chapter 6 we discuss an efficient, approximate but accurate modeling of a rectenna, i.e. an antenna connected to a rectifying element (diode), meant for collecting RF energy and transforming it to usable DC energy.
10
INTRODUCTION
Figure 1.6 Intravascular antenna, and optimization results. Left: antenna. Right: magnetic ﬁeld intensity calculated after optimization for local antenna ‘visibility’ (left), and calculated after optimization for maximum magnetic ﬁeld intensity at the position of the artery wall (right) for diﬀerent planes through the antenna.
Figure 1.7 Printed monopole antenna and results of analysis by an approximate model. Left: antenna conﬁguration. Right: calculated and measured return loss as a function of frequency for a particular conﬁguration.
We start by modeling the rectifying circuit with the aid of a largesignal equivalent model. Once the input impedance of this circuit has been determined, we use a modified cavity model for a rectangular microstrip patch antenna to find the complex conjugate impedance value. Thus we may directly match the antenna and the rectifying circuit. To complete the chapter, we discuss a means of using antennas for power and data exchange simultaneously, based on the concept of the Wilkinson power combiner (Figure 1.9).
ORGANIZATION OF THE BOOK
11
Figure 1.8 Linear array of reentrant folded dipoles. Left: array conﬁguration. Right: real and imaginary parts of the array input impedance as a function of frequency, calculated with the approximate model and with the method of moments.
Chapter 7 deals with ‘approximation’ in a different way. In this chapter, we use an approximation for large, planar array antennas. The approximation consists of considering the array antenna to be infinite in two directions in the transverse plane. This approximation allows us, for an array of identical radiating elements positioned in a regular lattice, to consider the array to be periodic and uniformly excited, and therefore we only have to analyze a single unit cell (Figure 1.10) that contains all of the information about the mutual couplings with the (infinite) environment. The approximation is applied to an array consisting of openended waveguide radiators with or without obstructions in the waveguides and with or without dielectric sheets in front of the waveguide apertures. The infinitearray approximation works best for very large array antennas where the majority of the elements experience an environment identical to that of an element in an infinite array. In practice, even arrays consisting of a few tens of elements may be approximated in this way. Although the material in this chapter dates back to the mid 1990s and a lot of work on this type of array antennas has been performed since [57–60], we find it appropriate to present a ‘classic’ modematching approach. The material here may aid in understanding new developments and may be relatively easy implemented in software for analyzing rectangular waveguide structures and infinite arrays of openended waveguides. Since the different chapters may be read independently, we have opted for a form where conclusions and references are given per chapter. Throughout the book, we indicate vectors by boldface characters, for example, A and b. Unit vectors are further denoted by hats, for example, uˆ x , uˆ y and uˆ z . The dB scale is defined as 1010 log x, where x is a normalized power. The definition 2010 log x is used when x is a normalized amplitude (electric field, voltage, magnetic field, current, etc.); 2010 log x = 1010 log x 2 . The natural numbers N
12
INTRODUCTION
Figure 1.9 Rectennas. Top left: rectenna feeding an LED, wirelessly powered by a GSM phone. Top right: antenna and powercombining network for simultaneously receiving power and data. Bottom left: even–odd mode analysis for power combiner with rectifying element. Bottom right: calculated and measured opensource voltage as a function of frequency across the rectifying element in the power combiner shown in the top right of the ﬁgure.
are the set {1, 2, 3, . . .} or {0, 1, 2, 3, . . .}. The inclusion of zero is a matter of definition [61]. Here we define N to include zero. Finally, a superscript number placed after a word indicates a footnote, for example, ‘example1’.
1.5
SUMMARY
Notwithstanding the progress in numerical electromagnetic analysis, the automated design of integrated antennas based on fullwave analysis is not yet feasible. In a twostage approach, where stochastic optimization techniques are used in combination with approximate models to generate predesigns and these predesigns are used as input for line search optimization in combination with fullwave modeling, automated antenna design is feasible. Therefore, a need exists for approximate antenna models for different classes of antennas.
13
REFERENCES
y
x t
Ω
z
s
Figure 1.10 Planar, inﬁnite, openended waveguide array antenna with the radiators arranged into a triangular grating, plus an indication of a single unit cell.
For oneofakind antenna designs, the iterative, manual use of a fullwave analysis program is advised. So, today, not only are fullwave models needed but also there still exists a need for approximate models. That both fullwave and approximate models are needed cannot be said more eloquently than Ronold W.P. King did in 2004 [62]: At this age of powerful computers, there are those who believe that numerical methods have made analytical formulas obsolete. Actually, the two approaches are not mutually exclusive but rather complementary. Numerical methods can provide accurate results within the resolution determined by the size of the subdivisions. Analytical formulas provide unrestricted resolution. Numerical results are a set of numbers for a specific set of parameters and variables. Analytical formulas constitute general relations that exhibit functional relationships among all relevant parameters and variables. They provide the broad insight into the relevant physical phenomena that is the basis of new knowledge. They permit correct frequency and dimensional scaling. Computer technology and mathematical physics are a powerful team in the creation of new knowledge.
REFERENCES 1. J.C. Maxwell, ‘A dynamical theory of the electromagnetic field’, Royal Society Transactions, Vol. 155, p. 459, 1865. 2. J.C. Maxwell, A Treatise on Electricity and Magnetism, Dover Publications, New York, 1954.
14
INTRODUCTION
3. R.S. Elliot, Electromagnetics: History, Theory and Applications, John Wiley & Sons, New York, 1999. 4. H.J. Visser, Array and Phased Array Antenna Basics, John Wiley & Sons, Chichester, UK, 2005. 5. G. Masini, Marconi, Marsilio, New York, 1995. 6. G.C. Corazza, ‘Marconi’s history’, Proceedings of the IEEE, Vol. 86, No. 7, pp. 1307– 1311, July 1998. 7. G. Marconi, ‘Wireless telegraphic communications’, Nobel Lectures in Physics, 1901– 21, Elsevier, 1967. 8. G. Pelosi, S. Selleri and B.A. Valotti, ‘Antennae’, IEEE Antennas and Propagation Magazine, Vol. 42, No. 1, pp. 61–63, February 2000. 9. B.A. Austin, ‘Wireless in the Boer War’, 100 Years of Radio, 5–7 September 1995, Conference Publication 411, IEE, pp. 44–50, 1995. 10. A.D. Olver, ‘Trends in antenna design over 100 years’, 100 Years of Radio, 5–7 September 1995, Conference Publication 411, IEE, pp. 83–88, 1995. 11. H.C. Pocklington, ‘Electrical oscillations in wires’, Proceedings of the Cambridge Philosophical Society, pp. 324–332, 25 October 1897. 12. J. Hamilton (ed.), They Made Our World; Five Centuries of Great Scientists and Inventors, Broadside Books, London, pp. 125–132, 1990. 13. W.F. Crosswell, ‘Some aspects of the genesis of radio engineering’, IEEE Antennas and Propagation Magazine, Vol. 35, No. 6, pp. 29–33, December 1993. 14. J. Ramsay, ‘Highlights of antenna history’, IEEE Communications Magazine, pp. 4–16, September 1981. 15. L.N. Ridenour, Radar System Engineering, Vol. 1 of MIT Radiation Laboratory Series, McGrawHill, New York, 1947. 16. J.S. Hall, Radar Aids to Navigation, Vol. 2 of MIT Radiation Laboratory Series, McGrawHill, New York, 1947. 17. A.R. Roberts, Radar Beacons, Vol. 3 of MIT Radiation Laboratory Series, McGrawHill, New York, 1947. 18. J.A. Pierce, A.A. McKenzie and R.H. Woodward, Loran, Vol. 4 of MIT Radiation Laboratory Series, McGrawHill, New York, 1948. 19. G.N. Glasoe and J.V. Lebacqz, Pulse Generators, Vol. 5 of MIT Radiation Laboratory Series, McGrawHill, New York, 1948.
REFERENCES
15
20. G.B. Collins, Microwave Magnetrons, Vol. 6 of MIT Radiation Laboratory Series, McGrawHill, New York, 1948. 21. D.R. Hamilton, J.K. Knipp and J.B. Horner Kuper, Klystrons and Microwave Triodes, Vol. 7 of MIT Radiation Laboratory Series, McGrawHill, New York, 1948. 22. C.G. Montgomery, R.H. Dicke and E.M. Purcell, Principles of Microwave Circuits, Vol. 8 of MIT Radiation Laboratory Series, McGrawHill, New York, 1948. 23. G.L. Ragan, Microwave Transmission Circuits, Vol. 9 of MIT Radiation Laboratory Series, McGrawHill, New York, 1948. 24. N. Marcuvitz, Waveguide Handbook, Vol. 10 of MIT Radiation Laboratory Series, McGrawHill, New York, 1951. 25. C.G. Montgomery, Technique of Microwave Measurements, Vol. 11 of MIT Radiation Laboratory Series, McGrawHill, New York, 1947. 26. S. Silver, Microwave Antenna Theory and Design, Vol. 12 of MIT Radiation Laboratory Series, McGrawHill, New York, 1949. 27. D.E. Kerr, Propagation of Short Radio Waves, Vol. 13 of MIT Radiation Laboratory Series, McGrawHill, New York, 1951. 28. L.D. Smullin and C.G. Montgomery, Microwave Duplexers, Vol. 14 of MIT Radiation Laboratory Series, McGrawHill, New York, 1948. 29. H.C. Torrey and C.A. Whitmer, Crystal Rectifiers, Vol. 15 of MIT Radiation Laboratory Series, McGrawHill, New York, 1948. 30. R.V. Pound, Microwave Mixers, Vol. 16 of MIT Radiation Laboratory Series, McGrawHill, New York, 1948. 31. J.F. Blackburn, Components Handbook, Vol. 17 of MIT Radiation Laboratory Series, McGrawHill, New York, 1949. 32. G.E. Valley Jr. and H. Wallman, Vacuum Tube Amplifiers, Vol. 18 of MIT Radiation Laboratory Series, McGrawHill, New York, 1948. 33. B. Chance, V. Hughes, E.F. MacNichol Jr., D. Sayre and F.C. Williams, Waveforms, Vol. 19 of MIT Radiation Laboratory Series, McGrawHill, New York, 1949. 34. B. Chance, R.I. Hulsizer, E.F. MacNichol, Jr. and F.C. Williams, Electronic Time Measurements, Vol. 20 of MIT Radiation Laboratory Series, McGrawHill, New York, 1949. 35. I.A. Greenwood Jr., J.V. Holdam Jr. and D. MacRae Jr., Electronic Instruments, Vol. 21 of MIT Radiation Laboratory Series, McGrawHill, New York, 1948. 36. T. Soller, M.A. Star and G.E. Valley Jr., Cathode Ray Tube Displays, Vol. 22 of MIT Radiation Laboratory Series, McGrawHill, New York, 1948.
16
INTRODUCTION
37. S.N. Van Voorhis, Microwave Receivers, Vol. 23 of MIT Radiation Laboratory Series, McGrawHill, New York, 1948. 38. J.L. Lawson and G.E. Uhlenbeck, Threshold Signals, Vol. 24 of MIT Radiation Laboratory Series, McGrawHill, New York, 1950. 39. H.M. James, N.B. Nichols and R.S. Phillips, Theory of Servomechanisms, Vol. 25 of MIT Radiation Laboratory Series, McGrawHill, New York, 1947. 40. W.M. Cady, M.B. Karelitz and L.A. Turner, Radar Scanners and Radomes, Vol. 26 of MIT Radiation Laboratory Series, McGrawHill, New York, 1948. 41. A. Svoboda, Computing Mechanisms and Linkages, Vol. 27 of MIT Radiation Laboratory Series, McGrawHill, New York, 1948. 42. K. Henney (ed.), Index, Vol. 28 of MIT Radiation Laboratory Series, McGrawHill, New York, 1953. 43. J. Kraus, Antennas, McGrawHill, New York, 1950. 44. S.A. Schelkunoff, Antennas, Theory and Practice, John Wiley & Sons, London, 1952. 45. R.W.P. King, Theory of Linear Antennas, Harvard University Press, Cambridge, MA, 1956. 46. R.S. Elliott, Antenna Theory and Design, Prentice Hall, Englewood Cliffs, 1981. 47. C.A. Balanis, Antenna Theory, Analysis and Design, John Wiley & Sons, New York, 1982. 48. R.C. Hansen, Microwave Scanning Antennas, Academic Press, New York, Vols. 1 and 2, 1964, Vol. 3, 1966. 49. N. Amitay, V. Galindo and C.P. Wu, Theory and Analysis of Phased Array Antennas, John Wiley & Sons, New York, 1972. 50. R.J. Mailloux, Phased Array Antenna Handbook, Artech House, 1980. 51. J.C. Rautio, ‘Planar electromagnetic analysis’, IEEE Microwave Magazine, pp. 35–41, March 2003. 52. R.F. Harrington, Field Computation by Moment Methods, Macmillan, New York, 1986. 53. R. Harrington, ‘Origin and developments of the method of moments for field computation’, IEEE Antennas and Propagation Magazine, Vol. 32, No. 3, pp. 31–35, June 1990. 54. A. Vasylchenko, Y. Schols, W. De Raedt and G.A.E. Vandenbosch, ‘A benchmarking of six software packages for fullwave analysis of microstrip antennas’, Proceedings of the 2nd European Conference on Antennas and Propagation, EuCAP2007, November 2007, Edinburgh, UK.
REFERENCES
17
55. A. Vasylchenko, Y. Schols, W. De Raedt and G.A.E. Vandenbosch, ‘Challenges in full wave electromagnetic simulation of very compact planar antennas’, Proceedings of the 2nd European Conference on Antennas and Propagation, EuCAP2007, November 2007, Edinburgh, UK. 56. A.G. Tijhuis, M.C. van Beurden, B.P. de Hon and H.J. Visser, ‘From engineering electromagnetics to electromagnetic engineering: Using computational electromagnetics for synthesis problems’, Elektrik, Turkish Journal of Electrical Engineering and Computer Sciences, Vol. 16, No. 1, pp. 7–19, 2008. 57. H.J. Visser and M. Guglielmi, ‘CAD of waveguide array antennas based on “filter” concepts’, IEEE Transactions on Antennas and Propagation, Vol. 47, No. 3, pp. 542– 548, March 1999. 58. D. Bakers, Finite Array Antennas: An Eigencurrent Approach, PhD thesis, Eindhoven University of Technology, 2004. 59. B. Morsink, Fast Modeling of Electromagnetic Fields for the Design of Phased Array Antennas in Radar Systems, PhD thesis, Eindhoven University of Technology, 2005. 60. S. Monni, Frequency Selective Surfaces Integrated with Phased Array Antennas: Analysis and Design Using Multimode Equivalent Networks, PhD thesis, Eindhoven University of Technology, 2005. 61. T.C. Collocot and A.B. Dobson (eds.), Dictionary of Science and Technology. Revised edition, Chambers, Edinburgh, UK, 1982. 62. R.W.P. King, ‘A review of analytically determined electric fields and currents induced in the human body when exposed to 50–60Hz electromagnetic fields’, IEEE Transactions on Antennas and Propagation, Vol. 52, No. 5, pp. 1186–1191, May 2004.
2 Intravascular MR Antennas: Loops and Solenoids1 The rapid developments in the field of (nuclear) magnetic resonance imaging ((N)MRI), especially the fast growth in temporal efficiency, and the development of ‘open’ MRI systems have contributed significantly to the feasibility of interventional MRI (iMRI). In this context, a need exists for intravascular MR antennas, to be used for either tracking of guide wires and catheters through blood vessels during surgery or for obtaining highresolution images of vessel walls, images that cannot be obtained by conventional MRI operation. Although various intravascular MR antenna concepts have already been investigated, an electromagnetic model – leading to fast calculations when implemented in a computer code – to quantitatively compare such concepts or even synthesize optimum antennas is needed. An approximate model, based on a quasistatic magneticfield computation, is developed here and thoroughly compared with exact solutions to assess its range of validity. With the thus verified approximate model, various antenna concepts for tracking and imaging are quantitatively compared and a selection of the ‘best’ antenna concepts is made. Next, in vitro tests are described, confirming the results obtained theoretically. Finally, we describe optimization using a genetic algorithm based on the approximate model, to synthesize antenna designs.
1 Parts of this chapter are the result of a cooperation between the Electromagnetics Department of the Faculty
of Electrical Engineering of Eindhoven University of Technology (TU/e) and the Image Science Institute of the University Medical Center Utrecht (UMC Utrecht), both in The Netherlands. Within this cooperation, two students from TU/e performed MSc thesis projects on intravascular MR antennas at UMC Utrecht, supervised by representatives of both universities. Approximate Antenna Analysis for CAD
Hubregt J. Visser
© 2009 John Wiley & Sons, Ltd. ISBN: 9780470512937
20 2.1
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
INTRODUCTION
This chapter addresses a recent development in medical imaging: magnetic resonance imaging (MRI). More specifically, it addresses means to expand the applications of MRI by intravascular collection of measurement data. MRI is one of many medical imaging techniques. Medical imaging (MI) is the process by which parts of the body, not normally visible, are examined and diagnosed, preferably by visualizing those parts. The bestknown imaging technique – skipping the obvious ‘tapping, feeling and interpreting’ of a physician – is that of radiology, employing Xrays. The classical Xray image, which can show bone fractures and pathological changes in the lungs, is a shadow image resulting from the attenuation of Xray photons by (parts of) the body. An extension of this technique is found in computed tomography (CT) and computed axial tomography (CAT). In a CT scan or CAT scan, many Xray images of a slice of the body are taken from different angles. These Xray images are then mathematically processed to produce a comprehensive image of the slice. A major disadvantage of these radiology techniques is that the use of ionizing radiation imposes a limit on the image acquisition time, especially for children. Ultrasound is a widely used, soundbased technique. Waves of highfrequency (2– 10 MHz) acoustic energy are radiated into the body. These waves are reflected by tissue to varying degrees, detected by an acoustic transducer and transformed into an image. These images are produced in real time, which is one of the advantages of the technique. Another advantage of ultrasound is that it is safe to use, as ultrasound does not seem to harm the patient. The major drawback is that an ultrasound image shows less detailed information than a CT or CAT scan. The resolution is directly related to the wavelength used. For Xrays, the wavelengths are of the order of 0.01 nm. For ultrasound, the frequencies are of the order of 4.5 Hz, but since the wave velocity is of the order of 1.5 × 103 m s−1 , the wavelengths are of the order of 0.3 mm. In nuclear medicine imaging (NMI), a radioactive source is injected into the patient. This radioactive source functions as a tracer and is ‘designed’ to tag molecules that seek specific sites in the body. A detector is positioned next to or around the patient and the radiation emitted from the body is measured. The technique is very similar to that of a CT or CAT scan, but with the difference that the radiation source is now internal and its distribution is unknown. The two most commonly employed types of NMI are singlephotonemission computed tomography (SPECT), which uses radiotracers that emit photons when decaying, and positron emission tomography (PET), which uses radiotracers that produce positron–electron pairs. The drawbacks of these techniques are the ones mentioned above for techniques employing ionizing radiation. Magnetic resonance imaging (MRI) does not employ ionizing radiation. A patient is positioned in a highintensity static magnetic field. The magnetic field causes the spinpossessing molecules in the body to align their magnetic moments with this field. When a radio frequency (RF) pulse is emitted, causing the main magnetic field to deflect, the molecules absorb energy, which is reradiated after the RF pulse has ceased to exist. This reradiation induces a current in a receiver coil, and this received signal is a measure of the tissue being excited. By applying a positiondependent gradient in the main magnetic field, it is possible to identify the spatial location of reemitted RF energy. As in a CT or CAT scan image, slices of patients are produced, but the image contrast that can be achieved in soft
21
INTRODUCTION
Figure 2.1
MR image of the human brain. Courtesy NASA/JPLCaltech.
matter by MRI is superior [1]. The radiation involved is nonionizing and roughly in the range 30–120 MHz. An example of an MR image is shown in Figure 2.1. The fast growth in the temporal efficiency of MRI systems has contributed significantly to the feasibility of interventional MRI (iMRI). In this context, a need exists for intravascular MR antennas, to be used for either tracking of guide wires and catheters through blood vessels during surgery and even for obtaining highresolution images of vessel walls, images that cannot be obtained by conventional MRI operation. For MRI operation, receiver coils are employed to detect the reradiated RF energy that is absorbed by molecules in the tissue when excited by an external RF pulse. The resolution of an image that can be formed is directly related to the signaltonoise ratio (SNR). When we wish to obtain detailed information about blood vessel walls, the commonly used receiver coils cannot produce the desired SNR. The employment of local receiver coils instead of surface coils to increase the SNR has been successful, for example, in the diagnosis of prostate cancer (endorectal coils, e.g. [2]) and in the detection of tumours of less than 1 cm3 volume (endovagin*l coils, e.g. [3]). A logical next step would be the employment of intravascular coils or antennas for detecting areas of stenosis, dissection, aneurysm or other vascular pathology. It should be noted, however, that the use of these intravascular antennas will only have practical value in combination with endovascular intervention, when an arterial puncture has already been made and the risks involved in endovascular intervention have already been assessed as being acceptable. Before moving on to the topic of intravascular antennas, we shall first give a brief overview of the basics of MRI. Subsequently, we shall present an overview of existing intravascularantenna concepts for tracking and imaging purposes and we shall compare these concepts in a qualitative way. The development of a static electromagnetic model for intravascular MR antennas will be discussed next. To assess the validity of the static model, comparisons will be made with results obtained from a dynamic, smallloop, uniformcurrent
22
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
antenna model. Since our reference is an approximate model, the validity of this approximate model is investigated first. Then, results obtained with the static model are compared with results obtained with the dynamic model for a loop antenna immersed in blood. After the model has been verified for a singleloop antenna, length restrictions on a multiturn loop are derived. After this model has been verified, the antenna concepts will be compared again, but now in a quantitative way. Test results for realized intravascular antennas are the next subject, followed by a discussion of synthesis of intravascular MR antennas. Then, patient safety issues related to the use of intravascular MR antennas are discussed, after which the conclusions of this chapter are presented.
2.2
MRI
MRI, magnetic resonance imaging, formerly known as nuclear magnetic resonance imaging (NMRI),2 uses magnetic properties of tissue to create internal anatomical images of people. To understand the basics of MRI, first the magnetic properties of atom nuclei must be understood [1, 4]. 2.2.1
Magnetic Properties of Atomic Nuclei
An atom may be envisaged as a nucleus consisting of positively charged protons and neutral neutrons, surrounded by negatively charged electrons that travel around the nucleus in orbitals. Every particle possesses a property called spin, a (fast) rotational motion around its axis. Any nucleus with either an odd atomic number or an odd atomic weight has a net spin, and we may regard that nucleus as a charged, spinning sphere (Figure 2.2). Since the nucleus has a net charge, the rotation induces a magnetic field or magnetic moment, with an axis that corresponds to the axis of spin, as shown in Figure 2.2. The amplitude of the magnetic moment is µM .3 For medical MRI, the most important nucleus that has a net spin is the hydrogen nucleus, or proton, owing to its ample occurrence in the human body. The magnetic moments of the nuclei in any volume of matter are oriented randomly. When an external static magnetic field B0 is applied, the magnetic moments will tend to align with this external field.4 The alignment, however, is not perfect. In the presence of an external static magnetic field, the nuclei experience a torque which causes the magnetic moments of these nuclei to rotate around the axis of the external field. This precession is analogous to 2 MRI has its roots in the chemical world, where nuclear magnetic resonance (NMR) has become the most important
analytical technique for the structural analysis of molecules in solution [1]. NMR imaging brought NMR technology into the medical world, and although the word ‘nuclear’ in NMR has nothing to do with radioactivity, this emotionladen word has been dropped by the medical community to avoid confusion or fear in (potential) patients. 3 The subscript M has been introduced to avoid confusion with the symbol µ that is used to represent the electromagnetic permeability. 4 Some of the nuclei will align with the external field and some will align against it. These alignments correspond to quantummechanical energy states, the one aligned with the external field corresponding to a lower energy state. Thus, for a large enough sample, there will be a net alignment with the external field. The amplitude of this net alignment is proportional to the field strength.
23
MRI
N S
Figure 2.2 Spinning nucleus, where N and S indicate the magnetic north pole and south pole, respectively.
G
G L
L
(a)
(b)
B0
B0 µ
µ N
N S
S (c)
(d)
Figure 2.3 Analogy between a spinning top in a gravitational ﬁeld and a spinning nucleus in an external magnetic ﬁeld. (a) Spinning top with angular momentum aligned with the gravitational ﬁeld. (b) Spinning top with angular momentum not aligned with the gravitational ﬁeld. (c) Spinning nucleus with magnetic moment aligned with the external static magnetic ﬁeld. (d) Spinning nucleus with magnetic moment not aligned with the external static magnetic ﬁeld.
the motion of a spinning top with angular momentum L in a gravitational field G, where the spinning top is not aligned with the gravitational field (Figure 2.3) [1].
24
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
In the situation depicted in Figure 2.3(b), the top precesses around the direction of the gravitational field with an angular frequency ω. For the situation depicted in Figure 2.3(d), the rate of precession is isotope and magneticfielddependent. The precession frequency ωL is given by (e.g. [1]) ωL = γ B0 , (2.1) where B0 = B0  is the strength of the external static magnetic field (in T), ωL (in MHz) is known as the Larmor frequency, and the gyromagnetic ratio γ is a constant for any particular nuclear isotope. The Larmor frequency is the frequency at which atomic nuclei respond when interrogated by RF radiation. So far, we have described the MR process on the atomic or microscopic level. On a macroscopic level, we only have to deal with the net results. So, on a macroscopic level, we observe – at equilibrium5 – a net magnetization aligned with the external static magnetic field. Let us assume that the direction of the external static field is the z direction of a Cartesian coordinate system. At equilibrium, the net magnetization is M = Mz uˆ z , where uˆ z is the unit vector in the z direction. 2.2.2
Signal Detection
To record a response from the net magnetic moment of the nuclei, the static magnetic field Mz uˆ z is distorted by a dynamic magnetic field MT with a direction that differs from that of the static field. Owing to this ‘distortion’ component, the nuclei will precess around the direction of the newly formed magnetization M (Figure 2.4). When the dynamic field ceases to exist, the net magnetization will be restored. This change in magnetic field, from the deflected field back to the zdirected magnetic field, may be recorded by virtue of induced currents in RF receiver coils. These receiver coils, which are positioned perpendicular to the direction of the static magnetic field, are sensitive only to dynamic magnetic fields (MT ) in the transverse plane shown in Figure 2.4. The angle ϑ between the net magnetization M and the main field B0 , known as the RF flip angle or RF pulse angle, is given by [1] ϑ = γ B1 t,
(2.2)
where B1 is the amplitude of the disturbing magnetic field and t is the duration of the RF pulse, i.e. the time for which the field MT has been turned on. Maximum signal reception is achieved for ϑ = π/2. When looking at a single nucleus, we see – at equilibrium – the magnetic moment precessing around the direction of the main magnetic field. On a macroscopic scale, we see a magnetization vector directed parallel to the main magnetic field; the individual transverse components cancel each other. When a distorting field (RF pulse) MT is now applied, deflecting the main magnetic field to ϑ = π/2, we see the precession axis move from the 5 For materials with atomic nuclei that possess a property called spin, this spin makes the nuclei behave as small
magnets. Applying a strong external magnetic field to these nuclei results in the precessing of the magnetic moments of the nuclei around the direction of the external field and thus in the forming of a net magnetization in the same direction as that of the external field. It takes a time denoted by T1 to develop this steadystate net magnetization.
25
MRI
z
B0 Mzûz θ
M y
MT x
Figure 2.4 Magnetization.
z
B0 M
y
x
Figure 2.5
Precession of net magnetization during an RF pulse having a ﬂip angle ϑ = π/2.
z direction to a transverse direction and, to an observer in the external laboratory frame of reference, the magnetization vector spirals down – the precession axis following the net magnetization – towards the xy plane (Figure 2.5). In a rotating frame of reference, the net magnetization vector simply rotates from the z direction to the transverse direction.
26
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
amplitude
1
1 1
3
5
7
time Figure 2.6 FID signal. Time and amplitude normalized.
When the RF pulse (at the Larmor frequency) has been transmitted, the RF energy absorbed by the protons (having made them jump to higher energy states) is retransmitted (again at the Larmor frequency). The magnetization starts to return to equilibrium and the protons begin to dephase. The recovery of the magnetization to its thermodynamic equilibrium value M0 with time is described by [4] (2.3) Mz = M0 (1 − e−t /T1 ). The time constant T1 is called the spin–lattice relaxation time.6 Spin–lattice relaxation is the process whereby the energy absorbed by excited protons or spins is released back into the surrounding lattice, reestablishing thermodynamic equilibrium. The dephasing is the result of proton–magneticfield interactions, also known as spin– spin interactions. The result of the dephasing is a decay in the magnitude of the transverse component of the net magnetization. The return of the transverse magnetization Mxy to its equilibrium value Mxy0 with time is described by [4] Mxy = Mxy0 e−t /T2 .
(2.4)
6 The exponential recovery is characterized by a time constant T , at which 63.2% of the magnetization has recovered 1
its alignment with the main magnetic field. The value of T1 is unique to every tissue.
27
INTRAVASCULAR MR ANTENNAS
RF frequency
z
patient dBz dx bandwidth Bz(x=0)
slice thickness
x
Figure 2.7 Magnetic ﬁeld gradient for sliceselective excitation. The ﬁgure shows, schematically, the projection of a patient on the xz plane. When an xdependent magnetic ﬁeld gradient is added to the static magnetic ﬁeld, the Larmor frequency becomes linearly dependent on x. Therefore, every frequency bandwidth (see the right vertical axis) selected in the received signal corresponds to a ‘slice’ of the patient. By choosing the central frequency, the position of the slice can be selected. The slice thickness may be decreased by selecting a smaller frequency bandwidth.
The time constant T2 is known as the spin–spin relaxation time.7 Spin–spin relaxation is a temporary, random interaction between two excited spins that causes a cumulative loss in phase, resulting in an overall loss of signal. In the absence of any gradient in the main magnetic field, the received MR signal (Figure 2.6), is known as the free induction decay (FID). The oscillation of the FID signal is due to the Larmor precession of the net magnetization around B0 . Since the Larmor frequency is directly related to the strength of the main magnetic field, adding a field gradient that depends on a direction orthogonal to the main field direction opens up the possibility to select slices of the patient to be imaged (Figure 2.7). Every slice will now return signals that correspond to a different Larmor frequency. The intensity will give information about the concentration of protons.
2.3
INTRAVASCULAR MR ANTENNAS
The application of intravascular receiver coils or, put more generally, intravascular antennas, will lead to an increase in SNR compared with the use of surface receiver coils. Intravascular antennas can be placed in close proximity to the specific target tissue, which results in a
7 The value of T is the time after excitation when the signal amplitude has been reduced to 36.8% of its original 2 value. The value of T2 is unique to every tissue.
28
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
(a)
(b)
(c)
(d)
(e)
(f)
(g)
(h)
(i)
(j)
(k)
(l)
Figure 2.8 Intravascularantenna concepts. (a) Antiparallel wires [5]. (b) Doublehelix wire [5]. (c) Opposed doublehelix wire [5]. (d) Single loop [6]. (e) Double loop. (f) Triple loop. (g) Solenoid [7]. (h) Dualopposed solenoids [7]. (i) Saddle coil. (j) Fourwire center return. (k) Fourwire birdcage. (l) Quadrature coil [8].
considerable reduction in the amount of received noise. An intravascular antenna can be used for imaging artery walls and may also be employed for tracking purposes, i.e. locating the position and/or orientation of a catheter or catheter tip. 2.3.1
Antenna Designs for Tracking
In Figure 2.8, we show some intravascularantenna concepts reported in the literature. Some of these antenna concepts are more suited for imaging purposes, and some are more suited for tracking purposes. For tracking purposes, the intravascular antenna needs to aid in visualizing the catheter. This may be accomplished semiactively [9], using resonant antennas to locally add gain to the RF magnetic field, or actively, where the antenna is detuned during the RF pulse with external circuitry. The antenna may, for example, be a loop mounted along the complete length of the catheter, used to induce locally, along the catheter, a magneticfield distortion. Examples of such antennas are the antiparallelwire antenna (Figure 2.8(a)), the doublehelix wire antenna (Figure 2.8(b)) and the opposeddoublehelix wire antenna (Figure 2.8(c)).
29
INTRAVASCULAR MR ANTENNAS
Table 2.1 Qualitative comparison of intravascularantenna concepts for tracking [9, 12]. Concept Resonant antenna Antiparallel wires Double helix Opposed double helix Guide wire Three dualopposed solenoids Perpendicular coils
Mechanics
Signal sensitivity
Orientation
Safety
+ + + + + + +
++ + + ++ +/− ++ ++
− −− ++ ++ − − ++
− − − − − − −
The main magneticfield distortion gives rise to a local dephasing, which will be visible as a deviation in the MRI image and thus acts as a position indicator for the catheter. The tip of a catheter may be detected by placing a small resonant antenna or coil at the tip of the catheter and using the detected MR frequency to steer the three orthogonal mainfield gradients to locally code the Larmor frequency of the protons. Thus the catheter tip location can be determined in three dimensions and may be projected onto MRI images [10]. If the catheter orientation needs to be determined as well, the singlecoil antenna may be replaced by multiple coils. This provides multiple highsignal locations. Instead of using the ‘alongthecatheter loop’ antenna mentioned above for detecting the catheter, the guide wire may also be used as a dipole antenna [11]. On the basis of intravascularantenna results reported in the open literature, a qualitative comparison of antenna concepts for imaging is given in Table 2.1 [9, 12]. Passive tracking methods such as using contrast agents or adding magnetic rings to the catheter are not considered here. These methods lack the possibility of dynamically compensating for signal loss for different catheter orientations as is possible with the application of intravascular antennas. First of all, the table shows that none of the antenna concepts is safe. An intravascular MR antenna is safe when it does not present an additional risk to the patient. The largest risk is presented not by the antenna itself, but by its electrical leads. Leads that have a length equal to or longer than half a wavelength (in the surrounding medium) may act as linear antennas and become resonant. This may result in heating of (parts of) the leads to temperature in excess of 70◦ C [10]. At such levels, the surrounding tissue will be destroyed. In section 2.8, we shall address safety issues more thoroughly. Furthermore, the table indicates that the antenna concepts to be studied in more detail are the opposeddoublehelix antenna, the threedualopposedsolenoids antenna (for determining catheter orientation) and the perpendicularcoils antenna. The doublehelix antenna will not be considered for further investigation, since it is outperformed by the opposeddoublehelix antenna. The resonant antenna will not be considered, since this antenna is part of a semiactive tracking system that has the same drawbacks as these mentioned for passive tracking methods. The perpendicularcoils antenna needs some explanation. Originally developed as a fiducial marker [13], i.e. not electrically connected to the MRI scanner hardware, but applied here as an antenna for active tracking [12], the perpendicularcoils antenna consists of two coils, wound on top of each other, the first coil making an angle α with the catheter axis, and the second making an angle π − α with the catheter axis (Figure 2.9). When the angle α is
30
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
α
πα
Figure 2.9 Perpendicularcoils antenna developed from two skewed coils.
chosen to be π/4, the two skewed coils are perpendicular to each other. In this configuration, the net magnetic field is concentrated in the center of the two orthogonal coils and is directed in the radial direction. 2.3.2
Antenna Designs for Imaging
The main requirements on an antenna for arterialwall imaging are a high sensitivity outside the catheter boundaries up to and beyond the artery wall, a radially hom*ogeneous sensitivity pattern and, preferably, a large longitudinal coverage for multislice imaging. On the basis of results reported in the open literature, a qualitative comparison of antenna concepts for imaging is given in Table 2.2 [9]. In this table, mechanics stands for size, rigidity and complexity, orientation stands for sensitivity to the antenna orientation and safety relates to the length of the antenna. In selecting antenna concepts worthwhile to be further investigated, we shall start by omitting all candidates that show a double minus sign in one or more column entries. That leaves us with the doubleloop antenna, the tripleloop antenna and the dualopposedsolenoids antenna to be evaluated in more detail. Although the saddlecoil antenna must be discarded owing to its complexity, we shall still investigate this concept, assuming a feasible construction method. 2.4
MR ANTENNA MODEL
To compare the different antenna concepts without actually constructing prototypes and performing in vitro and in vivo experiments8 with an MRI scanner, the availability of an
8 In vitro – Latin for ‘in glass’ – is an experimental technique where the experiments are performed outside a living
organism. Here, it means that experiments are performed within a phantom. In vivo – Latin for ‘in the living’ – indicates that the experiments are performed in the presence of a living organism.
31
MR ANTENNA MODEL
Table 2.2 Qualitative comparison of intravascularantenna concepts for imaging [9]. Concept Single loop Single loop Multiturn Double loop Triple loop Dipole Twin lead Dual opposed solenoids Saddle coil Center return Birdcage Quadrature coil
Mechanics Radial sensitivity Axial sensitivity Orientation Safety +
−
++
−−
+/−
− + − − ++ ++ −− −− −− −−
− + + +/− + ++ + + + −
+/− ++ ++ ++ + +/− ++ −− −− ++
− − − − − − − − − −
+/− +/− +/− −− −− − − − − −
electromagnetic model for calculating the fields is desirable. To this end, commercialofftheshelf (COTS), threedimensional, fullwave electromagnetic analysis software may be applied, for example [14] (finite element method), [15] (transmission line matrix method) and [16] (method of moments). To obtain analysis results very rapidly and possibly optimize antenna designs through repeated analyses, however, we prefer to develop an approximate model; this is feasible for antennas that are small compared with the wavelength [16]. Using equation (2.1) for protons (1 H), which have a gyromagnetic ratio of 42.58 MHz T−1 [1], an MRI scanner that produces a 1.5 T strong static magnetic field [9, 12], gives us a Larmor frequency of 63.87 MHz. Owing to the field gradient applied, the frequency will vary around this value and, for convenience, we shall therefore assume, from now on, a resonance frequency f = 64 MHz. The medium that surrounds the intravascular antenna will be mainly blood, which is characterized by a relative permeability µr ≈ 1, a relative permittivity εr ≈ 80 and a conductance σ ≈ 8 S m−1 at a temperature of 37◦C [17, 18]. The wavelength may thus be calculated as 1 c0 1 = √ = 0.52 m. (2.5) λ= √ µr ε r f f µ0 µr ε 0 ε r The large blood vessels for which MRI antennas are needed have a diameter between 4 and 6 mm [9, 18]. Therefore an antenna diameter of about 2 mm is anticipated, allowing the catheter to be maneuvered through the vascular system and preventing complete blood flow blockage.9 The far field for a small antenna [19] starts at a distance of λ/2π = 82.8 mm from the antenna. With the stated dimensions of intravascular antennas and vessels, the vessel wall position will be in the near field of the antenna.
9 The first prototypes, used for in vitro experiments, were made a little larger. A diameter of 4 mm was dictated by the
materials and construction facilities available at the time [9]. Later, in vitro experiments dedicated to intravascular antennas for tracking purposes were performed with antennas positioned on 5F catheters, which have a diameter of 1.67 mm [12].
32
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
z
z
P
P r
r
h
y
I0
I0
y
x
x (a)
(b)
Figure 2.10 Elementary radiators. (a) Elementary, or Hertzian, dipole. (b) Elementary loop, or magnetic dipole.
The radiated fields of an elementary, or Hertzian, dipole10 at an observation position P (r, ϑ, ϕ) (Figure 2.10(a)) may be calculated as follows [20]: kI0 h sin(ϑ) 1 1+ e−jkr , 4πr jkr 1 I0 h cos(ϑ) Er = η e−jkr , 1 + jkr 2πr 2 1 kI 0 h sin(ϑ) 1 − Eϑ = jη e−jkr , 1+ 4πr jkr (kr)2 Hr = Hϑ = Eϕ = 0, Hϕ = j
(2.6) (2.7) (2.8) (2.9)
where I0 is the amplitude of the uniform current, h is the length of the dipole, ω = 2πf is the radial frequency, η is the characteristic impedance of free space, ε = ε0 εr , µ = µ0 µr and k is the wave number. The radiated fields of an elementary magnetic dipole11 at an observation position P (r, ϑ, ϕ) (Figure 2.10(b)), may be obtained from the results for an elementary dipole by
10 A Hertzian dipole is a dipole antenna with a length much smaller than the wavelength, in fact so small that the
current may be considered to be uniform over the length. 11 An elementary magnetic dipole can be realized as an electriccurrent loop with a circumference that is so much
smaller than the wavelength that the current may be considered to be uniform over the loop.
33
MR ANTENNA MODEL
virtue of duality12 [20], or may be calculated as follows [21]: 1 I0 (ka)2 sin(ϑ) Eϕ = η 1+ e−jkr , 4r jkr 1 I0 ka2 cos(ϑ) e−jkr , Hr = j 1 + jkr 2r 2 1 1 I0 (ka)2 sin(ϑ) Hϑ = − 1+ − e−jkr , 4r jkr (kr)2 Er = Eϑ = Hϕ = 0,
(2.10) (2.11) (2.12) (2.13)
where a is the radius of the loop. Far away from these elementary antennas, the r −1 terms dominate; very close to the antenna, the r −3 terms are dominant. In between, the r −2 terms are dominant. If, for a very small but not elementary antenna, we can identify a region where the r −2 terms are clearly dominant and if the artery wall is in this region, then it is likely that we may be able to approximate the magnetic field in this region of interest by a quasistatic magnetic field. The Biot–Savart law [20] tells us that the static magnetic field produced by a steady current I0 shows an I0 r −2 dependence. Upon inspection of equations (2.8)–(2.11), we see that for dominating r −2 terms, the dynamic magnetic field also shows an I0 r −2 dependence. To verify this hypothesis, we shall calculate the magnetic field of a small loop antenna of radius a by employing equations (2.9) and (2.10) and compare these results with those obtained by applying the Biot–Savart law to a direct current in a loop. The loop will be approximated by a finite number of straight wire segments. We have specifically chosen a small loop antenna, since most concepts for intravascular antennas that have been demonstrated are based upon small loops. Before we start this comparison, we first need to verify our reference, i.e. the smallloop approximation, based on a constant direct current, which gave rise to equations (2.10)– (2.13). In [22] it was shown that, first of all, the smallloop approximations resulting in equations (2.10)–(2.13) may be obtained as a limiting case of general exact series representations for a uniform current loop. Secondly, it was demonstrated in [22] that the field component Hϑ in the plane of the loop, at a distance of half a wavelength from the loop center, as calculated by equation (2.12), is less than 5% in error compared with the exact solution for a uniform current for loop radii up to 0.11λ and less than 10% in error for loop radii up to 0.15λ. For practical purposes, therefore, we need to find up to what radius a loop antenna may be regarded as supporting a uniform current. In [23], it was demonstrated that for loop circumferences larger than a wavelength, the current may not be treated as uniform. To find limiting values for the loop radius, we shall look at the admittance of a loop.
12 The concept of duality states that exchanging H for E, E for H, µ for ε and ε for µ leaves Maxwell’s curl equations
for sourcefree regions unchanged. Thus solutions for a problem with an electric source can be adapted to problems with a magnetic source [20].
34
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
Iϕ
ϕ=0
2b
Ve0δ
2a
Figure 2.11
2.4.1
Circular loop antenna.
Admittance of a Loop
We start by considering a loop in air. The loop has a radius b and is made of a wire having a circular cross section defined by a radius a. The loop is excited at ϕ = 0 with a voltage deltagap generator V0e δ(ϕ) and carries a current Iϕ (Figure 2.11). Expanding the current in a Fourier series [24–26], ∞ V0e 1 cos(nϕ) +2 Iϕ = −j , (2.14) η0 π a0 an n=1 results in an input impedance Yin = where η0 =
∞ I (0) 1 1 1 = −j + 2 , V0e η0 π a0 a n=1 n
(2.15)
√ µ0 /ε0 is the characteristic impedance of free space and an =
k0 b n2 (Fn+1 + Fn−1 ) − Fn . 2 k0 b
√ Here k0 = ω ε0 µ0 is the freespace wave number, 1 8b 1 2k0 b F0 = ln [0 (x) + jJ0 (x)] dx − π a 2 0
(2.16)
(2.17)
and Fn = F−n =
1 na na 1 2k0 b [2n (x) + jJ2n (x)] dx, n = 0. (2.18) K0 I0 + Cn − π b b 2 0
In the above, J0 (x) is the Bessel function of the first kind and order zero with argument x, I0 (x) is the modified Bessel function of the first kind and order zero with argument x, K0 (x)
35
MR ANTENNA MODEL
is the modified Bessel function of the second kind and order zero with argument x, Cn = γ − 2
n−1
(2m + 1)−1 + ln(4),
(2.19)
m=0
where γ = 0.5772 . . . is Euler’s constant, and 1 π m (x) = sin[x sin(ϑ) − mϑ] dϑ, π 0
(2.20)
is the Lommel–Weber function of order m with argument x. For a loop antenna immersed in a dissipative medium (such as blood), the equations stated above still apply, but we need to replace ε0 by ε − jσ/ω, µ0 by µ and k0 by k [24–26], where √ k = β − jα = ω µε 1 − jp. (2.21) Here p = σ/ωε and
1 − jp = cosh
1 1 sinh−1 (p) − j sinh sinh−1 (p) . 2 2
(2.22)
Furthermore, ε = ε0 εr , where, for blood and at a frequency of 64 MHz, εr ≈ 80 [17, 18] and σ ≈ 8 S m−1 [17, 18]. The input admittance is then found as Yin = −j where
∞ 1 (1 − jα/β) 1 , +2 πη0 a0 a n=1 n
1 √ −1 α = ω µε sinh sinh (p) , 2 1 √ −1 β = ω µε cosh sinh (p) , 2 εr 1 −1 = cosh sinh (p) µr 2
and
(2.23)
(2.24) (2.25) (2.26)
σ . (2.27) ωε With the use of equations (2.15)–(2.20), the input impedance of a loop antenna in air with a thickness parameter = 2 ln (2πb/a) = 10 has been calculated as a function of the loop radius expressed in wavelengths. For the analysis, 20 Fourier terms were used. As shown in [25], that number of Fourier terms leads to convergent impedance values for loop radii satisfying βb ≤ 0.5, or b ≤ 0.08λ. The results for the resistance are shown in Figure 2.12, and the results for the modulus of the reactance are shown in Figure 2.13. The input impedance of a uniform current loop antenna as a function of the loop radius was also calculated using the same equations, but taking only the a0−1 term into account. The results for this smallloop p=
36
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
1000
1
0.9 10
0.8
resistance (Ω)
0.6
0.5 0.001
0.4
Fourier analysis Small loop Rel. difference 5% rel. difference
1e05
1e07
0.3
relative difference
0.7
0.1
0.2
0.1 0
1e09
Figure 2.12 radius.
0.01
0.02 0.03 0.04 loop radius (wavelengths)
0.05
0.1 0.06
Real part of the input impedance of a loop antenna as a function of the loop
approximation are shown in the same two figures. The figures agree with the results presented in [19], both for the 20term Fourier analysis and for the smallloop approximation. In Figures 2.12 and 2.13, the relative difference between the 20term Fourier analysis and the smallloop approximation is also shown as a function of loop radius. Accepting a 5% difference between the approximate and exact results and also taking account of the fact that the reactance dominates over the resistance leads to the commonly quoted rule of thumb, [19, 21,27], that the circumference of a loop antenna should be smaller than approximately a tenth of a wavelength for the uniformcurrent approximation to be valid. Having thus established the validity of the analysis methods in air, we took the final step of immersing the loop antenna in a dissipative medium, more specifically blood, for which we have used εr = 80 and σ = 8. Applying equation (2.23) would result in replacing the real integration limits in equations (2.17) and (2.18) by complex ones and having to deal with Bessel and Weber–Lommel functions of complex arguments. Since solving this problem would be beyond the scope of establishing the validity of our reference, we shall follow [28] and use a power series expansion for a small loop antenna. Taking only the first two terms in equation (2.23) into account, developing power series for the functions Fn (equations (2.17) and (2.18)) for complex k = β(1 − jα/β) and using
37
MR ANTENNA MODEL
1
1000
0.9
0.8
reactance (Ω)
0.6
0.5 0.4
Fourier analysis Small loop Rel. difference 5% rel. difference
10
relative difference
0.7
100
0.3
0.2
0.1 1
Figure 2.13 radius.
0.01
0.02
0.03 0.04 loop radius (wavelengths)
0.05
Imaginary part of the input impedance of a loop antenna as a function of the loop
only the first few terms of these power series leads to [28] (1 − jα/β) 1 1 2 Yin = −j + πη0 kb a0 a1 1 1 1 4 , = −j + 120π 2 βb F1 F0 + F2 − (2)/((kb)2 )F1 where
0 0.06
(2.28)
8b 1 4 32 1 1 1 2 4 6 3 5 F0 = ln (kb) − j kb − (kb) + (kb) , − 2(kb) − (kb) + π a π 9 675 3 20 (2.29) 8b 1 2 1 1 4 160 1 − 2 − (kb)2 + (kb)4 − (kb)6 − j (kb)3 − (kb)5 , F1 = ln π a π 3 15 4725 6 30 (2.30) 8b 2 1 4 1 8 − (kb)2 − (kb)4 + 0.0104(kb)6 F2 = ln − π a π 3 15 105 1 1 (kb)5 − (kb)7 , (2.31) −j 120 840
38
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
250
200
450
R, twoterm Fourier power series R, 20term Fourier X, twoterm Fourier power series X, 20term Fourier
400
350
150
250
200 100
150
reactance, X (Ω)
resistance, R (Ω)
300
100
50
50 0
0.05
0.1
0.15
0.2 loop circumference (wavelengths)
0.25
0 0.3
Figure 2.14 Normalized impedance of a circular loop antenna in a dissipative medium for which α/β = 1. Exact and approximate analysis.
and where use has been made of and
a0 = kbF1
(2.32)
1 F0 + F2 − F1 . a1 = kb 2 kb
(2.33)
is given in equation (2.26). In Figure 2.14, we show the normalized input impedance thus calculated as a function of loop circumference over the wavelength for α/β = 1. The normalization is with respect to . In the same figure, we show the results of a 20term Fourier analysis, taken from [24, 25]. The results apply to the problem at hand, since for blood (εr = 80 and σ = 8), the ratio of α and β is equal to 0.98. The figure shows that, up to relatively large loop sizes, the loop reactance is well modeled by the first two terms of the Fourier analysis. The approximation for the loop resistance, however, starts to deviate seriously from the exact value for loop circumferences exceeding 0.05 wavelengths. So, to find the loop radius limit for which a uniform current may be assumed, we must restrict our analysis to loop circumferences 2πb ≤ 0.05λ. In Figure 2.15, we show again the results of a 20term Fourier analysis and also an approximate twoterm Fourier analysis for the input impedance of a small loop antenna over a smaller circumferenceoverwavelength range than that in Figure 2.14. In this figure, we show also
39
MR ANTENNA MODEL
8
200
resistance, R (Ω)
6 5
180 160 140
120
100
4
80
3
60
reactance, X (Ω)
R, twoterm Fourier power series R, 20term Fourier R, oneterm Fourier power series R, interpolated X, twoterm Fourier power series X, 20term Fourier X, oneterm Fourier power series
7
2
40 1
20
0.02
0.04 0.06 0.08 loop circumference (wavelengths)
0 0.1
Figure 2.15 Normalized impedance of a circular loop antenna in a dissipative medium for which α/β = 1. Exact, approximate and uniformcurrent analysis.
the results of a singleterm Fourier analysis, i.e. the results of a uniformcurrent analysis. Finally, the figure shows the result of a rationalfunction interpolation [29] based on the two smallest twoterm Fourier analysis results and the four smallest 20term Fourier analysis results, bridging the gap between these two analyses. The figure again shows fair agreement over the entire circumference range between all simulation results for the reactance of the loop, but poor agreement between the resistance simulation results. For small loop circumferences or radii, the agreement between the exact reactance and the reactance based on a uniformcurrent approximation is excellent. The resistance for a uniformcurrent approximation approaches that for the twoterm Fourier analysis when the circumference becomes infinitely small. The resistance value, though, for small loops is outweighed by the reactance value.13 Taking this into account, we start by determining the 5% deviation between the one and twoterm Fourier analysis results for the loop reactance as a function of the circumference. As Figure 2.16 shows, this restricts us to loops of circumference smaller than 0.2λ. Next, somewhat arbitrarily but aiming at a fair assessment, we determine the circumference value below which the reactance is two to three orders of magnitude larger than the resistance, and take this value as our maximum circumference value that allows a 13 A small loop antenna carrying a uniform current may be considered as a radiating inductor [19].
40
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
0.2
0.18
rel. difference 5% rel. difference
0.16 relative difference
0.14
0.12
0.1
0.08
0.06
0.04
0.02
0 0
0.05
0.1
0.15
0.2
0.25
0.3
loop circumference (wavelengths) Figure 2.16 Relative diﬀerence between the approximate and the uniformcurrent analysis of the reactance of a circular loop antenna in a dissipative medium for which α/β = 1.
uniformcurrent approximation. Following this line of reasoning, we put a restriction on the circumference of a loop antenna in blood of 0.02 wavelengths. This means that the radius should not exceed 1.7 mm. Thus, we may consider a loop antenna immersed in blood and subject to a 1.5 T main MR magnetic field to carry a uniform current if the radius of the loop does not exceed 1.7 mm. 2.4.2
Sensitivity
In the remainder of this section, we shall calculate sensitivity patterns. A sensitivity pattern displays the sensitivity of an intravascular antenna to the magnetic field as a function of position (in the near field). Owing to the reciprocity of a passive antenna, this threedimensional pattern or twodimensional sections of this pattern may be calculated from the magneticfield amplitude as a function of the nearfield position, transmitted by the antenna when a current I flows through the wire. The sensitivity is defined as [9] 1 2 Bx + By2 (T A−1 ) (2.34) S= I where Bi = µHi , i = x, y. This definition is based on the argument that magnetic fields will be measured only in the transverse (xy) plane in an MR scanner where the main magnetic field is zdirected.
41
MR ANTENNA MODEL
dH
z
P r
R=rr0 r0
dI y
O x Figure 2.17 Magnetic ﬁeld induced by an inﬁnitesimal straight wire segment.
2.4.3
Biot–Savart Law
Ampère’s law relates the induced magnetic field of a general but stationary current path to that current path. Before Ampere formulated this relation, Biot and Savart derived a quantitative relation for the special case of a straight wire [20]. For a current element dI at position r0 relative to a chosen origin, the induced magnetic field dH at a position P = P (r) relative to the same origin (Figure 2.17), is given by dH(r) =
dI × R dl × R = I (r0 ) , 4πR 3 4πR 3
(2.35)
where dI = I (r0 )dl. Equation (2.35) is known as the Biot–Savart law, although, for the reasons mentioned above, it is also referred to as Ampère’s law [20].14 The total magnetic field H(r) of the current elements around a current path C is obtained by integrating this equation over the path: dl × R I (r0 ) . (2.36) H(r) = 4πR 3 C This integral is readily evaluated for observation positions on the axis of a circular loop, but off axis and for shapes more complex than a circular loop, this is difficult or impossible [31]. In order to use the Biot–Savart law with more complicated wire structures, it is necessary to subdivide the structure into segments that result in integrals that can be evaluated in closed form. To that end, it is desirable that the equation of such a line segment may be expressed in terms of a single parameter ζ [31], l = l(ζ ) = uˆx x(ζ ) + uˆy y(ζ ) + uˆz z(ζ ),
(2.37)
14 The Biot–Savart law is postulated in [20] and derived in [30] for the static situation, and may be derived for the
quasistatic situation as well, as is demonstrated in Appendix 2.A.
42
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
where the uˆi , i = x, y, z, are unit vectors in the x, y and z directions, respectively, of a Cartesian coordinate system. The infinitesimal segment in equation (2.36) is then dx(ζ ) dy(ζ ) dz(ζ ) dl(ζ ) dζ = uˆx + uˆy + uˆz dl = dζ. (2.38) dζ dζ dζ dζ The vector R in equation (2.36) is given by (Figure 2.17) R = uˆx [x(ζ ) − x] + uˆy [y(ζ ) − y] + uˆz [z(ζ ) − z],
(2.39)
where it is understood that P = P (x, y, z). For a straight wire segment between the positions (x1 , y1 , z1 ) and (x2 , y2 , z2 ), the functions x(ζ ), y(ζ ) and z(ζ ) are simply (x1 + (x2 − x1 )ζ ), (y1 + (y2 − y1 )ζ ) and (z1 + (z2 − z1 )ζ ), respectively, with 0 ≤ ζ ≤ 1. So, equations (2.38) and (2.39), for this straight wire segment, become dl = [uˆx (x2 − x1 ) + uˆy (y2 − y1 ) + uˆz (z2 − z1 )] dζ
(2.40)
and R = uˆx [(x1 − x) + (x2 − x1 )ζ ] + uˆy [(y1 − y) + (y2 − y1 )ζ ] + uˆz [(z1 − z) + (z2 − z1 )ζ ].
(2.41)
The cross product of dl and R may then be computed as dl × R
uˆx
(x2 − x1 ) dζ =
(x1 − x) + (x2 − x1 )ζ
uˆy (y2 − y1 ) dζ (y1 − y) + (y2 − y1 )ζ
uˆz
, (z2 − z1 ) dζ
(z1 − z) + (z2 − z1 )ζ
(2.42)
and R 3 = {[(x1 − x) + (x2 − x1 )ζ ]2 + [(y1 − y) + (y2 − y1 )ζ ]2 + [(z1 − z) + (z2 − z1 )ζ ]2}3/2 .
(2.43)
With the use of equation (2.36), the magnetic field at position P = P (x, y, z) due to a unit current I (r0 ) = 1 flowing in a straight wire segment between the points (x1 , y1 , z1 ) and (x2 , y2 , z2 ) is given by [31] dl × R H(r)I (r0 )=1 = 3 C 4πR 1 1 Dy Dx = uˆx dζ + uˆy dζ 2 )3/2 (A + Bζ + Cζ (A + Bζ + Cζ 2 )3/2 ζ =0 ζ =0 1 Dz + uˆz dζ, (2.44) 2 3/2 ζ =0 (A + Bζ + Cζ )
43
MR ANTENNA MODEL
where A = (x1 − x)2 + (y1 − y)2 + (z1 − z)2 ,
(2.45)
B = 2[(x1 − x)(x2 − x1 ) + (y1 − y)(y2 − y1 ) + (z1 − z)(z2 − z1 )],
(2.46)
C = (x2 − x1 )2 + (y2 − y1 )2 + (z2 − z1 )2 ,
(2.47)
Dx = (y2 − y1 )(z1 − z) − (z2 − z1 )(y1 − y), Dy = (z2 − z1 )(x1 − x) − (x2 − x1 )(z1 − z)
(2.48) (2.49)
and Dz = (x2 − x1 )(y1 − y) − (y2 − y1 )(x1 − x).
(2.50)
The magnetic field at position P produced by multiple straight wire segments is the sum of the contributions calculated for the isolated wire segments. 2.4.4
Model Veriﬁcation
To validate the applicability of the Biot–Savart model thus derived for our intravascular MR antennas subject to a 1.5 T main magnetic field, we shall compare the results obtained from the Biot–Savart model with results that can be obtained analytically for a small loop antenna. We have seen that, for the above static MR magnetic field, a small loop antenna immersed in blood may be considered to carry a uniform current as long as the radius of the loop does not exceed 1.7 mm. The radiated fields of such a loop antenna are obtained from those derived for a loop in air, stated in equations (2.10)–(2.13). The adaptation for the surrounding dispersive medium is accomplished by substituting k˜ for k, η˜ for η and ε˜ for ε in these equations, where [28, 32] ˜k = β 1 − j α , (2.51) β √ µ0 /εr ε0 (1/f (p)) , (2.52) η˜ = 1 − jα/β σ ε˜ = εr ε0 − j , (2.53) ω with p=
σ , ωε
f (p) = cosh and
(2.54) 1 sinh−1 (p) 2
1 α = tanh sinh−1 (p) . β 2
(2.55)
(2.56)
For blood, p = 28.09, f (p) = 3.81, λ = 0.52 m and α/β = 0.97. To verify the model, singleloop antennas of radii 0.5 mm, 1.0 mm and 1.5 mm were placed at an angle ϑ relative to the z axis of a rectangular coordinate system. The sensitivity
44
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
z
y
x
Figure 2.18
Singleloop antenna, positioned in a zdirected artery.
was calculated for different positions y relative to the loop center (Figure 2.18). In this figure, the position of the vascular wall is indicated by a circular cylinder, with its central axis along the z direction. For these special situations, we may equivalently keep the loop positioned parallel to and in the xy plane, and calculate the sensitivity for different rotation angles ϑ and for a unit current as S = Br2 + Bϕ2 = Br (Figure 2.19). With the aid of [21], the sensitivity along the y axis (ϕ = π/2) can be obtained from the Cartesian components of the magnetic flux density as S = Br = sin(ϑ)By + cos(ϑ)Bz . For ϑ = 0, the sensitivity is equal to the magnetic flux density on the axis of the loop antenna. For a loop carrying a uniform current I0 , the static magnetic flux density on the axis may be calculated in closed form as (Figure 2.18) [20] S = B = By = µ0
I0 a 2 , 2(a 2 + y 2 )3 /2
(2.57)
where a is the radius of the loop and y is the distance between the observation point on the axis of the loop and the center of the loop. In Figures 2.20–2.22, the sensitivities calculated using the dynamic loop model and using the Biot–Savart model for a segmented loop are shown as a function of the distance from the center of the loop. The results are shown for three different loop radii, all loops being at an angle ϑ = 0 with respect to the z axis. The analytic results for the static sensitivity on the loop axis (equation (2.51)) are also shown in these figures. For the calculation based on the Biot–Savart model, 40 straight segments were used to approximate the loop; this is large enough to represent a circular loop accurately [9].
45
MR ANTENNA MODEL
z
ûr
û û y
x
Figure 2.19 Alternative for calculating the sensitivity of a singleloop antenna rotated ϑ from the z axis.
0.001
1 Dynamic Static Static analytic relative difference
1e05
0.8
0.7
0.6
1e09
0.5
1e11
0.4
relative difference
S (T/A)
1e07
0.9
0.3
1e13
0.2
1e15
0.1
1e17 1
10
100
0 1000
r (mm)
Figure 2.20 Sensitivity versus distance from loop center for ϑ = 0, a = 0.5 mm.
Also, in the same figures, the relative difference δ between the dynamic sensitivity Sdyn and the static sensitivity Sstat is shown (δ = (Sdyn − Sstat )/Sdyn ). Note that the sensitivity and the distance are displayed on a logarithmic scale.
46
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
0.001
1 Dynamic Static Static analytic relative difference
1e05
0.8
0.7
0.6
1e09
0.5
1e11
0.4
relative difference
S (T/A)
1e07
0.9
0.3
1e13
0.2
1e15
0.1
1e17 1
10
100
0 1000
r (mm)
Sensitivity versus distance from loop center for ϑ = 0, a = 1.0 mm.
0.001
1 Dynamic Static Static analytic relative difference
1e05
S (T/A)
1e07
0.9
0.8
0.7
0.6
1e09
0.5
1e11
0.4
relative difference
Figure 2.21
0.3
1e13
0.2
1e15
0.1
1e17 1
10
100
0 1000
r (mm)
Figure 2.22
Sensitivity versus distance from loop center for ϑ = 0, a = 1.5 mm.
These figures clearly show that the sensitivity calculated by the Biot–Savart model for a segmented loop coincides with the static analytical onaxis results for all distance values.
47
MR ANTENNA MODEL
1
1
0.1
0.9
Dynamic Static relative difference
0.8
0.7
0.6
0.001
0.5
0.0001
0.4
relative difference
S (T/A)
0.01
0.3
1e05
0.2
1e06
0.1
1e07
1
2
3
4
5
6
7
8
9
10
r (mm)
Figure 2.23 Sensitivity versus distance from loop center for ϑ = 0, a = 0.5 mm.
The figures also show that very close to the loop, where the fields show an r −3 dependence, the dynamic sensitivity differs substantially from the static value. This difference is larger for loops with a larger radius. Far away from the loop, where the fields show an r −1 dependence, the dynamic sensitivity starts to differ substantially from the static value again, and very far away from the loop, the difference between the dynamic and the static sensitivity becomes very large, owing to the attenuation of the dynamic fields, which show an e−αr dependence, α being the attenuation constant. This attenuation is accounted for in the dynamic model by the multiplication e−jkr in equations (2.11) and (2.12) but is not taken into account in the static model.15 In between these areas of substantially different sensitivities, we observe an area of minimum relative difference. This area, where the dynamic (and static) fields show an r −2 dependence, seems to coincide with our area of interest: the position of the artery wall, which, for medium and large arteries, varies between 1.0 mm and 3.0 mm [18]. Figures 2.23–2.25 show the same dynamic and static sensitivities and relative differences between these two sensitivities, but now over a smaller distance range. We observe that, on axis, for a wellchosen loop radius, i.e. 0.5 mm or less, the static model approximates the dynamic sensitivity with less than 30% deviation in the region of interest. If we concentrate on large arteries only (radii between 2.0 mm and 3.0 mm), this deviation is less than 13%.
15 The same attenuation could be introduced into the sensitivity parameter calculated with the Biot–Savart model. This would make the model more realistic but, owing to the r −1 behavior of the dynamic fields, the dynamic and
static sensitivities would still diverge with the distance r. At distances not very far away from the loop center, the influence of the attenuation is negligible. Therefore the attenuation was not taken into account in the Biot–Savart model.
48
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
1
1
0.1
0.9
Dynamic Static relative difference
0.8
0.7
0.6
0.001
0.5
0.0001
0.4
relative difference
S (T/A)
0.01
0.3
1e05
0.2
1e06
0.1
1e07
1
2
3
4
5
6
7
8
9
10
r (mm)
Sensitivity versus distance from loop center for ϑ = 0, a = 1.0 mm.
1
1
0.1
0.9
Dynamic Static relative difference
0.8
S (T/A)
0.01
0.7
0.6
0.001
0.5
0.0001
0.4
relative difference
Figure 2.24
0.3
1e05
0.2
1e06
0.1
1e07
1
2
3
4
5
6
7
8
9
10
r (mm)
Figure 2.25
Sensitivity versus distance from loop center for ϑ = 0, a = 1.5 mm.
49
MR ANTENNA MODEL
1
1
0.1
0.9
Dynamic Static relative difference
0.8 0.7 0.6
0.001
0.5 0.0001
0.4
relative difference
S (T/A)
0.01
0.3
1e05
0.2 1e06
0.1
1e07
0 1
3
4
5 6 r (mm)
7
8
9
10
Sensitivity versus distance from loop center for ϑ = 30◦ , a = 0.5 mm.
1
1
0.1
0.9
Dynamic Static relative difference
0.8
S (T/A)
0.01
0.7 0.6
0.001
0.5 0.0001
0.4
relative difference
Figure 2.26
2
0.3
1e05
0.2 1e06
0.1
1e07
0 1
2
3
4
5
6
7
8
9
10
r (mm)
Figure 2.27
Sensitivity versus distance from loop center for ϑ = 30◦ , a = 1.0 mm.
50
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
1
1
0.1
0.9
Dynamic Static relative difference
0.8 0.7 0.6
0.001
0.5 0.0001
0.4
relative difference
S (T/A)
0.01
0.3
1e05
0.2 1e06
0.1
1e07
0 1
2
3
4
5 6 r (mm)
7
8
9
10
Figure 2.28 Sensitivity versus distance from loop center for ϑ = 30◦ , a = 1.5 mm.
1
1
0.1
0.9
Dynamic Static relative difference
0.8 0.7 0.6
0.001
0.5 0.0001
0.4
relative difference
S (T/A)
0.01
0.3
1e05
0.2 1e06
0.1
1e07
0 1
2
3
4
5
6
7
8
9
10
r (mm)
Figure 2.29 Sensitivity versus distance from loop center for ϑ = 45◦ , a = 0.5 mm.
51
MR ANTENNA MODEL
1
1
0.1
0.9
Dynamic Static relative difference
0.8 0.7 0.6
0.001
0.5 0.0001
0.4
relative difference
S (T/A)
0.01
0.3
1e05
0.2 1e06
0.1
1e07
0 1
3
4
5 6 r (mm)
7
8
9
10
Sensitivity versus distance from loop center for ϑ = 45◦ , a = 1.0 mm.
1
1
0.1
0.9
Dynamic Static relative difference
0.8
S (T/A)
0.01
0.7 0.6
0.001
0.5 0.0001
0.4
relative difference
Figure 2.30
2
0.3
1e05
0.2 1e06
0.1
1e07
0 1
2
3
4
5
6
7
8
9
10
r (mm)
Figure 2.31
Sensitivity versus distance from loop center for ϑ = 45◦ , a = 1.5 mm.
52
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
1
1
0.1
0.9
Dynamic Static relative difference
0.8 0.7 0.6
0.001
0.5 0.0001
0.4
relative difference
S (T/A)
0.01
0.3
1e05
0.2 1e06
0.1
1e07
0 1
2
3
4
5 6 r (mm)
7
8
9
10
Figure 2.32 Sensitivity versus distance from loop center for ϑ = 60◦ , a = 0.5 mm.
1
1
0.1
0.9
Dynamic Static relative difference
0.8 0.7 0.6
0.001
0.5 0.0001
0.4
relative difference
S (T/A)
0.01
0.3
1e05
0.2 1e06
0.1
1e07
0 1
2
3
4
5
6
7
8
9
10
r (mm)
Figure 2.33 Sensitivity versus distance from loop center for ϑ = 60◦ , a = 1.0 mm.
53
MR ANTENNA MODEL
1
1
0.1
0.9
Dynamic Static relative difference
0.8 0.7 0.6
0.001
0.5 0.0001
0.4
relative difference
S (T/A)
0.01
0.3
1e05
0.2 1e06
0.1
1e07
0 1
3
4
5 6 r (mm)
7
8
9
10
Sensitivity versus distance from loop center for ϑ = 60◦ , a = 1.5 mm.
1
1
0.1
0.9
Dynamic Static relative difference
0.8
S (T/A)
0.01
0.7 0.6
0.001
0.5 0.0001
0.4
relative difference
Figure 2.34
2
0.3
1e05
0.2 1e06
0.1
1e07
0 1
2
3
4
5
6
7
8
9
10
r (mm)
Figure 2.35
Sensitivity versus distance from loop center for ϑ = 85◦ , a = 0.5 mm.
54
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
1
1
0.1
0.9
Dynamic Static relative difference
0.8 0.7 0.6
0.001
0.5 0.0001
0.4
relative difference
S (T/A)
0.01
0.3
1e05
0.2 1e06
0.1
1e07
0 1
2
3
4
5 6 r (mm)
7
8
9
10
Figure 2.36 Sensitivity versus distance from loop center for ϑ = 85◦ , a = 1.0 mm.
1
1
0.1
0.9
Dynamic Static relative difference
0.8 0.7 0.6
0.001
0.5 0.0001
0.4
relative difference
S (T/A)
0.01
0.3
1e05
0.2 1e06
0.1
1e07
0 1
2
3
4
5
6
7
8
9
10
r (mm)
Figure 2.37 Sensitivity versus distance from loop center for ϑ = 85◦ , a = 1.5 mm.
MR ANTENNA MODEL
55
For angles ϑ = 30◦ , 45◦ , 60◦ and 85◦ , the dynamic and static sensitivities as a function of y (Figure 2.18) and the relative difference between these two values are shown for three loop radii, in Figures 2.26–2.28, 2.29–2.31, 2.32–2.34 and 2.35–2.37, respectively. For ϑ = 90◦ , both the dynamic and static sensitivity are calculated to be zero. These figures reveal that, especially for large arteries (radii between 2 mm and 3 mm [18]) and small loop antennas (a = 0.5 mm), sensitivities that deviate by less than 13% from the exact values may be calculated by the static method. Moreover, the static and dynamic sensitivities in the area of interest (i.e. in and around the position of the vascular wall) show similar behavior as a function of distance from the center of the (tilted) loop antenna. This means that, under welldefined conditions, the static model may be used to predict the absolute value of the sensitivity of a loop antenna with a reasonable accuracy, but – more importantly – the static model may be used to compare different designs with respect to sensitivity profiles. Now that we have shown the validity of the static model for singleloop antennas immersed in blood, the next question to be answered concerns the validity of employing this model for multiple loops, or, more generally, wire antennas where the length of the wire is larger than that in a single loop of radius 1.7 mm. Restricting ourselves for convenience, for the purpose of this discussion, to multipleloop antennas, two issues need to be examined. The first is whether the uniformcurrent assumption still applies and, if not, what the consequences are; the second is the mutual influence of closely spaced turns. We start with the issue of current uniformity. For single and multiturn loops in air, the restriction on the circumference mentioned in the literature (e.g. [19]) applies to the total length of the wire antenna. So N2π ≤ 0.1λ, where N is the number of turns. We have seen that this restriction may be translated into a maximum error of 5% in both the real and the imaginary part of the input impedance of the loop with respect to the exact value. It may be expected, however, that the restrictions on the radiated fields can be relaxed, owing to the averaging effect of the current integrations involved. Moreover, we may recognize that in a multiturn loop antenna, in which the conductor loss may be neglected and every individual loop satisfies the circumference restriction, every loop may be regarded as carrying a uniform current. However, phase differences exist between different turns. Since the turn spacing will be very small in terms of the wavelength, applying array theory to the multiturn loop will result in effectively having N turns at the same position. Of course, this reasoning is only valid if mutual coupling between the turns may be neglected, which, in air, especially for closely packed turns, is not true [19, 33]. For multiturn loop antennas in air, one could resort to numerical methods, as explained in, for example, [34] for circular loops or, as explained in [35], approximate methods for rectangular loops. For loops immersed in blood, the situation is different. Since a current now also flows into the medium, the resistive part of the input impedance increases. Therefore, as we have already observed in section 2.4.1, the maximum allowable loop radius that justifies a uniformcurrent approach will be smaller. The mutual coupling between two widely spaced small loops immersed in blood will be negligible compared with the selfcoupling. When the loops are brought closer together, the mutual coupling increases, but up to short distances the mutual coupling is still negligible compared with the selfcoupling [33]. In this respect, the situation differs from that for loops in air. Extrapolating the results stated in [33], the mutual coupling for two small loops immersed in blood, when brought very close together, will increase to
56
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
ρ1 ρ22 ρ0
d d Figure 2.38
Geometrical parameters of an insulated loop.
such a level that it will outweigh the selfcoupling by far. This is consistent with our earlier observation that a current flows into the medium surrounding the loops. So, for barewire multiturn loop antennas contained in a small volume, the quasistatic approach will fail. Fortunately, in practice, insulated wire is used to construct intravascular antennas and the insulation does not necessarily compromise our earlier theoretical derivations. For insulated wire, the conducting medium will now act as a shield, thus reducing the mutual coupling between the wires or turns of a multiturn loop [36]. In [37] it was demonstrated that the effect of adding a thin layer of insulation to a loop antenna is that the uniform current flow is maintained when the loop antenna is immersed in a conducting medium such as blood. As long as the insulation layer is thin, the impedance of the loop is equal to that of the bare loop immersed in the conducting medium [37]. A practical value for the layer thickness d is given by [37] d ≈ 0.2(ρ2 − ρ1 ),
(2.58)
where d, ρ1 and ρ2 are defined in Figure 2.38. Figures 2.39 and 2.40 [37] show the real and imaginary parts of the admittance of a loop for two loop radii ρ0 as a function of insulation thickness. These figures show the admittance for a oneterm current approximation, Y0 , and the admittance for a twoterm current approximation, Y1 . Judging from these figures, a better estimate for the insulation thickness is given by d ≈ x(ρ2 − ρ1 ), where 0.4 ≤ x ≤ 0.5. Using the data presented in [37], we shall now look at the situation in which = 10, εr3 = 0.2εr4 and k3 ρ0 = 0.1. The relative permittivities εr3 and εr4 are those of the insulating layer and of the surrounding medium, respectively. The wave number k3 is the wave number in the insulating layer. The graphs in [37] show that in this situation, a uniformcurrent approach and an unchanged impedance apply for a highly conducting medium and the insulation thickness given in equation (2.58).
57
MR ANTENNA MODEL
1.6
Re{Y0}, k3ρ0=0.1 Re{Y1}, k3ρ0=0.05 Re{Y1}, k3ρ0=0.1 Re{Y1}, k3ρ0=0.05
1.4 1.2
G (mS)
1 0.8 0.6 0.4 0.2 0 0.1
0.2
0.3
0.7 0.6 0.4 0.5 insulation thickness d/(ρ2ρ1)
0.8
0.9
1
Figure 2.39 Real part of the loop admittance, G, versus insulation thickness. 16 Re{Y0}, k3ρ0=0.1 Re{Y1}, k3ρ0=0.05 Re{Y1}, k3ρ0=0.1 Re{Y1}, k3ρ0=0.05
14 12
B (mS)
10 8 6 4 2 0 0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
insulation thickness d/(ρ2ρ1)
Figure 2.40
Imaginary part of the loop admittance, B, versus insulation thickness.
Substituting the relative permittivity of blood for εr4 (i.e. 80) and assuming the insulating √ layer to be lossless, so that k3 = 2π εr3 /λ0 , λ0 being the freespace wavelength, yields
58
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
ρ0 = 18.65 mm for the maximum loop radius. Provided that the mutual coupling between the turns of a multiturn loop may be neglected, this value puts a limit on the total length of a multiturn loop. This means that for turns with a radius a = 0.5 mm, we may employ up to 37 turns and still assume a uniform current to flow through the wire. When a uniform current flows through a loop, we have seen that the resistive part of the impedance vanishes and is outweighed by the reactive part. Therefore we now only have to compare the selfinductance and mutual inductance between two coaxial loops as a function of the distance c between them. The two loops have the same loop radius R and wire radius a. The selfinductance L11 and mutual inductance L12 are given by [20, 38] 8R L11 = µ0 R ln −2 (2.59) a
and L12 = µ0 R
2 2 − k K(k) − E(k) , k k
where k2 = and [38, 39]
4R 2 , 4R 2 + c2
π/2
K(k ) = 2
dφ
(2.60)
(2.61)
(2.62)
1 − k 2 sin2 (φ)
is the complete elliptic integral of the first kind and π/2 E(k 2 ) = 1 − k 2 sin2 (φ) dφ
(2.63)
is the complete elliptic integral of the second kind. The ratio L12 /L11 as a function of the distance between the coaxial loops is shown in Figure 2.41 for = 10 and loops for which R = 0.5 mm and R = 1.0 mm. This figure shows that, for very small loop separations, the mutual inductance becomes comparable to the selfinductance and may not be neglected. To be able to neglect the mutual coupling effects, an orderofmagnitude difference between the mutual and the selfinductance is advisable. The loops should therefore be separated by at least the radius of the loop.
2.5
ANTENNA EVALUATION
Now that we have demonstrated the validity of the static model, we may employ this model to compare different antenna concepts quantitatively. In section 2.3, we have already conducted a qualitative comparison, based on information collected from various literature sources. As in that section, we shall again separate the antenna concepts into antennas intended for active tracking and antennas intended for imaging. We shall compare the various antenna concepts on the basis of sensitivity profiles, i.e. twodimensional sections through the threedimensional sensitivity patterns (see section 2.4.2), calculated for antennas positioned along
59
ANTENNA EVALUATION
0.6
0.5
R 0.5mm R 1.0mm
L12/L11
0.4
0.3
0.2
0.1
0 0
0.1
0.2
0.3 0.4 0.5 0.6 0.7 coaxial loop distance (mm)
0.8
0.9
1
Figure 2.41 Mutual inductance normalized to the selfinductance versus distance between coaxial loops for R = 0.5 mm, R = 1.0 mm and = 10.
the direction of the main magnetic field of the MR scanner. None of the antennas evaluated was optimized for tracking or imaging purposes. However, all antennas were dimensioned such that they may be mounted on a circular cylinder of radius R = 1 mm. The heights of the antennas were 10 mm, and all coils, whether used as the antenna or as part of the antenna, had a height of 3 mm. 2.5.1
Antennas for Active Tracking
For activetracking purposes, an antenna – mounted on a catheter – needs to be detectable with a high degree of positional accuracy. Therefore, intravascular MR antennas meant for activetracking purposes need to have a very inhom*ogeneous sensitivity pattern, with the peak values at or very near the antenna position. In Figures 2.42–2.46, we consider, respectively, the antiparallelwire antenna, the doublehelix antenna, the opposeddoublehelix antenna, the center return antenna and the perpendicularcoils antenna. Each figure shows the antenna geometry and three perpendicular sensitivity profiles. The center return antenna originally evaluated for imaging purposes, has been added to the list of antenna concepts for tracking purposes (Figure 2.45). The specifics of the antennas and the positions of the sensitivity profiles are stated in the figure captions. With the exception of the antiparallelwire antenna, all antennas demonstrate a localized sensitivity pattern. To obtain better insight into the behavior of the sensitivity as a function of the perpendicular distance from the cylindrical antenna body, Figures 2.47 and 2.48 show the
60
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
y
S (T/A)
z (mm)
z
x x (mm)
(a)
x (mm)
S (T/A)
z (mm)
S (T/A)
y (mm)
(b)
y (mm)
(c)
(d)
Figure 2.42 Antiparallelwire antenna: geometry and sensitivity patterns in xz, xy and yz planes. Height 10 mm, radius 1 mm, wire separation 2 mm, 8 segments per circumference. (a) Antenna geometry. (b) Sensitivity pattern in xz plane, y = 0.05 mm. (c) Sensitivity pattern in xy plane, z = 5 mm. (d) Sensitivity pattern in yz plane, x = 0.05 mm.
y
S (T/A)
z (mm)
z
x x (mm)
S (T/A)
z (mm)
(b)
S (T/A)
y (mm)
(a)
x (mm)
y (mm)
(c)
(d)
Figure 2.43 Doublehelix antenna: geometry and sensitivity patterns in xz, xy and yz planes. Height 10 mm, radius 1 mm, 4 turns up and 4 turns down, 8 segments per circumference. (a) Antenna geometry. (b) Sensitivity pattern in xz plane, y = 0.05 mm. (c) Sensitivity pattern in xy plane, z = 5 mm. (d) Sensitivity pattern in yz plane, x = 0.05 mm.
61
ANTENNA EVALUATION
y
S (T/A)
z (mm)
z
x x (mm)
(a)
S (T/A)
z (mm)
S (T/A)
y (mm)
(b)
x (mm)
y (mm)
(c)
(d)
Figure 2.44 Opposeddoublehelix antenna: geometry and sensitivity patterns in xz, xy and yz planes. Height 10 mm, radius 1 mm, 4 turns up and 4 turns down, 8 segments per circumference. (a) Antenna geometry. (b) Sensitivity pattern in xz plane, y = 0.05 mm. (c) Sensitivity pattern in xy plane, z = 5 mm. (d) Sensitivity pattern in yz plane, x = 0.05 mm.
y
S (T/A)
z (mm)
z
x x (mm)
S (T/A)
z (mm)
(b)
S (T/A)
y (mm)
(a)
x (mm)
y (mm)
(c)
(d)
Figure 2.45 Center return antenna: geometry and sensitivity patterns in xz, xy and yz planes. Height 10 mm, radius 1 mm, 4 wires, 8 segments per circumference. (a) Antenna geometry. (b) Sensitivity pattern in xz plane, y = 0.05 mm. (c) Sensitivity pattern in xy plane, z = 5 mm. (d) Sensitivity pattern in yz plane, x = 0.05 mm.
62
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
y
S (T/A)
z (mm)
z
x x (mm)
S (T/A)
z (mm)
(b)
S (T/A)
y (mm)
(a)
x (mm)
y (mm)
(c)
(d)
Figure 2.46 Perpendicularcoil antenna: geometry and sensitivity patterns in xz, xy and yz planes. Height 3 mm, radius of inner coil 0.8 mm, radius of outer coil 0.9 mm, 15 turns up and 15 turns down, 8 segments per circumference, coils placed at an angle of π/4 with respect to the main magneticﬁeld direction. (a) Antenna geometry. (b) Sensitivity pattern in xz plane, y = 0.05 mm. (c) Sensitivity pattern in xy plane, z = 1.5 mm. (d) Sensitivity pattern in yz plane, x = 0.05 mm.
sensitivity as a function of the distance from the antenna body at antenna half height in the planes y = 0.05 mm and x = 0.05 mm, respectively. These figures reveal that the center return antenna shows a highly localized sensitivity on the cylinder axis. The doublehelix antenna and dualopposed helix antenna show – for the chosen zaxis position – good localized sensitivity near the antenna body along the x axis, but poor sensitivity along the y axis. The perpendicularcoils antenna maintains good localized sensitivity near the antenna body along both axes. This becomes more evident from cross sections of the sensitivity patterns when the sensitivity is plotted on a logarithmic scale (Figures 2.49 and 2.50). Figures 2.49 and 2.50 also show clearly that the sensitivity pattern of the antiparallelwire antenna is not very localized. Moreover, the sensitivity pattern of this antenna depends strongly on the observation angle in the transverse (xy) plane, which makes this antenna type unsuitable for imaging purposes also. The sensitivity pattern sections shown have demonstrated that our earlier selection of antenna concepts for active tracking, supplemented with the center return antenna, was a correct one. The quantitative comparison of the antenna concepts reveals that the center return antenna is the best suited for active tracking, judging from magneticfield considerations only. If we also take manufacturing aspects into account, meaning that we have a preference for an antenna geometry that is situated on the outside of a cylindrical body only, the perpendicularcoils antenna is best suited for the job. This antenna combines a localized sensitivity that is
63
ANTENNA EVALUATION
0.02 0.018
antiparallel wires double helix opposed double helix centre return orthogonal pi/4 coils
0.016 0.014 S (T/A)
0.012 0.01 0.008 0.006 0.004 0.002 0 3
2.5
2
1.5
1
0.5
0.5
1
1.5
2
2.5
3
x (mm) Figure 2.47 Cross section of sensitivity patterns of antennas at antenna half height along the x axis.
0.02 0.018
antiparallel wires double helix opposed double helix centre return orthogonal pi/4 coils
0.016 0.014 S (T/A)
0.012 0.01 0.008 0.006 0.004 0.002 0 3
2.5
2
1.5
1
0.5
0.5
1
1.5
2
2.5
3
x (mm) Figure 2.48 Cross section of sensitivity patterns of antennas at antenna half height along the y axis.
64
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
0.1 antiparallel wires double helix opposed double helix centre return orthogonal pi/4 coils
0.01
S (T/A)
0.001
0.0001
1e05
1e06 3
2.5
2
1.5
1
0.5
0.5
1
1.5
2
2.5
3
x (mm) Figure 2.49 Cross section of sensitivity patterns of antennas on a logarithmic scale at antenna half height along the x axis.
0.1 antiparallel wires double helix opposed double helix centre return orthogonal pi/4 coils
0.01
S (T/A)
0.001
0.0001
1e05
1e06 3
2.5
2
1.5
1
0.5
0.5
1
1.5
2
2.5
3
x (mm) Figure 2.50 Cross section of sensitivity patterns of antennas on a logarithmic scale at antenna half height along the y axis.
65
ANTENNA EVALUATION
y
S (T/A)
z (mm)
z
x x (mm)
x (mm)
(c)
S (T/A)
z (mm)
(b)
S (T/A)
y (mm)
(a)
y (mm)
(d)
Figure 2.51 Singleloop antenna: geometry and sensitivity patterns in xz, xy and yz planes. Height 10 mm, radius 1 mm. (a) Antenna geometry. (b) Sensitivity pattern in xz plane, y = 0.05 mm. (c) Sensitivity pattern in xy plane, z = 5 mm. (d) Sensitivity pattern in yz plane, x = 0.05 mm.
independent of the transverse observation angle with a geometry that is restricted to the outer surface of the antenna body. 2.5.2
Antennas for Intravascular Imaging
For intravascular imaging, the antenna should show a sensitivity that extends from the antenna body to the vascular wall and is, preferably, hom*ogeneous along the direction of the antenna body. Furthermore, the sensitivity should be independent of the observation angle in the transverse plane. In Figures 2.51–2.56 we show, respectively, the geometry of the singleloop antenna, the doubleloop antenna, the tripleloop antenna, the dualopposedsolenoids antenna, the saddle coil antenna and the birdcage antenna together with, for each antenna, three perpendicular sensitivity profiles. The specifics of the antennas and sensitivity profile positions are specified in the figure captions. All of the antennas, with the exception of the dualopposedsolenoids antenna, show a hom*ogeneous sensitivity along the antenna body. It should be noted that the behavior of the single, double and tripleloop antennas is very similar. This is not directly visible from Figures 2.51–2.53, owing to the fact that the sensitivity pattern sections parallel to the x axis are taken close to, distant from and close to loops, respectively, in the structure. Although the hom*ogeneity of the sensitivity along the antenna body for the dualopposedsolenoids antenna is less than that for the other antennas, Figure 2.54 reveals that the sensitivity in the radial direction exceeds those for the other antennas.
66
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
y
S (T/A)
z (mm)
z
x x (mm)
(a)
S (T/A)
z (mm)
S (T/A)
y (mm)
(b)
x (mm)
y (mm)
(c)
(d)
Figure 2.52 Doubleloop antenna: geometry and sensitivity patterns in xz, xy and yz planes. Height 10 mm, radius 1 mm, distance between adjacent loops 1.32 mm. (a) Antenna geometry. (b) Sensitivity pattern in xz plane, y = 0.05 mm. (c) Sensitivity pattern in xy plane, z = 5 mm. (d) Sensitivity pattern in yz plane, x = 0.05 mm.
y
S (T/A)
z (mm)
z
x x (mm)
S (T/A)
z (mm)
(b)
S (T/A)
y (mm)
(a)
x (mm)
y (mm)
(c)
(d)
Figure 2.53 Tripleloop antenna: geometry and sensitivity patterns in xz, xy and yz planes. Height 10 mm, radius 1 mm, distance between adjacent loops 0.66 mm. (a) Antenna geometry. (b) Sensitivity pattern in xz plane, y = 0.05 mm. (c) Sensitivity pattern in xy plane, z = 5 mm. (d) Sensitivity pattern in yz plane, x = 0.05 mm.
67
ANTENNA EVALUATION
y
S (T/A)
z (mm)
z
x x (mm)
(a)
x (mm)
S (T/A)
z (mm)
S (T/A)
y (mm)
(b)
y (mm)
(c)
(d)
Figure 2.54 Dualopposedsolenoids antenna: geometry and sensitivity patterns in xz, xy and yz planes. Height per coil 3 mm, gap between coils 3 mm, 15 turns per coil, radius 1 mm, 8 segments per circumference. (a) Antenna geometry. (b) Sensitivity pattern in xz plane, y = 0.05 mm. (c) Sensitivity pattern in xy plane, z = 4.5 mm. (d) Sensitivity pattern in yz plane, x = 0.05 mm.
y
S (T/A)
z (mm)
z
x x (mm)
S (T/A)
z (mm)
(b)
S (T/A)
z (mm)
(a)
y (mm)
y (mm)
(c)
(d)
Figure 2.55 Saddle coil antenna: geometry and sensitivity patterns in xz, xy and yz planes. Height 10 mm, radius 1 mm, distance between the two parts 0.8 mm, 8 segments per circumference. (a) Antenna geometry. (b) Sensitivity pattern in xz plane, y = 0.05 mm. (c) Sensitivity pattern in xy plane, z = 5 mm. (d) Sensitivity pattern in yz plane, x = 0.05 mm.
68
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
y
S (T/A)
z (mm)
z
x x (mm)
S (T/A)
z (mm)
(b)
S (T/A)
y (mm)
(a)
x (mm)
y (mm)
(c)
(d)
Figure 2.56 Birdcage antenna: geometry and sensitivity patterns in xz, xy and yz planes. Height 10 mm, radius 1 mm, 4 wires. (a) Antenna geometry. (b) Sensitivity pattern in xz plane, y = 0.05 mm. (c) Sensitivity pattern in xy plane, z = 5 mm. (d) Sensitivity pattern in yz plane, x = 0.05 mm.
For a better comparison of the various antenna concepts, we shall look again at the behavior of the sensitivity as a function of the perpendicular distance from the cylindrical antenna body. Figures 2.57 and 2.58 show the sensitivity as a function of the distance from the antenna body at antenna half height in the planes y = 0.05 mm and x = 0.05 mm, respectively. For the distances of interest, i.e. where we may expect the vessel wall (2–3 mm for the large arteries [18]), we see that the best antenna, judging from the magnetic field only, is the dualopposedsolenoids antenna, followed by the tripleloop antenna, the saddle coil antenna and the doubleloop antenna. The behavior of the last two antennas is nearly identical for the number of wires and loops chosen. To complete the quantitative comparison of the antenna concepts, we have to look at the hom*ogeneity of the sensitivity in the transverse plane. To that end, we have calculated the sensitivity at half the height of the antenna body as a function of the observation angle. The results are shown in Figures 2.59 and 2.60 for a radial distance from the antenna body axis of, respectively, 2 mm and 4 mm. The figures show that the dualopposedsolenoids antenna demonstrates the highest sensitivity levels, but that the tripleloop antenna outperforms the dualopposedsolenoids antenna with respect to sensitivity hom*ogeneity at distances closer to the antenna body. Further away from the antenna body, the sensitivity becomes comparable for all antennas. The manufacturing of both the dualopposedsolenoid antenna and the tripleloop antenna is expected to be equal in complexity. Therefore, both antenna types are regarded as suitable for imaging purposes.
69
ANTENNA EVALUATION
0.003 single loop double loop triple loop dual opposed solenoids saddle coil birdcage
0.0025
S (T/A)
0.002
0.0015
0.001
0.0005
0 3
2.5
2
1.5
1
0.5
0.5
1
1.5
2
2.5
3
x (mm) Figure 2.57 x axis.
Cross section of antenna sensitivity patterns at antenna half height along the
0.003 single loop double loop triple loop dual opposed solenoids saddle coil birdcage
0.0025
S (T/A)
0.002
0.0015
0.001
0.0005
0 3
2.5
2
1.5
1
0.5
0.5
1
1.5
2
2.5
3
y (mm) Figure 2.58 axis.
Cross section of antenna sensitivity patterns at antenna half height along the y
70
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
0.0004
0.00035
0.0003
S (T/A)
0.00025
0.0002
0.00015
0.0001 dual opposed solenoids double loop triple loop saddle coil
5e05
0 180 150 120 90 60 30
30
60
90
120 150 180
ϕ (degrees) Figure 2.59 Antenna sensitivity in the halfheight plane as a function of the observation angle at distance R = 2 mm from the antenna body axis.
0.0004
0.00035
0.0003
S (T/A)
0.00025
dual opposed solenoids double loop saddle coil triple loop
0.0002
0.00015
0.0001
5e05
0 180 150 120 90 60 30 0 ϕ (degrees)
30
60
90
120 150 180
Figure 2.60 Antenna sensitivity in the halfheight plane as a function of the observation angle at distance R = 4 mm from the antenna body axis.
71
ANTENNA EVALUATION
2.5.3
Antenna Rotation
Although the human vascular system is mainly ‘headtotoe’ directed, parts of the system will have different directions. The MR main magnetic field is also ‘headtotoe’ directed. For the active tracking of catheters through arteries that are not lined up with the main magnetic field or for imaging the walls of these arteries, it is important to compare the different antenna concepts with respect to antenna rotation. For all simulations described thus far, the antenna body directed along the direction of the main magnetic field and the sensitivity was calculated for a magneticfield distortion transverse with respect to the antenna body direction. To analyze the effects of antenna rotation, we make use of the rotation transformations [Rx (α)] for a rotation angle α around the x axis, [Ry (β)] for a rotation angle β around the y axis and [Rz (γ )] for a rotation angle γ around the z axis [40]:
1 0 0 [Rx (α)] = 0 cos(α) sin(α) , 0 − sin(α) cos(α) cos(β) 0 − sin(β) , 1 0 [Ry (β)] = 0 sin(β) 0 cos(β) cos(γ ) sin(γ ) 0 [Rz (γ )] = − sin(γ ) cos(γ ) 0 . 0 0 1
(2.64)
(2.65)
(2.66)
For an antenna rotated by the angles α, β and γ , we find the magneticfluxdensity components (Bx , By , Bz ) from the ‘unrotated’ components (Bx , By , Bz ) using Bx Rxx By = Ryx Rzx Bz
Rxy Ryy Rzy
Rxz Bx Ryz By , Rzz Bz
(2.67)
where the transformation matrix is obtained by multiplication of [Rx (α)], [Ry (β)] and [Rz (γ )]. This is equivalent to first rotating by γ , then rotating by β and finally rotating by α.16 The matrix elements Rij , i, j = x, y, z, are then found to be Rxx = cos(β) cos(γ ), Rxy = cos(β) sin(γ ),
(2.68) (2.69)
Rxz = − sin(β),
(2.70)
Ryx = − cos(α) sin(γ ) + sin(α) sin(β) cos(γ ),
(2.71)
16 These rotation operations are commutative and associative.
72
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
Ryy = cos(α) cos(γ ) + sin(α) sin(β) sin(γ ),
(2.72)
Ryz = sin(α) cos(β), Rzx = sin(α) sin(γ ) + cos(α) sin(β) cos(γ ),
(2.73) (2.74)
Rzy = − sin(α) cos(γ ) + cos(α) sin(β) sin(γ ), Rzz = cos(α) cos(β).
(2.75) (2.76)
The sensitivity of the rotated antenna is given by 1 2 2 S= Bx + By . I
(2.77)
To demonstrate the effects of antenna rotation, we shall rotate the antennas that we found best for tracking and the ones that we found best for imaging around the x axis. The sensitivity, expressed in terms of the unrotated magneticfluxdensity components, is then given by 1 S= Bx2 + (By cos(α) + Bz sin(α))2 , (2.78) I where α is the rotation angle. 2.5.3.1 Rotation of Antennas for Active Tracking In section 2.5.1, we found that the center return antenna and the perpendicularcoils antenna were best suited for activetracking purposes. In Figures 2.61 and 2.62, we show the sensitivity patterns of these antennas in the xy plane at half height for rotation angles of 0◦ (no rotation), 45◦ , 60◦ and 90◦ . The scaling of the sensitivity in both figures has been adjusted to maximize the visibility of the effects of rotation on the sensitivity pattern sections. These figures show that for both antennas, up to large angles, the sensitivity patterns remain hom*ogeneous. The perpendicularcoils antenna maintains a hom*ogeneous pattern even up to 90◦ rotation. This behavior, added to the ease of manufacturing, makes this antenna concept stand out for activetracking purposes.17 2.5.3.2 Rotation of Antennas for Imaging In section 2.5.2, we found that the tripleloop antenna and the dualopposedsolenoids antenna were the most promising for imaging purposes. In Figures 2.63 and 2.64, we show the sensitivity patterns of these antennas in the xy plane at half height for rotation angles of 0◦ (no rotation), 45◦ , 60◦ and 90◦ . The scaling of the sensitivity has again been adjusted to maximize the visibility of the effects of rotation on the sensitivity pattern sections. These figures show that for angles from 45◦ upwards, the radial sensitivity rapidly loses hom*ogeneity. For a rotation angle of 90◦ , the radial sensitivity of the tripleloop antenna has formed four distinct lobes and that of the dualopposedsolenoids antenna has assumed the form of a twolobe pattern. The increase in radial inhom*ogeneity with rotation angle seems to be more severe for the tripleloop antenna. Both antennas appear to be suitable for imaging up to rotation angles of 45◦ . 17 This antenna is positioned completely on the outside of the antenna body, as opposed to the center return antenna, which has an additional wire segment passing through the axis of the cylindrical antenna body.
73
S (T/A)
y (mm)
y (mm)
S (T/A)
ANTENNA EVALUATION
(b)
y (mm)
y (mm)
S (T/A)
x (mm)
(a)
S (T/A)
x (mm)
x (mm) (c)
x (mm) (d)
S (T/A)
y (mm)
y (mm)
S (T/A)
Figure 2.61 Center return antenna: sensitivity in the xy plane at half height for diﬀerent rotation angles. (a) α = 0◦ , (b) α = 45◦ , (c) α = 60◦ , (d) α = 90◦ .
(b)
y (mm)
y (mm)
S (T/A)
x (mm)
(a)
S (T/A)
x (mm)
x (mm)
x (mm)
(c)
(d)
Figure 2.62 Perpendicularcoils antenna: sensitivity in the xy plane at half height for diﬀerent rotation angles. (a) α = 0◦ , (b) α = 45◦ , (c) α = 60◦ , (d) α = 90◦ .
74
S (T/A)
y (mm)
y (mm)
S (T/A)
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
(b)
y (mm)
y (mm)
S (T/A)
x (mm)
(a)
S (T/A)
x (mm)
x (mm)
x (mm)
(c)
(d)
S (T/A)
y (mm)
y (mm)
S (T/A)
Figure 2.63 Tripleloop antenna: sensitivity in the xy plane at half height for diﬀerent rotation angles. (a) α = 0◦ , (b) α = 45◦ , (c) α = 60◦ , (d) α = 90◦ .
(a)
(b)
y (mm)
y (mm)
S (T/A)
x (mm)
S (T/A)
x (mm)
x (mm)
x (mm)
(c)
(d)
Figure 2.64 Dualopposedsolenoids antenna: sensitivity in the xy plane at half height for diﬀerent rotation angles. (a) α = 0◦ , (b) α = 45◦ , (c) α = 60◦ , (d) α = 90◦ .
75
IN VITRO TESTING
C2 R C1
50Ω@64MHz
L
Figure 2.65 Intravascular coil, depicted as a series circuit of an inductor and a resistor with tuning and impedancematching circuit.
2.6
IN VITRO TESTING
The final validation of the analytical model developed was delivered by comparing the calculated sensitivity patterns for the different antennas with images created by an MR system where the intravascular antennas were used for active tracking. To ensure maximum power transfer between an intravascular antenna and the external circuitry, the antenna needs to be resonant at 64 MHz and impedancematched to the transmission line that connects the antenna to the external circuitry. 2.6.1
Sensitivity Pattern
A number of antennas were constructed. The complex input impedances of these antennas were measured while the antennas were immersed in tap water, which served for this purpose as a bloodmimicking fluid.18 Next, a parallel and a series capacitor were added (Figure 2.65) to accomplish tuning at 64 MHz and impedance matching to a 50 coaxial transmission line. Owing to inaccuracies in the measurements, the unavailability of a welldefined bloodmimicking fluid, the availability of only a limited set of discretevalued capacitors and, most of all, the fact that the antennas were handmade, poorly reproducible products (Figure 2.66), the tuning and matching was not optimal for most antennas. Therefore the SNR realized for the antennas left room for improvement. The first set of prototype antennas were constructed with materials available at that moment and therefore did not have sizes suitable for a clinical application. A center return antenna consisting of four wires, equally spaced around the circumference of a cylindrical body, was constructed (Figure 2.66). The radius of the antenna body was 4 mm, and the height of the antenna was 16 mm. A tripleloop antenna was made on a cylindrical body 18 As we did not have the possibility to create a saline solution at the time of measurement, the use of tap water (σ ≈ 0.01 S m−1 [41]) was preferred over the use of distilled water (σ ≈ 0.0001 S m−1 [41]).
76
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
Figure 2.66 Realized intravascularantenna prototypes. Reproduced by permission on Nicole Op en Camp.
with a radius of 2.5 mm and height of 16 mm. The separation between two adjacent coils was 2 mm. Finally, an opposedsolenoids antenna was constructed (see Figure 2.66 again), consisting of two coils with a height of 3 mm and nine turns, separated by 3 mm and having a total height of 16 mm and a radius of 4 mm. In Figures 2.67–2.69, we show, for, the center return antenna, the tripleloop antenna and the dualopposedsolenoids antenna, respectively, the calculated sensitivity patterns and the MR images obtained for various antenna rotation angles [9]. Although the MR images are directly related to the sensitivity, the exact values have been lost in the signal processing. The calculated sensitivity profiles have been scaled to achieve a visual match with the MR images. The measurements were performed with the antennas in the setting of a phantom made of Perspex. The surrounding medium was a bloodmimicking fluid created by dissolving 2 mg of manganese chloride (MnCl2 ) per liter of water [42]. By choosing a thicker slice for the measurement, a higher SNR may be achieved. The slice thickness was chosen per measurement. The slice thickness for the 0◦ rotation angle in the measurement of the center return antenna was 7 mm. The thicknesses for the 45◦ and 90◦ rotation angles were 15 mm. The slice thicknesses for all measurements of the tripleloop antenna were 7 mm. The slice thickness for the 0◦ rotation angle in the measurement of the dualopposedsolenoids antenna was 30 mm. The thicknesses for the 45◦ and 90◦ rotation angles were 7 mm. To demonstrate the influence of the slice thickness, the MR image for a 0◦ rotation angle and the dualopposedsolenoids antenna is shown in Figure 2.70 for two slice thicknesses [9]. Taking the inaccuracies mentioned earlier into account and noting that in the construction of the center return antenna, the lumen of the antenna was filled with a contrast agent and that in the construction of the tripleloop antenna and the dualopposedsolenoids antenna, the lumen of the antenna was filled with a silicone gel [9], not accounted for by the analytical model, the calculations and measurements show good agreement. So, again, the validity of
77
S (T/A)
y (mm)
IN VITRO TESTING
x (mm)
S (T/A)
y (mm)
(a)
x (mm)
S (T/A)
y (mm)
(b)
x (mm)
(c) Figure 2.67 Calculated sensitivity proﬁles (left) and MR images (right) for the prototype center return antenna. (a) Rotation angle 0◦ . (b) Rotation angle 45◦ . (c) Rotation angle 90◦ .
the model developed has been proved. This leaves us with the task of investigating whether tracking works in practice. 2.6.2
Tracking
For the tracking experiments, a new prototype antenna was constructed based on the perpendicularcoils antenna. The antenna consisted of only one coil, of height 3 mm, at 45◦ from the antenna body axis, consisting of 15 turns wound around the tip of a 1.67 mm diameter catheter with 0.09 mm diameter insulated copper wire [12]. The reason for having only a single coil instead of two was the ease of realizing this onecoil antenna. The copper
78
S (T/A)
y (mm)
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
x (mm)
S (T/A)
y (mm)
(a)
x (mm)
S (T/A)
y (mm)
(b)
x (mm)
(c) Figure 2.68 Calculated sensitivity proﬁles (left) and MR images (right) for the prototype tripleloop antenna. (a) Rotation angle 0◦ . (b) Rotation angle 45◦ . (c) Rotation angle 90◦ .
wire leads were twisted over the length of the catheter, to minimize their influence on the magnetic field, and were connected to a coaxial transmission line at the end of the catheter [12]. With the catheter immersed in a phantom filled with a bloodmimicking fluid and the lumen of the catheter also filled with the same fluid, interactive MR scans with active tracking were performed for various rotation angles of the antenna. Measurements were taken over a period of one minute for every rotation angle. The measured antenna positions (indicated with dots) are shown in Figure 2.71 for this ‘45◦ coil antenna’ at the pixel level, where the squares in the figure represent the pixels of the underlying MR image [12]. The pixel size is 1.3722 mm × 1.3722 mm.
79
S (T/A)
y (mm)
IN VITRO TESTING
x (mm)
S (T/A)
y (mm)
(a)
x (mm)
S (T/A)
y (mm)
(b)
x (mm)
(c) Figure 2.69 Calculated sensitivity proﬁles (left) and MR images (right) for the prototype dualopposedsolenoids antenna. (a) Rotation angle 0◦ . (b) Rotation angle 45◦ . (c) Rotation angle 90◦ .
For reference, the same exercise was repeated with an ‘ordinary coil antenna’, the results of which are shown in Figure 2.72. The figures indicate the superiority of the 45◦ coil antenna over the ordinary coil antenna. Even better results may be expected from employing a perpendicularcoils antenna, since the sensitivity pattern of this antenna will be more concentrated, as explained in section 2.3.1. Finally, the 45◦ coil antenna was inserted into a humanabdomen vascular phantom. The catheter carrying the antenna was inserted via a guide wire into the phantom and then guided through the vessels. The catheter tip positions were measured during this movement, and snapshots of this process are shown in Figure 2.73 [12]. The catheter tip position is indicated
80
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
(a)
(b)
Figure 2.70 MR images for nonrotated prototype dualopposedsolenoids antenna. (a) Slice thickness 7 mm. (b) Slice thickness 30 mm.
by a white ‘+’ mark. The arrows in the figure have been added for clarity. Apart from the single horizontal error in Figure 2.73(b), the active tracking works well.
2.7
ANTENNA SYNTHESIS
With the availability of an analytical model that – when implemented in software – generates reliable results in a very short time, the possibility has been created to generate or synthesize antenna designs automatically within a reasonable time frame. With an optimization procedure that relies on function evaluations only, a design may be realized, subject to userdefined mechanical and electromagnetical constraints, within a few minutes on standard office computing equipment. Two examples of optimization procedures that need function evaluations only are simulated annealing [29] and genetic algorithms [43, 44]. Here we have opted specifically for the latter method, owing to its ‘natural’ appeal and its ease of software implementation. 2.7.1
GeneticAlgorithm Optimization
Genetic algorithms (GAs) are optimization methods based upon the principles of natural selection and evolution. The concepts used in the optimization process are genes, chromosomes, generations, populations, parents, children and fitness. A gene is a coded version of one of the parameters of the problem. A possible coding is a binary coding, making the gene a string of zeros and ones. A chromosome is a series of genes and is thus a solution of the problem. A gene is also known as an ‘individual’. A population is a set of individuals. A generation is a population iteratively formed from the previous one. A parent is an individual from the previous generation, and a child is an individual from the current generation. The fitness is a number assigned to an individual and is a measure of ‘how good’ this individual is.
ANTENNA SYNTHESIS
81
Figure 2.71 Position measurements made with a ‘45◦ coil antenna’. (a) Rotation angle 0◦ . (b) Rotation angle 30◦ . (c) Rotation angle 45◦ . (d) Rotation angle 60◦ . (e) Rotation angle 90◦ .
In a typical GA optimization problem, a starting population is created randomly, or intelligently if the general direction of the solution is known. In our intravascularantenna problem, we want to find the number of coils, the number of turns per coil, the coil heights and the turn directions that give the highest sensitivity at the antenna body surface or at a distance where we may expect the artery wall to be present. The population thus consists of sets of numbers, heights and directions. A fitness is assigned to every individual from this population. Here, this fitness could be the amplitude of the sensitivity parameter. Next, parents are selected from the population (several different selection processes exist) and, by means of crossover and mutation, children of a new generation are created. In the crossover process, the parameters of two antenna configurations are intermixed. In the mutation process, one or
82
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
Figure 2.72 Position measurement made with an ‘ordinary coil antenna’. (a) Rotation angle 0◦ . (b) Rotation angle 30◦ . (c) Rotation angle 45◦ . (d) Rotation angle 60◦ . (e) Rotation angle 90◦ .
a few of the parameters change randomly. The process is depicted in Figure 2.74 and, more specifically, in Figure 2.75 specifically for a fiveparameter problem [43, 44]. As an example, we shall look at an intravascular antenna consisting of a discrete number of coils wound around a cylindrical antenna body. The maximum number of coils was three, the height of every coil was allowed to vary between 0.1 mm and 4 mm, the gap between two adjacent coils could vary between 0.1 mm and 3 mm, the number of turns per coil could vary between one and 15, and every coil could be wound clockwise or counterclockwise. The radius of the antenna was 1 mm and one circumference was approximated by 12 straightline segments. We generated designs for tracking and for imaging.
83
ANTENNA SYNTHESIS
(a) (a)
(b) (b)
(c) (c)
(d) (d)
(e) (e)
(f) (f)
Figure 2.73 MR images obtained during active tracking using the 45◦ coil antenna in a humanabdomen phantom.
First, we generated an antenna for tracking purposes. As the fitness parameter, we used the minimum value of the sensitivity parameter, sampled on the axis of the antenna between 3 mm and 6 mm in height. The optimization process is a maximisation process, and so, by selecting the fitness parameter in this way, we demanded a high sensitivity on the axis between 3 and 6 mm measured from the base of the antenna. The optimization process generated (within a few minutes) a design consisting of three coils. The first coil started at 2.64 mm from the antenna base. The gaps between the first and second and between the second and third coil were, respectively, 2.91 mm and 2.68 mm. The heights of the coils, from bottom to top, were 3.76 mm, 3.75 mm and 3.13 mm, and the
84
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
old generation
old generation
selection mating
children new generation
Figure 2.74
Geneticalgorithm process.
parents 1 2 3 4 5
1 2 3 4 5
chromosomes reproduction children
1 2 3 4 5
mutation
1 2 3 4 5
crossover
Figure 2.75 Geneticalgorithm iteration for a ﬁveparameter problem.
numbers of turns were 13, 5 and 1. The first and third coils were wound counterclockwise, and the middle coil was wound clockwise. Sensitivity profiles in the xz, yz and xy planes were calculated and are shown in Figure 2.76. This figure shows that, between 3 mm and 6 mm from the antenna base, an increased sensitivity is indeed present on the axis of the antenna body. Of course, the sensitivity is still below the sensitivity obtained at the surface of the antenna body. So the figure shows that the optimization procedure works, but also that care must be taken in formulating the fitness parameter. Alternatively, for a tracking antenna, one could aim at an increased sensitivity on the antenna body surface only. As a second example, we generated an antenna for imaging purposes. As the fitness parameter, we used the minimum value of the sensitivity parameter, sampled over the outer surface of a cylinder with a radius of 2.5 mm encapsulating the antenna, between 3 mm and 6 mm in height. The optimization process generated (within a few minutes) a design consisting of three coils. The first coil started at 1.73 mm from the antenna base. The gaps between the first
85
S (T/A)
z (mm)
ANTENNA SYNTHESIS
x (mm)
S (T/A)
z (mm)
(a)
y (mm)
S (T/A)
y (mm)
(b)
x (mm)
(c) Figure 2.76 Sensitivity patterns in xz, xy and yz planes for a multiplecoil antenna optimized for tracking. (a) Sensitivity pattern in xz plane, y = 0.05 mm. (b) Sensitivity pattern in yz plane, x = 0.05 mm. (c) Sensitivity pattern in xy plane, z = 4.5 mm.
and second and between the second and third coil were, respectively, 0.28 mm and 0.19 mm. The heights of the coils, from bottom to top, were 0.41 mm, 0.12 mm and 3.20 mm, and the numbers of turns were 12, 8 and 11. All coils were wound counterclockwise. Sensitivity profiles in the xz, yz and xy planes were calculated and are shown in Figure 2.77. A strong, rotationally hom*ogeneous sensitivity is visible in the radial and axial range specified. The values of the heights and gap widths, however, show that a more accurate and reproducible construction method is needed than the one that has been used up to now. The handwork used for the construction of these prototype antennas thus far will no longer be
86
S (T/A)
z (mm)
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
x (mm)
S (T/A)
z (mm)
(a)
y (mm)
S (T/A)
y (mm)
(b)
x (mm)
(c) Figure 2.77 Sensitivity patterns in xz, xy and yz planes for a multiplecoil antenna optimized for imaging. (a) Sensitivity pattern in xz plane, y = 0.05 mm. (b) Sensitivity pattern in yz plane, x = 0.05 mm. (c) Sensitivity pattern in xy plane, z = 4.5 mm.
sufficient. The construction accuracy needs to be improved up to the level of the modeling accuracy. For the present, there is no reason to improve the modeling accuracy. The tracking antennas perform as expected, and the imaging antennas first need to undergo tests in an MR environment.
2.8
SAFETY ASPECTS
Thus far, we have been looking at intravascular antennas from a modeling or a constructional point of view, disregarding safety aspects. Since the ultimate goal is to employ these antennas
SAFETY ASPECTS
87
in living persons, we need to address these aspects as well. We have briefly mentioned the risk of heating of the antenna leads. For a complete treatment, however, we need to address all intrinsic potential sources of hazard in an MR environment [12, 45]: static magnetic fields and spatial gradients, pulsed gradient magnetic fields and, finally, pulsed RF fields and the associated heating problem. For a properly operating MR system, the hazards associated with direct interactions of these fields with the body are negligible. It is the interaction of these fields with medical devices placed within them that create concerns for safety [45]. Before we discuss the potential sources of hazard, we first need to define what we mean by the term ‘safe’. According to [45], the term ‘MR safe’ indicates that the device, when used in an MR environment, has been demonstrated to present no additional risk to the patient, but it may affect the quality of the diagnostic information. Closely connected with the definition of ‘MR safe’ is the definition of ‘MR compatible’. The term ‘MR compatible’ indicates that a device, when used in an MR environment, is ‘MR safe’ and has also been demonstrated neither to significantly affect the quality of the diagnostic information nor to have its operations affected by the MR device. Understanding now what is meant by safety, we may proceed with the potential sources of hazard. 2.8.1
Static Magnetic Fields and Spatial Gradients
A static magnetic field in the range of 0.2 T to 2.0 T, and possibly extending to 4 T or 5 T, is always present in an MR scanner, even when the scanner is not imaging [45]. This strong magnetic field decreases rapidly, on moving away from the magnet, producing a large spatial gradient. This large gradient may cause magnetizable objects to be accelerated, thus possibly causing injuries to patients and/or medical staff.19 In addition to the potential hazard of acceleration of magnetizable objects outside the patient, magnetizable objects inside the patient may undergo torque and displacement forces when brought into the MR main magnetic field, possibly resulting in the tearing of soft tissue.20 Furthermore, certain cardiac pacemakers are known to function erratically even in relatively weak magnetic fields.21
19 A pair of scissors was pulled out of a nurse’s hand as she entered a magnet room. The scissors hit a patient, causing
a cut on the patients head (8/2/93). A patient was struck by an oxygen bottle while being placed in a magnet bore. The patient received injuries requiring sutures (6/2/91). Two steel tines (parts of a fork lift truck) weighing 80 pounds each were accelerated by a magnet, striking a technician and knocking him a distance of over 15 feet, resulting in serious injury (6/5/86) [45]. 20 A patient with an implanted intracranial aneurysm clip died as a result of an attempt to scan her. The clip reportedly shifted when exposed to the magnetic field. The staff had apparently obtained information indicating that the material in this clip could be scanned safely (11/11/92). Dislodgement of an iron filing in a patient’s eye during MR imaging resulted in vision loss in that eye (1/8/85). A patient complained of double vision after an MR examination. The MR examination, as well as an Xray, revealed the presence of metal near the patient’s eye. The patient was sedated at the time of the examination and was not able to inform anyone of this condition (12/15/93) [45]. 21 A patient with an implanted cardiac pacemaker died during an MR examination (12/2/92). A patient with an implanted cardiac pacemaker died during or shortly after an MR examination. The coroner determined that the death was due to interruption of the pacemaker by the MR system (9/18/89) [45].
88 2.8.2
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
Pulsed Gradient Magnetic Fields
A pulsed gradient magnetic field is used for signal localization. During the rise time of the magnetic field, currents are induced in electrical conductors. In most MRI systems, the amplitudes of these currents, however, are about three orders of magnitude smaller than those induced by the pulsed RF field [45]. Therefore, thermal injuries due to pulsed gradient magnetic fields are not of great concern. More important are the biological effects due to pulsed gradient magnetic fields. One of these effects is the electrical stimulation of nerves and the generation of light flashes (magnetophosphenes), which may result from a slight torque exerted on the retinal cones [45]. Current limits on ∂B/∂t prevent painful peripheralnerve stimulation. 2.8.3
Pulsed RF Fields and Heating
Concerning pulsed RF fields, one needs to be aware of the production of heat in tissue and the production of heat by electrical currents induced in metal implants and medical equipment. The rate at which RF energy is deposited in tissue is measured by the specific absorption rate (SAR). The SAR is measured in watts per kilogram and is limited for wholebody exposure to avoid heating problems22 [45]. As we have already mentioned, one needs to be aware of the length of electrical leads. If this length is equal to or greater than half a wavelength (in the surrounding medium), standing (current) waves may be induced in the leads. Radiation will take place at the tips of the leads, causing an increase in temperature by dissipation in the surrounding medium, which may become harmful for the patient23 [12, 45]. In all intravascularantenna designs presented thus far, we have not paid attention to the length of the electrical leads. Our main concern was the development and validation of an antenna model. However, practical solutions for the problem of leads becoming too long have been reported in the open literature [46, 47]. To avoid the leads becoming resonant, quarterwavelength chokes or traps may be inserted into the cable. The drawback of these countermeasures is that the chokes or traps need to be designed for the correct resonance frequency, and they may give rise to local energy dissipation [48]. A better solution to the heating problem seems to be to divide the cable into sections that are too short to become resonant. This technique was employed in [48], where compact, inductive transformers were used to interconnect the cable sections, which ensured that there was a signal path without the risk of the electrical leads becoming resonant.
22 A patient received small blistered burns to the left thumb and left thigh. Reportedly, the operator input an inaccurate
patient weight, resulting in an incorrect SAR value (2/10/93) [45]. 23 An electrically conductive lead was looped and placed against bare skin, causing a burn on the patient’s upper
arm (5/19/95). A child received a burn to the right hand from an ECG cable while the patient was anesthetized. A skin graft was required to treat the affected area (1/26/95). A patient received a 1.5 inch × 4 inch blistered burn to the left side of the back near the pelvis from an ECG gating cable (9/23/91). A patient received blistered burns on a finger where a pulse oximeter was attached during MR scanning. A skin graft was required to treat the affected area (2/27/95) [45].
CONCLUSIONS
2.9
89
CONCLUSIONS
The formation of MR images is accomplished by trading off SNR, imaging speed and spatial resolution. For temporally efficient MRI, local receiver coils are being developed to improve the SNR without compromising imaging speed and spatial resolution. During intravascular interventions, passive methods may be employed to visualize catheter positions and orientations. However, these passive methods suffer from a severe time inefficiency, limiting their feasibility for intravascular, interventional MRI purposes. The visualization of catheter position and orientation is therefore expected to be accomplished best by employing active, intravascular devices (i.e. antennas). Taking the concept of local receiver coils one step further, intravascular imaging is expected to be feasible too, by employing intravascular receiver coils or antennas that will improve the SNR beyond levels feasible by employing local receiver coils outside the body. Although various intravascularantenna concepts have been described in the literature and have been evaluated by means of MR imaging, a quantitative comparison of the various concepts has not been conducted until recently [49]. For the purpose of such a quantitative comparison, a fast approximate model, based on the static magnetic field induced by a direct current in a straight wire segment, has been employed. This model was originally developed for the design of surface coils, for which it is now regarded as unsuitable, since the magnetic field at the positions of interest is not expected to behave as a static magnetic field, nor is the current in a surface coil expected to behave as a direct current. For intravascular antennas, though, positions in or near the artery wall are expected to be in the radiating near field of the antenna, where the fields are locally inversely proportional to the square of the distance. The static magnetic field induced by a direct current is also inversely proportional to the square of the distance. The current in a small intravascular antenna is expected to be well approximated by a uniform current, and therefore an approximation of the dynamic fields by static ones in and near the artery wall should yield acceptable results. To assess the validity of a static model, comparisons were made between the static model and the smallloop approximation for a loop antenna immersed in blood. Before this assessment was performed, the smallloop uniform current approximation was validated. It turns out that we may consider a bare loop antenna, immersed in blood and subject to a 1.5 T main MR magnetic field, to carry a uniform current for radii up to 1.7 mm. Having thus put a practical limit on the radius of our reference, we have verified the static model. For several different loop orientations, we compared the ‘static’ sensitivity with the ‘dynamic’ sensitivity, where the sensitivity S is defined by S = (1/I ) Bx2 + By2 . On the axis of the loop, the dynamic sensitivity is approximated to within 13% for small loop antennas (radius 0.5 mm) within the region of interest, i.e. a circular cylinder with a radius between 2 mm and 3 mm. This cylinder corresponds to a large artery. Moreover, the behavior of the sensitivity as a function of the distance from the loop center is similar for both the static and the dynamic model, which means that the static model may be employed for comparison of loop antenna designs. For bare wire antennas larger than a single loop and contained in a small volume, the static approach will fail owing to coupling effects. However, when a thin insulation layer is used a uniform current is maintained when the antenna is immersed in a highly conducting medium
90
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
such as blood. For multiturn loop antennas where the turns are not too closely spaced, up to about 35 turns with a radius of the order of 0.5 mm may be employed without compromising the model. A comparison of the sensitivity patterns of a number of intravascular antennas described in literature for tracking and imaging purposes, not keeping too strictly to the limits defined earlier, confirms the results obtained from our qualitative comparisons. The center return antenna is best suited for active tracking, judging from magneticfield considerations only. If we also take manufacturing aspects into account, the perpendicularcoils antenna may be better suited for the job, combining a localized sensitivity that is independent of the transverse angle of observation with a geometry that is restricted to the outer surface of the antenna body. Furthermore, the perpendicularcoils antenna performs better when it is rotated with respect to the MR main magnetic field. For imaging purposes, both the dualopposedsolenoids antenna and the tripleloop antenna are considered favorites. They exhibit comparable sensitivity profiles and the manufacturing of both antennas is expected to be equally complex. Neither antenna should be used for rotation angles with respect to the MR main magnetic field in excess of 45◦ . The calculated sensitivity profiles compare well with images created with an MR system for a number of realized prototype antennas, even though the geometrical limits were not observed too strictly and the use of contrast fluid in the antenna body lumen was not taken into account in the model. Having established the availability of a fast analytical model that is of practical use in the analysis of intravascular antennas, we have incorporated the model into a geneticalgorithm optimization environment. It has been demonstrated that antenna designs may be generated – subject to userdefined mechanical and electromagnetic constraints – within minutes, employing standard office computing equipment. For the realization of the antenna designs thus generated, more precise manufacturing methods are required than the handwork used for the construction of the prototype antennas so far. A preliminary investigation of creating copper strip patterns on a cylindrical dielectric body, by applying laser patterning, reveals that precise manufacturing is feasible. To prevent heating of the antenna leads, dissecting the transmission line connecting the antenna to the MR hardware into sections that are too short to become resonant at the Larmor frequency is recommended. A technique involving inductive coupling from transmission line section to transmission line section, as described in [48], could be employed for transferring signals between the antenna and the MR hardware.
APPENDIX 2.A. BIOT–SAVART LAW FOR QUASISTATIC SITUATION To derive the Biot–Savart law for the quasistatic situation, we start with Maxwell’s equations for a hom*ogeneous, lossless, isotropic medium, ∇ × E = −jωµH, ∇ × H = J + jωεE,
(2.A.1) (2.A.2)
91
APPENDIX 2.A. BIOT–SAVART LAW FOR QUASISTATIC SITUATION
where E is the electric field, H is the magnetic field, J is the current density, ω is the angular frequency, µ is the permeability of the medium and ε is the permittivity of the medium. Next, we need the continuity equation, which is given by ∇ · J = −jωρ,
(2.A.3)
where ρ is the charge density, and Gauss’s laws, which are given by ∇ · H = 0, ρ ∇·E= . ε
(2.A.4) (2.A.5)
Now, we assume that jωεE is negligible compared with J, so that equation (2.A.2) may be approximated by ∇ × H = J. (2.A.6) Since ∇ · ∇ × H = 0, equation (2.A.6) results in ∇ · J = 0, and this, when substituted in equation (2.A.3), means that ρ = 0 in equation (2.A.3). This is known as the quasistatic approach, where J is assumed to be ‘almost stationary’. A static charge in equation (2.A.5) remains possible. Next, we introduce the magnetic vector potential A through H = ∇ × A.
(2.A.7)
∇ × ∇ × A = ∇(∇ · A) − ∇ 2 A = J.
(2.A.8)
Then, from equation (2.A.6),
Assuming the Coulomb gauge ∇ · A = 0 then results in ∇ 2 A = −J, and [30] 1 A= 4π
Vsource
J(r ) dv . r − r 
(2.A.9)
(2.A.10)
In the above, primed coordinates are associated with the source volume, and unprimed coordinates refer to the observation point. The magnetic field may be written as ∇ × dA, (2.A.11) H= Vsource
where dA =
J(r ) dv . 4πR
(2.A.12)
Here R = r − r . For a currentcarrying wire, the product J(r ) dv may be written (Figure 2.A.1) as J(r ) dv = J (r ) dS d = I (r ) d ,
(2.A.13)
92
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
Figure 2.A.1 Currentcarrying wire.
so that
I (r ) 1 1 ∇ × dA = ∇ × d + ∇ × d . 4π R R
(2.A.14)
Since the nabla operator works on the observation point coordinates, the second term in the above equation equals zero, and since 1 R/R (2.A.15) ∇ =− 2 , R R equation (2.A.11) may be written as H= C
I (r ) d × R , 4πR 3
(2.A.16)
which is the Biot–Savart law stated in equation (2.36). C is the contour in Figure 2.A.1, carrying the current I (r ).
REFERENCES 1. B. Blümich, NMR Imaging of Materials, Clarendon Press, Oxford, 2000. 2. H. Hricak, S. White, D. Vigneron, J. Kurhanewickz, A. Cosko, D. Levin, J. Weiss, P. Narayan and P.R. Carroll, ‘Carcinoma of the prostate gland: MR imaging with pelvic
REFERENCES
93
phasedarray coils versus integrated endorectalpelvic phasedarray coils’, Radiology, Vol. 193, No. 3, pp. 703–709, December 1994. 3. N.M. deSouza, R. Dina, G.A. McIndoe and W.P. Soutter, ‘Cervical cancer: Value of an endovagin*l coil magnetic resonance imaging technique in detecting small volume disease and assessing parametrial extension’, Gynecologic Oncology, Vol. 102, pp. 80– 85, 2006. 4. J.P. Hornbak, The Basics of MRI, available at www.cis.rit.edu/htbooks/mri, 1996–2007. 5. A. Glowinski, J. Kürsch, G. Adam, A. Bücker, T.G. Noll and R.W. Günther, ‘Device visualization for interventional MRI using local magnetic fields: Basic theory and its application to catheter visualization’, IEEE Transactions on Medical Imaging, Vol. 17, No. 5, pp. 786–793, October 1998. 6. H.H. Quick, M.E. Ladd, G.G. ZimmermannPaul, P. Erhart, E. Hofmann, G.K. von Schulthess and J.F. Debatin, ‘Singleloop coil concepts for intravascular magnetic resonance imaging’, Magnetic Resonance in Medicine, Vol. 41, pp. 751–758, 1999. 7. P.A. Rivas, K.S. Nayak, G.C. Scott, M.V. McConnell, A.B. Kerr, D.G. Nishimura, J.M. Pauly and B.S. Hu, ‘In vivo realtime intravascular MRI’, Journal of Cardiovascular Magnetic Resonance, Vol. 4, No. 2, pp. 223–232, 2002. 8. H.H. Quick, J.M. Serfaty, H.K. Pannu, R. Genadry, C.J. Yeung and E. Atalar, ‘Endourethral MRI’, Magnetic Resonance in Medicine, Vol. 45, pp. 138–146, 2001. 9. N.A.A. Op Den Kamp, Analysis and Design of Intravascular MR Antennas, MSc thesis, Eindhoven University of Technology, Faculty of Electrical Engineering, February 2003. 10. L.W. Bartels, ‘MRI voor het Geleiden en Evalueren van Behandelingen van het Bloedvatstelsel’, NVS Nieuws, pp. 18–21, December 2001. 11. O. Ocali and E. Atalar, ‘Intravascular magnetic resonance imaging using a loopless catheter antenna’, Magnetic Resonance in Medicine, Vol. 37, pp. 112–118, 1997. 12. M.J.H. Aben, Aspects of Active Tracking in MRI, MSc thesis, Eindhoven University of Technology, Faculty of Electrical Engineering, June 2004. 13. S. Weiss, T. Kuehne and M. Zenge, ‘Switchable resonant fiducial marker for safe instrument localisation at all marker orientations’, Proceedings of the 10th ISMRM Scientific Meeting and Exhibition, p. 2245, 2002. 14. M. MohammadZadeh, H. SoltanianZadeh, M. ShahAdabi and A. Tavakkoli, ‘New doubleturn loop probe for intravascular MRI’, Proceedings of the 26th Annual International Conference of the IEEE EMBS, pp. 1151–1154, September 2004. 15. P.J. Cassidy, K. Clarke and D.J. Edwards, ‘Validation of the transmissionline modelling method for the electromagnetic characterization of magnetic resonance imaging radiofrequency coils’, Proceedings of the IEE Seminar on Validation of Computational Electromagnetics, pp. 37–41, March 2004.
94
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
16. J.M. Jin, ‘Electromagnetics in magnetic resonance imaging’, IEEE Antennas and Propagation Magazine, Vol. 40, No. 6, pp. 7–22, December 1998. 17. S.M. Michaelson and J.C. Lin, Biological Effects and Health Implications of Radiofrequency Radiation, Plenum Press, New York, p. 120, 1987. 18. J.D. Bronzino (ed.), The Biomedical Engineering Handbook, second edition, Vol. 1, CRC Press, Boca Raton, pp. 894–895, 2000. 19. R.J. Johnson, Antenna Engineering Handbook, third edition, McGrawHill, New York, 1993. 20. S. Ramo, J.R. Whinnery and T. Van Duzer, Fields and Waves in Communication Electronics, second edition, John Wiley & Sons, New York, 1984. 21. C.A. Balanis, Antenna Theory Analysis and Design, second edition, John Wiley & Sons, New York, 1997. 22. H. Werner, ‘An exact integration procedure for vector potentials of thin circular loop antennas’, IEEE Transactions on Antennas and Propagation, Vol. 44, No. 2, pp. 157– 165, February 1996. 23. E. Lepelaars, Transient Electromagnetic Excitation of Biological Media by Circular Loop Antennas, PhD thesis, Eindhoven University of Technology, 1997. 24. R.W.P. King, in R.E. Collin and F.J. Zucker (eds.), Antenna Theory, Chapter 11, Part I, McGrawHill, New York, pp. 458–482, 1969. 25. R.W.P. King, C.W. Harrison and D.G. Tingley, ‘The admittance of bare circular loop antennas in a dissipative medium’, IEEE Transactions on Antennas and Propagation, pp. 434–438, July 1965. 26. R.W.P. King, C.W. Harrison and D.G. Tingley, ‘The current in bare circular loop antennas in a dissipative medium’, IEEE Transactions on Antennas and Propagation, pp. 529–531, July 1965. 27. J.D. Kraus, Antennas, McGrawHill, New York, 1950. 28. C.L. Chen and R.W.P. King, ‘The small bare loop antenna immersed in a dissipative medium’, IEEE Transactions on Antennas and Propagation, pp. 266–269, May 1963. 29. W.H. Press, B.P. Flannery, S.A. Teukolsky and W.T. Vetterling, Numerical Recipes: The Art of Scientific Computing, Cambridge University Press, 1988. 30. C.T.A. Johnk, Engineering Electromagnetic Fields and Waves, John Wiley & Sons, New York, 1975. 31. J.H. Letcher, ‘Computerassisted design of surface coils used in magnetic resonance imaging. I. The calculation of the magnetic field’, Magnetic Resonance Imaging, Vol. 7, pp. 581–583, 1989.
REFERENCES
95
32. R.K. Moore, ‘Effects of a surrounding conducting medium on antenna analysis’, IEEE Transactions on Antennas and Propagation, pp. 216–225, May 1963. 33. K. Ilzuka, R.W.P. King and C.W. Harrison Jr., ‘Self and mutual admittances of two identical circular loop antennas in a conducting medium and in air’, IEEE Transactions on Antennas and Propagation, Vol. AP14, No. 4, pp. 440–450, July 1966. 34. C.D. Taylor and C.W. Harrison Jr., ‘On thinwire multiturn loop antennas’, IEEE Transactions on Antennas and Propagation, Vol. 22, No. 3, pp. 407–413, May 1974. 35. S.G. Pan, T. Becks, D. Heberling, P. Nevermann, H. Rosmann and I. Wolff, ‘Design of loop antennas and matching networks for lownoise RF receivers: Analytic formula approach’, IEE Proceedings, Part H, Vol. 144, No. 4, pp. 274–280, August 1997. 36. R.C. Hansen, ‘Radiation and reception with buried and submerged antennas’, IEEE Transactions on Antennas and Propagation, pp. 207–215, May 1963. 37. J. Galeijs, ‘Admittance of insulated loop antennas in a disipative medium’, IEEE Transactions on Antennas and Propagation, pp. 229–235, March 1965. 38. J.A. Stratton, Electromagnetic Theory, McGrawHill, New York, 1941. 39. M. Abramowitz and I.A. Stegun, Handbook of Mathematical Functions, Dover Publications, New York, 1965. 40. H. Reichardt (ed.), Kleine Enzyklopädie Mathematik, VEB Bibliographisches Institut Leipzig, 1986. 41. C.A. Balanis, Advanced Engineering Electromagnetics, John Wiley & Sons, New York, 1989. 42. S.E. Langerak, P.K. Kunz, H.W. Vliegen, J.W. Jukema, A.H. Zwinderman, P. Steendijk, H.J. Lamb, E.E. van der Wall and A. Roos, ‘MR flow mapping in coronary artery bypass grafts: A validation study with Doppler flow measurements’, Radiology, Vol. 122, No. 1, pp. 127–135, January 2002. 43. J.M. Johnson and Y. RahmattSamii, ‘Genetic algorithms in engineering electromagnetics’, IEEE Antennas and Propagation Magazine, Vol. 39, No. 4, pp. 7–11, August 1997. 44. Y. RahmattSamii and E. Michielsen, Electromagnetic Optimization by Genetic Algorithms, John Wiley & Sons, New York, 1999. 45. R.A. Philips and M. Skopec, A Primer on Medical Device Interactions with Magnetic Resonance Imaging Systems, draft document, US Department of Health and Human Services, Food and Drug Administration, Center for Devices and Radiological Health, February 7, 1997. 46. E. Atalar, ‘Safe coaxial cables for MRI’, Proceedings of ISMRM Annual Meeting, p. 1006, 1999.
96
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
47. M.E. Ladd and H.H. Quick, ‘Reduction of resonant RF heating in intravascular catheters using coaxial chokes’, Magnetic Resonance in Medicine, Vol. 43, pp. 615–619, 2000. 48. P. Vernickel, V. Schulz, S.N. Weiss and B. Gleich, ‘A safe transmission line for MRI’, IEEE Transactions on Biomedical Engineering, Vol. 52, No. 6, pp. 1094–1102, June 2005. 49. N.A.A. Op den Kamp, J.H. Seppenwoolde, H.J. Visser, A.G. Tijhuis and C.J.G. Bakker, ‘Intravascular MR antenna designs by simulation of sensitivity profiles’, Proceedings of the International Society of Magnetic Resonance in Medicine, p. 1187, July 2003.
3 PCB Antennas: Printed Monopoles The dipole antenna is one of the oldest antennas used in practice. Heinrich Hertz used a halfwave dipole antenna in the first ever radio experiment in 1886. A wire dipole antenna may be easily constructed from a twowire transmission line by bending the ends of the open transmission line outward by 90◦ . The halfwave dipole antenna has a ‘nearomnidirectional’ radiation pattern, i.e. the radiation pattern looks like a torus with a maximum in directions perpendicular to the antenna and ‘nulls’ in directions along the antenna. A monopole antenna may be derived from a dipole antenna by mounting one arm of the dipole above a ground plane. The monopole and its image in the ground plane then form a dipole antenna. The input impedance of this monopole antenna is equal to half that of the corresponding dipole antenna, and the radiation pattern above the (infinite) ground plane is identical to the upper half of the radiation pattern of the corresponding dipole antenna. Nearomnidirectional ultrawideband (UWB) antennas may be realized, starting from a dipole or monopole antenna, using physical reasoning and ‘trial and error’ employing fullwave analysis software. These antennas may be realized as planar printed circuit board (PCB) antennas. For less widefrequencyband applications, the effect of the ground plane on the behavior of printed monopole antennas will play an important role. For this class of antennas, it is worthwhile to develop analytical models to aid in the design process.
3.1
INTRODUCTION
The current trend in miniaturization of handheld mobile wireless devices puts high constraints on the antenna or antennas to be employed. The antenna has to be small and has to possess omnidirectional radiation and sensitivity characteristics. For a singlefrequencyband applications, these two requirements make the choice of a monopole antenna natural, limiting Approximate Antenna Analysis for CAD
Hubregt J. Visser
© 2009 John Wiley & Sons, Ltd. ISBN: 9780470512937
98
PCB ANTENNAS: PRINTED MONOPOLES
Figure 3.1 Printed (microstrip) monopole antenna integrated into a PCB. The monopole antenna is excited by a microstrip transmission line. The strip monopole starts at the rim of the microstrip ground plane and is a continuation of the top conductor of the microstrip.
the size to about a quarter of a wavelength at the center frequency. For aesthetic reasons, the antenna often needs to be placed inside the device, and the combination of this constraint with the everpresent pressure to reduce production costs leads to the choice of employing a printed monopole antenna. This printed monopole antenna needs to be integrated into the RF printed circuit board (PCB), as shown in Figure 3.1. The antenna shown in this figure is excited by a microstrip transmission line. Coplanarwaveguide (CPW) excitation is an alternative possible feeding mechanism. The antennas should preferably be realized on standard FR41 PCB material, instead of on special microwave laminates. Microwave laminates, although they have a very stable relative permittivity and – in general – low losses, are expensive and difficult to process. These antennas, as well as other antenna types, may be designed on the basis of physical reasoning and ‘educated’ trial and error employing a commercially offtheshelf (COTS) fullwave analysis program. This approach, however, is only recommended for the design of a oneofakind antenna. As soon as it is foreseen that similar but not identical antennas,2 i.e. a class of antennas, need to be designed, it is worthwhile to invest in the development of an analytical model for this class of antennas. In the end, this will speed up the entire design process. In the following, we shall demonstrate this by discussing the design of a printed UWB antenna [1] and the development of an analytical model for microstripexcited monopole
1 ‘FR’ here means ‘flameretardant’, and ‘4’ means fiber glass epoxy. 2 These could be antennas for similar applications but now for different frequency bands, on different dielectric
substrates or in different environments.
PRINTED UWB ANTENNAS
99
Figure 3.2 Evolution from a narrowband, thinwire dipole antenna to a broadband, spherical dipole antenna.
antennas of the kind shown in Figure 3.1. The design constraint for all antennas will be an impedance match to 50 .
3.2
PRINTED UWB ANTENNAS
Highdatarate wireless communications need wide bandwidths. In the UWB frequency band from 3.1 GHz to 10.6 GHz, information may be spread over a large bandwidth at low power levels, thus creating the possibility of sharing the spectrum with other users. To prevent interference with existing wireless systems, such as IEEE 802.11a WLAN, stop band characteristics are required from 5 GHz to 6 GHz. In general, a UWB system and thus its antenna should be small and inexpensive. These constraints, added to the low power levels, make the antenna a critical component, a fact often undervalued by electronic designers (even RF designers). 3.2.1
Ultrawideband Antennas
From the 1930s on, antenna engineers have been searching for wideband antenna elements. They soon discovered that, starting from a dipole or monopole antenna, thickening the arms resulted in an increased bandwidth. The reason for this is that for a thick dipole or monopole antenna, the current distribution is – unlike for the thin dipole and monopole – no longer sinusoidal. While this hardly affects the radiation pattern of the antenna, it severely influences the input impedance [2]. This bandwidening effect is even more severe if the thick dipole is given the shape of a biconical antenna. A further evolution may be found in dipole and monopole antennas formed from spheres or ellipsoids [3]. Figure 3.2 shows the evolution from a thinwire dipole antenna to a spherical dipole antenna. For practical, compact applications however, a planar antenna is preferred. One planar version of the biconical antenna (the third antenna from the left in Figure 3.2) is the bow
100
PCB ANTENNAS: PRINTED MONOPOLES
Figure 3.3 Twopenny dipole antenna. Front (left) and back (right).
tie antenna. The angular discontinuities in the bow tie antenna, however, make it difficult to create an impedance match over a large frequency bandwidth [3]. Therefore, a planar antenna structure with a curved outline is preferred. A planar version of the spherical dipole antenna may be found in the ‘twopenny dipole antenna’. 3.2.2
TwoPenny Dipole Antenna
A circular planar dipole antenna may be constructed using two US cents (‘pennies’) and a semirigid coaxial piece of transmission line [3, 4] (Figure 3.3). The measured return loss of this antenna as a function of frequency is presented in Figure 3.4. A good match to 50 (S11 < −10 dB) may be observed for the UWB frequency band (3.1 GHz to 10.6 GHz) and beyond. The measurements show the first resonance just above 3 GHz. Considering the size of the antenna – the diameter of a penny is 19 mm – this resonance may be attributed to the dipole3 (Figure 3.5(a)). For higher frequencies, the current is concentrated in the rims of the pennies and the good impedance match is now due to the fact that the dipole with circular elements has transitioned to a dualnotch horn antenna formed by the rims of the pennies [3] (Figure 3.5(b)). From the twopenny UWB antenna, it should be a relatively small step towards the design of a compact PCB UWB antenna. 3.2.3
PCB UWB Antenna Design
A PCB microstrip version of the twopenny antenna has been designed and manufactured. Here, the upper circular dipole arm was realized on the upper PCB plane and was connected 3 The ‘dipole’ arm length is 19 mm. The total length of the dipole – including a 1 mm gap between the arms – is thus 39 mm, corresponding to half a wavelength at resonance (for a thin dipole). The first resonance frequency thus equals 3.8 GHz. Since we are dealing not with a thin but with a thick dipole, the first resonance frequency should be a little lower than this value. This is demonstrated in Figure 3.4.
101
PRINTED UWB ANTENNAS
5
measurement
Return Loss (dB)
10
15
20
25
30
35
40
45 1
2
3
4
5
6
7
8
9 10 11 12 13 14 15 16 17 18
Frequency (GHz)
Figure 3.4 Measured return loss as a function of frequency for the antenna shown in Figure 3.3.
Figure 3.5 Twopenny dipole UWB antenna. (a) Dipole structure around 3 GHz. (b) Dualnotch horn structure for higher frequencies.
to a microstrip transmission line (Figure 3.6). The lower circular dipole arm was integrated with the microstrip ground plane, thus forming a pseudomonopole [5]. The lower circular arm was integrated into a rectangular ground plane. This does not need to disturb the dualnotch horn antenna behavior seriously, as long as the reflection level
102
PCB ANTENNAS: PRINTED MONOPOLES
s
Figure 3.6
Microstripexcited UWB antenna.
Table 3.1 Dimensions of the microstripexcited UWB antenna. Parameter
Value (mm)
W Ws t th H1 H2 H3 D εr tan δ
22 1.44 1.6 0.07 14.38 4.62 1.51 19 4.28 0.016
at the discontinuity formed by the circle and rectangle is low. This reflection condition may be controlled by a height parameter H2 (Figure 3.6). The antenna may be regarded as an evolution of the stripline version demonstrated in [6]. The microstrip version is less costly in production than the stripline version and easier to integrate into an existing RF PCB design. The simulated return loss as a function of frequency, after the dimensions indicated in Figure 3.6 had been optimized manually, is shown in Figure 3.7. For the optimization, use was made of the fullwave finiteintegration technique (FIT) software package Microwave Studio© from CST [7]. The dimensions used are stated in Table 3.1. The parameters r, tan δ and th are the relative permittivity of the PCB substrate, the loss tangent of the substrate and the thickness of the copper layers, respectively.
103
PRINTED UWB ANTENNAS
0 simulation measurement
Return Loss (dB)
5
10
15
20
25
30 1
2
3
4
5
6 7 8 Frequency (GHz)
9
10
11
12
Figure 3.7 Simulated and measured return loss as a function of frequency for the microstripexcited UWB antenna shown in Figure 3.6.
In Figure 3.7, we also show measurement results for the antenna that was constructed, shown in Figure 3.8. It was observed that cable currents greatly influenced the measurement results. This is the main reason for the differences between the simulation and measurement results visible in Figure 3.7. This phenomenon may also be observed, although it is not always explicitly mentioned, in recently published UWB antenna simulation and measurement results (e.g. [5, 8–10]). Nevertheless, the author is convinced that with proper actions for suppressing these currents [5, 11], close agreement may be reached. One has to bear in mind, though, that the antenna is intended for application in an integrated onPCB solution. 3.2.3.1 Feed Line The disadvantage of the centerfed dipole is that a transmission line must be brought to the gap between the dipole arms. Since the transmission line will be positioned inside the reactive near field of the antenna, it will be vulnerable to undesired sheet coupling. The radiation pattern of the antenna may be distorted owing to this coupling [3]. In [3], a solution to this possible problem was shown that consists of a strip transmission line feed and a tapered balun. In [6], a ‘hidden’ stripline feed was used. The stripline was positioned halfway between the two layers of the bottom dipole arm. Both dipole arms consisted of two metal layers on opposite sides of the PCB substrate, electrically connected through metallized vias located on the rims of the circular arms. This latter antenna will be used as a benchmark. Our pseudomonopole UWB antenna with a nonhidden microstrip feed, however, does not seem to be prone to the abovementioned negative effects. This is demonstrated in Figure 3.9, which shows the threedimensional radiation patterns for 3 GHz and 6 GHz. The gain value
104
PCB ANTENNAS: PRINTED MONOPOLES
Figure 3.8 The microstripexcited UWB antenna constructed. Top (left) and bottom (right).
Figure 3.9 Simulated threedimensional radiation patterns of the pseudomonopole printed UWB antenna for 3 GHz (top) and 6 GHz (bottom).
is indicated in the figure at the right. The antenna PCB was positioned parallel to and in the xy plane shown in Figure 3.9. For the stripline dipole antenna discussed in [6], as well as for our microstrip pseudomonopole antenna, the radiation patterns for frequencies above 6 GHz start to deviate
105
PRINTED UWB ANTENNAS
Table 3.2 Azimuthal gain and maximum variation of the azimuthal gain function for a stripline dipole (SL) and a microstrip pseudomonopole (MS) antenna. Frequency Azimuthal gain, Azimuthal gain, Maximum variation, Maximum variation, (GHz) SL (dBi) MS (dBi) SL (dBi) MS (dBi) 3 4 5 6 7 8 9 10
2.19 2.63 3.09 3.82 3.65 2.46 – –
2.46 3.02 3.63 4.02 4.24 2.16 −3.19 0.11
0.39 0.79 1.84 3.26 4.42 5.10 – –
0.84 1.69 2.54 3.35 5.22 8.16 9.41 15.91
seriously from the halfwave dipole patterns shown in Figure 3.9. The origin of this deviation will be explained in section 3.2.3.2. First, we shall take a closer look at the azimuthal (xzplane) radiation patterns, which demonstrate this deviation. In Table 3.2, the azimuthal gain and the maximum variation of the azimuthal gain function are shown for both antennas for a number of discrete frequencies. This table demonstrates, together with Figure 3.7, that the two antennas are comparable in behavior. The stripline antenna shows a slightly more uniform radiation pattern, close to that of a halfwave dipole antenna, but the microstrip antenna is easier and thus less costly to manufacture. To demonstrate how the radiation pattern changes with frequency, simulated azimuthal (xzplane) radiation patterns of the pseudomonopole antenna are shown in Figure 3.10 for frequencies from 3 GHz to 6 GHz and in Figure 3.11 for frequencies from 7 GHz to 10 GHz. Zero degrees coincides with x = 0. Figure 3.11 clearly shows how, for frequencies in excess of 6 GHz, the azimuthal pattern deviates seriously from that of a halfwave dipole antenna. Since the antenna behaves around 3 GHz as a halfwave dipole antenna, above 6 GHz the length of the antenna becomes larger than a whole wavelength and ‘elevational lobes’ will evolve with increasing frequency, which disturb the azimuthal sections of the radiation pattern. The occurrence of elevational lobes is demonstrated in Figure 3.12, which shows the threedimensional radiation patterns at 7 GHz and 10 GHz. A more halfwavedipolelike pattern over the whole UWB frequency band can thus be created by shortening the antenna. 3.2.3.2 Antenna Shortening If we shorten the antenna, the generation of elevational lobes will start at a higher frequency. The first resonance, however, will also occur at a higher frequency. Looking at the results for return loss versus frequency (simulation) in Figure 3.7, we observe that we still have some margin if we require a return loss of less than −10 dB over the frequency band from 3.1 GHz to 10.6 GHz. After, again, a manual optimization using a fullwaveanalysis software package, the length H1 = 14.38 mm (Figure 3.6) was replaced by H1 = 6.38 mm. The simulated and measured return loss as a function of frequency are shown in Figure 3.13.
106
PCB ANTENNAS: PRINTED MONOPOLES
4.5
4
3GHz 4GHz 5GHz 6GHz
3.5
Gain (dBi)
3
2.5
2
1.5
1
0.5
0 0
30
60
90
120 150 180 210 240 270 300 330 360 Azimuth angle (degrees)
Figure 3.10 Simulated gain of the pseudomonopole printed UWB antenna as a function of azimuthal angle for frequencies of 3 GHz, 4 GHz, 5 GHz and 6 GHz.
10 7GHz 8GHz 9GHz 10GHz
7.5
5
2.5
Gain (dBi)
2.5
5
7.5
10
12.5
15
17.5
20 0
30
60
90
120 150 180 210 240 270 300 330 360 Azimuth angle (degrees)
Figure 3.11 Simulated gain of the pseudomonopole printed UWB antenna as a function of azimuth angle for frequencies of 7 GHz, 8 GHz, 9 GHz and 10 GHz.
PRINTED UWB ANTENNAS
107
Figure 3.12 Simulated threedimensional radiation patterns of the pseudomonopole printed UWB antenna for 7 GHz (top) and 10 GHz (bottom).
The measurements again suffered from cable current effects. These effects are more severe than for the previous antenna, since this antenna is shorter. The main characteristics, i.e. a high return loss for frequencies below 3 GHz and a low return loss for frequencies above 3 GHz, are still present in the measurement results. Comparison of the simulation results with those shown in Figure 3.7 shows that it is possible to shorten the antenna without compromising the return loss characteristics. One has to be careful, though, for the return loss around 3 GHz. In Table 3.3, we compare the azimuthal gain characteristics of the shortened pseudomonopole antenna with those of the strip line dipole antenna of [6]. The simulated azimuthal (xzplane) radiation patterns of the shortened pseudomonopole antenna are shown in Figure 3.14 for frequencies from 3 GHz to 6 GHz and in Figure 3.15 for frequencies from 7 GHz to 10 GHz. Zero degrees coincides with x = 0. Table 3.3 and Figures 3.14 and 3.15 (to be compared with Table 3.2 and Figures 3.10 and 3.11, respectively) show that the azimuthal behavior of the gain has improved with respect to the original microstrip pseudomonopole antenna. This is demonstrated again in Figure 3.16, which shows the maximum of the gain function and the maximum variation of the gain function in the azimuthal (xz) plane for the original and the shortened pseudomonopole UWB antenna. This figure shows that the shortened antenna exhibits a more constant gain over the frequency band. A comparison of Table 3.3 with Table 3.2 reveals further that the shortened
108
PCB ANTENNAS: PRINTED MONOPOLES
0 simulation measurement
Return Loss (dB)
5
10
15
20
25
30 1
2
3
4
5
6
7
8
9
10
11
12
Frequency (GHz)
Figure 3.13 Simulated and measured return loss as a function of frequency for the shortened microstripexcited UWB antenna.
Table 3.3 Azimuthal gain and maximum variation of the azimuthal gain function for a stripline dipole (SL) and a shortened microstrip pseudomonopole (MS) antenna. Frequency Azimuthal gain, Azimuthal gain, Maximum variation, Maximum variation, (GHz) SL (dBi) MS (dBi) SL (dBi) MS (dBi) 3 4 5 6 7 8 9 10
2.19 2.63 3.09 3.82 3.65 2.46 – –
2.30 2.68 3.07 3.14 2.99 2.13 0.62 1.36
0.39 0.79 1.84 3.26 4.42 5.10 – –
0.554 1.17 1.90 2.84 5.58 6.50 6.45 3.68
microstrip UWB antenna shows a behavior more similar to that of the stripline UWB antenna. The dimensions, however, are smaller (22 mm×33 mm×1.6 mm versus 20.5 mm×40 mm× 1 mm [6]), and the antenna is easier to produce and does not have a critical feeding and transition region [12]. With this shortened microstrip pseudomonopole UWB antenna as a basis, we shall now look at measures to suppress signals in the 5 GHz to 6 GHz frequency band.
109
PRINTED UWB ANTENNAS
4.5
4
3GHz 4GHz 5GHz 6GHz
3.5
Gain (dBi)
3
2.5
2
1.5
1
0.5
0 0
30
60
90
120 150 180 210 240 270 300 330 360 Azimuth angle (degrees)
Figure 3.14 Simulated gain of the shortened pseudomonopole printed UWB antenna as a function of azimuth angle for frequencies of 3 GHz, 4 GHz, 5 GHz and 6 GHz.
Table 3.4 Slot dimensions of the microstrip UWB antenna.
3.2.4
Parameter
Value (mm)
WU1 WU2 WL LU1 LU2
7.2 1.2 1.2 15.1 5.9
BandStop Filter
To create a frequency band notch function, we may either change the current flow in the metal parts of the antenna or insert a filter before or in the feed line of the antenna. 3.2.4.1 Slot in Radiator To influence the current flow (in such a way that destructive interference would occur for frequencies between 5 GHz and 6 GHz), we introduced a slot into the upper arm of our antenna. Since, at higher frequencies, the current will be concentrated at the rims of the two circular arms, the slot has to be positioned in the neighbourhood of the rim of the circle. For ease of drawing, we chose a Ushaped slot as shown in Figure 3.17. The slot dimensions, after manual optimization, were as stated in Table 3.4. All other dimensions were those of the shortened UWB antenna.
110
PCB ANTENNAS: PRINTED MONOPOLES
10 7GHz 8GHz 9GHz 10GHz
7.5
5
2.5
Gain (dBi)
2.5
5
7.5
10
12.5
15
17.5
20 0
30
60
90
120 150 180 210 240 270 300 330 360 Azimuth angle (degrees)
Figure 3.15 Simulated gain of the shortened pseudomonopole printed UWB antenna as a function of azimuth angle for frequencies 7 GHz, 8 GHz, 9 GHz and 10 GHz.
20 Long, gain Short, gain Long, gain variation Short, gain variation
Max gain function (variation) (dBi)
17.5
15
12.5
10
7.5
5
2.5
2.5
5 3
4
5
6
7
8
9
10
Frequency (GHz)
Figure 3.16 Simulated maximum of the gain function and maximum variation of the gain function with azimuthal angle versus frequency for original (‘long’) and shortened (‘short’) microstrip UWB antenna.
PRINTED UWB ANTENNAS
Figure 3.17 function.
111
Microstrip UWB antenna with Ushaped slot to create a frequency band notch
Figure 3.18 Shortened microstrip UWB antenna with Ushaped slot to create a frequency band notch function.
The antenna constructed is shown in Figure 3.18. The simulated and measured return loss as a function of frequency are shown in Figure 3.19. Although the measurements were still hindered by cablecurrent effects, we can clearly observe that the Ushaped slot adds the desired frequency band notch functionality.
112
PCB ANTENNAS: PRINTED MONOPOLES
0 simulation measurement
5
Return Loss (dB)
10
15
20
25
30
35 1
2
3
4
5
6
7
8
9
10
11
12
Frequency (GHz) Figure 3.19 Simulated and measured return loss as a function of frequency for a shortened microstrip UWB antenna with a Ushaped slot.
3.2.4.2 Spurline Filter in Microstrip A microstrip spurline filter [13] acts as a bandstop filter. A nice feature of a spurline filter is that the physical structure is completely contained within the boundaries of the microstrip transmission line (Figure 3.20). The length L is equal to a quarter of the wavelength in the transmission line [14]. As an example, a spurline filter was incorporated into the microstrip transmission line of the planar UWB antenna shown in Figure 3.6. The widths S and G were taken to be S = G = 0.3 mm and the length L was 7 mm. The spurline filter was positioned symmetrically in the microstrip transmission line at a distance of 1 mm from the edge of the substrate. These values were found after several iterations employing a fullwave analysis program. Figure 3.21 shows the simulated return loss as a function of frequency for the original antenna, i.e. without a filter, and for the antenna incorporating a spurline filter. The figure clearly shows the stop band behavior between 5 GHz and 6 GHz. The figure also shows that additional optimization is needed to correct the return loss characteristics between 3 GHz and 5 GHz. Since the microstrip transmission line in the UWB antenna is part of the antenna (and the characteristic impedance is not equal to 50 ), this optimization may involve many lengthy fullwave iterations. Therefore, it may be advantageous to incorporate the spurline filter into a 50 microstrip transmission line that will be connected to the antenna. The reflection and transmission coefficients of the spurline filter may be calculated relatively easily using the closedform equations for the elements of the ABCD matrix given in [13]. In these equations, use may be made of the quasistatic even and oddmode effective
113
PRINTED UWB ANTENNAS
Figure 3.20 Microstrip spurline ﬁlter. The ﬁlter structure is ‘cut out’ of the top layer of the microstrip.
permittivities and characteristic impedances for parallel coupled microstrip lines that can be found in [15]. The ABCD matrix of the spurline filter shown in Figure 3.20 is given by [13] j(1/2)[Z0e sin(ϑe ) + Z0o tan(ϑo ) cos(ϑe )] A B cos(ϑe ) , (3.1) = j(2/Z0e ) sin(ϑe ) cos(ϑe ) − (Z0o /Z0e ) sin(ϑe ) tan(ϑo ) C D where Z0e and Z0o are the evenmode and oddmode characteristic impedances, respectively, and ϑe and ϑo are the evenmode and oddmode electrical lengths. These are given by
where
ϑe = βe L,
(3.2)
ϑo = βo L,
(3.3)
√ 2π εreffe , βe = λ0 √ 2π εreffo , βo = λ0
(3.4) (3.5)
Here, λ0 is the freespace wavelength and εreffe and εreffo are the effective relative permittivities of the even mode and odd mode, respectively. The ABCD matrix of the microstrip spurline filter is derived from the impedance matrix of a section of coupled microstrip transmission lines, applying the correct termination
114
PCB ANTENNAS: PRINTED MONOPOLES
0 No filter With spurline filter
Return Loss (dB)
5
10
15
20
25
30 1
2
3
4
5
6
7
8
9
10
11
12
Frequency (GHz) Figure 3.21 Simulated return loss versus frequency for original (nonshortened) pseudomonopole UWB antenna without and with an integrated spurline ﬁlter.
Figure 3.22 ﬁlter.
Applying termination conditions to coupled transmission lines to create a spurline
conditions, i.e. the two coupled microstrip transmission lines are connected together at one side while one of the transmission lines is left open at the other side (Figure 3.22). The derivation of the impedance matrix of a section of coupled microstrip transmission lines follows the derivation in [16, 17] for coupled TEM transmission lines, corrected for the nonTEM nature of a microstrip transmission line. This correction consists of employing different phase velocities for the even and odd modes. The even and oddmode characteristic impedances and effective permittivities follow from the treatment in [15], see also Figure 3.23. The normalized strip width and the normalized gap width (with respect to the substrate height) are given by W , h S g= . h u=
(3.6) (3.7)
115
PRINTED UWB ANTENNAS
Figure 3.23
Coupled microstrip transmission lines.
The mode characteristic impedances are given by Z0m (u, g) =
Z0 (u) , 1 − Z0 (u)φm (u, g)/η
(3.8)
where m = e (even mode) or m = o (odd mode) and where Z0 (u) = and
η u + 1.98u0.172
η=
µ0 . ε0 εreffm
(3.9)
(3.10)
Further, φe (u, g) =
(g){α(g)um(g)
ϕ(u) , + [1 − α(g)]u−m(g) }
ϕ(u) = 0.8645u0.172, g 2.09
g + , 1.45 3.95 α(g) = 0.5e−g , 20.36 6 −0.251 m(g) = 0.2175 + 4.113 + g 10 g 1 ln + , 323 1 + (g/13.8)10 θ (g) [β(g)u−n(g) ln(u)] e , φo (u, g) = φe (u, g) − (g) 0.627 , θ (g) = 1.729 + 1.175 ln 1 + g + 0.327g 2.17 (g) = 1 +
(3.11) (3.12) (3.13) (3.14)
(3.15) (3.16) (3.17)
116
PCB ANTENNAS: PRINTED MONOPOLES
g 10 1 ln β(g) = 0.2306 + 301.8 1 + (g/3.73)10 1 + ln[1 + 0.646g 1.175], 5.3 1 [−6.424−0.76 ln(g)−(g/0.23)5 ] +e n(g) = 17.7 10 + 68.3g 2 × ln , 1 + 32.5g 3.093 εr + 1 εr − 1 + Fm (u, g, εr ), εreffm (u, g, εr ) = 2 2 −a(u)b(εr) 10 Fe (u, g, εr ) = 1 + , µ(u, g)
(3.18)
(3.19) (3.20) (3.21)
20 + g 2 , 10 + g 2 4 u + (u/52)2 u 3 1 1 ln ln 1 + a(u) = 1 + , + 49 u4 + 0.432 18.7 18.1 εr − 0.9 0.053 , b(εr) = 0.564 εr + 3 10 −a(u)b(εr) Fo (u, g, εr ) = fo (u, g, εr ) 1 + , u µ(u, g) = ge−g + u
fo (u, g, εr ) = fo1 (g, εr )e[p(g) ln(u)+q(g) sin(πln(u)/ln(10))] , p(g) =
0.295 e−0.745g
cosh(g 0.68 )
q(g) = e−1.366−g , fo1 (g, εr ) = 1 − e
2 e1−(εr −1) /8.2 . r(g, εr ) = 1 + 0.15 1 − 1 + g −6
(3.23) (3.24) (3.25) (3.26) (3.27)
,
[−0.179g 0.15 −0.328g r(g,εr ) /ln[e+(g/7)2.8 ]]
(3.22)
(3.28) ,
(3.29) (3.30)
As an example, in Figure 3.24, the transmission coefficient is shown as a function of frequency for a spurline filter with W = 3.3 mm, t = 1.6 mm, εr = 4.28, th = 0.07 mm, tan δ = 0.016 (50 characteristic impedance), S = G = 0.3 mm and L = 7 mm. The transmission coefficient was calculated from the closedform expressions and is compared here methodofmoments simulation results. The figure shows that the calculations based on the quasistatic, closedform expressions result in transmission characteristics very close to those calculated with a fullwave method. The differences still present must be attributed to the fact that the gap (G; see Figure 3.20) is not accounted for in the quasistatic calculations. This gap may be accounted for by employing an effective spurline filter length. In Figure 3.24, employing an effective length Leff = 1.065L results in closer agreement between the quasistatic and fullwave simulation results. As will be shown in Chapter 5, the concept of an equivalent length may be employed
117
PRINTED STRIP MONOPOLE ANTENNAS
0 MoM Quasi static Quasi static, corr.
Transmission (dB)
5
10
15
20
25
30 1
2
3
4
5
6
7
8
9
10
11
12
Frequency (GHz)
Figure 3.24
Simulated transmission coeﬃcient as a function of frequency for a spurline ﬁlter.
to correct for resonance. A general length extension equation needs to be derived, but this is beyond the scope of this chapter. Nevertheless, it has been demonstrated how analytical models may help in speeding up the design process. The time needed to generate the quasistatic results is orders of magnitude smaller than the time needed by fullwave methods. An analytical model for the antenna itself would therefore help considerably in speeding up the design process.
3.3
PRINTED STRIP MONOPOLE ANTENNAS
Most printed UWB antennas reported in the literature may be considered as pseudomonopole antennas, acting as halfwave dipole antennas around 3 GHz and as two taperedslot antennas for higher frequencies. Since printed pseudomonopole antennas (Figure 3.1), may also be of interest for nonUWB applications, owing to their small size and easy integration into a PCB, the availability of a model that is fast when implemented in software but is also accurate would be advantageous. The modeling of a monopole antenna at the edge of an infinite sheet has been performed by several authors employing the dyadic Green’s function for a perfectly conducting wedge [18, 19]. Modeling of monopole antennas on the edge of finite halfsheets has been conducted in [20], amongst others. These models, however, still rely heavily on numerical methods and are not considered fit for our purpose. We seek an analytical model that is relatively easy to implement in software and generates results quickly. A model relying heavily on numerical methods could be employed to generate a database of analysis results for various configurations, after which, through interpolation and extrapolation, an antenna
118
PCB ANTENNAS: PRINTED MONOPOLES
design could be generated in a relatively short time. The additional advantage of employing an analytical model, next to the overall speed benefit, however, is that it provides insight by showing how different parameters are related to the analysis results. The price that we are prepared to pay is a limited but still acceptable accuracy that will allow us to use the model for generating initial designs that will, eventually, need to be finetuned, employing slower but more accurate methods. We have found such an analytical model. This analytical model is based on the ‘threeterm model’ for a cylindrical dipole antenna, where an imperfect conductor is modeled by means of a distributed impedance [21, 22]. We specifically chose an analytical dipole model in favor of a numerical model for the reasons mentioned above. By using a distributed inductance [23] for the distributed impedance, it becomes possible to model a cylindrical dipole antenna that has a dielectric or magnetic coating. Next, a strip dipole antenna on a dielectric slab is modeled as an equivalent magnetically coated cylindrical dipole antenna [24]. The input impedance of a strip monopole antenna is then found as half that of the corresponding strip dipole antenna; the radiation pattern above an infinite, perpendicular ground plane is identical to the upper half of the radiation pattern of the corresponding strip dipole antenna. Next, the (finite) ground plane is placed parallel to the strip monopole antenna as shown in Figure 3.1. In the following, we shall briefly discuss the model of an imperfectly conducting dipole antenna with a circular cross section, the introduction of a distributed inductance representing a magnetic coating and the use of a generalization of the concept of the equivalent radius to convert a strip dipole antenna to a magnetically coated wire dipole antenna with a circular cross section. 3.3.1
Model of an Imperfectly conducting Dipole Antenna
The admittance Y of a circularly cylindrical, imperfectly conducting dipole antenna with halflength h and cylinder radius a, excited centrally by a deltagap voltage generator V (Figure 3.25), is given by [25] 1 2πk0 sin(kh) + TU {1 − cos(kh)} + TD 1 − cos k0 h . (3.31) Y =j ξ0 kψdR cos(kh) 2 As will be shown, the distributed impedance is included in the wave number k. In the above equation, µ0 (3.32) ξ0 = ε0 is the characteristic impedance of free space; µ0 = 4π × 10−7 H m−1 is the permeability of free space and ε0 ≈ 8.854 × 10−12 F m−1 is the permittivity of free space. Also, in equation (3.31), √ k 0 = ω ε 0 µ0 (3.33) is the freespace wave number, where ω = 2πf , f being the frequency. The wave number k in equation (3.31) is defined by [25] k 2 = k02 1 − j
4πzi , k0 ξ0 kψdR
(3.34)
119
PRINTED STRIP MONOPOLE ANTENNAS
Figure 3.25
Circularly cylindrical dipole antenna of halflength h.
where zi is the distributed impedance. The expansion parameter ψdR is defined by ψdR
ψdR (0) = ψdR (h − λ/2)
if k0 h ≤ π/2 if π/2 ≤ k0 h ≤ 3π/2,
(3.35)
where λ is the wavelength and cos(k0 r) cos(k0 rh ) − dz , sin(k[h − z ]) ψdR (z) = csc(k[h − z]) r rh z =−h
h
where r= and
(z − z )2 + a 2
(3.36)
(3.37)
(h − z )2 + a 2 .
(3.38)
TU =
CV ED − CD EV CU ED − CD EU
(3.39)
TD =
CU EV − CV EU , CU ED − CD EU
(3.40)
rh = Further,
and
120
PCB ANTENNAS: PRINTED MONOPOLES
where k2 CU = 1 − 2 (ψdUR − ψdR )(1 − cos[kh]) k0 1 3 − cos k0 h + ψU (h), − ψdUR cos[kh] + jψdUI 4 2 1 3 CD = ψdD − cos k0 h 4 2 k2 1 − 1 − 2 ψdR 1 − cos k0 h + ψD (h), 2 k 0 1 3 − cos k0 h + ψV (h) , CV = − jψdI 4 2 2 k 1 1 EU = − 2 ψdUR cos[kh] − j ψdUI cos k0 h + ψU (h), 4 2 k0 1 1 ED = − ψdD cos k0 h + ψD (h), 4 2 1 1 EV = j ψdI cos k0 h − ψV (h). 4 2
(3.41)
(3.42) (3.43) (3.44) (3.45) (3.46)
In equations (3.41)–(3.46), ψV (h) =
h
z =−h h
sin[k(h − z )]
e−jk0 rh dz , rh
e−jk0 rh dz , rh z =−h −jk0 rh h e 1 1 dz , ψD (h) = cos k0 z − cos k0 h 2 2 r h z =−h
ψU (h) =
{cos[kz ] − cos[kh]}
ψdUR = {1 − cos[kh]}−1 h cos[k0 r0 ] cos[k0 rh ] dz , × {cos[kz ] − cos[kh]} − r0 rh z =−h ψdD = {1 − cos[kh]}−1 −jk0 r0 h e e−jk0 rh dz , × {cos[k0 z ] − cos[k0 h]} − r r 0 h z =−h −1 1 ψdI = − 1 − cos k0 h 2 h sin[k0 r0 ] sin[k0 rh ] dz , × sin[k(h − z )] − r r 0 h z =−h
(3.47) (3.48) (3.49)
(3.50)
(3.51)
(3.52)
PRINTED STRIP MONOPOLE ANTENNAS
−1 1 ψdUI = − 1 − cos k0 h 2 h sin[k0 r0 ] sin[k0 rh ] dz . × {cos[kz ] − cos[kh]} − r r 0 h z =−h
121
(3.53)
Equations (3.34), (3.35) and (3.36) are implicit, meaning that the expansion parameter ψdR is needed for the calculation of the wave number k and, equally, the wave number k is needed for the calculation of the expansion parameter ψdR . To obtain a solution, an iterative method was used. First, the expansion parameter ψdR was calculated with k0 (the freespace wave number) substituted for k in equation (3.36). The expansion parameter thus found was then used to obtain a better solution for k by substituting the value found into equation (3.34). With the newly found value for k, a better solution for ψdR was calculated, after which the whole procedure was repeated. Since the expansion parameter is relatively insensitive to the value of the wave number [25, 26], a stable solution was obtained, in general, after one or two iterations. 3.3.2
Dipole Antenna with Magnetic Coating
In this section, we shall briefly explain how any (analytical) expression for the current in a wire antenna where we have the facility to impose an arbitrary impedance per unit length may be used to obtain results for the same configuration when the wire has a magnetic coating. To this end, we start with the electricfield integrodifferential equation for the unknown current I () in the inner wire of a wire configuration where the wires are coated with a material having a relative permeability µr () and a relative permittivity εr () as a function of the position along the wire. The core is assumed to have a radius a() as a function of the position along the wire, and the radius of the cylindrical coating is b(). The expression is given by [23] 1 1 dI ( ) ∂ µr ( )ˆ · ˆ + Ga (, ) d εr ( ) k 2 d ∂ wires jωµ0 [µr ( ) − 1]ˆ · ˆ I ( ) − 4π wires 1 dI ( ) ∂ 1 − 1 + Gb (, ) d εr ( ) k 2 d ∂ = ˆ · Ei () − Zi ()I (),
jωµ0 4π
(3.54)
where Ei is the externally impressed electric field, Zi () is the intrinsic impedance per unit length of the inner conductor and ˆ and ˆ are unit vectors parallel to the wire at positions and , respectively. Ga (, ) and Gb (, ) are the Green’s functions for the inner and outer radii, respectively. Using Gab (, ) ≡ Ga (, ) − Gb (, ), the above equation may
122
PCB ANTENNAS: PRINTED MONOPOLES
be rewritten as
1 dI ( ) ∂ ˆ ˆ · I ( ) + 2 Ga (, ) d k d ∂ wires jωµ0 + [µr ( ) − 1]ˆ · ˆ I ( )Gab (, ) d 4π wires 1 dI ( ) ∂ 1 jωµ0 − 1 Gab (, ) d + 4π wires εr ( ) k 2 d ∂ = ˆ · Ei () − Zi ()I ().
jωµ0 4π
(3.55)
The above equation shows separate contributions for the bare wire (first term), the magnetic effect of the coating (second term) and the dielectric effect of the coating (third term). The equation shows that a magnetic coating is easier to handle than a dielectric coating. Therefore, in section 3.3.4 we shall discus how we can transform an equivalent dipole with a dielectric coating into another equivalent dipole with a magnetic coating. For a magnetic coating (εr ( ) = 1), the third term in the above equation vanishes. With
e−jkRp (, ) , Gp (, ) = Rp (, ) Rp (, ) = p2 + r() − r( )2 ,
p = a, b,
(3.56)
where r() is the position vector of the point on the wire structure and the source point at taken on the axis of the wire, Gab (, ) shows a contribution concentrated at and around = , so that the second term may be approximated by [23] jωµ0 b() jωµ0 . (3.57) I ()[µr () − 1] I ()[µr () − 1] ln Gab (, ) d
4π 2π a() wires Then, finally, the integral equation may be written as jωµ0 1 dI ( ) ∂ Ga (, ) d = ˆ · Ei () − [Zi () + Zm ()]I (), ˆ · ˆ I ( ) + 2 4π wires k d ∂ (3.58) where jωµ0 b() [µr () − 1] ln Zm () = ≡ jωL(). (3.59) 2π a() 3.3.3
Generalization of the Concept of Equivalent Radius
The equivalentradius theory of Hallén is based essentially on a twodimensional electrostatic approximation [24]. We first determine the capacitance per unit length of a twodimensional conductor with a cross section the same as that of the antenna with respect to some parallel reference conductor at a certain distance. Then we demand that the capacitance per unit length of the equivalent radius conductor with respect to the same reference conductor, at
PRINTED STRIP MONOPOLE ANTENNAS
123
Figure 3.26 Conversion from a cylindrical antenna in the presence of a substrate to an equivalent circularcrosssection antenna (PEC, perfect electric conductor).
the same distance, is equal to this value. This concept may be generalized to the case of a thin cylindrical antenna in the presence of arbitrarily shaped dielectric and/or magnetic materials (Figure 3.26). If we consider the metal part of the antenna shown on the left of Figure 3.26 as a perfect conductor, the longitudinal component of the electric field on the metallic surface will be zero. Also, the normal component of the magneticfield vector will be zero, and – owing to the assumed cylindrical shape – the longitudinal component of the magnetic field will be zero as well. Close to the antenna, the two field vectors may therefore considered as quasistatic in nature, and to be due to a current flowing along and a charge on an infinitely long cylinder. For the antenna shown on the right of Figure 3.26 to be equivalent to the one shown on the left, the charge per unit length Q and the current I should be identical [24]. Furthermore, the electric and magnetic fields at large distances from the antennas should also be identical, whereas the fields near the antennas will in general differ greatly. Next, we take a twoconductor system where both conductors are of the form shown on the left of Figure 3.26 and a twoconductor system where both conductors are of the form shown on the right of the same figure. We demand that the electrical energies per unit length We for the two systems corresponding to equal and opposite charges Q and −Q on the conductors be equal [24]: Weleft =
Q2 Q2 = W = . eright 2Cleft 2Cright
(3.60)
We also demand that the magnetic energies per unit length Wm for the two systems corresponding to equal and opposite currents I and −I in the conductors be equal [24]:
Wm left =
Lright I 2 Lleft I 2 = Wm right = . 2 2
(3.61)
124
PCB ANTENNAS: PRINTED MONOPOLES
Figure 3.27
. Geometry for the calculation of Cleft
With πε0 , ln(d/a) − (1 − 1/εer ) ln(b/a) d b µ0 µe − µ0 ln ln = + , π a π a
Cright =
(3.62)
Lright
(3.63)
where d b is the distance between the two conductors, we find that b 1 d πε0 ln 1− = ln − , εer a a Cleft π b d = . (µer − 1) ln Lleft − ln a µ0 a
(3.64) (3.65)
We now have two equations and four unknowns (a, b, εer and µer ), so two unknowns may be chosen for convenience. For a metallic strip on a dielectric slab, we choose εer = εr and µer = 1. Then, equation (3.64) leads to εr d πε0 b = exp ln − . (3.66) a εr − 1 a Cleft This choice of unknowns, when substituted into equation (3.65), leads to a formulation that applies to an antenna without a dielectric or magnetic covering. This implies that the equivalent radius a should be equal to onefourth of the strip width, i.e. a = w/4. is the capacitance between two identical electrically conducting The capacitance Cleft strips on the dielectric slab, displaced relative to each other a distance d, as shown in Figure 3.27. The capacitance value is calculated as [27] 16 εr2 cosh(xh) + εr sinh(xh) 2 xw 2 xd dx. sin sin 3 2 2 2 2 (εr + 1) sinh(xh) + 2εr cosh(xh) x=0 x w (3.67) −1 It is not possible to let the distance d go to infinity, since that would lead to Lleft , Lright, Cleft 1 1 = Cleft πεr ε0
∞
−1 and Cright becoming infinite. A value of 20 to 200 times the radius R in Figure 3.26 is advised in [24].
125
PRINTED STRIP MONOPOLE ANTENNAS
3.3.4
Equivalent Dipole with Magnetic Coating
To analyze a strip dipole or monopole antenna on a dielectric slab, we prefer to transform the equivalent dielectrically coated wire antenna into an equivalent wire antenna with a purely magnetic coating. The reason for performing this extra transformation lies in the fact that this will lead to a thinner coating and the theory is more accurate for thinner coatings [24]. For this transformation, we replace the parameters a, b, εr and µr by a , b , εr and µr , respectively, where [23, 24] b = b, µr = µr εr , 1/εr a . a =b b Substituting these transformed parameters into equation (3.59) gives b jωµ0 εr − 1 . Ze = ln 2π εr a
(3.68) (3.69) (3.70)
(3.71)
Then, with equation (3.66) substituted into equation (3.71) and the latter equation substituted into equation (3.34), we may calculate the input impedance of a strip dipole on a dielectric slab. 3.3.5
Validation
To validate the computer code based on the analysis techniques described above, a strip monopole antenna on a dielectric slab, placed perpendicularly on an infinite ground plane, was analyzed. The configuration and its dimensions are shown in Figure 3.28. The real and imaginary parts of the input admittance (Yin = G + jB) were calculated as a function of frequency and are shown, together with measured results from [24], in Figure 3.29. The admittance was calculated as twice the admittance of the corresponding dipole antenna. In the same figure, we also show the calculated results for a bare strip antenna, analyzed as an equivalent antenna of circular cross section.4 The figure shows, first of all, that the effects of the dielectric need to be included in the analysis. Furthermore, the agreement between the calculated and measured input admittance results is fair to good over the frequency band shown: the difference between the calculated and measured values of G relative to the measured maximum of G remains below 10% and the difference between the calculated and measured values of B relative to the measured maximum of B remains below 16%. Around resonance, these numbers are much lower. By replacing the numerical evaluation of a coated wire antenna by an analytical evaluation, we have simplified the total analysis without severely compromising the accuracy, as demonstrated in [24]. Next, we adapt the analysis technique to analyze planar printed monopoles of the kind shown in Figure 3.1. 4 In the calculation of the capacitance, the distance between the identical strips was varied between 10 and 200 times
the height of the dielectric. No significant difference was observed between the calculated admittance results.
126
PCB ANTENNAS: PRINTED MONOPOLES
Figure 3.28 Strip monopole antenna on a dielectric slab, placed perpendicularly on an inﬁnite ground plane. The width of the dielectric slab was 51 mm.
35
30
G model B model G measurement B measurement G bare strip, model B bare strip, model
25
G, B (mS)
20
15
10
5
5
10
15 180
200
220
240
260
280
300
320
340
360
380
400
Frequency (MHz)
Figure 3.29 Calculated and measured (from [24]) input admittance as a function of frequency for the strip monopole antenna shown in Figure 3.28, and calculated input admittance as a function of frequency for a bare strip monopole antenna.
PRINTED STRIP MONOPOLE ANTENNAS
Figure 3.30
127
Microstripexcited planar strip monopole antenna.
Figure 3.31 Microstripexcited planar strip monopole antenna considered as an asymmetrically driven strip dipole antenna that may be separated into two strip monopole antennas.
3.3.6
MicrostripExcited Planar Strip Monopole Antenna
To analyze microstripexcited planar monopole antennas of the kind shown in Figure 3.30, we make use of an approximate expression for the impedance of an asymmetrically driven antenna [28, 29]. To this end, we consider the structure shown in Figure 3.30 as an asymmetrically driven strip dipole antenna and then separate the structure into two grounded monopole antennas as shown in Figure 3.31, where one of the monopole antennas is the strip monopole and the other monopole antenna is formed by the ground plane of the microstrip transmission line. Both strip monopole antennas can be analyzed using the theory discussed above. We need to incorporate the microstrip transmission line into the analysis to transfer the input admittance, calculated at the junction between the microstrip and the strip, to the beginning of the microstrip transmission line on the underside of the printed circuit board. We employed copper tape, a knife and a ruler for the construction of prototype antennas [30]. Using
128
PCB ANTENNAS: PRINTED MONOPOLES
this technique, we could not realize 50 microstrip lines with high accuracy.5 Since we could measure the strip dimensions with high accuracy, however, it was necessary to have an accurate, preferably analytical, model for microstrip transmission lines at our disposal to accurately account for the section of microstrip transmission line. We used the model described in [31]. 3.3.6.1 Analysis of Microstrip Transmission Line A microstrip transmission line of strip width W , thickness d, positioned on a grounded dielectric slab of height h and relative permittivity εr , is characterized by a characteristic impedance Zc and a phase constant6 β, which are defined by, respectively, η0 h , Zc = √ εeff Weff √ β = k0 εeff ,
(3.72)
(3.73) √ √ where η0 = µ0 /ε0 is the characteristic impedance of free space and k0 = ω ε0 µ0 is the freespace wave number. In equations (3.72) and (3.73), an effective width Weff and an effective relative permittivity εeff have been used. The effective width is defined by Weff (f ) =
W + (Rw + Pw )1/3 − (Rw − Pw )1/3 , 3
(3.74)
where
Sw W Weff (0) − , 2 3 2 W Sw − , Qw = 3 3 Pw =
W 3
3
+
Rw = (Pw2 + Q2w )1/2 , Sw =
c02
, 4f 2 [εeff (f ) − 1] εr − εeff (0) , εeff (f ) = εr − 1+P
(3.75) (3.76) (3.77) (3.78) (3.79)
with P = P1 P2 {(0.1844 + P3 P4 )fn }1.5763, 0.525 P1 = 0.27488 + 0.6315 + u − 0.065683e−8.7513u, (1 + 0.0157fn)20 P2 = 0.33622{1 − e−0.03442εr },
(3.80) (3.81) (3.82)
5 The accuracy with which we could realize strips with this technique was, depending on the operator, about half a
millimeter. 6 Ignoring losses for the moment.
129
PRINTED STRIP MONOPOLE ANTENNAS
P3 = 0.0363e−4.6u{1 − e(fn /38.7) P4 = 1 + 2.751{1 − e fn = fh × 10 u=
W+
−6
(εr /15.916)8
4.97
},
(3.83)
},
(3.84) (3.85)
,
(W
− W )/εr , h
(3.86)
and where c0 is the speed of light in free space. The static (f = 0) effective width is defined by Weff (0) = where and
2πh , 1 + (2h/W )}
(3.87)
ln{hF/W +
F = 6 + (2π − 6)e−(4π
2 /3)(h/W )3/4
4 d 1 + ln W = W + π (d/ h)2 +
(1/π)2 (W/t +1.1)2
(3.88) .
(3.89)
The static relative permittivity is defined by εeff (0) =
1 {εr + 1 + (εr − 1)G}, 2
(3.90)
where 10h −AB ln(4) d G= 1+ − √ , W π Wh 4 (W/ h) + (W/52h)2 W 3 1 1 ln ln 1 + , + A=1+ 49 18.7 18.1h (W/ h)4 + 0.432 and
B = 0.564e−0.2/(εr+0.3) .
(3.91) (3.92)
(3.93)
With equations (3.74)–(3.93), we can calculate the transformation from the input admittance of the monopole antenna to the connector at the side of the PCB (Figure 3.30) using the wellknown transmission line equation Yin = Yc
YL + Yc tanh(γ ) . Yc + YL tanh(γ )
(3.94)
In the above equation, Yin , the input admittance, is the admittance at the edge of the PCB. YL , the load admittance, is the input admittance of the monopole antenna at the position where the microstrip continues as a strip without a ground plane. Yc = 1/Zc is the characteristic
130
PCB ANTENNAS: PRINTED MONOPOLES
admittance of the microstrip transmission line, is the length of the transmission line and γ is the propagation constant, which is given by γ = α + jβ.
(3.95)
Here α is the attenuation coefficient, which we have ignored so far. If losses cannot be ignored, the attenuation factor is given by α = αd + αcs + αcg ,
(3.96)
where εr εeff (f ) − 1 tan(δ), εeff (f ) εr − 1 αcs = αn Rss Fs Fs , αd = 0.5β
αcg = αn Rsg Fg , πf µ0 Rss = , σs πf µ0 , Rsg = σg 32 − (W / h)2 1 4πhZ (0) 32 + (W / h)2 c αn = √ 0.667W / h εeff (0) W + 2η0 Weff (0) h W / h + 1.444 2 Fs = 1 + arctan{1.4(Rsss σs )2 }, π 2 Fg = 1 + arctan{1.4(Rsgg σg )2 }, π W − W 1 2h . F =1+ 1− + W π t
(3.97) (3.98) (3.99) (3.100) (3.101) if
W <1 h
W if ≥ 1, h
(3.102)
(3.103) (3.104) (3.105)
In equations (3.100), (3.101), (3.103) and (3.104), σg is the conductivity of the ground plane, σs is the conductivity of the strip, g is the skin depth of the ground plane and s is the skin depth of the strip.7 3.3.6.2 Results To validate the theory, a number of prototype antennas were realized in the way described at the start of section 3.3.6. To mimic mobile wireless devices, PCBs were made as shown in Figure 3.30, where monopole strip lengths of 15 mm, 30 mm and 50 mm, ground plane lengths of 30 mm, 65 mm and 100 mm and ground plane widths of 30 mm, 7 The skin depth is the distance a wave must travel in a medium for its field amplitude values to be reduced by a √ factor e−1 = 0.37. For good conductors, the skin depth may be approximated by ≈ 2/(ωµσ ) [17]. So, for 7 −1 −6 copper (σ = 5.8 × 10 S m ) at 1 GHz, the skin depth is approximately 2.1 × 10 m.
131
PRINTED STRIP MONOPOLE ANTENNAS
2
4
S11 (dB)
6
8
10
12
14
16
Model Measurement
18
20 1
1.25
1.5
1.75
2
2.25
2.5
Frequency (GHz)
Figure 3.32 Calculated and measured return loss as a function of frequency for a strip monopole antenna of length 50 mm. The ground plane length was 100 mm, and the ground plane width was 50 mm. The strip width was 3.0 mm. The relative permittivity of the substrate was 4.28, the loss tangent was 0.016 and the thickness was 1.6 mm.
40 mm and 50 mm were used. The width of the monopole strip and microstrip was 3 mm in all prototypes. The return loss as a function of frequency was calculated and measured in the frequency range from 1 GHz to 6 GHz.8 We start with two examples where the match between the calculations and measurements was poor. The results are shown in Figures 3.32 and 3.33. The figure captions provide details of the antenna dimensions. Evaluation of all of the 27 combined sets of measurements and simulations reveals that: • For a ground plane width W = 50 mm, in all situations, the correspondence between calculations and measurements is poor. If we return to the basics of the theory, i.e. the threeterm dipole model, the radius should satisfy the condition that k0 a 1 [25], where k0 = 2π/λ. Taking for the radius a quarter of the ground plane width, k0 a for W = 30 mm, 40 mm and 50 mm at 3 GHz equals 0.47, 0.63 and 0.79, respectively. None of these values satisfies the radius condition completely, but the violation of the condition becomes worse for wider ground planes. Wider ground planes (larger equivalent radii) also lead to larger input impedances. So, with increasing ground plane width, the simulation results become more inaccurate and more dominant in the total antenna input impedance. 8 We anticipate use in the frequency bands for the Global System for Mobile Communications (GSM) around 0.9 GHz and around 1.8 GHz, for the Global Positioning System (GPS) around 1.5 GHz, for Bluetooth, ZigBee and wireless local area networks around 2.4 GHz, and for future WLANs around 3.7 and 5 GHz.
132
PCB ANTENNAS: PRINTED MONOPOLES
5
S11 (dB)
10
15
20
25
Model Measurement
30 3
3.25
3.5
3.75
4
4.25
4.5
4.75
5
Frequency (GHz)
Figure 3.33 Calculated and measured return loss as a function of frequency for a strip monopole antenna of length 15 mm. The ground plane length was 30 mm, and the ground plane width was 30 mm. The strip width was 3.0 mm. The relative permittivity of the substrate was 4.28, the loss tangent was 0.016 and the thickness was 1.6 mm.
• For a monopole length Lm = 15 mm, a ground plane length Lgp = 65 mm and, to an even greater extent for Lgp = 30 mm, the measured minimum in the reflection occurs at a lower frequency than simulated. The ground plane, which is wider and longer than the strip monopole antenna, has a strong contribution to the input impedance of the antenna. Owing to the unbalanced excitation of the antenna, a ground plane that is ‘too short’ will lead to currents flowing over the outer conductor of the coaxial cable connected to the antenna. These currents effectively lengthen the ‘too short’ ground plane, thus lowering the resonance frequency. The effect is hardly visible for ground planes 100 mm long, but becomes visible for ground planes 65 mm and 30 mm long. • For monopole antennas resonant in the frequency range from 3 to 5 GHz, a ripple in the reflection coefficient as a function of frequency may be visible. For these frequencies, the rim of the ground plane (at the junction of the microstrip with the monopole strip) may reach a length close to half a wavelength and become resonant. From these observations, we may distill the following set of rough design constraints for the width of the monopole strip w and the ground plane width W : w ≤ 0.075, W W k0 ≤ 0.65, 4 c0 fmax , (3.106) W ≤ 2
133
PRINTED STRIP MONOPOLE ANTENNAS 0
2
4
S11 (dB)
6
8
10
12
14
16
Model Measurement
18
20 1
1.25
1.5
1.75
2
2.25
2.5
Frequency (GHz)
Figure 3.34 Calculated and measured return loss as a function of frequency for a strip monopole antenna of length 50 mm. The ground plane length was 100 mm, and the ground plane width was 40 mm. The strip width was 3.0 mm. The relative permittivity of the substrate was 4.28, the loss tangent was 0.016 and the thickness was 1.6 mm.
where c0 is the speed of light in free space and fmax is the maximum frequency of operation for the antenna. Some typical results (where the relative difference in resonance frequency is well below 10%), obtained following these design guidelines, are shown in Figures 3.34–3.37.9 In [29], the separation into monopoles was analyzed using a method of moments applied to an asymmetrically driven, barewire dipole antenna in free space and applied to the two monopole antennas of equal radii that the structure was separated into. From comparing the complex input impedance of the asymmetrically driven dipole antenna with the sum of the input impedances of the two monopole antennas, it was concluded that the separation led to results that were too inaccurate, and correction terms were added. The reason for the differences between the results for the dipole and the separated monopoles was considered to be due to the lack of coupling between the monopoles in the model. The reason that the separation of the antenna structure into two monopole antennas works well for a strip monopole on a dielectric slab therefore lies in the fact that the dielectric slab confines the electromagnetic field to the dipole arms, making the coupling between the two dipole arms less critical than in the barewire dipole case.
9 Observing these figures, it appears that a systematic shift in resonance frequency and return loss level has occurred
in the simulations. However, other simulation and measurement results indicate that this is not true.
134
PCB ANTENNAS: PRINTED MONOPOLES
2
4
S11 (dB)
6
8
10
12
14
16
Model Measurement
18
20 1
1.25
1.5
1.75
2
2.25
2.5
Frequency (GHz)
Figure 3.35 Calculated and measured return loss as a function of frequency for a strip monopole antenna of length 50 mm. The ground plane length was 65 mm, and the ground plane width was 40 mm. The strip width was 3.0 mm. The relative permittivity of the substrate was 4.28, the loss tangent was 0.016 and the thickness was 1.6 mm.
2
4
S11 (dB)
6
8
10
12
14
16
Model Measurement
18
20 1.5
1.75
2
2.25
2.5
2.75
3
3.25
3.5
Frequency (GHz)
Figure 3.36 Calculated and measured return loss as a function of frequency for a strip monopole antenna of length 30 mm. The ground plane length was 65 mm, and the ground plane width was 30 mm. The strip width was 3.0 mm. The relative permittivity of the substrate was 4.28, the loss tangent was 0.016 and the thickness was 1.6 mm.
135
CONCLUSIONS 0
5
S11 (dB)
10
15
20
25
30 1.5
Model Measurement
1.75
2
2.25
2.5
2.75
3
3.25
3.5
Frequency (GHz)
Figure 3.37 Calculated and measured return loss as a function of frequency for a strip monopole antenna of length 30 mm. The ground plane length was 30 mm, and the ground plane width was 40 mm. The strip width was 3.0 mm. The relative permittivity of the substrate was 4.28, the loss tangent was 0.016 and the thickness was 1.6 mm.
3.4
CONCLUSIONS
When physical reasoning in itself is not accurate enough for designing an antenna, this reasoning may be used as a starting point for designing an antenna employing, in an iterative way, a commercial offtheshelf fullwave electromagnetic analysis program. The iterative use of the program replaces the trialanderror method in which test structures are physically realized. Owing to the timeconsuming character of this design strategy, it should only be applied to oneofakind designs. As soon as it is understood that a class of antennas needs to be dealt with, i.e. similar antennas need to be designed for different frequency bands or made of different materials, it is worthwhile to invest in the development of an (approximate) antenna model. It has been shown that the design of a planar, printed UWB antenna may be realized by employing a fullwave analysis program, starting from the understanding that the antenna is a combination of a ‘fat’ dipole antenna and two tapered slot antennas. The design of a UWB antenna that incorporates filtering structures for blocking a certain frequency band may be accelerated by applying analytical models to the filtering structure. For nonUWB planar, printed dipole or monopole antennas, which may be used in various kind of wireless devices using frequencies in the range from 1 GHz to 6 GHz (GSM, GPS, Bluetooth, ZigBee, and WLAN), the development of an approximate analytical model will help in the design of these antennas. By employing a model of an equivalentradius, magnetically coated wire dipole antenna and the separation of a microstripexcited strip monopole antenna into a strip monopole antenna and a monopole antenna consisting of the microstrip ground plane, an approximate but still accurate analytical model has been created. Provided that the width of
136
PCB ANTENNAS: PRINTED MONOPOLES
the ground plane does not violate the thinwire approximation too much and that the width is also small enough to prevent resonance from the rim of the ground plane in the frequency band of interest, this model may be used to create (pre)designs of internal monopole antennas.
REFERENCES 1. H.J. Visser, ‘Lowcost, compact UWB antenna with frequency bandnotch function’, Proceedings of the 2nd European Conference on Antennas and Propagation, EuCAP2007, Edinburgh, UK, pp. 1–4, November 2007. 2. C.A. Balanis, Antenna Theory, Analysis and Design, 2nd edition, John Wiley & Sons, New York, 1996. 3. H. Schantz, The Art and Science of Ultrawideband Antennas, Artech House, Boston, MA, 2005. 4. Timederivative, available at www.timederivative.com/PUBs2centantenna.pdf. 5. T.W. Hertel, ‘Cablecurrent effects of miniature UWB antennas’, Proceedings of the IEEE Antennas and Propagation Society International Symposium, pp. 524–527, 2005. 6. E. Gueguen, F. Thudor and P. Chambelin, ‘A low cost UWB printed dipole antenna with high performances’, Proceedings of the IEEE International Conference on UWB, pp. 89–92, 2005. 7. Available at www.cst.com. 8. J. Kim, T.L. Yoon, J. Kim and J. Choi, ‘Design of an ultra wideband printed monopole antenna using FDTD and genetic algorithm’, IEEE Microwave and Wireless Components Letters, Vol. 15, No. 6, pp. 395–397, June 2005. 9. T.S.P. See and Z.N. Chen, ‘A small UWB antenna for wireless USB’, Proceedings of the IEEE International Conference on UltraWideband, ICUWB2007, pp. 198–203, 2007. 10. K.M. Kim, S.K. Park, I.S. Na and C.B. Park, ‘A planar UWB antenna with band rejection characteristic’, Proceedings of the IEEE Region 10 Conference, TENCON2007, pp. 1–4, 2007. 11. D.H. Kwon and Y. Kim, ‘Suppression of cable leakage current for edgefed printed dipole UWB antennas using leakageblocking slots’, IEEE Antennas and Wireless Propagation Letters, Vol. 5, pp. 183–186, 2006. 12. Available at www.cst.com/Content/Applications/Article/UltraWideBand+Printed+ Circular+ Dipole+Antenna. 13. N. Nguyen, C. Hsieh and D.W. Ball, ‘Millimeter wave printed circuit spurline filters’, Proceedings of the IEEE MTTS International Microwave Symposium, pp. 98–100, 1983.
REFERENCES
137
14. T.C. Edwards, Foundations for Microstrip Circuit Design, John Wiley & Sons, Chichester, 1981. 15. E. Hammerstad and Ø. Jensen, ‘Accurate models for microstrip computer aided design’, Proceedings of the IEEE MTTS International Microwave Symposium, pp. 407–409, 1980. 16. E.M.T. Jones and J.T. Bolljahn, ‘Coupledstriptransmissionline filters and directional couplers’, IRE Transactions on Microwave Theory and Techniques, pp. 75–81, April 1956. 17. D.M. Pozar, Microwave Engineering, second edition, John Wiley & Sons, New York, 1998. 18. C.T. Tai, Dyadic Green Functions in Electromagnetic Theory, second edition, IEEE Press, New York, 1994. 19. K. Sawaya, T. Ishizone and Y. Mushiake, ‘A simplified expression for the dyadic Green’s function for a conducting halfsheet’, IEEE Transactions on Antennas and Propagation, Vol. AP29, pp. 749–756, September 1981. 20. D.M. Pozar and E.H. Newman, ‘Analysis of a monopole mounted near an edge or a vertex’, IEEE Transactions on Antennas and Propagation, Vol. AP30, No. 5, pp. 401– 408, May 1982. 21. R.W.P. King and T.T. Wu, ‘The imperfectly conducting cylindrical transmitting antenna’, IEEE Transactions on Antennas and Propagation, Vol. AP14, No. 5, pp. 524– 534, September 1966. 22. R.W.P. King, C.W. Harrison and E.A. Aronson, ‘The imperfectly conducting cylindrical transmitting antenna, numerical results’, IEEE Transactions on Antennas and Propagation, Vol. AP14, No. 5, pp. 535–542, September 1966. 23. J. Moore and M.A. West, ’Simplified analysis of coated wire antennas and scatterers’, IEE Proceedings, Vol. 142, Part H, No. 1, pp. 14–18, February 1995. 24. B.D. Popovic and A. Nesic, ‘Generalisation of the concept of equivalent radius of thin cylindrical antennas’, IEE Proceedings, Vol. 131, Part H, No. 3, pp. 153–158, June 1984. 25. R.W.P. King and T.T. Wu, ‘The imperfectly conducting cylindrical transmitting antenna’, IEEE Transactions on Antennas and Propagation, Vol. AP14, No. 5, pp. 524– 534, September 1966. 26. R.W.P. King, C.W. Harrison and E.A. Aronson, ‘The imperfectly conducting cylindrical transmitting antenna, numerical results’, IEEE Transactions on Antennas and Propagation, Vol. AP14, No. 5, pp. 535–542, September 1966. 27. E. Ymash*ta and S. Yamazaki, ‘Parallelstrip line embedded in or printed on a dielectric sheet’, IEEE Transactions on Microwave Theory and Techniques, pp. 972– 973, November 1968.
138
PCB ANTENNAS: PRINTED MONOPOLES
28. R.W.P. King, ‘Asymmetrically driven antennas and the sleeve dipole’, Proceedings of the IRE, pp. 1154–1164, October 1950. 29. K. Boyle, Antennas for MultiBand RF FrontEnd Modules, PhD thesis, Delft University Press, 2004. 30. H.J. Visser, Array and Phased Array Antenna Basics, John Wiley & Sons, Chichester, UK, 2005. 31. A. van de Capelle, ‘Transmission line model for rectangular microstrip antennas’, in J.R. James and P.S. Hall (eds.), Handbook of Microstrip Antennas, Peter Peregrinus, London, pp. 527–578, 1989.
4 RFID Antennas: Folded Dipoles Wire and strip foldeddipole antennas serve as examples of lowprofile antennas that can be impedancetuned relatively easily. This impedance tuning is necessary, since the output impedance levels of communication and RFID ICs are, in general, anything but 50 . Impedance tuning of the antenna makes the application of an impedancematching network between the transceiver and the antenna unnecessary, thus aiding in the system’s compactness, power efficiency and production cost minimisation. The tuning may be accomplished by separating the transmission line mode and the dipole mode of the foldeddipole antenna and operating on the transmission line mode only. Also, the addition of a parasitic dipole element parallel to the foldeddipole antenna may be used for impedance tuning. For more directivity or a wider frequency bandwidth, several reentrant foldeddipole elements may be added in series to create a (logperiodic) foldeddipole array antenna.
4.1
INTRODUCTION
The halfwave dipole antenna is a muchused antenna type, owing to its structural simplicity, nearomnidirectional radiation pattern and ease of housing for applications using frequencies above 1 GHz. Some drawbacks of the halfwave dipole antenna, which may possibly prevent its use, are its narrow frequency bandwidth and almost fixed input impedance of approximately (73 + j42.5) [1]. The latter means that hardly any geometrical features, except for the dipole (equivalent) radius, are at the designer’s disposal for controlling the input impedance of the antenna. By placing a shortcircuited dipole antenna parallel to a dipole antenna and connecting the two dipoles at the ends (Figure 4.1(a)–(c)), a narrow loop, or foldeddipole antenna, is
Approximate Antenna Analysis for CAD
Hubregt J. Visser
© 2009 John Wiley & Sons, Ltd. ISBN: 9780470512937
140
RFID ANTENNAS: FOLDED DIPOLES
(a)
(b)
(c)
(d)
Figure 4.1 Foldeddipole antenna. (a) Single circularly cylindrical dipole antenna. (b) Dipole antenna and parallel shortcircuited dipole antenna. (c) Foldeddipole antenna. (d) Equivalentradius dipole antenna with parallel shortcircuited twowire transmission line stubs.
formed that provides more means of controlling the input impedance. At the same time, the frequency bandwidth is increased. The need for controlling the input impedance of the antenna arises from the output impedance levels of commercially available communication and radio frequency identification (RFID) ICs, which, in general, are anything but 50 .1 If the impedance matching is accomplished directly at the antenna terminals, matching networks will not be necessary and more compact, powerefficient designs become possible. The fact that the frequency bandwidth of a foldeddipole antenna is increased with respect to a singledipole antenna of the same size can be easily understood by recognizing that the folded dipole in fact is a dipole antenna with parallel shortcircuited twowire transmission line stubs (Figure 4.1(d)). To demonstrate this, we first calculate the input impedance ZA of a circularly cylindrical dipole antenna of radius a and length 2 as follows [3]: ZA = [122.65 − 204.1k + 110(k)2] 2 2 − j 120 ln − 1 cot(k) − 162.5 + 140k − 40(k) , a
(4.1)
where k = 2π/λ, λ being the wavelength, 1.3 ≤ k ≤ 1.7 and 0.001588 ≤ a/λ ≤ 0.009525. The input impedance ZT of a shortcircuited lossless transmission line stub of length l follows from the impedance transfer equation
ZL + jZ0 tan(k)
= jZ0 tan(k), (4.2) ZT = Z0 Z0 + jZL tan(k) ZL =0 1 For example, the output impedance of the NORDIC nRF905 transceiver, operated at 868 MHz, is (225 − j210)
[2].
141
INTRODUCTION
100 dipole stub
X (Ω)
50
50
100 0.9
0.95
1 frequency (in units of f0)
1.05
1.1
Figure 4.2 Imaginary part of the input impedance of a circularly cylindrical dipole of radius a = 0.002λ0 and length 2 = λ0 /2 and the input impedance divided by 100 of a shortcircuited twowire transmission line stub of length , radius a and wire separation s = 0.01λ0 , as a function of the normalized frequency.
where ZL = 0 because of the short circuit, and Z0 is the characteristic impedance of the transmission line. For a twowire transmission line where the wire radius is a and the wire separation is s, this characteristic impedance is given by [1, 4] √ s/2 + (s/2)2 − a 2 µ0 /ε0 Z0 = ln . (4.3) π a The imaginary part of the input impedance of a dipole antenna of radius a and length 2 and the input impedance of a shorted stub of length are shown in Figure 4.2 as a function of √ frequency around f0 = 1/( ε0 µ0 λ0 ), where ε0 and µ0 are the permittivity and permeability, respectively, of free space. The length is such that 2 = λ0 /2, and a = 0.002λ0 and s = 0.01λ0 . The input impedance of the stub has been divided by a factor of 100 to improve the readability of the graph. Although the figure does not show the input impedance of a foldeddipole antenna as a function of frequency, it does show how the behavior of a dipole as a function of frequency is – to a certain extent – compensated for by the stub’s behavior. For frequencies below the dipole resonance, the dipole is capacitive while the stubs are inductive, and for frequencies above the dipole resonance, the dipole is inductive while the stubs are capacitive. Thus a
142
RFID ANTENNAS: FOLDED DIPOLES
frequency bandwidth that is increased with respect to the singledipole antenna has been realized. In the remainder of this chapter, we shall outline the analysis of the input impedance of a symmetric wire foldeddipole antenna, where we shall use the concept of separating the dipole mode and the transmission line mode. This analysis is followed by an analysis of an asymmetric wire foldeddipole antenna. The asymmetric wire foldeddipole antenna provides an additional means of impedance control through the choice of the diameters of the two wires. Then, new means of impedance control will be discussed. These means consist of placing additional short circuits into the arms of the folded dipole, placing parasitic dipole or foldeddipole elements parallel to the foldeddipole antenna or both. Next, the general – i.e. asymmetric – coplanarstrip foldeddipole antenna will be discussed. This foldeddipole antenna is a practical printedcircuitboard (PCB) implementation of the foldeddipole antenna discussed in the previous sections. Design equations will be developed for an asymmetric coplanarstrip foldeddipole antenna on a dielectric slab. Following this analysis, reentrant foldeddipole antenna elements will be combined in seriesfed linear array antennas. In the previous chapter, we desired an accuracy of a few percent in the amplitude of the reflection coefficient, since we were looking at antennas to be matched to a 50 impedance level. In this chapter, we are looking at means of tuning the input impedance of an antenna to a desired complex value. So now, we require an accuracy of a few percent in both the calculated real part of the input impedance and the calculated imaginary part of the input impedance. This requirement is, as we explained in the introductory chapter, much tighter.
4.2
WIRE FOLDEDDIPOLE ANTENNAS
Before discussing the asymmetric coplanarstrip foldeddipole antenna on a dielectric slab, which is the type of most practical use for RFID and handheldtelecommunication applications, we need to understand the basic operation of the foldeddipole antenna. Therefore, we start with the wire foldeddipole antenna. First, we shall discuss the special (easier) case of the symmetric wire folded dipole, followed by a discussion of the general, i.e. asymmetric, wire foldeddipole antenna. These wire foldeddipole antennas may find applications in broadcast radio and television reception, either as standalone antennas or as exciting elements in a Yagi–Uda antenna. The analyses to be presented assume that end effects may be ignored, and that the parallel wires of the foldeddipole antenna are positioned close to each other. A centertocenter spacing, assuming circularly cylindrical wires, of 0.01λ0 is frequently mentioned [5, 6, 9], and will be used later on as a rule of thumb for closely spaced wires. 4.2.1
Symmetric FoldedDipole Antenna
To obtain the input impedance of a foldeddipole antenna of the kind shown in Figure 4.1(c), where the driven and parasitic elements are circularly cylindrical, equalradius wires, we start by connecting a voltage source V to the input terminals of the driven element. The current flowing through the folded dipole may now be decomposed into a transmission line
143
WIRE FOLDEDDIPOLE ANTENNAS
D
L
I +
V 2
V 
d Folded dipole antenna
IT
IT
+


V +2
V 2
IA
IA
2
2
+
+


V 2
d=2a Transmission line mode
Antenna mode
Figure 4.3 Decomposition of an equalradiuswire foldeddipole antenna into a transmission line mode and an antenna mode.
mode current, or differentialmode current, and an antenna mode current, or common mode current [5, 10]. Throughout this chapter, we shall use the terms ‘transmission line mode current’ and ‘antenna’ (or ‘dipole’) ‘mode current’. The decomposition into a transmission line mode and an antenna mode is shown in Figure 4.3 for a foldeddipole antenna of length L, wire radius a and wire separation D, D ≤ 0.01λ0. The figure shows that the total current flowing through the left arm of the foldeddipole antenna, I , equals IT + (1/2)IA . The input impedance therefore is Zin =
V . IT + (1/2)IA
(4.4)
Looking at the equivalent transmission line circuit, we observe that the voltage applied to the upper shortcircuited transmission line stub equals V /2. The transmission line mode current IT , therefore, may be expressed as V IT = , (4.5) 2ZT where ZT is the impedance looking into the shortcircuited twowire transmission line stub. Looking at the equivalent dipole antenna circuit, we observe two equal currents in parallel connected to the same voltage source (Figure 4.4). This figure reveals that the antenna mode current IA is given by V IA = , (4.6) 2ZD where ZD is the impedance of an ordinary dipole of equivalent radius ae and length L. The equivalent radius is given by [5, 11] D 1 ln(ae ) = ln(a) + ln . (4.7) 2 a
144
RFID ANTENNAS: FOLDED DIPOLES
V 2
IA
IA
IA
IA
2
2
2
2
+
+


V 2
V 2
+
IA
V 2

IA
IA
2
2
+ IA
Figure 4.4 Equivalent dipole antenna circuit of a foldeddipole antenna.
The substitution of equations (4.5) and (4.6) in equation (4.4) results in the following expression for the input impedance of the foldeddipole antenna: Zin =
4ZTZD . ZT + 2ZD
(4.8)
The transmission line mode impedance ZT is calculated using equations (4.2) and (4.3): L ZT = jZc tan β , (4.9) 2 where β = 2π/λ and 1 Zc = π
D + D 2 − (2a)2 µ0 ln . ε0 2a
(4.10)
The antenna mode impedance ZD may be calculated using equation (4.1) for very thin wires. For wires that are not very thin, one might employ the empirical doublepolyfit equations for the King–Middleton secondorder solution as given in [3] and, if these equations fail owing to a wire radius that is too large, one should resort to numerical methods. An interesting feature of the symmetric foldeddipole antenna can be deduced from equations (4.8) and (4.9). If the length L of the foldeddipole antenna is chosen to be equal to half the operational wavelength, the absolute value of ZT becomes infinite and the input impedance of the foldeddipole antenna becomes equal to four times the input impedance of a single halfwave dipole antenna of radius ae . A twowire transmission line, encapsulated in a ribbonshaped isolator, having a characteristic impedance of 300 is commercially available. This characteristic impedance is about four times the resistive part of the input impedance of a halfwave dipole antenna. 4.2.2
Asymmetric FoldedDipole Antenna
If the two parallel wires or elements of a foldeddipole antenna do not have equal radii, the current in the antenna will be divided asymmetrically between the two elements. It may be
145
WIRE FOLDEDDIPOLE ANTENNAS
Figure 4.5 Decomposition of an asymmetric wire foldeddipole antenna into a transmission line mode and an antenna mode.
assumed that the current is divided inversely as the ratio of the characteristic impedances of the wire elements [10], i.e. Zc1 I2 = = q. (4.11) I1 Zc2 The transmission line voltage will be divided according to the ratio of the characteristic impedances of the elements: Zc2 1 V2 (4.12) = = . V1 Zc1 q The decomposition into the transmission line mode and the antenna mode is illustrated in Figure 4.5 for an asymmetric wire foldeddipole antenna of length L, current ratio q and wire separation D, D ≤ 0.01λ0 . The radii of the wires are a1 and a2 . From this figure, we obtain the total current flowing in the left arm of the foldeddipole antenna as IT +IA . The total voltage supplied to the left arm is (1+b)V . The input impedance is therefore calculated as (1 + q)V Zin = . (4.13) IT + IA Looking at the equivalent transmission line circuit, we observe that the voltage applied to the upper shortcircuited transmission line stub equals (1/2)(1 + q)V . The transmission line mode current IT may therefore be expressed as IT =
(1 + q)V , 2ZT
(4.14)
where ZT is the impedance looking into the shortcircuited twowire transmission line stub. ZT is given by equation (4.9), where the characteristic impedance Zc of the asymmetric twowire transmission line could be calculated using equation (4.10) in which a is replaced by √ a1 a2 [11].
146
RFID ANTENNAS: FOLDED DIPOLES
The equivalent dipole antenna circuit may be regarded as two currents, IA and qIA , in parallel connected to the same voltage source V . The antenna current is therefore IA =
V . (1 + q)ZA
(4.15)
Here, ZA is the input impedance of a circularly cylindrical dipole antenna of equivalent radius ae . The equivalent radius is given by [11] ln (ae ) = ln (a1 ) +
1 (1 + a2 /a1 )2
a2 a1
2
ln
a2 a1
+2
a2 a1
ln
D a1
.
(4.16)
The impedance of the antenna may be calculated as outlined in section 4.2.1. Substitution of equations (4.14) and (4.15) in equation (4.13) results in the input impedance of an asymmetric foldeddipole antenna, expressed in terms of the stub impedance, the equivalent dipole impedance and the impedance stepup ratio: Zin =
2(1 + q)2 ZA ZT . (1 + q)2 ZA + 2ZT
(4.17)
The factor (1 + q)2 is known as the impedance stepup ratio [5, 11].
4.3
IMPEDANCE CONTROL
From a practical point of view, for wire implementations, symmetric foldeddipole antennas are preferred over asymmetric ones. Symmetric wire foldeddipole antennas can be constructed using a single piece of tubing, wire or electrically conducting thread. The latter allows the possibility of embroidering foldeddipole antennas into clothing [12]. Although we do have some means of controlling the input impedance (length, wire radius and wire separation), we would like to have additional means of controlling this impedance, without resorting to employing dielectric substrates and/or superstrates. These means are offered by placing short circuits in the arms of the foldeddipole antenna and/or by positioning parasitic dipole elements next to the antenna. To test these means of input impedance control, we have made use of a symmetric wire foldeddipole antenna, half a meter long, made out of a circularly cylindrical wire of radius 0.0001 m and with a wire separation of 0.005 m. The antenna was analyzed using the transmission line theory outlined in the previous sections. Figure 4.6 shows the real and imaginary parts of the input impedance as a function of frequency, calculated using the transmission line method (TL) and (for the propose of verification) calculated with a method of moments (MoM), employing an axialcurrent method and deltagap voltage excitation. The impedance of the dipole was calculated using the empirical doublepolyfit equations for the King–Middleton secondorder solution as given in [3], which are valid for 0.0016 m ≤ ae ≤ 0.01 m and 248 MHz ≤ f ≤ 325 MHz. We observe fair agreement between the two analysis results around resonance, even though the real part of the TL impedance differs by up to 20% from the MoM value at the highest frequency. The deviations between the TL
147
IMPEDANCE CONTROL
2000
1500
RMoM XMoM RTL XTL
R,X (Ω)
1000
500
500
1000 240
260
280
300 320 frequency (MHz)
340
360
Figure 4.6 Real part (R) and imaginary part (X) of the input impedance of a symmetric wire foldeddipole antenna as a function of frequency. Antenna dimensions: L = 0.5 m, a = 0.0001 m, D = 0.005 m.
results and MoM results at the higher frequencies within the frequency range in which the methods are valid are attributed to inaccuracies in the method of moments used. The accuracy of the transmission line method was considered to be acceptable for design purposes, as will be explained below. 4.3.1
Power Waves
Although we compare calculated input impedance results in Figure 4.6, the parameter that really matters is another one related to the impedance; this is the amount of power delivered to the antenna in the case of transmission or the amount of power taken from the antenna in the case of reception. For a system where a generator with internal impedance ZR is connected to a load ZL (Figure 4.7), the standard (voltage) reflection coefficient =
ZL − ZR , ZL + ZR
(4.18)
is a direct measure of the reflected power only when ZR is real [13]. When ZR is complex and conjugate matching is applied to maximize the power transfer from the generator to the load, the standard reflection coefficient is nonzero. Therefore, the standard definition of the reflection coefficient – based on physical waves propagating along a transmission line – does not in general represent the reflection of power. Power waves – mathematical constructs
148
RFID ANTENNAS: FOLDED DIPOLES
ZR
ZL
Figure 4.7 A source with internal impedance ZR is connected to a load impedance ZL .
from the circuit world – have been introduced to correctly describe the reflection of power in systems with complex reference impedances. Looking at the terminals of the circuit shown in Figure 4.7, we assume a nonpowerreflecting voltage wave V + propagating to the right. Associated with this voltage is a current I + . The reflected wave is defined by V − and I − . Since the wave traveling to the right is not reflected in terms of power, it must experience an impedance ZR∗ [13], where the superscript ∗ denotes the complex conjugate. The reflected wave must experience an impedance ZR . Thus V + = ZR∗ I +
(4.19)
V − = ZR I − .
(4.20)
and
The total voltage and current are given by V = V + + V −
(4.21)
I = I + − I − ,
(4.22)
and
respectively, and are related through the load impedance by V = ZL I .
(4.23)
From equations (4.19)–(4.22), it follows that ZR∗ (V + ZR I ), 2(ZR ) ZR (V − ZR∗ I ), = 2(ZR )
V + =
(4.24)
V −
(4.25)
149
IMPEDANCE CONTROL
I + =
1 (V + ZR I ), 2(ZR )
(4.26)
I − =
1 (V − ZR∗ I ), 2(ZR )
(4.27)
where (x) denotes the real part of the complex argument x. Using the power waves defined by [13] a=
V + ZR I √ , 2 (ZR )
(4.28)
b=
V − ZR∗ I √ , 2 (ZR )
(4.29)
the total voltage and current may be written as [13] V =
ZR∗ a + ZR b √ , (ZR )
a−b . I = √ (ZR )
(4.30) (4.31)
The average power delivered to the load, P , is then P =
1 a2 b2 (V I ∗ ) = − , 2 2 2
(4.32)
and the reflection coefficient for power waves is given by p =
ZL − ZR∗ b = . a ZL + ZR
(4.33)
Using the concept of power waves, we are now able to asses the influence of the deviation of the complex impedance calculated with the TL method from the values obtained with the MoM, as shown in Figure 4.6. Taking the MoM results as a reference, we assume, for every frequency, the reference impedance ZR to be equal to the complex conjugate of the MoMcalculated input impedance. Thus we shall see how the deviations of the TL values are translated into power reflection factors. In Figure 4.8, we show 20 log(p ), that is, the fraction of reflected power with respect to the maximum power that can be delivered to the load, as a function of frequency. The figure shows that the deviation from the MoM results of the input impedance calculated with the TL method results in a worstcase additional reflection of −19 dB at 325 MHz. This means that, owing to inaccuracies, an uncertainty of (much) less than 11% will be added to the power reflection. So, a deviation in the real part of the input impedance of about 20% translates to a deviation in the reflected power of about 10%. We consider this worstcase value good enough for design purposes.
150
RFID ANTENNAS: FOLDED DIPOLES
5
10
Reflection (dB)
15
20
25
30
35
40
45
50 240
260
280
300
320
340
360
frequency (MHz) Figure 4.8 Deviation of the input impedance of a foldeddipole, calculated by the TL method, with respect to the MoM results, translated into reﬂected power.
4.3.2
Short Circuits
As we have seen, the foldeddipole antenna may be analyzed by decomposing the current flowing through the antenna into a transmission line mode and an antenna or dipole mode. This decomposition opens up the possibility to act either on the transmission line mode or on the antenna mode to tune the input impedance. We may act on the transmission line mode by placing short circuits in the arms of the foldeddipole antenna (Figure 4.9). In doing so, we shorten the stub lengths but keep the dipole antenna mode intact.2 In the TL analysis, we need only to replace the stub length L/2 in equation (4.19) with L /2 (see Figure 4.9). To validate the above, a number of configurations were analyzed and compared with MoM analysis results. Figures 4.10 and 4.11 show two examples. Both example antennas were based on the foldeddipole antenna without additional short circuits (L = L in Figure 4.9) for which analysis results were shown in Figure 4.6. Figures 4.6, 4.10 and 4.11 show that indeed the additional short circuits operate mainly on the transmission line mode (there is no shift in resonance frequency). Furthermore, the figures show that the input impedance can be adjusted over a large dynamic range. Comparison with the MoM analysis results shows that the TL analysis method may also be employed for these antenna structures.
2 We shall consider only the symmetric case, i.e. the additional short circuits are placed symmetrically with respect to the horizontal foldeddipole axis.
151
IMPEDANCE CONTROL
L′
Figure 4.9
L
Foldeddipole antenna with additional short circuits.
2000
1500
RMoM XMoM RTL XTL
R,X (Ω)
1000
500
500
1000 240
260
280
300 320 frequency (MHz)
340
360
Figure 4.10 Real and imaginary parts of the input impedance of a symmetric, shortcircuited wire foldeddipole antenna as a function of frequency. Antenna dimensions: L = 0.5 m, a = 0.0001 m, D = 0.005 m, L = 0.3 m.
Note that no attempt has yet been made to tune the input impedance to any given (complex) value. The graphs are presented to show the possibility of tuning by positioning short circuits in the arms of a foldeddipole antenna. Another tuning mechanism may be provided by acting on the antenna mode instead of on the transmissionline mode. By placing parasitic elements in close proximity to a folded dipoleantenna, one may also tune the input impedance.
152
RFID ANTENNAS: FOLDED DIPOLES
500
RMoM XMoM RTL XTL
450
400
350
R,X (Ω)
300 250 200 150
100
50
0 240
260
280
320 300 frequency (MHz)
340
360
Figure 4.11 Real and imaginary parts of the input impedance of a symmetric, shortcircuited wire foldeddipole antenna as a function of frequency. Antenna dimensions: L = 0.5 m, a = 0.0001 m, D = 0.005 m, L = 0.1 m.
4.3.3
Parasitic Elements
We may tune the input impedance of a foldeddipole antenna by placing a parasitic element, for example, a shortcircuited dipole or foldeddipole antenna, parallel to and in close proximity to the foldeddipole antenna as shown in Figure 4.12. The input impedance of the driven foldeddipole antenna shown in either Figure 4.12(a) or Figure 4.12(b) is3 Zin = Z11 −
Z12 Z21 , Z22
(4.34)
where Z11 is the input impedance of the isolated foldeddipole antenna on the left in Figure 4.12(a) or (b), Z22 is the input impedance of the isolated antenna on the right in Figure 4.12(a) or (b), and Z12 = Z21 is the mutual impedance between the antenna on the left and the antenna on the right. For the situation depicted in Figure 4.12(a), the mutual impedance is given by [14] Z12 = Z21 = 2Z21dipoletodipole ,
(4.35)
3 The input impedance may be calculated from
V1 = Z11 I1 + Z12 I2 , V2 = Z21 I1 + Z22 I2 , where the subscripts 1 and 2 indicate the driven antenna on the left and the shortcircuited antenna on the right, respectively. Thus, V2 = 0. The input impedance is given by Zin = V1 /I1 .
ASYMMETRIC COPLANARSTRIP FOLDEDDIPOLE ANTENNA ON A DIELECTRIC SLAB
(a)
153
(b)
Figure 4.12 Foldeddipole antenna with parasitic elements. (a) Shortcircuited dipole coupled to a foldeddipole antenna. (b) Shortcircuited folded dipole coupled to a foldeddipole antenna.
where Z21dipoletodipole is the mutual impedance between two ‘ordinary’ dipole antennas. For thin wires, this mutual impedance may be calculated using the closedform equations given in [15]. For the situation depicted in Figure 4.12(b), the mutual impedance is given by [14] Z12 = Z21 = 4Z21dipoletodipole ,
(4.36)
where Z21dipoletodipole again is the mutual impedance between two ‘ordinary’ dipole antennas. To validate the above TL analysis method, a number of configurations were analyzed and the results were compared with those of an MoM analysis. Here, the length of the driven foldeddipole antenna is L, the length of the parasitic, shortcircuited dipole antenna is Ld and the centertocenter spacing between the foldeddipole antenna and the parasitic dipole antenna is Dc . All wires have an equal radius a (Figure 4.13). For the mutual coupling between two ‘ordinary’ dipole antennas, the closedform equations of [15] were used. Figures 4.14 and 4.15 show two examples of results for a foldeddipole antenna with a parasitic, shortcircuited dipole antenna. Again, the foldeddipole antenna is the one for which analysis results were shown in Figure 4.6. Figures 4.14 and 4.15 therefore should be compared with Figure 4.6. These figures show that the TL analysis method may be used for analyzing the configurations shown for frequencies around resonance. The figures also show that the input impedance of a (folded) dipole antenna may be adjusted by positioning parasitic elements close to the driven element, using variation of the length and the distance to the driven element to obtain the required result. Note that, again, no attempt was made to tune the input impedance to a specific value. 4.4
ASYMMETRIC COPLANARSTRIP FOLDEDDIPOLE ANTENNA ON A DIELECTRIC SLAB
The resonant foldeddipole antenna is known for its improved frequency bandwidth over that of an ‘ordinary’ dipole antenna [1]. We have seen, though, that its main attraction for
154
RFID ANTENNAS: FOLDED DIPOLES
D
L
2a
2a
Ld
Dc
Figure 4.13
Foldeddipole antenna with parasitic, shortcircuited dipole antenna.
2000
1500
RMoM XMoM RTL XTL
R,X (Ω)
1000
500
500
1000 240
260
280
300 320 frequency (MHz)
340
360
Figure 4.14 Real and imaginary parts of the input impedance of a symmetric wire foldeddipole antenna with a parasitic, shortcircuited dipole antenna, as a function of frequency. Antenna dimensions: L = 0.5 m, Ld = 0.5 m, a = 0.0001 m, D = 0.005 m, Dc = 0.01 m.
155
ASYMMETRIC COPLANARSTRIP FOLDEDDIPOLE ANTENNA ON A DIELECTRIC SLAB
2000
1500
RMoM XMoM RTL XTL
R,X (Ω)
1000
500
500
1000 240
260
280
300 320 frequency (MHz)
340
360
Figure 4.15 Real and imaginary parts of the input impedance of a symmetric wire foldeddipole antenna with a parasitic, shortcircuited dipole antenna, as a function of frequency. Antenna dimensions: L = 0.5 m, Ld = 0.4 m, a = 0.0001 m, D = 0.005 m, Dc = 0.02 m.
wireless communication and RFID applications lies in the possibility to adjust the input impedance over a wide range of values. Adjustment of the impedance characteristics has been demonstrated for freestanding wire foldeddipole antennas. We now want to adapt the impedance control measures introduced in the preceding sections to a more practical implementation of the foldeddipole antenna. This practical implementation takes the form of an asymmetric coplanarstrip (CPS) foldeddipole antenna on a dielectric slab (Figure 4.16), which allows the integration of the antenna into a PCB design. 4.4.1
Lampe Model
Design equations for the input impedance of an asymmetric coplanarstrip foldeddipole antenna were developed by Lampe [6,7]. These equations give three means of controlling the input impedance of the antenna: the input impedance of a dipole antenna of equivalent radius; the stepup impedance ratio, which depends on the widths of the two arms of the planar foldeddipole antenna; and the characteristic impedance of the coplanarstrip transmission line formed by these two arms. The input impedance of the antenna is given by [6, 7]
Zin =
2(1 + q)2ZD ZX , (1 + q)2 ZD + 2ZX
(4.37)
156
RFID ANTENNAS: FOLDED DIPOLES
C 0 C x
L
W1
d
W2
t
Figure 4.16 Asymmetric coplanarstrip foldeddipole antenna on a dielectric slab.
where ZD is the input impedance of an equivalent dipole, i.e. a circular cylindrical dipole of equivalent radius ρe . ZX is the input impedance of a shortcircuited piece of transmission line of length L/2 (Figure 4.16), and (1 + q)2 is the stepup impedance ratio. The impedance of the shortcircuited piece of asymmetric CPS of length L/2 is given by 120π K(k) L ZX = j √ tan β . (4.38) εr K (k) 2 The expression in the square brackets is the characteristic impedance of the CPS, embedded in a hom*ogeneous medium of relative permittivity εr .4 K(k) is the complete elliptic function of the first kind, and K (k) = K(k ), where k 2 = 1 − k 2 . β is the wave number in the medium. The ratio of complete elliptic functions in equation (4.22) may be approximated by [16] √ √ 1 K 1+k+ 4 k 1 for 1 ≤ ≤ ∞, √ ≤ k ≤ 1 ln 2 √ √ 4 K(k) 2π K 1+k− k 2 ≈ (4.39) 1 K 2π K(k ) √ √ ≤ 1, 0 ≤ k ≤ √ for 0 ≤ . √ √ K ln[2( 1 + k + 4 k/ 1 + k − 4 k)] 2 The argument k is given by, [6, 7] k=
(d/2)[1 + e(d/2 + W1 )] , d/2 + W1 + e(d/2)2
(4.40)
4 We assume, for the moment, that the material of the dielectric slab and the surrounding medium in Figure 4.16 are
identical.
ASYMMETRIC COPLANARSTRIP FOLDEDDIPOLE ANTENNA ON A DIELECTRIC SLAB
where e=
√ W1 W2 + (d/2)(W1 + W2 ) − W1 W2 (d + W1 )(d + W2 ) . (d/2)2 (W1 − W2 )
157
(4.41)
The stepup impedance ratio, (1 + b)2 , is calculated from b=
ln{4C + 2[(2C)2 − (W1 /2)2 ]1/2 } − ln{W1 } , ln{4C + 2[(2C)2 − (W2 /2)2 ]1/2 } − ln{W2 }
(4.42)
and the equivalent radius is given by ρe =
W1 4
1/(1+b) C+
C2
−
W2 4
2 b/(1+b) ,
(4.43)
where C is defined in Figure 4.16. As an example, the real and imaginary parts of the input impedance of an asymmetric coplanarstrip folded dipole are shown in Figure 4.17 as a function of frequency. The input impedance was calculated with a fullwave method (the finite integration (FI) technique, CST Microwave Studio© ) and with the equations given above (the transmission line (TL) method). The dimensions of the antenna were, with reference to Figure 4.16, W1 = 3 mm, W2 = 1 mm, d = 1 mm, L = 62.5 mm and εr = 1. The input impedance of the equivalent circularly cylindrical dipole was calculated by applying the empirical doublepolyfit equations for the King–Middleton secondorder solution as given in [3]. Fair agreement is shown – bearing in mind our previous discussion concerning power waves and conjugate matching – between the two simulation results around resonance. Thus, the usefulness of the Lampe model has been demonstrated. In Figure 4.18, we show FIsimulated results for a strip foldeddipole antenna of the same dimensions, but positioned on a dielectric slab of thickness t = 1.6 mm, having a relative permittivity εr = 4.28 and a loss tangent tan δ = 0.016 (FR4). The TLsimulated results for a strip foldeddipole antenna in free space are shown in the same Figure [17]. This figure shows that the input impedance of the antenna on a dielectric slab as a function of frequency is very different from that of the same antenna in free space. Therefore it is necessary to adapt the TL model to take account of the effects of the dielectric slab. The dielectric slab affects both the transmission line mode and the antenna mode. Since we expect that the dielectric slab will have a larger impact on the transmission line mode than on the antenna mode, we shall start by looking at an asymmetric coplanarstrip transmission line. The influence of the dielectric slab on the antenna or dipole mode is expected to be manifested mainly in a lower resonance frequency. 4.4.2
Asymmetric CoplanarStrip Transmission Line
Closedform equations for the characteristic impedance of asymmetric coplanarstrip transmission lines on a dielectric slab of finite thickness are not readily available. For symmetric CPS transmission lines, analytic formulas can be found in, for example, [18, 19]. In our first attempt to derive the required analytic equations, we shall modify the equation for the characteristic impedance Z0 of an asymmetric CPS in a uniform medium of relative
158
RFID ANTENNAS: FOLDED DIPOLES
1000 750 500
R,X (Ω)
250 0 250 500 RFI XFI RTL XTL
750 1000 1
1.2
1.4
1.6
1.8
2
2.2
2.4
2.6
2.8
3
frequency (GHz) Figure 4.17 Calculated (FI and TL) real and imaginary parts of the input impedance versus frequency for an asymmetric CPS foldeddipole antenna in free space. W1 = 3 mm, W2 = 1 mm, d = 1 mm, L = 62.5 mm and εr = 1.
permittivity εr (equation (4.38)): 120π K(k) . Z0 = √ εr K (k)
(4.44)
4.4.2.1 Uniform Medium First, in a very crude approximation, we shall substitute the relative permittivity of the dielectric slab for εr in equation (4.44). This means that we shall assume the coplanar strips to be present in a uniform medium with a relative permittivity equal to that of the (finitethickness) dielectric slab. The characteristic impedance was calculated in this way for several different values of the permittivity of the dielectric slab, the height, the strip separation and strip width. Results for symmetric CPS transmission lines were compared with fullwave simulation results reported in [19] (Table 4.1). In Table 4.1, t is the thickness of the dielectric slab and d is the separation of the identical strips of width W = W1 = W2 (see also Figure 4.16). The relative error is taken with respect to the fullwave value of the characteristic impedance. The table reveals that the relative difference may be as high as 32%. The impact of the approximation chosen for the determination of the characteristic impedance of the CPS on the input impedance of an asymmetric coplanarstrip foldeddipole antenna on a dielectric slab is shown in Figure 4.19. The geometrical and electrical characteristics of this antenna
159
ASYMMETRIC COPLANARSTRIP FOLDEDDIPOLE ANTENNA ON A DIELECTRIC SLAB
1000 750 500
R,X (Ω)
250 0 250 500 RFI XFI RTL free space XTL free space
750 1000 1
1.2
1.4
1.6
1.8
2
2.2
2.4
2.6
2.8
3
frequency (GHz) Figure 4.18 Calculated (FI and TL) real and imaginary parts of the input impedance versus frequency for an asymmetric CPS foldeddipole antenna on an FR4 dielectric slab and in free space. W1 = 3 mm, W2 = 1 mm, d = 1 mm, L = 62.5 mm, εr = 4.28 (FI), εr = 1 (TL) and tan δ = 0.016 (FI).
Table 4.1 Characteristic impedance of symmetric CPS transmission lines, calculated by a fullwave method and by the TL method, where, in the latter method, the dielectric slab was approximated by a uniform medium. εr 2.20 2.20 9.90 9.90 12.90 12.90 50.00 50.00
t (mm) d (mm) W (mm) Z0 , full wave () Z0 , analytic () Relative error (%) 0.79 0.79 0.64 0.64 0.25 0.25 0.25 0.25
0.10 0.30 0.04 0.37 0.026 0.15 0.031 0.030
1.52 0.76 1.27 0.51 0.38 0.13 0.20 0.025
100.07 149.79 49.91 99.98 50.00 100.05 30.03 50.03
82.60 125.56 33.99 69.84 34.23 70.51 20.71 35.63
17.46 16.18 31.90 30.15 31.54 29.53 31.04 28.78
were W1 = 3 mm, W2 = 1 mm, d = 1 mm, L = 62.5 mm, t = 1.6 mm, εr = 4.28 and tan δ = 0.016; see Figure 4.16 for the definition of the geometrical parameters.
160
RFID ANTENNAS: FOLDED DIPOLES
1000 750 500
R,X (Ω)
250 0 250 500 RFI XFI RTL epsr=4.28 XTL epsr=4.28
750 1000 1
1.2
1.4
1.6
1.8
2
2.2
2.4
2.6
2.8
3
frequency (GHz) Figure 4.19 Calculated (FI and TL) real and imaginary parts of the input impedance versus frequency for an asymmetric CPS foldeddipole antenna on an FR4 dielectric slab and in a uniform medium that has the same relative permittivity as the dielectric slab. W1 = 3 mm, W2 = 1 mm, d = 1 mm, L = 62.5 mm, εr = 4.28 and tan δ = 0.016.
It should be noted that the dielectric slab was taken into account only for the CPS transmission line mode of the antenna. As mentioned before, the effect on the antenna mode is expected to be a shift in frequency and a scaling of the levels of the impedance curves, not a change in shape. Therefore, it suffices for the moment to take only the transmission line mode into account. Although the impedance curves in Figure 4.19 show a distinct improvement – bearing the above in mind – with respect to those shown in Figure 4.18, there is still room for improvement. It is expected that much could be gained by improving the accuracy of the characteristic impedance of the CPS transmission line to well above 68%. Another reason for concentrating on the accuracy of the characteristic impedance of the CPS is that the antenna will most likely be connected to a transceiver by a length of CPS transmission line. Obviously, the approximation where the entire surroundings of the antenna are considered to consist of a uniform medium with a relative permittivity equal to that of the dielectric slab is too coarse. A refinement may be made by considering halfspaces.
161
ASYMMETRIC COPLANARSTRIP FOLDEDDIPOLE ANTENNA ON A DIELECTRIC SLAB
Table 4.2 Characteristic impedance of symmetric CPS transmission lines, calculated by a fullwave method and by the TL method, where, in the latter method, the dielectric slab was approximated by a halfspace. εr 2.20 2.20 9.90 9.90 12.90 12.90 50.00 50.00
t (mm) d (mm) W (mm) Z0 , full wave () Z0 , analytic () Relative error (%) 0.79 0.79 0.64 0.64 0.25 0.25 0.25 0.25
0.10 0.30 0.04 0.37 0.026 0.15 0.031 0.030
1.52 0.76 1.27 0.51 0.38 0.13 0.20 0.025
100.07 149.79 49.91 99.98 50.00 100.05 30.03 50.03
96.86 147.24 45.80 94.13 46.63 96.06 29.00 49.89
3.21 1.70 8.23 5.85 6.74 3.99 3.43 0.28
4.4.2.2 HalfSpaces A more realistic approximation than assuming the entire space to be filled with the slab dielectric is to assume that the dielectric slab fills up a halfspace. The antenna is then positioned at the interface between this dielectric halfspace and free space. We then replace εr in equation (4.44) with the arithmetic average of the relative permittivities of the two halfspaces on the two sides of the antenna [20], εr =
εrslab + 1 , 2
(4.45)
where εrslab is the relative permittivity of the dielectric slab. Characteristic impedances calculated for symmetric CPS transmission lines, where this effective relative permittivity was used in equation (5.44), were compared with fullwave analysis results reported in [19] (Table 4.2). The table shows that the relative error (with respect to the fullwave simulation results) is now less than 8.5%. The impact of the new approximation for the determination of the characteristic impedance of the CPS on the input impedance of an asymmetric coplanarstrip foldeddipole antenna on a dielectric slab is shown in Figure 4.20. The foldeddipole antenna was the same one as that analyzed in section 4.5.2.1. Again, it should be noted that the dielectric slab was taken into account only for the CPS transmission line mode of the antenna. Upon close inspection of Figures 4.19 and 4.20, we see that the impedance curves in Figure 4.20 calculated with the transmission line model are – apart from the anticipated shift in frequency – in closer agreement than those shown in Figure 4.19 with the fullwave simulation results shown in both figures. Our aim was to developclosed form equations for the input impedance of asymmetric coplanarstrip foldeddipole antennas with an accuracy that is sufficient to allow antenna design. Now that we have arrived at this point, the accuracy of the transmissionline model for calculating the characteristic impedance of coplanarstrip transmission lines on dielectric slabs seems to be sufficient. However, instead of moving on to the calculation of the effects of the dielectric slab on the equivalent circularly cylindrical dipole antenna, we chose to take the calculation of the characteristic impedance of the CPS one step further and improve again on the accuracy. The reason for this exercise is that, as explained before, CPS transmission
162
RFID ANTENNAS: FOLDED DIPOLES
1000 750 500
R,X (Ω)
250 0 250 500 RFI XFI RTL half space XTL half space
750 1000 1
1.2
1.4
1.6
1.8
2
2.2
2.4
2.6
2.8
3
frequency (GHz) Figure 4.20 Calculated (FI and TL) real and imaginary parts of the input impedance versus frequency for an asymmetric CPS foldeddipole antenna on an FR4 dielectric slab and on a halfspace that has the same relative permittivity as the dielectric slab. W1 = 3 mm, W2 = 1 mm, d = 1 mm, L = 62.5 mm, εr = 4.28 and tan δ = 0.016.
lines not only will be encountered within a foldeddipole antenna but also will be employed as real transmission lines between the antenna and the rest of an RFID or telecommunication system. 4.4.2.3 Analogy with Asymmetric Coplanar Waveguide In this section, we shall derive a closedform expression for the characteristic impedance of an asymmetric CPS transmission line, based on an analogy with the asymmetric coplanar waveguide (CPW). The characteristic impedance of a symmetric CPS on a dielectric slab of height t, strip width W , strip separation d and relative permittivity εr (Figure 4.16) is given by [18] 120π K(k) Z0 = √ , εeff K (k)
(4.46)
where εeff = 1 +
εr − 1 K(k ) K(k2 ) 2 K(k) K(k2 )
(4.47)
ASYMMETRIC COPLANARSTRIP FOLDEDDIPOLE ANTENNA ON A DIELECTRIC SLAB
W1 Figure 4.21
163
t d
W2
Asymmetric coplanar waveguide on a dielectric slab.
and d , d + 2W sinh (πd/4t) k2 = . sinh ((π/2t)[d/t + W ]) k=
(4.48) (4.49)
Figure 4.21 shows an asymmetric CPW with slot widths W1 and W2 and slot separation d on a dielectric slab of thickness t. The characteristic impedance of an asymmetric CPW on a halfspace dielectric slab (t → ∞) of relative permittivity εr is given by [21] 30π K (k) , Z0 = √ εeff K(k)
(4.50)
where k is given by equation (4.40) and εeff is given by equation (4.45). For a CPW on a dielectric slab of finite thickness t, the characteristic impedance is still given by equation (4.50), but εeff is now given by [21] εeff = 1 +
εr − 1 K(k ) K(k2 ) , 2 K(k) K(k2 )
(4.51)
where k is given by equation (4.40) and where WA (1 + αWB ) , WB + αWA2 πd WA = sinh , 4t k2 =
(4.52) (4.53)
164
RFID ANTENNAS: FOLDED DIPOLES
π d + W2 , 2t 2 π d + W1 WE = − sinh 2t 2
WB = sinh
and
2 WB2 WE WB WE 1 −1 − α= − − 1 − 1 . WB + WE WA2 WA2 WA2
(4.54) (4.55)
(4.56)
Given the analogy between a CPW and a coplanarstrip transmission line, we may easily transform the equations for the characteristic impedance of an asymmetric CPW on a finitethickness dielectric slab to those for a CPS. The characteristic impedance of an asymmetric coplanarstrip transmission line on a finitethickness dielectric slab is thus given by equation (4.46), where k is calculated with equations (4.40) and (4.41) and εeff is calculated with equations (4.51)–(4.56). The ratio of complete elliptic functions of the first kind is calculated with equation (4.39). To summarize, 120π K(k) , Z0 = √ εeff K (k) where εeff = 1 +
(4.57)
εr − 1 K(k ) K(k2 ) , 2 K(k) K(k2 )
(4.58)
and √ √ 1 1+k+ 4 k ln 2 √ √ K(k) 2π 1+k− 4 k ≈ 2π K(k ) √ √ √ √ 4 ln[2( 1 + k + k)/( 1 + k − 4 k)]
1 K ≤ ∞, √ ≤ k ≤ 1 K 2 1 K for 0 ≤ ≤ 1, 0 ≤ k ≤ √ , K 2 (4.59)
for 1 ≤
d/2[1 + e(d/2 + W1 )] , d/2 + W1 + e(d/2)2 √ W1 W2 + (d/2)(W1 + W2 ) − W1 W2 (d + W1 )(d + W2 ) , e= (d/2)2 (W1 − W2 ) WA (1 + αWB ) k2 = , WB + αWA2 πd , WA = sinh 4t π d + W2 , WB = sinh 2t 2 π d + W1 WE = − sinh 2t 2 k=
(4.60) (4.61) (4.62) (4.63) (4.64) (4.65)
165
ASYMMETRIC COPLANARSTRIP FOLDEDDIPOLE ANTENNA ON A DIELECTRIC SLAB
Table 4.3 Characteristic impedance of symmetric CPS transmission lines, calculated by a fullwave method and by the TL method, where, in the latter method, the dielectric slab was fully accounted for. εr 2.20 2.20 9.90 9.90 12.90 12.90 50.00 50.00
and
t (mm) d (mm) W (mm) Z0 , full wave () Z0 , analytic () Relative error (%) 0.79 0.79 0.64 0.64 0.25 0.25 0.25 0.25
0.10 0.30 0.04 0.37 0.026 0.15 0.031 0.030
1.52 0.76 1.27 0.51 0.38 0.13 0.20 0.025
100.07 149.79 49.91 99.98 50.00 100.05 30.03 50.03
100.96 151.08 49.94 99.83 49.92 99.90 29.97 49.99
0.89 0.86 0.06 0.15 0.16 0.15 0.20 0.08
2 WE WB2 1 WB WE α= − − 1 − 1 . −1 − WB + WE WA2 WA2 WA2
(4.66)
Furthermore, K (k) = K(k ), where k 2 = 1 − k 2 . The relative error in the characteristic impedance thus calculated for symmetric CPS transmission lines compared with fullwave analysis results [19] remains well below 1%, as is demonstrated in Table 4.3. The impact of the characteristic impedance of a CPS thus calculated on the input impedance of an asymmetric coplanarstrip foldeddipole antenna having the same dimensions and electrical characteristics as in the previous paragraphs is shown in Figure 4.22. Again, it should be noted that the dielectric slab has been taken into account only for the CPS transmission line mode of the antenna. The figure shows an improvement, although slight, over the results obtained with the twohalfspaces approximation (Figure 4.20).5 For designing an asymmetric coplanarstrip foldeddipole antenna on a dielectric slab, we may either use the twohalfspaces approximation for determining the characteristic impedance of the CPS or use the method described above. The latter has the disadvantage of a slightly more cumbersome implementation in software but gives excellent characteristicimpedance results; the former method has the advantage of an easy implementation in software, resulting in a fair approximation to the characteristic impedance. In the following section, we shall briefly outline the way to analyze the dipole mode to obtain results with the best accuracy for the input impedance of a foldeddipole antenna and shall discuss in more detail a less sophisticated approximate method for getting fair results, good enough for design purposes. For the latter, we may use both methods for obtaining the characteristic impedance of the CPS. We have chosen the last method described in this section.
5 Shifting the TL graphs and adjusting the levels would result in closer agreement over a wider range of frequencies for the TL graphs shown in Figure 4.20.
166
RFID ANTENNAS: FOLDED DIPOLES
1000 750 500
R,X (Ω)
250 0 250 500 RFI XFI RTL slab XTL slab
750 1000 1
1.2
1.4
1.6
1.8
2
2.2
2.4
2.6
2.8
3
frequency (GHz) Figure 4.22 Calculated (FI and TL) real and imaginary parts of the input impedance versus frequency for an asymmetric CPS foldeddipole antenna on an FR4 dielectric slab. W1 = 3 mm, W2 = 1 mm, d = 1 mm, L = 62.5 mm, εr = 4.28 and tan δ = 0.016.
4.4.3
Dipole Mode Analysis
Although it is possible to account accurately for the strip dipole on the dielectric slab, we shall develop an approximate method based on correction terms applied to the freespace analysis of a foldeddipole antenna. If higher accuracy in the end results is required, an accurate analysis of the transmission line mode of the antenna needs to be used together with an accurate analysis of the dipole mode. Such an accurate method for the analysis of the dipole mode has been used in the previous chapter, on printed monopole antennas, and will be outlined again below. 4.4.3.1 Analysis of Strip Dipole In the analysis of a foldeddipole antenna, use is made of an equivalent radius ρe (equation (4.43)). First, this equivalent radius is transformed to an equivalent strip width, using We = 4ρe [1]. To account accurately for the fact that the (equivalent) strip dipole antenna is positioned on a dielectric slab, we start with a threeterm model for a cylindrical dipole antenna that models an imperfect conductor by means of a distributed impedance [22, 23]. By virtue of this distributed impedance, it is possible to model a dielectric or magnetic coating on a cylindrical dipole antenna. To that end, a distributed inductance is substituted into the distributed impedance [24]. A strip dipole on a dielectric slab is now modeled as an equivalent magnetically coated, circularly cylindrical
ASYMMETRIC COPLANARSTRIP FOLDEDDIPOLE ANTENNA ON A DIELECTRIC SLAB
167
dipole antenna [25]. In this analysis [25], the static capacitance of a coupled strip transmission line is needed, where the strip widths are equal to the equivalent dipole strip width. This capacitance value may be calculated by the method described in [26]. Further details may be found in the previous chapter. This analysis method, however, will not be applied to the problem at hand now. Instead, we shall attempt to correct the impedance curves resulting from accounting for the dielectric slab in the transmission line mode only. This correction is accomplished by introducing correction terms applied to the freespace dipole length and equivalent radius. 4.4.3.2 Approximation for Strip Dipole We have seen that accounting for the dielectric slab in the transmission line mode of an asymmetric coplanar foldeddipole antenna led to an improvement in the impedance versus frequency curves compared with the freespace and uniformdielectric cases. The curves resemble the ones obtained from fullwave analyses, apart from a frequency shift and an overall change in impedance level. We know that one of the main effects of a dielectric on a dipole antenna will be a lowering of the resonance frequency. Therefore, we could try, by lengthening the dipole in the antenna mode part of the analysis of the folded dipole, to make the resonance frequency coincide with that obtained by fullwave analysis. Further, by increasing the equivalent radius, we could try to make the impedance levels coincide. Thus, in the dipole mode part of the analysis of the foldeddipole antenna, we may substitute L by L and ρe by ρe , where L = κL
(4.67)
ρe = τρe .
(4.68)
and
The correction factors κ and τ have been determined for a large number of asymmetric coplanarstrip foldeddipole antennas, with different dimensions, and positioned on dielectric slabs of different heights and having different relative permittivities. Figure 4.23 shows a typical example of the impedance curves thus calculated for a foldeddipole antenna together with fullwave analysis results. Figure 4.24 shows a good example of the impedance curves thus calculated. The agreement between the values of the input impedance calculated as above and the values obtained through fullwave simulations tends to get better for decreasing strip widths and strip separations. For the frequency range tested (1 GHz–6 GHz), the correction factors appear to be frequencyindependent and may be approximated by κ ≈ (1 + t × 103 log(εr ))0.45 , τ ≈ 1.90.
(4.69) (4.70)
These approximations were derived on the basis of inspection of the graphs of κ as a function of t and εr .
168
RFID ANTENNAS: FOLDED DIPOLES
1000 750 500
R,X (Ω)
250 0 250 500
RFI XFI RTL corrected XTL corrected
750 1000 1
1.2
1.4
1.8
1.6
2
2.2
2.4
2.6
2.8
3
frequency (GHz)
Figure 4.23 Calculated (FI and TL) real and imaginary parts of the input impedance versus frequency for an asymmetric CPS foldeddipole antenna on a dielectric slab. W1 = 1 mm, W2 = 1 mm, d = 0.5 mm, L = 62.5 mm, t = 1.6 mm, εr = 4.28, tan δ = 0.016, κ = 1.30 and τ = 1.90. 1000 750 500
R,X (Ω)
250 0 250 500 RFI XFI RTL corrected XTL corrected
750 1000 3
3.2 3.4 3.6 3.8
4
4.2 4.4 4.6 4.8
5
5.2 5.4 5.6 5.8
6
frequency (GHz)
Figure 4.24 Calculated (FI and TL) real and imaginary parts of the input impedance versus frequency for an asymmetric CPS foldeddipole antenna on a dielectric slab. W1 = 0.75 mm, W2 = 0.25 mm, d = 0.25 mm, L = 30 mm, t = 5.6 mm, εr = 4.28, tan δ = 0.016, κ = 1.45 and τ = 1.90.
169
FOLDEDDIPOLE ARRAY ANTENNAS
2a
L1
D Figure 4.25
4.5
L2
s
L
D
Example of a seriesfed foldeddipole array antenna.
FOLDEDDIPOLE ARRAY ANTENNAS
To broaden the frequency bandwidth achievable with a single foldeddipole antenna, one can employ multiple foldeddipole elements, each of them resonant at a different frequency. When the resonance frequencies of the individual elements are close enough together, a broad bandwidth for the array may be realized. Such an antenna may be used, for example, for broadcast applications [27] or ultrawideband (UWB) applications where OFDMlike modulation schemes are used [28]. A practical way of connecting the individual elements in the array is by series feeding, as shown in the example in Figure 4.25. This figure shows a series twoelement array of (wire) folded dipoles. The lengths of the foldeddipole elements are L1 and L2 , respectively. Both of the foldeddipole elements have a wire separation D, and the spacing between the two elements is L, as indicated in the figure. The twowire transmission line has a wire separation s, and the foldeddipole array antenna is constructed out of a single wire of radius a. The foldeddipole array antenna shown in Figure 4.25 will be used later for model verification. With appropriate scaling and separation of the foldeddipole elements, where the required 180◦ phase difference is ensured through the use of the folded dipoles, the seriesfed foldeddipole array antenna becomes a logperiodic foldeddipole array antenna, invented by John W. Greiser in 1964 [29]. Logperiodic antennas belong to the class of socalled frequencyindependent antennas, known for their wideband characteristics, which are constructed in the case of logperiodic antennas using a particular scaling of their dimensions [11]. A linear, seriesfed array of foldeddipole antennas does not consist of folded dipoles of the kind discussed before, except for the last element. The common element for these arrays is a reentrant foldeddipole antenna. A reentrant foldeddipole antenna is a foldeddipole antenna equipped with an additional port (Figure 4.26). A reentrant foldeddipole antenna may be analyzed, however, using the results derived for an ‘ordinary’ foldeddipole antenna of the kind shown in Figure 4.1(c) [30].
170
RFID ANTENNAS: FOLDED DIPOLES
V1
I1
I2
1
2
1′
V2 22′
Figure 4.26 Reentrant foldeddipole antenna: geometry and deﬁnitions of voltages and currents.
4.5.1
Reentrant FoldedDipole Antenna
Figure 4.26 shows the geometry of a reentrant foldeddipole antenna and the definitions of the voltages and currents in it. The relations between the currents and voltages are given by I1 = Y11 V1 + Y12 V2 , I2 = Y21 V1 + Y22 V2 . The admittance matrix element Y11 is calculated from
I1
1 1 Y11 = = YT + YD , V1 V2 =0 2 4
(4.71) (4.72)
(4.73)
−1 where use has been made of equation (4.8), where YT = ZT−1 and YD = ZD . The admittance matrix element Y22 follows from shortcircuiting port 1–1 in Figure 4.26 (V1 = 0) and exciting port 2–2. This results in Y22 = Y11 . (4.74)
The admittance matrix element Y12 is calculated by equalizing the twoport voltages, i.e. V1 = V2 . Only the antenna mode will be excited now. Upon substitution of equation (4.73) in equation (4.71), we find for the antenna current 1 1 YT + YD + Y12 V1 . (4.75) I1 = 2 4 From equation (4.6), we obtain the input impedance of the antenna, Yin = YD =
2I1 . V1
(4.76)
The substitution of equation (4.76) in equation (4.75) results in 1 1 Y12 = Y21 = − YT + YD . 2 4
(4.77)
171
FOLDEDDIPOLE ARRAY ANTENNAS
Note that we have assumed a symmetric wire foldeddipole antenna, meaning that the impedance stepup ratio is equal to 4 (in equation (4.17), q = 1).6 4.5.2
SeriesFed Linear Array of Folded Dipoles
If, for the moment, we neglect the effects of mutual coupling between the foldeddipole elements within the array antenna, the input impedance of the linear array antenna may be best calculated using ABCD matrices. For the nth foldeddipole element in an array consisting of N foldeddipole elements, the ABCD matrix is given by An Cn
1 Bn Y22n =− Dn Y21n Y11n Y22n − Y21n Y12n
1
Y11n
,
(4.79)
where the admittance matrix elements Yijn , i, j = 1, 2, n = 1, 2, . . . , N, are given by equations (4.73), (4.74) and (4.77). The ABCD matrix of a length of transmission line between two foldeddipole elements is given by [31] Z0p sinh (γp p ) cosh (γp p ) Ap Bp , (4.80) = Cp Dp cosh (γp p ) sinh (γp p )/Z0p for p = 1, 2, . . . , N −1. Here p is the length of transmission line p, Z0p is the characteristic impedance of this transmission line and γp is the propagation constant. The ABCD matrix of a series load YL (e.g. a short circuit, but not necessarily so) at the end of the array is given by 1 0 AL BL . (4.81) = CL DL YL 1 The complete array is characterized by
A C
B = D
2N n=1
An Cn
Bn . Dn
(4.82)
The impedance matrix of the seriesfed, linear array of folded dipoles is
Z11 Z21
1 A AD − BC Z12 = . Z22 D C 1
(4.83)
6 For an asymmetric wire foldeddipole antenna, the analysis is similar, but we have to use equation (4.17) instead of
equation (4.8) and equation (4.15) instead of equation (4.6). The final result is 1 1 1 1 Y − + Y YT + Y A T A 2 2 (1 + q)2 (1 + q)2 [Y ] = . 1 1 1 1 − YT + Y + Y Y A T A 2 2 2 2 (1 + q) (1 + q)
(4.78)
172
RFID ANTENNAS: FOLDED DIPOLES
2000
RMoM XMoM RTL XTL
1750
1500
1250
1000 R,X (Ω)
750
500 250 0 250
500
750
1000 240
260
280
300
320
340
360
frequency (MHz)
Figure 4.27 Calculated (MoM and TL) real part (R) and imaginary part (X) of the input impedance of a twoelement foldeddipole array antenna versus frequency. L1 = 0.4 m, L2 = 0.5 m, a = 0.0001 m, D = 0.005 m, s = 0.005 m, L = 0.2 m.
4.5.3
Model Veriﬁcation
To verify the model, a number of configurations of the kind shown in Figure 4.25 were analyzed using the transmission line model developed. Since we aimed only at verification of the model, no attempt was made to design an antenna to given specifications. In all calculations, the wire was assumed to be a perfect electric conductor. MoM analyses, using the axialcurrent method and a voltage deltagap feed, of the same structures using copper wires gave results nearly identical to the ones shown below. Figure 4.27 shows impedance curves calculated with the TL method and the MoM method for a twoelement foldeddipole array antenna where the separation of the folded dipoles L was 0.2 m. In Figures 4.28 and 4.29, impedance curves are shown for similar foldeddipole array antennas, but now for separation of foldeddipole elements of L = 0.4 m and 0.8 m, respectively. The figures show that the analysis results get better with increasing separation of the foldeddipole elements at the seriesfed linear array antenna. This is consistent with the fact that the effects of mutual coupling are not accounted for in the TL analysis. Furthermore, the analysis results have been shown for rather thin wires. For thick wires, the agreement with the MoM results is not as good as that in the graphs shown here. Replacement of the dipole analysis in the TL method with an analysis better fitted to handling thicker wires, for example in MoM analysis, should be capable of curing this problem. Nevertheless, the current TL method is considered accurate enough for creating initial designs for foldeddipole array antennas.
173
FOLDEDDIPOLE ARRAY ANTENNAS
2000
RMoM XMoM RTL XTL
1750
1500
1250
1000 R,X (Ω)
750
500 250 0 250
500
750
1000 240
260
280
300
320
340
360
frequency (MHz)
Figure 4.28 Calculated (MoM and TL) real part (R) and imaginary part (X) of the input impedance of a twoelement foldeddipole array antenna versus frequency. L1 = 0.4 m, L2 = 0.5 m, a = 0.0001 m, D = 0.005 m, s = 0.005 m, L = 0.4 m. 2000 RMoM XMoM RTL XTL
1750 1500 1250
1000 R,X (Ω)
750
500
250
250
500 750 1000 240
260
280
300
320
340
360
frequency (MHz)
Figure 4.29 Calculated (MoM and TL) real part (R) and imaginary part (X) of the input impedance of a twoelement foldeddipole array antenna versus frequency. L1 = 0.4 m, L2 = 0.5 m, a = 0.0001 m, D = 0.005 m, s = 0.005 m, L = 0.8 m.
174 4.5.4
RFID ANTENNAS: FOLDED DIPOLES
Inclusion of Eﬀects of Mutual Coupling
If we wish to calculate the input impedance of a linear array antenna without neglecting the effects of mutual coupling between the dipole elements, the ABCD matrix is no longer the preferred tool for accomplishing this task. Instead, the construction of an overall admittance matrix for the array antenna allows the inclusion of there effects in a relatively easy way. To include the effects of mutual coupling, the radiating elements and the feeding structure are separated as described in [32, 33]. The admittance matrix of an Nelement foldeddipole array may be written as a 2N × 2N array [Y ], [Y ] = [YF ] + [YA ],
(4.84)
where [YF ] is the admittance matrix of the feed network, i.e. the collection of interconnecting pieces of twowire transmission line, and where [YA ] is the admittance matrix of the collection of reentrant foldeddipole elements. The admittance matrix of the feed network is given by 0 0 0 ... 0 0 0 [YF1 ] 0 ... 0 0 0 0 [YF2 ] . . . 0 0 .. . . 0 0 0 0 0 (4.85) [YF ] = .. .. .. .. .. . . . ... . . 0 0 0 . . . [YF(N−1) ] 0 0 0 0 ... 0 YL In the above equation, YL is the load admittance of the array. For the array shown in Figure 4.25, which is terminated by a short circuit, YL = ∞. The 2 × 2 submatrices [YFi ], i = 1, 2, . . . , N − 1, are defined by [31] YFi11 = YFi22 = −jY0i cot(k0 li ), YFi12 = YFi21 = jY0i csc(k0 li ),
(4.86)
where Y0i and li are the characteristic admittance and length, respectively, of the piece of twowire transmission line i. The admittance matrix of the collection of reentrant foldeddipole elements is given by [YA11 ] [CA12 ] . . . [CA1N ] [CA21 ] [YA22 ] . . . [CA2N ] (4.87) [YA ] = . .. .. , .. .. . . . [CAN1 ] [CAN2 ] . . . [YANN ] where the 2 × 2 submatrices [YAii ], i = 1, 2, . . . , N, are defined by [30] 1 1 YTi + YDi , 2 4 1 1 = − YTi + YDi . 2 4
YAii11 = YAii22 = YAii12 = YAii21
(4.88)
175
FOLDEDDIPOLE ARRAY ANTENNAS
In the above equations, YTi and YDi are the transmission line stub admittance and the equivalent dipole admittance, respectively, of reentrant folded dipole i. The values are calculated using equations (4.1), (4.9) and (4.10). The 2 × 2 coupling submatrices [CAij ], i, j = 1, 2, . . . , N, i = j , contain the mutual admittances between the reentrant foldeddipole elements. They are defined by CAij11 = CAij12 = CAij21 = CAij22 =
1 Yij , 4
(4.89)
where [14] Yij = Y21dipoletodipole .
(4.90)
The mutual admittance between two thin ‘ordinary’ dipoles may be calculated using the closedform equations in [15].7 The mutual impedance between two nonstaggered thin dipoles of halflengths 1 and 2 , separated by a distance d, is given by Z12 = 30R12 + j30X12,
(4.92)
where R12 = cos{k0 (1 + 2 )} × [Ci(u0 ) + Ci(v0 ) − Ci(u1 ) − Ci(v1 ) − Ci(w1 ) − Ci(y1 ) + 2Ci(k0 d)] + cos{k0 (1 − 2 )} × [Ci(u0 ) + Ci(v0 ) − Ci(u1 ) − Ci(v1 ) − Ci(w1 ) − Ci(y1 ) + 2Ci(k0 d)] + sin{k0 (1 + 2 )} × [−Si(u0 ) + Si(v0 ) + Si(u1 ) − Si(v1 ) − Si(w1 ) + Si(y1 )] + sin{k0 (1 − 2 )} × [−Si(u0 ) + Si(v0 ) + Si(u1 ) − Si(v1 ) − Si(w1 ) − Si(y1 )]
(4.93)
7 For reentrant folded dipoles with different wire radii, the coupling submatrices are defined by
CAij
11
CAij
12
CAij
21
CAij
22
=
1 1 Yij , 1 + qi 1 + qj
=
1 1 Yij , 1 + qi 1 + 1/qj
=
1 1 Yij , 1 + 1/qi 1 + qj
=
1 1 Yij , 1 + 1/qi 1 + 1/qj
(4.91)
where qi,j is the impedance stepup ratio of the reentrant folded dipoles i and j. For equal wire radii, qi,j = 1.
176
RFID ANTENNAS: FOLDED DIPOLES
and X12 = cos{k0 (1 + 2 )} × [−Si(u0 ) − Si(v0 ) + Si(u1 ) + Si(v1 ) + Si(w1 ) + Si(y1 ) − 2Si(k0 d)] + cos{k0 (1 − 2 )} × [−Si(u0 ) − Si(v0 ) + Si(u1 ) + Si(v1 ) + Si(w1 ) + Si(y1 ) − 2Si(k0 d)] + sin{k0 (1 + 2 )} × [−Ci(u0 ) + Ci(v0 ) + Ci(u1 ) − Ci(v1 ) − Ci(w1 ) + Ci(y1 )] + sin{k0 (1 − 2 )} × [−Ci(u0 ) + Ci(v0 ) + Ci(u1 ) − Ci(v1 ) − Ci(w1 ) − Ci(y1 )].
(4.94)
In equations (4.93) and (4.94), Si(x) and Ci(x) are, respectively, the sine and cosine integrals of argument x, and $ # u0 = k0 d 2 + (1 + 2 )2 − (1 + 2 ) , # $ v0 = k0 d 2 + (1 + 2 )2 + (1 + 2 ) , # $ u0 = k0 d 2 + (1 − 2 )2 − (1 − 2 ) , $ # v0 = k0 d 2 + (1 − 2 )2 + (1 + 2 ) , # $ u1 = k0 d 2 + 21 − 1 , # $ v1 = k0 d 2 + 21 + 1 , # $ w1 = k0 d 2 + 22 + 2 , # $ y1 = k0 d 2 + 22 − 2 . (4.95) The mutual admittance is obtained from the mutual impedance through Y12 =
Z12 , ZD1 ZD2 − Z12
(4.96)
where ZD1 and ZD2 are the selfimpedances of the two coupled dipole antennas. 4.5.5
Veriﬁcation of Modeling of Mutual Coupling
The analysis of mutual coupling described in section 4.5.1 was implemented in software and several wire foldeddipole array antennas were analyzed. Comparisons were made with results obtained by a methodofmoments analysis. As a typical example, we shall show the results for a fourelement foldeddipole array antenna, shown in Figure 4.30. The array dimensions were: L1 = 11 mm, L2 = 15 mm, L3 = 19 mm, L4 = 23 mm, D = 0.15 mm, T L = 10 mm and a = 3 µm. The analysis was performed for frequencies
177
FOLDEDDIPOLE ARRAY ANTENNAS
Figure 4.30 Fourelement series array of reentrant foldeddipole antennas.
1800 MoM TL
1600
1400
1200 R (Ω)
1000
800
600
400
200
0 9
9.5
10
10.5
11
11.5
12
frequency (GHz) Figure 4.31 Real part of the input impedance as a function of frequency for a fourelement foldeddipole array antenna.
178
RFID ANTENNAS: FOLDED DIPOLES
400
200
X (Ω)
200
400
600
800 MoM TL
1000
1200 9
9.5
10
10.5
11
11.5
12
frequency (GHz) Figure 4.32 Imaginary part of the input impedance as a function of frequency for a fourelement foldeddipole array antenna.
ranging from 9 GHz to 12 GHz. In Figure 4.31, the real part of the input impedance vs. frequency is shown, as calculated with the TL method. MoM analysis results are shown for comparison. In Figure 4.32, the imaginary part of the input impedance vs. frequency is shown. These two figures show good agreement between the analysis results obtained with the TL method with inclusion of the effects of mutual coupling and the results obtained with the MoM. Although the MoM is more accurate and more versatile, the TL method, by virtue of being a dedicated method for this type of antenna, is much faster in producing analysis results. Therefore the TL method is more suited for synthesis problems, especially for generating predesigns.
4.6
CONCLUSIONS
An accurate analytical model has been derived for a linear array of wire foldeddipole antennas. The model combines closedform analytical equations for a foldeddipole antenna, a reentrant foldeddipole antenna, a twowire transmission line, the mutual coupling between two foldeddipole antennas and the mutual coupling between two thin dipole antennas. A technique in which the analysis of the antennas and that of the feeding network are separated has been successfully applied in order to combine all of these closedform equations. This analysis is expected to be extremely useful in optimization schemes to synthesize desired input impedance characteristics. To aid further in obtaining desired input impedance
REFERENCES
179
characteristics, measures for tuning the impedance of a folded dipole have been derived and verified. These measures consist of placing short circuits in the ‘arms’ of the folded dipole, placing parasitic shortcircuited dipoles or folded dipoles close to the foldeddipole antenna, or both.
REFERENCES 1. C.A. Balanis, Antenna Theory, Analysis and Design, second edition, John Wiley & Sons, New York, 1997. 2. NORDIC Semiconductor, Single Chip 433/868/915 MHz Transceiver nRF905, NORDIC Semiconductor, January 2005. 3. R.S. Elliot, Antenna Theory and Design, revised edition, John Wiley & Sons, New York, 2003. 4. C.T.A. Johnk, Engineering Electromagnetic Fields and Waves, John Wiley & Sons, 1975. 5. G.A. Thiele, E.P. Ekelman Jr. and L.W. Henderson, ‘On the accuracy of the transmission line model of the folded dipole’, IEEE Transactions on Antennas and Propagation, Vol. AP28, No. 5, pp. 700–703, September 1980. 6. R.W. Lampe, ‘Design formulas for an asymmetric coplanar strip folded dipole’, IEEE Transactions on Antennas and Propagation, Vol. AP33, No. 9, pp. 1028–1031, September 1985. 7. R.W. Lampe, ‘Correction to design formulas for an asymmetric coplanar strip folded dipole’, IEEE Transactions on Antennas and Propagation, Vol. AP34, No. 4, p. 611, April 1986. 8. A.R. Clark and A.P.C. Fourie, ‘An improvement to the transmission line model of the folded dipole’, IEE Proceedings H, Vol. 138, No. 6, pp. 577–579, December 1991. 9. J.A. Flint and J.C. Vardaxoglou, ‘Exploitation of nonradiating modes in asymmetric coplanar strip folded dipoles’, IEE Proceedings on Microwaves, Antennas and Propagation, Vol. 151, No. 4, pp. 307–310, August 2004. 10. E.C. Jordan, Electromagnetic Waves and Radiating Systems, Prentice Hall, Englewood Cliffs, 1950. 11. R.C. Johnson, Antenna Engineering Handbook, third edition, McGrawHill, New York, 1983. 12. H.J. Visser and A.C.F. Reniers, ‘Textile antennas: A practical approach’, Proceedings of the 2nd European Conference on Antennas and Propagation, Edinburgh, UK, November 2007.
180
RFID ANTENNAS: FOLDED DIPOLES
13. J. Rahola, ‘Power waves and conjugate matching’, IEEE Transactions on Circuits and Systems II: Express Briefs, (in press). 14. A.R. Clark and A.P.C. Fourie, ‘Mutual impedance and the folded dipole’, Proceedings of the 2nd International Conference on Computation in Electromagnetics, pp. 347–350, April 1994. 15. H.E. King, ‘Mutual impedance of unequal length antennas in echelon’, IRE Transactions on Antennas and Propagation, pp. 306–313, July 1957. 16. W. Hilberg, ‘From approximations to exact relations for characteristic impedances’, IEEE Transactions on Microwave Theory and Techniques, Vol. MTT17, No. 5, pp. 259– 265, May 1969. 17. H.J. Visser, ‘Improved design equations for asymmetric coplanar strip folded dipoles on a dielectric slab’, Proceedings of the 2nd European Conference on Antennas and Propagation, Edinburgh, UK, November 2007. 18. M.Y. Frankel, R.H. Voelkner and J.N. Hilfiker, ‘Coplanar transmission lines on thin substrates for highspeed lowloss propagation’, IEEE Transactions on Microwave Theory and Techniques, Vol. 42, No. 3, pp. 396–402, March 1994. 19. T.Q. Deng, M.S. Leong, P.S. Kooi and T.S. Yeo, ‘Synthesis formulas for coplanar lines in hybrid and monolithic MICs’, Electronics Letters, Vol. 32, No. 24, pp. 2253–2254, November 1996. 20. S.B. Cohn, ‘Slot line on a dielectric substrate’, IEEE Transactions on Microwave Theory and Techniques, Vol. MTT17, No. 10, pp. 768–778, October 1969. 21. V.F. Hanna and D. Thebault, ‘Theoretical and experimental verification of asymmetric coplanar waveguides’, IEEE Transactions on Microwave Theory and Techniques, Vol. MTT32, No. 12, pp. 1649–1651, December 1984. 22. R.W.P. King and T.T. Wu, ‘The imperfectly conducting cylindrical transmitting antenna’, IEEE Transactions on Antennas and Propagation, Vol. AP14, No. 5, pp. 524– 534, September 1966. 23. R.W.P. King, C.W. Harrison and E.A. Aronson, ‘The imperfectly conducting cylindrical transmitting antenna, numerical results’, IEEE Transactions on Antennas and Propagation, Vol. AP14, No. 5, pp. 535–542, September 1966. 24. J. Moore and M.A. West, ‘Simplified analysis of coated wire antennas and scatterers’, IEE Proceedings on Microwaves, Antennas and Propagation, Vol. 142, No. 1, pp. 14– 18, February 1995. 25. B.D. Popovi´c and A. Nesi´c, ‘Generalisation of the concept of equivalent radius of thin cylindrical antennas’, IEE Proceedings, Vol. 131, Part H, No. 3, pp. 153–158, June 1984.
REFERENCES
181
26. E. Ymash*ta and S. Yamazaki, ‘Parallel strip line embedded in or printed on a dielectric sheet’, IEEE Transactions on Microwave Theory and Techniques, pp. 972– 973, November 1968. 27. R. Wilensky, ‘Highfrequency antennas’, in R.C. Johnson (ed.), Antenna Engineering Handbook, third edition, McGrawHill, pp. 261–2642, 1993. 28. W. Sögel, C. Waldschnidt and W. Wiesbeck, ‘Transient responses of a Vivaldi antenna and a logarithmic periodic dipole array for ultra wideband communication’, Proceedings of the IEEE Antennas and Propagation Society International Symposium, pp. 592–595, June 2003. 29. J.W. Greiser, ‘A new class of logperiodic antennas’, Proceedings of the IEEE, pp. 617– 618, May 1964. 30. H. Shnitkin, ‘Analysis of logperiodic folded dipole array’, IEEE Antennas and Propagation International Symposium Digest, pp. 2105–2108, July 1992. 31. K.C. Gupta, R. Garg and R. Chadha, Computer Aided Design of Microwave Circuits, Artech House, Dedham, MA, 1981. 32. J. Mautz and R. Harrington, ‘Modal analysis of loaded Nport scatterers’, IEEE Transactions on Antennas and Propagation, Vol. 21, No. 2, pp. 188–199, March 1973. 33. E. Newman and J. Tehan, ‘Analysis of a microstrip array and feed network’, IEEE Transactions on Antennas and Propagation, Vol. 33, No. 4, pp. 397–403, April 1985.
5 Rectennas: Microstrip Patch Antennas Rectennas are used for converting wireless RF power into DC power. The challenge lies in maximizing the power conversion efficiency for low input power and – at the same time – minimizing the dimensions of the rectenna. By conjugately matching a rectifying circuit directly to a microstrip patch antenna, a matching and filtering network between the antenna and rectifying circuit can be avoided. With the aid of analytical models for the antenna and the rectifying circuit, singlelayer, internally matched and filtered PCB rectennas may be designed for low input power. An efficiency of 52% for 0 dBm input power has been realized at 2.45 GHz for a rectenna on a standard PCB material (FR4). A series connection of these rectennas is able to power a standard household electric wall clock. If the resistor in a Wilkinson power combiner is replaced by a rectifying circuit, the simultaneous transmission of power and data, employing the same antennas for power and data transmission, becomes feasible. A survey of expected power density levels distant from GSM900 and GSM1800 base stations, WLAN routers and GSM mobile phones reveals that a single GSM telephone can produce sufficient energy for wirelessly powering small applications at moderate distances.
5.1
INTRODUCTION
Within the framework of the miniaturization of autonomous sensors, wireless data transmission and power management are two of the key features that determine success or failure. For the purpose of minimizing the maintenance of sensors, the combination of wireless technology for both data transmission and power transmission would be an attractive feature. The reception of microwave power is maximized by enlarging the collecting aperture. However, even for small apertures, the reception of microwave power may be advantageous. Approximate Antenna Analysis for CAD
Hubregt J. Visser
© 2009 John Wiley & Sons, Ltd. ISBN: 9780470512937
184
RECTENNAS: MICROSTRIP PATCH ANTENNAS
Figure 5.1 Schematic representation of a general RF power supply.
The large duty cycles that characterize a substantial class of miniature, autonomous sensors – e.g. sensors that need to communicate a temperature once every few hours – make a miniature RF powerharvesting device, used to charge a battery or capacitor, feasible. The history of power transmission by radio waves dates back to the experiments of Heinrich Hertz in the 1880s to prove Maxwell’s theory of electromagnetism. The modern history of freespace power transmission may be considered to have its origin in the late 1950s [1] with applications in microwavepowered aircraft and the Solar Power Satellite Concept. After a period of relatively little activity in the field of freespace power transmission in the 1980s and 1990s, we can now observe a renewed interest in this field. This interest seems to have been initiated by shortrange (< 2 m [2]) radio frequency identification (RFID) applications, focusing on the available industry, science and medical (ISM) frequency bands around 0.9 GHz, 2.4 GHz, 5.8 GHz and higher. Especially for the higher frequencies, the wavelengths are small enough for the realization of miniature wireless products. The power supply for these systems may consist of an antenna coupled to a rectifying component. The combination of a rectifier and an antenna is commonly called a rectenna. A rectenna is a device to be operated in the far field of a transmitting antenna. Therefore, it is substantially different from powercollecting devices employing inductively coupled coils, which need to be positioned very close together and in line with each other. A ‘traditional’ RF power supply is shown schematically in Figure 5.1. RF power is collected by the antenna. The antenna is impedancematched to the rectifier (usually a Schottky diode or diode circuit) for maximum power transfer. The diode circuit generates a voltage that is dependent of the input power and the output load [3,4]. A capacitor is placed in parallel with the load resistor. The values of the resistor and capacitor are such that significant RF feedthrough to the output port is prevented [3]. A diode becomes a more efficient rectifier at higher input power levels. In [5], a power conversion efficiency exceeding 80% for an input power level of 20 dBm was reported for a rectenna. Whenever the use of a battery is not possible or a physical connection for externally feeding an application is not available, and distance is not critical, i.e. the radiating source will be close to the rectenna, such a highinputpowersolution is preferred. For feeding wireless applications at larger distances, however, using transmissionpowerrestricted ISM frequency bands, we encounter three main challenges. The first challenge is to maximize the power conversion efficiency for low input power levels to the rectenna. The second challenge is to minimize the dimensions of the wireless battery.
RECTENNA DESIGN IMPROVEMENTS
185
The third challenge lies in integrating antenna and rectenna operation into the same device. We want to be able to transmit power and data simultaneously over a wireless link by using the same structure for both the antenna and the rectenna. Here, we shall discuss solutions to these three challenges. In the following, we shall first discuss the design of the traditional wireless battery and same means to improve this design. Then, analytical models developed for antenna and rectifier circuits will be discussed, followed by verification of these models. A 2.45 GHz wirelessbattery design will be discussed, as well as the application of eight of these batteries in series to power a household electric wall clock. Finally, we shall discuss a twoantenna array with an RFtoDCconverting Wilkinson powercombining network, and we shall end with our conclusions.
5.2
RECTENNA DESIGN IMPROVEMENTS
In a ‘traditional’ rectenna or wireless battery, the receiving antenna is connected to the rectifying circuit through an impedancematching network (e.g. [6]). The impedancematching network, which may take the form of a transmission line and a stub, was used in [6] to match the 50 antenna impedance to the rectifying circuit. In previous work on deriving analytical models for rectenna design [7], we decided to abandon the 50 subsystem impedance and choose internal conjugate matching. To that end, we devised a method for analytically calculating the input impedance of the rectifying circuit. Then, by properly choosing the position of the input probe of a microstrip patch antenna, we were able to conjugately match the receiving antenna to the rectifying circuit. We aimed, in doing so, at improving the overall power conversion efficiency. The layout of the resulting structure is shown in Figure 5.2. The antenna and the rectifying circuit are placed on both sides of a shared ground plane. Although an impedancematching network has become unnecessary, we still observe a network between the antenna and the rectifying circuit (a diode). The purpose of this network, consisting of a transmission line and two radial stubs, is to prevent the second harmonic, generated by the diode, from being reradiated by the microstrip patch antenna. Additional radial stubs are placed between the diode and the load. One of the stubs prevents the signal at the operating frequency from being dissipated in the load, and the other prevents the signal at the second harmonic from being dissipated [7]. This rectenna may be improved with respect to complexity. The analysis may be simplified without compromising the accuracy. The first improvement with respect to complexity consists of employing a microstrip edge feed instead of a probe feed for exciting the microstrip antenna. Thus, the whole of the antenna and the rectifying circuit may be realized on a single grounded printed circuit board (PCB). Next, the rectifying circuit (again a diode) is conjugately matched directly to the microstrip antenna. Since higherorder harmonics generated by the diode will be severely mismatched to the smallband microstrip patch antenna, the patch antenna itself will act as a higherorderharmonic filter, making the transmission line stub network shown in Figure 5.2 superfluous. The impedance of the diode dictates the impedance of the microstrip patch antenna. The microstrip patch antenna will therefore not be operated at resonance.
186
RECTENNAS: MICROSTRIP PATCH ANTENNAS
Figure 5.2 Layout of a stacked rectenna. A rectangular patch is placed on top of a grounded dielectric sheet. The ground plane of the antenna acts as the ground plane of a microstrip structure that is placed at the bottom of the antenna. The microstrip patch antenna is fed by a probe. The microstrip structure couples the rectiﬁer to the antenna through an impedancematching and ﬁltering network. The rectiﬁer output is coupled to a load through a second microstrip ﬁltering network.
Finally, we may omit the filtering network between the diode and the load, provided that the performance of the rectenna system will not be seriously degraded owing to fundamental and higherorder harmonic frequency components. This assumption has to be verified of course. The improvements in the analysis consist of using a robust, yet sufficiently accurate analytical model for the diode that does not depend on numerical evaluation of an integral [7], and the use of a broadband analytical model for the microstrip patch antenna. These analytical models are the subject of the next section. That section will be followed by a section where these analysis techniques are employed for the design of a compact, efficient microstrip patch rectenna.
ANALYTICAL MODELS
5.3
187
ANALYTICAL MODELS
To conjugately match the microstrip patch antenna to the rectifying circuit, we need to be able to predict the input impedances of both devices. The input impedance of the rectifying circuit needs to be determined at the operating frequency. Once this impedance has been determined, we need to find the dimensions and excitation location of a rectangular microstrip patch antenna that will yield the conjugate value of this impedance at the operating frequency. We shall thus not use the microstrip patch antenna at resonance. As a consequence, we shall need an analytical model for the microstrip patch antenna that results in accurate impedance values over a relatively broad range of frequencies. When the analysis results for the subsystems are accurate to within a few percent, we shall be able to use our analytical models for design purposes. 5.3.1
Model of Rectangular Microstrip Patch antenna
In our recent work on rectennas [8], we applied a cavity model for a rectangular microstrip patch antenna [9, 10]. Although effective lengths and widths were employed to account for the fringe fields [10], the results proved to be too inaccurate for our needs. Therefore, in [8], we used the following method to find better effective lengths and widths. We start by employing the cavity model of [9,10] to find the dimensions and feed location that result in the desired (complex) input impedance. Then the dimensions and material characteristics of the microstrip patch antenna are used as input for a fullwave analysis program. The differences between the analysis results from the fullwave analysis program and from our cavity model are attributed to inaccurate length and width extensions in the cavity model. Next, for every radiating mode, the length and width extensions in the cavity model are determined such that the results of the cavity model analysis coincide with those of the fullwave analysis. With the ‘corrected’ length and width extensions thus found, the cavity model is applied once more to design the rectangular microstrip patch antenna that meets our requirements. As a final check, the structure thus obtained is analyzed once again with the fullwave analysis software and if differences between the results of the cavity model analysis and the fullwave analysis still exist, the procedure is repeated again. In practice, for all situations tested, one fullwave analysis iteration proved to be sufficient. The design procedure outlined above has the advantage of speeding up the design process by employing both a cavity model and a fullwave analysis program. The fast, dedicated microstrip patch antenna model is employed for the multiiteration design process, while the slow, generalpurpose, fullwave analysis program is employed for a singleiteration – or at most dualiteration – verification process. Notwithstanding this advantage, the need to employ a fullwave analysis program at all is regarded as undesirable. Ideally, the design should rely completely on fast analytical methods. That would create the possibility of automating the design process, where, through optimization, designs are created within an acceptable amount of time. An analytical model that accurately predicts the input impedance of a rectangular microstrip patch antenna over a wide band of frequencies was found in [11]. This model is based on the transmission line model of [12] and involves a series combination of transmission lines for different modes. For every mode, equivalent dimensions are introduced.
188
RECTENNAS: MICROSTRIP PATCH ANTENNAS
Figure 5.3 Geometry of a rectangular microstrip patch antenna.
Consider a rectangular microstrip patch antenna of length L and width W , as indicated in Figure 5.3. The feed is positioned at x = xf , y = yf , where the origin of the coordinate system is positioned at the bottom left corner of the patch. If the feed is a microstrip edge feed, one of the feed coordinates will be zero. The patch is positioned on a grounded dielectric slab of height h and relative permittivity εr . The input impedance of the microstrip patch antenna is given by [11]
1 1 1 1 1 Zin = + + + + Y10 Y01 Y11 Y21 Y12
−1 ,
(5.1)
where the Ymn , m, n = 0, 1, 2, are the TMmn mode admittances.1 Ymn is given by Ymn = 2Y0
Y02 + Yt2 + 2Ys Y0 coth(γ Leq ) − 2Ym Y0 csch(γ Leq ) (Y02 + Yt2 ) coth(γ Leq ) + (Y02 − Yt2 ) cosh(2γ eq )csch(γ Leq ) + 2Ys Y0
, (5.2)
where2 Yt2 = Ys2 − Ym2
(5.3)
1 The TM , TM and TM modes are harmonics of the TM , TM and TM modes, respectively and are 20 02 22 10 01 11
already accounted for in the transmission line model [11]. 2 In [11], Y 2 was erroneously reproduced from [12]. The plus sign should be a minus sign. t
189
ANALYTICAL MODELS
and Leq = √ eq =
xoff yoff
LW + n2 L2 xoff yoff
m2 W 2
(5.4)
,
m2 2yoff + n2 2xoff
L
= − xf
, 2
W
= − yf
. 2
(5.5)
,
(5.6) (5.7)
In equation (5.1), Y0 is the characteristic impedance of a microstrip transmission line of width W , and γ = α + jβ is the propagation constant. The microstrip patch antenna is regarded as a microstrip transmission line of length L, connecting two radiating slots that operate by virtue of fringe fields at the edges of the patch. Ys in equation (6.2) is the selfadmittance of such a radiating slot, and Ym is the mutual admittance between the slots. The characteristic admittance is given by [13] ε0 √ Weff , (5.8) Y0 = εreff µ0 h and the phase constant is
√ √ β = ω ε0 µ0 εreff .
(5.9)
To calculate the effective width as a function of frequency, Weff , and the effective relative permittivity as a function of frequency, εreff , we first need to calculate the static effective width Weff (0) and static effective relative permittivity εreff (0). These static parameters have been calculated as follows [13]: Weff (0) = where and
2πh , ln{hF/W + 1 + (2h/W )2 }
F = 6 + (2π − 6)e−(4π
(5.10)
2 /3)(h/W )3/4
4 th 1 + ln W =W+ π (t/ h)2 +
(5.11)
(1/π)2 (W/t +1.1)2
.
(5.12)
In the last equation, th is the thickness of the metal patch. Also, εreff (0) = where
1 {εr + 1 + (εr − 1)G}, 2
10h −AB ln(4) t − √ G= 1+ W π Wh
(5.13)
(5.14)
190
RECTENNAS: MICROSTRIP PATCH ANTENNAS
and 3 1 (W/ h)4 + (W/52h)2 W 1 A=1+ ln ln 1 + , + 49 (W/ h)4 + 0.432 18.7 18.1h
(5.15)
B = 0.564e−0.2/(εr +0.3) .
(5.16)
The frequencydependent effective relative permittivity is then calculated as [13] εreff (f ) = εr −
εr − εreff (0) , 1+P
(5.17)
where P = P1 P2 {(0.1844 + P3 P4 )fn }1.5763, 0.525 P1 = 0.27488 + 0.6315 + u − 0.065683e−8.7513u, (1 + 0.0157fn)20
(5.18)
P2 = 0.33622{1 − e−0.03442εr }, P3 = 0.0363e−4.6u{1 − e−(fn /38.7) P4 = 1 + 2.751{1 − e fn = fh × 10 u=
W
−6
−(εr /15.916)8
4.97
},
},
,
+ (W
− W )/εr . h
(5.19)
The frequencydependent effective width is calculated as [13] Weff (f ) =
W + (RW + PW )1/3 − (RW − PW )1/3 , 3
(5.20)
where SW W Weff (0) − , 2 3 2 W SW − = , 3 3 2 + Q3 , = PW W
PW = QW RW
SW =
W 3
3
+
c02 , 4f 2 [εeff (f ) − 1]
(5.21)
c0 being the speed of light in free space. The selfadmittance of the radiating slot is given by [12, 13] Ys = Gs + jBs ,
(5.22)
191
ANALYTICAL MODELS
where Gs =
s2 sin(w) 1 + cos(w) − 2 1 − wSi(w) + √ w 24 π µ0 /ε0 2 s 1 cos(w) sin(w) + + − 12 3 w2 w3
(5.23)
and √
Bs = Y0 tan(βl).
(5.24)
In the above equations, w = ω ε0 µ0 Weff , and l is the width of the radiating slot. l is given by ξ1 ξ3 ξ5 l = h , (5.25) ξ4 where ξ1 = 0.434907
εr0.81 + 0.26 (W/ h)0.8544 + 0.236 eff , εr0.81 − 0.189 (W/ h)0.8544 + 0.87 eff
(W/ h)0.371 , 2.358εr + 1 0.5274 arctan{0.084(W/ h)1.9413/ξ2 } ξ3 = 1 + , εr0.9236 eff 1.456 W {6 − 5e0.036(1−εr) }, ξ4 = 1 + 0.0377 arctan 0.067 h ξ2 = 1 +
ξ5 = 1 − 0.218e−7.5W/ h.
(5.26)
The mutual admittance between the radiating slots is given by [12, 13] Ym = Gm + jBm ,
(5.27)
Gm = Gs Fg Kg , Bm = Bs Fs Kb
(5.28)
where
and s2 J2 (l), 24 − s 2 Y0 (l) + (s 2 /(24 − s 2 ))Y2 (l) π , Fb = 2 ln(s/2) + Ce − 3/2 + (s 2 /12)/(24 − s 2 ) Kg = 1, Fg = J0 (l) +
Kb = 1 − e−0.21w .
(5.29)
(5.30)
192
RECTENNAS: MICROSTRIP PATCH ANTENNAS
√ √ In the above, s = ω ε0 µ0 l, l = ω ε0 µ0 (L + l) and Ce is Euler’s constant, equal to 0.57216 . . . Ji (x) is the ithorder Bessel function of the first kind, and Yi (x) is the ithorder Bessel function of the second kind. Finally, the attenuation constant α is given by α = αd + αcs + αcg ,
(5.31)
where αd = 0.5β
εr εreff (f ) − 1 tan δ, εreff (f ) εr − 1
αcs = αn Rss Fs Fs , αcg = αn Rsg Fg . In the above, tan δ is the loss tangent of the dielectric. Further, πf µ0 Rss = , σs πf µ0 , Rsg = σg
(5.32)
(5.33)
where σs and σg are the conductance of the patch and of the ground plane, respectively. In addition, Y0 32 − (W / h)2 W <1 for 4πh 32 + (W / h)2 h αn = εreff (0) W 0.667W / h W √ for + ≥ 1, W / h + 1.444 h 2 (µ0 /ε0 )Weff (0) h 2 Fs = 1 + arctan{1.4(Rsss σs )2 }, π 2 Fg = 1 + arctan{1.4(Rsgg σg )2 }, π W − W 2h 1 Fs = 1 + 1 − + , (5.34) W π t where s and g are the rms surface roughness of the patch and of the ground plane, respectively. To demonstrate the accuracy of the ‘modified’ cavity model (i.e. using a single fullwave iteration to correct the length extension of the cavity model) and the transmission line model of [11], we calculated the input impedance of a dualresonant microstrip patch antenna. For such an antenna, the ‘standard’ cavity model will fail to accurately predict the input impedance for frequencies below and above the first resonance frequency [14]. The parameters used for the microstrip patch antenna were L = 30.8 mm, W = 27.7 mm, h = 1.6 mm, εr = 4.28, tan δ = 0.016, xf = 0, yf = 0.4 mm, th = 0.070 mm,
193
ANALYTICAL MODELS
250 MoM mod. Cavity TL
200
R (Ω)
150
100
50
0 2
2.1
2.2
2.3
2.4
2.5
2.6
2.7
2.8
Frequency (GHz)
Figure 5.4 Real part of the input impedance versus frequency, calculated with three diﬀerent models for a dualresonant rectangular microstrip patch antenna.
σs = σg = 0.556 × 108 S m−1 and s = g = 0.0015 mm. Figure 5.4 shows the real part of the input impedance as a function of frequency, calculated with the method of moments (MoM), with the modified cavity model and with the transmission line model. Figure 5.5 shows the calculated results for the imaginary part of the input impedance. In comparing the calculated input impedance results, we shall take the MoM analysis results as a reference. Over recent decades, commercial offtheshelf (COTS) MoM analysis software has become very reliable, and its results in general agree very well with measurements performed on real structures such as that under investigation here. Figures 5.4 and 5.5 show that both for the modified cavity model and for the transmission line model, the real part of the input impedance is within a few percent of the reference value around 2.45 GHz. The imaginary part of the input impedance differs by up to 50% of the reference value around this frequency. Notwithstanding this large difference in the imaginary part of the input impedance, the powerwave reflection coefficient still is of the order of 0.20, an acceptable value for design purposes. Both the modified cavity model and the transmission line model thus generate results that are in good agreement with the fullwave analysis results. The modified cavity model is more accurate but has the drawback that it relies on a fullwave analysis program. The modified cavity model and the transmission line model have comparable accuracy in the region of interest (i.e. off resonance). 5.3.2 Model of Rectifying Circuit The main component of a rectifying circuit is a diode. We may use a single diode in the rectifying circuit or use two diodes in a voltage doubler configuration [5]. The diode or diodes
194
RECTENNAS: MICROSTRIP PATCH ANTENNAS
200 MoM mod. Cavity TL
150 100
X (Ω)
50 0 50 100 150 2
2.1
2.2
2.3
2.4
2.5
2.6
2.7
2.8
Frequency (GHz)
Figure 5.5 Imaginary part of the input impedance versus frequency, calculated with three diﬀerent models for a dualresonant rectangular microstrip patch antenna.
Figure 5.6 Equivalent electric circuit of a diode.
are used unbiased, since we have no power available other than the RF power delivered by the antenna. The diode considered here is of the Schottky type, because of its low forward bias voltage compared with other diode types. The diode can be represented by the equivalent electric circuit shown in Figure 5.6. A Schottky diode consists of a semiconductor substrate attached to a metal contact. In the electric circuit in Figure 5.6, the resistance Rs represents the conduction losses in the substrate. The depletion layer in the substrate forms an insulating barrier between the two diode contacts and thus forms a junction capacitance Cj . The value of Cj depends on the
195
ANALYTICAL MODELS
Figure 5.7 Equivalent electric circuit of a packaged diode.
applied voltage vd according to [15] Cj 0 , Cj (vd ) = √ 1 − vd /φ
(5.35)
where vd is the voltage across the (bare) diode (Figure 5.6), Cj 0 is the zerobias differential barrier capacitance and φ is the barrier potential of the diode. Values for the latter two parameters may be found in a diode’s data sheet. The nonlinear voltage–current characteristic of the diode in Figure 5.6 is given by id = Is (e(q/nkT)vd − 1),
(5.36)
where id is the current through the (nonlinear) diode, Is is the saturation current, q is the electron charge, k is Boltzmann’s constant, equal to 1.3806504 × 10−23 J K−1 , T is the temperature in kelvin and n is the ideality factor of the diode. The ideality factor, a number close to one for most diodes, may be found in a diode’s data sheet as well. The above applies to the bare diode. In our rectennas, we have made use of packaged diodes. The equivalent electric circuit of a packaged diode is shown in Figure 5.7. In this circuit, we see an additional parasitic capacitance Cp and a parasitic inductance Lp , the values of which may be found in the diode’s data sheet. Several methods exist for analyzing the nonlinear behavior of a diode. Linearization of the voltage–current characteristic around a working point will not suffice for our purposes. We shall be using the diode in the largesignal region, meaning that the nonlinear behavior will show in the working region and a linearization therefore cannot be performed unambiguously. In a ‘harmonic balance’ analysis, the circuit is separated into a linear and a nonlinear subcircuit. The linear subcircuit is analyzed in the frequency domain, while the nonlinear subcircuit is analyzed in the time domain, after which the results of the latter analysis are Fourier transformed to the frequency domain. A solution is found if the currents through the interconnections between the linear and nonlinear subcircuits are the same for both subcircuits. The currents in the linear and nonlinear subcircuits through these interconnections have to be balanced at every harmonic. Application of an efficient harmonic balance method for analyzing a nonlinear diode has proven to generate accurate results [7]. A drawback of this method is that for more than one higher harmonic, closedform analytical solutions are no longer available.
196
RECTENNAS: MICROSTRIP PATCH ANTENNAS
Figure 5.8 Circuit for determining the impedance of a packaged diode.
For this reason, we chose to use a ‘timemarching’ algorithm, which is robust and, for our purposes, where only one nonlinear element needs to be analyzed, still timeefficient. In this method, the system is first discretized into a set of nonlinear firstorder ordinary differential equations. A starting condition is chosen and a (timedomain) input signal is impressed on the system. An integral over the differential equations, using a small time step, is then applied to determine the system state after this small time step, i.e. yn+1 = yn + t · f (tn , yn ),
(5.37)
where yn is the system state at time t = tn , yn+1 is the system state at t = tn+1 and the function f (tn , yn ) represents the derivative of y with respect to t at t = tn . t is the time step, i.e. t = tn+1 − tn . To determine the impedance of the diode, the packaged diode is embedded in the circuit shown in Figure 5.8. In this figure, Rg is the generator impedance, and the source voltage Vg is given by Vg = Vg  cos(ωt).
(5.38)
The voltages and currents in the circuit shown in Figure 5.8 are described by the following circuit equations: Vg = Ig Rg + Lp
dIg + VC p , dt
VCp = Vd + VRs , VRs = Rs IRs = (ICj + Id )Rs ,
197
ANALYTICAL MODELS
dVd , dt Id = Is (e(q/nkT)Vd − 1),
ICj = Cj
Ig = Id + ICj + ICp , dVCp . ICp = Cp dt
(5.39)
From these equations, we easily obtain the following set of coupled firstorder ordinary differential equations: dI g dV d + Vd + Rg Ig + Rs Cj + Rs Is (e(q/nkT)Vd − 1), dt dt dV g dI g dV d dI X Ig = Is (e(q/nkT)Vd − 1) + Cj + Cp − Cp Rg − Cp Lp , dt dt dt dt
Vg = Lp
(5.40)
where IX =
dIg dt
(5.41)
has been introduced to avoid the occurrence of secondorder derivatives. Vd and Ig are the unknowns. The first step in solving an ordinary differential equation usually consists of rewriting the equation as a set of coupled firstorder differential equations [16]. Having accomplished this, in the next step we rewrite the set of coupled firstorder ordinary differential equations in a form wherein every unknown derivative is expressed in terms of known parameters: 1 dV d = [Vg − Rg Ig − Vd − Lp IX − Rs Is (e(q/nkT)Vd − 1)], dt Rs Cj dVg 1 dV d dI X = + Cp − Cp Rg IX − Ig ], [Is (e(q/nkT)Vd − 1) + Cj dt Cp Lp dt dt dI g = IX . dt
(5.42)
The system parameters Vd , Ig and IX are known at t = t0 , as is dV g /dt, the time derivative of the impressed generator voltage (at t = t0 , we choose Vd = Ig = IX = dVg /dt = 0). A fourthorder Runge–Kutta method (RK4) [16] was applied to step through time. To make the method robust, an adaptivestepsize algorithm was applied. Timestepping results are shown in Figure 5.9 for an Agilent HSMS2852 Schottky diode [3]. In this figure, the diode voltage Vd is shown, together with the generator voltage Vg . The frequency of the generator was 2.45 GHz and the input power was 0 dBm. By applying RK4, we may find the diode junction voltage and the generator current and, from equation (6.40), the diode current. After applying a fast Fourier transform (FFT) to the diode voltage and current, the diode impedance is given, for each harmonic of the working frequency, by Vd Zdf = f . (5.43) Idf
198
RECTENNAS: MICROSTRIP PATCH ANTENNAS
0.8
0.6
Vg Vd
Vd, Vg (V )
0.4 0.2 0 0.2
0.4
0.6
0.8 1.25
1.35
1.45
1.55
1.65 1.75 time (ns)
1.85
1.95
2.05
Figure 5.9 RK4 simulation results for an Agilent HSMS2852 Schottky diode.
The incident power follows from the applied generator voltage and the generator impedance: Pinc =
Vg 2 . 8Rg
(5.44)
We now have an efficient analytical timedomain method to determine the voltages and currents in a packaged diode circuit, which – after Fourier transformation – gives us the diode circuit’s input impedance as a function of frequency. With an accurate transmission line model, we may determine the complex input impedance of a rectangular microstrip patch antenna as a function of frequency for any position of the excitation of the patch. So, we have the tools for, first, determining the rectifying circuit’s input impedance and, second, determining the feed position of the microstrip patch for creating a conjugate impedance match to this circuit. Before we make our first design, we shall first verify the rectifier circuit model.
5.4
MODEL VERIFICATION
We have already used fullwave analysis results for a microstrip patch antenna, using (proven) COTS software, to verify the analytical models that we have developed. For the diode, the analytical models were verified by comparison with measurement results at the operating frequency and by comparison with results of COTS harmonic balance simulation for higher harmonics.
MODEL VERIFICATION
199
Figure 5.10 Setup for measurement of the impedance of a diode.
First, the impedance of an Agilent HSMS2852 Schottky diode was calculated and measured, at operational frequencies in the range from 100 MHz to 4.1 GHz and for different input power levels. To that end, a packaged diode was mounted directly on a subminiature versionA (SMA) connector. This SMA connector was connected to the input port of a vector network analyzer (VNA) that was equipped with a DC block and a 50 load at the DC output port, of the VNA. The measurement setup is shown in Figure 5.10. Bearing in mind that the VNA is calibrated so that the reference plane is at the input of the packaged diode and that the RF choke isolates the DC from the AC signal, the measurement setup of Figure 5.10 is identical to that shown in Figure 5.8. Therefore, the calculated results may be compared directly with the measured results. However, the HSMS2852 in fact consists of two diodes, sharing a common pin of the threepin SOT23 package [17], where the anode of one diode and the cathode of the other are connected. When one diode is measured, the other diode, which is connected on one side and floats at the other side, will give rise to a capacitive coupling. The value of the coupled capacitor was estimated to be 0.3 pF. The value of the parasitic capacitance of a (single) diode was Cp = 0.08 pF. The other parameters were Cj 0 = 0.18 pF, Lp = 2 nH, Rs = 25 , Is = 3 µA, n = 1.06 and φ = 0.35 V [17]. The input impedance obtained from the measured reflection coefficient, corrected for the 0.3 pF capacitive coupling, is shown in Figures 5.11 and 5.12 together with the calculated results for an input power of 0 dBm. Figure 5.11 shows the real part of the input impedance, and figure 5.12 shows the imaginary part. Other input powers in the range from −10 dBm to +10 dBm show similar results. The ‘noise’ that is visible in the measurement data can be explained by the fact that the VNA was not able to fully suppress the higher harmonics at its input. Nevertheless, the figures show that with the RK4 analysis, we are able to predict the input impedance of the packaged diode to within a few percent at frequencies around 2.45 GHz and higher.
200
RECTENNAS: MICROSTRIP PATCH ANTENNAS
220 200
measurement model
180 160
R (Ω)
140 120 100 80 60 40 20 0 0.1
0.5
0.9
1.3
1.7 2.1 2.5 Frequency (GHz)
2.9
3.3
3.7
4.1
Figure 5.11 Real part of measured and simulated input impedance versus frequency for an Agilent HSMS2852 Schottky diode, for 0 dBm input power.
Since we did not have access to equipment for measuring the input impedance of the diode at higher harmonics, we compared the results calculated with RK4 with results obtained from AWR’s harmonic balance analysis program APLAC . In Figure 5.13, the voltage Vd is shown for the harmonic frequencies, as calculated by our RK4 method and as obtained from a harmonic balance analysis for a 0 dBm input signal at 2.45 GHz. We see, for the higher harmonics also, results that are within a few percent of the expected values. For the problem at hand, which results in a limited set of nonlinear ordinary differential equations, the Runge–Kutta method proves to be computationally more efficient (by one order of magnitude) than the harmonic balance method, while yielding results of comparable accuracy. Now that we have verified the models of the rectangular microstrip patch antenna and of the diode, we have completed developing the tools for designing a wireless battery.
5.5
WIRELESS BATTERY
The analysis tools described above were used to design a single rectenna, or wireless battery. After realization and characterization, this wireless battery was applied in cascade to power an electric wall clock.
201
WIRELESS BATTERY
0 10
measurement model
20 30
X (Ω)
40 50 60 70 80 90 100 0.1
0.5
0.9
1.3
1.7 2.1 2.5 Frequency (GHz)
2.9
3.3
3.7
4.1
Figure 5.12 Imaginary part of measured and simulated input impedance versus frequency for an Agilent HSMS2852 Schottky diode, for 0 dBm input power.
500 450 400 Diode voltage (mV)
350 300
Re Harm. bal. Im Harm. bal. Re RK4 Im RK4
250 200 150 100 50 0 50 100 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8 Frequency (GHz)
Figure 5.13 Real and imaginary parts of the diode voltage as calculated by RK4 and harmonic balance for a 0 dBm input signal at 2.45 GHz for a packaged Agilent HSMS2852 diode.
202
RECTENNAS: MICROSTRIP PATCH ANTENNAS
Figure 5.14 Voltage doubler diode rectifying circuit. C1 aids in creating a DC voltage across D2 . In the rectenna, the microstrip patch antenna serves this purpose.
5.5.1
Single Rectenna
In the single rectenna designed, we made use of two diodes in a voltage doubler configuration (Figure 5.14) [15]. From Figure 5.14, we may observe that at microwave frequencies, the capacitors will behave as short circuits and, consequently, the input impedance of the rectifying circuit will be that of an antiparallel pair of diodes. Hence, the impedance will be half of that of a single diode. For DC, the capacitor will behave as an open circuit and the two diodes will now act as sources that are connected in series. The output voltage is therefore doubled in comparison with a singlediode rectifying circuit. The design of a wireless battery starts by deciding on the operating frequency, in our case 2.45 GHz. Next, the input impedance of the diode to be used – alone or in a voltage doubler configuration – is determined for a chosen input power level, and the DC load impedance is chosen. The input power level was chosen to be 0 dBm, a value considered to be representative for the power level received by a microstrip patch antenna in an environment containing one or more ISM radio sources. With the aid of Figures 5.11 and 5.12, the input impedance of an Agilent HSMS2852 diode under these conditions is found to be (80 − j90) . The RF input impedance for a voltage doubler configuration thus equals (40 − j45) . Next, a rectangular microstrip patch antenna with a microstrip edge feed was designed for an input impedance of (40 + j45) at 2.45 GHz. The real and imaginary parts of the input impedance as a function of frequency are shown in Figures 5.4 and 5.5. The values show that the dimensions of the rectangular microstrip patch antenna were such that a conjugate match with the voltagedoubling circuit had been realized. Thus the need for an impedancematching network was circumvented. Also, a filtering structure for preventing the reradiation of higher harmonics was not necessary, since the microstrip patch antenna was mismatched to these higher harmonic signals. Finally, to suppress the presence of the fundamental frequency (and higher harmonics) in the output signal, it is common practice to use a microstripline stub filter between the rectifying circuit and the load (Figure 5.15(a)). Since the capacitor C in the voltage doubler circuit (Figure 5.14), takes over this role, this filtering network could be omitted here. Measurements showed that this hardly degrades the performance of the system. So, finally, with the structure shown in Figure 5.15(b), a wireless battery that was hardly larger than the microstrip patch antenna was realized.
203
WIRELESS BATTERY
Figure 5.15 Singlelayer, grounded PCB rectenna consisting of a microstrip patch antenna, a diode voltage doubler and an SMD capacitor, connected directly to the edge of the patch. (a) With ﬁltering structure, (b) without ﬁltering structure.
5.5.2
Characterization of Rectenna
To verify the rectenna design, the unloaded output voltage as a function of the input power was calculated and measured. Since the input power coupled into the rectifying circuit was not easily accessible, a relative measurement was performed. A poweradjustable transmitter was connected to a directive antenna. The rectenna was placed at a fixed distance, and the unloaded output voltage was measured as a function of the transmitter power. For one transmitter power level, the input power to the rectifying circuit that would result in the observed unloaded output voltage was calculated. For all other transmitter power levels, the corresponding input power levels were scaled relative to this calculated value (i.e. if the transmitter power level increases by 10 dB, the input power to the rectifying circuit increases by the same amount). The results at 2.45 GHz calculated and measured in this way are shown in Figure 5.16. First of all, the excellent agreement between the measured and calculated output voltages – the differences are within a few percent – demonstrates the validity of the design tools and of the design itself. Furthermore, it shows that for a highimpedance load, relatively high voltages may be obtained for input powers at or below 0 dBm. To determine the efficiency of the rectenna, the distance between the transmitter and rectenna was first determined for 0 dBm input power. The Friis transmission equation was used to this end, with the gain of the transmitter antenna and the gain of the rectangular microstrip antenna as input. Then the (resistive) load was varied to maximize the dissipated power. The maximum dissipated power turned out to be 516 µW for a 900 load. The efficiency η of the rectenna was then given by η=
516 PDC × 100% = 52%, × 100% = Pinc 1000
(5.45)
where Pinc , the input power, is 1 mW. The efficiency of this rectenna has improved by more than 10% compared with a rectenna with a filtering structure [7] evaluated at the same input power level, and the structure has become more compact.
204
RECTENNAS: MICROSTRIP PATCH ANTENNAS
0.7 0.6
Measurement Model
Vout (V )
0.5 0.4 0.3 0.2 0.1 0 35
Figure 5.16
5.5.3
30
25
15 20 Pinc (dBm)
10
5
Unloaded voltage doubler output voltage as a function of input power.
Cascaded Rectennas
Next, it was decided to use the wireless battery in a practical application. As may be concluded from Figure 5.16, the output voltage of the wireless battery decreases with the received power and thus with distance from the RF source. By placing several batteries in series, an acceptable output voltage can be obtained at an acceptable distance from the RF source. Then, to aid further in applicability, it was decided to use a series connection of rectennas to power a lowpower, lowdutycycle application. Here, power is collected by the wireless batteries and the application employs the collected power only at discrete time intervals. Such an application was found in a common household electric wall clock, requiring 1.2 V continuously and a very small current every second. The clock was powered by eight wireless batteries connected in series, as shown in Figure 5.17. The use of a 20 dBm transmitter, connected to a 5 dB gain horn antenna, positioned at the focal point of a 51 cm diameter parabolic reflector, enabled us to power this clock at a distance of up to 6 m.
5.6
POWER AND DATA TRANSFER
In the preceding sections, we have discussed the use of an antenna to collect RF power, and converting this RF power into DC power. In doing so, the antenna is no longer available for telecommunication purposes, i.e. the transfer of data. Instead of using two separate systems,
205
POWER AND DATA TRANSFER
Figure 5.17 Wirelessly powered electric wall clock; (a) front view, (b) back view. The additional circuit shown in (b) is a voltage protection circuit.
Figure 5.18
Linear array antenna with parallel (corporate) feeding network.
Figure 5.19
Microstrip Wilkinson power combiner.
each with their own antennas, we would prefer to have a system where the same antennas are used for both power and data transfer. If array antennas with corporate feeds are employed (Figure 5.18), this becomes possible. To this end, we employed Wilkinson power combiners for the Tjunctions in Figure 5.18 [14, 18]. A Wilkinson power combiner is a splitT combiner with a resistive element placed between the two input arms, which has all three ports matched and at the same time provides isolation between the two input ports [19, 20]. A microstrip Wilkinson power combiner is shown in Figure 5.19. Inequalities between the antenna elements in the array, due either to tolerance effects or to phase differences in the received signals, will cause a voltage difference to occur across
206
RECTENNAS: MICROSTRIP PATCH ANTENNAS
the resistor and thus make a current flow through it. Therefore, through dissipation, isolation between the antenna elements is created. If we replace the resistors by rectifying circuits, the inequalities in the received signals may be used for power conversion while maintaining the isolation between the antenna elements. The DC power output will be larger for larger inequalities between the antenna elements. So, the antenna elements may – or, rather, must – be imperfectly matched to the feeding network, but the elements’ imperfect matches may not be identical. Through even–oddmode analysis for the ‘standard’ Wilkinson power combiner [19, 20], the lengths of the two input arms can be determined to be equal to a quarter of a wavelength √ at the center frequency. The characteristic impedances can be determined to be equal to 2Z0 , and the value of the resistor is found to be 2Z0 . In the even–oddmode analysis [21] of the combiner shown in Figure 5.19, we start with the equivalent transmission line circuit (Figure 5.20(a)). Next, this network is drawn in a normalized, symmetric form (Figure 5.20(b)). For this equivalent circuit, two excitation modes are defined: an even or inphase mode, for which Vg2 = Vg3 = 2 V, and an odd or outofphase mode, for which Vg2 = −Vg3 = 2 V. By superposition of the even and odd modes, an effective excitation of Vg2 = 4 V and Vg3 = 0 is created. From this excitation, the S parameters of the network can be derived [21]. If we now replace the resistor by a given parallel circuit of a resistor R and a capacitor C, we may no longer keep the lengths of the arms equal to a quarter of a wavelength at the center frequency. Figure 5.21 shows one half of the bisected circuits of the combiner for even and oddmode excitation. √ For α = 1, R = Z0 , C = 0, Z01 = 2Z0 and l = λ/4, the standard Wilkinson power combiner is obtained. For C = 0, we need a length = λ/4+nλ, n = 0, 1, 2, . . . . Then, from the evenmode analysis, we find that α = 2 and Z01 = 2Z0 . From the oddmode analysis, we find that Z0 = R/4 and 1 1 = arctan , (5.46) β ωRC where β is the wave number of the microstrip transmission line. Quarterlambda impedance transformers are then needed to transform the impedances of all ports to the desired value of Z0 . If the impedance Zrect of the rectifying circuit is given by Zrect = A − jB
(5.47)
(Figures 5.11 and 5.12), the values of R and C are given by 2 B , R =A 1+ A
C=
B/A . ωA[1 + (B/A)2 ]
(5.48)
The scattering matrix of the power combiner may be obtained following the procedure outlined in [20] for the unnormalized scattering matrix, after which the results are transformed
207
POWER AND DATA TRANSFER
Figure 5.20 Even–oddmode analysis of Wilkinson power combiner. (a) Equivalent transmission line circuit. (b) Equivalent transmission line circuit in normalized, symmetric form.
to the normalized scattering matrix using [14, 22]: e−j 2βl , 1 + 1 2 e−j2βl 1 (1 + 1 )(1 + 2 )e−jβl = S13 = S31 = √ , 1 + 1 2 e−j2βl α Z2e 1 = + − 1, Z2e + αZ0 1 + αY02 Z0 Z2e 1 = − , Z2e + αZ0 1 + αY02 Z0
S11 = 1 + (1 − 12 )2
(5.49)
S12 = S21
(5.50)
S22 = S33 S23 = S32
(5.51) (5.52)
where Z01 − 2Z0 , Z01 + 2Z0 αZ0 − Z01 , 2 = αZ0 + Z01 1 =
(5.53) (5.54)
208
RECTENNAS: MICROSTRIP PATCH ANTENNAS
Figure 5.21 Bisections of the combiner circuit. (a) Evenmode excitation. Port 3 is excited by Ve . (b) Oddmode excitation. Port 3 is excited by minus −Vo .
2Z0 + jZ01 tan(βl) , Z01 + j2Z0 tan(βl) 1 1 = + Z jZ 01 tan(βl)
Z2e = Z01
(5.55)
Y02
(5.56)
and Z=
R . 2(1 + jωRC)
(5.57)
β may be calculated using the quasiTEM equations for a microstrip given in [19]. In an initial experiment, we took a ‘standard’ Wilkinson power combiner and replaced Wilkinson the resistor by a rectifying circuit. If we replace the 100 resistor in a Z0 = 50 √ power combiner by a parallel RC circuit, we need to substitute α = 1, Z01 = 2Z0 and = λ/4 at the center frequency in the above equations. For a power combiner matched to the rectifying circuit, we need to substitute α = 2, Z0 = R/4, Z01 = R/2 and the length l given by equation (5.47), calculated for the center frequency. For the purposes of feasibility demonstration and model validation, a modulated transmission system was constructed (Figure 5.22). For this feasibility demonstration, an amplitude modulation (AM) scheme was chosen, where a lowfrequency (1 kHz) block wave was superimposed on a 2.45 GHz carrier signal. The reason for this choice lay in the ease of demodulation, which could be performed with a single diode, not requiring an external power supply.
POWER AND DATA TRANSFER
209
Figure 5.22 Modulated transmission system for demonstrating simultaneous transmission of power and data.
The hardware realization of the power/data receiver is shown in Figure 5.23. For this first demonstrator, a ‘standard’ 50 microstrip Wilkinson power combiner was used, where the 100 resistor was replaced by a voltagedoubling rectifying circuit using Agilent HSMS2852 diodes. The microstrip patch antennas, with dimensions 58.3 mm × 29.2 mm, were dual linearly polarized. The (50 ) feed positions were 20 mm from the corner on the long side. The width of the 50 microstrip transmission lines was 3.1 mm. The patches were separated by a gap of 17 mm. The dielectric was FR4, with a thickness of 1.6 mm, a relative permittivity of 4.28 and a loss tangent of 0.016. To create an inequality between the signals at the combiner input ports, the transmission lines from the patches to the combiner input ports differed in length. The lengths of the transmission lines were, more or less arbitrarily, chosen to be 48.1 mm and 66.4 mm. For the chosen dimensions, the maximum sensitivity would be found at an angle of 26◦ from broadside in the plane of the array. However, the broad element patterns and the limited number of array elements had the effect that the gain variation between broadside and 26◦ off broadside remained within a few decibels, which was acceptable for this demonstration.
210
RECTENNAS: MICROSTRIP PATCH ANTENNAS
Figure 5.23 Hardware realization of AM receiver for demonstrating simultaneous transmission of power and data. The port on the left is for DC power, and the port at the bottom is for the demodulated signal.
Figure 5.24
Calculated and measured unloaded output voltage of the rectifying circuit.
RF ENERGY SCAVENGING
211
To calculate the unloaded output voltage of the rectifying circuit, use was made of the models developed for the power combiner and the voltage doubler circuit. Since, for the particular microstrip patch antennas used here (with a side ratio of approximately 2), the transmission line model fails to calculate accurate input impedance results, fullwave simulation results were used. Then, on the basis of the impedance of the microstrip patch antenna as a function of frequency and the frequencydependent input impedances of the input ports of the power combiner, loaded at the output port with a single Agilent HSMS2852 diode, the power accepted by the rectifying circuit was calculated. From that value, the unloaded output voltage of the rectifying circuit was calculated. Figure 5.24 shows the results, together with measurement results obtained when 3 mW of input power was delivered by the antennas. In the same figure, we also show simulation results based on fullwave simulations of the antennas and fullwave simulations of the power combiner with an opencircuited output port, thus neglecting the power dissipated in the demodulator. In these fullwave simulations, the impedance of the voltage doubler rectifier was linearized over a small frequency range around 2.45 GHz to create an Sparameter list for a black box to be inserted into the combiner configuration. The fullwave simulation results are within 10% of the measured results. The results from the analytical model are within the same order of accuracy, which is adequate for generating preliminary designs. Fullwave analysis of a matched power combiner, using the analytical model for the rectifier circuit, indicated that the unloaded output voltage level increased to a level beyond 1.2 V. At the data output of the receiver, the 1 kHz block wave was retrieved after passing through a series capacitor and a 50 Hz bandstop filter.
5.7
RF ENERGY SCAVENGING
With the ability to analyze and design rectennas, the question arises as to whether it is possible to employ ambient RF energy for powering miniature wireless applications. By ambient RF energy, we mean RF energy not specifically introduced for wirelessly powering an application [7, 8], but RF energy available through public telecommunication services. Our main interest is in telecommunication services operating in the microwave region of the frequency spectrum, especially the Global System for Mobile Communications (GSM) and wireless local area networks (WLANs). For any of these services, printed antennas can be made with dimensions of the order of a few square centimeters, satisfying our constraint for miniature sensors. To assess the feasibility of employing the ambient RF energy supplied by the above systems, we need to assess the power density levels in various environments (e.g. inner city areas, outer city areas, industrial areas, indoors and outdoors for GSM) and various situations (e.g. with respect to traffic, i.e. peak hours and offpeak hours, for GSM and WLANs). 5.7.1 GSM and WLAN Power Density Levels Owing to a growing concern about a potential relation between GSM nonionizing radiation and health risks, a number of national [23] and European initiatives have been taken, dealing
212
RECTENNAS: MICROSTRIP PATCH ANTENNAS
Power density (mW/m^2)
100
IC0 IC1 IC2 OC0 OC1 OC2 OC3 IR0 IR1 IR2 ST2 R0 R1 R2 R3
1
0.01
0.0001
1e06 0
25
50
75
100
125
150
175
200
225
250
Distance (m)
Figure 5.25 Measured GSM900 peak power density levels as a function of distance from the nearest base station. The data is taken from [26]. The code ‘XYa’ indicates the characteristics of the area and measurement site. XY: IC = inner city; OC = outer country, IR = industrial area; ST = small town; R = rural or countryside area. a: 0 = outdoors on the ground; 1 = outdoors on roof, terrace or balcony; 2 = indoors, close to windows, 1.5 m or less; 3 = indoors, not close to windows.
with several aspects of RF exposure caused by the GSM system. The most important of the European initiatives are COST281, ‘Potential Health Implications from Mobile Communication Systems’ [24], and the ‘European Information System on Electromagnetic Fields Exposure and Health Impacts’ [25]. Within these initiatives, ample use has been made of measurement data from several countries, gathered together in COST Action 244bis, ‘Biomedical Effects of Electromagnetic Fields’ [26]. We have used this data to assess the power density levels supplied by GSM base stations. We have made a distinction between the GSM900 (downlink 935 MHz–960 MHz) and GSM1800 (downlink 1805 MHz–1880 MHz) systems. For these communication systems, we have selected published power density measurements and made these visible in graphs showing the power density as a function of distance from the nearest GSM base station. 5.7.1.1 GSM900 In Figure 5.25, the measured peak power densities as a function of the distance from the base station are shown. The data was measured in Austria, Germany and Hungary [26]. Measured data from France and Sweden was not used, since the distance from the nearest base station was not known for this data. All measurements used for this figure were singlefrequency spot measurements in the range 935 MHz–960 MHz. The measurements were taken in either two or three orthogonal directions, employing a bicone or logperiodic receiving antenna. The traffic density and the
213
RF ENERGY SCAVENGING
100 IC1 IC2 OC1 OC2 OC3 IR1 IR2 ST2 R1 R2 R3
Power density (mW/m^2)
10
1
0.1
0.01
0.001 0
25
50
75
100
125
150
175
200
225
250
Distance (m)
Figure 5.26 Measured GSM900 peak power density levels as a function of distance from the nearest base station. The data is taken from [26]. The ﬁgure is the same as Figure 5.25 but with the data taken at ground level removed.
transmission power of the nearest base station were unknown [26]. Since exposure limits differ between different countries, we do not expect the base station transmission powers to be identical in all measurements. Therefore, the power density levels should be regarded as indicative. The codes ‘XYa’ associated with the various measurement sets indicate the characteristics of the area and the measurement site, as explained in the figure caption. We may expect the largest variation in power density at ground level, on passing through the base station beam. Indoors and at an elevated level, these effects should be smaller. Indeed, in Figure 5.25 we observe a large variation in power density in the data sets IC0, OC0, IR0 and R0, all measured outdoors at ground level. If we remove these data sets from the figure, we get the results shown in Figure 5.26. From this figure, we may conclude that between 25 m and 100 m from a GSM900 base station, we may expect – either indoors or outdoors at an elevated level – a power density between 0.1 mW m−2 and 1.0 mW m−2 (10−5 –10−4 mW cm−2 ). For measurements taken over the frequency band as a whole (935 MHz–960 MHz), the summed power density as a function of distance from the nearest base station is shown in Figure 5.27. From this figure, we may conclude that between 25 m and 100 m from a base station, we may expect – either indoors or outdoors at an elevated level – a summed power density between 0.3 mW m−2 and 3.0 mW m−2 (3 × 10−5 − 3 × 10−4 mW cm−2 ). Of course, the measurements over the entire frequency band depend strongly on the traffic density at the moment of measurement. A closer examination of the measurement data supplied by different countries [26] revealed that the power density levels for measurements over the entire frequency band may differ by a factor of one to ten from singlefrequency
214
RECTENNAS: MICROSTRIP PATCH ANTENNAS
100 IC1 IC2 OC1 OC2 OC3 IR1 IR2 ST2 R2 R3
Power density (mW/m^2)
10
1
0.1
0.01
0.001 0
25
50
75
100
125
150
175
200
225
250
Distance (m)
Figure 5.27 Measured GSM900 summed power density levels as a function of distance from the nearest base station, without data taken at ground level. The data is taken from [26].
measurements. Details of GSM traffic at the moments of measurement are not available in [26]. 5.7.1.2 GSM1800 At the time the data shown in Figures 5.25–5.27 was gathered (November 1996 to November 2000 [26]), GSM900 was dominant over GSM1800. Therefore, more measurement data is available for GSM900 exposure than for GSM1800 exposure. Although the frequency of GSM1800 is double that of GSM900 and therefore the freespace loss has quadrupled, the ICNIRP exposure limit has only doubled [27]. Since actual exposure levels for GSM900 are well beyond ICNIRP limits, we may expect power density levels for GSM1800 similar to these for GSM900. This assumption appears to be justified by measurements conducted in the UK at 118 locations in 17 sites for a mix of GSM900 and GSM1800 base stations [28], and by recent measurements in Australia for 60 base stations in five cities [29] (Figure 5.28). The exposure limits in Australia follow those of ICNIRP. The measurements in [28, 29] (see Figure 5.28 again) show that the power density levels received from GSM1800 base stations are at up to 100 m, of the same order of magnitude as those received from GSM900 base stations at a single frequency or summed for lowtrafficdensity situations. From the assessment of the power density levels produced by GSM base stations, we may conclude that it will be very difficult, at the very least, to wirelessly power or charge a small sensor. 5.7.1.3 WLANs A WLAN router (base station) will transmit less power than a GSM base station. But since WLANs are more or less confined to indoor environments and the distance
215
RF ENERGY SCAVENGING
100 GSM900 GSM1800
Power density (mW/m^2)
10
1
0.1
0.01
0.001
0.0001 0
50
100 150 200 250 300 350 400 450 500 550 600 Distance (m)
Figure 5.28 Measured summed power density levels as a function of distance from the nearest GSM900 or GSM1800 base station. The data is taken from [29].
from a router is usually small, reflections and a low path loss could help in establishing higher power density levels. Initial measurements show, however, that the power density levels are at least one order of magnitude lower than those obtained close to a GSM base station. Therefore our assumptions regarding WLAN power densities prove to be wrong. It has to be noted, though, that during these measurements the traffic density was extremely low (on average two connections at a time). However, it seems unlikely that small sensors can be powered for a reasonably long time in a WLAN environment. The idea of getting closer to an RF source, as proposed in the concept of WLAN RF energy scavenging, may be applied, though, to a GSM mobile phone. 5.7.2
GSM Mobile Phone as RF Source
In [30], the use of a GSM mobile phone was proposed as a source for RF energy scavenging. For a specific antenna, a scavenged power of 1.9 mW was predicted at a distance of 1 m from a GSM900 mobile phone, transmitting at a power level of 2 W. This number, however, is rather optimistic and should be regarded as a theoretical maximum, since it assumes a 3 dB gain mobile phone antenna, polarization and gain alignment of the phone antenna and the rectenna, and a rectenna efficiency of 100%. Nevertheless, the concept seemed to be feasible, and therefore a microstrip patch rectenna for GSM1800 was designed following the procedure in [8]. The rectenna was realized on FR4 and connected to an LED for demonstration purposes. A GSM1800 mobile phone (with a maximum allowed transmission
216
RECTENNAS: MICROSTRIP PATCH ANTENNAS
Figure 5.29 Rectenna feeding an LED, wirelessly powered by a GSM1800 mobile phone, transmitting at a peak power level of 1 W.
power of 1 W) was able to make the LED illuminate at up to a distance of 20 cm, proving the concept. The system is shown in Figure 5.29.
5.8
CONCLUSIONS
Analytical models for microstrip patch antennas, diodes and Wilkinson power combiners have been applied and derived to design efficient, compact, lowpower rectenna systems. If a microstrip patch antenna is employed and this antenna is conjugately matched directly to a Schottky diode or a voltagedoubling circuit of Schottky diodes, the need for an impedance and filtering network no longer exists. This leads to an efficiency improvement of more than 10% with respect to a ‘traditional’ lowpower rectenna. Rectennas may be regarded as wireless batteries, and by connecting them in series, lowpower, lowdutycycle devices may be wirelessly powered. Modifying a Wilkinson power combiner by replacing the resistive element with a rectifying circuit allows simultaneous wireless transfer of power and data. Although the feasibility of simultaneous wireless power and data transfer has been demonstrated, there is still room for improvement. The analytical tools developed and discussed here may be of help in this process. For distances ranging from 25 m to 100 m from a GSM base station, power density levels ranging from 0.1 mW m−2 to 1.0 mW m−2 may be expected for single frequencies. For the total GSM downlink frequency bands, these levels may be elevated by a factor of between one and three, depending on the traffic density. Initial measurements in a WLAN environment indicate power density values that are at least one order of magnitude lower. Therefore, neither GSM nor a WLAN is likely to produce enough ambient RF energy for wirelessly powering miniature sensors. However, by employing rectenna arrays [31], this ambient RF energy may be transformed to usable DC energy. A single GSM telephone has proven to deliver enough energy to wirelessly power small applications at moderate distances.
REFERENCES
217
REFERENCES 1. W.C. Brown, ‘The history of power transmission by radio waves’, IEEE Transactions on Microwave Theory and Techniques, Vol. MTT32, No. 9, pp. 1230–1242, September 1984. 2. P. Green, ‘RFID tracking in manufacturing plants, the promise and the reality’, Bellhawk Systems Corporation White Papers. Available at http://www.bellhawk.com/PDF/whitepapers/RFID.pdf. 3. Agilent Technologies, Designing the Virtual Battery, Application Note 1088, Agilent Technologies, Inc., 1999. 4. R.G. Harrison and X. Le Polozec, ‘Nonsquarelaw behavior of diode detectors analyzed by the Ritz–Galerkin method’, IEEE Transactions on Microwave Theory and Techniques, Vol. 42, No. 5, pp. 840–846, May 1994. 5. Y.H. Suh and K. Chang, ‘A highefficiency dualfrequency rectenna for 2.45 and 5.8GHz wireless power transmission’, IEEE Transactions on Microwave Theory and Techniques, Vol. 50, No. 7, pp. 1784–1789, July 2002. 6. J. Heikkinen, P. Salonen and M. Kivikoski, ‘Planar rectennas for 2.45GHz wireless power transfer’, Proceedings of the Radio and Wireless Conference, Denver, CO, pp. 63–66, 2000. 7. J.A.G. Akkermans, M.C. van Beurden, G.J.N. Doodeman and H.J. Visser, ‘Analytical models for lowpower rectenna design’, IEEE Antennas and Wireless Propagation Letters, Vol. 4, pp. 187–190, 2005. 8. J.A.C. Theeuwes, H.J. Visser, M.C. van Beurden and G.J.N. Doodeman, ‘Efficient, compact, wireless battery design’, Proceedings of the European Conference on Wireless Technology, ECWT2007, Munich, Germany, pp. 233–236, October 2007. 9. K.R. Carver and J.W. Mink, ‘Microstrip antenna technology’, IEEE Transactions on Antennas and Propagation, Vol. AP29, No. 1, pp. 2–24, January 1981. 10. W.F. Richards, Y.T. Lo and D.D. Harrison, ‘An improved theory for microstrip antennas and applications’, IEEE Transactions on Antennas and Propagation, Vol. AP29, No. 1, pp. 38–46, January 1981. 11. R.W. Dearnley and A.R.F. Barel, ‘A broadband transmission line model for a rectangular microstrip antenna’, IEEE Transactions on Antennas and Propagation, Vol. 37, No. 1, pp. 6–15, January 1989. 12. H. Pues and A. Van de Capelle, ‘Accurate transmissionline model for the rectangular microstrip antenna’, Proceedings of the IEE, Vol. 131, Part H, No. 6, pp. 334–340, 1984. 13. A. Van de Capelle, ‘Transmissionline model for rectangular microstrip antennas’, in J. James and P.D. Hall (eds.), Handbook of Microstrip Antennas, Peter Peregrinus, pp. 527–578, 1990.
218
RECTENNAS: MICROSTRIP PATCH ANTENNAS
14. H.J. Visser, Array and Phased Array Antenna Basics, John Wiley & Sons, Chichester, 2005. 15. D.A. Fleri and L.D. Cohen, ‘Nonlinear analysis of the Schottkybarrier mixer diode’, IEEE Transactions on Microwave Theory and Techniques, Vol. 21, No. 1, pp. 39–43, January 1973. 16. W.H. Press, B.P. Flannery, S.A. Teukolsky and W.T. Vetterling, Numerical Recipes: The Art of Scientific Computing, Cambridge University Press, Cambridge, 1988. 17. Agilent Technologies, Surface Mount Zero Bias Schottky Detector Diodes, Technical Data, HSMS285x Series, available at http://www.agilent.com/semiconductors. 18. R.C. Hansen, Phased Array Antennas, John Wiley & Sons, New York, 1998. 19. D.M. Pozar, Microwave Engineering, second edition, John Wiley & Sons, New York, 1998. 20. L.I. Parad and R.L. Moynihan, ‘Splittee power divider’, IEEE Transactions on Microwave Theory and Techniques, Vol. 13, No. 1, pp. 91–95, January 1965. 21. J. Reed and G.J. Wheeler, ‘A method of analysis of symmetrical fourport networks’, IRE Transactions on Microwave Theory and Techniques, Vol. 4, No. 4, pp. 246–252, October 1956. 22. C.G. Montgomery (ed.), Technique of Microwave Measurements, Vol. 11 of MIT Radiation Laboratory Series, McGrawHill, New York, 1947. 23. A.P.M. Zwamborn, S.H.J.A. Vossen, B. van Leersum, M.A. Ouwens and W.N. Mäkel, Effects of Global Communication System Radiofrequency Fields on Well Being and Cognitive Functions of Human Subjects with and without Subjective Complaints, TNO Physics and Electronics Laboratory, The Hague, The Netherlands, Report FEL03C148, 2003. 24. COST281, Potential Health Implications from Mobile Communication Systems, available at http://www.cost281.org/ 25. C. del Pozo and D. Papameletiou, European Information System on Electromagnetic Fields Exposure and Health Impacts – Country Reports on EMF and Health: Sources, Regulations, and Risk Communication Approaches, European Commission, Joint Research Centre, Institute for Health and Consumer Protection, Physical and Chemical Exposure Unit, December 2005. 26. U. Bergqvist et al., Mobile Telecommunication Base Stations: Exposure to Electromagnetic Fields, Report of a Short Term Mission within COST244bis, COST244bis Short Term Mission on Base Station Exposure, 2000. 27. ICNIRP, Guidelines for Limiting Exposure to TimeVarying Electric, Magnetic and Electromagnetic Fields (up to 300 GHz), International Commission on NonIonizing Radiation Protection, Oberschleissheim, Germany, 1998.
REFERENCES
219
28. S.M. Mann, T.G. Cooper, S.G. Allen, R.P. Blackwell and A.J. Lowe, Exposure to Radio Waves near Mobile Phone Base Stations, National Radiological Protection Board, Report NRPBR321, June 2000. 29. S.I. Henderson and M.J. Bangay, ‘Survey of RF exposure levels from mobile telephone base stations in Australia’, Bioelectromagnetics, Vol. 27, No. 1, pp. 73–76, January 2006. 30. P. Ancey, ‘Ambient functionality in MIMOSA from technology to services’, Proceedings of the Joint sOcEUSAI Conference, Grenoble, 2005. 31. J.A. Hagerty, F.B. Helmbrecht, W.H. McCalpin, R. Zane and Z.B. Popovic, ‘Recycling ambient microwave energy with broadband rectenna arrays’, IEEE Transactions on Microwave Theory and Techniques, Vol. 52, No. 3, pp. 1014–1024, March 2004.
6 Large Array Antennas: OpenEnded RectangularWaveguide Radiators The accurate analysis of large planar phased array antennas is a very timeconsuming task owing to the incorporation of effects of the mutual coupling that exists between the radiating elements. For very large phased array antennas, i.e. for arrays consisting of roughly more than a few tens of radiating elements, the analysis time may be sped up considerably by approximating the phased array antenna by an infinite, uniformly excited phased array antenna. This is permissible because, in a very large phased array antenna, the majority of elements experience an environment identical to that of an element in a periodic phased array antenna. The analysis may now be restricted to a single unit cell only. The infinitearray concept will be applied to phased array antennas consisting of openended waveguide radiators, employing a modematching technique. To allow for impedancematching structures inside the waveguide radiators, a modematching technique will be developed for waveguidetounitcell junctions and waveguidetowaveguide junctions. After a thorough validation of the results for cascaded waveguidetowaveguide junctions, the results for the radiation into a unit cell will be validated by comparison with published analysis results and measurements on a large array antenna. For the evaluation of large array antennas analysis of the reflection coefficient of a unit cell suffices, since this reflection coefficient is directly related to the scan element pattern, which exhibits all features of a scanned phased array antenna. The material presented in this chapter is not state of the art, and dates back to the mid 1990s. Since then, a lot of progress in analyzing these kinds of antennas has been made. Nevertheless, we find it appropriate to present a ‘classic’ modematching approach in detail. This material may aid in understanding new developments and may be implemented relatively easily in software for analyzing rectangularwaveguide structures and infinite arrays of openended waveguides. Approximate Antenna Analysis for CAD
Hubregt J. Visser
© 2009 John Wiley & Sons, Ltd. ISBN: 9780470512937
222 6.1
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
INTRODUCTION
Array and phased array antennas are gaining in popularity. They seem no longer to be of interest only for military systems (radar), but are encountered today in many civilian systems, for example in mobilecommunication base stations. Advances in electronics have made fastswitching phase shifters feasible, and increased computer power has made realtime signal processing feasible. These two developments have helped in the realization of large phased array antenna systems. In order to design (large) phased array antennas consisting of regularly positioned, identical radiating elements, an analysis technique is required that takes the mutual coupling between the radiating elements into account in addition to the characteristics of the radiating elements. The mutual coupling is largely responsible for the unique characteristics of phased arrays [1]. Only for relatively small array antennas, having (far) fewer than one hundred elements, is it possible to calculate coupling characteristics within a reasonable amount of time using a ‘brute force’ technique. For larger array antennas, a different approach must be taken. This chapter deals with the application of a ‘brute force’ technique to large array antennas. The development of the model dates back to the mid 1990s. Since then, a lot of progress in analyzing large array antennas has been made [2–7]. Notwithstanding these developments, a presentation in detail of this ‘brute force’ technique (a ‘classic’ modematching technique) is still appropriate, since it is easy to understand, simple to implement in software and valuable for analyzing waveguide structures and infinite arrays of openended waveguide radiators. To facilitate the analysis of large or periodic planar array antennas, we assume the array antenna to be infinitely large in the two planar directions and to be uniformly illuminated. Thus, edge effects due to the boundedness of the large but finite structure are neglected. Owing to the periodicity of the configuration, the effects of coupling between the radiators are essentially the same for each of the radiators or elements, except for phase differences. Therefore, it is permissible to place fictitious walls, perpendicular to the face of the array, between neighboring elements, thereby setting up unit cells into which each element individually radiates. A unit cell may be regarded as a rectangular waveguide with phaseshifting walls. Following this approach allows us to solve the infinitearray problem for a single unit cell only. Of course, in the case of a finite array antenna, the effects of coupling are not the same for each element. However, for arrays consisting of more than a few hundred elements, the majority of the (inner) elements behave locally as if embedded in an infinite array [8]. Openended waveguide array antennas are attractive owing to the inherent characteristics of openended waveguide radiators such as wide frequency bandwidth, good crosspolarization behavior and highpass filtering characteristics due to the cutoff frequency of the waveguide modes [9, 10]. Therefore, large phased array antennas consisting of openended waveguide radiators are still being considered for new developments [4, 11, 13]. 6.1.1
Mode Matching and Generalized Scattering Matrices
Pioneering work in the field of analyzing openended waveguide (phased) array antennas was performed by a number of researchers in the 1960s and 1970s [14–20]. As a result, mode matching is by now generally accepted as an analysis method that is very suitable
INTRODUCTION
223
Figure 6.1 Waveguide structure consisting of cascaded rectangularwaveguide discontinuities and transmission lines.
for rectangularwaveguide structures. From the 1980s onwards, mode matching, applied for constructing generalized scattering matrices (GSMs)1 to describe waveguide junctions, has been used with success for the characterization of filterlike structures [21–28]. Because of this success of this method with filterlike structures, we have adopted the method for the analysis of planar, infinite, openended waveguide (phased) array antennas. Mode matching is employed to derive GSMs for waveguide discontinuities, including the transition from a rectangular waveguide to a rectangular unit cell, i.e. the transition from a waveguide to free space in a periodic array of waveguides.2 A waveguide structure, as shown in Figure 6.1, is analyzed by cascading the GSMs of the discontinuities and rectangularwaveguide transmission lines, which results in the overall GSM of the structure. From this overall GSM, relevant parameters may be extracted. The dimensions of the GSM are infinite in principle, but by application of a justified truncation, desired parameters having an acceptable accuracy may be extracted. The example shown in Figure 6.1 shows an element within a unit cell of a planar, openended waveguide array antenna. Going from the left to the right in this figure, we encounter a rectangular waveguide, an asymmetric twodimensional crosssectional reduction, a second rectangular waveguide, a symmetric twodimensional crosssectional enlargement, a third rectangular waveguide, a symmetric onedimensional crosssectional reduction, a fourth waveguide, a junction between a waveguide and a freespace unit cell, a dielectric sheet,
1 In this chapter, a GSM is a scattering matrix, containing, besides the fundamentalmode scattering coefficients,
higherordermode scattering coefficients. 2 To calculate the GSM of a transition from an isolated rectangular waveguide to free space, the modematching
technique presented in [29] may be applied.
224
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
a dielectric step, a second dielectric sheet and, finally, a dielectric step. So, for the structure shown in Figure 6.1, we need to cascade 11 GSMs to obtain the overall GSM. The analysis method based upon cascading GSM’s is very versatile. It is, for instance, capable of analyzing infinite planar arrays consisting of openended waveguide radiators that employ finitethickness irises in the waveguide apertures (Figure 6.1). In order to obtain a wideangle impedance match (WAIM), aperture irises and a dielectric sheet in front of the apertures may be employed [18]. Previously developed analysis methods could deal only with infinitely thin aperture irises [8,30]. More recent developments, however, have indicated that this WAIM may be improved by using one or more irises with a finite thickness [31, 32]. Since our main concern is the analysis and design of planar, openended, rectangularwaveguide (phased) array antennas, in this chapter we shall derive only the GSM’s for the building blocks necessary for analyzing these antennas. These building blocks are: • crosssectional reductions and enlargements of rectangular waveguides; • junctions between a rectangular waveguide and a freespace unit cell; • finitelength waveguides; • finitethickness dielectric sheets in a unit cell; • dielectric steps in a unit cell. Since we are dealing only with electromagnetic fields inside rectangular waveguides and inside rectangular phaseshiftwall waveguides (unit cells), we start with the derivation of field quantities within these domains. Next, we derive the GSM’s of a crosssectional reduction or enlargement in a rectangular waveguide, a junction of a rectangular waveguide to a rectangular phaseshiftwall waveguide, a dielectric step in a unit cell, and a transmission line (a waveguide with an electrically conducting wall or phase shift wall) of finite length. Every type of discontinuity will be the subject of a separate section. After the cascading of these GSM’s has been discussed, a software implementation is thoroughly validated by comparing simulation results for filterlike structures with results obtained from the open literature. Simulation results for planar, openended waveguide array antennas are compared with results obtained from an independently developed and validated computer code [31].
6.2
WAVEGUIDE FIELDS
A semiinfinite rectangular waveguide with crosssectional dimensions a and b is shown in Figure 6.2. A rectangular coordinate system (ˆix , ˆiy , ˆiz ), as indicated in the figure, is assumed. The origin of this coordinate system is positioned at the center of the waveguide cross section, following [8, 33]. The waveguide walls are assumed to be perfectly electrically conducting.
225
WAVEGUIDE FIELDS
iˆy
b/2
b ˆ a/2 ix
iˆz a/2
b/2 a
Figure 6.2
Semiinﬁnite waveguide of rectangular cross section, and coordinate system.
The electric field E = E(r) and magnetic field H = H(r) satisfy the sourcefree Maxwell equations ∇ × H − jωεr ε0 E = 0, ∇ × E + jωµr µ0 H = 0,
(6.1) (6.2)
∇ · E = 0, ∇ · H = 0,
(6.3) (6.4)
where ε0 is the permittivity of free space and µ0 is the permeability of free space.3 The position vector r is given by (6.5) r = x ˆix + y ˆiy + zˆiz , and a harmonic time factor ejωt for the fields is assumed and suppressed. The z dependence of every wave or mode is of the form e−jγ z for waves propagating in the direction of increasing z and of the form ejγ z for waves propagating in the direction of decreasing z. The wave number γ here is positive real for propagating waves and negative imaginary for evanescent waves. The evanescent waves therefore decay exponentially. To simplify the model, we restrict ourselves to lossfree media. For airfilled waveguides and lossfree or thin WAIM sheets, this is a valid restriction. Next, we shall discuss the decomposition of the electric and magnetic fields into modes. We shall consider the axial field components ez± and h± z as fundamental unknowns.
3 Strictly speaking, the last two equations are not part of the Maxwell equations for the sourcefree situation. They follow from the first two equations.
226
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
We start by decomposing the fields into transverse (with respect to z) and axial components E(r) = e± (rT )e∓jγ z ,
(6.6)
±
(6.7)
H(r) = h (rT )e
∓jγ z
,
where e+ (rT ) and h+ (rT ) are the electric field vector and magnetic field vector, respectively, for a wave traveling in the positive z direction; e− (rT ) and h− (rT ) are the electric field vector and magnetic field vector, respectively, for a wave traveling in the negative z direction. The transverse position vector rT is given by rT = x ˆix + y ˆiy .
(6.8)
Substitution of equations (6.6) and (6.7) into the sourcefree Maxwell equations (6.1) and (6.2) results in ± ± ˆ kT2 e± T = jωµr µ0 (iz × ∇T hz ) ∓ jγ ∇T ez ,
kT2 h± T
(6.9)
= −jωεr ε0 (ˆiz × ∇T ez± ) ∓ jγ ∇T h± z ,
(6.10)
± where e± T and hT are the transverse electric and magnetic field vectors;
∇T ≡
∂ ˆ ∂ ix + ˆiy ∂x ∂y
(6.11)
and where
kT2 = kT · kT = ω2 ε0 εr µ0 µr − γ 2
(6.12)
kT = kx iˆx + ky iˆy
(6.13)
is the transverse wave vector. The axial wave number γ is related to the square of the transverse wave number, kT2 , through 2 ω2 ε0 εr µ0 µr ≥ kT2 ω 2 ε 0 ε r µ0 µr − k T , γ = (6.14) −j −ω2 ε0 εr µ0 µr + k 2 , ω2 ε0 εr µ0 µr < k 2 , T
T
where the square root is nonnegative. Equations (6.9) and (6.10) indicate that a knowledge of the axial field components of a mode suffices to determine the transverse field components of that mode. Furthermore, the problem of solving equations (6.9) and (6.10) can be separated into solving two scalar problems in the transverse plane. For transverse electric (TE) modes, we assume that ez± = 0 ± ± and h± z = 0. For transverse magnetic (TM) modes, we assume that hz = 0 and ez = 0. For obvious reasons, TE modes are also known as H modes and TM modes are also known as ± E modes. Transverse electromagnetic (TEM) waves, where h± z = 0 and ez = 0, cannot exist in the waveguide.
227
WAVEGUIDE FIELDS
6.2.1
TE Modes
For TE modes, ez± = 0 and h± z = 0. All field components are now determined from the scalar quantity h± . z Taking the curl of equation (6.2), using the vector identity ∇×∇×A = −∇·∇A+∇(∇·A), results in (6.15) ∇ × ∇ × H = −∇ · ∇H + ∇(∇ · H) = jωε0 εr ∇ × E. Next, substituting equations (6.2) and (6.4) into equation (6.15), using ω2 ε0 εr µ0 µr = k 2 , and rearranging terms gives (∇ · ∇ + k 2 )H = 0. (6.16) This vectorial equation may be separated into its scalar components. So, after substitution of equation (6.7) into equation (6.16), we find for the z component of the magnetic field ∂ 2 ± ∓jγ z ∂2 ∓jγ z ∓jγ z hz e + 2 h± e∓jγ z − γ 2 h± + k 2 h± = 0. ze ze 2 ∂x ∂y z
(6.17)
Using equation (6.12) and realizing that equation (6.17) should apply for all values of z leads finally to the twodimensional Helmholtz equation, 2 2 ± (∇T · ∇T + kT2 )h± z = (∇T + kT )hz = 0.
(6.18)
As will be shown, nontrivial solutions to equation (6.18) exist only for certain values of kT2 . These values depend on the transverse geometry of the waveguide. The solutions are called modes. Applying the method of separation of variables [34] leads to the following solution for equation (6.18): a b ± = A cos k cos k , (6.19) x + y + h± x y z 2 2 where A± is the mode amplitude coefficient (as yet unknown) and the separation constants kx and ky are related to the transverse and axial wave numbers through kT2 = kx2 + ky2 = ω2 ε0 εr µ0 µr − γ 2 ,
(6.20)
as can be found by substitution of equation (6.19) into equation (6.18). Next, we use the boundary conditions that the modulus of h± z reaches its maximum at x = ±a/2 and y = ±b/2. These boundary conditions are the results of our earlier assumption of perfectly electrically conducting waveguide walls, which forces ez± to be zero at the walls. Applying these boundary conditions leads to4 kx =
mπ , a
m ∈ N,
(6.21)
4 The natural numbers N are the set {1, 2, 3, . . .} or {0, 1, 2, 3, . . .}. The inclusion of zero is a matter of definition [35].
Here we define N to include zero.
228
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
and
nπ , n ∈ N. (6.22) b Since negative values of kx and ky will result in the same solutions as for positive values of kx and ky , we shall restrict ourselves from now on to nonnegative values of m and n. The indices m and n are known as the mode indices. So, for the TEmn mode, we have, using equation (6.20), 2 mπ 2 nπ 2 + , m, n ∈ N ∧ (m, n) = (0, 0). (6.23) kTEmn = a b ky =
Substitution of equations (6.21) and (6.22) into equation (6.19) gives a nπ b mπ ± h± x + cos y + , = A cos zmn mn a 2 b 2
(6.24)
and upon substitution of equation (6.23) into equation (6.14), the wave number of the axial mode is given by 2 2 mπ 2 nπ mπ 2 nπ 2 k02 εr µr − − , k ε µ ≥ + r r 0 a b a b γmn = (6.25) 2 2 2 2 mπ nπ mπ nπ + , k02 εr µr < + , −j −k02 εr µr + a b a b √ where the square root is greater than or equal to zero and k0 = ω ε0 µ0 is the freespace wave number. The modal wave admittance YTEmn for the TEmn mode follows from substitution of equation (6.26) [36], ± ˆ (6.26) h± Tmn = ±YTEmn (iz × eTmn ), into equations (6.9) and (6.10). This results in YTEmn =
γmn . ωµ0 µr
(6.27)
The wave impedance for the TEmn mode is then, by default, ZTEmn = 1/YTEmn . 6.2.2
TM Modes
± For TM modes, h± z = 0 and ez = 0. All field components are now determined from a scalar quantity ez± that satisfies the twodimensional Helmholtz equation
(∇T · ∇T + kT2 )ez± = 0.
(6.28)
As in the TEmode situation, nontrivial solutions to equation (6.28) exist only for certain values – different from those for TE modes – of kT2 . Applying the method of separation of variables leads to a b sin ky y + , (6.29) ez± = A± sin kx x + 2 2
229
WAVEGUIDE FIELDS
where A± is the mode amplitude coefficient (as yet unknown).5 Using the boundary conditions that ez± is zero at the waveguide boundaries x = ±a/2 and y = ±b/2 (we assume perfectly electrically conducting waveguide walls) leads to the following expressions for the separation constants: mπ , m ∈ N ∧ m = 0, a nπ ky = , n ∈ N ∧ n = 0. b kx =
(6.30) (6.31)
Again, taking only positive values for kx and ky into account, the following relation applies for the transverse wave number of the TMmn mode: 2 mπ 2 nπ 2 kTMmn = + , m, n ∈ N, m = 0; n = 0. (6.32) a b Substitution of equation (6.32) into equation (6.29) yields a nπ b mπ ± ± x+ sin y+ , ezmn = Amn sin a 2 b 2
(6.33)
and the axial wave number is given by equation (6.25). The wave impedance ZTMmn for the TMmn mode follows from substitution of equation (6.34) [36], ± ˆ (6.34) e± Tmn = ±ZTMmn (hTmn × iz ), into equations (6.9) and (6.10), which results in ZTMmn =
γmn . ωε0 εr
(6.35)
The wave admittance for the TMmn mode is, by definition, YTMmn = 1/ZTMmn . 6.2.3
Transverse Field Components
The complete transverse fields ET and HT in the waveguide, at a crosssectional position z, consist of a linear superposition of modal field distributions ET (r) = HT (r) =
∞ ∞ m=0 n=0 ∞ ∞
eT±mn (rT )e∓jγmn z ,
(6.36)
∓jγmn z h± , Tmn (rT )e
(6.37)
m=0 n=0
where it is implicitly understood that (m, n) = (0, 0) for both TE and TM modes, and m = 0 and n = 0 for TMmode contributions. 5 To distinguish between TE and TMmode solutions in the remainder of this chapter, TEmode amplitude coefficients will be unprimed and TMmode amplitude coefficients will be primed.
230
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
Equations (6.36) and (6.37) can be separated into four scalar equations for the transverse field components: ∞ ∞
Ex (r) =
m=0 n=0 ∞ ∞
Ey (r) =
m=0 n=0 ∞ ∞
Hx (r) =
m=0 n=0 ∞ ∞
Hy (r) =
ex±mn (x, y)e∓jγmnz ,
(6.38)
ey±mn (x, y)e∓jγmnz ,
(6.39)
∓j γmn z h± , xmn (x, y)e
(6.40)
∓jγmn z h± , ymn (x, y)e
(6.41)
m=0 n=0
where ± ± ± ex±mn = A± mn exTEmn + Amn exTMmn ,
ey±mn h± xmn h± ymn
= = =
± A± mn eyTEmn ± A± mn hxTEmn ± A± mn hyTEmn
(6.42)
± + A± mn eyTMmn , ± + A± mn hxTMmn , ± + A± mn hyTMmn .
(6.43) (6.44) (6.45)
With the aid of equations (6.9), (6.10), (6.24) and (6.33), the mode functions can be found as j nπ x y ex±TEmn = 2 ωµ0 µr Cm Sn , (6.46) b a b kTmn mπ x y j ± Sm Cn , (6.47) eyTEmn = − 2 ωµ0 µr a a b kTmn j mπ x y ± exTMmn = ∓ 2 γmn Cm Sn , (6.48) a a b kTmn nπ x y j Sm Cn , (6.49) ey±TMmn = ∓ 2 γmn b a b kT mn
and, using equations (6.27) and (6.35), ± h± xTEmn = ∓YTEmn eyTEmn ,
h± yTEmn ± hxTMmn h± yTMmn where
= = =
1 , C (ξ ) = cos π ξ + 2
(6.50)
±YTEmn ex±TEmn , ∓YTMmn ey±TMmn , ±YTMmn ex±TMmn , = 0, 1, 2, . . . ,
(6.51) (6.52) (6.53)
−
1 1 ≤ξ ≤ , 2 2
(6.54)
231
UNIT CELL FIELDS
Figure 6.3 Planar, inﬁnite openended waveguide array antenna with the radiators arranged in a triangular grating, plus an indication of a single unit cell.
and
6.3
1 S (ξ ) = sin π ξ + , 2
= 1, 2, 3, . . . ,
−
1 1 ≤ξ ≤ . 2 2
(6.55)
UNIT CELL FIELDS
By virtue of the infinite planar extent and periodicity of an openended waveguide array, the resulting twodimensional grating may be decomposed into an infinite sequence of unit cells (Figure 6.3). In this figure, the radiators are arranged in a general triangular lattice that is defined by the distances s and t between elements and the angle between the s and t directions.6 One unit cell is indicated in a lighter shade. The analysis of the fields is similar to that for a rectangular waveguide with electrically conducting side walls. So equations (6.1)–(6.14) apply equally well to the unit cell. However, in analysing the TE and TMmode fields, we have to take the transverse unboundedness and the periodicity of the structure into consideration.
6 For = π/2, we have a rectangular lattice.
232 6.3.1
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
TE Modes
For TE modes, ez± = 0 and h± z = 0. All field components are determined from the scalar ± satisfies the twodimensional Helmholtz equation quantity h± , where h z z (∇T · ∇T + kT2 )h± z = 0.
(6.56)
Nontrivial solutions to this equation exist only for certain values of kT2 . In the situation where we were dealing with a rectangular waveguide with perfectly electrically conducting side walls, we chose, in the method of separation of variables, a sample solution in the form of a product of cosines. Now, owing to the transverse unboundedness of the structure, we choose a sample solution that has an exponential form, ± −jkT0 ·rT = B ± f (x, y)e−j(kx0 x+ky0 y) , h± z = B f (x, y)e
(6.57)
where B ± is the mode amplitude coefficient (as yet unknown), f (x, y) is a periodic function and the separation constants are related to the sphericalcoordinate angles ϑ and ϕ through kx0 = k sin(ϑ) cos(ϕ),
(6.58)
ky0 = k sin(ϑ) sin(ϕ),
(6.59)
√ with k = k0 εr µr being the wave number in the medium immediately outside the radiating apertures. The periodicity of the grating is accounted for in the function f (x, y). Since all unit cells are identical, we see from inspection of Figure 6.3 that f (x, y) must satisfy [37] f (x + s, y) = f (x, y)
(6.60)
f (x + t cos(), y + t sin()) = f (x, y).
(6.61)
and The function that satisfies these conditions is given by f (x, y) = e−jp2πx/s ejp2πy/[s tan()] e−jq2πy/[t sin()] ,
p, q ∈ Z,
(6.62)
which results in, for the TEpq mode, ± −j(up0 x+vpq y) , h± pq = Bpq e
where up0 = k sin(ϑ) cos(ϕ) + and vpq = k sin(ϑ) sin(ϕ) −
2pπ , s
(6.63) p ∈ Z,
(6.64)
2qπ 2pπ + , p, q ∈ Z. (6.65) s tan() t sin() mode, kTpq , which now accounts for the periodicity
The transverse wave number of the TEpq of the structure, is given by 2 , kTpq = u2p0 + vpq
p, q ∈ Z.
(6.66)
233
UNIT CELL FIELDS
y x
z
s
tsin(Ω)
Figure 6.4
Rectangular unit cell, centered above openended waveguide radiator.
Each set of values (p, q) specifies a plane wave,7 radiated into the direction (ϑ, ϕ), called a space harmonic, grating mode or Floquet mode. As long as we do not modify the periodicity, we may deform and shift the position of the unit cell shown in Figure 6.3. For convenience, we therefore assume rectangular unit cells, centered above the openended waveguide apertures, as shown in Figure 6.4. The dimensions of the rectangular unit cell are s and t sin() in the x and y coordinates, respectively. It is now apparent that we may treat the unit cell as a rectangular waveguide where the walls of this waveguide do not have reflective properties, but instead phaseshifting properties, as shown by equations (6.64)–(6.66). The axial wave number follows from equation (6.14) as
γpq
2 2 2 2 k02 εr µr ≥ u2p0 + vpq k0 εr µr − up0 − vpq , = −j −k 2 ε µ + u2 + v 2 , k 2 ε µ < u2 + v 2 , pq pq 0 r r p0 0 r r p0
(6.67)
where the square root is greater than or equal to zero.
7 Contrary to the situation for a rectangular waveguide with a metallic wall, negative values need to be taken into account here, since they represent solutions that differ from their positive counterparts.
234
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
The modal wave admittance YTEpq for the TEpq mode follows from the substitution of equation (6.68) [36], ± ˆ h± (6.68) Tpq = ±YTEpq (iz × eTpq ), into equations (6.9) and (6.10). This results in YTEpq =
γpq . ωµ0 µr
(6.69)
The wave impedance for the TEpq mode is ZTEpq = 1/YTEpq . 6.3.2
TM Modes
± ± For TM modes, h± z = 0 and ez = 0. All field components are determined by ez , which satisfies the twodimensional wave equation
(∇T · ∇T + kT2 )ez± = 0.
(6.70)
Solutions to equation (6.70) exist only for certain values of kT2 . In applying the method of separation of variables, we take as a sample solution to equation (6.70) ± ez±pq = Bpq f (x, y)e−jkT0 ·rT ,
(6.71)
where B ± is the mode amplitude coefficient (as yet unknown) and f (x, y) is a function related to the periodicity of the structure. Following the same procedure as outlined in the section 6.3.1 for TE modes, we finally find for the TMpq mode ± −j(up0 x+vpq y) ez±pq = Bpq e ,
(6.72)
where up0 and vpq are defined by equations (6.64) and (6.65), respectively. The transverse wave number is defined by equation (6.76) and the axial wave number is defined by equation (6.67). The wave impedance ZTMpq for the TMpq mode follows from substitution of equation (6.73) [36], ± ˆ (6.73) e± Tpq = ±ZTMpq (hTpq × iz ), into equations (6.9) and (6.10), which results in γpq . ωε0 εr mode is YTMpq = 1/ZTMpq . ZTMpq =
The wave admittance for the TMpq 6.3.3
(6.74)
Transverse Field Components
The complete transverse fields in the unit cell consist of a superposition of mode contributions ET (r) = HT (r) =
∞
∞
p=−∞ q=−∞ ∞ ∞ p=−∞ q=−∞
∓γpq z e± , Tpq (rT )e
(6.75)
∓γpq z h± . Tpq (rT )e
(6.76)
235
UNIT CELL FIELDS
Equations (6.75) and (6.76) can be separated into four scalar equations for the transverse field components: ∞ ∞ Ex (r) = ex±pq (x, y)e∓γpq z , (6.77) Ey (r) = Hx (r) = Hy (r) =
p=−∞ q=−∞ ∞ ∞ p=−∞ q=−∞ ∞ ∞ p=−∞ q=−∞ ∞ ∞
ey±pq (x, y)e∓γpq z ,
(6.78)
∓γpq z h± , xpq (x, y)e
(6.79)
∓γpq z h± , ypq (x, y)e
(6.80)
p=−∞ q=−∞
where ± ± ± ± exTEpq + Bpq exTMpq , ex±pq = Bpq
(6.81)
± ± ± ± eyTEpq + Bpq eyTMpq , ey±pq = Bpq
(6.82)
h± xpq
± ± + Bpq hxTMpq ,
(6.83)
± ± ± ± h± ypq = Bpq hyTEpq + Bpq hyTMpq .
(6.84)
=
± ± Bpq hxTEpq
With the aid of equations (6.9), (6.10), (6.63) and (6.72), the mode functions can be found as 1 ex±TEpq = − 2 ωµ0 µr vpq Xpq , (6.85) kTpq ey±TEpq =
1 kT2 pq
ex±TMpq = ∓ ey±TMpq = ∓
ωµ0 µr up0 Xpq ,
1
γpq up0 Xpq , kT2 pq 1 KT2pq
γpq vpq Xpq ,
(6.86) (6.87) (6.88)
and, using equations (6.69) and (6.74), ± h± xTEpq = ∓YTEpq eyTEpq ,
(6.89)
± h± yTEpq = ±YTEpq exTEpq ,
(6.90)
h± xTMpq
=
∓YTMpq ey±TMpq ,
(6.91)
=
±YTMpq ex±TMpq ,
(6.92)
h± yTMpq where
Xpq = e−j(up0 x+vpq y)
(6.93)
2 . kT2 pq = kTpq · kTpq = u2p0 + vpq
(6.94)
and
236
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
y
(xc,yc,0)
z II
z>0 b
b⬘
x
z=0
(0,0,0)
I
z<0 a⬘ a Figure 6.5 Crosssectional step (reduction) in a rectangular waveguide, in the plane z = 0.
6.4
CROSSSECTIONAL STEP IN A RECTANGULAR WAVEGUIDE
In this section, we shall derive the GSM for a one or twodimensional crosssectional step in a rectangular waveguide. Seen from the input waveguide, we may regard this step as a waveguide reduction or a waveguide enlargement. When at least one of the two transverse dimensions of the output waveguide is larger than the corresponding dimension of the input waveguide, we define the step as a waveguide enlargement. Otherwise, we define the step as a waveguide reduction. Figure 6.5 shows a junction of two rectangular waveguides. The input waveguide (z < 0) is designated as waveguide I, and the output waveguide (z > 0) is designated as waveguide II. According to the definition given above, the configuration shown in the figure is a waveguide reduction.8 The input waveguide has crosssectional dimensions a and b. The output waveguide has crosssectional dimensions a and b , where a ≤ a and b ≤ b. The step is positioned in the plane z = 0. A rectangular coordinate system is centered with respect to the input waveguide. The output waveguide has the center of its cross section at (x, y, z) = (xc , yc , 0), where xc  +
a a ≤ 2 2
(6.95)
and
b b ≤ , (6.96) 2 2 such that the cross section of the output waveguide falls entirely within that of the input waveguide. yc  +
8 A change in the medium filling a rectangular waveguide with a uniform cross section will also be considered a crosssectional step.
CROSSSECTIONAL STEP IN A RECTANGULAR WAVEGUIDE
237
Also, S I is the area of the cross section of waveguide I (−a/2 ≤ x ≤ a/2, −b/2 ≤ y ≤ b/2), S II is the area of the crosssection of waveguide II (−a /2 ≤ x − xc ≤ a /2, −b /2 ≤ y − yc ≤ b /2) and S is the area of the cross section of waveguide I , excluding the cross section of waveguide II. S is the complement of the intersection of S I and S II . S is assumed to be perfectly electrically conducting. In accordance with the analysis in terms of modes presented in previous sections, the fields satisfy the boundary conditions at the transverse boundaries of the waveguides. By enforcing the boundary conditions across the interface at z = 0 – the tangential components of the electric and magnetic fields must be continuous across the interface – a matrix formulation for this step discontinuity may be derived, as shown in the remainder of this section. 6.4.1
Boundary Conditions Across the Interface
The continuity of the tangential fields across the interface at z = 0, written for the transverse Cartesian components of the electric and magnetic fields, is specified by lim ExI = lim ExII ,
(6.97)
lim EyI = lim EyII ,
(6.98)
lim HxI = lim HxII ,
(6.99)
lim HyI = lim HyII ,
(6.100)
z↑0
z↓0
z↑0
z↓0
z↑0
z↓0
z↑0
z↓0
where the superscripts I and II relate to the field components in waveguides I and II, respectively. Using equations (6.38)–(6.41), equations (6.42)–(6.45), equations (6.50)–(6.53) and ex−TEmn = ex+TEmn ,
(6.101)
ey−TEmn ex−TMmn ey−TMmn
(6.102)
= = =
ey+TEmn , −ex+TMmn , −ey+TMmn ,
(6.103) (6.104)
within the boundaries −a/2 ≤ x ≤ a/2, −b/2 ≤ y ≤ b/2, results in M N %&
& + ' + ' ( + − I − I + − A AImn + AImn exI TEmn + AI e mn mn xTMmn
m=0 n=0
L W %& ' + ' + ( & + + − − AII exIITE + AII exIITM , + AII − AII lw lw lw lw lw lw l=0 w=0 = a a b b − + xc ≤ x ≤ + xc ∧ − + yc ≤ y ≤ + yc , 2 2 2 2 0, otherwise,
(6.105)
238
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS M N %&
& + ' + ' ( + − I − I + AImn + AImn eyI TEmn + AI mn − Amn eyTMmn
m=0 n=0
=
M
L W %& & + ' ' ( + II − II + II II − II + + A + A − A AII e e yTElw yTMlw , lw lw lw lw l=0 w=0
− 0,
N %&
(6.106)
a a b b + xc ≤ x ≤ + xc ∧ − + yc ≤ y ≤ + yc , 2 2 2 2 otherwise,
' ' ( & + − + + I I − I I+ −AImn + AImn YTE eI + −AI mn − Amn YTMmn eyTMmn mn yTEmn
m=0 n=0
=
L W %&
' + I − II II + −AI + A lw lw YTElw eyTE
lw
l=0 w=0 a
' ( & + II − II II + Y , + −AII − A e TMlw yTM lw lw lw
a b b + xc ∧ − + yc ≤ y ≤ + yc , 2 2 2 2 M N %& ' ' ( & + + − I I+ I I − I I+ AImn − AImn YTE Y e + A + A e TMmn xTMmn mn mn mn xTEmn −
+ xc ≤ x ≤
(6.107)
m=0 n=0
=
L W %&
' + I − II II + AI lw − Alw YTElw exTE
l=0 w=0 a
−
2
+ xc ≤ x ≤ +
lw
& + ' ( II − II II + + AII , lw + Alw YTMlw exTM
a b b + xc ∧ − + yc ≤ y ≤ + yc . 2 2 2 −
lw
(6.108)
In the above equations, AImn and AImn are the mode amplitude coefficients in waveguide I for the TEmn mode traveling in the increasing and the decreasing z direction, respectively. + I − AI mn and Amn are the mode amplitude coefficients in waveguide I for the TMmn mode + II − traveling in the increasing and the decreasing z direction, respectively. AII lw and Alw are the mode amplitude coefficients in waveguide II for the TElw mode traveling in the + II − increasing and the decreasing z direction, respectively, and AII lw and Alw are the mode amplitude coefficients in this waveguide for the TMlw mode traveling in the increasing and the decreasing z direction, respectively. The mode functions (solutions of the twodimensional Helmholtz equation) and mode admittances are superscripted with a ‘I ’ or a ‘II’, depending on which waveguide is being considered. For waveguide I , the mode functions are given by equations (6.46)–(6.49); for waveguide II, the same equations are used, but a is substituted for a, b is substituted for b, l is substituted for m and w is substituted for n. The TE and TM admittance functions for waveguide I are given by equations (6.27) and (6.35). For waveguide II, these functions are obtained by the same equations upon substitution of l for m and w for n. Equations (6.105)–(6.108) are exact for M = ∞, N = ∞ and L = ∞, W = ∞. For the exact equations, the matching procedure leads to an infinite system of linear equations ± ± II ± II ± with AImn , AI mn , Amn and Amn as unknowns. Fortunately, it turns out that not all modes contribute equally strongly to the GSM characterization of a waveguide junction. In fact,
CROSSSECTIONAL STEP IN A RECTANGULAR WAVEGUIDE
239
only the fundamental waveguide modes and a limited set of higherorder modes suffice for an accurate characterization. In the next subsection, the creation of a finite system of linear equations will be discussed. 6.4.2
Creation of a Finite System of Linear Equations
We start by truncating the infinite summations to finite ones, as has been done in equations (6.105)–(6.108). We now have a finite system of linear equations. However, these equations must be satisfied simultaneously at all positions (x, y). Therefore, the system cannot be solved at this point. To get a system of K (K ∈ N) linear, independent equations with K unknowns that can be solved, we need to establish tangentialfield equalities in a weak form. To that end, we must, integrate the tangential electric and magnetic fields over the cross section. By applying an appropriate weighting function, we can arrive at the required finite system of linear equations. The appropriate weighting functions are provided by the transverse mode functions of the rectangular waveguide, owing to their orthogonality properties [9]. By utilizing these orthogonality properties (see appendix 6.10), we are able to reduce a double summation in the integration to a single term only, as will be shown in the remainder of this section. We start by taking the vector product of both sides of the electricfield continuity equations (equations (6.97) and (6.98)) with the complex conjugate of the magneticfield mode vector of waveguide I . Then a dot product is formed WITH the unit vector in the direction of propagation and, finally, the resulting equation is integrated over the cross section of the interface. Remembering that S I is the cross section of waveguide I , S II is the cross section of waveguide II and S is the crosssectional area of waveguide I , excluding the crosssectional area of waveguide II, we obtain # $ ∗ EIT × hITij · ˆiz dS = 0, i, j ∈ N, (6.109) S # # $ $ ∗ I∗ ˆ EIT × hITij · ˆiz dS = EII × h i, j ∈ N, (6.110) T Tij · iz dS, S II
S II
where EIT = ExI ˆix + EyI ˆiy , EII T and
=
ExII ˆix
+ EyII ˆiy
hITij = hIxij ˆix + hIyij ˆiy ,
i, j ∈ N.
(6.111) (6.112) (6.113)
The superscript asterisk denotes the complex conjugate.9 9 Equation (6.109) states in effect that the integral of the kernel over the crosssectional surface of waveguide I is equal to the integral of the same kernel over the crosssectional surface of waveguide II. This does not mean, however, that the choice of the modal weighting function is arbitrary. We have to choose the modal weighting function as that of waveguide I and understand that equation (6.109) is implicitly incorporated into equation (6.110). If we had taken the modal weighting function as that of waveguide II, than we would only have enforced the continuity of the electric field over the common aperture, and not over the complete discontinuity as we have done now [38].
240
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
Next, we take the vector product of both sides of the magneticfield continuity equation (equations (6.99) and (6.100)) with the complex conjugate of the electricfield mode vector of waveguide II. Then a dot product is formed with the unit vector in the direction of propagation and the resulting equation is integrated over the cross section of the interface: # # $ $ I II ∗ II ∗ ˆ ˆ HT × eTij · iz dS = HII T × eTij · iz dS, S II
S II
i, j ∈ N,
(6.114)
where HIT = HxI ˆix + HyI ˆiy ,
(6.115)
II ˆ II ˆ HII T = Hx ix + Hy iy
(6.116)
and eTIIij = exIIij ˆix + eyIIij ˆiy ,
i, j ∈ N.
(6.117)
Substitution of equations (6.42)–(6.45) into equations (6.110) and (6.114) leads to the following set of scalar equations:10 # $ +∗ +∗ ExI exI TE + EyI eyI TE dS ij ij I S# $ ∗ ∗ + + ExI exI TM + EyI eyI TM dS ij ij I S # ∗ ∗ $ + + HxI eyIITE − HyI exIITE dS ij ij S II # $ ∗ ∗ + + HxI eyIITM − HyI exIITM dS S II
ij
ij
= = = =
# II
S # S II
# II
S # S II
+∗
$
+∗
ExII exI TE + EyII eyI TE ij
+∗
dS,
ij
$
+∗
ExII exI TM + EyII eyI TM ij
ij
+∗
+∗
ij
ij
+∗
+∗
HxII eyIITE − HyII exIITE
HxII eyIITM − HyII exIITM ij
(6.118)
dS,
(6.119)
dS,
(6.120)
$
ij
$ dS,
(6.121)
where i, j ∈ N and where, in equations (6.118) and (6.119), use has been made of equation (6.109). Upon expanding the transverse field components in the above equations into mode contributions, the orthogonality characteristics of the modes (see Appendix 6.10) show that nontrivial solutions exist only for i = m and j = n in domain I and for i = l and j = w in domain II.
10 Equations (6.110) and (6.114) ensure conservation of complex power across the discontinuity. Another
implementation of the modematching method that is employed does not perform weighting with complex conjugate mode functions, but with the mode functions themselves. In this implementation, selfreaction across the discontinuity is ensured. The two formulations are equivalent as long as the structure is lossless, because the mode functions are then realvalued [38].
241
CROSSSECTIONAL STEP IN A RECTANGULAR WAVEGUIDE
Use of equations (6.38)–(6.41), (6.42)–(6.45), (6.46)–(6.49) and (6.50)–(6.53) results in &
+
−
AImn + AImn
'
+ L W II − II + II − [H1mnlw + H3mnlw ][AII lw + Alw ] + [H2mnlw + H4mnlw ][Alw − Alw ] , = [O1 (m, n, a, b) + O5 (m, n, a, b)] l=0 w=0 (6.122) & + ' − I AI mn − Amn + L W II − II + II − [H5mnlw + H7mnlw ][AII lw + Alw ] + [H6mnlw + H8mnlw ][Alw − Alw ] = , [O4 (m, n, a, b) + O8 (m, n, a, b)] l=0 w=0 (6.123)
where m = 0, 1, 2, . . . , M and n = 0, 1, 2, . . . , N, and &
+
−
II AII lw − Alw
'
=
I+ I− M N YI TEmn [H9lwmn + H11lwmn ][Amn − Amn ] m=0 n=0
II [O (l, w, a , b ) + O (l, w, a , b )] YTE 1 5 lw +
+ &
+ AII lw
− + AII lw
'
=
−
I I YTM [H10lwmn + H12lwmn ][AI mn + Amn ] mn II [O (l, w, a , b ) + O (l, w, a , b )] YTE 1 5 lw
,
(6.124)
I+ I− M N YI TEmn [H13lwmn + H15lwmn ][Amn − Amn ] m=0 n=0
II YTM [O4 (l, w, a , b ) + O8 (l, w, a , b )] lw +
+
−
I I YTM [H14lwmn + H16lwmn ][AI mn + Amn ] mn II YTM [O4 (l, w, a , b ) + O8 (l, w, a , b )] lw
,
(6.125)
where l = 0, 1, 2, . . . , L and w = 0, 1, 2, . . . , W . The functions Himnlw , i = 1, 2, . . . , 8, and Hjlwmn , j = 9, 10, . . . , 16, are the modecoupling integrals H1mnlw = H2mnlw = H3mnlw = H4mnlw =
a /2+xc x=−a /2+xc a /2+xc x=−a /2+x
c
a /2+xc x=−a /2+x
c
a /2+xc x=−a /2+xc
b /2+yc
+∗
y=−b /2+yc b /2+yc y=−b /2+y b /2+yc y=−b /2+y
lw
+∗
c
b /2+yc y=−b /2+yc
+
exI TEmn exIITM dx dy, lw
+∗
c
+
exI TEmn exIITE dx dy,
+
eyI TEmn eyIITE dx dy, lw
+∗
+
eyI TEmn eyIITM dx dy, lw
(6.126) (6.127) (6.128) (6.129)
242
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
H5mnlw = H6mnlw = H7mnlw = H8mnlw = H9lwmn = H10lwmn = H11lwmn = H12lwmn = H13lwmn = H14lwmn = H15lwmn = H16lwmn =
a /2+xc x=−a /2+xc a /2+xc x=−a /2+x
c
a /2+xc x=−a /2+x
c
a /2+xc x=−a /2+xc a /2+xc x=−a /2+xc a /2+xc x=−a /2+x
c
a /2+xc x=−a /2+x
c
a /2+xc x=−a /2+x
c
a /2+xc x=−a /2+xc a /2+xc x=−a /2+xc a /2+xc x=−a /2+x
c
a /2+xc x=−a /2+x
c
b /2+yc
+∗
y=−b /2+yc b /2+yc y=−b /2+y
c
c
b /2+yc
b /2+yc c
+
eyI TMmn eyIITM dx dy, lw
+
eyIITE eyI TEmn dx dy, lw
+
eyIITE eyI TMmn dx dy, lw
+∗
c
b /2+yc
+
exIITE exI TEmn dx dy, lw
+∗
c
b /2+yc
+
exIITE exI TMmn dx dy, lw
+∗
y=−b /2+yc b /2+yc
+
eyIITM eyI TEmn dx dy, lw
+∗
y=−b /2+yc b /2+yc
+
eyIITM eyI TMmn dx dy, lw
+∗
c
b /2+yc y=−b /2+y
lw
+∗
b /2+yc
y=−b /2+y
+
eyI TMmn eyIITE dx dy,
+∗
y=−b /2+yc
y=−b /2+y
lw
+∗
y=−b /2+yc
y=−b /2+y
+
exI TMmn exIITM dx dy, +∗
b /2+yc
y=−b /2+y
lw
+∗
b /2+yc y=−b /2+y
+
exI TMmn exIITE dx dy,
+
exIITM exI TEmn dx dy, lw
+∗
c
+
exIITM exI TMmn dx dy. lw
(6.130) (6.131) (6.132) (6.133) (6.134) (6.135) (6.136) (6.137) (6.138) (6.139) (6.140) (6.141)
The coupling integrals are calculated in Appendix 6.A. The normalization constants O1 , O4 , O5 and O8 are defined in Appendix 6.A. Before we can form the generalized scattering matrix of the waveguide junction, we first cast equations (6.122)–(6.125) into a more convenient form: &
L W % ' & + ' & + '( + − II − II II − Hamnlw AII + H A , AImn + AImn = + A − A b mnlw lw lw lw lw l=0 w=0
&
+
AI mn
m = 0, 1, . . . , M, n = 0, 1, . . . , N, (6.142) L W % ' & + ' & + '( − II − II II − Hcmnlw AII , − AI mn = lw + Alw + Hdmnlw Alw − Alw l=0 w=0
m = 0, 1, . . . , M,
n = 0, 1, . . . , N,
(6.143)
243
CROSSSECTIONAL STEP IN A RECTANGULAR WAVEGUIDE
&
+
−
II AII lw − Alw
'
=
M N %
& + ' & + '( − I − Helwmn AImn − AImn + Hflwmn AI , mn + Amn
m=0 n=0
&
+
' −
II AII lw + Alw
=
N % M
l = 0, 1, . . . , L, w = 0, 1, . . . , W, (6.144) & + ' & '( − + I − Hglwmn AImn − AImn + Hhlwmn AI , mn + Amn
m=0 n=0
l = 0, 1, . . . , L,
w = 0, 1, . . . , W,
(6.145)
where H1mnlw + H3mnlw , O1 (m, n, a, b) + O5 (m, n, a, b) H2mnlw + H4mnlw , = O1 (m, n, a, b) + O5 (m, n, a, b) H5mnlw + H7mnlw , = O4 (m, n, a, b) + O8 (m, n, a, b) H6mnlw + H8mnlw , = O4 (m, n, a, b) + O8 (m, n, a, b) I YTE H9lwmn + H11lwmn , = II mn YTElw O1 (l, w, a , b ) + O4 (l, w, a , b )
Hamnlw =
(6.146)
Hbmnlw
(6.147)
Hcmnlw Hdmnlw Helwmn
Hflwmn = Hglwmn = Hhlwmn =
I YTM mn
H10lwmn + H12lwmn , II O (l, w, a , b ) + O4 (l, w, a , b ) YTE 1 lw
I YTE mn
H13lwmn + H15lwmn , II YTMlw O4 (l, w, a , b ) + O8 (l, w, a , b ) I YTM H14lwmn + H16lwmn mn . II YTMlw O4 (l, w, a , b ) + O8 (l, w, a , b )
(6.148) (6.149) (6.150) (6.151) (6.152) (6.153)
The auxiliary functions defined by equations (6.146)–(6.153) are calculated in Appendix 6.A. 6.4.3
Matrix Formulation and GSM Derivation
Equations (6.142)–(6.145) can be written in matrix form as follows: I − II + II − II + II − [AI ]+ TE + [A ]TE = [Ha ]{[A ]TE + [A ]TE } + [Hb ]{[A ]TM − [A ]TM }, I − [AI ]+ TM − [A ]TM II − [AII ]+ TE − [A ]TE II − [AII ]+ TM + [A ]TM
= = =
II − II + II − [Hc ]{[AII ]+ TE + [A ]TE } + [Hd ]{[A ]TM − [A ]TM }, I − I + I − [He ]{[AI ]+ TE − [A ]TE } + [Hf ]{[A ]TM + [A ]TM }, I − I + I − [Hg ]{[AI ]+ TE − [A ]TE } + [Hh ]{[A ]TM + [A ]TM }.
(6.154) (6.155) (6.156) (6.157)
II ± In the above equations, [AI ]± TE and [A ]TE are column matrices (vectors) containing the (m, n) mode amplitude coefficients for TE waves traveling in the increasing and the decreasing z direction for waveguide I and waveguide II, respectively. Similarly, [AI ]± TM
244
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
and [AII ]± TM are column matrices containing the (m, n) mode amplitude coefficients for TM waves traveling in the increasing and the decreasing z direction for waveguide I and waveguide II, respectively. [Ha ], [Hb ], [Hc ] and [Hd ] are matrices containing the TEmn toTElw , TEmn toTMlw , TMmn toTElw and TMmn toTMlw mode coupling coefficients, respectively, going from waveguide I to waveguide II. So, in these matrices, the (m, n) variations are in the columns and the (l, w) variations are in the rows. [He ], [Hf ], [Hg ] and [Hh ] are matrices containing the TElw toTEmn , TElw toTMmn , TMlw toTEmn and TMlw toTMmn modecoupling coefficients, respectively, going from waveguide II to waveguide I . So, in these matrices, the (l, w) variations are in the columns and the (m, n) variations are in the rows. Using the composite matrices ) I ±
[A ] = [A ] =
* ,
±
) II ±
±
[AI ]TE [AI ]TM ±
[AII ]TE ±
[AII ]TM
* ,
equations (6.154)–(6.157) can be written as +
−
+
−
II −
I +
I −
[AI ] + [U I ][AI ] = [G1 ][AII ] + [G1 ][U I ][AII ] , II +
[A ] − [U II ][A ] = [G3 ][A ] − [G3 ][U II ][A ] ,
(6.158) (6.159)
where [G1 ] =
[Ha ] [Hb ] [He ] [Hf ] , [G3 ] = , [Hc ] [Hd ] [Hg ] [Hh ]
and [U I ] and [U II ] are modified unit matrices for waveguide I and waveguide II, respectively: [I ] [0] . [U I,I I ] = [0] −[I ] The difference between [U I ] and [U II ] is in the dimensions of the matrices. The minus sign before the bottom right identity submatrix is due to propagation to the left. The submatrices of the generalized scattering matrix for a reduction in a rectangular waveguide are related to the mode amplitude coefficients through −
+
−
II +
I +
II −
[AI ] = [S11 ]r [AI ] + [S12 ]r [AII ] , [A ] = [S21 ]r [A ] + [S22 ]r [A ] , where [S]r =
[S11 ] [S12 ] . [S21 ] [S22 ]
(6.160) (6.161)
JUNCTION BETWEEN A RECTANGULAR WAVEGUIDE AND A UNIT CELL
245
The substitution of equations (6.160) and (6.161) into equations (6.158) and (6.159) finally yields [S11 ]r = [U I ]([G1 ][G3 ] + [I ])−1 ([G1 ][G3 ] − [I ]),
(6.162)
[S12 ]r = 2[U I ]([G1][G3 ] + [I ])−1 [G1 ][U II ],
(6.163)
[S21 ] = [G3 ]([I ] − [U ][S11 ] ),
(6.164)
[S22 ] = [U ] − [G3 ][U ][S12 ] ,
(6.165)
r r
r
I
II
I
r
where [I ] is the unit matrix. The submatrices of the GSM for a waveguide crosssectional enlargement, [S]e , are constructed from the above submatrices by virtue of reciprocity:
6.5
[S11 ]e = [S22 ]r , [S12 ]e = [S21 ]r ,
(6.166) (6.167)
[S21 ]e = [S12 ]r , [S22 ]e = [S11 ]r .
(6.168) (6.169)
JUNCTION BETWEEN A RECTANGULAR WAVEGUIDE AND A UNIT CELL
In this section, we shall derive the generalized scattering matrix for a transition from one rectangular waveguide (out of the infinite number of rectangular waveguides that make up an infinite array antenna) to a unit cell, i.e. to a hom*ogeneous halfspace. The waveguide apertures of the array antenna are assumed to be positioned in an electrically conducting ground plane. The rectangular waveguide, with metallic walls, is assumed to be positioned centrally in a rectangular unit cell (which is a rectangular waveguide with phase shift walls) as shown in Figure 6.4. The transition is shown again in Figure 6.6. The input is the rectangular waveguide (with metallic walls), designated as waveguide I (z < 0). The crosssectional dimensions of this waveguide are a and b. The output is the rectangular unit cell (with phase shift walls), designated as waveguide II (z > 0). The crosssectional dimensions of the unit cell are s and t sin(). The crosssectional dimensions of the two waveguides are such that s ≥ a and t sin() ≥ b. The junction of the two waveguides is positioned at z = 0 in a rectangular coordinate system (Figure 6.6). Furthermore, S I is the area of the cross section of waveguide I (−a/2 ≤ x ≤ a/2, −b/2 ≤ y ≤ b/2), S II is the area of the cross section of waveguide II(−s/2 ≤ x ≤ s/2, −t/2 sin() ≤ y ≤ t/2 sin()) and S is the area of the cross section of waveguide II, excluding the cross section of waveguide I . S is assumed to be perfectly electrically conducting. The derivation of the generalized scattering matrix for the rectangularwaveguidetounitcell junction proceeds along the same lines as explained in detail in section 6.4 for a rectangularwaveguidetorectangularwaveguide junction. The only difference in comparison with that situation is in the modes in waveguide II. Therefore, in this section, we shall only outline the major steps in the derivation of the GSM.
246
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
y
z
s II z>0
tsin(Ω) x
b
(0,0,0) z>0 I
z=0 z<0 a
Figure 6.6 Rectangularwaveguidetounitcell junction in the plane z = 0.
6.5.1
GSM Derivation
Establishing tangentialfield continuity in a weak form across the waveguide aperture at z = 0 leads to &
M N % ' & + ' & + '( + II − I I− I I − AII = W A + W A , + A + A − A a b pqmn pqmn pq pq mn mn mn mn m=0 n=0
p = −P , . . . , −1, 0, 1, . . . , P ,
&
+
AII pq
q = −Q, . . . , −1, 0, 1, . . . , Q, (6.170) M N % ' & + ' & + '( − − I − = Wcpqmn AImn + AImn + Wdpqmn AI , − AII pq mn − Amn m=0 n=0
p = −P , . . . , −1, 0, 1, . . . , P , &
P ' + − AImn − AImn =
Q % & + Wemnpq AII pq
q = −Q, . . . , −1, 0, 1, . . . , Q, (6.171) ' & + '( − II II − − AII , pq + Wfmnpq Apq + Apq
p=−P q=−Q
& + AI mn
m = 0, 1, 2, . . . , M, n = 0, 1, 2, . . . , N, (6.172) Q P ' & + ' & + '( % − II − II II − Wgmnpq AII , + AI mn = pq − Apq + Whmnpq Apq + Apq p=−P q=−Q
m = 0, 1, 2, . . . , M,
n = 0, 1, 2, . . . , N.
(6.173)
247
JUNCTION BETWEEN A RECTANGULAR WAVEGUIDE AND A UNIT CELL +
−
In the above equations, which are exact for M, N, P , Q = ∞, AImn and AImn are the mode amplitude coefficients in waveguide I for the TEmn mode traveling in the increasing and the + I − decreasing z direction, respectively. AI mn and Amn are the mode amplitude coefficients in waveguide I for the TMmn mode traveling in the increasing and the decreasing z direction, + II − respectively. AII pq and Apq are the (Floquet) mode amplitude coefficients in waveguide II for the TEpqn mode traveling in the increasing and the decreasing z direction, respectively, and + II − AII pq and Apq are the (Floquet) mode amplitude coefficients in this waveguide for the TMpq mode traveling in the increasing and the decreasing z direction, respectively. The auxiliary functions Wapqmn , Wbpqmn , Wcpqmn , Wdpqmn , Wemnpq , Wfmnpq , Wgmnpq and Whmnpq are defined and calculated in Appendix 6.D, using Appendix 6.C. These equations may be written in matrix form as ( ( % % II − I + I − I + I − [AII ]+ TE + [A ]TE = [Wa ] [A ]TE + [A ]TE + [Wb ] [A ]TM − [A ]TM , ( ( % % II − I + I − I + I − [AII ]+ + [W − [A ] = [W ] [A ] + [A ] ] [A ] − [A ] c d TM TM TE TE TM TM , % % ( ( I − II + II − II + II − [AI ]+ TE − [A ]TE = [We ] [A ]TE − [A ]TE + [Wf ] [A ]TM + [A ]TM , ( ( % % I − II + II − II + II − + [W [AI ]+ + [A ] = [W ] [A ] − [A ] ] [A ] + [A ] g h TM TM TE TE TM TM ,
(6.174) (6.175) (6.176) (6.177)
II ± where [AI ]± TE and [A ]TE are column matrices (vectors) containing the (m, n) and (p, q) mode amplitude coefficients for TE waves traveling in the increasing and the decreasing z II ± directions in waveguide I and waveguide II, respectively. Similarly, [AI ]± TM and [A ]TM contain the (m, n) and (p, q) mode amplitude coefficients for TM waves traveling in the increasing and the decreasing z directions in waveguide I and waveguide II, respectively. [Wa ], [Wb ], [Wc ] and [Wd ] are matrices containing the TEmn toTEpq , TEmn toTMpq , TMmn toTEpq and TMmn toTMpq modecoupling coefficients, respectively, going from waveguide I to waveguide II. [We ], [Wf ], [Wg ] and [Wh ] are matrices containing the TEpq toTEmn , TEpq toTMmn , TMpq toTEmn and TMpq toTMmn mode coupling coefficients, respectively, going from waveguide II to waveguide I . Using the composite matrices
±
[AI ] = ) II ±
[A ] =
±
[AI ]TE ±
[AI ]TM ±
[AII ]TE ±
[AII ]TM
,
(6.178)
* (6.179)
,
equations (6.174)–(6.177) can be written as [AII ]+ + [U II ][AII ]− = [V1 ][AI ]+ + [V1 ][U I ][AI ]− , I +
I −
II +
II −
[A ] − [U ][A ] = [V3 ][A ] − [V3 ][U ][A ] , I
II
(6.180) (6.181)
248
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
where [Wa ] [Wb ] , [V1 ] = [Wc ] [Wd ] [We ] [Wf ] [V3 ] = , [Wg ] [Wh ]
(6.182) (6.183)
and [U I ] and [U II ] are modified unit matrices for waveguide I and waveguide II, respectively: [I ] [0] I,I I ]= . (6.184) [U [0] −[I ] The submatrices of the generalized scattering matrix for a rectangularwaveguidetounitcell junction are related to the mode amplitude coefficients through [AI ]− = [S11 ]u [AI ]+ + [S12 ]u [AII ]− , II +
I +
II −
[A ] = [S21 ] [A ] + [S22 ] [A ] . u
u
(6.185) (6.186)
Combining equations (6.180) and (6.181) with equations (6.185) and (6.186) finally results in [S11 ]u = [U I ]([I ] + [V3 ][V1])−1 ([I ] − [V3 ][V1 ]), u
[S12 ] = 2[U ]([I ] + [V3 ][V1 ]) I
−1
II
[V3 ][U ],
(6.187) (6.188)
[S21 ] = [V1 ]([I ] + [U ][S11 ] ),
(6.189)
[S22 ] = [V1 ][U ][S12 ] − [U ],
(6.190)
u u
I
u
I
u
II
where [I ] is the unit matrix. The submatrices of the GSM for a unitcelltorectangularwaveguide transition, [S]w , are constructed from the above submatrices by virtue of reciprocity: [S11 ]w = [S22 ]u , w
6.6
u
(6.191)
[S12 ] = [S21 ] , [S21 ]w = [S12 ]u ,
(6.192) (6.193)
[S22 ]w = [S11 ]u .
(6.194)
DIELECTRIC STEP IN A UNIT CELL
Only the GSM of a dielectric step in a unit cell will be treated explicitly. For a dielectric step in a metallicwall rectangular waveguide, we can use the GSM of a crosssectional enlargement or reduction in a waveguide. Since there is no such thing for a phaseshiftwall waveguide – the unit cell dimensions are uniquely related to the array lattice – we have to treat a step in the dielectric explicitly. A dielectric step in a unit cell will in practice be a step in the permittivity, but a step in the permeability or both is also allowed. The geometry of the configuration under consideration is shown in Figure 6.7.
249
DIELECTRIC STEP IN A UNIT CELL
y
z
II z>0
t sin (Ω)
x
(0,0,0) z=0 I z<0 s Figure 6.7 Dielectric step in rectangular unit cell.
The crosssectional dimensions of the unit cell are s and t sin(). A rectangular coordinate system is assumed, centrally positioned at the step interface in the substrate. In domain I , the relative permittivity and permeability of the substrate are denoted by εrI and µIr , respectively. The corresponding quantities in domain II are εrII and µII r . The derivation of the GSM follows the by know wellknown modematching procedure. It leads to a diagonal GSM matrix, the elements of which may be stored in a vector. 6.6.1
GSM Derivation
Establishing tangentialfield continuity in a weak form over the discontinuity leads to
&
' ' µIr & I + + II − I− AII pq + Apq = II Apq + Apq , µr p = −P , . . . , −1, 0, 1, . . . , P ,
q = −Q, . . . , −1, 0, 1, . . . , Q, (6.195)
250 &
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
+
−
II AII pq − Apq
'
=
I & ' γpq I + I − A − A pq pq , II γpq
p = −P , . . . , −1, 0, 1, . . . , P , &
q = −Q, . . . , −1, 0, 1, . . . , Q, (6.196)
' γ II & + ' + − pq II − AIpq − AIpq = I AII − A pq , γpq pq p = −P , . . . , −1, 0, 1, . . . , P ,
q = −Q, . . . , −1, 0, 1, . . . , Q, (6.197)
& + ' εII & + ' r I − II II − AI pq + Apq = I Apq + Apq , εr p = −P , . . . , −1, 0, 1, . . . , P ,
q = −Q, . . . , −1, 0, 1, . . . , Q, (6.198)
±
where the mode amplitude coefficients AI,II pq are as defined in previous sections. These equations may be written in matrix form as II − [AII ]+ TE + [A ]TE =
µIr I − {[AI ]+ TE + [A ]TE }, µII r
(6.199)
II − I + I − [AII ]+ TM − [A ]TM = [J1 ]{[A ]TM − [A ]TM },
[AI ]+ TE
− [AI ]− TE
= [J1 ]
I − [AI ]+ TM + [A ]TM =
−1
{[AII ]+ TE
εrII {[AII ]+ TM εrI
−
(6.200)
[AII ]− TE },
+ [AII ]− TM },
(6.201) (6.202)
where the column matrices (vectors) [AI,I I ]± TE,TM have their usual meaning, explained in previous sections, and [J1 ] is a diagonal matrix containing the TMtoTM coupling coefficients going from domain II to domain I . Using the usual composite column matrices, these equations may be written as +
−
+
I +
I −
II +
−
[AII ] + [U ][AII ] = [R1 ][AI ] + [R1 ][U ][AI ] , II −
[A ] − [U ][A ] = [R3 ][A ] − [U ][R3 ][A ] ,
(6.203) (6.204)
where I II (µr /µr )[I ] [0] [R1 ] = , [0] [J1 ] [J1 ]−1 [0] [R3 ] = , II [0] (εr /εrI )[I ] and where [U ] =
[I ] [0]
[0] . −[I ]
(6.205) (6.206)
(6.207)
251
FINITELENGTH TRANSMISSION LINE
The submatrices of the generalized scattering matrix for a dielectric step in a unit cell are related to the mode amplitude coefficients through −
+
−
[AI ] = [S11 ]s [AI ] + [S12 ]s [AII ] , II +
I +
II −
[A ] = [S21 ] [A ] + [S22 ] [A ] . s
s
(6.208) (6.209)
Combining equations (6.203) and (6.204) with equations (6.208) and (6.209) finally results in11 [S11 ]s = [U ]([I ] + [R3 ][R1 ])−1 ([I ] − [R3 ][R1 ]),
6.7
−1
(6.210)
[S12 ] = 2[U ]([I ] + [R3 ][R1 ]) [R3 ][U ], [S21 ]s = [R1 ]([I ] + [U ][S11 ]s ),
(6.211) (6.212)
[S22 ]s = [R1 ][U ][S12 ]s − [U ].
(6.213)
s
FINITELENGTH TRANSMISSION LINE
The last building block will be a finitelength transmission line. This transmission line can be either a piece of rectangular waveguide with metallic walls or a piece of rectangular waveguide with phase shift walls (a unit cell). The geometry is shown in Figure 6.8. The transmission line has crosssectional dimensions a and b if a rectangularwaveguide transmission line is considered and crosssectional dimensions s and t sin() if a ‘unit cell transmission line’ is considered. A rectangular coordinate system as shown in Figure 6.8 is assumed, where the origin is positioned at the center of the cross section of the transmission line at z = 0. The transmission line has a length d. The beginning of the transmission line (z = 0) is designated as domain I , and the end of the transmission line (z = d) is designated as domain II. In the following, we shall restrict ourselves to the derivation of the GSM for a rectangularwaveguide transmission line with metallic walls. The GSM for a ‘unit cell transmission line’ can then be obtained by inspection from the GSM for a waveguide transmission line, because of the similarity. The field components in domain II are obtained from those in domain I upon substitution I of a phase factor e−jγmn d in equations (6.38)–(6.41). To apply a modematching procedure in the plane z = 0, a coordinate transformation needs to be applied first to the field components in domain II. This coordinate transformation is implemented by replacing z by z − d in the descriptions of the fields for domain II.
11 Since the matrices [U ], [R ] and [R ] are all diagonal matrices, the scatteringmatrix elements may be calculated 1 3
directly (and stored in a vector) when implemented in software.
252
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
y
z
z=d (II) b; t sin(Ω)
x (0,0,0) z=0 (I)
z<0 a; s Figure 6.8 Finitelength transmission line.
6.7.1
GSM Derivation
Establishing tangentialfield continuity in a weak form across the interface leads to &
' & + I ' + − − −jγ I d jγmn d mn + AII , AImn + AImn = AII mn e mn e
m = 0, 1, 2, . . . , M, n = 0, 1, 2, . . . , N, ' & + I ' & + − I d I II jγmn d II − −jγmn = A , AI − A e − A e mn mn mn mn n = 0, 1, 2, . . . , N, ' ,
(6.215)
m = 0, 1, 2, . . . , M, n = 0, 1, 2, . . . , N, ' & + I ' & + − I d I II jγmn d II − −jγmn = A , AI + A e + A e mn mn mn mn
(6.216)
&
I+
I−
'
&
m = 0, 1, 2, . . . , M,
(6.214)
II +
Amn − Amn = Amn e
I d jγmn
II −
− Amn e
I d −jγmn
m = 0, 1, 2, . . . , M,
n = 0, 1, 2, . . . , N.
(6.217)
These equations may be written in matrix form as +
−
+
−
[AI ]TE + [AI ]TE = [D1 ][AII ]TE + [D2 ][AII ]TE ,
(6.218)
253
FINITELENGTH TRANSMISSION LINE +
−
+
−
+
−
[AI ]TE − [AI ]TE = [D1 ][AII ]TE − [D2 ][AII ]TE , + [AI ]TM + [AI ]TM
− + [AI ]TM − − [AI ]TM
= [D3 ][AII ]TM + [D4 ][AII ]TM , =
+ [D3 ][AII ]TM
− − [D4 ][AII ]TM ,
(6.219) (6.220) (6.221)
where the column matrices [AI,I I ]± TE,TM have their usual meaning, explained in previous sections, and [D1 ] and [D2 ] are diagonal matrices containing the TEtoTE coupling I coefficients. The nontrivial elements of [D1 ] are ejγmn d , and the nontrivial elements of [D2 ] I are e−jγmn d , for m = 0, 1, 2, . . . , M and n = 0, 1, 2, . . . , N. [D3 ] and [D4 ] are diagonal matrices containing the TMtoTM coupling coefficients. The nontrivial elements of [D3 ] I I are ejγmn d , and the nontrivial elements of [D4 ] are e−jγmn d , for m = 0, 1, 2, . . . , M and n = 0, 1, 2, . . . , N. [D1 ], and [D3 ], and similarly [D2 ] and [D4 ], differ only in size. Using the usual composite column matrices, the above equations may be written as +
−
+
−
I +
I −
II +
II −
[AI ] + [AI ] = [H1][AII ] + [H2 ][AII ] , [A ] − [A ] = [H1][A ] − [H2 ][A ] , [D1 ] [H1] = [0]
where
[0] [D3 ]
(6.223)
(6.224)
[D2 ] [0] [H2 ] = . [0] [D4 ]
and
(6.222)
(6.225)
The submatrices of the generalized scattering matrix for a finitelength rectangularwaveguide transmission line are related to the mode amplitude coefficients through −
+
−
+
I +
−
[AI ] = [S11 ]w [AI ] + [S12 ]w [AII ] , [AII ] = [S21 ]w [A ] + [S22 ]w [AII ] .
(6.226) (6.227)
Combining equations (6.222) and (6.223) with equations (6.226) and (6.227) finally results in [S11 ]w = [S22 ]w = [0], [S12 ]w = [S21 ]w = [H2 ].
(6.228) (6.229)
The derivation of the GSM of a finitelength ‘unit cell transmission line’, [S]d , is similar to the derivation of the GSM of a finitelength rectangularwaveguide transmission line. Therefore, for a ‘unit cell transmission line’ of length d, the results can be obtained by inspection from equations (6.228) and (6.229): [S11 ]d = [S22 ]d = [0],
(6.230)
[S12 ] = [S21 ] = [L],
(6.231)
d
d
where [L] is a diagonal matrix with nontrivial elements e−jγpq d for p = −P , . . . , −1, 0, 1, . . . , P and q = −Q, . . . , −1, 0, 1, . . . , Q. Since the submatrices are diagonal matrices, the nontrivial elements may be stored in a single vector. I
254
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
Figure 6.9 Waveguide structure consisting of cascaded rectangularwaveguide discontinuities and transmission lines.
6.8
OVERALL GSM OF A CASCADED RECTANGULARWAVEGUIDE STRUCTURE
Now that we have learned how to calculate the generalized scattering matrices of discontinuities in rectangular waveguides and unit cells, as well as GSMs for waveguide and unit cell transmission lines, it is necessary to develop a method for obtaining the GSM of a cascade of these transmission lines and discontinuities. An example of such a cascade is given in Figure 6.9. The example shown in the figure consists of a rectangular waveguide connected to an asymmetrically positioned smaller rectangular waveguide, which is connected symmetrically to a larger waveguide. This waveguide terminates into a symmetrical inductive iris [9] of finite thickness, which radiates into a unit cell that is filled with two different layers of dielectric material, of finite thickness. The last junction consists of a dielectric step to free space. The cascaded structure consists of six transmission lines (four waveguide transmission lines and two ‘unit cell transmission lines’) and six discontinuities. Of the several possible ways to combine the local GSMs, the most obvious one is to ‘directly’ cascade these GSM’s, i.e. to repeatedly construct a combined GSM from two local GSM’s. We have chosen this method, since alternative methods – aimed at improving the cascading process by reducing the number of matrix inversion operations – impose restrictions on the length of the transmission line for the applicability of the cascading process. Mansour and Macphie [28], for example, proposed a transmission matrix formulation to reduce the number of matrix inversion operations. Van Schaik [8, 33] also used transmission matrices. The major drawback of using a transmission matrix formulation is that diagonal matrices will evolve that have nontrivial elements of the form ejγ d , where d is the distance
255
OVERALL GSM OF A CASCADED RECTANGULARWAVEGUIDE STRUCTURE
a1 α
a1 β
a2 α
a2 β
α
β
b1α
b1β
b2 α
b2β
a1 γ
a2 γ
γ b1γ
b2 γ
Figure 6.10
Cascading of two 2ports.
between two discontinuities. If this distance becomes too large, it may give rise to exploding exponentials. If GSMs are cascaded ‘directly’, only diagonal matrices that have nontrivial elements of the form e−jγ d (see section 6.7) will evolve. These matrices will not give rise to numerical instabilities with increasing distance d. Besides numerical stability, a further advantage of cascading ‘directly’ is that it is not necessary to have an equal number of modes in the waveguides on both sides of a discontinuity [28, 39]. The admittance matrix formulation, as proposed by Alessandri et al. [40], is restricted to structures where the cascaded discontinuities are first of the enlargement type and second of the reduction type, thus not providing the degree of freedom desired for analyzing structures such as those shown in Figure 6.9. So, although the method of ‘directly’ cascading the local GSMs has the drawback of needing one matrix inversion operation per discontinuity, it guarantees numerical stability with respect to exponentials and general applicability. Truncation effects will be discussed in the next section. To visualize the cascading process, we consider two 2ports to be cascaded, as shown in Figure 6.10. The 2ports α and β are cascaded, resulting in the 2port γ . The scattering equations for the three 2ports in Figure 6.10 are, considering outgoing waves to be a result of ingoing waves, α α α α a1 + S12 a2 , b1α = S11 α α α α α b2 = S21 a1 + S22 a2 , β
β
β
β
β
β
(6.232) (6.233)
β
(6.234)
β
(6.235)
b1 = S11 a1 + S12 a2 , α a2 b2α = S21 a1 + S22
256
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
and γ
γ
γ
γ
γ
b1 = S11 a1 + S12 a2 ,
(6.236)
γ b2
(6.237)
=
γ γ S21 a1
+
γ γ S22 a2 .
Combining equations (6.232)–(6.237) leads to the scattering parameters of the 2port γ , expressed in terms of the (known) scattering parameters of the 2ports α and β. Owing to the direct relation between 2port scattering parameters and the scattering submatrices of an Nport [41], we find the following for the overall GSM of γ , which is the result of cascading the two local GSM’s of α and β: [S11 ]γ = [S11 ]α + [S12 ]α ([I ] − [S11 ]β [S22 ]α )−1 [S11 ]β [S21 ]α , α −1
[S12 ] = [S12 ] ([I ] − [S11 ] [S22 ] ) γ
α
β
[S12 ] ,
(6.239) α −1
[S21 ] = [S21 ] ([I ] + [S22 ] ([I ] − [S11 ] [S22 ] ) γ
β
α
β
α −1
[S22 ] = [S21 ] [S22 ] ([I ] − [S11 ] [S22 ] ) γ
β
α
β
(6.238)
β
[S11 ] )[S21 ] , β
α
[S12 ] + [S22 ] . β
β
(6.240) (6.241)
The above equations show that every cascading of two GSM’s requires one matrix inversion. The overall GSM of a complete waveguide structure (see for example Figure 6.9) is obtained by repeatedly using the cascading operation, reducing the number of local GSM’s in every iteration. From the overall GSM, we may find the reflection coefficients for all incident waves. The element GSM11 , for example, is the reflection coefficient for an incident TE10 wave.
6.9
VALIDATION
The theory discussed in the previous sections has been implemented in a computer code. To validate the theory and its implementation in code, analysis results will be compared, for several different structures, with simulation results from the literature and simulation results obtained with an independently developed GSM modematching code [31]. The validation will be obtained in two steps. First, waveguide structures without open boundaries will be analyzed. We shall refer to these structures as filter structures. Next, waveguide structures with open boundaries in the form of unit cells will be analyzed. These structures will be referred to as array antenna structures. Before we start with the validation, we need to discuss the choice of modes and the number of modes. To be able to perform analyses at all, we need to truncate the various infinite series. The way in which we truncate the series at both sides of a waveguide junction cannot be chosen arbitrarily. A phenomenon called relative convergence (RC) has to be taken into account. 6.9.1
Initial Choice of Modes
In general, a waveguide structure is designed for operation in the fundamental (TE10) input waveguide mode. When the structure is excited with this mode, the geometrical features of the crosssectional steps in the waveguide will dictate which modes should be taken into account
257
VALIDATION
TE10
TE10
TE10
TE10
(a)
(b)
(c)
(d)
Figure 6.11 Crosssectional discontinuities in a rectangular waveguide. (a) Symmetric inductive step. (b) Symmetric capacitive step. (c) Symmetric dual step. (d) Asymmetric dual step. Table 6.1 Mode preselection criteria for rectangularwaveguide discontinuities subject to an incident TE10 mode. Waveguide step
Excited modes
Restrictions
Symmetric inductive Symmetric capacitive Symmetric dual Asymmetric dual
TEm0 TE1n , TM1n TEmn , TMmn TEmn , TMmn
m odd None m odd, n even None
in an analysis of the discontinuity. Using only these modes in the analysis of the junction will enhance the efficiency of the computer code. Several possible crosssectional discontinuities in a rectangular waveguide are shown in Figure 6.11: a symmetric inductive step, a symmetric capacitive step, a symmetric dual step and an asymmetric dual step. Through rigorous but straightforward analysis of the TEtoTE, TEtoTM, TMtoTE and TMtoTM coupling coefficients given in equations (6.154)–(6.157), using the coupling integrals in Appendix 6.10, the modes excited at the discontinuities shown in Figure 6.11 may be found. For a TE10 mode incident into a rectangular waveguide that terminates in a symmetric inductive step (Figure 6.11(a)), only TEm0 modes will be excited, where m ∈ N and m is odd. For a TE10 mode incident into a rectangular waveguide that terminates in a symmetric capacitive step (Figure 6.11(b)), TE1n and TM1n modes will be excited, where n ∈ N. A TE10 mode incident into a rectangular waveguide that terminates in a symmetric dual step (Figure 6.11(c)) gives rise to TEmn and TMmn modes, where m, n ∈ N and m is odd and n is even. When the rectangular waveguide terminates in an asymmetrical dual step, all TE and TM modes are excited. Use of these mode preselection criteria, which are tabulated in Table 6.1 for convenience, will help in developing an efficient computer code for analyzing waveguide structures, avoiding the use of unnecessary matrix elements and thus memory.
258
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
What remains to be discussed now is how to appropriately truncate the infinite series encountered in the theory and how the truncated series at the two sides of a crosssectional discontinuity (step) in a waveguide relate to one another. 6.9.2
Relative Convergence and Choice of Modes
As explained in the foregoing, we have to truncate the infinite series to obtain a numerical solution to the problem of cascaded waveguide discontinuities. Because a digital computer will nearly always give a solution, even if the problem to be solved is not well posed or does not have a unique solution, we have to be careful about applying these truncations. It is only permitted to truncate an infinite series if the series is convergent. To validate the convergence, we have to look at the desired analysis parameters, and we consider the series at hand to be convergent – and thus the truncation to be appropriate – if the change in the parameters is smaller than certain criteria when the number of modes applied in the truncation is increased. This convergence criterion in itself, however, is not enough to ensure correct solutions for the analysis method that we are using! In the analysis of waveguide discontinuities by mode matching, as described in previous sections, we have to truncate two or more infinite series on both sides of the discontinuity simultaneously. It is known that when mode matching is applied, the numerical results may converge to different values depending on the way the series are truncated [42]. This phenomenon of relative convergence was first studied in [15] and was believed to originate from violation of field distributions at the edge of a conductor at a boundary, an explanation also found in [42]. In [43], it was explained that using the ‘cure’ for the RC problem stated in [15] indeed results in a bounded solution, but that it is not necessary to satisfy an explicit edge condition. It was shown that the origin of the RC phenomenon is in the behavior of the linear system to be solved. It was proved that the numerical results converge to the exact solution if the linear system is well conditioned. The ‘cure’ described in [15], which consists of taking the ratio of the numbers of modes on the two sides of a crosssectional discontinuity to be approximately equal to the ratio of the waveguide dimensions on the two sides, ensures a wellconditioned linear system. Thus a a1 1 if ≥1 ≥ a a M1 2 2 (6.242) a1 a1 M2 ≤ if ≤1 a2 a2 and
b 1 ≥ b2 N1 N2 ≤ b1 b2
if
b1 ≥1 b2
b1 if ≤ 1, b2
(6.243)
where the inequality sign applies to the most appropriate choice. In the above equations, M1 and M2 are the maximum mode indices m in waveguide 1 and waveguide 2, respectively, and N1 and N2 are the maximum mode indices n in waveguide 1 and waveguide 2, respectively,
259
VALIDATION
a2 b2
b1 a1
Figure 6.12
Asymmetric junction between two rectangular waveguides.
for TEmn and TMmn modes, where m, n ∈ N. The horizontal and vertical dimensions of waveguides 1 and 2 are a1 , a2 and b1 , b2 , respectively. The ‘RC condition’ stated in equations (6.242) and (6.243) ensures that the highest cutoff wave numbers on the two sides of the discontinuity are (almost) equal (see also equation (6.23)). Therefore the fastest transversely fluctuating signal at either side of the waveguide crosssectional discontinuity is supported by the signals at the other side, which is not true when the RC condition is violated. So, the spectral content in the Fourier representations is equal for the two sides. When the equality sign cannot be realized, we choose to have more modes in the larger waveguide, thus ensuring that the fastest fluctuation with the lowest mode number, on both sides of the junction, is accounted for. To visualize the RC phenomenon, we have analyzed an asymmetric junction between two rectangular waveguides where both transverse dimensions are different (Figure 6.12). The dimensions of the larger rectangular waveguide were a1 = 15.8 mm and b1 = 7.9 mm; those of the smaller waveguide were a1 = 11.85 mm and b1 = 6.0 mm. If we choose the maximum wave numbers in the larger guide, M1 and N1 , to be (M1 , N1 ) = (1, 1)12 and increase the maximum wave numbers in the smaller guide, M2 and N2 , we seem to reach convergence in the fundamentalmode reflection coefficient of the large waveguide as a function of frequency after only a few iterations (Figure 6.13). Although convergence is reached for our choice of input waveguide modes, the reflection coefficient results are not correct. We have clearly violated the RC condition, so that rapidly fluctuating transverse fields in the smaller waveguide are not supported by the fields in the larger one. Taking the RC condition into account, i.e. choosing the maximum mode numbers in accordance with the ratio of transverse waveguide dimensions, leads to convergence by increasing the maximum mode numbers at both sides of the junction (Figure 6.14). We observe that the reflection coefficient as a function of frequency now converges to different (correct) values. 12 This means that we allow the TE , TE , TE and TM modes to be present in the waveguide with the largest 01 10 11 11
cross section.
260
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
8
10
12
Reflection (dB)
14
16
18
20
22
24
(M2,N2)=(1,1) (M2,N2)=(2,2) (M2,N2)=(3,3) (M2,N2)=(4,4)
26
28
30 13.5 14 14.5 15 15.5 16 16.5 17 17.5 18 18.5 19 19.5 20 20.5 21 frequency (GHz) Figure 6.13 TE10 mode reﬂection coeﬃcient of the largercrosssectional waveguide versus frequency as a function of the maximum mode numbers in the smaller guide. (M1 , N1 ) = (1, 1).
As long as the system is well conditioned, increasing the number of modes in the input and output waveguides, regardless of the RC condition, eventually leads to convergence to the correct solution. However, choosing the modes according to equations (6.242) and (6.243) leads to the fastest convergence and prevents an overflow of memory. So, in all situations, the RC condition should be met. By meeting the RC condition, we ensure that the highest cutoff wave numbers on the two sides of the junction are identical, as can be concluded from inspection of equations (6.23), (6.32), (6.242) and (6.243), thus ensuring that the fastestfluctuating field along the transverse directions in either of the two guides is supported by the field in the other guide. It is this observation that will help us in selecting the Floquet modes when dealing with a junction formed by a metallicwall waveguide radiating into a freespace unit cell, for which situation no such relations as stated in equations (6.242) and (6.243) can be easily derived. From the above observation, a guideline can be found for selecting the Floquet modes necessary for ensuring a wellposed problem, thus eliminating the need for numerical experiments. First, the waveguide modes needed to represent the waveguide aperture are selected. Next, the cutoff wave numbers of the Floquet modes for a given scan angle (ϑ0 , ϕ0 ) are computed, and assembled in increasing order. All Floquet modes that have a cutoff wave number less than or equal to that of the cutoff wave number of the highest waveguide mode are used for the expansion in the external halfspace.
261
VALIDATION
8
10
Reflection (dB)
12
14
16
18
20
22
(M1,N1)=(2,2), (M2,N2)=(1,1) (M1,N1)=(4,4), (M2,N2)=(3,3) (M1,N1)=(7,7), (M2,N2)=(5,5) (M1,N1)=(11,11), (M2,N2)=(8,8)
24 13.5 14 14.5 15 15.5 16 16.5 17 17.5 18 18.5 19 19.5 20 20.5 21 frequency (GHz) Figure 6.14 TE10 mode reﬂection coeﬃcient of the largercrosssectional waveguide versus frequency as a function of the maximum mode numbers in both waveguides, chosen according to the RC condition.
We have now achieved a situation, by using the RC condition and the Floquet mode selection guideline, where ensuring the validity of an analysis of a waveguide structure is restricted to a convergence check only. We start by selecting a set of modes in an input rectangular waveguide. Next, the modes in the subsequent waveguide sections are chosen according to the RC condition, utilizing only the ratios of the transverse dimensions of the waveguide sections. Floquet modes in phaseshiftwall waveguides are chosen according to the described selection guideline described above. Next, we increase the set of modes in the input waveguide and change the modes in the subsequent waveguide sections and in free space, with use of the RC condition and the Floquet mode selection guideline, until the change in the parameter(s) obtained from the analysis is smaller than some chosen criterion, which depends on the required accuracy. By using the waveguide mode preselection criteria described in section 6.9.1, we may speed up the numerical analysis. Note that using the preselection criteria influences only the computational efficiency; it does not influence the accuracy of the analysis. The computational efficiency may be improved further by leaving out the highestorder modes in relatively long (with respect to wavelength) waveguide sections. Since these modes will decay rapidly with distance, their contribution may become negligible for a long enough waveguide section.
262
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS aout
Output waveguide
w10
Iris 2
Iris 2 Cavity 1 Iris 1
a9 w9
w3 l1
a2 w2
Input waveguide
l1 t
(a)
a1 w1 ain
(b)
Figure 6.15 Ninepole, Hstep rectangularwaveguide ﬁlter. (a) Layout. (b) Top view and dimensions.
With the mode preselection scheme, the convergence scheme described above and the additional measures to improve computational efficiency, we may now analyze waveguide structures. We shall start by demonstrating the validity of the method for filter structures. Then, we expand the cascaded metallicwall waveguide sections with a phaseshiftwall waveguide, thus demonstrating the validity of the method for infinite array antennas. 6.9.3
Filter Structures
For all of the metalwall waveguide structures for which analysis results will be shown in this section, convergence checks as discussed above were performed. Therefore, in general, the numbers of modes employed in the analyses were slightly more than necessary. We start with rectangularwaveguide filter structures employing symmetric inductive irises of finite thickness (i.e. length). The filter structures analyzed were derived from a ninepole waveguide filter described in [44]. The layout of the filter is shown in Figure 6.15, and the dimensions are stated in Table 6.2. In Figure 6.16, the scattering parameter S21 is shown as a function of frequency for the first cavity of the ninepole filter. The measurement results obtained from [44] are also shown in this figure. Since only symmetric inductive irises are present in the structure, only TEm0 modes, m ∈ N, were used in the analysis; 45 modes were used in the input waveguide. Figure 6.17 shows the analysis results for the first two cavities of the filter. For this analysis, 90 modes were used in the input waveguide. Both figures show fair to good agreement between the simulation and measurement results (the difference is less than 0.2 db). The differences are believed to be largely due to extracting the measurement results from the graphs in [44] and to waveguide wall losses that were not incorporated into the analysis.
263
VALIDATION
Table 6.2 Dimensions of the ninepole ﬁlter shown in Figure 6.15. The height of the waveguide sections was 4.318 mm (WR42 waveguide), and the thickness of the irises t was 2.0 mm. Cavity widths (mm) ain a1 a2 a3 a4 a5 a6 a7 a8 a9 aout
= 10.68 = 7.50 = 8.00 = 8.50 = 9.00 = 9.50 = 10.00 = 10.20 = 10.40 = 10.60 = 10.68
Iris widths (mm) w1 w2 w3 w4 w5 w6 w7 w8 w9 w10
Cavity lengths (mm)
= 5.364 = 3.675 = 3.183 = 3.010 = 2.924 = 2.879 = 2.872 = 2.919 = 3.181 = 4.964
l1 l2 l3 l4 l5 l6 l7 l8 l9
= 9.818 = 9.663 = 8.998 = 8.478 = 8.094 = 7.800 = 7.689 = 7.523 = 6.678
8 Visser Guglielmi
Reflection (dB)
8.5
9
9.5
10
10.5
11 23.15
23.2
23.25
23.3
23.35
23.4
23.45
frequency (GHz) Figure 6.16 Simulation and measurement results for the ﬁrst cavity of the ninepole ﬁlter of Figure 6.15.
The analysis results for the complete ninepole waveguide filter are shown in Figure 6.18. The agreement between the simulation and measurement results is fair to good, considering the low reflection levels and the rather coarse frequency resolution (f = 5.0 MHz). Next, we take a look at symmetrical dual steps. In Figure 6.19, the TE10mode reflection coefficient is shown as a function of frequency for a KubandtoXband concentric step. Measurement results obtained from [26] are also shown in this figure. For this simulation, we did not use mode preselection, but we did choose maximum mode numbers such as to meet
264
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
10
10.1
Visser Guglielmi
10.2
Reflection (dB)
10.3
10.4
10.5
10.6
10.7
10.8
10.9
11 23.2 23.22 23.24 23.26 23.28 23.3 23.32 23.34 23.36 23.38 23.4 23.42 23.44 frequency (GHz) Figure 6.17 Simulation and measurement results for the ﬁrst two cavities of the ninepole ﬁlter of Figure 6.15.
the RC condition. For the Kuband waveguide (15.8 mm × 7.9 mm), we chose (M1 , N1 ) = (9, 6), and for the Xband waveguide (22.86 mm×10.16 mm,) we chose (M2 , N2 ) = (13, 9). We observe a good match between the simulations and measurements from 12.5 GHz onwards (the difference is less than 0.2 db). The large difference at 10 GHz between the measurement and simulations is also present in [26], which casts suspicion on the measurement value. The other differences are believed to originate from extracting the measurement results from the graphs in [26]. This assumption is strengthened by the good agreement between the simulation and measurement results (from [26]) over the whole frequency band for a resonant iris, shown in Figure 6.20. Again, we did not use mode preselection. For the waveguide (15.8 mm × 7.9 mm), we chose (M1 , N1 ) = (13, 9) and for the iris (11.17 mm × 5.59 mm, thickness 2 mm),we chose (M1 , N1 ) = (9, 6). To check the validity of the calculated TE10 mode reflection coefficient in terms of both amplitude and phase, we looked at an H plane step discontinuity as analyzed in [45]. Simulation results, together with the simulation results from [45], are shown in Figure 6.21. The width of the larger waveguide was 22.86 mm, and that of the smaller waveguide was 15 mm. The height of both waveguides was 10.16 mm. In the larger waveguide, seven modes were used (TE10, TE20 , TE30 , TE40 , TE50 , TE60 and TE70), and in the smaller waveguide, five modes were used (TE10, TE20, TE30, TE40 and TE50 ), meeting the RC condition. The results match very well: the difference in amplitude is less than 0.2 dB, and the difference in phase is less than 2◦ . As a last validation check for filter structures, we looked at an asymmetric iris (11.85 mm × 6.0 mm) of finite thickness (3.0 mm), placed in a Kuband waveguide
265
VALIDATION
0 S11, Visse r S21, Visse r S11, Guglielm i S21, Guglielm i
10
S11, S21 (dB)
20 30 40 50
60
70
80 23.1
23.15
23.2
23.25 23.3 23.35 frequency (GHz)
23.4
23.45
23.5
Figure 6.18 Simulation and measurement results for the complete ninepole ﬁlter of Figure 6.15.
(15.8 mm × 7.9 mm) (Figure 6.22). The analysis results for the TE10mode reflection coefficient are compared with measurement results from [26] in Figure 6.22. With the exception of the values below −25 db, the simulation and measurement data agree reasonably well (the differences are less than 1.5 db). Here also, errors were introduced by extracting measurement data from the graphs in [26]. Since the simulated results in [26] for levels below −25 db differ in the same way from the measurement results as in Figure 6.22, we may attribute these differences to measurement errors around the resonance frequency. Now that the method and its software implementation have been thoroughly validated for filter structures, we have to demonstrate validation when radiation into a unit cell is incorporated. 6.9.4
Array Antenna Structures
For the evaluation of large, uniformly excited phased array antennas consisting of openended waveguide radiators, it suffices to analyze the reflection coefficient of the fundamental mode of the feeding waveguide. First of all, the reflection coefficient tells us directly about the impedance matching and, secondly, through the scan element pattern, we can learn directly about the radiation pattern of the phased array. 6.9.4.1 Scan Element Pattern As explained in [37, 46] for a linear array, the gain of a fully, uniformly excited array scanned to the angular position (ϑ0 , ϕ0 ), in that direction,
266
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
0 Visser Patzelt & Arndt
5
10
15
20
25 11
12
13
14
15
16
17
18
19
20
frequency (GHz) Figure 6.19 Simulated and measured reﬂection coeﬃcient versus frequency for a concentric KutoXband step.
5
Visser Patzelt & Arndt
10
15
20
25
30
35
40
45 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 frequency (GHz) Figure 6.20
Simulated and measured reﬂection coeﬃcient versus frequency for a resonant iris.
267
VALIDATION
2
170 Visser ampl Weisshaar ampl . Visser phase Weisshaar phase
2
4
6
150
130
)
8
10
110
12
14
90
16
18
70
20
22
50
8.5
9
9.5 10 10.5 11 11.5 12 12.5 13 13.5 14 14.5 15 frequency (GHz)
Figure 6.21 Simulated and measured reﬂection coeﬃcient versus frequency for an H plane step discontinuity.
Ga (ϑ0 , ϕ0 ), is given by Ga (ϑ0 , ϕ0 ) ≈ KGsk (ϑ0 , ϕ0 ),
(6.244)
where K is the number of elements and Gsk (ϑ0 , ϕ0 ) is the scan element pattern for the angular position (ϑ0 , ϕ0 ). The scan element pattern is the radiation pattern of a single excited element in its array environment, when all other elements are terminated into matched loads. The scan element pattern is given by (6.245) Gsk (ϑ, ϕ) = Ge (ϑ, ϕ)[1 − (ϑ, ϕ)2 ], where (ϑ, ϕ) is the scan reflection coefficient and Ge (ϑ, ϕ) is the gain of an isolated element. Since all coupling effects are contained in the scan reflection coefficient, the pattern of the phased array antenna may now be constructed by multiplying the array factor by the scan element pattern. All anomalies in the scan reflection coefficient as a function of scan angle – such as ‘dips’ or ‘nulls’ – will appear in the scan element pattern and thus in the pattern of the phased array antenna. 6.9.4.2 Infinite OpenEnded Waveguide Arrays To validate the theory and the computer code that we developed with results from the open literature,13 we refer to the array layout
13 Most large, openended waveguide phased array antennas have been designed and realized for military radar and
therefore only a limited amount of publicly available analysis and measurement data exists.
268
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
0 Visser Patzelt & Arndt
5
10
15
20
25
30
35
40 10.5 11 11.5 12 12.5 13 13.5 14 14.5 15 15.5 16 16.5 17 17.5 18 frequency (GHz) Figure 6.22 Simulated and measured reﬂection coeﬃcient versus frequency for a ﬁnitethickness asymmetric iris placed in a waveguide.
shown in Figure 6.23. The grid here is triangular in general, the rectangular grid ( = π/2) being a special case. The dielectric sheets, combined with the aperture iris, serve the purpose of wideangle impedance matching. We shall make our first comparisons with analysis results for the CAISSA antenna [8,33], for which simulation and measurement data is publicly available. For the CAISSA antenna (consisting of 849 irisloaded rectangularwaveguide elements and an external matching sheet), s = 0.942λ0, t = 0.5444λ0, = 30.097◦, a = 0.659λ0, b = d = 0.217λ0, c = 0.650λ0, t = 0.000, τ1 = τ2 = 0.0942λ0, ε1 = 1.000 and ε2 = 2.300, where λ0 is the freespace wavelength. The calculated reflection coefficients for a number of scan positions in the H plane (ϕ0 = 0), D plane (ϕ0 = 45◦ ) and E plane (ϕ0 = 90◦)14 are shown in Table 6.3, as calculated with the theory of [47], the theory of [17], the theory of [8,33] and the current theory, for an infinite antenna without a matching structure, i.e. without irises and a matching sheet. In Table 6.4, results are shown for the CAISSA antenna without a matching sheet, for different iris sizes, with the beam scanned to (ϕ0 , ϑ0 ) = (45◦ , 60◦ ). (M1 , N1 ) give the maximum mode numbers in the feeding waveguide, and (M2 , N2 ) give the maximum mode numbers in the aperture. Finally, in Table 6.5, results for the CAISSA antenna with its internal and external matching structures are given.
14 The x axis is parallel to the long side of the waveguides, and the y axis is parallel to the short side.
269
VALIDATION
s tsin(Ω) Ω b
d c
a
εr1
εr2
τ1
τ2
t
Figure 6.23 Inﬁnite array of openended waveguide radiators, having a ﬁnitethickness iris in the waveguide aperture and two dielectric sheets in front of the array aperture. Table 6.3 Calculated reﬂection coeﬃcients for the CAISSA antenna with c = b and ε2 = 1.000. ϕ0 0 0 0 0 45 45 45 90 90
ϑ0
[47]
[17]
[8]
0 20 40 60 20 40 60 40 60
0.19∠7.29◦
0.18∠9.63◦
0.18∠13.21◦ 0.26∠ −4.53◦ 0.48∠ −24.36◦ 0.71∠ −34.79◦ 0.20∠1.46◦ 0.27∠ −23.74◦ 0.31∠ −55.52◦ 0.04∠6.97◦ 0.17∠ −163.96◦
0.18∠10.79◦ 0.27∠ −5.26◦ 0.48∠ −24.12◦ 0.71∠ −36.64◦ 0.21∠0.18◦ 0.27∠ −23.64◦ 0.31∠ −54.92◦ 0.05∠4.01◦ 0.16∠ −164.78◦
0.27∠ −6.49◦ 0.48∠ −23.67◦ 0.70∠ −33.98◦ 0.21∠ −1.99◦ 0.28∠ −23.80◦ 0.31∠ −53.93◦ 0.06∠ −3.85◦ 0.15∠ −162.90◦
0.27∠ −6.16◦ 0.48∠ −24.72◦ 0.71∠ −35.11◦ 0.21∠ −0.81◦ 0.28∠ −24.22◦ 0.32∠ −54.81◦ 0.06∠2.73◦ 0.15∠ −165.26◦
Looking at Table 6.4, we see a very poor match, especially in the phase. The reason must be due to the infinitelythiniris model employed in [8, 17]. However, in Tables 6.3 and 6.5, we observe a much better correspondence between the results of the various calculations. In general, we can say that our modematching method appears to produce correct results, but at the same time we see that it is very difficult to make this judgment on a pointtopoint basis. It is better to make the comparisons over a range of scan angles for a geometrically fixed array antenna system. In the remainder, the comparisons that we shall make will be of this type. To start with, we analyze a square grid array and compare the results with analysis results and measurement results for a 19 × 19 element array reported in [19]. The relevant array
270
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
Table 6.4 Calculated reﬂection coeﬃcients for the CAISSA antenna without sheet, scanned to (45◦ , 60◦ ). c/a 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3
(M1 , N1 ) (7,2) (7,2) (8,2) (10,2) (10,2) (10,2) (10,2) (10,2)
(M2 , N2 )
[17]
[8]
(7,2) (6,2) (6,2) (7,2) (6,2) (5,2) (4,2) (3,2)
0.32∠304.86◦ 0.29∠315.29◦ 0.25∠343.46◦ 0.27∠25.80◦ 0.47∠81.08◦ 0.86∠136.80◦ 0.91∠145.05◦ 0.89∠148.06◦
0.31∠304.35◦ 0.29∠308.53◦ 0.26∠324.20◦ 0.23∠2.62◦ 0.37∠63.33◦ 0.59∠98.50◦ 0.87∠137.51◦ 0.84∠140.08◦
0.31∠10.79◦ 0.27∠354.74◦ 0.23∠335.88◦ 0.26∠323.36◦ 0.50∠0.18◦ 0.72∠336.36◦ 0.93∠305.08◦ 0.97∠4.010◦
Table 6.5 Calculated reﬂection coeﬃcients for the CAISSA antenna with its matching structures. ϕ0 0 0 0 0 45 45 45 90 90 90
ϑ0
[17]
[8]
0 20 40 60 20 40 60 20 40 60
0.27∠102.27◦ 0.24∠90.72◦ 0.18∠51.16◦ 0.19∠334.41◦ 0.23∠98.79◦ 0.08∠66.58◦ 0.18∠287.71◦ 0.22∠108.03◦ 0.10∠129.35◦ 0.05∠142.14◦
0.24∠105.50◦ 0.21∠90.25◦ 0.17∠39.90◦ 0.21∠329.60◦ 0.19∠102.21◦ 0.04∠56.79◦ 0.21∠279.87◦ 0.19∠115.31◦ 0.10∠158.96◦ 0.08∠194.71◦
0.30∠103.42◦ 0.27∠92.69◦ 0.21∠59.38◦ 0.18∠337.53◦ 0.26∠100.20◦ 0.11∠76.78◦ 0.14∠294.48◦ 0.25∠108.71◦ 0.12∠124.51◦ 0.07∠131.33◦
dimensions, referring to Figure 6.23, are a = b = c = d = 0.5354λ0, s = t = 0.5714λ0, = 90◦ and εr1 = εr2 = 1.000, where λ0 is the freespace wavelength. Figure 6.24 shows the amplitude of the H plane scan reflection coefficient as calculated in [19], as calculated with the theory described in this chapter and as measured in [19]. The phase, as calculated in [19] and as calculated with the modematching method of this chapter, is shown in Figure 6.25. The maximum number of modes applied in the waveguide aperture is given by (M, N) = (5, 5). The figures show fair agreement up to 48◦ in the amplitude (the difference is less than 1.5 db) for both simulation methods. The measurements fail to register the deep ‘null’ around 48◦ . Most likely the null is (partly) filled owing to coupling effects in the finite array. These couplings are also visible in the fluctuations of the measured amplitude of the reflection coefficient as a function of the scan angle around the values for the reflection coefficient of the infinite array. The relatively high reflection coefficients measured for scan angles beyond 48◦ are probably caused by the same effect that has filled up the null. Obviously, the 19 × 19 array is not ‘infinite enough’ to realize the low reflection coefficient amplitudes predicted by the infinitearray antenna model. The phase as a function of the scan angle calculated with the current method agrees well with the value calculated in [19]; the difference over all scan angles is less than 5◦ .
271
VALIDATION
0 Visser Wu & Galindo theory Wu & Galindo measurement
Reflection (dB)
5
10
15
20
25
30 0
5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 Scan angle (degrees)
Figure 6.24 Amplitude of reﬂection coeﬃcient for a squaregrid, openended waveguide array, as calculated for an inﬁnite array and measured for a 19 × 19 element array.
To compare the analysis results for openended waveguide array antennas with thick irises in the apertures (t = 0 in Figure 6.23), we have to resort to published analysis results [48]. We start with an infinite array of openended waveguide radiators without a matching structure. The relevant parameters for this geometry (Figure 6.23) are a = c = 0.660λ0, b = d = 0.159λ0, s = 0.856λ0, t = 0.495λ0, = 32.440◦ and εr1 = εr2 = 1.000, where λ0 is the freespace wavelength. The amplitude of the reflection coefficient for an H plane scan (ϕ0 = 0) and for an Eplane scan (ϕ0 = 90◦ ) is shown in Figure 6.26, as calculated in [48] and as calculated by the modematching method described in this chapter. The number of waveguide modes was set to 42, with (M, N) = (8, 2), identical to the number of modes used in [48]. The phase as a function of the scan angle is shown in Figure 6.27. Considering the difficulty of extracting the analysis data from the graphs in [48], the results of our analysis match those of [31] well. The difference in amplitude is less than 1.7 dB, except around resonance, and the difference in phase is less than 4.5◦, excluding the last Eplane analysis result from [48]. Finally, we shall compare analysis results for an infinite array of openended waveguides, arranged in a triangular lattice, having a thick iris in the waveguide aperture. The relevant dimensions are a = 0.660λ0, b = d = 0.159λ0, c = 0.500λ0, s = 0.856λ0, t = 0.495λ0, = 32.440◦, t = 0.015λ0 and εr1 = εr2 = 1.000, where λ0 is the freespace wavelength. The array was analyzed at f = 1.17f0 , where f0 is the frequency corresponding to the wavelength λ0 . The amplitude of the reflection coefficient for an H plane scan (ϕ0 = 0) and for an Eplane scan (ϕ0 = 90◦) is shown in Figure 6.28, as calculated in [48] and as calculated
272
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
180
Argument of Reflection (degrees)
170
160
150
140
130
120
110
100
90
80
Visser Wu & Galindo
70
60
50 0
5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 Scan angle (degrees)
Figure 6.25 Phase of reﬂection coeﬃcient for a squaregrid, openended waveguide array, as calculated for an inﬁnite array.
by the modematching method described in this chapter, for f = 1.17f0. The number of waveguide aperture modes was set to 42, with (M, N) = (8, 2), identical to the number of modes used in [48]. The number of waveguide modes in the feeding waveguide then met the RC condition, with (M, N) = (11, 2). The phase as a function of the scan angle is shown in Figure 6.29. Again considering the difficulty of extracting the analysis data from the graphs in [48], the results of our analysis match those of [31] well. The difference in amplitude is less than 2.5 db over all scan angles, and the difference in phase is less than 5.0◦, excluding the last Eplane analysis result from [48]. Our method and its software implementation have now also been thoroughly validated for planar, infinite, openended waveguide array antennas.
6.10
CONCLUSIONS
A versatile analysis method has been developed for waveguide structures and infinite waveguide phased array antennas. The method is based on mode matching, which is used to construct generalized scattering matrices for waveguidetowaveguide and waveguidetounitcell junctions. The code developed has been thoroughly validated by comparing analysis results with analysis and measurement results from the open literature. A mode preselection scheme, a convergence strategy and a Floquet mode selection scheme have been developed and have been applied to the structures analyzed. The convergence strategy, which avoids the
APPENDIX 6.A. WAVEGUIDE MODE ORTHOGONALITY AND NORMALIZATION FUNCTIONS
273
5
10
Reflection (dB)
15
20
25
30
35
40
Visser, Hplane Chan et al., Hplane Visser, Eplane Chan et al., Eplane
45
50 0
5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 Scan angle (degrees)
Figure 6.26 Amplitude of reﬂection coeﬃcient for an inﬁnite array of openended, rectangular waveguides in a triangular grid.
occurrence of relative convergence, together with the Floquet mode selection scheme based on the maximum wave number in the waveguide aperture, makes it possible for the code to be used by nonspecialists. The theory presented in this chapter may be regarded as a ‘classic’ modematching approach for analyzing large arrays of openended waveguides, stemming from work done in the mid 1990s. Although valuable in itself, as demonstrated in this chapter, it may also serve in understanding the progress that has been made since then in analyzing this kind of antennas, amongst others, the realization of uniform convergence by virtue of the concept of the multimode equivalent network [2–7].
APPENDIX 6.A. WAVEGUIDE MODE ORTHOGONALITY AND NORMALIZATION FUNCTIONS In section 6.2, we decomposed the electric and magnetic fields in a rectangular waveguide into modes. The mode vectors, eTTEmn for the electric field of a TE mode and eTTMmn for the electric field of a TM mode, possess orthogonality characteristics, i.e. each mode vector eTTEmn or eTTMmn is orthogonal to all other mode vectors [9, 36]:
eTTEmn · eTTEij dS =
eTTMmn
1 · eTTMij dS = 0
for m = i ∧ n = j for m = i ∨ n = j
274
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
20
Argument of reflection (degrees )
20
40
60
80
100
120
140 Visser, Hplane Chan et al., Hplane Visser, Eplane Chan et al., Eplane
160
180
200 0
5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 Scan angle (degrees)
Figure 6.27 Phase of reﬂection coeﬃcient for an inﬁnite array of openended, rectangular waveguides in a triangular grid.
and
eTTEmn · eTTMij dS = 0.
(6.A.1)
The fields of the magnetic modes, hTTEmn and hTTMmn , follow from the fields of the electric modes through [9] hTTEmn = ˆiz × eTTEmn (6.A.2) and
hTTMmn = ˆiz × eTTMmn .
(6.A.3)
Taking the mode functions of section 6.2, equations (6.45)–(6.49), into account, these orthogonality relations can be written in terms of the following pair of integrals, NC(ξ, η) = NS(ξ, η) =
1/2
−1/2 1/2 −1/2
Cξ (κ)Cη (κ) dκ,
(6.A.4)
Sξ (κ)Sη (κ) dκ,
(6.A.5)
where the cosine function Cξ (κ) and sine function Sξ (κ) are defined by equations (6.54) and (6.55), respectively: 1 (6.A.6) Cξ (κ) = cos ξ π κ + 2
275
APPENDIX 6.A. WAVEGUIDE MODE ORTHOGONALITY AND NORMALIZATION FUNCTIONS
5
Reflection (dB)
10
15
20
25
Visser, Hplane Chan et al., Hplane Visser, Eplane Chan et al., Eplane
30
35 0
5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 Scan angle (degrees)
Figure 6.28 Amplitude of reﬂection coeﬃcient for an inﬁnite array of openended, rectangular waveguides with a thick aperture iris in a triangular grid, evaluated at f = 1.17f0 .
1 . Sξ (κ) = sin ξ π κ + 2 Calculation of the integrals NC(ξ, η) and NS(ξ, η) yields 1 if ξ = η = 0 1 NC(ξ, η) = if ξ = η = 0 2 0 if ξ = η, 1 if ξ = η = 0 1 NS(ξ, η) = if ξ = η = 0 2 0 if ξ = η. and
(6.A.7)
(6.A.8)
(6.A.9)
Using the above, we can calculate the normalization functions Oi (ξ, η, α, β), i = 1, 2, . . . , 8, ξ, η ∈ N, α, β ∈ R, as α/2 β/2 ∗ O1 (ξ, η, α, β) = ex+TE ex+TE dx dy x=−α/2 y=−β/2
=
ηπ αβ (ωµ0 µr )2 β kT4 ξη
ij
ξη
2 NC(i, ξ )NS(j, η),
(6.A.10)
276
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
Argument of reflection (degrees )
200 Visser, Hplane Chan et al., Hplane Visser, Eplane Chan et al., Eplane
150
100
50
50 0
5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 Scan angle (degrees)
Figure 6.29 Phase of reﬂection coeﬃcient for an inﬁnite array of openended, rectangular waveguides with a thick aperture iris in a triangular grid, evaluated at f = 1.17f0 .
α/2
O2 (ξ, η, α, β) =
β/2
∗
x=−α/2 y=−β/2
ex+TE ex+TM dx dy ij
ξη
αβ ηπ ξπ = − 4 (ωµ0 µr )γξ η NC(i, ξ )NS(j, η), α β kTξη α/2 β/2 ∗ O3 (ξ, η, α, β) = ex+TM ex+TE dx dy ij ξη x=−α/2 y=−β/2 αβ ξπ ηπ = − 4 (ωµ0 µr )γξ∗η NC(i, ξ )NS(j, η), α β kTξη α/2 β/2 ∗ ex+TM ex+TM dx dy O4 (ξ, η, α, β) =
ij
x=−α/2 y=−β/2
= O5 (ξ, η, α, β) =
αβ ∗ 2 ξ π γξ η  α kT4 ξη α/2 β/2
(6.A.11)
(6.A.12)
ξη
2 NC(i, ξ )NS(j, η),
(6.A.13)
∗
x=−α/2 y=−β/2
ey+TE ey+TE dx dy
=−
ξπ αβ (ωµ0 µr )2 α kT4 ξη
ij
ξη
2 NS(i, ξ )NC(j, η),
(6.A.14)
APPENDIX 6.B. MODECOUPLING INTEGRALS FOR WAVEGUIDETOWAVEGUIDE JUNCTION
α/2
β/2
∗
ey+TE ey+TM dx dy ij ξη x=−α/2 y=−β/2 ξπ αβ ηπ = 4 (ωµ0 µr )γξ η NS(i, ξ )NC(j, η), α β kTξη α/2 β/2 ∗ ey+TM ey+TE dx dy O7 (ξ, η, α, β) = O6 (ξ, η, α, β) =
ij
x=−α/2 y=−β/2
=
ij
x=−α/2 y=−β/2
=
αβ kT4 ξη
γξ∗η 2
ηπ β
(6.A.15)
ξη
αβ ηπ ∗ ξπ NS(i, ξ )NC(j, η), (ωµ µ )γ 0 r ξη α β kT4 ξη α/2 β/2 ∗ O8 (ξ, η, α, β) = ey+TM ey+TM dx dy
277
(6.A.16)
ξη
2 NS(i, ξ )NC(j, η).
(6.A.17)
In the above, i, j ∈ N. Solutions exist only for i = ξ and j = η. Note that O2 (ξ, η, α, β) + O6 (ξ, η, α, β) = 0 and O3 (ξ, η, α, β) + O7 (ξ, η, α, β) = 0.
APPENDIX 6.B. MODECOUPLING INTEGRALS FOR WAVEGUIDETOWAVEGUIDE JUNCTION In the modematching procedure for a rectangularwaveguidetowaveguide junction we encountered, in section 6.4.2, the modecoupling integrals Himnlw , i = 1, 2, . . . , 8 and Hjlwmn , j = 9, 10, . . . , 16. These coupling integrals can be expressed in terms of the following set of integrals: 1/2α+β Cξ (κ)Cη (α[κ − β]) dκ (6.B.1) CC(ξ, η, α, β) = κ=−1/2α+β
and
SS(ξ, η, α, β) =
1/2α+β
Sξ (κ)Sη (α[κ − β]) dκ.
(6.B.2)
κ=−1/2α+β
In the above two equations, ξ, η ∈ N and α, β ∈ R. The cosine function Cξ (κ) and sine function Sξ (κ) are defined by equations (6.54) and (6.55), respectively:
and
1 Cξ (κ) = cos ξ π κ + 2
(6.B.3)
1 Sξ (κ) = sin ξ π κ + . 2
(6.B.4)
278
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
The calculation of the integrals CC(η, ξ, α, β) and CC(η, ξ, α, β) results in CC(η, ξ, α, β) 1 α ξ 1 α + 1 + 2αβ η (−1) sin ξ π π ξ 2 − (ηα)2 2α = α − 1 + 2αβ − sin ξ π 2α α − 1 + 2αβ 1 cos ηπ 2α 2 and
SS(η, ξ, α, β) =
0 ηα
CC(η, ξ, α, β) ξ CC(η, ξ, α, β)
if ξ = ηα = 0
if ξ = ηα
(6.B.5)
if ξ = ηα = 0
if ξ = 0 ∨ η = 0 if ξ = ηα ∧ ξ = 0
(6.B.6)
if ξ = ηα = 0.
The coupling integrals are then found as (ωµ0 )2 µIr µII r (nπ/b)(wπ/b )ab [(mπ/a)2 + (nπ/b)2 ][(lπ/a )2 + (wπ/b )2 ] a xc b yc × CC m, l, , SS n, w, , , a a b b II lb −YTElw H1 if w = 0 = wa mnlw 0 if w = 0,
H1mnlw =
H2mnlw
(ωµ0 )2 µIr µII r (mπ/a)(lπ/a )ab [(mπ/a)2 + (nπ/b)2 ][(lπ/a )2 + (wπ/b )2 ] b yc a xc × CC n, w, , SS m, l, , , b b a a II wa YTElw H3mnlw if l = 0 = lb 0 if l = 0, −Y I ∗ mb H if n = 0 1 TEmn = na mnlw 0 if n = 0, Y I ∗ Y II mlbb H if n = 0 ∧ w = 0 1 TEmn TElw nwaa mnlw = 0 if n = 0 ∨ w = 0,
(6.B.7)
(6.B.8)
H3mnlw =
H4mnlw
H5mnlw
H6mnlw
(6.B.9)
(6.B.10)
(6.B.11)
(6.B.12)
APPENDIX 6.B. MODECOUPLING INTEGRALS FOR WAVEGUIDETOWAVEGUIDE JUNCTION
H7mnlw =
H8mnlw
∗ Y I
TEmn
na H3 mb mnlw
if m = 0
0 if m = 0, Y I ∗ Y II nwaa H if m = 0 ∧ l = 0 3 TEmn TElw mlbb mnlw = 0 if m = 0 ∨ l = 0,
H9lwmn = H3mnlw , na Y I H3mnlw TEmn mb H10lwmn = 0
279
(6.B.13)
(6.B.14) (6.B.15)
if m = 0
(6.B.16)
if m = 0,
H11lwmn = H1mnlw , −Y I mb H1 if n = 0 TEmn na mnlw H12lwmn = 0 if n = 0, Y II ∗ wa H 3mnlw if l = 0 TElw lb H13lwmn = 0 if l = 0, Y I Y II ∗ nwaa H if m = 0 ∧ l = 0 3 TEmn TElw mlbb mnlw sH14lwmn = 0 if m = 0 ∨ l = 0, −Y II ∗ lb H if w = 0 1 TElw wa mnlw H15lwmn = 0 if w = 0, Y I Y II ∗ mlbb H if n = 0 ∧ w = 0 1 TEmn TElw nwaa mnlw H16lwmn = 0 if n = 0 ∨ w = 0.
(6.B.17) (6.B.18)
(6.B.19)
(6.B.20)
(6.B.21)
(6.B.22)
The coupling integrals, combined and normalized, give rise to the following functions, which are used as matrix elements in section 6.4.2: H1mnlw + H3mnlw O1 (m, n, a, b) + O5 (m, n, a, b) 2 b nπ 2kT4 mn a 2 = [H1mnlw + H3mnlw ] ab(ωµ0 µIr )2 mπ 2 2 kTmn
Hamnlw =
if m = 0 ∧ n = 0 if m = 0 ∧ n = 0 if m = 0 ∧ n = 0,
(6.B.23)
280
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
Hbmnlw =
lb H2mnlw + H4mnlw wa II = − H1mnlw + H3mnlw YTE lw O1 (m, n, a, b) + O5 (m, n, a, b) wa lb 0 if l = 0 ∧ w = 0 2 b if l = 0 ∧ w = 0 ∧ m = 0 ∧ n = 0 nπ 4 2kTmn × (6.B.24) a 2 I 2 ab(ωµ0 µr ) if l = 0 ∧ w = 0 ∧ m = 0 ∧ n = 0 mπ 2 if l = 0 ∧ w = 0 ∧ m = 0 ∧ n = 0, 2 kTmn
H5mnlw + H7mnlw O4 (m, n, a, b) + O8 (m, n, a, b) 4kT2 mn na mb 0 if m = 0 ∨ n = 0 I∗ H3 (6.B.25) = − H1mnlw + Y I 2 1 if m = 0 ∧ n = 0, na mb mnlw TEmn abγmn mlbb H6mnlw + H8mnlw nwaa = − = H + H 1 3 O4 (m, n, a, b) + O8 (m, n, a, b) nwaa mnlw mlbb mnlw 4kT2 mn 0 if m = 0 ∨ n = 0 ∨ l = 0 ∨ w = 0 I∗ II × YTE Y (6.B.26) mn TElw I 2 abγmn  1 if m = 0 ∧ n = 0 ∧ l = 0 ∧ w = 0,
Hcmnlw =
Hdmnlw
I YTE mn
H9lwmn + H11lwmn II O (l, w, a , b ) + O5 (l, w, a , b ) YTE 1 lw 2 b wπ I 2 2kT4 lw YTE a mn = [H1mnlw + H3mnlw ] 2 Y II a b (ωµ0 µII ) lπ r TElw 2 2 kTlw
Helwmn =
Hflwmn =
I YTM mn II YTE lw
if l = 0 ∧ w = 0 if l = 0 ∧ w = 0 if l = 0 ∧ w = 0, (6.B.27)
mb H10lwmn + H12lwmn na = − H1mnlw + H3 O1 (l, w, a , b ) + O5 (l, w, a , b ) na mb mnlw 0 if m = 0 ∧ n = 0 2 b if m = 0 ∧ n = 0 ∧ l = 0 ∧ w = 0 wπ I 4 YTMmn 2kTmn 2 I × YTE a mn II II 2 a b (ωµ µ ) YTElw if m = 0 ∧ n = 0 ∧ l = 0 ∧ w = 0 0 r lπ 2 if m = 0 ∧ n = 0 ∧ l = 0 ∧ w = 0, 2 kTlw (6.B.28)
APPENDIX 6.C. UNIT CELL MODE ORTHOGONALITY AND NORMALIZATION FUNCTIONS I YTE mn
H13lwmn + H15lwmn O4 (l, w, a , b ) + O8 (l, w, a , b ) I 4kT2 lw YTE lb wa 0 mn II ∗ = − H1mnlw + H3mnlw YTElw II II 2 1 wa lb YTMlw a b γlw
Hglwmn =
281
II YTM lw
if l = 0 ∨ w = 0 if l = 0 ∧ w = 0, (6.B.29)
I YTM mn
H14lwmn + H16lwmn O4 (l, w, a , b ) + O8 (l, w, a , b ) nwaa mlbb H1 + H3 = − nwaa mnlw mlbb mnlw I 4kT2 lw YTM 0 if m = 0 ∨ n = 0 ∨ l = 0 ∨ w = 0 mn I II ∗ × YTEmn YTElw II II 2 1 if m = 0 ∧ n = 0 ∧ l = 0 ∧ w = 0. YTMlw a b γlw (6.B.30)
Hhlwmn =
II YTM lw
APPENDIX 6.C. UNIT CELL MODE ORTHOGONALITY AND NORMALIZATION FUNCTIONS The modematching procedure for a rectangularwaveguidetounitcell junction gives rise to a number of normalization functions, denoted by Ui (ξ, η, α, β), i = 1, 2, . . . , 8, ξ, η ∈ N, α, β ∈ R. These normalization functions incorporate two normalization integrals: α/2 NEU(α, ξ, ζ ) = e−j(2ξ π/α−2ζ π/α)κ dκ (6.C.1) κ=−α/2
and
NEV(β, ξ, η, ζ, ς ) =
β/2
e−j(2ηπ/β−2ςπ/β+2βζ π−2βξ π)κ dκ,
(6.C.2)
κ=−β/2
where ζ, ς ∈ N. These normalization integrals possess the mode orthogonality characteristics and can be calculated as α if ξ = ζ NEU(α, ξ, ζ ) = (6.C.3) 0 if ξ = ζ, β if η = ς ∧ ξ = ζ NEV(β, ξ, η, ζ, ς ) = (6.C.4) 0 if η = ς ∨ ξ = ζ. The normalization functions are defined and calculated as follows: α/2 ∗ ex+TE ex+TE dx dy U1 (ξ, η, α, β) = x=−α/2
=
ij
ξη
1 (ωµ0 µr )2 vξ2η NEU(α, ξ, i)NEV(β, ξ, η, i, j ), kT4 ξη
(6.C.5)
282
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
U2 (ξ, η, α, β) =
α/2 x=−α/2
=
1 kT4 ξη
U3 (ξ, η, α, β) =
α/2
1 kT4 ξη
U4 (ξ, η, α, β) =
ij
ξη
(ωµ0 µr )γξ η uξ 0 vξ η NEU(α, ξ, i)NEV(β, ξ, η, i, j ),
x=−α/2
=
∗
ex+TE ex+TM dx dy
∗
ex+TM ex+TE dx dy ij
ξη
(ωµ0 µr )γξ∗η uξ 0 vξ η NEU(α, ξ, i)NEV(β, ξ, η, i, j ),
α/2 x=−α/2
∗
ij
ξη
1 γξ η 2 u2ξ 0 NEU(α, ξ, i)NEV(β, ξ, η, i, j ), kT4 ξη α/2 ∗ U5 (ξ, η, α, β) = ey+TE ey+TE dx dy ij
x=−α/2
1 kT4 ξη
U6 (ξ, η, α, β) = U7 (ξ, η, α, β) = U8 (ξ, η, α, β) =
x=−α/2 α/2 x=−α/2 α/2 x=−α/2
=
∗
ey+TE ey+TM dx dy = −U2 (ξ, η, α, β), ij
ξη
∗
ey+TM ey+TE dx dy = −U3 (ξ, η, α, β), ij
(6.C.8)
ξη
(ωµ0 µr )2 u2ξ 0 NEU(α, ξ, i)NEV(β, ξ, η, i, j ),
α/2
(6.C.7)
ex+TM ex+TM dx dy
=
=
(6.C.6)
ξη
(6.C.9) (6.C.10) (6.C.11)
∗
ey+TM ey+TM dx dy ij
ξη
1 γξ η 2 vξ2η NEU(α, ξ, i)NEV(β, ξ, η, i, j ). kT4 ξη
(6.C.12)
In the above, i, j ∈ N. Nontrivial solutions exist only for i = ξ and j = η.
APPENDIX 6.D. MODECOUPLING INTEGRALS FOR RECTANGULARWAVEGUIDETOUNITCELL JUNCTION In the modematching procedure for a rectangularwaveguidetounitcell junction we encountered, in section 6.4.2, the modecoupling integrals Wipqmn , i = 1, 2, . . . , 8, and Wjmnpq , j = 9, 10, . . . , 16. These coupling integrals make use of the following set of (Fourier) integrals: FCU(α, ξ, η) =
α/2
Cξ κ=−α/2
κ juη0 κ dκ, e α
(6.D.1)
283
APPENDIX 6.D. MODECOUPLING INTEGRALS
κ −juη0 κ FCUm(α, ξ, η) = Cξ dκ, e α κ=−α/2 α/2 κ juη0 κ FSU(α, ξ, η) = Sξ dκ, e α κ=−α/2 α/2 κ −juη0 κ FSUm(α, ξ, η) = Sξ dκ, e α κ=−α/2 α/2 κ jvητ κ FCV(α, ξ, η, τ ) = Cξ dκ, e α κ=−α/2 α/2 κ −jvητ κ FCVm(α, ξ, η, τ ) = Cξ dκ, e α κ=−α/2 α/2 κ jvητ κ FSV(α, ξ, η, τ ) = Sξ dκ, e α κ=−α/2 α/2 κ −jvητ κ FSVm(α, ξ, η, τ ) = Sξ dκ, e α κ=−α/2
α/2
(6.D.2) (6.D.3) (6.D.4) (6.D.5) (6.D.6) (6.D.7) (6.D.8)
where ξ, η, τ ∈ N and α ∈ R. The transversewavenumber components uη0 and vητ are defined by equations (6.64) and (6.65), respectively. The cosine function Cξ (κ) and sine function Sξ (κ) are defined by equations (6.54) and (6.55), respectively. We repeat these expressions here for completeness: 1 Cξ (κ) = cos ξ π κ + (6.D.9) 2 and
1 Sξ (κ) = sin ξ π κ + . 2
(6.D.10)
Calculation of the above integrals gives α if ξ = uη0 = 0 ξπ α −jξ π/2 e = uη0 if FCU(α, ξ, η) = 2 α uη0 ξπ uη0 , = {(−1)ξ ejuη0 α/2 − e−juη0 α/2 } if j α (ξ π /α)2 − u2η0 (6.D.11) FCUm(α, ξ, η) = (−1)ξ F CU (α, ξ, η), 0 α −jξ π/2 j e FSU(α, ξ, η) = 2 (ξ π/α) {(−1)ξ ejuη0 α/2 − e−juη0 α/2 } − (ξ π /α)2 − u2η0
(6.D.12) if ξ = 0 ξπ if = uη0 = 0 α ξπ if = uη0 , α (6.D.13)
284
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
FSUm(α, ξ, η) = −(−1)ξ FSU(α, ξ, η), α α −jξ π/2 FCV(α, ξ, η, τ ) = 2 e vητ ξ jv α/2 − e−jvητ α/2 } j (ξ π /α)2 − v 2 {(−1) e ητ ητ FCVm(α, ξ, η, τ ) = (−1)ξ FCV(α, ξ, η, τ ), 0 α −jξ π/2 FSV(α, ξ, η, τ ) = 2 e (ξ π /α) {(−1)ξ ejvητ α/2 − e−jvητ α/2 } − 2 (ξ π/α)2 − vητ
(6.D.14) if ξ = vητ = 0 ξπ = vητ if α ξπ = vητ , if α (6.D.15) (6.D.16) if ξ = 0 ξπ if = vητ = 0 α ξπ if = vητ , α (6.D.17)
FSVm(α, ξ, η, τ ) = −(−1)ξ FSV(α, ξ, η, τ ).
(6.D.18)
The coupling integrals are then found as
W1pqmn = = W2pqmn = =
=
b/2
−Y I
W3pqmn =
W4pqmn =
+∗
+
exIITEpq exI TEmn dx dy
x=−a/2 y=−b/2 (ωµ0 )2 µIr µII r (nπ /b)vpq FCU(a, m, p)FSV(b, n, p, q), −j 2 2 ] [(mπ/a) + (nπ /b)2][u2p0 + vpq a/2 b/2 +∗ + exIITEpq exI TMmn dx dy x=−a/2 y=−b/2
=
a/2
TEmn
0 a/2
mb W1pqmn na
if n = 0
TEmn
(6.D.20)
if n = 0, b/2
+∗
+
eyIITEpq eyI TEmn dx dy
x=−a/2 y=−b/2 (ωµ0 )2 µIr µII r (mπ/a)up0 FSU(a, m, p)FCV(b, n, p, q), −j 2 ] [(mπ/a)2 + (nπ/b)2 ][u2p0 + vpq a/2 b/2 +∗ + eyIITEpq eyI TMmn dx dy x=−a/2 y=−b/2
Y I
(6.D.19)
na W3pqmn mb
if m = 0 if m = 0,
(6.D.21)
(6.D.22)
285
APPENDIX 6.D. MODECOUPLING INTEGRALS
W5pqmn = = W6pqmn = =
a/2
= W8pqmn = =
−Y I
+∗
+
exIITMpq exI TEmn dx dy
TEmn
0 a/2
mb W5pqmn na
if n = 0
Y I
W10mnpq = W11mnpq = W12mnpq =
TEmn
0 a/2
b/2
(6.D.24)
+∗
+
eyIITMpq eyI TEmn dx dy
na W7pqmn mb
b/2
b/2
x=−a/2 y=−b/2 a/2
b/2
x=−a/2 y=−b/2 a/2
b/2
x=−a/2 y=−b/2 a/2
if m = 0
(6.D.26)
b/2
+∗
+
+∗
+
eyI TEmn eyIITEpq dx dy = (−1)m+n W3pqmn ,
eyI TEmn eyIITMpq dx dy = (−1)m+n +∗
W7pqmn , II ∗
II γpq
γpq
+
exI TEmn exIITEpq dx dy = (−1)m+n W1pqmn , ∗
+ + exI TEmn exIITMpq dx dy +∗
= (−1)
m+n
W5pqmn , II ∗
II γpq
γpq
(6.D.27) (6.D.28) (6.D.29) (6.D.30)
+
eyI TMmn eyIITEpq dx dy x=−a/2 y=−b/2 na (−1)m+n Y I ∗ W3pqmn if m = 0 TEmn = mb 0 if m = 0, a/2 b/2 ∗ + + = eyI TMmn eyIITMpq dx dy x=−a/2 y=−b/2 II na (−1)m+n γpq Y I ∗ W7pqmn if m = 0 TEmn II ∗ mb = γpq 0 if m = 0,
W13mnpq =
(6.D.25)
if m = 0,
x=−a/2 y=−b/2 a/2
(6.D.23)
if n = 0,
x=−a/2 y=−b/2 II ∗ (mπ/a)v ωµ0 µIr γpq pq FSU(a, m, p)FCV(b, n, p, q), j 2 2 2 2 ] [(mπ/a) + (nπ /b) ][up0 + vpq a/2 b/2 +∗ + eyIITMpq eyI TMmn dx dy x=−a/2 y=−b/2
W9mnpq =
W14mnpq
b/2
x=−a/2 y=−b/2 II ∗ (nπ/b)u ωµ0 µIr γpq p0 FCU(a, m, p)FSV(b, n, p, q), −j 2 2 2 2 ] [(mπ/a) + (nπ /b) ][up0 + vpq a/2 b/2 +∗ + exIITMpq exI TMmn dx dy x=−a/2 y=−b/2
W7pqmn =
(6.D.31)
(6.D.32)
286
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
b/2
+∗
+
exI TMmn exIITEpq dx dy x=−a/2 y=−b/2 mb (−1)m+n Y I ∗ W1pqmn if n = 0 TEmn na = 0 if n = 0, a/2 b/2 +∗ + = exI TMmn exIITMpq dx dy x=−a/2 y=−b/2 II mb −(−1)m+n γpq Y I ∗ W5pqmn if n = 0 ∗ TEmn II na = γpq 0 if n = 0.
W15mnpq =
W16mnpq
a/2
(6.D.33)
(6.D.34)
Combining the coupling integrals gives rise to the following functions, which are used as matrix elements in section 6.5.1: W1pqmn + W3pqmn
Wapqmn =
Wbpqmn
U1 (p, q, s, t sin()) + U5 (p, q, s, t sin()) kT2 pq = {W1pqmn + W3pqmn }, 2 (ωµ0 µII r ) st sin() W2pqmn + W4pqmn = U1 (p, q, s, t sin()) + U5 (p, q, s, t sin()) I kT2 pq YTE mn = 2 (ωµ0 µII r ) st sin() na − mb W W3pqmn if m = 0 ∧ n = 0 1pqmn + na mb × 0 if m = 0 ∨ n = 0, U4 (p, q, s, t sin()) + U8 (p, q, s, t sin()) kT2 pq = II 2 {W5pqmn + W7pqmn }, γpq  st sin() W6pqmn + W8pqmn = U4 (p, q, s, t sin()) + U8 (p, q, s, t sin()) I kT2 pq YTE mn = II 2 γpq  st sin() na − mb W W if m = 0 ∧ n = 0 + 5pqmn 7pqmn na mb × 0 if m = 0 ∨ n = 0,
Wemnpq =
(6.D.36)
W5pqmn + W7pqmn
Wcpqmn =
Wdpqmn
(6.D.35)
II YTE pq
W9mnpq + W11mnpq
I YTE mn
O1 (m, n, a, b) + O5 (m, n, a, b)
(6.D.37)
(6.D.38)
287
APPENDIX 6.D. MODECOUPLING INTEGRALS
=2
II YTE pq
kT4 mn
(−1)m+n {W1pqmn + W3pqmn } I I )2 ab (ωµ µ YTE 0 r mn 2 b if m = 0 ∧ n = 0 nπ a 2 × if m = 0 ∧ n = 0 mπ 2 if m = 0 ∧ n = 0, 2 kTmn
Wfmnpq = =
II YTM pq
W10mnpq + W12mnpq I YTEmn O1 (m, n, a, b) + O5 (m, n, a, b) II II YTM kT4 mn γpq pq (−1)m+n {W5pqmn 2 I II ∗ (ωµ µI )2 ab YTEmn γpq 0 r
2 b nπ a 2 × mπ 2 2 kTmn Wgmnpq = =
+ W7pqmn }
if m = 0 ∧ n = 0 if m = 0 ∧ n = 0
(6.D.40)
if m = 0 ∧ n = 0,
II YTE pq
W13mnpq + W15mnpq I YTMmn O4 (m, n, a, b) + O8 (m, n, a, b) II YTE kT2 mn pq I∗ (−1)m+n 4 I YTE mn I 2 ab γmn YTMmn
mb na × − W1pqmn + W3pqmn na mb 1 if m = 0 ∧ n = 0 × 0 if m = 0 ∨ n = 0, Whmnpq =
(6.D.39)
II YTM pq
(6.D.41)
W14mnpq + W16mnpq
I O4 (m, n, a, b) + O8 (m, n, a, b) YTM mn II II YTM kT2 mn γpq pq I∗ (−1)m+n YTE =4 I mn II ∗ γ I 2 ab γpq YTMmn mn mb na × − W5pqmn + W7pqmn na mb 1 if m = 0 ∧ n = 0 × 0 if m = 0 ∨ n = 0.
(6.D.42)
288
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
REFERENCES 1. R.C. Hansen, Phased Array Antennas, John Wiley & Sons, New York, 1998. 2. H.J. Visser and M. Guglielmi, ‘CAD of waveguide array antennas based on “filter” concepts’, IEEE Transactions on Antennas and Propagation, Vol. 47, No. 3, pp. 542– 548, March 1999. 3. D. Bakers, Finite Array Antennas: An Eigencurrent Approach, PhD thesis, Eindhoven University of Technology, 2004. 4. B. Morsink, Fast Modeling of Electromagnetic Fields for the Design of Phased Array Antennas in Radar Systems, PhD thesis, Eindhoven University of Technology, 2005. 5. S. Monni, Frequency Selective Surfaces Integrated with Phased Array Antennas: Analysis and Design Using Multimode Equivalent Networks, PhD thesis, Eindhoven University of Technology, 2005. 6. D.J. Bekers, S.J.L. van Eijndhoven, A.A.F. van de Ven, P.P. Borsboom and A.G. Tijhuis, ‘Eigencurrent analysis of resonant behavior in finite array antennas’, IEEE Transactions on Microwave Theory and Techniques, Vol. 54, No. 6, pp. 2821–2829, June 2006. 7. S. Monni, G. Gerini, A. Neto and A.G. Tijhuis, ‘Multimode equivalent networks for the design and analysis of frequency selective surfaces’, IEEE Transactions on Antennas and Propagation, Vol. 55, No. 10, pp. 2824–2835, October 2007. 8. H.J. van Schaik, Theory and Performance of a SpaceFed, Planar, Phased Array Antenna with 849 IrisLoaded Rectangular Waveguide Elements and External Matching Sheet, PhD thesis, Delft University of Technology, 1979. 9. N. Marcuvitz, Waveguide Handbook, McGrawHill, New York, 1951. 10. W.P.M.N. Keizer, A.P. de Hek and A.B. Smolders, ‘Theoretical and experimental performance of a wideband widescan angle rectangular waveguide array antenna’, Proceedings of the IEEE Antennas and Propagation International Symposium, APS1991, pp. 1724–1727, 24–28 June 1991. 11. G. Gerini, ‘Phased arrays of compact waveguide radiators and frequency selective surfaces: A multimode equivalent network approach’, Proceedings of the 12th International Conference on Antennas and Propagation, pp. 454–457, March–April 2003. 12. B.J. Morsink, G.H.C. van Werkhoven, A.G. Tijhuis and S.W. Rienstra, ‘An accelerated coupled feedradiatorfrequency selective surface model for the next generation phased array systems’, Proceedings of the IEEE International Symposium on Phased Array Systems and Technology, pp. 470–475, October 2003. 13. A. Ellgardt, ‘Study of rectangular waveguide elements for planar wideanglescanning phased array antennas’, Proceedings of the IEEE Antennas and Propagation Society International Symposium, pp. 815–818, July 2005.
REFERENCES
289
14. E.G. Magill and H.A. Wheeler, ‘Wideangle impedance matching of a planar array antenna by a dielectric sheet’, IEEE Transactions on Antennas and Propagation, Vol. AP14, pp. 49–53, January 1966. 15. R. Mittra and S.W. Lee, Analytical Techniques in the Theory of Guided Waves, Macmillan, New York, 1971. 16. J.J. Campbell and B.V. Popovich, ‘A broadband wideangle scan matching technique for large environmentally restricted phased arrays’, IEEE Transactions on Antennas and Propagation, Vol. AP20, pp. 421–427, July 1972. 17. S.W. Lee and W.R. Jones, ‘On the suppression of radiation nulls and broadband impedance matching of rectangular waveguide phased arrays’, IEEE Transactions on Antennas and Propagation, Vol. AP18, pp. 41–51, January 1971. 18. C.C. Chen, ‘Broadband impedance matching of rectangular waveguide phased arrays’, IEEE Transactions on Antennas and Propagation, Vol. AP21, pp. 298–302, May 1973. 19. V. Galindo and C.P. Wu, ‘Numerical solutions for an infinite phased array of rectangular waveguides with thick walls’, IEEE Transactions on Antennas and Propagation, Vol. AP14, pp. 652–654, September 1966. 20. V. Galindo and C.P. Wu, ‘Properties of a phased array of rectangular waveguides with thick walls’, IEEE Transactions on Antennas and Propagation, Vol. AP14, pp. 163– 173, March 1966. 21. G.A. Gesell and I.R. Ciric, ‘Recurrence modal analysis for multiple waveguide discontinuities and its application to circular structures’, IEEE Transactions on Microwave Theory and Techniques, Vol. 41, No. 3, pp. 484–490, March 1993. 22. U. Papziner and F. Arndt, ‘Field theoretical computeraided design of rectangular and circular iris coupled rectangular or circular waveguide cavity filters’, IEEE Transactions on Microwave Theory and Techniques, Vol. 41, No. 3, pp. 462–471, March 1993. 23. U. Tucholke, F. Arndt and T. Wriedt, ‘Field theory design of square waveguide iris polarizers’, IEEE Transactions on Microwave Theory and Techniques, Vol. 34, No. 1, pp. 156–160, January 1986. 24. G.L. James, ‘Analysis and design of TE11toHE11 corrugated cylindrical waveguide mode converters’, IEEE Transactions on Microwave Theory and Techniques, Vol. 29, No. 10, pp. 1059–1066, October 1981. 25. O.P. Franza and W.C.C. Chew, ‘Recursive mode matching method for multiple waveguide junction modeling’, IEEE Transactions on Microwave Theory and Techniques, Vol. 44, No. 1, pp. 87–92, January 1996. 26. H. Patzelt and F. Arndt, ‘Doubleplane steps in rectangular waveguides and their application for transformers, irises and filters’, IEEE Transactions on Microwave Theory and Techniques, Vol. 30, No. 5, pp. 771–776, May 1982.
290
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
27. P. Foster and S.M. Tun, ‘Modelling waveguide components using modematching techniques’, Microwave Engineering Europe, pp. 23–25, April 1996. 28. R.R. Mansour and R.H. Macphie, ‘An improved transmission matrix formulation of cascaded discontinuities and its applications to Eplane circuits’, IEEE Transactions on Microwave Theory and Techniques, Vol. 34, No. 12, pp. 1490–1498, December 1986. 29. A. Moumen and L.P. Ligthart, ‘Radiation pattern and aperture admittance of dielectric filled waveguides with rectangular cross section using the mode matching technique’, Proceedings of the 28th European Microwave Conference, Amsterdam, pp. 587–592, 1998. 30. H.J. Visser, ‘Waveguide phased array analysis and synthesis’, Proceedings of RADAR 94, Paris, pp. 707–712, 1994. 31. K.K. Chan and R.M. Turner, ‘Modal design of broadband widescan waveguide phased array’, Proceedings of RADAR 94, Paris, pp. 25–30, 1994. 32. K.K. Chan, H.J. Visser, R.M. Turner and W.P.M.N. Keizer, ‘Dual polarized waveguide phased array for SAR applications’, Proceedings of PIERS 94, The Netherlands, 1994. 33. H.J. van Schaik, ‘The performance of an irisloaded planar phased array antenna of rectangular waveguides with an external dielectric sheet’, IEEE Transactions on Antennas and Propagation, Vol. AP26, No. 3, pp. 413–419, May 1978. 34. R.W.P. King, Transmission Line Theory, McGrawHill, New York, 1955. 35. T.C. Collocot and A.B. Dobson (eds.), Dictionary of Science and Technology, revised edition, Chambers, Edinburgh, UK, 1982. 36. R.F. Harrington, TimeHarmonic Electromagnetic Fields, John Wiley and Sons, New York, 2001. 37. H.J. Visser, Array and Phased Array Antenna Basics, John Wiley & Sons, Chichester, 2005. 38. G.V. Eleftheriades, A.S. Omar, L.P.B. Katehi and G.M. Rebeiz, ‘Some important properties of waveguide junction generalized scattering matrices in the context of the mode matching techniques’, IEEE Transactions on Microwave Theory and Techniques, Vol. 42, No. 10, pp. 1896–1903, October 1994. 39. A.S. Omar and K. Schünemann, ‘Transmission matrix representation of finline discontinuities’, IEEE Transactions on Microwave Theory and Techniques, Vol. MTT33, No. 9, pp. 765–770, September 1985. 40. F. Alessandri, G. Bartolucci and R. Sorrentino, ‘Admittance matrix formulation of waveguide discontinuity problems: Computeraided design of branch guide directional couplers’, IEEE Transactions on Microwave Theory and Techniques, Vol. 36, No. 2, pp. 394–403, February 1988.
REFERENCES
291
41. G.H. Golub and C.F. van Loan, Matrix Computations, John Hopkins University Press, Baltimore, 1989. 42. T. Itoh (ed.), Numerical Techniques for Microwave and MillimeterWave Passive Structures, John Wiley & Sons, New York, 1989. 43. M. Leroy, ‘On the convergence of the solution of numerical results in modal analysis’, IEEE Transactions on Antennas and Propagation, Vol. AP31, pp. 655–659, July 1983. 44. M. Guglielmi, ‘Simple CAD procedure for microwave filters and multiplexers’, IEEE Transactions on Microwave Theory and Techniques, Vol. 42, No. 7, pp. 1347–1352, July 1994. 45. A. Weisshaar, M. Mongiardo and V.K. Tripathi, ‘CADoriented equivalent circuit modeling of step discontinuities in rectangular waveguides’, IEEE Microwave and Guided Wave Letters, Vol. 6, No. 4, pp. 171–173, April 1996. 46. D.M. Pozar, ‘The active element pattern’, IEEE Transactions on Antennas and Propagation, Vol. 42, No. 8, pp. 1176–1178, August 1994. 47. B.L. Diamond, ‘A generalized approach to the analysis of infinite planar array antennas’, Proceedings of the IEEE, Vol. 56, No. 11, pp. 1837–1851, November 1968. 48. K.K. Chan, R.M. Turner and K. Chadwick, ‘Modal analysis of rectangular waveguide phased arrays’, Proceedings of the IEEE Antennas and Propagation International Symposium, pp. 1400–1403, June 1995.
7 Summary and Conclusions
Even in this age of powerful computers and numerical methods, a need still exists for approximate antenna models. The models are needed in a twostage approach to synthesizing antenna designs that also involves the use of fullwave analysis methods, and they have a right to exist in themselves. Approximate antenna models have been developed for five classes of antennas, having specific applications in mind: loops and solenoids for intravascular MR antennas, printed monopoles for integration on printed circuit boards, foldeddipole antennas and arrays thereof for RFID use, microstrip patch antennas for use in rectennas, and infinite planar arrays of openended waveguide radiators.
7.1
FULLWAVE AND APPROXIMATE ANTENNA ANALYSIS
Since the end of World War II, a lot of effort has been put into the development of numerical electromagnetic analysis and, for sure, fullwave numerical solvers have come a long way. Nevertheless, the automated design of integrated antennas based on fullwave analysis is not yet feasible. An automated design would require some form of stochastic optimization, and the time that each fullwave analysis iteration would take is still too long. A twostage approach involving both approximate and fullwave analysis seems more feasible [1]. In this approach, a stochastic optimization is used in combination with an approximate analysis. Then, the outcome of this optimization is used as input in a line search optimization in combination with fullwave modeling. So, approximate antenna models are needed for automated antenna design. But approximate antenna models also have a right to exist in themselves.
Approximate Antenna Analysis for CAD
Hubregt J. Visser
© 2009 John Wiley & Sons, Ltd. ISBN: 9780470512937
294
SUMMARY AND CONCLUSIONS
As a recent benchmarking of commercially distributed fullwave analysis programs has shown [2, 3], an antenna designer needs to be careful to select the right fullwave analysis method for the right problem. Although the strong advice to use at least two different fullwave analysis methods is fully supported by the present author, it is also understood that not many companies or institutions can afford to purchase or lease multiple commercially distributed fullwave analysis programs. In fact, for many small companies it may be difficult even to get access to a single fullwave analysis program. The availability of approximate models or a method to develop approximate models may be a good alternative. Approximate models result in fast calculations and have the additional advantage that they provide insight into the physical phenomenon that is the basis of the problem. They relate all relevant parameters and variables and allow frequency and dimensional scaling. The accuracy may be less than that obtained with a fullwave analysis but can be sufficient for design purposes. The development of an approximate antenna model is only beneficial if a fullwave analysis program is not available; it is even more beneficial in such a situation if the antenna to be designed belongs to a class of antennas, i.e. when similar antenna designs are foreseen in the future but for different frequency ranges, materials or environments. When a fullwave analysis program is available and the antenna to be designed is known to be a oneofakind design, an educated software version of a trialanderror procedure using that program is advised. Both fullwave analysis software and approximate analysis software has to be used by an antenna expert, with carefully considered input of parameter values and interpretation of the analysis results. In this book, we have described approximate models for five classes of antennas: 1. Loops and solenoids in a conducting medium for intravascular use have been analyzed. These antennas were intended to function as receiving antennas in a magnetic resonance imaging system. Since we were interested in the magnetic field strength close to the antenna, we analyzed the antennas using a quasistatic approximation. 2. Printed monopole antennas for integration on a PCB when limited space is available have been analyzed. An approximate model for this class of antennas was derived using separation of an asymmetrically driven dipole antenna into two grounded monopole antennas. One of the monopole antennas consisted of a microstripexcited printed monopole antenna, and the other was formed by the ground plane of the microstrip. The two monopole antennas were analyzed using an equivalentradius dipole antenna with a magnetic covering. To demonstrate the use of a fullwave analysis program, a oneofakind printed UWB antenna was designed. 3. Closely related to PCB antennas are foldeddipole antennas on dielectric slabs. Foldeddipole antennas with additional short circuits in the arms and/or parasitic radiators positioned parallel to the foldeddipole antenna offer the possibility to control the input impedance while maintaining a nearomnidirectional radiation pattern. This makes them very suitable for RFID applications, where the chip impedance in general is anything but 50 and the antenna may be designed to be conjugately matched to the RFID chip. Analytical dipole antenna models and transmission line theory have been applied to model both thinwire and strip foldeddipole antenna structures and linear arrays of reentrant foldeddipole elements.
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
295
4. The concept of designing antennas to be conjugately matched to the RF front end can also be applied to the design of compact, efficient, lowpower rectennas, i.e. rectifying antennas. First, the rectifying circuit was modeled using a largesignal equivalent circuit. Next, a modified transmission line model for a rectangular microstrip patch antenna was applied to find the edge feed position that corresponded to an antenna input impedance that was the complex conjugate of the impedance of the rectifying circuit. The concept of conjugate matching, in combination with an even–odd mode analysis, was applied to the analysis of a Wilkinson power combiner where the resistive element was replaced by a rectifying circuit. Thus a system of two microstrip antennas was designed, where the combined RF data input was brought to the output of the power combiner and the mismatch between the two antennas was converted to DC energy. 5. The final class of antennas for which an approximate model has been developed is a little different, in the sense that the radiators were modeled in a fullwave manner and the approximation was in the size of an array consisting of these radiators. A large planar array, consisting of radiators placed in a regular grid, was approximated as being infinite in extent in two directions in the transverse plane. Furthermore, the elements were assumed to be excited uniformly. By making this approximation, the array could now be considered to be periodic and analysis could be restricted to a single unit cell, the analysis of which included all of the information about the mutual coupling with the infinite environment.
7.2
INTRAVASCULAR MR ANTENNAS: LOOPS AND SOLENOIDS
The forming of magnetic resonance images is accomplished by trading off signaltonoise ratio, imaging speed and spatial resolution. By replacing the receiver coils outside the body in a ‘standard’ MRI setup by intravascular coils or antennas, the SNR can be improved considerably, making the visualization of catheter positions and orientations during surgery feasible. Even imaging of the condition of the walls of the large arteries should become feasible. The idea of employing intravascular antennas is not new, and some qualitative comparisons of different intravascularantenna concepts have even been performed. To perform the first quantitative comparison of various antenna concepts for intravascular catheter tracking and artery wall imaging, an approximate model for intravascular wire antennas has been developed. This quasistatic, approximate model calculates the magnetic field intensity induced by the wire structure. The dimensions of an intravascular antenna and the frequency1 are such that the artery walls will be positioned in the radiating near field of the antenna, where the field amplitudes locally are inversely proportional to the square of the distance, justifying a quasistatic approximation. To assess the validity of the quasistatic approximation, the results were compared with analytical expressions for the fields of a smallloop antenna carrying a uniform current. To be 1 The frequency is the Larmor frequency which, for a static magnetic field of 1.5 T, is about 64 MHz.
296
SUMMARY AND CONCLUSIONS
able to perform this comparison, the smallloop, uniformcurrent approximation was first validated. It turns out that we may consider a bare loop antenna immersed in blood and subject to a 1.5 T main magnetic field to carry a uniform current for radii up to 1.7 mm. Next, the quasistatic model was verified by looking at the dynamic ‘sensitivity’, which was defined as the transverse magnetic field intensity normalized to the uniform current density. The maximum relative error in the dynamic sensitivity for a single bare loop of radius 0.5 mm immersed in blood was 30% on the axis of the loop and 13% for all other loop orientations, calculated in the region of interest. This region of interest was a horizontal section through the center of the antenna and a coaxial, vertical circular cylinder with a radius between 2 mm and 3 mm, corresponding to the size of the large arteries. Moreover, the sensitivity as a function of distance from the loop center showed a similar appearance for the quasistatic and for the dynamic model. This means that we may employ our quasistatic model for comparison of loop antenna designs. For a multiturn loop antenna immersed in blood, up to 35 turns may be employed for a radius of the order of 0.5 mm without compromising the model. The uniform current is maintained, provided that a thin insulation layer is applied to the wire. If this insulation layer is not applied, the quasistatic approximation will fail for wire antennas larger than a single loop. After we had established the validity of the quasistatic approximate model, use of the model revealed that a center return antenna was best suited for tracking purposes. A perpendicularcoils antenna was preferred, however, owing to easier manufacture. This preferred antenna also performed better than the first one when rotated with respect to the main magnetic field. For imaging purposes, both a dualopposedsolenoids antenna and a tripleloop antenna were preferred. These exhibit comparable sensitivity profiles, and the manufacturing complexity of the two antennas is expected to be equal. Neither antenna should be rotated in excess of 45◦ with respect to the main MR magnetic field. It has been demonstrated that intravascularantenna designs may be created automatically, employing geneticalgorithm optimization in combination with approximate antenna analysis. These designs, subject to userdefined electromagnetic and geometrical constraints, can be generated within minutes, employing standard office computing equipment. Since the designs thus created are very sensitive to the threedimensional geometry, precise manufacturing techniques are required for realizing these designs. Last but not least, there is the issue of patient safety. In the modeling and design, the focus was on the intravascular antennas. The signals received by the antennas need to be transported to the MR hardware, however, and therefore transmission lines are employed. These transmission lines go from the MR hardware, which is exterior to the patient, through the patient’s vascular system to the intravascular antennas. A real danger exists that the lengths of these transmission lines may be such that they will be resonant at the Larmor frequency. Radiation will then take place at the tips of the transmission line leads and will be dissipated in the tissue surrounding the lines. The temperature may thus increase up to and beyond a level that will be harmful for the patient. Dissecting the transmission line into sections that are too short to become resonant at the Larmor frequency is recommended as a solution to this problem. A technique involving inductive coupling between transmission line
PCB ANTENNAS: PRINTED MONOPOLES
297
sections as described in [4] could be employed for transferring signals from the antenna to the MR hardware.
7.3
PCB ANTENNAS: PRINTED MONOPOLES
As explained before, approximate antenna models, if not already available, should be developed only for classes of antennas. If the design of a certain type of antenna is not to be restricted to a single application but, instead, designs of similar antennas, maybe for different frequency bands, materials or environments, are foreseen for the (near) future, it is worthwhile to invest in the development of an approximate model. If a oneofakind antenna needs to be designed, or time is critical, an educated trialanderror design methodology may be followed. The iterative realization and redesign of the antenna may be replaced by the use of a commercially distributed fullwave electromagnetic analysis program. The educated part of the trialanderror process consists of choosing a starting antenna configuration, the operation of which can be understood on basis of physical reasoning. This process has been demonstrated for a printed UWB antenna, the operation of which may be explained by seeing the antenna as a combination of a ‘fat’ dipole antenna and two tapered slot antennas. The antenna was upgraded, following the same process, with a slot in one of the radiating arms to block out WLAN frequencies between 5 GHz and 6 GHz. When the original UWB antenna was upgraded for the same reason with a filtering structure in the microstrip transmission line exciting the antenna, it was found that applying an approximate model for the filtering structure sped up the design process. Next, for the design of nonUWB, planar printed monopole antennas, an approximate model was developed. Printed monopole antennas may be used for various applications, ranging from GSM through GPS, Bluetooth and ZigBee to WLANs, covering frequency bands from around 1 GHz to around 6 GHz. Printed monopole antennas may be incorporated on the rim of FR4based PCB designs and are especially attractive when not much space is available. The approximate model was based on a model of an equivalentradius, magnetically coated, circularly cylindrical, wire dipole antenna. This model was applied both to a microstripexcited monopole antenna and to a monopole antenna that was formed by the ground plane of a microstrip transmission line. If the width of the ground plane is small enough to allow a thinequivalentwire approximation and to prevent halfwave resonance effects from the rim of the ground plane, the model may be employed to create (pre)designs of planar printed monopole antennas.
7.4
RFID ANTENNAS: FOLDED DIPOLES
The foldeddipole antenna is an attractive antenna for RFID applications. First, the radiation characteristics of a foldeddipole antenna are identical to those of an ordinary dipole antenna, i.e. the radiation is nearomnidirectional. Then, the geometry of the foldeddipole antenna offers opportunities to tune the input impedance to a desired complex value. This makes it possible – through complexconjugate matching – to connect the antenna directly to the RFID chip, without having a matching network in between. The antenna may be modeled
298
SUMMARY AND CONCLUSIONS
on the basis of a dipole antenna model and a twowire transmission line model. The tuning can be achieved by placing additional short circuits in the arms of the foldeddipole antenna, which affect mostly the transmission line characteristics of the antenna and not the dipole characteristics, or by placing parasitic dipoles next to the antenna, which affect mostly the dipole characteristics of the antenna and not the transmission line characteristics, or by a combination of both techniques. These impedancetuning techniques have been verified when applied to thinwire foldeddipole antennas. A wire transmission line model for the foldeddipole antenna was applied to these antennas and modified to take account of the tuning mechanisms. For the more realistic application of a strip foldeddipole antenna on a dielectric slab, a model of an asymmetric strip foldeddipole antenna for an antenna in a uniform medium was modified to take account of a finitethickness dielectric slab. To design linear arrays of foldeddipole antennas, which give additional degrees of freedom for obtaining a desired input impedance, an approximate but accurate (wire) model was developed. This model combines closedform analytical equations for a foldeddipole antenna, a reentrant foldeddipole antenna, a twowire transmission line, the mutual coupling between two foldeddipole antennas and the mutual coupling between two thinwire dipole antennas. The model successfully applies separate analysis of the feeding network and of the antennas. A relative error of less than one percent in both the real and the imaginary part of the input impedance was demonstrated over a frequency band of 10%.
7.5
RECTENNAS: MICROSTRIP PATCH ANTENNAS
Wireless batteries, or rectennas – rectifying antennas – are intended for converting wireless RF power into DC power. Although power conversion efficiencies exceeding 80% have been reported for high (20 dBm) input power levels to the rectenna, wireless batteries will be most beneficial at large distances from sources that radiate at power levels limited by national and international regulations. Therefore, the challenge lies in maximizing the power conversion efficiency of wireless batteries for low input power levels, i.e. 0 dBm and below. If a rectifying circuit is conjugately matched directly, to a microstrip patch antenna, the need for a matching network between the two no longer exists. Thus the efficiency of the wireless battery will improve. Moreover, this matching technique automatically suppresses the reradiation of harmonics by the microstrip patch antenna, since the harmonics will be mismatched. Thus, the impedancematching and filtering network encountered in traditional wirelessbattery designs has become obsolete. With the aid of analytical models developed for the antenna and the rectifier, singlelayer, internally matched, filtered PCB rectennas have been designed for low input power levels. A largesignal model for the rectifying circuit was employed; this model is analyzed in the time domain and gives, after fast Fourier transformation, the input impedance in the frequency domain. A rectangular microstrip patch antenna was modeled using a modified, multimodal cavity model. After the complex input impedance of the rectifying circuit was determined, the edge feed position on the microstrip patch antenna was determined such that it would give an input impedance of the patch that was the complex conjugate of the impedance of the rectifying circuit. An efficiency of 52% for 0 dBm input power was achieved at 2.45 GHz for a wireless battery constructed on FR4,
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
299
showing an improvement of more than 10% over a traditional rectenna design, in addition to a reduction in size and complexity. A series connection of these wireless batteries was shown to be able to power a standard household wall clock over a distance of a few meters. If the resistor in a Wilkinson power combiner connected to two microstrip patch antennas is replaced by a rectifying circuit, the simultaneous reception of data and power becomes feasible, a feature not realizable with a standard rectenna. On the basis of the complex input impedance of the rectifying circuit, a modified Wilkinson power combiner has been designed, based on an even–oddmode analysis. The calculated unloaded output voltage as a function of frequency remained within 10% of the measured values. A setup for the simultaneous reception of amplitudemodulated data and power was realized. For power generation, an inequality between the two antennas was created by insertion of an additional section of microstrip transmission line between one antenna and the input port of the modified Wilkinson power combiner. The use of ambient RF power generated by GSM base stations is feasible only if relatively large collecting apertures are employed for the rectenna. For distances ranging from 25 m to 100 m from such a station, power density levels ranging from 0.1 mW m−2 to 1.0 mW m−2 may be expected at single frequencies. For the total GSM downlink frequency bands, these levels may be elevated by a factor of between one and three, depending on the traffic density. Initial measurements in a WLAN environment indicate power density levels that are at least one order of magnitude lower. A single GSM telephone has been demonstrated to deliver enough energy to wirelessly power small applications at distances of a few decimeters. 7.6
LARGE ARRAY ANTENNAS: OPENENDED RECTANGULARWAVEGUIDE RADIATORS
The by now classical modematching analysis method for rectangularwaveguide structures and infinite arrays of openended rectangularwaveguide radiators has been described in this book. Although, since the development of this method about 15 years ago, more efficient analysis methods have been derived, the classical modematching approach may still be of use, especially for analyzing structures that are not overcomplicated. Therefore a thorough description of the method that can be straightforwardly implemented into software may be of value to engineers in the field. As said, the analysis of waveguides and waveguide discontinuities in this method is based on mode matching. For every discontinuity, waveguidetowaveguide and waveguidetounitcell, a generalized scattering matrix is formed. Generalized scattering matrices of discontinuities are cascaded to create an overall generalized scattering matrix of the complete structure to be analyzed. The code implemented has been thoroughly validated by comparing analysis results with analysis and measurement results reported in the open literature. A mode preselection scheme, a convergence strategy and a Floquet mode selection scheme have been developed and have been applied to the structures analyzed. REFERENCES 1. A.G. Tijhuis, M.C. van Beurden, B.P. de Hon and H.J. Visser, ‘From engineering electromagnetics to electromagnetic engineering: Using computational electromagnetics
300
SUMMARY AND CONCLUSIONS
for synthesis problems’, Turkish Journal of Electrical Engineering and Computer Sciences, Vol. 16, No. 1, pp. 7–19, 2008. 2. A. Vasylchenko, Y. Schols, W. De Raedt and G.A.E. Vandenbosch, ‘A benchmarking of six software packages for fullwave analysis of microstrip antennas’, Proceedings of the 2nd European Conference on Antennas and Propagation, EuCAP2007, Edinburgh, UK, November 2007. 3. A. Vasylchenko, Y. Schols, W. De Raedt and G.A.E. Vandenbosch, ‘Challenges in full wave electromagnetic simulation of very compact planar antennas’, Proceedings of the 2nd European Conference on Antennas and Propagation, EuCAP2007, Edinburgh, UK, November 2007. 4. P. Vernickel, V. Schulz, S. Weiss and B. Gleich, ‘A safe transmission line for MRI’, IEEE Transactions on Biomedical Engineering, Vol. 52, No. 6, pp. 1094–1102, June 2005.
Index
AC, 199 acoustic transducer, 20 aerial, 3 ambient RF energy, 211 ambient RF power, 299 Amitay, N., 4 Ampère’s law, 41 antenna, 3 antenna mode, 143, 145, 157, 170 antenna rotation, 71, 72 antiparallel wire antenna, 28, 59, 62 approximate analysis, 6, 9, 293, 294 approximate model, xi, 7–9, 12, 13, 19, 22, 31, 89, 294–297 artery, 9, 28, 30, 33, 47, 81, 89, 295 Ascoli, Professor, 3 asymmetric coplanar waveguide, 162 asymmetric coplanarstrip foldeddipole antenna, 142, 155, 158, 161, 165, 167 asymmetric coplanarstrip transmission line, 157, 164 asymmetric foldeddipole antenna, 146 asymmetric strip foldeddipole antenna, 298 asymmetric wire foldeddipole antenna, 145 atom, 22 audion, 4 Approximate Antenna Analysis for CAD
Hubregt J. Visser
© 2009 John Wiley & Sons, Ltd. ISBN: 9780470512937
Balanis, Constantine A., 4 band notch, 109, 111 Bessel function, 34, 192 biconical antenna, 99 Biot and Savart law, 33, 41, 90 Biot and Savart model, 43, 44, 46, 47 birdcage antenna, 65 blood vessel, 19, 21, 31 Boot, Henry, 4 bow tie antenna, 100 British Broadcasting Corporation (BBC), 4 cable currents, 103 cascade, 254 CAT scan, 20 cavity magnetron, 4 cavity model, 10, 187, 192, 193, 298 center return antenna, 59, 62, 72, 75, 76, 90, 296 common mode, 143 complete elliptic function of the first kind, 156 conjugate matching, 185, 216, 295, 297, 298 conjugately matching, 183 COTS, 5, 31, 98, 135, 193, 198 CPS, 155–158, 160–162, 164, 165 CPW, 98, 162–164
302 crosssectional step, 236 CT scan, 20 DC, 9, 183, 185, 199, 202, 204, 206, 216, 295, 298 de Forest, Lee, 4 dephasing, 26, 29 dielectric coating, 122 dielectric step in a unit cell, 248, 251 differential mode, 143 diode, 4, 9, 184, 185, 193, 208 dipole antenna, 2, 3, 97, 105, 118, 139, 141 dipole mode, 139, 142, 150, 165 displacement current, 1 doublehelix antenna, 59, 62 doublehelix wire antenna, 28 doubleloop antenna, 30, 65, 68 dualnotch horn antenna, 100 dualopposedhelix antenna, 62 dualopposedsolenoids antenna, 30, 65, 68, 72, 76, 90, 296 duality, 33 edge feed, 185, 188, 202, 298 elevated wire, 3 Elliott, Robert S., 4 endovascular intervention, 21 equivalent radius, 143, 146, 156, 157, 166, 294, 297 evenmode, 113 FFT, 197 FID, 27 finetuning, 8 finitelength transmission line, 251 FIT, 102, 157 Fleming, Ambrose, 3, 4 Floquet mode, 233, 260, 261, 272, 273, 299 foldeddipole antenna, 139–143, 153, 165, 178, 297 foldeddipole array antenna, 139, 169, 176 Fourier series, 34 FR, 98, 157, 183, 209, 215, 297, 298 free induction decay, 27 Galindo, V., 4 generalized scattering matrix, 242, 244, 245, 248, 251, 253, 299 genetic algorithm (GA), 19, 80, 81 GPS, 131, 135, 297 gradient, 20, 27, 29, 31, 87, 88
INDEX
grating mode, 233 GSM, 131, 135, 183, 211–216, 223, 224, 236, 238, 245, 248, 249, 251, 253–256, 297, 299 half space, 161 Hansen, Robert C., 4 harmonic balance analysis, 195 Harrington, Roger, 4 Hertz, Heinrich Rudolf, 2 Hertzian dipole, 32 higher harmonics, 199, 202 higherorder harmonics, 185 horn, 4 hydrogen, 22 image, 3 imaging, 21, 28, 30, 58, 59, 62, 65, 72, 82, 84, 90, 295, 296 impedance stepup ratio, 146, 175 impedance tuning, 139 in vitro, 31 induction coil, 2 infinite array, 221, 222 input admittance, 35, 125, 127, 129 input impedance, 8–10, 34, 35, 38, 55, 75, 97, 99, 118, 125, 131–133, 139–147, 149–153, 155–158, 161, 165, 167, 170, 171, 174, 178, 185, 187, 188, 192, 193, 198–200, 202, 211, 294, 295, 297–299 input impedance control, 146 interventional MRI (iMRI), 19, 21 intravascular antenna, 9, 21, 22, 27–29, 31, 33, 40, 56, 75, 81, 82, 86, 88–90, 295, 296 intravascular intervention, 89 inverted L antenna, 3 King, Ronold W. P., 4 Kraus, John, 4 Larmor frequency, 24, 26, 27, 29, 31, 90, 295, 296 line search optimization, 293 logperiodic foldeddipole array antenna, 169 Lommel–Weber function, 35 loss tangent, 102, 157, 192, 209 magnetic coating, 122, 125, 166 magnetic moment, 22 magnetic resonance imaging, 20, 294 Mailloux, Robert J., 4
303
INDEX
Marconi, Guglielmo, 2 Massachusetts Institute of Technology Radiation Laboratory (MIT Rad Lab), 4 Maxwell, J. C., 1 medical imaging techniques, 20 MEN, 273 method of moments, 4, 133, 146, 176, 193 microstrip patch antenna, 183, 185, 187, 188, 198, 202, 211, 216, 295, 298 microstrip transmission line, 98, 101, 112, 114, 127, 128, 189, 206, 297 mode matching, 221–223, 240, 249, 251, 258, 269–273, 299 modified Bessel function, 34, 35 MoM, 4, 146, 147, 149, 150, 153, 172, 178, 193 monopole antenna, 3, 97, 117, 297 MR antenna, 19, 21, 22, 29, 43, 59, 293 MRI, 9, 19–22, 29–31, 88, 89, 295 multiturn loop antenna, 55, 56, 90, 296 mutual admittance, 175, 176, 189, 191 mutual coupling, 11, 55, 56, 58, 153, 171, 172, 174, 176, 178, 221, 222, 295, 298 mutual impedance, 152, 153, 175, 176 neutron, 22 NMI, 20 NMRI, 22 nuclear medicine imaging (NMI), 20 nucleus, 22, 24 numerical electromagnetics code (NEC), 5 numerical stability, 255 oddmode, 113 OFDM, 169 openended waveguide, 222 openended waveguide array antenna, 223 opposeddoublehelix antenna, 29, 59 opposeddoublehelix wire antenna, 28 opposedsolenoids antenna, 76 parasitic dipole, 139, 142, 146, 153 PCB, 97, 98, 100, 102–104, 117, 129, 130, 142, 155, 183, 185, 294, 297, 298 PEC, 172 permeability, 22, 31, 91, 118, 121, 141, 225, 248, 249 permittivity, 31, 57, 91, 98, 102, 113, 118, 121, 128, 129, 141, 156–158, 160–163, 188–190, 209, 225, 248, 249
perpendicularcoils antenna, 29, 59, 62, 72, 77, 90, 296 PET, 20 phased array antenna, 221 plane wave, 233 Pocklington, H. C., 3 power density, 213, 214, 216 power series, 36, 37 power waves, 149, 157 precession, 22 preliminary design, 6 printed monopole antenna, 98 printed UWB antenna, 98 proton, 22, 26, 27, 29, 31 pseudo monopole, 101, 103–105, 107, 108, 117 quasistatic, 41, 56, 90, 91, 112, 116, 117, 295, 296 approximation, 294 magnetic field, 33 model, 296 radiating slots, 189 radiation pattern, 97, 99, 103–105, 107, 118, 139, 265, 267, 294 radio system, 2 radioactive source, 20 radiology, 20 Randall, John, 4 Rautio, James, 5 RC, 256, 258–261, 264, 272 reentrant foldeddipole antenna, 169, 170, 178, 298 receiver, 2 receiver coil, 20, 21, 24, 27, 89, 295 rectangular waveguide, 223, 224, 299 rectangularwaveguidetounitcell junction, 245, 248 rectenna, 9, 183–185, 200, 203, 298 rectifying circuit, 10, 183, 185, 187, 193, 198, 202, 203, 206, 208, 209, 211, 216, 295, 298, 299 reflection coefficient, 9, 132, 142, 147, 149, 193, 199, 221, 256, 259, 263–265, 268, 270, 271 reflection level, 8, 101 reflector antenna, 4 relative convergence, 256, 258 Runge–Kutta method, 197, 200
304
saddle coil antenna, 30, 65, 68 safety, 86, 296 scan element pattern, 267 scan reflection coefficient, 267, 270 Schelkunoff, S. A., 4 Schottky diode, 184, 197, 199, 216 self coupling, 55 sensitivity pattern, 40, 59, 62, 65, 72 separation of variables, 227, 232, 234 short circuit, 141, 171 signaltonoise ratio, 295 Silver, Samuel, 4 singleloop antenna, 22, 43, 55, 65 smallloop antenna, 33, 36, 38, 39, 43, 55, 89, 295 smallloop approximation, 33, 35, 36, 89 space harmonic, 233 spark gap, 2 specific absorption rate, 88 SPECT, 20 spherical dipole antenna, 99 spin, 22 spin–lattice relaxation time, 26 spin–spin interaction, 26 spin–spin relaxation time, 27 splitT combiner, 205 spurline, 112, 113, 116 static magnetic field, 20, 22, 24, 31, 87, 89 stepup impedance ratio, 155–157 stochastic optimization, 6, 8, 9, 12, 293 strip foldeddipole antenna, 298 strip monopole antenna, 118, 125 stub, 140, 141, 143, 145, 146, 150, 175, 185 synthesis, 5–7, 22, 178 thermionic valve, 4 thinwire dipole antenna, 99 three dualopposedsolenoids antenna, 29
INDEX
threeterm model, 118, 166 Tizard, Sir Henry, 4 tracer, 20 tracking, 21, 28, 58, 59, 62, 72, 75, 77, 82–84, 90, 296 transmission line mode, 139, 142, 143, 145, 150, 157, 160, 166, 167 transmission line model, 172, 187, 188, 192, 193, 198, 295, 298 transmission matrix, 254 transverse electric (TE) modes, 226 transverse magnetic (TM) modes, 226 triangular lattice, 231 triode, 4 tripleloop antenna, 30, 65, 68, 72, 75, 76, 90, 296 twopenny dipole antenna, 100 twowire transmission line, 141, 143, 144, 178, 298 ultrasound, 20 ultrawideband (UWB), 99 antenna, 294, 297 uniformcurrent approach, 55, 56 uniformcurrent approximation, 36, 39, 40, 89, 296 unit cell, 11, 221, 222, 233, 295 vector network analyzer, 199 voltage doubler, 202, 211 waveguide to freespace unit cell junction, 223 wideangle impedance match, 224 Wilkinson power combiner, 183, 205, 206, 208, 209, 295, 299 wireless battery, 185, 200, 202, 204, 298 wireless telegraphy, 2 Wu, C. P., 4 Xray, 20