Generalized Filter

Design by Computer
Optimization

Djuradj Budimir
Generalized Filter Design
by Computer Optimization
For a complete listing of the Artech House Microwave Library,
turn to the back of this book.
Generalized Filter Design
by Computer Optimization

Djuradj Budimir

Artech House
Boston ∙ London
Library of Congress Cataloging-in- Publication Data
Budimir, Djuradj
Generalized filter design by computer optimization ⁄ Djuradj Budimir
p. cm. — (Artech House microwave library)
Includes bibliographical references and index.
ISBN 0-89006-579-9 (alk. paper)
1. Electric filters— Design and construction — Data processing.
2. Computer-aided design. I. Title. II. Series.
TK7872.F5B82 1998
621.3815’324’028551—dc21 97-39465
CIP

Generalized filter design by computer optimization

(Artech House microwave library)
1 . Electric fιlters-Design and construction-Data processing
I. Title
621.3’815324’0285

ISBN 0-89006-579-9

Cover design by Jennifer L. Stuart

© 1998 ARTECH HOUSE, INC.

685 Canton Street
Norwood, MA 02062

All rights reserved. Printed and bound in the United States of America. No part of this
book may be reproduced or utilized in any form or by any means, electronic or me-
chanical, including photocopying, recording, or by any information storage and re-
trieval system, without permission in writing from the author
All terms mentioned in this book that are known to be trademarks or service marks
have been appropriately capitalized. Artech House cannot attest to the accuracy of this
information. Use of a term in this book should not be regarded as affecting the validity
of any trademark or service mark.

International Standard Book Number: 0-89006-579-9

Library of Congress Catalog Card Number: 97-39465

109 8 76 5 4 3 2 1
 To
Aleksandar and Slobodan
Contents

Preface xiii

1 Introduction to Computer-Aided Filter

Design 1
1.1 Electromagnetic Simulation 3
1.2 Computer-Aided Analysis 4
1.3 Computer Optimization 5
1.4 Outline of This Book 7
References 8
Selected Bibliography 9

2 Transmission Lines, Lumped Elements,

and Resonators 11
2.1 Transmission Lines 11
2.1.1 Coaxial Line 12
2.1.2 Rectangular Waveguide 13
2.1.3 Circular Waveguide 17
2.1.4 Ridged Waveguide 17
2.1.5 Rectangular Dielectric-Loaded Waveguide 19
2.1.6 E-Plane Circuit (Finline) 19
2.1.7 Dielectric Waveguide 19

vii
viii Generalized Filter Design by Computer Optimization

2.1.8 Metal-Pipe Waveguide 20

2.1.9 Stripline 20
2.1.10 Microstrip Line 21
2.1.11 Coplanar Line 21
2.1.12 Multilayer Coplanar Waveguide 22
2.1.13 Microshield Line 22
2.2 Comparison of Various Transmission
Lines 23
2.3 Lumped Elements 23
2.3.1 Inductors 24
2.3.2 Capacitors 25
2.4 Resonators 25
2.4.1 Dielectric Resonator 26
2.4.2 Dielectric Material 29
References 32
Selected Bibliography 37

3 Characterization of Discontinuities 39
3.1 Analysis of Filter Discontinuities by the
Mode-Matching Method 42
3.1.1 Discontinuities in Conventional E-Plane
Filters 42
3.1.2 Discontinuities in Ridged Waveguide
Filters 48
3.2 Convergence 62
3.3 Coplanar Waveguide Discontinuities 68
3.3.1 Introduction to Coplanar Waveguide
Modeling 68
TM TM
3.3.2 Introduction to Xgeom and Em
Software 70
References 73
Selected Bibliography 74
Appendix 3A 75
 Contents ix

Appendix 3B: Scattering Parameters for

Bifurcated Waveguides 78
Appendix References 80

4 Optimization-Based Filter Design 81

4.1 Filter Circuit Parameters 81
4.2 Filter Design 83
4.3 Filter Optimization 85
4.4 Description of the Algorithm 91
4.4.1 Symmetrical Case 91
4.4.2 Asymmetrical Case 97
References 101
Selected Bibliography 102

5 Design of Lumped-Element Filters by

Optimization 103
5.1 An Approximate Synthesis-Based Design
Procedure 103
5.1.1 Element Values of Chebyshev Lowpass
Prototype Filters 104
5.1.2 Filter Design Procedure 108
5.2 Numerical Implementation of Equal-
Ripple Optimization 120
5.3 Numerical Results 121
References 122
Selected Bibliography 123

6 Design of E-Plane Filters by Optimization 125

6.1 An Approximate Synthesis-Based
Procedure for the Design of Direct-
Coupled Cavity Filters 128
6.2 A Cad Algorithm Using the Passband
Correction Method 134
X Generalized Filter Design by Computer Optimization

6.2.1 Design Procedure Using Passband

Correction Method 137
6.3 Numerical Implementation of Equal-
Ripple Optimization 138
6.3.1 Calculation of Insertion Loss 138
6.3.2 Solution of Nonlinear Equations 139
6.4 Numerical and Experimental Results 141
References 145
Selected Bibliography 145

7 Design of Ridged Waveguide Filters by

Optimization 147
7.1 Proposed Filter Configuration 148
7.2 Circuit Representation and Design
Procedure 149
7.2.1 Circuit Representation 149
7.2.2 Design Procedure 151
7.3 Numerical Implementation of Equal-
Ripple Optimization 155
7.4 Numerical and Experimental Results 158
References 166
Selected Bibliography 167
Appendix 7A: Derivation of the Impedance
Inverter Equations 168

8 Design of Coplanar Waveguide Filters

by Optimization 171
8.1 An Approximate Synthesis-Based Design
Procedure 172
8.1.1 Circuit Representation 172
8.1.2 Design Procedure 173
8.2 Numerical Implementation of Equal-
Ripple Optimization 177
8.3 Numerical and Experimental Results 179
 Contents xi

References 180
Selected Bibliography 181

9 CAD Programs 183

9.1 The LCFILTER Program 183
9.2 The EPFILTER Program 195
Selected Bibliography 198

Appendix A Parameters 199

Appendix B 215

List of Principal Symbols and

Abbreviations 217

About the Author 223

Index 225
Preface

Recent advances in microwave computer-aided filter design technology suggest

the feasibility of interfacing electromagnetic simulations directly to sophisticated
optimization systems. With the availability of powerful computers and RISC
workstations, this optimization-based approach to the design of microwave
filters becomes a desirable tool.
The central theme of this book is the optimization-oriented approach
for the accurate design of radio frequency, microwave, and millimeter-wave
filters. It presents computer-aided filter design algorithms and provides examples
of their applications. This book is divided into nine chapters. Chapter 1 gives
an introduction. Chapter 2 considers transmission lines, lumped elements, and
resonators. Chapter 3 describes the characterization of discontinuities.
Chapter 4 presents an optimization method used in filter design. Chapters 5
through 8 concentrate on the design of lumped element, E-plane metal insert,
ridged waveguide, and coplanar waveguide filters. The final chapter describes
computer-aided filter design programs.
This book is intended for advanced undergraduate and graduate courses,
for design engineers, and for research and development specialists who are
interested in computer-aided filter design of radio frequency, microwave, and
millimeter-wave filters. The material represents state-of-the-art technology at
the time of writing. The reader is assumed to have completed courses at the
basic level in filter theory, electromagnetics, and microwaves.
The material in this book is based on my own work at the University
of Leeds and the University of London over the last seven years. I wish to give
particular thanks to Professor J. David Rhodes of Filtronic Components and
the Department of Electronic and Electrical Engineering, University of Leeds,
who provided expert advice throughout the course of this work. Further thanks

xiii
xiv Generalized Filter Design by Computer Optimization

are due to Dr. Vasil Postoyalko, Dr. John R. Richardson and Dr. Stavros
Iezekiel for their encouragement and during part of this work, Professor Roger
Pollard for useful technical discussions and assisting in the microwave measure-
ment, and Mr. Terry Moseley for expert machining of the waveguide housing,
all of whom are at the Department of Electronic and Electrical Engineering,
University of Leeds. Furthermore, I would like to acknowledge Professor Alec
Cullen, Department of Electronic and Electrical Engineering, University Col-
lege, University of London, for useful technical suggestions. Thanks are also
due to Dr. Ian Robertson for permission to undertake part of this work in the
Department of Electronic and Electrical Engineering, King’s College, University
of London.
1
Introduction to Computer-Aided Filter
Design

Computer-aided design (CAD) of radio frequency (RF), microwave, and

millimeter-wave (mm-wave) filter structures must be capable of handling the
restrictions of a widespread application of low-cost precision fabrication meth-
ods, such as computer-controlled milling, spark eroding, or photolithographic
etching techniques, in which postassembly tuning is no longer economical or
feasible. The filter design also must meet the demands of the expanding utiliza-
tion of higher frequency bands (up to mm-waves), which need tighter tolerances.
These conditions require CAD, rigorous electromagnetic simulation techniques,
and efficient computer optimization methods that allow the computer-aided
filter design to take into account all the significant design parameters.
CAD strictly interpreted can be taken to mean any design process where
the computer is used as a tool. However, usually the term CAD implies that
without the computer as a tool, that particular design process would have been
impossible or much more difficult, more expensive, and more time consuming.
A typical flow diagram for circuit CAD is shown in Figure 1.1. To implement
a conventional design procedure to meet the given filter specifications success-
fully, a CAD approach becomes necessary. When the CAD approach is used,
experimental modifications of the filter (which is unavoidable in the conven-
tional design procedure) are replaced by a computer-based optimization of the
initial design.
A typical design process usually begins with a given set of specifications
or design goals for the filter. Synthesis methods and available design data help
to arrive at the initial filter design. The performance of the initial filter design
is evaluated by computer-aided filter analysis. Numerical methods needed for

1
2 Generalized Filter Design by Computer Optimization

FILTER SPECIFICATIONS

ARE FILTER PARAMETERS KNOWN ?

NO
YES

SYNTHESIS

E M SIMULATOR

ANALYSIS OPTIMIZER

NO
ARE SPECS MET ?

YES

FABRICATION

Figure 1.1 Flow chart for a CAD procedure.

 Introduction to Computer-Aided Filter Design 3

the analysis of filter structures are called from the library of subroutines devel-
oped for that purpose. Filter characteristics obtained as a result of the analysis
are compared with the given specifications. If the results fail to satisfy the
desired specifications, the designable parameters of the filter are altered in a
systematic manner. The optimization is especially important for mm-wave
filters because they cannot easily be tuned or trimmed after fabrication. The
sequence of filter analysis, comparison with the desired performance, and
parameter modification is performed iteratively until the specifications are met
or the optimum performance of the filter is achieved. If the specifications are
not met, the design may have to be revised. When the design satisfies the
specifications, a filter can be built and tested. If the measured results meet the
specifications, the design process is completed; otherwise, the design needs to
be repeated. The filter then is fabricated and experimental measurements are
carried out. Some modifications may still be required if the modeling has not
been accurate enough. Ideally, the modifications will be very small, and the
aim of the CAD method is to minimize the experimental iterations as far as
practicable. Thus, CAD can greatly decrease the time and the cost of design
while enhancing its quality.
The process of CAD, as outlined in the preceding paragraphs, consists
of three important segments:

• Electromagnetic simulation;
• Computer-aided analysis;
• Computer optimization.

1.1 Electromagnetic Simulation

Electromagnetic simulation involves the characterization of various filter discon-

tinuities, to the extent of providing a numerical model that can be handled by
the computer. An accurate and reliable characterization of filter discontinuities is
one of the basic prerequisites of successful CAD. The degree of accuracy to
which the performance of filters can be predicted depends on the accuracy of
the characterization and simulating of those discontinuities. S parameters are
convenient for characterization of filter structures. Detailed numerical simulat-
ing of filter structures becomes involved and time consuming. Characterization
and simulating of various types of waveguide (rectangular and ridged) disconti-
nuities such as step discontinuity rectangular waveguide to metal septum in
rectangular waveguide and step discontinuity metal septum in rectangular
waveguide to ridged waveguide by the mode-matching method are discussed
in Chapter 3. The effects of those discontinuities become more and more
4 Generalized Filter Design by Computer Optimization

significant as one moves from the microwave frequency range to mm-waves

with the need for lower tolerances. Such effects require rigorous electromagnetic
simulation methods that allow the accurate computer-aided component design
to take into account all significant factors, such as finite thickness of septa.
Much research aimed at characterization for CAD purposes has been reported
for waveguide discontinuities [ l - 5 ] . Sonnet software’s Em™ package uses the
method of moments used for electromagnetic simulation of various types of
coplanar waveguide discontinuities, as discussed in Chapter 3 [6]. Table 1.1
summarizes some commercially available electromagnetic simulators for
computer-aided filter design purposes [6-1 5] .

1 .2 Computer-Aided Analysis
Computer-aided analysis constitutes the key step in the CAD procedure.
Because the analysis forms a part of the optimization loop, the analysis subpro-

Table 1.1
Some Commercially Available Electromagnetic Simulators

Company Product* Type

HP-EEsof (HP range) Momentum 3D planar electromagnetic

HFSS 3D arbitrary
electromagnetic
Compact Software Microwave Explorer 3D planar electromagnetic
Sonnet Software Em 3D planar electromagnetic
Xgeom Layout entry
Emvu Current display
Jansen Microwave Unisym/Sfpmic 3D planar electromagnetic
Ansoft Corporation Maxwell Strata 3D planar electromagnetic
Maxwell Eminence 3D arbitrary
MicroWaveLab 3D arbitrary
electromagnetic
ArguMens Stingray 3D planar electromagnetic
Zeland Software IE3D 3D planar electromagnetic
Optimization System Empipe Sonnet, Ansoft, HP
Associates optimization
Empipe3D⁄EmpipeExpress Geometry capture front end
MacNeal-Schwendler Corp. MicroWaveLab 3D arbitrary
electromagnetic
Kimberley Communications Micro-Stripes 3D arbitrary
Consultants electromagnetic (TLM)
Boulder Microwave Tech. Ensemble 3D planar
Computer System Mafia 3D planar
Technologies

*AII trademarks acknowledged.

 Introduction to Computer-Aided Filter Design 5

gram is executed again and again for a specific filter design. For that reason,
an efficient analysis algorithm constitutes the backbone of any CAD package.
Computer-aided analysis provides the response of a specified filter configuration
to a given set of inputs. Computer-aided analysis is perhaps the most developed
and most widely used aspect of CAD. The filters considered in this book can be
expressed as a cascade combination of two-port sections. The analysis problem in
that case may be stated as follows: Characterize the filter sections (say, in terms
of individual scattering, ABCD matrices, or even or odd mode impedances)
and find the scattering, ABCD matrices, or, by choice, even and odd mode
impedances of the overall filter structure. Computer-aided analysis is the process
of evaluating the filter performance, as follows:

1. The lumped-element lowpass filter is decomposed into building blocks,

such as inductor and capacitor, while the conventional E-plane band-
pass filter structure is decomposed into appropriate key building blocks,
such as homogeneous rectangular waveguide and step discontinuity
rectangular waveguide to metal septum in rectangular waveguide.
2. The ridged waveguide bandpass filter structure is decomposed into
building blocks such as homogeneous rectangular waveguide, step
discontinuity rectangular waveguide to metal septum in rectangular
waveguide, step discontinuity metal septum in rectangular waveguide
to ridged waveguide, and the ridged waveguide.
3. Meanwhile, the coplanar waveguide bandpass filter structure is decom-
posed into building blocks such as coplanar waveguide, edge coupled
coplanar waveguide discontinuity, and end-coupled coplanar wave-
guide discontinuity.
4. Microwave analysis of the conventional E-plane bandpass filter struc-
tures involves evaluation of even and odd mode impedances of the
overall filter structure in terms of the given even and odd mode
impedances of individual sections of the filter, while microwave analysis
of the lumped-element, ridged-waveguide and coplanar-waveguide
filter structures involves evaluation of ABCD parameters of the overall
filter structure in terms of the given ABCD parameters of individual
sections of the filter.

1 .3 Computer Optimization

An important class of problems in the CAD of microwave filters concerns

algorithms for adjusting circuit parameters to minimize the deviations between
the circuit performance achieved at some stage of the design and the desired
6 Generalized Filter Design by Computer Optimization

specifications. As shown in the flow diagram in Figure 1.1, one starts with a
given set of filter specifications and an initial filter design. Filter characteristics
obtained from the analysis are compared with the given specifications. If the
results fail to satisfy the desired specifications, the designable parameters of
the filter are altered in a systematic manner. The sequence of filter analysis,
comparison with the designed performance, and parameter modification is
performed iteratively until the optimum performance of the filter is achieved.
The process is termed optimization. Most optimization techniques employed
in RF, microwave, and mm-wave CAD employ general forms of error minimiza-
tion algorithms [16-25] (Table 1.2). The accuracy of the algorithms is related
to the number of discrete frequency points used to define the error vector.
Cohn introduced an optimization algorithm that requires fewer sampling points
to achieve convergence than generalized error minimization algorithms [26] .
The intention here is to provide the reader with a numerical optimization
method [27], based on Cohn’s, that gives an accurate equal-ripple response
within the passband for filters based on Chebyshev, generalized Chebyshev,
and elliptic function prototypes. It is desirable to have good starting values (filter
dimensions) for this optimization method. Using the conventional synthesis
procedures, starting values of the optimization routine can be obtained. The
advantages of this method are that the problems of local minima are avoided,
it requires fewer sampling points to achieve convergence than generalized error

Table 1.2
Some Commercially Available Filter Optimization Tools

Company Product* Type

HP-EEsof (HP range) MDS Integrated package

Series IV/PC Integrated package
HP-EEsof (EEsof range) Touchstone Linear
Linecalc Physical parameters
E-Syn Filter synthesis
Compact Software SuperCompact Linear
Jansen Microwave Linmic+ Linear (integrated suite)
ArguMens Octopus Linear
Optimization System Hope Linear/nonlinear
Associates
Empipe Sonnet, Ansoft, HP
optimization
Empipe3D⁄EmpipeExpress Geometry capture front end
Optotek MMICAD Linear
Eagleware M/FILTER Linear

*AII trademarks are acknowledged.

 Introduction to Computer-Aided Filter Design 7

minimization algorithms, and the Chebychev criteria are satisfied [28] . This
method can handle symmetrical and asymmetrical lowpass, highpass, and band-
pass Chebyshev, generalized Chebyshev, and elliptic filters [26,27,29,30-34].
To illustrate the application of the method described in this book, the designs
of lumped-element lowpass, E-plane metal-insert bandpass, ridged-waveguide
bandpass, and coplanar waveguide bandpass filters are considered as examples.
When implemented around electromagnetic simulators, the method can be
used to include all the effects of discontinuities, junctions, and so on, to reduce
the amount of tuning required in the final filter.

1.4 Outline of This Book

The need to develop an efficient and accurate method for the design of RF,
microwave, and mm-wave filters has already been discussed. It is the main aim
of this book to make a contribution to this area. Key stages in the filter design
are addressed, namely, electromagnetic simulation of filter discontinuities,
determination of the starting points for the optimization algorithm, and
the development of a numerical optimization algorithm. Toward that end,
Chapter 2 discusses characterization of various types of transmission lines,
lumped elements, and resonator structures. Chapter 3 deals with the electromag-
netic analysis of the discontinuities in E-plane metal-insert filter structures
such as the metal septum in rectangular waveguide and the metal septum in
rectangular waveguide between two ridged waveguides with equal and different
gaps by the rigorous full-wave mode-matching method. In the case of longitudi-
nally symmetrical filter structures such as conventional E-plane metal insert
filters, we need consider only the structures with the electric and magnetic
walls at the plane of symmetry (half-filter structure). Each E-plane septum is
itself symmetrical and can be electrically represented by normalized even and
odd mode impedances. The mode-matching method, where up to 140 modes
may be included, has been used for calculation of the even and odd mode
impedances. Analysis of ridged waveguide by the generalized transverse reso-
nance method is described. The convergence mechanism has been studied to
achieve a reliable solution with minimal numerical computation.
Chapter 4 presents an equal-ripple approach to numerical optimization
of E-plane metal-insert filters. The advantages of the proposed optimization
method over generalized methods available in software packages like EEsof
Touchstone™ [13] and Compact Software SuperCompact™ [35] are discussed.
Chapter 5 presents the design of lumped element filters by computer optimiza-
tion. Chapter 6 deals with a new approach to the design of conventional
E-plane metal-insert bandpass filters by equal-ripple optimization. Determina-
8 Generalized Filter Design by Computer Optimization

tion of the starting point for the optimization algorithm is described. The
analysis of those filters has neglected higher-order mode coupling between E-
plane septa. For the design example considered, that was shown not to be
important. When higher-order mode coupling needs to be taken into account,
the numerically efficient procedure for the cascading of filter sections, which
involves only real scalar arithmetic, no longer can be applied. The optimization
aspect of the presented design method is still applicable. To confirm the
accuracy of the design procedure, a five-resonator X-band conventional E-
plane bandpass filter was fabricated. Specific characteristics of the filter are
discussed and measurement results of given filter responses presented.
Chapter 7 examines the design of ridged-waveguide bandpass filters by
computer optimization. The modified design procedure, which should include
the concept of impedance inverters and impedance scaling of the impedance
levels of the prototype filter for design of ridged-waveguide bandpass filters,
is presented. A five-resonator X-band E-plane ridged waveguide bandpass filter
is designed and fabricated. Chapter 8 describes the design of coplanar waveguide
bandpass filters by computer optimization. Finally, Chapter 9 describes CAD
programs.

References

[1] Itoh, T., Numerical Techniques for Microwave and Millimeter-Wave Passive Structures,
New York: Wiley, 1989.
[2] Collin, R. E., Field Theory of Guided Waves, New York: McGraw-Hill, 1960.
[3] Schwinger, J., and D . Saxon, Discontinuities in Waveguide (Documents on Modem Physics),
New York: Gordon and Greach, 1968.
[4] Wexler, A., “Solution of Waveguide Discontinuities by Modal Analysis,” IEEE Trans.
Microwave Theory & Tech., Vol. MTT-15, September 1967, pp. 508-517.
[5] Mittra, R., and S. W. Lee, Analytical Techniques in the Theory of Guided Waves, New
York: Macmillan, 1971.
[6] Em User’s Manual, Vol. 1, Release 4.0, Sonnet Software Inc., Liverpool, NY, 1996.
[7] HFSS Reference Manual, Release 2.0, Hewlett-Packard Co., Palo Alto, CA, 1992.
[8] Micro-Stripes User’s Manual, Release 2.3, Kimberley Communications Consultants Ltd.,
Nottingham, England, 1994.
[9] Maxwell Strata User’s Manual, Ansoft Corp., Pittsburgh, PA, 1996.
[10] Maxwell Eminence User’s Manual, Ansoft Corp., Pittsburgh, PA, 1996.
[11] MicroWaveLab User’s Manual, Ansoft Corp., Pittsburgh, PA, 1996.
[12] MicroWaveLab User’s Manual, MacNeal-Schwendler Corp., Milwaukee, WI, 1996.
[13] Mafia User’s Manual, Version 3.2, Computer Simulation Technology, Darmstadt,
Germany, 1995.
 Introduction to Computer-Aided Filter Design 9

[14] IE3D User's Manual, Zeland Software Inc., Fremont, CA, 1995.
[15] Microwave Explorer User's Manual, Compact Softwaree Inc., Paterson, NJ, 1994.
[16] Bandler, J. W., Computer-Aided Circuit Optimization, in Modem Filter Theory and Design,
G . C . Temes and S. K. Mitra, eds. New York: Wiley, 1973, pp. 211-271.
[17] Bandler, J. W., and S. H . Chen, “Circuit Optimization: The State of the Art,” IEEE
Trans. Microwave Theory & Tech., Vol. MTT-36, February 1988, pp. 424-443.
[18] OSA9Q⁄hope Reference Manual, Version 3.5, Optimization System Associates Inc., Canada,
1995.
[19] Touchstone Reference Manual, Version 3.0, EEsof Inc., Westlake Village, CA, 1991.
[20] MDS Reference Manual, Release 6.0, Hewlett-Packard Co., Palo Alto, CA, 1994.
[21] Series IV/PC Reference Manual, Version 6.0, Hewlett-Packard Co., Palo Alto, CA, 1995.
[22] Super-Compact User's Manual, Rev. 6.5, Compact Softwaree Inc., Paterson, NJ, 1994.
[23] LINMIC+ User Manual, Version 2.1, Jansen Microwave, Germany, 1989.
[24] M/FILTER Reference Manual, Eagleware Corp., USA, 1993.
[25] MMICAD Reference Manual, Version 2, Optotek Ltd., Canada, 1996.
[26] Cohn, S. B., “Generalized Design of Bandpass and Other Filters by Computer Optimiza-
tion,” IEEE MTT-S Int. Microwave Symp. Dig, June 1974, pp. 272-274.
[27] Postoyalko, V. and D . Budimir, “Design of Waveguide E-Plane Filters With All-Metal
Inserts by Equal Ripple Optimization,” IEEE Trans. Microwave Theory & Tech.,
Vol. 42, No. 2, February 1994, pp. 217-222.
[28] Hasler, M., and J. Neiryuck, Electrical Filters, Dedham, MA: Artech House, 1986.
[29] Cohn, S. B., “Synthesis of Commensurate Comb-Line Band-Pass Filters With Half-
Length Capacitor Lines, and Comparison to Equal-Length and Lumped-Capacitor Cases,”
IEEE MTT-S Int. Microwave Symp. Dig, May 1980, pp. 135-137.
[30] Hunton, J. K., “Novel Contributions to Microwave Circuit Design,” IEEE MTT-S Int.
Microwave Symp. Dig, 1989, pp. 753-755.
[31] Budimir, D., and V. Postoyalko, “EPFILTER: A CAD Package for E-Plane Filters,
Microwave Journal, August 1996, pp. 110-114.
[32] Budimir, D., “Optimized E-Plane Bandpass Filters With Improved Stop Band Perfor-
mance,” IEEE Trans. Microwave Theory & Tech., February 1997, pp. 212-220.
[33] DBFILTER Reference Manual, Tesla Communications Ltd., London, England.
[34] Parry, R., “Optimisation of Microwave Filters,” Colloquium on Filters in RF and Microwave
Communications, Digest No. 1992/220, University of Bradford, Bradford, England,
December 1992, pp. 7/ 1-7/5.
[35] Empipe Reference Manual, Version 3.1, Optimization System Associates Inc., Canada,
1995.

Selected Bibliography
Bahl, I. J., and P. Bhartia, Microwave Solid State Circuit Design, New York: Wiley, 1988.
Empipe3D Reference Manual, Version 3.5, Optimization System Associates Inc., Canada, 1996.
10 Generalized Filter Design by Computer Optimization

Gupta, K. C., R. Garg, and R. Chadha, Computer-Aided Design of Microwave Circuits, Norwood,
MA: Artech House, 1981.
Temes, G. C., and D. A. Calahan, “Computer-Aided Network Optimization the State of the
Art,” Proc. IEEE, Vol. 55, 1967, pp. 1832-1863.
2
Transmission Lines, Lumped Elements,
and Resonators

This chapter summarizes the information for transmission lines (e.g., coaxial
lines, waveguides, striplines, microstrip lines, suspended striplines, coplanar
waveguides (CPWs) i fmlines, image guides, microshield lines), lumped elements,
and resonators (e.g., dielectric resonator) that is most often needed in filter
design. Because of the limited scope of this book, no attempt at completeness
has been made. It is hoped that the references included at the end of the
chapter will direct the interested reader to sources of more detailed information
on particular subjects.

2.1 Transmission Lines

A collection of some transmission line structures commonly used as filter

elements in RF, microwave, and mm-wave filters are shown in Figure 2.1. The
main applications of these filters can be classified under the following [1]:

• Cellular;
• Cordless;
• Personal communication systems (PCSs);
• Personal communication networks (PCNs);
• Wireless local area networks (WLANs);
• Microwave links;
• Satellite communications;

11
12 Generalized Filter Design by Computer Optimization

Circular Coaxial Line

Square Coaxial Line

Rectangular Coaxial Line

Figure 2.1 Transmission line structures commonly used in RF, microwave, and mm-wave
filters.

• Radar systems;
• Electronic warfare systems;
• Automotive electronics;
• Microwave instrumentation;
• Troposcatter systems.

The actual choice of transmission line depends on several factors, including

the type of filter and its operating frequency. Each type of transmission line
has potential advantages for various applications and therefore is briefly dis-
cussed here. Frequency band designations (according to radar standard recom-
mendations) are listed in Table 2.1.

2.1.1 Coaxial Line

The basic structure consists of a hollow outer conductor and a center conductor
supported by means of a dielectric placed in the space between the outer and
 Transmission Lines, Lumped Elements, and Resonators 13

Rectangular Waveguide Rectangular Dielectric

Waveguide

Circular Waveguide Circular Dielectric

Wavegude

Ridged Waveguide Dielectric Loaded

Waveguide

Bilateral Finline Image Line

Figure 2.1 (continued).

the center conductor. The cross-section of a coaxial line may be circular, square,
or rectangular, as shown in Figure 2.1. They often are used in low-power
applications at L-band, S-band, C-band, X-band, Ku-band and in some cases
as high as Ka-band (see Table 2.1). The dominant mode of propagation is the
transverse electromagnetic (TEM) mode. The TEM mode has neither an electric
nor a magnetic field in the direction of propagation. The mode has no cut-
off frequency. The references [2-9] provide a detailed treatment of coaxial
lines and include a wide range of basic design details as well as information
on discontinuities and applications.

2.1.2 Rectangular Waveguide

Probably the most widely used filter elements in high-power and low-loss
microwave and mm-wave filters are rectangular waveguides (see Figure 2.1).
14 Generalized Filter Design by Computer Optimization

Stripline Coupled Striplines

Broadside Coupled Suspended Stripline

Striplines

Microstrip Line
Suspended Microstrip
Line

Coupled Microstrip
Lines Coplanar Strips

Coplanar Waveguide

Figure 2.1 (continued).

The waveguide supports transverse electric (TE) and transverse magnetic (TM)
modes. TE modes have a magnetic field (7¾) but no electric field (2⅛) in the
direction of propagation. They also are referred to as H-modes, or modes of
magnetic type.
For TE ww modes, the electromagnetic field components are given by
r
ιkq∣Qnπ ⁄ mπx∖ , ⁄ nπy
Ex = ------j----- A mn cos ------ sin —7— (2.1)
b ∖ a ∖ b

ιkqT↓ Qmπ ' ⁄ mπx∖ ⁄ nπy

Eyv = ----------- A rnn sin ------ I cos —τ~ (2.2)
a ∖ a ∖ b
 Transmission Lines, Lumped Elements, and Resonators 15

t&l

Metal-pipe on Wafer
Waveguide

ayerl (M3)
ayer2 (M2)
ayer3 (M 1)
Polyimide
&1

Multilayer Coplanar
Waveguide

Membrane

Substrate ∈r

Microshield Line

Figure 2.1 (continued).

Ez = 0 (2.3)

rr jβmnmττ . ⁄ mπx∖ ⁄ nπy

H x = ----------- A mn sin — — cos —7- (2.4)
a ∖ a j ∖ b

τr jβmn n π ( rnπx∖ . ⁄ nπy

Hy = b~— Aδ m n cosl —— ) sιnl — (2.5)

2 ⁄ mτrx∖ . ( n π y
H z = k c A mn cos ------- ) sin —-j— (2.6)
∖ a I ∖ b
16 Generalized Filter Design by Computer Optimization

Table 2.1
F r e q u e n c y B a n d Designation

Band Designation Frequency (GHz)

VHF 0.10-0.30
UHF 0.30-1.00
L 1.00-2.00
S 2.00-4.00
C 4.00-8.00
X 8.00-12.40
Ku 12.40-18.00
K 18.00-26.50
Ka 26.50-40.00
U 40.00-60.50
V 50.00-75.00
E 60.50-92.00
W 75.00-110.00
F 92.00-140.00
D 110.00-170.00
G 140.00-220.00
Y 170.00-260.00
J 220.00-300.00

Modes that have an electric field ( Ez ) but no magnetic field ( j⅛) in the
direction of propagation are known as TM and also are referred to as E-modes,
or modes of electric type. The electromagnetic field components of TM ww
modes are given by

· ~jβmn m π (mπx∖ . (nπy∖

Ec x - Bn m n cos sin , j (2.7)

-jβ m n mτ d . (mπx∖ ⁄ nπy∖

Ec7v = -------.----- Brnn sin ------ cos —γ~ (2.8)
& ∖ a ⁄ ∖ & ⁄

⁄ mττx∖ . ⁄ nπy∖
Ez = kc Brnn sinl —— Isinl— -∣ (2.9)

jk riTT . ⁄ mτrx∖ ⁄ nπy

sin -
Hx =
~bη Bmn
( Λ- ) cθs(~Γ (2.10)

-jk mπ ⁄ mτrx∖ , ⁄ nπy

yj =
ar mn ∞S ∖ a~ ⁄I Sin∖I -j— (2.11)
E)
 Transmission Lines, Lumped Elements, and Resonators 17

Hz =Q (2.12)

where A mn and Bm n are amplitude coefficients, and

ω
yJP'Q e Q (2.13)

7 (2.14)
7θ
*o
.2 , 2
mτr∖ I nττ∖x
(2.15)
a ~b }

The modes that enable single-mode operation in the natural state are
usually referred to as dominant or fundamental, while all other modes are
known as high-order modes. The TEιo is the dominant mode of the rectangular
waveguide. It has the lowest cutoff frequency of all possible rectangular wave-
guide modes. Table 2.2 summarizes the properties of standard air-filled rectan-
gular waveguides. Formulas and some design data for them, as well as
information on discontinuities and applications, are available in several of the
references [2-1 9] . Because Chapters 3 and 6 of this book are devoted to the
study of rectangular waveguide discontinuities and applications, further details
are not given here.

2.1.3 Circular Waveguide

A cross-section of a circular waveguide is given in Figure 2.1. Like rectangular

waveguides, circular waveguides support TE and TM modes. In the circular
waveguide, the dominant mode is the TE∣1 , because it has the lowest cutoff
frequency. The circular waveguide is well suited for application in long-distance
communication at frequencies between 40 and 110 GHz because it has, if
oversized and excited in the TEθι mode, extremely low attenuation. However,
the TEθι is not a dominant mode, thus necessitating the incorporation of
mode filters. Additional details on circular waveguides, as well as information
on discontinuities and applications, can be found in the references [2—19].

2.1.4 Ridged Waveguide

Ridged waveguides are commonly used in some communication systems for

wideband operation. A cross-section of a ridged waveguide is given in
Figure 2.1. The introduction of a ridge to the waveguide reduces the cutoff
18 Generalized Filter Design by Computer Optimization

Table 2.2
Properties of Rectangular Waveguides

Theoretical CW
Cutoff Power Rating,
Recommended Frequency Lowest to
EIA WG Operating for TEoι Highest
Designation Range for TEq1 Mode Frequency Inside Dimensions in mm
WR (WG) Mode (GHz) (GHz) (MW) (Inches in parentheses)

650 (6) 1.12-1.70 0.908 11.90-17.20 165.100-82.550 (6.500-3.250)

510 (7) 1.45-2.20 1.157 7.50-10.70 129.540-64.770 (5.100-2.550)
430 (8) 1.70-2.60 1.372 5.20-7.50 109.220-54.610 (4.300-2.150)
340 (9A) 2.20-3.30 1.736 3.10-4.50 86.360-43.180 (3.400-1.700)
284 (10) 2.60-3.95 2.078 2.20-3.20 72.140-34.040 (2.840-1.340)
229 (11 A ) 3.30-4.90 2.570 1.60-2.20 58.170-29.083 (2.290-1.145)
187 (12) 3.95-5.85 3.152 1.40-2.00 47.550-22.149 (1.872-0.872)
159 (13) 4.90-7.05 3.711 0.79-1.00 40.390-20.193 (1.590-0.795)
137 (14) 5.85-8.20 4.301 0.56-0.71 34.850-15.799 (1.372-0.622)
112 (15) 7.05-10.00 5.259 0.35-0.46 28.499-12.624 (1.122-0.497)
90 (16) 8.20-12.40 6.557 0.20-0.29 22.860-10.160 (0.900-0.400)
75 (17) 10.00-15.00 7.868 0.17-0.23 19.050-9.525 (0.750-0.375)
62 (18) 12.40-18.00 9.486 0.12-0.16 15.799-7.899 (0.622-0.311)
51 (19) 15.00-22.00 11.574 0.08-0.107 12.954-6.477 (0.510-0.255)
42 (20) 18.00-26.50 14.047 0.043-0.058 10.666-4.318 (0.420-0.170)
34 (21) 22.00-33.00 17.328 0.034-0.048 8.636-4.318 (0.340-0.170)
28 (22) 26.50-40.00 21.081 0.022-0.031 7.112-3.556 (0.280-0.140)
22 (23) 33.00-50.00 26.342 0.014-0.020 5.690-2.845 (0.224-0.112)
19 (24) 40.00-60.00 31.357 0.011-0.015 4.775-2.388 (0.188-0.094)
15 (25) 50.00-75.00 39.863 0.0063-0.0090 3.759-1.880 (0.148-0.074)
12 (26) 60.00-90.00 48.350 0.0042-0.0060 3.099-1.549 (0.122-0.061)
10 (27) 75.00-110.00 59.010 0.0030-0.0041 2.540-1.270 (0.100-0.050)
8 (28) 90.00-140.00 73.840 0.0018-0.0026 2.032-1.016 (0.080-0.040)
7 (136) 110.00-170.00 90.840 0.0012-0.0017 1.651-0.8255 (0.065-0.0325)
5 (135) 40.00-220.00 115.750 0.00071-0.00107 1.2954-0.6477 (0.051-0.0255)
4 (137) 170.00-260.00 137.520 0.00052-0.00075 1.0922-0.5461 (0.043-0.0215)
3 (139) 220.00-325.00 173.280 0.00035-0.00047 0.8636-0.4318 (0.034-0.0170)

frequency o f the fundamental mode more than the cutoff o f the first higher
order mode. Thus, the single-mode operation in such a waveguide can be
expanded. By s u i t a b l e selection o f the geometry o f the ridge, the bandwidth
of the ridged waveguide can be c o n t r o l l e d . For more details, the reader is
referred to t h e references [2-9, 1 1 ,20-34] . Because Chapters 3 and 7 of this book
are devoted to the study o f ridged waveguides’ discontinuities and applications,
further d e t a i l s are n o t given here.
 Transmission Lines, Lumped Elements, and Resonators 19

2.1.5 Rectangular Dielectric-Loaded Waveguide

Rectangular dielectric-loaded waveguides are basically a form of capacitively

loaded rectangular waveguides. Figure 2.1 shows a cross-sectional view of this
type of waveguide, which is used in some communication systems for wideband
operation. Like ridged waveguide, the usable bandwidth of dielectric-loaded
waveguide is greater than that of an ordinary rectangular waveguide. However
it has higher conductor losses and lower power-handling capacity than an
ordinary rectangular waveguide. By suitable selection of the geometry of the
dielectric slab, the bandwidth of the dielectric-loaded waveguide can be con-
trolled. For more details, the reader is referred to the references [35-37] .

2.1.6 E-Plane Circuit (Finline)

A dielectric-loaded waveguide with fins is called a printed E-plane circuit. It

can be viewed as a slot line inserted into the E-plane of a rectangular waveguide
or as a ridged waveguide with thin ridges backed by dielectric substrate.
It includes all-metal E-plane circuit and fmlines (see Figure 2.1). The
advantages of low insertion loss, simple fabrication, and wide single-mode
bandwidth have made this circuit widely utilized in low- and medium-power
mm-wave applications. The mode of propagation is a hybrid mode consisting
of a combination of TE and TM modes. In Beyer [38], the line parameters
of these waveguides are determined numerically as well as experimentally for
the first time considering the metallization thickness and the influence of the
longitudinal slits in the mount grove. A substantial amount of work on these
structures has been reported in the literature [1 1,35,38 44]. Because Chapters
3, 6, and 7 of this book are devoted to the study of all-metal E-plane circuits,
further details are not given here.

2.1.7 Dielectric Waveguide

At mm-wave frequencies (up to around 140 GHz), where precise fabrication

of ordinary rectangular waveguides, suspended striplines, and fmlines becomes
extremely difficult, dielectric waveguide structures offer an alternative approach
with the potential for lower losses and relaxed tolerances. There are several
different dielectric waveguide structures, such as dielectric-rod, dielectric-slab
(image line), trapped image line, insulated image line, and inverted stripline.
Cross-sectional views of dielectric-rod waveguide and image line are shown in
Figure 2.1. The dielectric-rod waveguide consists of a rectangular dielectric
rod surrounded by an infinite air medium. The dominant mode of a dielectric-
rod waveguide has no cutoff frequency, so in theory its operational bandwidth
is unlimited. The image line consists of a rectangular dielectric slab placed on
20 Generalized Hlter Design by Computer Optimization

a backing sheet of a perfectly conducting ground plane. The structure behaves

like a dielectric-rod waveguide, which supports hybrid modes. The main draw-
back of the structures is the radiation loss at junctions, discontinuities, and
bends. Also, the dielectric/metal adhesives are very lossy, which reduces the
practical, loaded Q-factor. Because of those difficulties, some further low-cost
contenders have been considered for operation at frequencies around and
exceeding 100 GHz. A detailed discussion of those structures has been reported
in the literature [2,10,1 1,35-37, 45].

2.1.8 Metal-Pipe Waveguide

Monolithic transmission line for submm-wave and terahertz frequency applica-
tions may be realized by considering variants of the early dielectric waveguides.
The metal-pipe on-wafer waveguides are constructed from dielectric materials
and structures, which are available in monolithic technology, so the use of
them in integrated circuits is possible. A cross-sectional view of such a transmis-
sion line is shown in Figure 2.1. These structures are used not only as transmis-
sion lines but also as filter elements [46] . For more informations about these
lines, the reader is referred to the references [46—49].

2.1.9 Stripline
One of the most commonly used homogeneous planar transmission lines in
a microwave low-power transmission system is stripline. A cross-sectional view
of stripline is given in Figure 2.1. The dominant mode of propagation in this
type of line is the TEM mode. The basic structure consists of the flat strip
conductor situated symmetrically between two large ground planes. The electric
and magnetic fields are concentrated around the strip conductor and away
from the strip fields that decay rapidly with distance. Formulas and some
design data for these lines and applications are available in the literature
[6-9,1 l,5O-55].
Suspended and coupled striplines are the most useful variants of stripline.
Basically, the suspended stripline (see Figure 2.1) is an inhomogeneous line
in which the substrate carrying the strip conductor is placed symmetrically
between the two ground planes, thereby leaving an air gap on either side of
the substrate. Reduced thickness of the substrate partly decreases the dielectric
losses. However, the considerable conduction losses, particularly at the edges
of the strips, remain unchanged. The shielding eliminates transverse coupling
and radiation but contributes to additional costs. The possibility of breakage
of the thin substrate is a disadvantage that cannot be disregarded. The configura-
tion of parallel coupled stripline is shown in Fig 2.1. The principal application
areas of parallel coupled striplines are filters and a variety of other useful circuits.
 Transmission Lines, Lumped Elements, and Resonators 21

2.1.10 Microstrip Line

Microstrip is the second basic type of planar transmission line, developed

immediately after stripline. Unlike stripline, microstrip is an inhomogeneous
structure with quasi-TEM as the dominant mode of propagation. A variant
of this line, suspended microstrip, incorporates an air gap between the substrate
and the ground (see Figure 2.1). It provides a higher Q (500 to 1,500) than
microstrip. The wide range of impedance values achievable makes these lines
particularly suitable for filters.
A coupled microstrip line configuration consists of two transmission lines
placed parallel to each other and in close proximity. Coupled lines are utilized
extensively as basic elements for filters and a variety of other useful circuits.
Because of the coupling of electromagnetic fields, a pair of coupled lines
can support two different modes of propagation. These modes have different
characteristic impedances. When the lines are imbedded in a homogeneous
dielectric medium the velocity of propagation of these two modes is equal.
However, for transmission lines such as coupled microstrip lines the dielectric
medium is not homogeneous. In the configuration for the parallel coupled
microstrip lines, a part of the field extends into the air above the substrate (see
Figure 2.1). This fraction of total field is different for the two modes of the
coupled lines. Consequently, the effective dielectric constant and the phase
velocities are not equal for the two modes. This nonsynchronous feature reduces
the performance of circuits using these types of coupled lines. Numerous papers
and a few books [3,7,8,10,11,44,50,56-74] deal with the analysis of single,
suspended, and coupled microstrip lines, characterization of discontinuities,
and various other aspects such as losses, dispersive behavior, and applications.

2.1.11 Coplanar Line

Coplanar waveguide and coplanar strip structures (see Figure 2.1) belong to
the category of coplanar lines. All conductors are on the same side of a dielectric
substrate. The dominant mode of propagation in the coplanar lines is a quasi-
TEM mode, because of different dielectrics above and below the conductors.
A CPW consists of a strip conductor placed on a dielectric substrate with two
ground planes adjacent and parallel to the strip, while a coplanar strip structure
comprising a pair of coplanar strips is placed on a dielectric substrate. Coplanar
strips are used more in high-speed digital circuits then in microwave circuits.
A detailed discussion of these lines, characterization of discontinuities, and
various other aspects such as losses, dispersive behavior, and applications can
be found in the references [3, 7,8,10, 50, 56, 57, 72,74].
22 Generalized Filter Design by Computer Optimization

2.1.12 Multilayer Coplanar Waveguide

With dielectric and conductor layers, additional degrees of freedom arise for
filter design, like an extended range of impedances. In the case of Monolithic
Microwave Integrated Circuit (MMIC) filters, the additional layer can be added
using a thin film of polyimide or silicon nitride. For example, a fifth-order
Iowpass CPW multilayer MMIC filter was tested (Figure 2.2). It consists of
three low-impedance V-shaped multilayer and high-impedance elevated CPW
transmission lines on a GaAs substrate. With reference to Figure 2.1, the
top, middle, and bottom sections of the structure are made from aluminum
metallization layers, M3, M2, and M l , respectively. Layers of polyimide are
employed for separation of the two metal layers M3⁄M2 and M2⁄M1. A
photograph of a V-shaped multilayer CPW is shown in Figure 2.3. A description
of this transmission structure is available in the literature [75—81].

2.1.13 Microshield Line

Microshield line was introduced as an alternative transmission medium to
CPW for mm-wave and submm-wave applications. Dib et al. describes this

CPULPF DB/KCL/BS

Figure 2.2 Layout of lowpass multilayer coplanar waveguide filter.

Figure 2.3 Photomicrograph of multilayer coplanar waveguide.

 Transmission Lines, Lumped Elements, and Resonators 23

type of waveguide extensively [83] . It is a half-shielded geometry, quasi-planar

transmission line similar to CPW, which uses a thin (around 1.5 μm) dielectric
membrane to support the conducting lines and upper ground planes above a
metallized shielding cavity. That configuration allows single-mode TEM wave
propagation over a broad bandwidth with zero dielectric loss and minimal
dispersion. A description of these and other advantages of the microshield
geometry can be found in the literature [82-86] .

2.2 Comparison of Various Transmission Lines

The preceding sections briefly described several different types of transmission

line structures. For a given frequency range, the selection of a particular transmis-
sion line structure is based on several considerations, some of which are presented
in Table 2.3.

2.3 Lumped Elements

This section briefly describes commonly used lumped elements (as filter ele-
ments). Detailed discussion of lumped elements has been reported in the
literature [7,8,56,87-9 1]. Filter elements can be both lumped elements (dimen-
sions <0.1 wavelength) and distribution elements that are composed of sections
of transmission lines and waveguides. The choice of filter elements depends
on the frequency of operation. At lower frequencies, the lumped element filters
that have a lower Q than distributed element filters have the advantage of
lower cost, smaller size, and wideband performances. However, it is often

Table 2.3
Impedance Ranges, Unloaded Q-Factors, and Frequency Ranges for Various
Transmission Line Structures

Characteristic
Transmission Line Impedance Range Unloaded Q-Factor (Qu ) Frequency Range
Structure (∩) at 30 GHz (GHz)

Microstrip 20-125 250 up to 110

CPW 25-150 100 up to 100
Suspended stripline 40-150 up to 600 up to 110
Suspended microstrip 40-150 500-1,500 up to 110
Finline 10-400 500 up to 110
Image line ≈26 2,700 up to 110
24 Generalized Filter Design by Computer Optimization

difficult to realize a truly lumped element even at lower frequencies because

of the parasitics to ground associated with thin substrates. With the advent of
new techniques [91], the fabrication of lumped elements can now be extended
from X-band to about 100 GHz. The two basic building blocks for filter design
available in lumped form are inductors and capacitors. The design of lumped
element filters by computer optimization (described in Chapter 5) requires a
complete and accurate characterization of lumped elements at high frequencies.
That necessitates the development of electromagnetic models, which take into
account the presence of ground planes, proximity effect, fringing fields, parasit-
ics, and so on.

2.3.1 Inductors

Depending on the inductance required and the frequency of operation, induc-

tors can be realized in a few types and all are achieved with a single-layer
metallization scheme. High-impedance line section, meander inductor, and
square spiral inductor are illustrated in Figure 2.4. Table 2.4 compares various
inductors.

High-Impedance
Line Section

π∏∏∏-
Meander Inductor

Lead-out
bridge

Square Spiral Inductor

Figure 2.4 Planar inductor configurations.

 Transmission Lines, Lumped Elements, and Resonators 25

Table 2.4
Comparison of Various Planar Inductors

Inductance Range Unloaded Q Factor

Inductor Type (nH) (fl 0 ) at 10 GHz

Straight section of line 2-3 50

Spiral (circular, square, or rectangular) >3 >50

2.3.2 Capacitors
A lumped capacitor can be realized by the use of a single metallization scheme
and a two-level metallization technology in conjunction with dielectric films.
Figure 2.5 shows a low-impedance transmission line, end-coupled, and interdig-
ital capacitor. Choosing from these capacitors depends on the capacitance
required, the frequency of operation, and the processing technology available.
Table 2.5 compares three types of capacitors. Figure 2.2 illustrated a lowpass
filter realized by cascaded sections of high-impedance multilayer CPWs (induc-
tors) and low-impedance multilayer CPWs (capacitors) (see Figure 2.1).

2.4 Resonators
Structures extensively used as filter elements in the realization of various band-
pass and bandstop filters are resonant structures. At low frequencies, resonant

Low-Impedance
Line Section

End Coupled

Interdigital Capacitor

Figure 2.5 Planar capacitor configurations.

26 Generalized Filter Design by Computer Optimization

Table 2.5
Comparison of Various Planar Capacitors

Capacitance Range Unloaded Q Factor

Capacitor Type (pF) (fl u ) at 10 GHz

Straight section of line <0.2 50

End coupled <0.2 50
Interdigital 0.05-0.5 50
Metal insulator metal (overlay) 0.1-30 50-100

structures invariably are composed of lumped elements. At higher frequencies,

however, lumped elements in general cannot be employed because of low Q
and difficulties in realizing inductances and capacitances at high frequencies.
Distributed elements are widely used to overcome those limitations. The reso-
nant structures commonly used in various bandpass and bandstop filters can
be realized by several forms, such as lumped element resonators, cavity resonators
(Figure 2.6), planar resonators (Figure 2.7), and dielectric resonators
(Figure 2.8). Choosing from these forms depends on the mechanical size of
the resonator at the resonant frequency, the unloaded Q, and the temperature
stability and electrical tunability of the frequency. This section presents a brief
description of the dielectric resonators commonly used in various bandpass
and bandstop filters. A detailed description of resonant structures can be found
in the literature [2-4,8,56,92-98] .

2.4.1 Dielectric Resonator

A dielectric resonator consists of a solid, low-loss, temperature-stable, high-

permitivity (100 > 6r > 30), and high- Q dielectric material, formed in a few
geometrical shapes, as shown in Figure 2.8. It resonates in various modes at

ε0 ε0

Rectangular Cavity
Cylindrical Cavity

Figure 2.6 Cavity resonators.

 Transmission Lines, Lumped Elements, and Resonators 27

Half-Wavelength Microstrip Ring

Microstrip Resonator Resonator

Rectangular Microstrip Circular Microstrip

Resonator resonator

Figure 2.7 Planar resonators.

frequencies determined by resonator dimensions and shielding conditions.

Typical geometrical shapes used for dielectric resonators are as follows:

• Disc type;
• Cylinder type;
• Dielectric-loaded type;
• TEM coaxial dielectric resonator;
• Distributed Bragg reflector resonator.

With their high unloaded Q ( Q w ), high dielectric constant, and high-

temperature-stability dielectric resonators are used in place of waveguide compo-
nents in a large variety of microwave systems to reduce their size by a factor
of approximately l⁄y∣~er for equal electrical performances. The intrinsic loss of
a dielectric material can be expressed in terms of its loss tangent (tan <5). If
the RF field is totally contained within the dielectric, then the resonator
unloaded Q is given as

qu =
- ■ J7 (2.16)

If the field extends outside the dielectric (as it will in practice), it will excite
currents and produce ohmic loss in the surrounding enclosure. In that case,
28 Generalized Filter Design by Computer Optimization

Dielectric Resonator
(Cylinderical Type)
Dielectric Resonator
with Support

Dielectric Resonator
(Disk Type)
Dielectric Resonator
(Coaxial Type)

Dielectric Resonator
(Rectangular) Dielectric Distributed
Bragg Reflector

Figure 2.8 Dielectric resonators.

J _____]_ 1
+ (2.17)
Q.u Qd Qc

where Q e is the Q factor due to conductor losses only. It is apparent that the
high-resonator Q possible with low-loss dielectrics is achievable only if Q f can be
maximized. In practical cases, the enclosure losses may dominate the resonator Q
or determine the maximum amount of miniaturization that can be achieved
for a given Q u . The actual loss of a dielectric resonator will be determined by
the following:

• Mode of operation, which is the field configuration at resonance and

is determined by the geometry and mounting of the dielectric on its
enclosure;
 Transmission Lines, Lumped Elements, and Resonators 29

• Loss tangent of support material: many modes of operation require

the dielectric resonator to be positioned in the center of a cavity, which
is achieved in practice by positioning the puck on a low dielectric
constant support;
• RF loss of adhesive used to secure dielectric;
• Resistivity and size of metallic enclosure;
• Tuning and coupling structures.

The resonator bandwidth is inversely proportional to the Q factor; thus,

high Q-factor resonators have narrow bandwidths.

2.4.2 Dielectric Material

The important properties of the dielectric material used in dielectric resonators
are as follows:

• Q factor, which is equal to the inverse of the loss tangent. The Q

factor of the dielectric resonator decreases with increasing in frequency.
• Dielectric constant.
• The temperature coefficient at the resonant frequency, Tp which
includes the combined effects of the temperature coefficient of the
dielectric constant and the thermal expression of the dielectric. The
temperature coefficient, 7χ, of the resonator can be controlled in some
materials by modifying the composition.

Manufacturers’ data from a selection of available materials are given in

Tables 2.6-2.10. The loss is given as the dielectric Q, which is the reciprocal

Table 2.6
Properties of Dielectric Materials Manufactured by NTK Piezoelectric Ceramics [97]

a Operating Frequency (GHz)

A material 12.6 18,000 13.0

C material 21.0 12,000 6.0
F material 34.0 8,000 8.0
E material 45.0 10,600 4.0
T material 80.0 1,500 2.7
L material 93.0 1,350 2.7
S material 6.40 2,200 10.0
30 Generalized Filter Design by Computer Optimization

Table 2.7
Properties of Dielectric Materials Manufactured by Tekelec Components [98]

€/> a Operating Frequency (GHz)

E 2000 series 36.9-37.2 9,500 5.0

E 9000 series 36.0-37.0 40,000 1.0

Table 2.8
Properties of Some Typical Dielectric Materials [99]

€f Q Operating Frequency (GHz)

Sapphire 9.394 450,000 13.2

Sapphire 9.394 650,000 9.0

of the loss tangent. The loss tangent normally is linear with frequency, so that
for a given material the product of Q and frequency is a constant. That enables
the loss to be extrapolated to other frequencies. The main purpose of
Tables 2.6-2. 10 is to give an idea of what is typically known about the properties
of materials used in dielectric resonators.
 Transmission Lines, Lumped Elements, and Resonators 31

Table 2.9
Properties of Dielectric Materials Manufactured by Alpha/Trans-Tech, Inc. [100]

8300 Series e r = 35-36.5 Q > 28,000 at 850 MHz

Composition: Barium titanate
Frequency range: Disc type: 800-13,800 MHz
Frequency range: Cylinder type (ring): 800-9010 MHz
Temperature coefficient of TE0 ι mode resonant frequency, T f (ppm⁄C) ± (2.0 or 1.0)

8600 Series ε r = 80.0 Q > 3,000 at 3.000 GHz

Composition: BaLnTi Oxide
Frequency range: Disc type: 700-3618 MHz
Frequency range: Cylinder type (ring): 967-3618 MHz
Temperature coefficient of TE0 u mode resonant frequency, 7} (ppm⁄C) ± (2.0 or 1.0)

8700 Series ef = 27.6-30.6 Q > 10,000 at 10.0 GHz

Composition: BaZnTaTi Oxide
Frequency range: Disc type: 5550-32150 MHz
Frequency range: Cylinder type (ring): 5550-9870 MHz
Temperature coefficient of TE0 u mode resonant frequency, 7> (ppm⁄C) ± (2.0 or 1.0)

8800 Series e r = 36.6-38.3 Q > 6,000 at 4.5 GHz

Composition: BaTitanium Oxide
Frequency range: Disc type: 790-5210 MHz
Frequency range: Cylinder type (ring): 1390-5210 MHz
Temperature coefficient of TE0 u mode resonant frequency, Tf (ppm⁄C) ± (2.0 or 1.0)

9000 Series e r = 85.5 Q > 1,500 at 1.0 GHz

Frequency range: TEM mode resonator of rod: 320-5000 MHz
32 Generalized Filter Design by Computer Optimization

Table 2.10
Properties of Some Typical Dielectric Materials Manufactured by Murata Manufacturing
Co., Ltd. [101]

Brand: Resomics-U series er = 36.0-40.0 Q = 6,000 at 7.0 GHz

Composition: (Zr,Sn) Ti04
Frequency range: 1-12 GHz

Brand: Resomics-R series er = 36.0-40.0 Q = 6,000 at 7.0 GHz

Composition: (Zr,Sn) Ti04
Frequency range: 1-12 GHz

Brand: Resomics-R series eγ = 36.0-40.0 Q = 6,000 at 7.0 GHz

Composition: (Zr,Sn) Ti04
Frequency range: 1-12 GHz

Brand: Resomics-E series ef = 36.0-40.0 Q = 6,000 at 7.0 GHz

Composition: (Zr,Sn) Ti04
Frequency range: 1-12 GHz

Brand: Resomics-F series (Disc type) er = 24 Q = 35,000 at 10 GHz

Frequency range: 10-25 GHz

Brand: Resomics-F series (Ring type) βf = 24 Q = 35,000 at 10 GHz

Frequency range: 10-19.5 GHz

Brand: Resomics-TEM mode resonator of rod (silver-plated) material:

U er = 37-39 Q = 300 at 2500 MHz
K 6f = 91-93 Q = 500 at 1500 MHz
Frequency range: 400-4800 MHz

Brand: Resomics-TEM mode resonator of rod (copper-plated) material:

K 6f = 91-93 1-5 Q = 500 at 1500 MHz
Frequency range: 500-3000 MHz

References
[1] Special Issue on Emerging Commercial and Consumer Circuits, Systems, and Their
Applications, IEEE Trans. Microwave Theory & Tech., Vol. MTT-43, July 1995.
[2] Ramo, S., J. R. Whinnery, and T. Van Duzer, Fields and Waves in Communication
Electronics,3rd ed., New York: Wiley, 1994.
[3] Collin, R. E., Foundations for Microwave Engineering, 2nd ed., New York: McGraw-
Hill, 1992.
[4] Rizzi, P. A., Microwave Engineering Passive Circuits, Englewood Cliffs, NJ: Prentice-
Hall, 1988.
[5] Pozar, D. M., Microwave Engineering, Reading, MA: Addison-Wesley, 1990.
 Transmission Lines, Lumped Elements, and Resonators 33

[6] Matthaei, G., L. Young, and E. M . T. Yones, Microwave Filters, Impedance Matching
Networks, and Coupling Structures, 2nd ed., Norwood, MA: Artech House, 1980.
[7] Gupta, K. C., R. Garg, and R. Chadha, Computer-Aided Design of Microwave Circuits,
Norwood, MA: Artech House, 1981.
[8] Bahl, I. J., and P. Bhartia, Microwave Solid State Circuit Design, New York: Wiley,
1988.
[9] Saad, T. S., Microwave Engineers Handbook, Vol. 1, Dedham, MA: Artech House, 1971.
[10] Bhartia, P., and I. J. Bahl, Millimeter Wave Engineering and Applications, New York:
Wiley, 1984.
[11] Chang, K., ed., Handbook of Microwave and Optical Components, Vol. 1, New York:
Wiley, 1989.
[12] Marcuvitz, N., Waveguide Handbook, Rad. Lab. Series, Vol. 10, New York: McGraw-
Hill, 1951.
[13] Schwinger, J., and D. Saxon, Discontinuities in Wave Guide (Documents on Modem
Physics), New York: Gordon and Greach, 1968.
[14] Collin, R. E., Field Theory of Guided Waves, 2nd ed., Piscataway, NJ: IEEE Press, 1990.
[15] Wexler, A., “Solution of Waveguide Discontinuities by Modal Analysis,” IEEE Trans.
Microwave Theory & Tech., Vol. MTT-15, September 1967, pp. 508-517.
[16] Kuhn, E., “A Mode-Matching Method for Solving Field Problems in Waveguide and
Resonator Circuits,” AEU, Band Tl , Heft 12, 1973, pp. 511-518.
[17] Patzelt, H., and F. Arndt, “Double-Plane Steps in Rectangular Waveguide and Their
Application for Transformers, Irises and Filters,” IEEE Trans. Microwave Theory &
Tech., Vol. MTT-30, May 1982, pp. 771-776.
[18] Shih, Y. C.,andK. Gray, “Convergence of Numerical Solutions of Step-Type Waveguide
Discontinuity Problems by Modal Analysis,” IEEE MTT-S Int. Microwave Symp. Dig.,
1983, pp. 233-235.
[19] Tao, J. W., and H . Baudrand, “Multimodal Variational Analysis of Uniaxial Waveguide
Discontinuities,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-39, March 1991,
pp. 506-516.
[20] Cohn, S. B., “Properties of Ridge Waveguide,” Proc. IRE, Vol. 35, August 1947,
pp. 783-788.
[21] Hopfer, S., “The Design of Ridged Waveguide,” IRE Trans. Microwave Theory & Tech.,
Vol. MTT-3, October 1955, pp. 20-29.
[22] Pyle, J. R., “The Cutoff Wavelength of the TE10 Mode in Ridged Rectangular Wave-
guide of Any Aspect Ratio,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-14,
April 1966, pp. 175-183.
[23] Montgomery, J. P., “ On the Complete Eigenvalue Solution or Ridged Waveguide,”
IEEE Trans. Microwave Theory & Tech., Vol. MTT-19, June 1971, pp. 547-555.
[24] Dasgupta, D., and P. K. Saha, “Eigenvalue Spectrum of Rectangular Waveguide With
Two Symmetrically Placed Double Ridges,” IEEE Trans. Microwave Theory & Tech.,
Vol. MTT-29, January 1981, pp. 47-51.
[25] Hoefer, W. J. R., and M. R. Burtin, “Closed-Form Expressions for the Parameters of
Finned and Ridged Waveguide,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-30,
December 1982, pp. 2190-2194.
34 Generalized Filter Design by Computer Optimization

[26] Utsumi, Y., “Variation Analysis of Ridged Waveguide Modes,” IEEE Trans. Microwave
Theory & Tech., Vol. MTT-33, February 1985, pp. 111-120.
[27] Tao, J. W., and H . Baudrand, “Rigorous Analysis of Triple-Ridge Waveguides,” Electron-
ics Letters, Vol. 24, No. 13, June 1988, pp. 820-821.
[28] Bornemann, J., and F. Arndt, “Transverse Resonance, Standing Wave and Resonator
Formulations of the Ridge Waveguide Eigenvalue Problem and Its Application to the
Design of E-Plane Finned Waveguide Filters,” IEEE Trans. Microwave Theory & Tech.,
Vol. MTT-38, August 1990, pp. 1104-1113.
[29] Bornemann, J., “Comparison Between Different Formulations of the Transverse Reso-
nance Field-Matching Technique for the Three Dimensional Analysis of Metal-Fined
Waveguide Resonators,” Int. J. Numerical Modelling, Vol. 4 , March 1991, pp. 63-73.
[30] Fan, P., and D . Fan, “Computer Aided Design of E-Plane Waveguide Passive Compo-
nents,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-37, February 1989,
pp. 335-339.
[31] Gebauer, A., and F. Hernandez-Gil, “Analysis and Design of Waveguide Multiplexers
Using the Finite Element Method,” Proc. 18th European Microwave Conf., Stockholm,
Sweden, 1988, pp. 521-524.
[32] Omar, A. S., and K. Schunemann, “Application of the Generalized Spectral-Domain
Technique to the Analysis of Rectangular Waveguides With Rectangular and Circular
Metal Inserts,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-39, June 1991,
pp. 944-952.
[33] Getsinger, W. J., “Ridged Waveguide Field Description and Application to Direction
Couplers,” IRE Trans. Microwave Theory & Tech., Vol. MTT-10, January 1962,
pp. 41-51.
[34] Konishi, Y., and H . Matsumura, “Short End Effect of Ridge Guide With Planar Circuit
Mounted in Waveguide,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-26,
October 1978, pp. 716-719.
[35] Mansur, R. R., R. S. K. Tong, and R. H . Macphie, “Simplified Description of the
Field Distribution in Finlines and Ridge Waveguide and Its Application to the Analysis
of E-Plane Discontinuities,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-36,
December 1988, pp. 1825-1832.
[36] Knox, R., “Dielectric Waveguide Microwave Integrated Circuits— An Overview,” IEEE
Trans. Microwave Theory & Tech., Vol. MTT-24, No. 1 1, November 1976, pp. 806-814.
[37] Solbach, K., “The Fabrication of Dielectric Image Lines Using Casting Resins and the
Properties of the Lines in the Millimeter Wave Range,” IEEE Trans. Microwave Theory &
Tech., Vol. MTT-24, November 1976, pp. 879-881.
[38] Beyer, A., “Analysis of the Characteristics on an Earthed Fin Line,” IEEE Trans.
Microwave Theory & Tech., Vol. MTT-29, July 1981, pp. 676—680.
[39] Meier, P., “Integrated Fin-Line Millimeter Components,” IEEE Trans. Microwave
Theory & Tech., Vol. MTT- 22, No. 12, December 1974, pp. 1209-1216.
[40] Solbach, K., “The Status of Printed Millimeter Wave E-Plane Circuits,” IEEE Trans.
Microwave Theory & Tech., Vol. MTT-31, February 1983, pp. 107-121.
[41] Bhat, B., and S. Koul, “Analysis Design and Applications of Finlines” Norwood, MA:
Artech House, 1987.
 Transmission Lines, Lumped Elements, and Resonators 35

[42] Omar, A. S., and K. Schunemann, “Transmission Matrix Representation of Finline

Discontinuities,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-33, September
1985, pp. 765-770.
[43] Bornemann, J., and F. Arndt, “Modal S-Matrix Design of Optimum Stepped Ridged
and Finned Waveguide Transformers,” IEEE Trans. Microwave Theory & Tech.,
Vol. MTT-35, June 1987, pp. 561-567.
[44] Gupta, K. C., et al., Microstrip Lines and Slotlines, 2nd ed., Norwood, MA: Artech
House, 1996.
[45] Gelsthorpe, R. V., N. Williams, and N. M . Davey, “Dielectric Waveguide: A Low Cost
Technology for Millimeter Wave Integrated Circuits,” Radio and Electronics Engineer,
Vol. 52, November-December 1982, pp. 522-528.
[46] Brown, D . A., A. S. Treen, and N. J. Cronin, “Micromachining of Terahertz Waveguide
Components With Integrated Active Devices,” 19th Int. Conf. Infrared and Millimetre-
waves, Sendai, Japan, 1994.
[47] Yeh, C., F. I. Shimabukuro, and J. Chu, “Dielectric Ribbon Waveguides: An Optimum
Configuration for Ultra-Low-Loss Millimetre/Submillimeter Dielectric Waveguide,”
IEEE Trans. Microwave Theory & Tech., Vol. MTT-38, June 1990.
[48] Engel, A. G., Jr., and L. P. B. Katehi, “Low- Loss Monolithic Transmission Lines for
Submillimeter and Teraherz Frequency Applications,” IEEE Trans. Microwave Theory &
Tech., Vol. MTT-39, November 1991, pp. 1847-1854.
[49] Treen, A. S., and N. J. Cronin, “Terahertz Metal-Pipe Waveguides,” 18th Int. Conf.
Infrared and Millimetre-Waves, Essex, UK, 1993, pp. 470-471.
[50] Hoffman, R. K., Handbook of Microwave Integrated Circuits, Norwood, MA: Artech
House, 1987.
[51] Lucyszyn, S., et al., “Design of Compact Monolithic Dielectric Filled Metal-Pipe
Rectangular Waveguide for Millimetre-Wave Applications,” IEE Proc. Microwave, Anten-
nas and Propagation, Part H , Vol. 143, No. 5, October 1996.
[52] Bahl, I. J., and R. Garg, “Designer’s Guide to Stripline Circuits,” Microwaves, Vol. 17,
January 1978, pp. 90-96.
[53] Cohn, S. B., “Characteristic Impedance of Shielded Strip Transmission Line,” IRE
Trans. Microwave Theory Tech., Vol. MTT-2, July 1954, pp. 52-55.
[54] Howe, H., Jr., Stripline Circuits Design, Dedham, MA: Artech House, 1974.
[55] Barrett, R. M., “Microwave Printed Circuits— The Early Years,” IEEE Trans. Microwave
Theory & Tech., Vol. MTT-32, No. 9, September 1984, pp. 983-990.
[56] Konishi, Y., ed., Microwave Integrated Circuits, New York: Dekker, 1991.
[57] Howe, H., Jr., “Microwave Integrated Circuits-An Historical Perspective,” IEEE Trans.
Microwave Theory & Tech., Vol. MTT-32, No. 9, September 1984, pp. 991-996.
[58] Edwards, T., Foundation for Microstrip Circuit Design, 2nd ed., Chichester, England:
Wiley, 1992.
[59] Gardiol, F., Microstrip Circuit, New York: Wiley, 1994.
[60] Frey, J., and K. Bhasin, eds., Microwave Integrated Circuits, Norwood, MA: Artech
House, 1985.
[61] Schneider, M. V., “Microstrip Lines for Microwave Integrated Circuits,” BellSyst. Tech.
J., Vol. 48, May-June 1969, pp. 1421-1444.
36 Generalized Filter Design by Computer Optimization

[62] Denlinger, E. J., “A Frequency Dependent Solution for Microstrip Transmission Lines,”
IEEE Trans. Microwave Theory & Tech., Vol. MTT-19, No. 1, Jan. 1971, pp. 30-39.
[63] Bryant, T. G., and J. A. Weiss, “Parameters of Microstrip Transmission Lines and of
Coupled Pairs of Microstrip Lines,” IEEE Trans. Microwave Theory & Tech., Vol.
MTT-16, No. 12, December. 1968, pp. 1021-1027.
[64] Silvester, P., “TEM Wave Properties of Microstrip Transmission Lines,” Proc. IEEE,
Vol. 115, 1968, pp. 43-48.
[65] Schneider, M. V., “Microstrip Lines for Microwave Integrated Circuits,” Bell System
Tech. J., Vol. 48, 1969, pp. 1421-1444.
[66] Garg, R., “Design Equations for Coupled Microstrip Lines,” Int. J. Electron., Vol. 47,
1979, pp. 587-591.
[67] Garg, R., and I. J. Bahl,, “Characteristics of Coupled Microstriplines,” IEEE Trans.
Microwave Theory & Tech., Vol. MTT-27, July 1979, pp. 700-705∙
[68] Getsinger, W. J., “Dispersion of Parallel-Coupled Microstrips,” IEEE Trans. Microwave
Theory & Tech., Vol. MTT-21, 1973, pp. 144—145.
[69] Akhtarzad, S., et al, “The Design of Coupled Microstrip Lines,” IEEE Trans. Microwave
Theory & Tech., Vol. MTT-23, June 1975, pp. 486-492.
[70] Hinton, J. H., “O n Design of Coupled Microstrip Lines,” IEEE Trans. Microwave
Theory & Tech., Vol. MTT- 28, March 1980, pp. Tl'l.
[71] Vendelin, G . D., “Limitations on Stripline Q,” Microwave J., Vol. 13, May 1970,
pp. 63-69.
[72] Mittra, R., and T. Itoh, “Analysis of Microstrip Transmission Lines,” in Advances in
Microwaves, Vol. 8, New York: Academic Press, 1974.
[73] Itoh, T., “Overview of Quasi-Planar Transmission Lines,” IEEE Trans. Microwave
Theory & Tech., Vol. MTT-37, February 1989, pp. 275-280.
[74] Chu, T. S., and T . Itoh, “Comparative Study of Mode-Matching Formulations for
Microstrip Discontinuity Problems,” IEEE Trans. Microwave Theory & Tech.,
Vol. MTT-33, October 1985, pp. 1018-1023.
[75] Wen, C . P., “Coplanar Waveguide: A Surface Strip Transmission Line Suitable for
Nonreciprocal Gyromagnetic Device Application,” IEEE Trans. Microwave Theory &
Tech., Vol. MTT-17, No. 12, December 1969, pp. 1087-1090.
[76] Budimir, D., et al., “V-Shaped CPW Transmission Lines for Multilayer MMICs,” IEE
Electronics Letters, October 1995, pp. 1928-1929.
[77] Wang, Q., et al., “Fabrication and Microwave Characterisation of Multilayer Circuits
for M M I C Applications,” IEE Proceedings, Part H , June 1996.
[78] Budimir, D., et al., “Low Loss Multilayer Coplanar Waveguide Transmission Lines on
Silicon Substrate for MMICs,” Proc. 26th European Microwave Conf., Prague, Czech
Republic, September 1996.
[79] Budimir, D., et al., “Novel Transmission-Lines With V-shaped Conductors for HTS
Circuit and M M I C Applications,” IEE Proc. Microwave, Antennas and Propagation, Part
H , 1997.
[80] Budimir, D., et al., “CPW Lowpass Filters on Silicon Substrate for Multilayer MMICs,”
Microwave and Optical Technology Letters, 1997.
 Transmission Lines, Lumped Elements, and Resonators 37

[81] Gokdemir, T., et al., “Multilayer Passive Components for Uniplanar Si/SiGe MMICs,”
IEEE MTT-S Int. Microwave Symp. Digest, June 1997.
[82] Dib, N. I., et al., “Study of a Novel Planar Transmission Line,” IEEE MTT-S Int.
Microwave Symp. Digest, 1991, pp. 623-626.
[83] Robertson, S. V., L. P. B. Katehi, and G . M. Rebeiz, “W-Band Microshield Low-Pass
Filters,” IEEE MTT-S Int. Microwave Symp. Digest, 1994, pp. 625-628.
[84] Weller, T. M., and L. P. B. Katehi, “Miniature Stub and Filter Designs Using the
Microshield Transmission Line,” IEEE MTT-S Int. Microwave Symp. Digest, 1995, pp.
675-678.
[85] Weller, T. M., L. P. B. Katehi, and G. M. Rebeiz, “High Performance Microshield
Line Components,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-43, March
1995, pp. 534-543.
[86] Weller, T. M., L. P. B. Katehi, and G. M. Rebeiz, “Novel Micromachined Approaches
to MMICs Using Low-Parasitic, High-Performance Transmission Media and Environ-
ments,” IEEE MTT-S Int. Microwave Symp. Digest, 1996, pp. 1145-1148.
[87] Caulton, M., “Lumped Elements in Microwave Integrated Circuits,” in Advances in
Microwaves, Vol. 8, L. Young and H . Sobol, eds., New York: Academic Press, 1974.
[88] Aitchison, C. S., et al., “Lumped-Circuit Elements at Microwave Frequencies,” IEEE
Trans. Microwave Theory & Tech., Vol. MTT-19, December 1971, pp. 928-937.
[89] Daly, D . A., et al., “Lumped Elements in Microwave Integrated Circuits,” IEEE Trans.
Microwave Theory dr Tech., Vol. MTT-15, December 1967, pp. 713-721.
[90] Pucel, R. Q., “Design Consideration for Monolithic Microwave Circuits,” IEEE Trans.
Microwave Theory & Tech., Vol. MTT-29, No. 6, June 1981, pp. 513-534.
[91] Chi, C-Y., and G . Rebeiz, “Planar Microwave and Millimeter-Wave Lumped Elements
and Coupled-Line Filters Using Micro-Machining Techniques,” IEEE Trans. Microwave
Theory & Tech., Vol. MTT-43, April 1995, pp. 730-738.
[92] Kajfez, D., and P. Guillon, Dielectric Resonators, Norwood, MA: Artech House, 1986.
[93] Cohn, S. B., “Microwave Bandpass Filters Containing High-Q Dielectric Resonators,”
IEEE Trans. Microwave Theory & Tech., Vol. MTT-16, April 1968, pp. 218-227.
[94] Harrison, W. H., “A Miniature High-Q Bandpass Filter Employing Dielectric Resona-
tors,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-16, April 1968, pp. 210-218.
[95] ESA/ESTEC Workshop, Microwave Filters for Space Applications, June 7-8, 1990,
Noordwijk, The Netherlands.
[96] Wakino, K., T. Nishikawa, and Y. Ishikawa, “Miniaturization Techniques of Dielectric
Resonator Filters for Mobile Communications,” IEEE Trans. Microwave Theory & Tech.,
Vol. MTT-42, July 1994, pp. 1295-1300.
[97] NTK Production Guide, NTK Piezoelectric Ceramics Div. Eng., Japan, 1997.
[98] Tekelec Production Guide, Tekelec Components, France, 1997.
[99] Flory, A., and R. C. Taber, “High Performance Distributed Bragg Reflector Microwave
Resonator,” IEEE Trans. Ultrasonics, Feroelectronics, and Freq. Control, Vol. 44, March
1997, pp. 486-495.
[100] Alpha/Trans-Tech Production Guide, Alpha/Trans-Tech, Inc, USA, 1997.
[101] Murata Production Guide, Murata Manufacturing Co., Ltd., Japan, 1997.
38 Generalized Filter Design by Computer Optimization

Selected Bibliography
Gardiol, F. H., “Higher-Order Modes in Dielectrically Loaded Rectangular Waveguides,” IEEE
Trans. Microwave Theory & Tech., Vol. MTT-16, November 1968, pp. 919-924.
Vartanian, P. H., W. P. Ayres, and A. L. Helgesson, “Propagation in Dielectric Slab Loaded
Rectangular Waveguide,” IRE Trans. Microwave Theory & Tech., Vol. MTT-6, April 1958,
pp. 215-222.
Characterization of Discontinuities

This chapter discusses the electromagnetic field analysis of discontinuities in

E-plane, ridged waveguide, and CPW filters. An electromagnetic field analysis
predicts the electrical behavior of discontinuities with a specified geometry by
solving Maxwell’s equations for the appropriate boundary conditions. The
geometry of the commonly used discontinuity employed in conventional (rect-
angular waveguide) E-plane filters such as the metal septum in rectangular
waveguide is shown in Figure 3.1. Some of the commonly used discontinuities
in ridged waveguide E-plane filters such as the metal septum in ridged wave-
guide, the metal septum between two ridged waveguides with different gaps
and the metal septum between rectangular and ridged waveguide are shown
in Figure 3.2(a) . In this book, only a finite length septum is considered. The
septum is assumed to be located at the center of the guide and lossless. Figure
3.2(b) shows the geometry of the CPW coupling gap, a commonly used
discontinuity employed in CPW filters.
Electromagnetic analysis of the septum in rectangular waveguide is rela-
tively simple, while the analysis of the septum in ridged waveguide is more

Figure 3.1 Discontinuity step rectangular waveguide to bifurcated rectangular waveguide.

39
40 Generalized Filter Design by Computer Optimization

Figure 3.2 (a) Step-ridged waveguide to bifurcated rectangular waveguide discontinuity and
(b) geometry of the coplanar waveguide coupling gap.

complicated. The major complexity of this type of discontinuity is in determin-

ing the field distribution of the eigenmodes in the ridged waveguide region.
To carry out accurate design of E-plane filters, it is important to characterize
the discontinuities that are required to be incorporated in them.
E-plane discontinuities have been studied using different rigorous meth-
ods, including the variational method [1-3], general purpose methods [4,5],
and the mode-matching method [6-9] . A good choice of the numerical methods
is the product of tradeoffs among accuracy, speed, storage requirement, versatil-
ity, and so on, and is quite structure dependent. General purpose methods,
 Characterization of Discontinuities 41

Figure 3.2 (continued).

for instance the finite-element method [4,5], can be applied to problems with
nearly arbitrary geometries or structures of complicated shape. Postprocessing
must be done to obtain the S parameters. A serious drawback, particularly
with the variational method, is that the formulation for a given structure may
call for considerable mathematical ability on the part of the user.
However, among these numerical methods, the mode-matching, or
modal, analysis originally presented by Wexler [7] proves faster and more
efficient for electromagnetic analysis of E-plane discontinuities [10] . The main
advantage of this method is its easy implementation on a modern computer
system. Mode matching is one of the oldest, most popular, and most frequently
employed rigorous full-wave methods to solve the scattering problem due to
various discontinuities in waveguide [6,7,11].
The first step in the mode-matching procedure entails the expansion of
unknown fields in the individual regions in terms of their respective normal
modes. Because the functional form of the normal modes is known, the problem
reduces to that of determining the set of modal coefficients (amplitudes)
associated with the field expansions in various regions. That procedure, in
conjunction with the orthogonality property of the normal modes, leads to a
set of linear simultaneous equations for the unknown modal coefficients. The
equations then are set up by enforcing the continuity condition for the tangential
42 Generalized Filter Design by Computer Optimization

electric and magnetic fields. The introduction of a matrix representation and

the particular choice of the family of modes helps simplify the mathematical
efforts and reduce computational time. This method has several different possi-
ble formulations, all theoretically equivalent (mathematically valid), although
they may be different numerically because of the different manipulations of
the basic equations and their individual implementation in a software package.
Because of the singular behavior of the magnetic field at the edges of the septa,
a large number of modes need to be included in the field expansions to ensure
good convergence.
The application of the mode-matching method to the electromagnetic
analysis of discontinuities in conventional E-plane filters (metal septum in
rectangular waveguide) is discussed in Section 3.1. Simplifications that arise
when evanescent modes between adjacent discontinuities do not interact and
the convergence problem also are covered. Section 3.2 analyzes discontinuities
in ridged waveguide E-plane filters, such as the metal septum between two
ridged waveguides with identical and different gaps, and metal septa between
rectangular and ridged waveguides. That same section also examines ridged
waveguide, the convergence problem, and results of the analysis of the septum
in ridged waveguide associated with the numerical mode-matching method.
Section 3.3 introduces electromagnetic analysis of the CPW discontinuities
such as CPW coupling gap [see Figure 3.2(b)] and edge-coupled CPW lines
by Em™ software [12].

3.1 Analysis of Filter Discontinuities by the Mode-Matching

Method

3.1.1 Discontinuities in Conventional E-Plane Filters

This section applies the mode-matching method to the analysis of discontinu-

ities in conventional E-plane filters such as the metal septum in rectangular
waveguide. The septum is assumed to be of finite thickness, of finite length,
and lossless. The geometry of the problem and the coordinate system used are
shown in Figure 3.3. The septum is assumed to be centered in the waveguide
with the energy propagating in the z direction. The approach used in this
section is based on the modal analysis method of Wexler [7].

3.1.1.1 Metal Septum in Rectangular Waveguide

To apply the modal analysis method, we consider the standard waveguide,
denoted as region a in Figure 3.3, while the reduced waveguide is denoted as
region b. In view of the longitudinal symmetry, we can simplify the problem,
 Characterization of Discontinuities 43

plane of symmetry

(a-t)⁄2

region a

(a-t)⁄2 ∣

magnetic
or electric wall

Figure 3.3 The geometry of the metal septum in rectangular waveguide.

44 Generalized Filter Design by Computer Optimization

considering even/odd excitation corresponding to placing a magnetic/electric

wall at z - LU. Therefore, a symmetrical septum can be analyzed by considering
the left half of the structure, with the symmetry plane representing an open
circuit for even excitation and a short circuit for odd excitation. When applying
the mode-matching method, it is important to know which components of
the electromagnetic field are involved. They are determined by the field compo-
nents of the exciting mode (e.g., the dominant waveguide mode) plus those
added by all discontinuities. Because usually all components tangential to the
discontinuity plane need to be matched, the number of field components
directly affects the matrix sizes and the CPU time required for the analysis of
a discontinuity.
At the discontinuity plane, a dominant mode TEιo excitation will not
introduce field components other than those of the incident wave such as Ey ,
H x , H z , because the discontinuity is uniform along thej axis. That discontinuity
can be characterized by TE w o modes, where m has all possible positive odd
integers, 1, 3, 5, . . . in standard waveguide, while m can take on all positive
integers, m = 1, 2, 3, . . . , in the reduced waveguide region. Due to the singular
behavior of the magnetic field at the edge of the metal septum in rectangular
waveguide, numerical problems can arise.
Consider a single propagating mode with mode coefficient a↑ emanating
from a matched source in guide a, and impinging on reduced waveguide b at
junction z - 0. At junction z = 0, let p be the reflection coefficient of this
mode, and let z⅛ where i = 2, 3, . . . be the mode coefficients of the scattered
modes. The total transverse fields that are just to the left of junction z = 0 (in
region d} can be expressed in terms of those modes to yield
M
a e a e
Eta = (1 + P ∖ a∖ + ∑ i ai (3. la)
z=2

M
H ta = (1 ~ p}a x h ax - ∑ a i h ai (3. lb)
z=2

Just to the right of junction z = 0 (in region A), the total transverse fields can
be expressed as
TV ⁄V
e + s
Etb ~ bj Jb b b j (3.2a)
j=1 Ď k=1 ⁄

TV N

Htb = y bl h b j - y Sjk h bk j (3.2b)

y=1 Ď k-∖ ⁄
 Characterization of Discontinuities 45

Expressions for the transverse fields in region a are

ea ι = sin/- — ) (3∙3a)
∖ a j

h a i = Y ai s i n ( - —) (3.3b)
∖ a )

The wave admittance of the zth mode is defined by

ι
⅛-4M W V -(⅛4
∖2 a J <3 · 4 >

where Λq is the wavelength in free space at f), and p = 2z - 1. Modes are

numbered consecutively, that is, i = 1, 2, 3, . . . , M.
Expressions for the transverse fields in region b are

e
(2qττx∖
bjj = sin∖a-------
—t ⁄ (3.5a)

hbj = Ybj sin⁄fff∖ (3.5b)

Equation (3.2) holds in the range 0 < x < (a - z) ⁄ 2 and is zero across the vane.
The wave admittance of the /th mode is

2
⅛ = √⅛ √' - B) ™
where q = 1, 2, . . . , N One should substitute (1 — xld) for (xla) when
(a + i)H < x < a.
To solve for the unknown parameters p, <¾ and bp the boundary condi-
tions at the discontinuity must be satisfied. Boundary conditions to be satisfied
at the discontinuity are as follows: Continuity of the total transverse electric
and magnetic fields across the aperture and the zero tangential electric field at
the surface that contains the septum. We presume that sρζ due to the second
discontinuity is known; otherwise, it must be evaluated by solving the second
discontinuity first. That leads to tedious computations, which can be avoided
46 Generalized Filter Design by Computer Optimization

for this problem because the discontinuity is symmetrical about the z = Z/2
plane. By using symmetrical and asymmetrical excitations, we can find the
normalized reactive impedances (⅛ ¾) without a complete set of scattering
coefficients. Symmetrical excitation of the septum is obtained by having two
modes in the waveguide, one traveling in the +z direction and the other in
the — z direction, such that the E fields are in phase in the z = LU plane;
antisymmetrical excitation is obtained if the modes are 180 degrees out of
phase. For symmetrical excitation, an open circuit appears at the symmetry
plane; antisymmetrical excitation produces a short circuit. With the open circuit
at the central plane (Figure 3.4), with Sjk = +1, the normalized reactive even-
mode impedance z e is given by

+
Pe
Ze ~ i (3.7)
1 - Pe

With the short circuit at the central plane (see Figure 3.4), with Sj = —1, the
normalized reactive odd-mode impedance z 0 is given by

1 + Po
Z (3.8)
° ~ ∖1 - Po
p

Mathematical details of the derivations’ normalized reactive impedances

(z e, z 0 ) are given in Appendix 3A.

(a-t)⁄2

(a-t)⁄2

z—0

Figure 3.4 Symmetrical finite length metal septum in rectangular waveguide.

 Characterization of Discontinuities 47

3.1.1.2 Simplification When Evanescent Modes Between Adjacent

Discontinuities Do Not Interact
In conventional E-plane filters, if the length, 4 of the resonator is comparable
to the internal waveguide width, a, the effect of higher-order mode coupling
along the filter section (resonator) can be neglected. Under those conditions,
only the dominant mode is responsible for the coupling between septa (disconti-
nuities). In the case of longitudinally symmetric structures, such as the conven-
tional E-plane bandpass filters under consideration, we can significantly simplify
the analysis. In that case, the structure is analyzed by placing electric and
magnetic walls at the plane of symmetry. By using Bartlett’s bisection theorem
[13], the insertion loss (Z√) and return loss (Z, ) of a symmetrical E-plane filter
can be expressed as

1
z, . 20log,0 < Λ⅛< .⅛ (3.9)

and

⅛-2θlog, 0 θ (3.10)

where jz e ( β } is the normalized input impedance of the two identical 1-ports

formed by placing a magnetic (electric) wall at the plane of symmetry, z e
can be calculated by transforming an open (short) circuit placed at the plane
of symmetry through the filter sections (resonators and E-plane septa) located
to the left of the plane of symmetry. Each E-plane septum is itself symmetrical
and can be electrically represented by normalized even- and odd-mode imped-
ances

% = j(×si + 2χ pz∙) (3.11)

z
oi = j x si (3.12)

where x5ι and x pz∙ are the normalized reactances of a symmetrical normalized
equivalent T circuit of the metal septum in waveguide (Figure 3.5).
For a normalized reactive impedance jz y an E-plane septum performs the
normalized impedance transformation jz => j¾ p where ¾ 2 is given by

z{zei ÷ z 0 j) ÷ 2z ei z oi
(3.13)
2z ÷ (% + z 0 -)
48 Generalized Filter Design by Computer Optimization

jXs jXs

jXp

Figure 3.5 A symmetrical normalized equivalent T circuit of the metal septum.

A resonator section, that is, a length of guide, performs the normalized

impedance transformation jz => jz ιn > where z ιn is given by

z ÷ tan β I
z
in = (3.14)
1 - z tan β I

where β(= 2ττlλ is the propagation constant, and I is the length of the
resonator. By applying (3.13) and (3.14), it is possible to calculate z e and z 0
starting at the center of the filter and working outward. The process involves
no matrix manipulation and uses only real arithmetic.

3.1.2 Discontinuities in Ridged Waveguide Filters

This section applies the mode-matching method to the analysis of the disconti-
nuities in ridged waveguide E-plane filters. At the discontinuity plane of the
conventional E-plane filters discussed so far, a dominant mode TEjq excitation
does not introduce field components other than those of the incident wave,
such as Ey , H x , H z , In contrast, in the case of discontinuities in E-plane ridged
waveguide filters, at the discontinuity plane shown in Figure 3.2, a dominant
mode TEiq excitation introduces field components other than those of the
incident wave, such as Ex , Ey , H x , H y , H z . These types of discontinuities can
be characterized by TE ww modes only where m has all odd possible positive
values, 1, 3, 5, . . . , and n all even possible positive values, 0, 2, 4, 6, . . . in
the ridged waveguide region (region ), while m and n have all possible positive
values, 1, 2, 3, . . . in the reduced waveguide region (region b) and need to
be considered.
First, a number of TE modes will be evaluated in the ridged waveguide
sections as a function of the geometrical parameters of the guide. This is
performed by applying the mode-matching method along the transversal direc-
 Characterization of Discontinuities 49

tion. The cutoff frequencies of ridged waveguide modes are found by an iterative
algorithm that searches for the zeros of a function corresponding to transverse
resonance conditions [14] . Then the scattering parameters of several discontinu-
ities in ridged waveguide filters are determined.

3.1.2.1 Analysis of Ridged Waveguide

Figure 3.6 illustrates the cross-sectional shape of the ridged waveguide with
the coordinates, where t denotes the thickness of the ridge, s the gap in the
ridge, 1 the region of 0 < x < r⁄2, and 2 the region of tH < x < all. The
width and the height of the waveguide cross-section are given by a and b y
respectively. Considering the odd TE modes, whose symmetry plane has been
assumed to be the y coordinate axis (x = 0) in this analysis, the x = 0 plane
can be regarded as a magnetic wall. In the case of even modes, a similar analysis
can be accomplished, regarding the x = 0 plane as an electric wall. Therefore,
attention will be focused on TE modes (fields) with a magnetic symmetry
plane about the y coordinate axis (x = 0), and only the fields in regions 1 and
2 need to be characterized.
The TE fields in ridged waveguide can be derived from the magnetic
Hertzian potential Q, which satisfies the Helmholtz equation in the transverse
plane:

magnetic
symmetry
wall

Figure 3.6 Geometry of ridged waveguide.

50 Generalized Filter Design by Computer Optimization

Δ⅛ + *⅛=O (3.15)

with

⅛ = ω 2 μ 0 e0 - β 2 (3.16)

where β is the propagation constant, ω is the angular frequency, ⁄zθ is the

permeability of free space, and 6θ is the permittivity of free space. The scalar
potential function is related to the electric and magnetic fields [15] by the
equations

Et = ez · Δ,Q (3.17)

7⁄ f = - - - Δ fQ (3.18)

where ez is a unit vector of the direction. Applying the TE boundary conditions,

we find the following expressions for Q.
In region 1 (0 < x < /72),
N

Qi =
∑ 7
7iw sm(k x∖ n x) cos k
y∖n ( y - (3.19)
w=0

where

Π7Γ
kχ∖n ~ y∣k2 — ky∖ n k
y = ~
(3.20)

In region 2 (z72 < x < all),

zλ
⁄ ∣1
M
0.2 = 7,V2m sin k⅛2wl
r
⁄
x
a r, 2I (3.21)
x. ~ 2 COS
m=0

where

=
mπ
kχlm ~ ∖∣l⅛ — ky2m kylm (3.22)

Performing the operations in (3.17) and (3.18) on (3.19) and (3.21), we

find
 Characterization of Discontinuities 51

N
5
E t∖ = ~X V∖n ky χ n sin(£x i„x) sin
2
n=t}

+ k χ χ n CQs{k χ χ n X) COS e
y
(3.23)

in region 1 and
M

E 2 ~ ~~ Vim j kylm c
θ s sin , j⁄2wl⁄y _ 2 )1 x e

w=0 I

~ k x2m sin COS kylm{ J (3.24)

in region 2.
Let Ega p{y) = Ey {t!2, y) be the y component of the electric field at the
boundary between regions 1 and 2. Because Ega p(y) must be continuous across
this interface,

cos nττ⁄ s
Egap{y) = -'∑V↑nkχln ( V ) cos VP z
2
n=0 ∖ ⁄
M
⁄ t —a mπ[ b∖
Vlm xlm s n
¾2wl - ~f~ COS (3.25)
w=0 ~τγ~η

Equation (3.26) yields the Fourier coefficients η χ n and Vlm> that is,
√2

Ega pfy) cos -

£ nV ∖n s kχ∖n c
θ s

√2
mπ( E , ( t
~ a
⅛ z e
m' fllml 3 k x 2 m sin 2
W"2 = Rχlm∖

(3.27)

in which e s is the Neumann number; (β j = 1 if 5 = 0; otherwise, 6 s = 2). Note

that in (3.27) the integrations embrace only the interval -s⁄2 < y < s/2 because
Ega p(y) = 0 for s/2 < y < b!2. Continuity of the scalar potential in conjunction
with (3.26) and (3.27) results in the integral equation
52 Generalized Filter Design by Computer Optimization

d- t j/2
M ∞t k x2m
' ∖l Γ /
COS ky2τ∏∖ Egap (y) cos ky2m (y - 4
w=0 kχlr∏e nJ?

⁄t √2
xXn
' \2 S
COS J ⅞√9 ,) cos k y λ n y 2
dy

-s⁄2
(3.28)

The eigenvalues kc for the TE mode can be extracted from (3.28).

However, the procedure is complicated by the fact that concomitantly one
must find E a p{y)i the y component of the TE electric field at the region
1 /region 2 boundary. An exact analytical solution is not known. However, a
numerical solution may be obtained. One technique for doing that is the Ritz-
Galerkin method [14]. The Ritz-Galerkin method consists of expanding the
unknown boundary field E a p{y) by a suitably chosen family of functions; it
then requires that the resultant equation be orthogonal to each expansion
function. A matrix equation is obtained and may be solved by suitably developed
matrix theory. In the case of ridged waveguide, because the edge of the ridge has
a right angle, the y component of an electric field near the edge is approximately
proportional to ∆j⁄ [16], where ∆y represents the distance between the
edge and an observational point. Therefore, by taking into account the edge
conditions mentioned by Mittra [11], it is convenient to expand E a p{y) in
the eigenfunctions of region 1:

2
i 2
-y (3.29)
⁄=0

Substitution of (3.29) into (3.28) and after integration, multiplying

⁄ ⁄ 7ΓĎ Ď
through by cosl I 6— 1((√2) - y) j, followed by integration over [-s⁄2, s⁄2],

yields the matrix equation

[M¼)] ∙ [C] = [0] (3.30)

with the matrix elements H l j given by

 Characterization of Discontinuities 53

tan ⁄ , t ⁄ d
~ t
∣ kx λ i - M cot
x2wl 2
-
Hi]{⅛c) =
ij i ' s
7
κ
x∖i
X Pi m Pjm ----- (3.31)
m=Q

where
s/2
2 1/3
f iτr Λ 2 mττ⁄ b
COS - - - y COS (3.32)
2 ~ W"2

√2
2 -1/3
ιπ Λ 2 mπ⁄ b∖
Pjm =
1 COS --- - y cos y (3.33)
S 2/ ~T∖ ~ι)
-s/2

and

[C] = [C0 C 1 C 2 . . . Q ] ' (3.34)

In (3.31), ∆ zy is the Kronecker delta (∆ zy = 1 if i = j∖ Δ⅛ = 0 if i_j). Equation

(3.30) is a generalized matrix eigenvalue problem. The TE eigenvalues kc are
solutions of the nonlinear equation

det∣Z⁄(¼)l = 0 (3.35)

The convergence of the dominant as well as higher order eigenvalues is

discussed in subsection 3. 2. 1.1. The vector [C], which is a solution of (3.30)
for a particular kc , is the associated eigenvector.
3.1.2.2 Metal Septum in Rectangular Waveguide Between Two
Different Ridged Waveguides
This section analyzes the metal septum in rectangular waveguide between two
different ridged waveguide by the mode-matching method. The original mode-
matching method formulation [5] is adapted to the case of the metal septum
in ridged waveguide. The metal septum is assumed to be located at the center
of the guide to reduce the excitation of the even-order modes, and the structure
is assumed to be lossless, as shown in Figure 3.7.
Basically, the analysis consists of two parts. The first part deals with a
scattering problem in a ridged waveguide bifurcated by a septum of finite
thickness. In the second part, two junctions, as in the first part, are joined
back-to-back to form a finite-width septum. The process takes into account
the interaction between junctions of both the dominant mode and higher order
modes.
54 Generalized Filter Design by Computer Optimization

5
2

ridged ridged
a waveguide 1 ∣metal septum ⅝ waveguide 2

z=O z=L

Figure 3.7 Metal septum inserted in rectangular waveguide between two different ridged
waveguides.

Ridged Waveguide: Metal Septum Step

Consider the step-ridged waveguide to bifurcated rectangular waveguide discon-
tinuity shown in Figure 3.8, where the septum is located at the center along
the E-plane. Because of symmetry, we can place a magnetic wall at the center
along the z-axis and consider only one-half of the structure. The problem
becomes a transverse step junction between ridged waveguide and rectangular
waveguide with widths zz/2 and (α - ⁄)⁄2, respectively, as shown in Figure
3.8(b). Assuming excitation by the TEjq dominant mode, the only types of
higher order modes that are excited (generated) at the step discontinuity are
TE wλ2 modes, where m has all odd possible positive values, 1, 3, 5, . . . ,
n = 0, 2, 4, . . . in the ridged waveguide region, and m = 1, 2, 3, . . . and
n = 0, 1, 2, 3, . . . in the reduced waveguide region. Due to the singular
 a
(a)

region a
I
∣ re gion b (a-t)⁄2<
I
ridged
waveguide

b
b

z=0

magnetic wall
(b)
ridged
waveguide ∣metal septum t/2
÷ + V
a
a/2 (a-t)⁄2

z=0

Figure 3.8 Configuration for the rigorous field theory treatment: (a) forward and backward
waves at the discontinuity step and (b) ridged waveguide to bifurcated rectangular
waveguide step with magnetic wall symmetry.
56 Generalized Filter Design by Computer Optimization

behavior of the fields at the edges of the bifurcated rectangular waveguide and
the ridges in ridged waveguide, numerical problems can arise.
The mode-matching procedure begins with expanding the tangential
components of the electric and magnetic fields at the discontinuity in terms
of the normal modes in each region. The amplitudes of the normal modes are
chosen such that the boundary conditions are satisfied at the discontinuity.
With reference to Figure 3.7(b), the total transverse fields in each region
at the discontinuity can be written [6] as:
region a{z < 0)
M
Eta. ~ Φ am[a m exp(-⁄β am⅜ + a
m exp(+jβ (3.36)
m- 1
M
H t a = Σ YamΦam⅛m exp(-jβ am z) - Λm Cxp(+jβ a m Z)} (3.37)
m- 1

where

zo2 ω 2z 6 /2
Pam = Z '0 0 “ k c (3.38)

and

y β am
* am ~ , ,,, (3.39)
ωμ 0

The normalized mode functions operating in (3.36) and (3.37) can be

obtained from the eigenmode analysis of the ridged waveguide. In the case of
the reduced waveguide, in region b {z > 0, 0 < x(a — i)H∖ the fields are given
by
K
E t b = ∑ΦbpWp exp(-jβι, p z) + b p exp(+jβ b p z)]} (3.40)
?=1
K
+
Htb = y Ybp Φ bp {∖Φp e×p(-jβb p z) ⅞ exp(+yβ z)]} (3.41)
⁄>=1

where

z?2 ω 2 k, 2
βbp = ~ cbp (3.42)
 Characterization of Discontinuities 57

k2 (2mπ∖ ⁄ nπ∖
∙Ď~ t) l l τ ) (3.43)

r
(3.44)
ωμo

and

ω⁄z 0 nπ ⁄2mττx∖ . ⁄ nπy

Φ bp ~ j p κ cos∣-------∣ sin —τ~
k∖cbp ~b ∖a - tJ y b
ωμo ⁄2mτr∖ , ⁄2mπx∖ ⁄ nπy∖
— p — ∣ ------ sm∣------- cos —~r~ (3.45)
κk∣ ∖d - ∖Ďa - t ⁄) ∖b
cbp L Ď t ⁄ Ď ⁄

with

kcbt>y∣2ε rrfi n
Ap = ∣ — ■■■ = (3.46)
∖{ a - i ) b ω μ βp

where the indices p are related to the waveguide modes m>n by rearranging
them with respect to increasing cutoff frequencies; kc is a ridged waveguide
eigenvalue that is related to the cutoff frequency, which can be easily calculated
by solving the related eigenvalue problem; φ arn , and φfy are the normalized
mode functions in regions a and b, with phase propagation constants bam and
bty respectively; Y arn and Y⅛ are wave admittances in regions a and b·, and
a↑n and b~p are the amplitudes of the waves traveling in the positive (forward)
z-direction for regions a and b, while d m and b are the amplitudes of the
waves traveling in the negative (backward) -direction in regions a and b,
respectively. It should be noted that according to the voltages and currents on
transmission lines, the electric and magnetic field components are given as the
sum and the difference of the forward and backward traveling waves, respec-
tively. The boundary conditions necessary for matching the field components
at the discontinuity (z = 0) are:

E
a —t a ÷t
ta = 0 ~y ≤ × ≤ -y

E
a- t
ta ~ E tb 0 ≤ X≤ “

a —t
H ta = H tb 0 ≤ x≤ — (3.47)
58 Generalized Filter Design by Computer Optimization

and from (3.36), (3.37), (3.40), and (3.41), it follows that

K
M a-t
+ + x
(⅛ τn)Φam ~ ( p p)Φbp θ — — (3.48)
m=1 p-1
K
M a- t
χ (z7w drr YamΦ am ~ y (bp - b p )YbpΦbp θ ≤ × ≤ ~f~ (3.49)
m-∖ p-∖

By using the property of mode orthogonality, we can derive a system of

equations involving the unknown coefficients from (3.48) and (3.49). The
next step is to eliminate the x dependence in (3.48) and (3.49). Following
Wexler’s procedure, we enforce continuity of the transverse electromagnetic
field at junction {z = 0) by (3.48) and (3.49), taking the cross-product with
φ am , and integrating with respect to x and y from 0 to (a - i)!1 and 0 to b,
respectively, we find
K
a + a
m ~m = ∑ H m p(bp + bp) m = ∖, 2 , . . . , M (3.50)
⁄=1
K
Y am m + <4) = ∑ Ybm H m p(bp + b~p) m = l , 2 , . . . , M (3.51)

This is a first set of linear equations, where H rnn is a coupling integral

defined at the plane of the discontinuity as
(a-t)H b

Hm p = Φ amΦ bp d χ ⅛ (3.52)
0 0

A second set of linear equations can be obtained by multiplying (3.5a)

and (3.5b) by and integrating with respect to x and y from 0 to
,
(a - i)∣l and 0 to b, respectively. This yields
M
∑H-pm⅛p + ⅛) = b m+ + b~m m = 1,2, . . . , K (3.53)
?=1
M
Yn p mY a p{ap - dp ) = Ybm (b*m - b~m ) m = 1, 2, . . . , K (3.54)
⁄>=ι
 Characterization of Discontinuities 59

From this point, several approaches are possible. It is easier to handle

(3.50) to (3.54) simultaneously in matrix form. The matrix form of those
equations is written as:

E t ·. a + a = EI(JE + b~) (3.55)

H t ·. a - a = Y ~a λ HYh( b+ - b~} (3.56)

Et ·. H t(a+ + a ) = b + + b~ (3.57)

H t ·. Y ~b H∖a - a ) = b* - b~ (3.58)

where / / i s a matrix of size M by ⁄fwith element H rn p as defined by (3.52).

The superscript t denotes the transpose operation, and Yff = a,b) are diagonal
matrices with diagonal elements Y am and The coefficient vectors are defined
as
-
+
a
∖ ~b~x ~
a +
+ 2
a = b~ = (3.59)

_a M_ _b~K_

'b↑'
a
2
a = b* = (3.60)
a
M bκ

a and ⁄~ are column vectors of the excitation terms, and a and ⁄ + are
column vectors of unknown modal coefficients. Two independent vectors are
required to solve for two unknown vectors. Hence, for four pairs of equations
[i.e., (3.55) and (3.56), (3.56) and (3.57), (3.57) and (3.58), and (3.58) and
(3.55)], substituting one equation into the other in the same pair, we have
eight ways of obtaining a solution for a and ⁄ - .
The final solutions are expressed by scattering parameters defined by

(3.61)
60 Generalized Filter Design by Computer Optimization

where S zy (z = 1 ,2; y = 1,2) are the generalized scattering parameters (submatrices)

representing the amplitude of the scattered field at port i due to the unit
incident field at port j. The detailed expressions for scattering parameters are
included in Appendix 3B. In this research work, the formulation [see (3.62)
to (3.65)] was employed because of the best satisfaction of the boundary
conditions [7], which is given by

5 2 2 = [⁄ + Y~b l H t Ya H ] ~ ' [ I - Y ~b λ H t Y a H] (3.62)

⅛ = 2[I + Y ~b λ H t Ya HV λ Y ~b H t Ya (3.63)

S 1 2 = H[I ÷ S2 2 ] (3.64)

5n = HS2x - I (3.65)

Finite-Length Septum
With the knowledge of the scattering parameters for a single discontinuity
(junction), the overall composite scattering matrix can be obtained by a network
constitution in terms of the generalized scattering matrices [6] . Let S 1 and S 2
represent the scattering matrices for the isolated junctions at z = 0 and
z = £, respectively (Figure 3.9).
The combination of matrices S 1 , T, and S 2 results in the composite
scattering matrix 5 for the overall structure given by (3.66a) to (3.66d) below
for the septum of length Z,, with reference planes located at z = 0 and
z = L.

5ιι = S ι ι ÷ 5i 2 7 ' 5 i i ( ⁄ - T s∖ 2 T s∖ x )~ λ T S 2xx (3.66a)

λ
5i 2 = 5} 2 ( ⁄ - T S 1
22 T S n Y λTS∖2 (3.66b)

2 λ
52 i = S 2 ↑(I- TS n T S ' 22 Γ i T S 21 (3.66c)

5 2 2 = S 22 2 ÷ S22λ T S22
x
{I - T S 2λ ι T 5 2 2 ) 1
T Sj 2 (3.66d)

In (3.66a) to (3.66d), I is the unity matrix, and the transmission matrix

T represents the wave propagating (for propagating modes) or attenuating (for
evanescent modes) for a distance of L in guide region b. This is a diagonal
matrix whose diagonal elements are

T{n,n) = ~ jβ n = 1, 2, 3, . . . (3.67)
 Characterization of Discontinuities 61

ridged ⅝metal septum ridged

waveguide 1 waveguide 2

—I------------------------μ-
z-0 z=L

(a)

(b)

(c)

Figure 3.9 (a) Septum of length L; (b) scattering network representation; and (c) the composite
scattering matrix after network combination.

So that the mode-matching formulation can be employed, it must be

possible to split the structure to be analyzed into well-defined regions for which
the mode functions are known. For a given discontinuity, all the mode types
that can be excited by it must be included so the field continuity equations
can be satisfied. The program structure is shown in Figure 3.10. It can be seen
that once written, the program can be easily modified for the analysis of
different structures. All that is necessary is that a subroutine be written for
computation of the required mode functions that returns the normalization
factors and propagation constants for the required numbers and types of modes.
The array of coupling integrals H also must be computed. The remaining parts
of the program are unchanged.
62 Generalized Filter Design by Computer Optimization

INPUT DATA
(dimensions o f the structure,
operation frequency, etc.)

COMPUTE
Ya, Yb and H

COMPUTE
S and T matrices

OUTPUT DATA

Figure 3.10 The program structure.

3.1.2.3 Metal Septum in Rectangular Waveguide Between Rectangular

and Ridged Waveguides
Using the procedure of field matching at z = 0 and z = Z, the elements of the
S matrix of the structure shown in Figure 3.11 are given by (3.62) to (3.65).
Because the characteristics of those two discontinuities are essentially the same
except for the different waveguide eigenvalues (⅛f , s), we obtain the scattering
parameters for both matrices from (3.62) to (3.65). From (3.66a) to (3.66d),
we obtain the element values for the overall structure.

3.1.2.4 Metal Septum in Ridged Waveguide

Using the procedure of field matching at z = 0 and z = Z, the elements of the
S matrix of the structure shown in Figure 3.12 are given by (3.62) to (3.65).
Because the characteristics of those two discontinuities are essentially the same
except for the different ridged waveguide eigenvalues ( ⁄s), we obtain the
scattering parameters for both matrices from (3.62) to (3.65). From (3.66a)
to (3.66d), we obtain the element values for the overall structure.

3.2 Convergence

For the electromagnetic analysis of E-plane septa in a rectangular waveguide,

the mode-matching method is a popular choice. However, due to the singular
 Characterization of Discontinuities 63

z=0 7 =L

a
ridged
waveguide EetaUeotuπ rectangular
waveguide

z=0 z=L

Figure 3.11 Metal septum between rectangular and ridged waveguides.

behavior of the magnetic field at the edges of the septa in a rectangular

waveguide, a large number of modes need to be included in the field expansions
to ensure good convergence. That point is illustrated by Tables 3.1 and 3.2.
Table 3.1 shows the slow convergence of the normalized even- and odd-mode
impedances of an E-plane septum with a length of 0.8∖⅛Galp and a thickness
of 0.01094 at a center frequency of 1.4478 f c { f c = TEjq cutoff frequency in
rectangular waveguide) as the number of the modes used in the field expansions
is increased. Table 3.2 shows the convergence of the insertion loss of a symmetri-
cal X-band fifth-degree E-plane bandpass filter with an insert thickness of
0.01094 , septa lengths given by d↑ = 0.1528 ⁄τr, ⅛ = 0.8146tf⁄ττ,
⅛ = 0.9985 ⁄τr, and resonator lengths given by l↑ = 2,2Q55al ιτ y =
2.2568zz⁄ π, and /3 = 2.2603 ⁄ 7Γ, atacenter frequency of 1.4453 fc , where fc is the
TE 1 o cutoff frequency in rectangular waveguide. The calculation of this insertion
loss requires the evaluation of the normalized even- and odd-mode impedances of
64 Generalized Filter Design by Computer Optimization

z=O z =L

a ridged ridged
waveguide 1 waveguide 1

z=O z=L

Figure 3.12 Metal septum inserted in rectangular waveguide between two equal ridged
waveguides.

three E-plane septa. In all three cases, the same number of modes is used. The calcu-
lations were performed on a SUN workstation (SPARC 10). About 0 . 5 minute of
CPU time is required for the calculation of the insertion loss at one frequency using
1 0 0 modes.
Tables 3.3 and 3 . 4 show the slow convergence of the dominant and the
higher order eigenvalues of a double-ridged waveguide with b/a = 0.80,
si b = 0.275, and tla = 0.20, as the number of terms used in the field expansions in
region 1 (TV), and region 2 (Λ⁄) is varied. Note that convergence of higher order
eigenvalues as a function of order of the number of terms is achieved even more rap-
idly than the dominant eigenvalue. That is generally the case with higher order
eigenvalues because the deviation from the ridgeless case becomes smaller. In the
actual computations of (3.47), the summations over ΛfandΛ⁄are terminated when
 Characterization of Discontinuities 65

Table 3.1
Convergence of Normalized Even- a n d O d d - M o d e I m p e d a n c e s
S e p t u m length = 0.8146a⁄τr
S e p t u m t h i c k n e s s = 0.01094a
Frequency = 1.4478 f c , f c - T∈iq cutoff f r e q u e n c y i n r e c t a n g u l a r w a v e g u i d e

Normalized Even-Mode Normalized Odd-Mode

Number of Modes Impedance Impedance

20 0.5923 0.2662
40 0.5909 0.2657
60 0.5919 0.2661
80 0.5923 0.2663
100 0.5921 0.2662
120 0.5919 0.2661
140 0.5919 0.2662
160 0.5921 0.2662
180 0.5921 0.2662

Table 3.2
Convergence of Insertion Loss
Frequency = 1.4478fc , f c = TE-∣o cutoff f r e q u e n c y i n r e c t a n g u l a r w a v e g u i d e
I n s e r t t h i c k n e s s = 0.01094a
S e p t u m lengths Resonator lengths
d↑ = 0.1528a⁄τr ⁄ 1 = 2.2055 a⁄τr
d 2 = 0.8146a⁄τr ⁄ 2 = 2.2568a⁄ττ
(⅛ = 0.9985 a⁄ 77 ⅛ = 2.2603a⁄π

Number of M o d e s Insertion Loss (dB)

2
20 0.9602 × 10
2
40 0.6994 × 10
2
60 0.8659 × 10
2
80 0.9354 × 10
-2
100 0.8993 × 10
2
120 0.8583 × 10
2
140 0.8646 × 10
-2
160 0.8848 × 10
2
180 0.8888 × 10
66 Generalized Filter Design by Computer Optimization

Table 3.3
Variation of Dominant Eigenvalue With Number of Terms in Region 1 (N) and
Region 2 (M) Parameters: a - 12.7 mm, b = 10.16 mm, t = 2.54 mm, s = 2.794 mm

M N kc (rad⁄mm)

5 1 0.1439
5 5 0.1454
5 10 0.1468
5 15 0.1482
5 20 0.1503
5 30 0.1628
10 1 0.1449
10 5 0.1462
20 5 0.1439
30 5 0.1438

Table 3.4
Variation of Higher Eigenvalue With Number of Terms in Region 1 (N) and Region 2 (M)
Parameters: a = 12.7 mm, b = 10.16 mm, t = 2.54 mm, s = 2.794 mm

M N kc (rad⁄mm)

5 1 9.3642
5 5 9.3637
5 10 9.3119
5 15 9.3645
5 20 9.3645
5 30 9.3645
10 1 9.3639
10 5 9.3628
20 5 9.3629
30 5 9.3628

N= 5 and M = 30 with enough accuracy when examining the convergence of the

variation of the eigenvalue as a function of 7Vand M. For example, by comparing
the eigenvalues of this proposed analysis with those of [15,17,18] for double-
ridged waveguide (Table 3.5), it can be seen that all eigenvalues exhibit a good
agreement.
Table 3.6 shows the convergence of the reflection coefficient magnitude of a
metal septum with a length of 0.6996 ⁄⁄> and thickness of 0.0043745tf placed
between two different ridged waveguides, with gaps of 9.00 and 1.00 mm, respec-
tively, as the number of the modes used in the fields expansions is increased. The
 Characterization of Discontinuities 67

Table 3.5
Cutoff Eigenvalues k c ( r a d ⁄ m m ) of t h e Five M o d e s i n a Double-Ridged W a v e g u i d e
Parameters: a = 12.7 mm, b = 10.16 m m , t = 2.54 mm, s = 2.794 m m

Mode 1 2 3 4 5

Present method 0.1438 0.3162 0.6190 0.6712 0.6972

Ref. [18] 0.1438 0.3155 0.6215 0.6707 0.6971
Ref. [19] 0.1437 0.3166 0.6190 0.6712 0.6973
Ref. [15] 0.1439 0.3166 0.6191 0.6711 0.6974

Table 3.6
Convergence of M o d u l e of Reflection Coefficients
S e p t u m length = 0.6996a⁄τr
S e p t u m t h i c k n e s s = 0.0043745a
Frequency = 1.4453 f c , f c = TEjo cutoff f r e q u e n c y i n r e c t a n g u l a r w a v e g u i d e
Gap of t h e r i d g e d w a v e g u i d e s (mm) = 9.00/1.00

Number of Modes l⅜l

1 0.84375
5 0.97795
20 0.97907
40 0.97724
60 0.97928
80 0.97929
100 0.97928
120 0.97927
140 0.97926

convergence of the insertion loss of an asymmetrical fifth-degree ridged waveguide

E-plane bandpass filter (nonuniform convergence) is presented in Table 3.7. As
can be seen, the convergence of the reflection coefficient magnitude of the metal
septum (Table 3.6) is very good. However, the convergence of the insertion loss of
an asymmetrical fifth-degree ridged waveguide filter (Table 3.7) is not as good.
Therefore, to ensure good accuracy, more than 140 modes need to be included in
the field expansions. That is similar to the situation for the septum in rectangular
waveguide and is due to the singular behavior of the magnetic field at the edges of
the septum. The calculation of insertion loss requires the evaluation of the ABCD
parameters of six E-plane septa. The same number of modes is used in all six cases.
About 17 hours of CPU time on a SUN computer (SPARC station 10) are required
for the calculation of the insertion loss of the filter at one frequency using 100
68 Generalized Filter Design by Computer Optimization

Table 3.7
Convergence of Insertion Loss
Insert thickness = 0.0043745a
Frequency = 1.4453 f c , f c = T∈iq cutoff frequency i n rectangular waveguide
Gap of the ridged waveguides (mm) = 9.00/1.00/8.00/2.00/7.00
Septum lengths Resonator lengths
= 0.19643a⁄p ⁄ 1 = 2.17314a⁄p
= 0.69969a⁄p ⁄ 2 = 1.89891alp
cS* rS

= 0.88734a⁄p ⅛ = 2.22128a⁄p
= 0.88467alp ⁄ 4 = 2.00274a⁄p
= 0.64905a⁄p ⁄ 5 = 2.15683a⁄p
= 0.15275a⁄p

Number of Modes Insertion Loss (dB)

l
1 0.2261 × 10
l
5 0.3590 × 10
l
20 0.3748 × 10
1
40 0.3985 × 10
,
60 0.4564 × 10
,
80 0.4621 × 10
,
100 0.4508 × 10
,
120 0.4364 × 10
,
140 0.4239 × 10

modes. Because the calculation of the insertion loss at one frequency using 1 0 0
modes requires about 1 7 hours of CPU time, the overall numerical effort can easily
become exceedingly heavy. Therefore, working with a large number of modes
places a heavy demand on computing resources.

3.3 Coplanar Waveguide Discontinuities

3.3.1 Introduction to Coplanar Waveguide Modeling

To design filters using CPWs [ 1 3,20] as the main transmission line, it is necessary
to characterize as many standard subsections o f the layout as possible. This section
examines, through a variety of simulations and measurements, Em software, by
Sonnet, which calculates S parameters for predominantly planar geometries using
the method of moments. A wide range o f electromagnetic simulators is available
for CPW circuits. A list of some commercial electromagnetic simulators is given in
Table 3.8. The listed programs are intended to solve for the S parameters of arbi-
trarily shaped CPW structures. By limiting the problem to predominantly planar
 Characterization of Discontinuities 69

Table 3.8
Some Electromagnetic Simulators

Product (all trademarks

Company acknowledged) Type

HP-EEsof (HP range) Momentum 3D planar electromagnetic

HFSS (HP/Ansoft) 3D arbitrary electromagnetic
Sonnet Software Em 3D planar electromagnetic
Jansen Microwave Unisym/Sfpmic 3D planar electromagnetic
Ansoft Corporation Maxwell-Strata 3D planar electromagnetic
Maxwell SI Eminence 3D arbitrary electromagnetic
ArguMens Stingray 3D planar electromagnetic
Optimization System Empipe Sonnet, Ansoft, HP optimization
Associates
Zeland Software IE3D 3D arbitrary electromagnetic
MacNeal-Schwendler Corp. MicroWaveLab 3D arbitrary electromagnetic
Kimberley Communications Micro-Stripes 3D arbitrary electromagnetic
Consultants (TLM)

structures, the speed of analysis is improved dramatically. Usually, the circuit con-
ductors are divided into subsections and the method of moments is used for the
electromagnetic analysis. An example of this type of simulator is Em. Often, many
metal and dielectric layers can be handled, but these are assumed to be planar, so the
term 2 . 5 D has been coined (2D currents but 3 D fields). These simulators cannot
analyze true 3 D structures, such as microstrip-to-stripline transitions, where the
dielectrics are not planar. Yet when the current on the third dimension can be used
to represent, for example, vias, but the analysis allows only layered dielectrics, the
term 3 D planar is more appropriate [2 1 ] . The speed of planar electromagnetic sim-
ulators makes them practical for carefully investigating nonstandard structures but
conventional optimization is not really feasible. For all those numerical methods,
the most time-consuming step is the solving of the matrix involving O ( n 3 ) number
of operations, where n is the number of elements in the matrix, while the building
of the matrix encompasses O ( n 2 ) operations. Recently, however, a fast iterative
matrix solver in which solution time increases only as the square of the number of
unknowns rather than the traditional cubic technique has been developed by
Ansoft Corporation for its 3 D planar Maxwell-Strata™ electromagnetic simula-
tor. This method, coined ALPS, is an adaptive procedure that allows wideband S
parameters data and radiation to be computed from a handful of single-frequency
analyses at the dominant poles and zeros.
OSA (Canada), however, has made a breakthrough in electromagnetic opti-
mization with the development of the Space-Mapping™ optimization technique
70 Generalized Filter Design by Computer Optimization

[22] used in the Empipe™ family. For 3 D planar electromagnetic simulators, a

minimum of 1 6 M B of RAM generally is required, and at least 32 M B ofswap space
is needed in the hard disk. Additionally, around 1 0 0 M B are needed to load and
store the program

3.3.2 Introduction to Xgeom™and Em™Software

The remainder of this chapter briefly describes how the software packages Xgeom
and Em can be used for modeling CPW discontinuities. Xgeom is an X-Window
system mouse-based program used to capture the CPW input geometry to the elec-
tromagnetic analysis program Em. The output file is saved with the .geo extension.
The software program Xgeom is for the precise analysis of passive microwave cir-
cuits. Xgeom provides a straightforward user interface that allows specification of
all needed information concerning the circuit to be analyzed. After capturing the
circuit with Xgeom and saving the resulting file, the user can run Em using the .geo
file as input. Em automatically subsections the circuit and performs (all effects
included) electromagnetic analysis. Em saves the resulting S parameters ready for
input to major microwave design programs such as DBFILTER or Touchstone.
Xgeom can capture (and Em can analyze) circuits with any number of dielectric
layers with any thickness, dielectric constant, and loss tangent. Em is a 3 D predom-
inantly planar electromagnetic analysis tool that can calculate S parameters for pla-
nar geometries. It takes into account dispersion, stray coupling, discontinuities,
surface waves, moding, metallization loss, dielectric loss, and radiation loss. The
electromagnetic analysis of CPW or any other 3 D planar geometry is done by solv-
ing the current distribution in the circuit metallization using the Method of
Moments. It analyzes planar structures inside a shielding box. Em can calculate
ultra-precise S parameters of specific discontinuities or groups of interacting dis-
continuities, and can grant design validation, thus eliminating the expensive
design iterations of a circuit. Em can produce microwave package evaluation, pro-
viding an analysis of a package prior to fabrication. Also , a wide variety of CPW dis-
continuities can be evaluated using Em. Em also can handle metallization between
any or all layers, including the effect of DC and skin effect loss as well as circuits
with any number of ports. A new feature of this software is the push-pull (balanced,
o r odd-mode) ports that allows CPW structures to be analyzed. More information
can be found in the Em User s Manual [ 1 2] .

3.3.2.1 Analysis of CPW Discontinuities with Xgeom and Em Software

Circuits with any number of ports can be analyzed with Em, without limit, and the
discontinuities between them can be evaluated precisely. That has an important
application for filters, to calculate the discontinuity and incorporate it in circuit
theory programs. The option is explained in detail in Section 1 2 of the Em Users
 Characterization of Discontinuities 71

Manual 1 2] . CPW filters can be examined using Em by means of level-to-level vias

to form air bridges and balanced ports. That is done in Xgeom by having negative
labels i n the ports, in the submenu Objects/Ports/Renumber. Em sums the total cur-
rent going out of all the ports with the same negative port number. I t is important
that the ground lines touch the sidewall only at the location of negative port num-
bers. Balanced ports are also called push-pull ports.
Figure 3 . 1 3 shows the layout of a CPW transmission line element being dis-
played in Xgeom’s window with push-pull ports (tabled —1, +1, —1, etc.) and
arrows marking the reference planes, which are employed for de-embedding (de-
embedding removes unwanted effects occurring at the box edges). Figure 3 . 1 3
also shows vias going between the upper metal layer and the lower metal underpass
ground straps. Figure 3 . 1 4 shows the layout of the CPW discontinuities such as
edge-coupled CPW lines (Figure 3.14a) and the CPW coupling gap (Figure
3.14b). That layout and the other layouts in this section were used as building
blocks in CPW filter design, which is described in detail in Chapter 8 . The CPW

Figure 3.13 Drawing of the CPW transmission line (CPW resonator) subsection being dis-
played in Xgeom’s window with push-pull ports and reference planes that are
used for de-embedding.
72 Generalized Filter Design by Computer Optimization

(a)

(b)

Figure 3.14 (a) Edge-coupled CPW lines and (b) CPW coupling gap.
 Characterization of Discontinuities 73

line, which has a thin metal underpass connecting both coplanar ground planes,
was characterized using an Em simulator. The analysis of this structure is well
within the capability of most workstations, with a run time of a few minutes per fre-
quency point. I t should be pointed out, however, that there is a practical limit to
the complexity of a layout that can be analyzed. Em is both a memory-intensive and
a computationally intensive program. Small circuits are analyzed quickly, but the
computing time and memory required to analyze a circuit increases dramatically

References

[1] Rozzi, T., et al., “Accurate Full Band Equivalent Circuits of Inductive Posts in Rectangular
Waveguide,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-40, May 1992,
pp. 1000-1009.
[2] Tao,J. W.,andH. Baudrand, “Multimodal Variational Analysis ofUniaxial Waveguide Dis-
continuities,” IEEE Trans. Microwave Theory &Tech.,\ ol.MTT-39 , March 1991, pp. 506-
516.
[3] Konishi, Y., and H . Matsumura, “Short End Effect of Ridge Guide With Planar Circuit
Mounted in Waveguide,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-26, October
1978, pp. 716-719.
[4] Ney, M . M., M . Chenier, and G . J. Costache, “Investigation on the Power Handling Capac-
ity of a Glass of E-Plane Millimetre-Wave Filters Using Finite Element Modelling,” Int.J.
Numerical Modelling, Vol. 2, 1989, pp. 93-102.
[5] Gebauer, A., and F. Hernandez-Gil, “Analysis and Design ofW aveguide Multiplexers Using
the Finite Element Method,” Proc. 18th European Microwave Conf, Stockholm, Sweden,
1988, pp. 521-524.
[6] Itoh, T., Numerical Techniques for Microwave and Millimeter-Wave Passive Structures, New
York: Wiley, 1989.
[7] Wexler, A., “Solution of Waveguide Discontinuities by Modal Analysis,” IEEE Trans.
Microwave Theory & Tech., Vol. MTT-15, September 1967, pp. 508-517.
[8] Bornemann,J., “Comparison Between Different Formulations of the Transverse Resonance
Field-Matching Technique for the Three Dimensional Analysis of Metal-Fined Waveguide
Resonators,” Int.J. Numerical Modelling, Vol. 4, March 1991 , pp. 63-73.
[9] Mansur, R. R., and R. H . Macphie, “An Improved Transmission Matrix Formulation ofCas-
caded Discontinuities and Its Application to E-Plane Circuits,” IEEE Trans. Microwave The-
ory & Tech., Vol. MTT-34, December 1974, pp. 1490-1498.
[10] Beyer, A., “Calculation of Discontinuities in Grounded Finlines Taking into Account the
Metallization Thickness and the Influence of the Mount-Slits,” Proc. 12th European Micro-
wave Conference, Helsinki, Finland, 1982, pp. 681-686.
[11] Mittra, R., and S. W. Lee, Analytical Techniques in the Theory of Guided Waves, New York:
Macmillan, 1971.
[12] Em User’s Manual, Vol. 1, Release 4.0, Sonnet Software Inc., Liverpool, NY, 1996.
74 Generalized Filter Design by Computer Optimization

[13] Gupta, K. C., et al., Microstrip Lines and Slotlines, 2nd ed., Norwood, MA: Artech House,
1996.
[14] J. P. Montgomery, “ O n the Complete Eigenvalue Solution or Ridged Waveguide,” IEEE
Trans. Microwave Theory & Tech., Vol. MTT-19,June 1971, pp. 547-555.
[15] Y. Utsumi, “Variation Analysis of Ridged Waveguide Modes,” IEEE Trans. Microwave The-
ory & Tech., Vol. MTT-33, February 1985, pp. 1 1 1-120.
[16] Collin, R. E., Field Theory of Guided Waves, New York: McGraw-Hill, 1960.
[17] Dasgupta, D., and P. K. Saha, “Eigenvalue Spectrum of Rectangular Waveguide With Two
Symmetrically Placed Double Ridges,” IEEE Trans. Microwave Theory & Tech., Vol.
MTT-29, January 1981, pp. 47-51.
[18] Hoefer, W. J. R., and M . R. Burtin, “Closed-Form Expressions for the Parameters of Finned
and Ridged Waveguide,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-30, December
1982, pp. 2190-2194.
[19] Pyle, J. R., “The Cutoff Wavelength of theTE10 Mode in Ridged Rectangular Waveguide
ofAny Aspect Ratio,” IEEE Trans. Microwave Theory & Tech.,No∖. MTT - 14, April 1966, pp.
175-183.
[20] Wen, C. P., “Coplanar Waveguide: A Surface Strip Transmission Line Suitable for Nonre-
ciprocal Gyromagnetic Device Application,” IEEE Trans. Microwave Theory Tech., Vol.
MTT-17, No. 12, December 1969, pp. 1087-1090.
[21] Rautio, J. C., “Some Comments on Electromagnetic Dimensionality,” IEEE Microwave
Theory & Tech. Soc. Newsletter,Winter 1992, pp. 23.
[22] Bandler, J. W., et al., “Electromagnetic Optimization Exploiting Aggressive Space Map-
ping,” IEEE Trans, on Microwave Theory & Tech., Vol. 43, December 1995,
pp. 2874-2882.

Selected Bibliography
Bornemann, J., and F. Arndt, “Modal S-Matrix Design of Optimum Stepped Ridged and Finned
Waveguide Transformers,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-35, June 1987,
pp. 561-567.
Bornemann, J., and F. Arndt, “Transverse Resonance, StandingWave, and Resonator Formulations
of the Ridge Waveguide Eigenvalue Problem and Its Application to the Design ofE-Plane Finned
Waveguide Filters,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-38, August 1990, pp.
1104-1113.
Budimir, D., et al., “V-Shaped CPW Transmission Lines for Multilayer MMICs,” IEE Electronics
Letters,October 1995, pp. 1928-1929.
Chu, T. S., and T. Itoh, “Comparative Study of Mode-Matching Formulations for Microstrip Dis-
continuity Problems,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-33, October 1985,
pp. 1018-1023.
Cohn, S. B., “Properties of Ridge Waveguide,” Proc. IRE,Vq∖. 35, August 1947, pp. 783-788.
Empipe Reference Manual, Version 3.1, Optimization System Associates Inc., Canada, 1995.
Fan, P., and D . Fan, “Computer Aided Design of E-Plane Waveguide Passive Components,” IEEE
Trans. Microwave Theory & Tech., Vol. MTT-37, February 1989, pp. 335-339.
Getsinger, W. J., “Ridged Waveguide Field Description and Application to Direction Couplers,”
IRE Trans. Microwave Theory drTech., Vol. MTT- 10, January 1962, pp. 41-51.
 Characterization of Discontinuities 75

HFSS Reference Manual, Release 2.0, Hewlett-Packard Co., Palo Alto, CA, 1992.
Hopfer, S., “The Design of Ridged Waveguide,” IRE Trans. Microwave Theory & Tech., Vol.
MTT-3, October 1955, pp. 20-29.
Konishi, Y., and K. Uenakada, “The Design of a Band-Pass Filter With Inductive Strip-Planar Cir-
cuit Mounted in Waveguide,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-22, October
1974, pp. 869-873.
Kuhn, E., “A Mode-Matching Method for Solving Field Problems in Waveguide and Resonator Cir-
cuits,” AEU, Band 27, Heft 12, 1973, pp. 51 1-518.
LINMIC+ User Manual, Version 2. 1, Jansen Microwave, Germany, 1989.
Mansur, R. R., R. S. K. Tong, and R. H . Macphie, “Simplified Description of the Field Distribution
in Finlines and Ridge Waveguide and Its Application to the Analysis ofE-Plane Discontinuities,”
IEEE Trans. Microwave Theory & Tech., Vol. MTT-36, December 1988, pp. 1825-1832.
Omar, A. S., and K. Schunemann, “Transmission Matrix Representation of Finline Discontinu-
ities,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-33, September 1985, pp. 765-770.
Omar, A. S., and K. Schunemann, “Application of the Generalized Spectral-Domain Technique to
the Analysis of Rectangular Waveguides With Rectangular and Circular Metal Inserts,” IEEE
Trans. Microwave Theory & Tech., Vol. MTT-39,June 1991, pp. 944—952.
Patzelt, H., and F. Arndt, “Double-Plane Steps in Rectangular Waveguide and Their Application for
Transformers, Irises and Filters,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-30, May
1982, pp. 771-776.
Ramo, S., J. R. Whinnery, and T. VanDuzer, Fieldsand Waves in Communication Electronics, New
York: Wiley, 1984.
Schwinger, J., and D . Saxon, Discontinuities in Wave Guide (Documents on Modem Physics), New
York: Gordon and Greach, 1968.
Shih, Y. C., and K. Gray, “Convergence of Numerical Solutions of Step-Type Waveguide Disconti-
nuity Problems by Modal Analysis,” IEEE MTT-S Int. Microwave Symp. Dig., 1983, pp. 233-
235.
Shih, Y. C., and T. Itoh, “E-Plane Filters With Finite-Thickness Septa,” IEEE Trans. Microwave
Theory & Tech., Vol. MTT-31, December 1983, pp. 1009-1013.
Shih, Y. C., T. Itoh, and L. Q. Bui, “Computer-Aid Design of Millimeter-Wave E-Plane Filters,”
IEEE Trans. Microwave Theory & Tech., Vol. MTT-31 , February 1983, pp. 135-142.
Si-fan, L., and C. Yi-yuan, “CAD of Rectangular and Ridged Waveguide Bandpass Filters,” IEEE
MTT-S Int. Microwave Symp. Dig, 1982, pp. 532-534.
Tao, J. W., and H . Baudrand, “Rigorous Analysis of Triple- Ridge Waveguides,” IEE Electronic Let-
ters, Vol. 24, No. 13, June 1988, pp. 820-821.
Vahldieck, R., et al., “Optimized Waveguide E-Plane Metal Insert Filters for Millimeter-Wave
Application,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-31, January 1983,
pp. 65-69.

Appendix 3A

Following Wexler’s procedure [ 1 ] , we enforce continuity of the transverse electric

field at junction {z= 0) by equating (3A.1) with (3A.2), taking the cross-product
with h am and integrating, keeping in mind the orthogonality relation [2, p. 230]
76 Generalized Filter Design by Computer Optimization

e
ai ’ a,me z s
~ θ (3A.1)
a

when ⁄≠ m. The surface integral extends over the entire cross-section of the region
a. The results are
N
pU↑ e ↑ · a h a ↑e z ds = ∑ b j bj ’ h a ↑e z ds

a b
N
(3A.2)
k=∖ J
b

for m = 1 and
N
a e f e s =
m I am * am z aτn 2

a b
N
(3A.3)
k=∖ J
b

for m ≠ 1.

the orthogonality relation

(3A.4)

when n ≠j, the result is

M
c
(1 — p) Λ∖ I ey n * h a ↑e z ds — al I e n * h a l·e z ds =
J i=2 J
b b
N
∖ C
e
( b∏ ~ Σjfyjn) I bn * hbn e z d ,s (3A. 5)
Ď j =1 ⁄j

Ifwe nowsubstitute (3A.3) into (3A.5) and change rato z, the coefficients
are eliminated, and we obtain the equation
 Characterization of Discontinuities 77

n
c
I e∣ j h ∖e ds +
y
,
bk ’ a∖ z
a z
e e s

n 1 mJ k=ι J
v j V J ____________b _____________
e
ai * hai e zds
b
b

e
bn ' hai e z d ,S + (
aι
f e
bn ’ bne z s
~ ∣ e bn ' a∖ e z ,s (3A.6)
∖ ⁄J
b b b

Equation (3A.6) is essentially N linear equations corresponding to

n = 1, 2, 3, . . . , N. There are 7V+ 1 unknowns, namely, pand the Λrmodal coef-
ficients in region b{b∖la↑), (⅛⁄ ι), . . . , {bj∖∣la↑)i but by dividing (3A.2) by a∖
and rearranging, we have

I e
bj ’ ha∖ e zds

b b
N
e
a∖ ’ ha∖ e z d ,S (3A.7)
k- 1 J
b a

N+ 1 unknowns. The integrations are fairly straightforward for this problem and
are as follows:

e
ai ’ h aι e z ds - ·5aY a i (3A.8)
a

(3A.9)
b

and

sin( ff
I e j * ba jez ds — 0.5 Ya (a t) (3A.10)
b

where
78 Generalized Filter Design by Computer Optimization

0.5√ -a — t - q 77 (3A.11)

and

(3A.12)

The results in (3A.8) to (3A.12) are substituted into (3A.6) and (3A.7) when
performing calculations. By solving the system of equations [i.e., (3A.6) and
(3A.7)] for /9, using both symmetrical and antisymmetrical excitations, equivalent
/"-circuit parameters (Figure 3.5) are determined. Assuming an open circuit at the
central plane (Figure 3.4), p = p can be computed from a system of N + 1 linear
equations using Sj∣ζ = +1 and from it the normalized reactive even-mode impedance
z e ∖s given by

z 1 + Pe
e ~ -------- (3A.13)
1 - Pe

With the short circuit at the central plane (Figure 3.4) , with Sj⁄ ζ = - 1, the nor-
malized reactive odd-mode impedance z 0 is given by

1 ÷ Po
(3A.14)
' ■ 1 - P,

Appendix 3B: Scattering Parameters for Bifurcated

Waveguides

Formulation 1A

χ x χ
S 12 = { γ - GY d+ GT (γ- GY d -G}

S2 1 = 2 ( Y a x GY d + G)~ x
512 = G{I + 5 2 2 )
5ιι = C52 i - I
 Characterization of Discontinuities 79

Formulation I B

Sn = ( γ - χ GYd + G T ∖γ - a x GYd - G )
⅛ = 2(Y~a x GYd + Gf x

Si 2 = Y~a x GYd ( J - S 22 )

511 = ⁄ - Y~ x GYd S2 l

Formulation 1C
x t t t t
S1 1 = (Y d G Ya ÷ G f∖y ~ d G Ya - G )
Sn = 2(Y d x G t Ya + GT 1

⅛ = G t(J + S]i)
S2 2 = G eS12 ~ I

Formulation I D
t
5 1 1 = GY~d G Ya + G t f∖Y~}G t Ya - G t }
5 1 2 = 2 ( r √ 1 <7¾ + Gt f x

¾ = Y~d x G Y f l - S xλ }

S2 2 = I - Y~dGYa Sn

Formulation 2A

‰ = (⁄ + γ-d λG t Ya G f ∖l - Y~d G t Ya G)
Sl x = 2 ( ⁄ + γ-d x G t Ya GT x Y~d x G t Ya
12 - G(I + ¾2)
S11 = GS2∖ - I

Formulation 2B
x t
S1 1 = {GY-d G Ya + DΓ∖GY~d x G t Ya - Γ)
Sn =2(GY~d x G t Ya + IΓ x G
⅛ = Y-d x G t YaG - S xx )

‰ = ⁄-
80 Generalized Filter Design by Computer Optimization

Formulation 2C

‰ = (G t Y~ x GYd + IΓ∖G t γ-a λ GYd - I)

⅛ = 2(Gt Y~ x GYd + I)~ x G t
⅛ = Y~a x GYd { I - S22 )
51 1 = 1 - Y~a x GYd S2 i

Formulation 2D

51 1 = (⁄ + γ- χ GYd G t γ∖l - Y~ x GYd G t )

‰ = 2 ( ⁄ + Y~ x GYd G t Γ x γ- χ GYd
<⅞ι = G t (I + S∏)
⅛2 = <J,'S12 - I

Appendix References
[1] Wexler, A., “Solution ofWaveguide Discontinuities by Modal Analysis,” IEEE Trans. Micro-
wave Theory & Tech., Vol. MTT-15, September 1967, pp. 508-517.
[2] Itoh, T., Numerical Techniques for Microwave and Millimeter-Wave Passive Structures, New
York: Wiley, 1989.
Optimization-Based Filter Design

This chapter attempts to show how problems within the scope of filter design
can be formulated effectively as optimization problems. The differences between
optimizations are explained, and an appropriate method indicated. Optimiza-
tion can be implemented in situations when the classical synthesis approach
is inappropriate.

4.1 Filter Circuit Parameters

Figure 4.1 shows a two-port lossless reciprocal network as a filter. It operates

between a resistive source and a resistive load. This network is characterized
by its scattering matrix [S(α>)] (4. la) normalized to the source impedance of
1∩ and the load impedance R[Ω.

Sn(ω) ¾ι(ω)
[5(ω)] = (4.1a)
⅛(ω) ⅛(ω)

Two-port
Lossless
Network

Figure 4.1 Normalized two-port network.

81
82 Generalized Filter Design by Computer Optimization

where

S ij {ω·) = ∖S ij {ω·)∖e (4. lb)

Some of the commonly used terms (see Figure 4.2) in filter design are defined
here [1].
Center frequency (for bandpass filters) is defined as

⁄o = √⁄½ <4.1c)

where fι and fy are the upper and lower bandedge frequencies, respectively.
Degree is defined as the number of elements or resonator sections used
in a filter.
Bandwidth (in megahertz or percentage) is defined as the frequency
difference between specified attenuation levels, generally 1 dB or 3 dB.
Insertion loss, Lj (in decibels), is defined as

Li = 10 logκ⁄- — (4.1d)
Ďl⅛ι I ⁄

Z/is a measure of the total attenuation of a signal after the signal passes through
the filter. It includes dissipation (or resistive) loss and the effects of input/output
mismatch loss. In the case of a bandpass filter, insertion loss is proportional to
the center frequency and degree and inversely proportional to bandwidth and
the unloaded Q of the medium.

Lι⁄dB

Frequency

Figure 4.2 Insertion loss response of bandpass filter.

 Optimization- Based Filter Design 83

Ripple, Z,zr, is defined as the variation of insertion loss amplitude across

a defined frequency range.
Return loss, (in decibels), is defined by

Rl = 10 loglθ(-7-2) (4.1e)
Ďl-*11 1 ⁄

Rl is a measure of the input and output match to the characteristic

impedance of the medium.
Selectivity is a measure of the rate of increase of attenuation at the band
edge.
Rejection is the through attenuation to unwanted signals.
Group Delay, ⁄√(in seconds), is the time taken for an item of information
to transmit from the input to the output of the device. Group delay is propor-
tional to the rate of change of electrical phase, as shown by

(4.1F)
aω

where φ is the transmission phase (in radians), and ω is the angular frequency
(in radians per second).
Power handling is the amount of RF power that can be handled by the
filter without voltage breakdown. The power-handling level must be specified
so the appropriate filter design is selected. Waveguide and coaxial filters have
to be used for higher power applications, while printed circuit filters usually
are used for lower power applications (up to a few hundred watts).
Temperature range is the ambient temperature range over which the filter
must meet the full electrical specification.

4.2 Filter Design

Selection of the prototype characteristic function that most closely approximates

the required filter characteristic is the first step of any filter design. Several
transfer functions can be used. Figure 4.3 shows (a) Chebyshev, (b) generalized
Chebyshev or quasi-elliptic, and (c) elliptic function responses. The second
step is synthesis of a lowpass prototype network. Once the degree and type of
transfer function required have been determined, the prototype network is
mathematically transformed into lowpass, highpass, bandpass, and bandstop
filters, either in distributed or lumped form. Sometimes the final filter can be
synthesized directly using either distributed or lumped techniques.
84 Generalized Filter Design by Computer Optimization

L∣⁄dB

(a)
Llr

f< f

Frequency

Lι⁄dB

Llr

Frequency

Lι⁄dB

(c)

Llr

Frequency

Figure 4.3 (a) Chebyshev function response; (b) quasi-elliptic function response; (c) elliptic
function response.
 Optimization-Based Filter Design 85

The next step is the realization phase, in which the element values, that
is, transmission lines, inductance, and capacitance are converted into realizable
entities. The following parameters are the most important of any filter:

• Frequency;
• Bandwidth;
• Insertion loss;
• Selectivity;
• Rejection;
• Group Delay;
• Mechanical consideration.

The desired frequency and bandwidth of filter tend to dictate the medium
in which a filter is best fabricated. Some applications and media are described
in Chapter 2.
Classification of filter networks is shown in Figure 4.4. Table 4.1 shows
characteristics of various filter realization media [2].

4.3 Filter Optimization

When a common approach to the design of filters (initial filter design in Figure
4.5) results in a design passband that differs considerably from that which is
specified, optimization is required to tune the filter dimensions to achieve a
design that meets certain requirements. Most microwave filters have not yielded
exact optimum synthesis. Taking into account parasitic effects, high-frequency
operation, frequency-dependent elements, a narrow range of element values,
and so on, a common approach to design provides, at best, only approximate
answers. Not infrequently, a common approach can be used to great advantage
in providing the initial points for optimization.
This book introduces an optimization procedure based on Cohn’s equal-
ripple optimization to optimize filters based on Chebyshev function prototype.

the ripple levels at those points to have specified values. The method requires
knowledge of the filter insertion or return loss at those points. The method
generates a set of equations that are solved to give a new set of parameter
values. The cycle then is repeated until the filter characteristic is within an
arbitrarily close value to the desired specification. The technique requires less
86 Generalized Filter Design by Computer Optimization

FILTER SPECIFICATIONS

APROXIMATION
(Selection of Prototype
Characteristic Function)

SYNTHESIS

REALIZATION
INITIAL FILTER (Lumped and/or
DESIGN Distributed Filter
Elements)

ANALYSIS OPTIMIZATION

NO
ARE SPECS MET ?

YES

FABRICATION

Figure 4.4 Classification of filters.

optimization routines so far applied. Cohn’s technique optimizes the passband

of a filter with respect to the Chebyshev (or minimax) criterion. The Chebyshev
criterion can be defined as follows [3]:

An approximating function F(x) constitutes an optimal Chebyshev

approximation of a target function f(x), on an interval [a,b], when
 Optimization-Based Filter Design 87

Table 4.1
Characteristics of Various Filter Realization Media

Media Frequency (GHz) Bandwidth (%) Q (at 10 GHz)

Coaxial 0.10-40.00 1.0-30 2,000

Waveguide 1.00-100.0 0.1-20 5,000
Stripline 0.10-20.00 5-octave 150
Microstrip 0.10-100.0 5-octave 200
Suspended microstrip 1.00-200.0 2.0-20.0 1,000
Finline 20.0-200.0 2.0-50.0 500
Lumped elements 0.01-10.00 (hybrid) 20-octave 200
0.10-60.0 (monolithic) 20-octave 100
Dielectric resonator 0.9-40.00 0.2-20 10,000

RF&MICROWAVE&MM-WAVE FILTERS

FILTERS WITH FILTERS WITH

DISTRIBUTED ELEMENTS LUMPED ELEMENTS

CAVITY COMBLINE TRANSMISSION MMIC LC HYBRID LC

AND LINE
INTERDIGITAL

WAVEGUIDE COAXIAL DIELECTRIC

RESONTOR

RECTANGULAR HELICAL STRIPLINE

MICROSTRIP
RIDGED
SUSPENDED
CIRCULAR SUBSTRATE

COPLANAR
WAVEGUIDE

Figure 4.5 Flow chart for direct search optimization method.

88 Generalized Filter Design by Computer Optimization

E = max∣F(x) -⁄(x)∣ x cz [a, b] (4.1g)

is minimal with respect to all the accepted functions E(x). The

variable x is real. Whether the requirements concern attenuation
or phase response, it is usually required that the real characteristic
does not depart from a target function by more than a maximum
deviation within a certain interval.

Here, E is the error function, which relates directly to the way filters are
specified in practice. The advantages of this method are that the problems of
local minima are avoided, it requires fewer sampling points to achieve conver-
gence than generalized error-minimization algorithms, and the Chebyshev crite-
rion is satisfied. This method can handle symmetrical and asymmetrical
bandpass [4-9], lowpass [10], and highpass filters based on Chebyshev, general-
ized Chebyshev, and elliptic function prototypes.
General purpose optimization techniques based on least y>th, least-square,
minimax, or worst-case objective functions use general forms of error-minimiza-
tion algorithms [11], which simply force the filter transfer characteristic to
be within specified constraints, whereas a filter must have a specified ripple
characteristic, for example, the Chebyshev function. Usually the response of
an optimizable filter is sampled at a number of equally spaced frequencies, and
the error between that sampled response and the desired response is computed
at each frequency. The optimization program, through an iterative process,
reduces the error to a minimum, arriving at a final filter design in terms of the
optimized filter parameters. These optimization techniques (such as: gradient,
gradient minimax, least ⁄>th, minimax, quasi-newton, random, random mini-
max, random maximizer) cannot be guaranteed to satisfy filter specifications
and may even converge to a local minimum.
Two steps in the optimization algorithm are fundamental: the determina-
tion of a search direction and the search for the minimum in that direction.
The determination of the search direction is the most difficult part of optimiza-
tion. There are two different ways of carrying out the determination of the search
direction: gradient methods and direct search methods. Gradient methods use
information about derivatives of the performance functions (with respect to
designable parameters) for arriving at the modified set of parameters. Figure
4.6 shows a flow chart for that method. The direct search algorithms do not
use gradient information, and parameter modification is carried out by searching
for the optimum in a systematic manner. A flow chart for that optimization
method is shown in Figure 4.7.
Cohn’s optimization technique seeks to make an approximately equal-
ripple circuit response function exactly equal ripple by iterative adjustment of
 Optimization-Based. Filter Design 89

FILTER SPECIFICATIONS

INITIAL FILTER
DESIGN

ANALYSIS OPTIMIZER

NO GRADIENT
ARE SPECS MET ? INFORMATION

YES

FABRICATION

Figure 4.6 Flow chart for gradient optimization method.

the values of the circuit elements. Because the equal-ripple bandpass response
for stepped impedance half-wave filters is optimum, in the sense of providing
the minimum degree solution for a given passband and stop band specification
[3], it seems appropriate to design filters by the application of Cohn’s optimiza-
tion technique to an approximate Chebyshev design. For a symmetrical filter
with an equal ripple passband the characteristic function g, employing the
terminology used by Gupta in his paper [12] on the design of multivariable
lowpass equal-ripple filters, exhibits, within the passband, equal amplitude
ripples that alternate in sign. The function

⁄ g (4.2)
P =
90 Generalized Filter Design by Computer Optimization

FILTER SPECIFICATIONS

INITIAL FILTER
DESIGN

ANALYSIS OPTIMIZER

NO
ARE SPECS MET ?

YES

FABRICATION

Figure 4.7 Flow chart for filter design.

also exhibits that property. Cohn assumed that for a sufficiently good approxi-
mate filter design g and p would still exhibit ripples of alternating sign, but
not necessarily equal amplitude. Cohn thus formulated the design of symmetri-
cal filters by equal ripple optimization in terms of p ' , stating that g could just
as well be used. Cohn stressed the importance of having a response function
that exhibited ripples of alternate sign, p ' , as pointed out by Cohn, has the
property that

l⁄l = l⅛∣ (4.3)

where 5∏ is the input reflection coefficient of the filter. For a symmetrical

filter, ∣S1 11 will have the same number of ripples as p' and g, but the ripples
will all be positive.
 Optimization- Based Filter Design 91

∣$111 will, of course, be zero when p' and g are zero. Since in equal-
ripple optimization it is necessary only to make the ripple amplitudes equal
to some prescribed value,∣Su∣could just as well be employed as p' or g y it
is not necessary to have a response function that will exhibit ripples of alternate
sign. Any suitable function of∣S∏∣ also could be employed.
A similar equal-ripple optimization method has been described by Gupta
for the design of multivariable lowpass equal ripple filters. Gupta’s method
[12] differs from Cohn’s method as follows.

• Gupta’s set of simultaneous equations includes equations to solve for

the fι (maxima frequencies) within the passband. Thus, the general
bandpass filter with specified bandedges requires 2 n simultaneous equa-
tions to be solved, rather than the n + 1 used in Cohn’s method.
• He treats only lowpass symmetrical filters, which have {n — 1)⁄2
independent , excluding the bandedge, and {n + 1)⁄2 independent
parameters, requiring n simultaneous equations compared to
( n + 1)⁄2 in Cohn’s method.
• Gupta computes his response function from the ABCD matrix of the
total filter, rather than that of its bisected half, as in Cohn’s method.

By making the optimization algorithm force the zeros as well as the peaks
of the equal-ripple error function, this method can be used for the asymmetrical
case as well. If n peaks are present, an additional set of n + 1 sample functions
will be needed for the zeros and the same number of additional equations will
be required. Fortunately, the asymmetrical filter has the same number of
additional optimization parameters.
Formulation of the equal-ripple optimization in the context of the design
of symmetrical and asymmetrical filters, in terms of insertion loss, is given in
Section 4.4. The numerical implementation of equal-ripple optimization, in
the context of the design of lowpass, highpass, and bandpass filters based on
the Chebyshev function prototype is presented in Chapters 5, 6, 7, and 8,
respectively.

4.4 Description of the Algorithm

4.4.1 Symmetrical Case

To determine the degree of the filter (i.e., the number of resonators) required
to satisfy a given bandpass filter specification, the equal-ripple bandpass charac-
92 Generalized Filter Design by Computer Optimization

teristic proposed in [4] is frequently used in reactance coupled filters. This

characteristic in terms of insertion loss, Z√, is given by

(4.4a)

where T n is the mh degree Chebyshev polynomial of the first kind; 6 defines

the passband ripple level Z,zr; fpj y and ft are the upper and lower bandedge
frequencies; and

⁄o - √ΛA <4.4b)

In general, approximate methods based on the synthesis of a Chebyshev

prototype to the design of a symmetrical filter will not meet the specifications
satisfied by (4.4). Assume that an mh degree symmetrical bandpass filter has
an insertion loss response Lj of the form shown in Figure 4.8. It exhibits
m — l ( m = n + 1) zeros and m — 2 ripples, the maxima of which occur at the
frequencies , , . . . , f m -
For a symmetrical bandpass filter, all those m — 2 frequencies lie within
the specified passband fi=½ fi j ∙ The deviation of a ripple maximum from the
maximum allowed insertion loss in the passband, Ljr , is a function of the
m = n + 1 symmetrical filter parameter values required to specify the bandpass
filter. There are n - 1 such functions for the symmetrical case:

Lι⁄dB

Frequency

Figure 4.8 Scheme for the numerical optimization of symmetrical filter.

 Optimization-Based Filter Design 93

E} = Lj{fι) — Lf r i = 2, 3, . . . , n (4.5)

E↑ and Em are defined by:

E i = L1 ( f l ) - L lr (4.6)

Em = L1 {f h ) - Llr (4.7)

E↑ , Em are also functions of the m = n + 1 parameter values of the symmetrical

filter.
The specification

L1 (f) ≤ Llr , f ≤ f≤ f h (4.8)

is satisfied when

El = 0 i = 1, 2, 3, . . . , m (4.9)

This is a system of m = n + 1 nonlinear equations in m = n + 1 variables

for the symmetrical case. Solving (4.9) gives the parameter values of a filter
satisfying (4.8). The E i = 1, . . . , ni} can be regarded as the components of
an -dimensional error vector. Optimization is carried out by equating each
of the components to zero (a vector process) rather than minimizing the
magnitude of the vector (a scalar process). Thus, equal-ripple optimization can
be regarded as a vector procedure, whereas general purpose optimization rou-
tines are scalar procedures. Usually the convergence criterion applied in general
purpose optimization routines is that the gradient, with respect to the filter
elements, of the magnitude of the error vector is zero. However, a zero gradient
may correspond to a local minimum and the error may not be truly minimized.
The convergence criterion applied in equal-ripple optimization is that
each component of the error vector is zero. Thus, on convergence the error is
reduced to zero. The problem of local minima does not arise. Only the passband
is optimized in equal-ripple optimization. Fixing the number and the amplitude
of the ripples recovers the stop band performance of (4.4), except in the upper
portion of the waveguide band, where, due to the frequency dependence of
the electrical parameters of the E-plane septa, the assumptions made in [13]
no longer are valid. In practice, (4.9) needs to be solved iteratively. To apply
an iterative nonlinear equation solver in the case of symmetrical filters, it is
necessary for a given set of filter parameter values to know the insertion loss
only at the bandedge frequencies and at the ripple maxima. However, the
frequencies at which the ripple maxima occur are unknown and are functions
94 Generalized Filter Design by Computer Optimization

of the filter parameter values. Those frequencies can be approximately located

by calculating the insertion loss on a coarse sample of frequency points in the
passband for a given set of filter parameter values. Figure 4.9 shows the ripple
maxima. Figure 4.9(a) shows fi correctly centered at the maximum with

L1 (f i - ∆ f ) , L1 (fi+ ∆f) < L1 { β (4.10)

for maximum. In Figure 4.9(b), the sample frequency, fi, is a little off, so the
function is sampled at frequencies slightly higher and lower.

Lι⁄dB
(a)

Z-∆⁄ ⁄, Z+∆⁄
Frequency

Lι⁄dB

(b)

f.-tf f>
Frequency

Figure 4.9 (a) f l on maximum and (b) f, off maximum.

 Optimization- Based Filter Design 95

By finding an equivalent parabola

Lj = a ÷ bf + cf 1 (4.11)

passing through the three points at — ∆fif· y and + Δ f> a correction is

derived that can be applied to the frequency to bring it closer to the extreme
(maxima). The necessary condition for the maximum of Lj(f) is that

d
= b + 2cf = 0 (4.12)
J

that is,

f-~Ta <4 ' 1 3 >

where f locates the maximum of Lj{fY The sufficiency condition for the
maximum of Lj{f) is that

⅜<θ
d
<4 1 4 >
J

The correct location and the amplitude of the ripple maxima can be
found, by using the above procedure (quadratic interpolation [14]) in the last
few iterations.

4.4.1.1 Algorithm for Solving System of Nonlinear Equations

The Newton-Raphson method [15] is a rapidly convergent technique for the
solution of a system of nonlinear equations if a good initial approximation is
available. The number of times the function is evaluated in the process of
finding its root is the usual measure of computational effort. That includes
function evaluations required to calculate derivatives numerically.
By using finite difference, the Jacobian matrix of the n ÷ 1 nonlinear
functions defined by (4.5) to (4.8) can be calculated numerically. For a given
set of filter parameters, the finite difference calculation of the Jacobian matrix
requires the evaluation of

f z∙(x1 , χ 2 > ∙ ∙ ∙ > 5 + <5*7 , ∙ ∙ ∙ > ⅞ + ι)

i = 1, 2, . . . , n + 1,
j = 1, 2, . . . , n ÷ 1 (4.15)
96 Generalized Filter Design by Computer Optimization

where x∖, ×2, . . . , ⅞ + ι denote the n + 1 parameters required to specify a

symmetrical filter.
By using sampling and quadratic interpolation, the evaluation of {E
generates as a by-product the value of

f i {x λ , x 2 , . . . , ×j + δx j , . . . , x„+ 1 )
z = 1, 2, . . . , n— 1, (4.16)
j = 1, 2, . . . , n + 1

≠ i i = 1, 2, . . . , n - 1, j = 1, 2, . . . , n + 1 (4.17)
oxy

thus can be readily evaluated using finite difference. Denoting by x and E the
n + 1 dimensional vectors with components x z∙(z = 1 , 2 , . . . , w + l ) and
2⅛(z = 1, . . . , n + 1), the Newton-Raphson method has the general form [12]

x k = X k ~ x - r∖x k
- λ )E{xk ~ λ } (4.18)

1
where k is the iteration number (k = 1, 2, . . . ) and ⁄ is the inverse of the
k—\ k
m × m Jacobian matrix evaluated at x . Once x has been calculated using
(4.18), fi(x i )(i = 1, 2, . . . , n - 1) can be approximated by

fi(x k ) = fi(×k ~ λ } + y (a 7 - a 7 1) (4 ∙ i 9 )
y+1 Ď J'χ=χ

Equation (4.19) identifies the regions within the passband that need to
be sampled to calculate E(x ), as well as J(x ). The response and the errors
after each iteration are computed again with the new corrected parameters,
until the errors are judged to be sufficiently small. A subroutine, EROPTIM,
has been developed and tested for several examples given in Chapters 5, 6,
and 7.

4.4.1.2 Evaluation of the Jacobian

In the Newton-Raphson method [15], the most complex task is evaluating the
Jacobian J of £(x) in (4.9). That can be done either numerically or, when
possible, analytically. In this work, a numerical approach has been adopted
because we do not have analytic expressions for the functions. In general, an
 Optimization-Based Filter Design 97

zw-dimensional system of equations requires m + 1 function evaluations to

calculate £(x) and J numerically using finite difference. The Jacobian matrix
of the m nonlinear functions is defined by

∂E↑ ∂E↑ ∂E↑

⅛ ⅜ ∂y m
∂E2 ∂E2 ∂E2
J _ ⅛1 ⅛ 2 3λ , (4.20)

θ⁄m lfm ∂fm

_dy } ∂y 2 dy m _

All blocks defined by (4.20) can be calculated numerically using finite

difference for a given set of filter parameter values.

4.4.2 Asymmetrical Case

Figure 4.10 shows the form of insertion loss response Lj of an mh degree

asymmetrical E-plane bandpass filter designed using approximate method. It
exhibits {m — l)⁄2(m = 2w + 1) nonzero minima, the minima of which occur
at the frequencies , . . . , f m -∖. There are (m - 3)⁄2 ripples, the maxima
of which occur at the frequencies , , . . . , f rn-2· All those m — 2 frequencies
lie within the specified passband fi j .

Lι⁄dB

hr

fl f1 F E. fi
Frequency

Figure 4.10 Scheme for the numerical optimization of asymmetrical filter.

98 Generalized Filter Design by Computer Optimization

The deviation of a ripple maximum from the maximum allowed insertion

loss in the passband, Ljry is a function of the m = 2w + 1 values required to
specify the bandpass filter. There are 2 n - 1 such functions

Ei = Lj(fy - Ljr i = 3, 5, . . . , I n - 1
Ei = L1 (f-) z'= 2, 4, . . . , 2w (4.21)

that have to be zero to satisfy the filter specification.

E↑ and Em are defined by

E λ = L1 {fι) - Llr (4.22)

E m = Lj{ fh) - Llr (4.23)

E x , Em are also functions of the m = 2 n + 1 parameter values of the asymmetrical

filter.
The specification

L1 ≤L lr J l <f<f h (4.24)

L1 ( f ) . 0 , f ≤ f ≤ f , (4.25)

is satisfied when

Et = 0 z' = 1, 2, 3, . . . , m (4.26)

This is a system of m = I n ÷ 1 nonlinear equations in m = 2 n + 1 variables

that in practice needs to be solved iteratively. The parameter values of a filter
satisfying (4.24) and (4.25) can be obtained by solving (4.26). The
E i = 1, . . . , m) can be regarded as the components of an -dimensional
error vector. By equating each of those components to zero (a vector process)
rather than minimizing the magnitude of the vector (a scalar process), optimiza-
tion is carried out.
To apply an iterative nonlinear equation solver, it is necessary for a given
set of filter parameter values to know the insertion loss only at the bandedge
frequencies and at the ripple maxima and minima. However, the frequencies
at which the ripple maxima and minima occur are unknown and are functions
of the filter parameter values. For a given set of filter parameter values, those
frequencies can be approximately located by calculating the insertion loss on
a coarse sample of frequency points in the passband. The ripple maxima and
minima are shown in Figures 4.11 and 4.12, respectively.
 Optimization-Based Filter Design

Lι⁄dB

(a)

Z-δ⁄ ⁄ Z+δ⁄
Frequency

Lι⁄dB

(b)

f i -∆f f f+∆f
Frequency

Figure 4.11 (a) f∣on maximum and (b) f∣off maximum.

Figures 4.11 (a) and 4.11 (b) show f l correctly centered at the maximum
and minimum with (4.9) for the maximum and

L1 ( f i - Δ ⁄ ) , L1{fi + Δ⁄) > L1 (fi (4.27)

for the minimum.

In Figures 4.12(a) and 4.12(b), the sample frequency, fa is a little off,
so the function is sampled at frequencies slightly higher and lower. By finding
an equivalent parabola (4.11) passing through the three points ⅛t — ∆fi fa
and + Δ fi a correction is derived that can be applied to the frequency ft to
100 Generalized Filter Design by Computer Optimization

Lι⁄dB

f, Z+δ⁄
Frequency

Lι⁄dB

(b)

f l -tf f. ⁄+∆⁄
Frequency

Figure 4.12 (a) f∣on maximum and (b) f∣off maximum.

bring it closer to the extreme (minima or maxima). The necessary condition

for the maximum (or the minimum) of L∣(f) is given by (4.12) and (4.13).
The sufficiency condition for the maximum of Z√( f) is given by (4.14) and

(4.28)

for the minimum.

 Optimization-Based Filter Design 101

By using that procedure (quadratic interpolation [14]) in the last few

iterations, the correct location and amplitude of the ripple maxima and minima
can be found.
In the asymmetrical case, the same algorithm can be used for solving the
system of nonlinear equations as in the symmetrical case. The difference is
only in the number of nonlinear functions. Instead of n + 1 nonlinear functions
considered in the symmetrical case, in the asymmetrical case 2w + 1 nonlinear
functions defined by (4.21) to (4.25) will be considered. The Jacobian matrix
of the 2 w + 1 nonlinear functions defined by (4.21) to (4.25) can be calculated
numerically using finite difference, as in the symmetrical case.

References

[1] Fikronic Components Product Guide, Filtronic Components Ltd., England.

[2] Bahl, I. J., and P. Bhartia, Microwave Solid State Circuit Design, New York: Wiley, 1988.
[3] Hasler, M., and J. Neiryuck, Electrical Filters, Dedham, MA: Artech House, 1986.
[4] Cohn, S. B., “Synthesis of Commensurate Comb-Line Band-Pass Filters With Half-
Length Capacitor Lines, and Comparison to Equal-Length and Lumped-Capacitor Cases,”
IEEE MTT-S Int. Microwave Symp. Dig, May 1980, pp. 135-137.
[5] Hunton, J. K., “Novel Contributions to Microwave Circuit Design,” IEEE MTT-S Int.
Microwave Symp. Dig, 1989, pp. 753-755.
[6] Postoyalko, V., and D. Budimir, “Design of Waveguide E-Plane Filters With All-Metal
Inserts by Equal Ripple Optimization,” IEEE Trans. Microwave Theory and Techniques,
Vol. 42, No. 2, February 1994, pp. 217-222.
[7] Budimir, D., and V. Postoyalko, “EPFILTER: A CAD Package for E-Plane Filters,”
Microwave J., August 1996, pp. 110-114.
[8] Budimir, D., “Optimized E-Plane Bandpass Filters With Improved Stop Band Perfor-
mance,” IEEE Trans. Microwave Theory dr Tech., February 1997, pp. 212-220.
[9] Parry, R., “Optimisation of Microwave Filters,” ESA Workshop on Advanced CAD for
Microwave Filters and Passive Devices, ESTEC/XRM, The Netherlands, November 6-8,
1995, pp. 265-271.
[10] Parry, R., “Optimisation of Microwave Filters,” Colloquium on Filters in RF and Microwave
Communications, Digest No.: 1992/220, University of Bradford, Bradford, England,
December 1992, pp. 7 / 1 - 7 /5.
[11] Bandler, J. W., “Optimization Methods for Computer-Aided Design,” IEEE Trans.
Microwave Theory & Tech., Vol. MTT-17, August 1969, pp. 533-552.
[12] Gupta, G . P., “A Numerical Algorithm to Design Multivariable Low-Pass Equal-Ripple
Filters,” IEEE Trans. Circuit Theory, Vol. CT-20, 1973, pp. 161-164.
[13] Rhodes, J. D., “Microwave Circuit Realizations,” in Microwave Solid State Devices and
Applications, D . V. Morgan and M. J. Howes, eds., England: Peregrinus, 1980,
pp. 49-57.
102 Generalized Filter Design by Computer Optimization

[14] Gupta, K. C., R. Garg, and R. Chadha, G4D of Microwave Circuits, Dedham, MA:
Artech House, 1981.
[15] Ortega, J. M., and W. C. Rheinboldt, Iterative Solution of Nonlinear Equations in Several
Variables. New York: Academic Press, 1970.

Selected Bibliography
Bandler, J. W., “Computer-Aided Circuit Optimization,” in Modem Filter Theory and Design,
G. C . Temes and S. K. Mitra, eds., New York: Wiley, 1973, pp. 211-271.
Bandler, J. W., and S. H . Chen, “Circuit Optimization: The State of the Art,” IEEE Trans.
Microwave Theory & Tech., Vol. MTT-36, February 1988, pp. 424-443.
Bandler, J. W., et al., “Efficient Optimization With Integrated Gradient Approximations,” IEEE
Trans. Microwave Theory & Tech., Vol. MTT-36, February 1988, pp. 444-455.
Broyden, C. G., “A Class of Methods for Solving Nonlinear Simultaneous Equations,” Mathemat-
ics of Computation, Vol. 19, 1965, pp. 577-593.
Cohn, S. B., “Generalized Design of Bandpass and Other Filters by Computer Optimization,”
IEEE MTT-S Int. Microwave Symp. Dig, June 1974, pp. TT1-T7⅛.
DBFILTER Reference Manual, Tesla Communications Ltd., London, England.
FILTER Reference Manual, Eagleware Corp., USA, 1993.
Levy, R., “Theory of Direct Coupled Cavity Filters,” IEEE Trans. Microwave Theory & Tech.,
Vol. MTT-15, June 1967, pp. 340-348.
Lim, J. B., C. W. Lee, and T. Itoh, “An Accurate CAD Algorithm for E-Plane Type Bandpass
Filters Using a New Passband Correction Method Combined With the Synthesis Procedures,”
IEEE MTT-S Int. Microwave Symp. Dig, June 1990, pp. 1 179-1 182.
LINMIC+ Reference Manual, Jansen Microwave, Germany, 1989.
Matthaei, G., L. Young, and E. M. T. Jones, Microwave Filters, Impedance-Matching Networks
and Coupling Structures, Dedham, MA: Artech House, 1980.
MDS Reference Manual, Hewlett-Packard Co., USA, 1994.
MMICAD Reference Manual, Optotek Ltd., Canada, 1995.
OSA90⁄HOPE Reference Manual, Optimization System Associates Inc., Canada, 1995.
SERIES IV/PC Reference Manual, Hewlett-Packard Co., USA, 1995.
Shih, Y. C., “Design of Waveguide E-Plane Filters With All Metal Inserts,” IEEE Trans.
Microwave Theory & Tech., Vol. MTT-32, July 1984, pp. 695-704.
SUPER COMPACT Reference Manual, Compact Software Inc., USA, 1994.
Temes, G . C., and D. A. Calahan, “Computer-Aided Network Optimization the State of the
Art,” Proc. IEEE, Vol. 55, 1967, pp. 1832-1863.
TOUCHSTONE Reference Manual, EEsofInc., USA, 1991.
Design of Lumped-Element Filters by
Optimization

Because small size is an important parameter in some communication systems,

lumped-element filters should be employed whenever possible. Besides size,
they offer some other advantages over distributed filters, such as lower cost and
broad stopbands free of spurious responses. Lumped-element filters constructed
using air-wound inductors soldered into a small housing [1,2] and parallel
plate chip capacitors, conventional thin-film techniques [3], thin-film/HTS
techniques [4-6], and micromachining techniques [7] have been reported in
the literature. These filters have been designed by approximate methods [8,9] .
Results have been usually good, but errors can be quite large in broadband
filters, and optimization is then required to tune the filter dimensions to satisfy
the design specification.
Section 5.1 describes an approximate synthesis-based design procedure
of lumped-element lowpass, highpass, bandpass, and bandstop filters. The
numerical implementation of equal-ripple optimization, in the context of the
design of lumped-element lowpass filters, is presented in Section 5.2. A filter
that employs five elements with three inductors and two capacitors, or two
inductors and three capacitors, with a cut-off frequency of 2 GHz is presented
as a design example in Section 5.3.

5.1 An Approximate Synthesis-Based Design Procedure

This section examines a design approach for passive lumped-element filters.

Figure 5.1 shows a two-port lossless reciprocal network, which operates between

103
104 Generalized Filter Design by Computer Optimization

Two-port
Lossless
Network

Figure 5.1 Normalized two-port lossless network.

a resistive source and a resistive load, as the simplest type of filter. The two-
port lossless reciprocal network is characterized by its scattering matrix [S(ω)],
as shown in (5. la), normalized to the source impedance of 1∩ and the load
impedance of R[Ω.

Sn(ω) ⅛ι(ω)
?
⅞ 1 (ω) ‰(ω)J °

where

Sy(ω) =∣5ξz( ω )∣√ (5. lb)

S21(ω) represents the scattering transmission coefficient of a two-port

lossless network. In the complex frequency plane, the transfer function ⅛ι (5)
can be defined as

⅛v) - 2)ζ5>) (5∙lc)

The realizability condition of a two-port lossless network constrains N(s) to

be a Hurwitz polynomial (i.e., all zeros of N(s) must lie in the left 5,-half plane).
Of course, D(s) is a strictly Hurwitz polynomial.

5.1.1 Element Values of Chebyshev Lowpass Prototype Filters

A lowpass filter with ω' = 1 as normalized cut-off frequency and 1∩ as source

and load resistors is defined as the lowpass filter prototype (Figure 5.2). For
Chebyshev (equireripple passband and maximally flat stopband) response, the
insertion loss (Figure 5.3) can be expressed as
 Design of Lumped-Element Filters by Optimization 105

1Ω
Lowpass
1Ω
Filter

ω' = 7

Figure 5.2 Lowpass filter prototype.

L f ⁄dB

ω,

Figure 5.3 Chebyshev lowpass filter response.

Li = 10 log 1 0 f [dB] (5.2)

Ďl⅛ιl ⁄

with

∣⅛ι∣2 = - 1 2 (5.3)
1 + 6 T n( ω )

where T n ( ω ' ) is the zzth-degree Chebyshev polynomial of the first kind defined
by

T n ( ω ' ) = cosh[w Arcosh(ω , )], for ∣α>'∣≥ 1 (5.4a)

Tfω') = cos[w arccos(ω , )], for 0 < ω' ≤ 1 (5.4b)

and 6 is the ripple level defined by

106 Generalized Filter Design by Computer Optimization

2
f . 1 0 l ⅛' ,
°l - 1 (5.5)

where Ljr is the passband ripple in decibels.

This response can be realized by a ladder network with inductors as series
elements and capacitors as shunt elements. Several possible forms of a lowpass
filter prototype are shown in Figure 5.4. The total number of elements required
is given by the value of n in (5.2). The prototypes are easily changed to other
impedances levels and frequency scales by using transformations applied to the
filter elements.
The element values g∣ζ (inductance if a series element and capacitance if
a shunt element) for a Chebyshev lowpass prototype can be calculated from
the following equations:

8ι

n even

8ι

n odd

Figure 5.4 Several possible forms of a lowpass filter prototype.

 Design of Lumped-Element Filters by Optimization 107

n even

81 8n

n odd

Figure 5.4 (continued).

1
go =
77
2n
=
. / 1 . . 1 1Ď
sinh∣- s i n h -e ∣
∖n ⁄
. ⁄2z - 1 Ď ⁄2z + 1 Ď
4 sin — ---- 77 sin — ---- 771
∖ 2n I ∖ 2n I
gigi+i = ---------7-.1 ---------1ττ ----------γ~l7τ i = 1, 2, . . . , (« - 1)
• l2⁄ ∙ . -1 Ď ∙ 2/ Ď
sinh I -sinh - I + sin —
∖n € ∖n
108 Generalized Filter Design by Computer Optimization

<gw+l = 1> f° r n
θdd

‰ 1 = --------,1 for n even (5.6)

2
(e ÷ √1TS

5.1.2 Filter Design Procedure

A common approach to the design of the lumped-element lowpass, highpass,

bandpass, and bandstop filters is well described in the literature [8,9] . Here,
only the most important steps in the design procedure are presented.

5.1.2.1 Lowpass Filters

For a given filter specification such as the cutoff frequency yielding ω c , passband
return loss (£/?), stopband attenuation (Z⁄) at frequency f s , the design procedure
is summarized as follows:

1 . Determine the passband ripple level 6 from the minimum passband

return loss, which is defined as:

Lr= 10 log 1 0 f 1 ÷⅛ (5.7)

Ď 6 ⁄

2. Determine the number of elements n at the designated stopband

frequency f s from (5.2) where application of the frequency transforma-
tion

, ω
ω =—
ω
(5.8)
c

that will transform the cut-off frequency from ω' = 1 to ω' = w c will
be used.
3. Determine the prototype element values gl from (5.6).
4. Determine the lumped-element values.

To determine the lumped-element values, the transformations of lowpass

prototype into lowpass filter with arbitrary source and load impedances, Zq
(Figure 5.5), are used.
These values are given by
 Design of Lumped-Element Filters by Optimization 109

Figure 5.5 Lumped-element lowpass filter.

l s (5 9)
∙ ■∙f'(∣
∖ c⁄')
ω -

for series inductors, and

c β
∙ ■ (⅛ t) (5.10)

for shunt capacitors.

5.1.2.2 Highpass Filters

For a given filter specification such as the cutoff frequency yielding ω c , passband
return loss (£/?), stopband attenuation (Z,⁄) at frequency fs , the design procedure
is summarized as follows:

1. Determine the passband ripple level 6 from (5.7).

2. Determine the number of elements n at the designated stopband
frequency fs from (5.2) where application of the frequency transforma-
tion

ω~c
ω' - (5.11)
ω

that will transform the lowpass prototype with cutoff frequency of

ω' = 1 into a highpass filter with a cut-off frequency of ω' = w c will
be used.
110 Generalized Filter Design by Computer Optimization

3. Determine the prototype element values gl from (5.6).

4. Determine the lumped-element values.

To determine the lumped-element values the transformations of lowpass

prototype into highpass filter with arbitrary source and load impedances, Zθ
(Figure 5.6), are used.
Those values (Figure 5.7) are given by

G = (5.12)

for series capacitors, and by

L,-⅛ (5.13)
giω c

for shunt inductors.

Figure 5.6 Lumped-element highpass filter.

Lowpass Highpass

Figure 5.7 Lowpass-to-highpass transformation of filter elements.

 Design of Lumped- Element Filters by Optimization 111

5.1.2.3 Bandpass Filters

For a given filter specification such as the two passband edge frequencies
yielding a>[ and ω⁄⁄, passband return loss {Lp) i stopband attenuation (Z,⁄) at
frequency f s , the design procedure is summarized as follows:

1. Determine the passband ripple level 6 from (5.7).

2. Determine the number of elements n at the designated stopband
frequency from (5.2) where application of the frequency transforma-
tion

. 1⁄ ω <y∩Ď
ω - —∣ -----------∣ (5.14)
<5yα>0 ω )

with

δ = mh- (O L
(5.15)
(t>Q

and

(5.16)

that will transform the lowpass prototype with cutoff frequency of

ω' = 1 into a bandpass filter with the two passband edge frequencies
yielding ω ι and α>⁄⁄, will be used.
3. Determine the prototype element values gl from (5.6).
4. Determine the lumped-element values.

To determine the lumped-element values, the transformations of lowpass

prototype into bandpass filter with arbitrary source and load impedances, Zq
(Figure 5.8), are used.
Those values (Figure 5.9) are given by

Zo 2τr∆ω

for series-tuned series elements, and by

112 Generalized Filter Design by Computer Optimization

Figure 5.8 Resonator ladder lumped-element bandpass filter.

Lowpass Bandpass
∆ω
ω Lω∖

IP
∆ω
2
Cω 0

C
∆ω

Figure 5.9 Lowpass-to-bandpass transformation of filter elements.

2ττZ Δ,ω gi
2 (5.18)
ω
gι 0 2ττiZQ∆.ω

for shunt-tuned shunt elements, where

∆ ω = (Off — (Of (5.19)

and

ω (5.20)
0 = y∣ω H ω L
 Design of Lumped-Element Filters by Optimization 113

5.1. 2.4 Bandstop Filters

For a given filter specification such as the two bandstop edge frequencies
yielding and ωχ, passband return loss (Lf), bandpass attenuation (Z√) at
frequency f p the design procedure is summarized as follows:

1. Determine the passband ripple level € from (5.7).

2. Determine the number of elements n at the designated bandpass
frequency from (5.2) where application of the frequency transforma-
tion

1
(5.21)
1 ⁄ <y <ι)Q
<5Ďωo ω

with

§ = (*>H~ ω L
(5.22a)
ω0

and

ω
o = ∖∣ω H ω L (5.22b)

that will transform the lowpass prototype with cutoff frequency of

<y' = 1 into a bandstop filter with the two bandstop edge frequencies
yielding (ι)[ and ω⁄y, will be used.
3. Determine the prototype element values gi from (5.6).
4. Determine the lumped-element values. To determine the lumped-
element values, the transformations of lowpass prototype into bandstop
filter (Figure 5.10a) with arbitrary source and load impedances, Zq,
are used.
Those values (Figure 5.1 Ob) are given by

1 27rz ω
r∙ - τ - osA (5.23)
t , t
2πZQg i Δω ω 2

for series-tuned series elements and by

2πg i ∆ ω = Zq
Q = (5.24)
Z 0 ω⅛ ’ ' 2τr i gi ∆ ω
114 Generalized Filter Design by Computer Optimization

Lowpass Bandstop

L ∆ω
ω0

1
L ∆ω

1 C ∆ω
C∆ω r .j

Figure 5.10 (a) Lumped-element bandstop filter and (b) lowpass-to-bandstop transformation
of filter elements.

for series-tuned shunt elements, with

ω
∆td = ω H ~ L (5.25)

and

ω 0 = y∣ H ω L (5.26)

5.1.2.5 Immittance Inverters

The design procedure of the bandpass lumped-element filters described in the
preceding sections is useful for broadband design. However, if a narrowband
bandpass filter is required, then an alternative design procedure is more suitable.
 Design of Lumped- Element Filters by Optimization 115

Such a design procedure includes the concept of immittance inverters with

one kind of lumped element. The impedance inverter is defined by its ABCD
matrix

- 0 jk
[ABCD] (5.27)
j!K 0

The function of the inverter is illustrated in Figure 5.11. Terminating

the inverter by the load impedance (Z ) or admittance (F2), the input imped-
ance (Z)„) or input admittance (F)„) of the inverter is given by

2
K
(5.28)

(a)

+⁄-90'

y ιn

(b)

+⁄-90o

Z in

Figure 5.11 Inverters: (a) admittance inverter and (b) impedance inverter.
116 Generalized Filter Design by Computer Optimization

for K inverter, and by

(5.29)

for /inverter, where K{Y) is the characteristic impedance (admittance) of the

inverter.
For example, by terminating the inverter in an inductor of impedance
Z[ the input impedance of the inductively terminated inverter is the impedance
of a capacitor. Thus, an LC ladder network (see Figure 5.5) is equivalent to
a network consisting of series inductors separated by impedance inverters, as
shown in Figure 5.12. The element values of this alternative prototype network
are given by

2 . Γ ( 2 ⁄ - 1)7Γ-
— cm ------------ I = 1, 2, 3, 4, . . . , n (5.30a)

and

sin(⁄τr⁄ ri)
1, 2, 3, 4, . . . , n — 1 (5.30b)
y

with

y • κΓ -1 sinh
sinh ∙ k - √ -1 M (5.31)
n ∖e

Figure 5.12 Lumped-element lowpass filter using impedance inverters.

 Design of Lumped-Element Filters by Optimization 117

Similarly, a bandpass filter can be realized by series inductors-capacitors

(Z-C) and series resonant circuits separated by impedance inverters
(Figure 5.13), or parallel inductors-capacitors (Z,-C) and parallel resonant cir-
cuits separated by admittance inverters (Figure 5.14). Thus, the use of immit-
tance inverters for filter design is a practical approach, since microwave filters
realized by alternate series and parallel resonators are difficult to realize. The
element values of this modified prototype network using impedance inverters
are given by

LplCpl = ~f= ⁄ = 1> 2, 3, 4, . . . , w (5.32a)

V<y0

(5.32b)

7+l
Ki,M = ω 0 δ∖Γ ⁄ = 1, 2, 3, 4, . . . , - 1 (5.32c)
V g⁄g⁄ + ι

Figure 5.13 Lumped-element bandpass filter using impedance inverters.

Figure 5.14 Lumped-element bandpass filter using admittance inverters.

118 Generalized Filter Design by Computer Optimization

ω
uδLp n Zo
72,W+1 = (5.32d)
gn

with

δ = ω∏- ω k
(5.32e)
ω 0

and

ω
ω 0 = yJω∏ L (5.32f)

The prototype element values are obtained from (5.6), while Lp t , Zθ,
and Cpj can be chosen as desired. Four types of circuits that can be used as K
or /inverters for the filter design are shown in Figure 5.15. Once the elements
of the filter have been found, the frequency response of the overall filter at
each frequency can be simulated by cascading the ABCD matrices of the filter

C K = l⁄ωC

Figure 5.15 Approximation to immittance inverters.

 Design of Lumped-Element Filters by Optimization 119

K-ωL

J = 1/cdL

Figure 5.15 (continued).

elements. To illustrate the application of this procedure to the design of a

lumped-element lowpass filter, the design of a lowpass filter with the specifica-
tions given in Section 5.3 is considered. Figure 5.18 shows the calculated
passband return loss (before optimization: dashed line) designed using that
procedure. As can be seen, the design specification still is unsatisfactory, and
optimization is often required in practice for the accurate design of the filters.

Figure 5.16 Lumped-element lowpass filter with series inductor as first filter element.
120 Generalized Filter Design by Computer Optimization

Figure 5.17 Lumped-element lowpass filter with shunt capacitor as first filter element.

0.00 —I
Return Loss (dB)

-20.00 -

-40.00 -

— Before optim.
----- After optim.

-60.00 ------
0.00 2.00 4.00 6.00
Frequency (GHz)

Figure 5.18 Calculated return loss before (dashed line) and after (solid line) optimization of
the lowpass lumped-element filter.

5.2 Numerical Implementation of Equal-Ripple Optimization

To apply the equal-ripple optimization technique described in [10,11] to the
design of lumped-element lowpass filters, it is necessary, for a given set of filter
element values, to be able to calculate the insertion loss on a sample of frequency
points within the specified passband. For a lumped-element lowpass filter (see
Figure 5.5), the insertion loss and the return loss can be expressed in terms
of an ABCD matrix. The matrix representation of the whole filter is

' βl3} t Γ βc< ’ A Ln j β L∏

(5.33)
JCl3 [j⅛4 C4 j C Ln D L n
 Design of Lumped-Element Filters by Optimization 121

in which An> Bn, Cn> and Dn are elements of the ABCD matrix of the series
inductor as filter element defined by

1 ωL·
(5.34)
0 1

and √4q∙, Bq, Cq, and Dq are elements of the ABCD matrix of the shunt
capacitor as filter element defined by

(5.35)

The overall filter response— insertion loss (Z√) and return loss (Z, )—
can be expressed in terms of elements of the total ABCD matrix of the filter
at each frequency (by directly combining the ABCD matrices of the individual
filter elements) as

Lj = 20 logιo (5.36)
2

⁄A + B + C + D
(5.37)
Lr=2Q ogl0
Ď√4 + 5 - C - D

A good approximate design of a lumped-element lowpass filter can be

obtained by the procedure described in Section 5.1. It is therefore adopted in
this chapter as a means of generating a starting point for the optimization.

5.3 Numerical Results

To illustrate our approach, a fifth-order lumped-element lowpass filter has

been designed, with the following specifications:

• Cutoff frequency: 2 GHz;

• Passband return loss: 20 dB;
• Insertion loss: 50 dB at 4 GHz;
• Source and load impedances: 50∩.

The filter can be described by three parameters: inductors (LI, L3) and
capacitor (C2), as marked in Figure 5∙16, or capacitors ( C l , C3) and (L2), as
122 Generalized Filter Design by Computer Optimization

marked in Figure 5.17. We used equal-ripple optimization with LI, L3, and
C2 as variables for the filter shown in Figure 5.16, and C l , C3, and L2 for
the filter shown in Figure 5.17.
The optimization variables before and after optimization are listed in
Tables 5.1 and 5.2. The dashed line in Figure 5.18 shows the calculated
passband return loss of filter using the approximate method. This approximate
design was used as a starting point for equal-ripple optimization. The passband
return loss calculated using the filter elements obtained on convergence are
indicated by the solid line in Figure 5.18.

Table 5.1
Lumped-Element L o w p a s s Filter With Series I n d u c t o r a s First Filter Element

Parameters Before Optimization After Optimization

Li (nH): 5.000 3.900

C2 (pF): 2.000 2.200
L 3 (nH): 8.000 7.200
C4 (pF): 2.000 2.200
L 5 (nH): 5.000 3.900

Table 5.2
Lumped-Element L o w p a s s Filter With Shunt Capacitor a s First Filter Element

Parameters Before Optimization After Optimization

c 1 (pF): 2.000 1.600

L 2 (nH): 5.000 5.500
C3 (pF): 3.000 2.900
U (nH): 5.000 5.500
C5 (pF): 2.000 1.600

References

[1] Levy, R., “Design Considerations for Lumped-Element Microwave Filters,” Microwave
Journal., February 1988, pp. 183-192.
[2] Geffe, P, R., “The Design of Single-Layer Solenoids for RF Filters,” Microwave Journal,
December 1996, pp. 70-76.
[3] Swanson D. G., “Thin-Film Lumped Element Microwave Filters,” IEEE MTT-S Int.
Microwave Symp. Dig., 1989, pp. 671-674.
 Design of Lumped- Element Filters by Optimization 123

[4] Swanson D . G., R. Porse, and B. J. L. Nilsson, “A 10 G H z Thin-Film Lumped Element

High Temperature Superconductor Filter,” IEEE MTT-S Int. Microwave Symp. Dig,
1989, pp. 671-674.
[5] Zhang, D., et al., “Narrowband Lumped-Element Microstrip Filters Using Capacitively-
Loaded Inductors,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-43, No. 12,
December 1995, pp. 3030-3036.
[6] Matthaei, G. L., S. M. Rohlfmg, and R. J. Forse, “Design of HTS, Lumped-Element,
Manifold-Type Microwave Multiplexers,” IEEE Trans. Microwave Theory & Tech., Vol.
MTT-43, No. 7, July 1996, pp. 1313-1321.
[7] Chi, C-Y., and G . Rebeiz, “Planar Microwave and Millimeter-Wave Lumped Elements
and Coupled-Line Filters Using Micro-Machining Techniques,” IEEE Trans. Microwave
Theory & Tech., Vol. MTT-43, April 1995, pp. 730-738.
[8] Cohn, S. B., “Direct-Coupled-Resonator Filters,” Proc. IRE, Vol. 45, February 1957,
pp. 187-196.
[9] Matthaei, G., L. Young, and E. M. T. Jones, Microwave Filters, Impedance-Matching
Networks and Coupling Structures, Dedham, MA: Artech House, 1980.
[10] Postoyalko, V., and D . Budimir, “Design of Waveguide E-Plane Filters With All-Metal
Inserts by Equal-Ripple Optimization,” IEEE Trans. Microwave Theory & Tech., Vol.
MTT-42, February 1994, pp. 217-222.
[11] DBFILTER Reference Manual, Tesla Communications Ltd., London, England.

Selected Bibliography
Bahl, I. J., and P. Bhartia, Microwave Solid State Circuit Design, New York: Wiley, 1988.
Bandler, J. W., and S. H . Chen, “Circuit Optimization: The State of the Art,” IEEE Trans.
Microwave Theory & Tech., Vol. MTT-36, 1988, pp. 424-443.
Chang, K., ed., Handbook of Microwave and Optical Components, Vol. 1, New York: Wiley,
1989.
Gaiewski, W. R., L. P. Dunleavy, and L. A. Geis, “Hybrid Inductor Modeling for Successful
Filter Design,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-42, No. 7, July 1994,
pp. 1426-1429.
Gupta, K. C., R. Gary, and R. Chadha, Computer-Aided Design of Microwave Circuits, Dedham,
MA: Artech House, 1981.
Hasler, M., and J. Neiryuck, Electrical Filters, Dedham, MA: Artech House, 1986.
Helszajn, J., Microwave Planar Passive Circuits and Filters, Chichester, England: Wiley, 1994.
Puglia, K., and M . Goldfarb, “Designing Lumped Element Bandpass Filters Using Coupled
Resonators,” Microwave J., September 1989, pp. 197-201.
Design of E-Plane Filters by
Optimization

Metallic structures placed in the E-plane of a waveguide along the waveguide

axis are widely employed in different microwave devices, in particular, frequency
selective units. The simplest such structures are waveguide E-plane metal insert
bandpass filters (Figure 6.1), with half-wave resonators coupled with one
another by means of longitudinally oriented, inductive diaphragms with the
rectangular cross-sections (metal septa). These filters offer the potential of
realizing low-cost, mass-producible, low-dissipation-loss mm-wave filters [1,2].
E-plane filters can be represented by a normalized equivalent circuit (Figure 6.2),
assuming TEjθ propagation only and neglecting higher order mode coupling
between adjacent E-plane septa. Normalization is with respect to the guide
impedance. The normalized reactances xs x pz∙ corresponding to the ⅛h septum
(z = 1, 2, . . . , n + 1) are functions of the length of the septum (⅛). Usually
the design of these filters is based on half-wave prototypes, with the septa being
related to impedance inverters.
The symmetrical T circuit, corresponding to a septum, symmetrically
embedded in a length of guide operates as an impedance inverter (Figure 6.3).
When the electrical length φ is chosen so that

φ = -tan~ 1 (2x⅛ + xa ) — tan 1 (xz2 ) (6.1)

the normalized ABCD matrix takes the form

0 jK
(6.2)
j/K 0

125
126 Generalized Filter Design by Computer Optimization

∣ure6.1 Structure of E-plane bandpass filters.

 Design of E-Plane Filters by Optimization 127

Figure 6.2 Normalized dominant mode equivalent circuit of E-plane filter.

0/2

Figure 6.3 7" circuit embedded in length of guide.

where

1 2*b
tan(2 tan K) = χλ
(6.3)
1 + 2xb* a + a

In the case of an E-plane septum, 7ζ defined by (6.3), has a nonlinear frequency

dependence and is not constant, although (6.2) has the form of the normalized
ABCD matrix of an impedance inverter. Applying (6.1) and (6.3) at the
“center” frequency of the specified passband to calculate the ⁄ z and ⅛ which
correspond to the impedance inverters given by Rhodes’ synthesis equations
for optimum Chebyshev distributed half-wave prototypes [3], is a common
method [1,4-6] in the design of E-plane filters. This approximate treatment
of the frequency dependence of (6.1) and (6.3) can result in a designed passband
that differs considerably from that which is specified. Therefore, to satisfy the
design specification [4], optimization is required to tune the filter dimensions.
The accurate calculation of the electrical parameters of E-plane septa requires
large computing resources, which makes the optimization of E-plane filters
128 Generalized Filter Design by Computer Optimization

prohibitive. Therefore, in this chapter, a new CAD algorithm is proposed that

employs the equal-ripple optimization method described in Chapter 4 for the
accurate design of E-plane metal insert bandpass filters. This method is more
suited to the problem than standard optimization because it requires less
frequency sampling. The elements of the equivalent circuit are calculated using
the mode-matching method described in Chapter 3.
Section 6.1 presents the procedure developed by Rhodes for the design
of direct-coupled cavity filter. The way in which that procedure attempts
to incorporate a linear frequency dependence of the cavity couplings also is
discussed.
Lim et al. [1] have attempted to include in an implicit way the actual
nonlinear frequency dependence of E-plane septa by applying a passband
correction scheme in conjunction with the Rhodes synthesis procedure. That
scheme is described in Section 6.2. Section 6.3 describes the application of
the equal-ripple optimization method (discussed in Chapter 4) to the design
of E-plane filters. The passband correction scheme described in Section 6.2 is
used to generate a starting point for the optimization.
A procedure for doing this, requiring only real scalar arithmetic, and
numerical aspects of equal-ripple optimization are discussed in Section 6.3.
Section 6.5 considers a design example and presents experimental results to
confirm the accuracy of the proposed optimization-based design procedure.
All numerical results presented in this work have been obtained using
EPFILTER, which is a field theory-based CAD program especially developed
for this kind of filter. The software [7] calculates the overall filter response in
terms of the even and odd input impedances of the filter structure at each
frequency and optimizes the filter parameters. The CAD synthesis software
EPFILTER was implemented on SUN workstations (SPARC 10) and IBM
PCs. The validity of the new method was confirmed by computer simulations
and experimental measurements of the filters designed by this method.

6.1 An Approximate Synthesis-Based Procedure for the

Design of Direct-Coupled Cavity Filters

The design procedure outlined in this section is based on a formulation proposed

by Rhodes for an optimum Chebyshev distributed stepped impedance lowpass
prototype [3]. It is a distributed filter (Figure 6.4a) consisting of a cascade of
n line elements (unit elements); each element corresponds to a resonator in
the conventional filter design. The elements, having characteristic impedances
Z r (r = 1, 2, . . . , ri) are assumed to have an equal length of λ ⁄2, where
is the guide wavelength of the line at the center frequency. The electrical
 Design of E-Plane Filters by Optimization 129

(a) o-----o-----o-----o ------o-----o-----o

Z 0 Z⁄ Z 2 Z< n Z n+⁄

(b)

Z' l 01 Z' Z'. n,n+l

(c)

12 K n.n+1

θl θι θ,

Figure 6.4 (a) Cascade of unit elements; (b) cascade of unit elements and inverters; (c) final
network after impedance scaling; (d) E-plane bandpass filter with metal insert;
and (e) equivalent circuit using Γ-equivalent for metal septum.

response of this transmission line structure depends on the impedances of

the unit elements. For electromagnetic waves propagating along the line, the
impedance differences between the unit elements yield reflected waves that,
after appropriate arrangement, cancel each other at the desired frequencies.
However, if a uniform waveguide is used to implement the circuit, all the unit
elements are of the same impedance; the necessary wave reflections must be
produced by inserting some sort of discontinuities between the unit elements.
One such example is where impedance inverters are inserted as the discontinu-
ities; the unit elements are assumed to have unit impedance as shown in
Figure 6.4b. For a cascade of unit elements (see Figure 6.4a), an optimum
equal ripple bandpass response occurs around θ = π when
130 Generalized Filter Design by Computer Optimization

j x sl jx χ ∣ j x s2 j∖2 rsn + 1, j jx ,
sn + 2

l-rl

J x l>ι J x l ,2 pn+l

(d)

H 12
(e) d1

Figure 6.4 (continued).

(6.4)
1 + 6 2 7' 2 (λ sin ff)

where

θ (6.5)

and

T tl(x) = cosh[w Arcosh(x)] for ∣x∣> 1 (6.6a)

T n {x) = cos[w arccos(x)] for 0 < x ≤ 1 (6.6b)

with

x = x sin θ (6.7)
 Design of E-Plane Filters by Optimization 131

is the mh degree Chebyshev polynomial of the first kind, a defines the passband
bandwidth (scaling parameter), 6 defines the passband ripple level, is the
guide wavelength at the center frequency, and λ g is the guide wavelength. In
this discussion, the discontinuities are assumed to be frequency independent,
which is hardly true in practice. The frequency-dependent behavior of the
discontinuity has significant effects on the filter performance. In an earlier
paper, Levy [8] studied in detail the reactance coupled filters and concluded
that the response in (6.4) should be modified as

I 12∣2 = (6.8)
~a Af
.2 1'τ-'2
n ——~ sin

to take into account the linear frequency dependence of the discontinuities.

To design a filter, the designer normally is given the two passband
edge frequencies yielding and A ⁄⁄, passband return loss (A/?), stopband
attenuation (Z√), the waveguide housing dimensions (a,b), and the metal-
septum thickness (r). The design procedure is as follows:

1. From and A ⁄⁄, determine the parameters a and A θ from the

following nonlinear equations, which readily can be solved numerically
[3]:

6Γ Ag£ /77 A βĎ
—---- sinl — ---- = 1
⅞0 ∖ AgL y
a
⅛H g . ⁄ 7r λ H∖
—; ---- sin —----- J = -1 (6.9)
A
¼ ∖ gH ⁄

where λ g χ and λ g ↑ are the guide wavelengths at the upper and lower
bandedge frequencies, respectively. Equation (6.9) comes from the
property of the Chebyshev polynomial of the first kind, which oscillates
between ÷1 and - 1 . Adding the two parts of (6.9) gives
a
⅛L . ⁄ 7r zĎgθĎ a λ g H ⁄π A Ď
■77777 sin — ---- + -77777- sin — ---- = 0 (6.10)
Δ(⁄1) ∖ λ g l, ) ∆(f 2 ) ∖ λ g H )

If A ι ≈ A 2 ≈ A ), then we can approximate (6.10) as

a A
gL ( ⅞<Λ g a A H
⁄ ⅞ 0∖
A ⁄ C∖ 1 Ď I ÷ A ⁄ C∖ I Ď I =
0 (6.11)
Δ(⁄1) Ď ∆(f 2 ) ∖ ⅛⁄
132 Generalized Filter Design by Computer Optimization

giving

¼ + X gH
¼= 2 (6.12)

However, for broad bandwidths, this will not be sufficiently accurate.

A better solution is obtained by applying the Newton-Raphson tech-
nique, as follows. Let

⁄ 7r ΛkΛ ⁄ 7γ z½oĎ
7
*(A gθ) = λ s i n ( - -— Ď+ λ sin(- — Ď (6.13)
∖ gl~, J ∖ g ⁄

Then differentiating w.r.t. λ g q gives

⁄ 7 Γ λ jχ)∖ / 7 7 λ f o∖
7
*(Agθ) = λ g L sin∣—7λ ----
l
I + λ g H sinl — ---- I (6.14)
∖ Z ⁄ ∖ A H
g ⁄

and the modified value of λ g Q will be

new
F*(λ f0 )
ι x ™ (r6 ι c∖
- ⅜ < ∙15 )

2. Determine the passband ripple level from the minimum passband

return loss, which is defined as:

Z,/? = 10 logi(⁄ 1 ÷ (6.16)

3. Determine the number of resonators n from (6.17). That is accom-

plished by finding the minimum value of n for which the most severe
constraint on the stopband insertion loss level (Z√) satisfies

ex λ g ⁄ 77
Lj = 10 log1 (J 1 + β 2 T 1n — - sin — — (6.17)
A
_ & ∖ g

at the designated stopband frequency f s. In (6.17), λ g is the guide

wavelength at f s.

4. For a cascade of unit elements with the transfer characteristic given

by (6.4), we can synthesize the network to obtain the characteristic
 Design of E- Plane Filters by Optimization 133

impedances of the unit elements. However, for up to moderate band-

widths, we can use explicit formulas for the element values [9]. Modi-
fying the network by introducing impedance inverters of the
characteristic impedances 7Cr>r+ i, as shown in Figure 6.4b, the explicit
formulas for the impedance values are

■2 . 2 rπ∖
v + sin —
⁄ 2 a . Γ ( 2 r - 1) π
Z r = — sin -------------
ι ∖ n
⁄
y 4 ay (2r + l)77
sin------ ------
2n

2 2 . { r - 1)77“|
y + sin -----------
n
r = 1, 2, . . . , n (6.18)
(2r - 3) τr
sin ------ ------
2n _

where

∣2 · 2 ⁄ z77Λ
A∖⁄·⁄ zy + sin ∖ —n I
K ,r r+ ↑ = ----------------------- r = 0, 1, . . . , n (6.19)

with

jy
• √ -1 sinh
= sinh ∙ k-ι -1 (6.20)
∖n €

5. In the rectangular waveguide realization, because the waveguide is

uniform, we must scale the internal impedance level, as shown in
Figure 6.4c, where

id
K...,- — r , o , 1, . . . , » (6.21)
' √4Λi

with

l
Z n = Z' n +χ = 1 (6.22)

6. Determine the ⅛h septum length, ⅛ (Figure 6.4d) by solving (6.3),

so that the required impedance inverter is realized. The normalized
134 Generalized Filter Design by Computer Optimization

reactances xa and λ⅛ (Figure 6.4e) are a function of the septum width.

Because these functions are not available explicitly, we must implement
a root-seeking routine to find the value of width that is provided by
the required impedance value K and the angle f for each impedance
inverter.
7. Finally, the length of the zth resonator (⁄ z∙) (see Figure 6.4e) formed
by zth and (z + l)th septa is given by

7τ
~ 2 Φi + Φi+l> (6.23)

where <⁄> z is given by (6.1), and the electrical distance, 77, corresponds
to the physical distance, λ φ⁄2.

The range of validity for the E-plane filter design is limited because of
the inaccurate approximation of the E-plane septa made in the derivation of
the design procedure. This problem can be reduced by using the method
proposed by Lim et al. [1]. That method is described in Section 6.2.

6.2 A CAD Algorithm Using the Passband Correction Method

This section describes the passband correction scheme proposed by Lim et al.
[4] to incorporate the actual nonlinear frequency dependence of the E-plane
septa into the Rhodes synthesis procedure.
This scheme compensates the linear frequency dependence of E-plane
septa to reduce the passband deviation problem associated with the procedure
described in Section 6.1. Equation (6.8) predicts the response of E-plane
bandpass filters more accurately if a modification of linear frequency dependence
by correction factors is introduced. Those correction factors implicitly include
the actual frequency dependence of the E-plane septa. In E-plane filters, the
correction factor cannot be expressed in an analytical form, because such a
quantity is a complicated function of both frequency and E-plane septa widths.
To define the passband correction factor, we used a numerical approach instead
of an analytical approach. Starting with a filter predesigned by the method
described in Section 6.1, we can express the passband correction factor as a
function of frequency only. In this algorithm, only two passband correction
factors are required and are calculated from the actual insertion losses (Z√) of
the predesigned filter at the specified bandedge frequencies (/£ and/#). When
the passband ripple, the lower and upper bandedge frequencies, and/#, and
 Design of E-Plane Filters by Optimization 135

the number of resonators are given, the insertion loss characteristic of a wave-
guide bandpass filter with a Chebyshev characteristic [1] is given by

.2 τ 2 *
Li = 10 k>gιo (6.24)
”\ m

where m = is the frequency dependence of the K-inverters. The factor

a and the guide-wavelength at the center frequency are determined by the
procedure given in Section 6.1.
The actual insertion losses and Z√⁄⁄ (in decibels) at fι and are
found by calculating the filter response by a field theoretical method, specifically
the mode-matching method developed in this work. In practice, the difference
between the septa width of a predesigned filter and those of a final filter is found
to be so small that the frequency dependence is assumed to be independent of
the septa width variations. Thus, starting with a filter predesigned by the
method given in Section 6.1, we can represent the passband correction factor
as a function of frequency only as described next.
It is assumed that the actual frequency dependence of the E-plane disconti-
nuity is M = m ∆ ( f ) , where Δ ( ⁄ ) is defined as the passband correction factor
that we try to find. Then the actual insertion losses (Z√) of the predesigned
filter are represented by

I e 2 2Γ a . ⁄∏⅛∖
Li actual = l θ ∙θglθi 1 + 7 m
'{,w' ( 4 ■_)
α / 7 7 AgφĎ
6 1
= l θ l°gl<J 1 + τ n (6.25)
W sin
Ď~V ⁄

where and a have the same values as in (6.24). At the specified bandedge
frequencies, (6.25) must satisfy the following condition:

10 ∙θgιo Lj actual i = L, H (6.26)

From (6.26), we can calculate the passband correction factors, Δ( ff) and Δ(⁄⅛),
at the specified bandedge frequencies as

Δ( ) = --------j---------- ----------------- i = L , H (6.27)

coshl - ln(c¾ ÷ y∣cst - 1 I
136 Generalized Filter Design by Computer Optimization

where

ιoo∙1Λ - ι
csti = -------z----- i = L> H (6.28)

Because our final goal is that the actual insertion loss-characteristics Z√ actua l
of the final E-plane filter have the same passband characteristics as the Z√ of
(6.24), which is the given specification, λ g Q and a in (6.25) must be recalculated
from (6.34) and (6.35), so that (6.25) satisfies the specified passband ripple,
10 logιθ(l + β 2 ), at both ⁄ jr and J⅛

□f λ gk . ⁄ '77' zĎgθĎ -gH . ⁄

(6.29)
Δ(⁄J ∖ λgL ) Δ(fH) AgH )

If λ g L ≈ λ g H ≈ λ g Q, then we can approximate as

z z
λgL ⁄ gθĎ -gH ( 4g()Ď
(6.30)
∆(⁄z) Ď 1
Af ⁄J +
{fH) y 1
λ g ∏)

giving

λ g L ∆(fκ) + λ g n ∆(⁄ z )
= (6.31)
*° ∆(⁄ z) + Δ(⁄ h )
For broad bandwidths, however, that is not sufficiently accurate. A better
solution is obtained by applying the Newton-Raphson technique, as follows.
Let

E∙*M A _ -AL s i∙n ( π λ Ď A H

g · ⁄∑⅛
F
⅝> " Δ(Λ) ( λgL ) + Δ(⁄h) sm ( λ gH (6.32)

Then differentiating w.r.t. Λ o gives

Γf 7 1 π
( π ⅛>∖ π

<⅛>1 = Λ<rd ∞ s
(τςr) ÷ ∆ζ⅛ c,,s
⁄ 77 ⅛∖
(⅛) (6.33)

and the modified values of and a are

ι new
gL (6.34)
F‰-)
 Design of E-Plane Filters by Optimization 137

a
new ∆(⁄z)
¼ . ⁄ 7r ⅛"'
-----
∖new
sin I —:-----
I A 1
ΛgO ∖ ζL ,

⅛fH)
(6.35)
λgH , (7τλ n
----- sin —: -----
∖n e w-----I Aχ lj
A g0 ∖ gH t

With the new and α, (6.25) accurately predicts the insertion loss characteris-
tics of the final filter around the passband, which are coincident with the
prediction of (6.24).

6.2.1 Design Procedure Using Passband Correction Method

For a given specification, the design procedure of E-plane bandpass filters using
the proposed passband correction method is summarized as follows:

1. Predesign the E-plane bandpass filter using the procedures given in

Section 6.1.
2. Using a field theoretical analysis method, calculate the actual insertion
losses Ljl and Z√⁄⁄ of the predesigned E-plane bandpass filter at the
specified bandedge frequencies ft and f∏.
3. Calculate the passband correction factors Δ(⁄J and Δ(⁄⅛), using
(6.27).
4. Calculate the new and the new a from (6.34) and (6.35), respec-
tively.
5. Finally, complete the design of the E-plane bandpass filter using the
procedure given in Section 6.1 with the new and a.

To illustrate the application of this procedure to the design of E-plane filters,

the design of an E-plane bandpass filter with the specifications given in Section
6.6 is considered. Figure 6.5 shows the calculated passband insertion loss of
an E-plane bandpass filter designed using the procedure. As can be seen, the
design specification still is unsatisfactory, and optimization often is required
in practice for the accurate design of these filters.
138 Generalized Filter Design by Computer Optimization

Insert thickness = 0.25 mm

0.10 septum lengths resonator lengths

Insertion loss (dB)

0.09 d1=0.9937 mm L1=16.0226 mm
0.08 d2=5.7368 mm L2=1 6.4227 mm
0.07 d3=7.2106 mm L2=1 6.4434 mm
0.06

0.05

0.04

0.03

0.02

0.01
0.00 - - hAAnj
9.2 9.3 9.4 9.5 9.6 9.7 9.8

Frequency (GHz)

Figure 6.5 Calculated insertion loss before optimization.

6.3 Numerical Implementation of Equal-Ripple Optimization

6.3.1 Calculation of Insertion Loss

To design E-plane filters for a given set of metal insert dimensions using the
equal-ripple optimization method described in Chapter 4, it is necessary to be
able to calculate the insertion loss on a sample of frequency points within the
specified passband. In the case of longitudinally symmetrical structures, such
as conventional E-plane bandpass filters, the insertion loss (Z√) can be expressed
in terms of normalized even- and odd-mode impedances as

0
L, - 20 log∣0 *z z
' *°> (6.36a)
e o

where jz e is the normalized input impedance of the two identical one-ports

formed by placing a magnetic (electric) wall at the plane of symmetry. By
transforming an open (short) circuit placed at the plane of symmetry through
the filter sections (resonators and E-plane septa) located to the left of the plane
of symmetry, z e ( 0 ) can be calculated. Each E-plane septum is itself symmetrical
and can be electrically represented by normalized even- and odd-mode imped-
ances

¾ = y(⅞∙ + 2x?/ ) (636b)

Z0i = j ×si (6.36c)

 Design of E' Plane Filters by Optimization 139

For a normalized reactive impedance jz, an E-plane septum performs the

normalized impedance transformation jz => y⅜, where Z{ n is given by

z (6.36d)
in =
1Z + (¾ + ⅞ z )

A resonator section, that is, a length of guide, performs the normalized imped-
ance transformation jz => j¾ 2, where z±n is given by

z + tan βl
zin = (6.36e)
1 — z tan βl

where β (= 2τrlλ is the propagation constant, and I is the length of the

resonator. By applying (6.36d) and (6.36e), it is possible to calculate z e and
z 0 starting at the center of the filter and working outward. That process involves
no matrix manipulation and uses only real arithmetic. For the analysis of
E-plane septa in a rectangular waveguide, the mode-matching method described
in Section 3.1 was used.
Neither accurate numerically fitted closed- form expressions nor accurate
design tables for the electrical parameters of E-plane septa in terms of dimensions
(length and thickness) and frequency are yet available [5,10]. Therefore, the
accurate design of E-plane filters requires the direct calculation of the electrical
parameters of E-plane septa and highlights the need for the design of
E-plane filters by optimization techniques. Optimization techniques minimize
the number of calculations of the electrical parameters of E-plane septa.

6.3.2 Solution of Nonlinear Equations

For the solution of a system of nonlinear equations, the Newton-Raphson

method is a rapidly convergent technique if a good initial approximation is
available. A good approximate design of an E-plane filter can be obtained by
the procedure given in Section 6.2. It is therefore adopted in this work as a
means of generating a starting point for the solution of a system of n + 1
nonlinear equations in n + 1 variables [see (4.13)] by means of the Newton-
Raphson method.
The number of times the function is evaluated in the process of finding
its root is the usual measure of computational effort. That includes function
evaluations required to calculate derivatives numerically. The most complex
task in the Newton-Raphson method is evaluating the Jacobian / of jE(x). The
Jacobian matrix of the n + 1 nonlinear functions can be calculated numerically
using finite difference as
140 Generalized Filter Design by Computer Optimization

3 Ei Ei(×j + δ xβ - Ei (xj)
r r 1
∣Λ∣■ aτ ≡ -----------« 7 ----------
Ď√ ∙Ay VZ *Ay

i = 1, 2, . . . , n + 1 ; ; = 1, 2, . . . , n + 1 (6.37)

The finite difference calculation of the Jacobian matrix for a given set of filter
parameters requires the evaluation of

Λz∙(x1 , x 2 > ∙ ∙ ∙ > 5 + <5×j, . . . » ⅞ + l)

i = 1, 2, . . . , n + 1; j' = 1, 2, . . . , n + 1 (6.38)

where %ι, %2> ∙ ∙ ∙ > ⅞+ι denote the n + 1 dimensions required to specify a
symmetrical E-plane filter. For odd n∖

. n +1
×i = d± ι = 1, 2, . . . , —~— (6.39)

(w+l) n +3 n +5
(6.40)

For even n:

xl = d l z = l , 2, . . . , - + l (6.41)

( n + ∖) . n n
×i = 4 - — 2— z = - ÷ 2 , - + 3, . . . , w + l (6.42)

Denoting by x and E the n + 1 dimensional vectors with components

x z∙(z = 1, 2, . . . , n + 1) and 2⅛(⁄ = 1, . . . , n + 1), the Newton-Raphson
method has the general form [11]

k 1 1 1
X = X (√" )E(Λ (6.43)

where k is the iteration number {k = 1, 2, . . .) and J ~ is the inverse of the

w-by-zw Jacobian matrix evaluated at x 1 . After each iteration, the response
and errors are computed again with the new corrected parameters, until the
errors are judged to be sufficiently small.
To minimize the CPU time required to evaluate the Jacobian matrix,
the normalized even- and odd-mode impedances corresponding to the septum
 Design of E-Plane Filters by Optimization 141

lengths at the start of a Newton-Raphson iteration should be calculated at the

sample frequencies and stored. That avoids unnecessary recalculation of
the parameters when the insert dimensions are perturbed in turn to calculate
the elements of the Jacobian matrix.

6.4 Numerical and Experimental Results

To verify the accuracy of the developed method, the design of an E-plane

bandpass filter with the following specifications was considered:

• Waveguide WG16 (WR90) internal dimension: 22.86 mm by

10.16 mm;
• Midband frequency: 9.5 GHz;
• Passband: 9.25-9.75 GHz;
• Ripple level: 0.05 dB;
• Number of resonators: 5;
• Metal insert thickness: 0.250 mm;
• Passband return loss: 20 dB min;
• Filter characteristic: Chebyshev.

The calculated passband insertion loss of an E-plane filter designed using

the approximate method given in Section 6.2 is shown in Figure 6.5. Also
included in that figure are dimensions of the E-plane insert. In both the design
and the calculations, mode matching with 100 modes was used. As a starting
point for equal-ripple optimization, this approximate design was used. The
passband insertion loss calculated using the insert dimensions obtained on
convergence is given in Figure 6.6. This took four Newton-Raphson iterations.
Mode matching with 100 modes was used throughout the optimization.
The calculated insertion loss of the final design over the whole waveguide
band is shown in Figure 6.7 (dashed line). This figure (solid line) also shows
a plot of (6.8) with n = 5, β 2 = 10, 0θ5 - 1, = 43.808 mm and a = 6.344.
Although the passband to stopband discrimination of the designed filter is
similar to that predicted by (6.8), it has inferior stopband attenuation in the
upper part of the waveguide band. That limitation of conventional E-plane
filters is well known [2,7,10]. Improved stopband attenuation can be achieved
by using solutions presented in Chapter 7. To evaluate the effect of higher
order mode interaction on this design, the passband insertion loss was recalcul-
ated, again using the dimensions given in Figure 6.6. In that recalculation of
142 Generalized Filter Design by Computer Optimization

Insert thickness = 0.25 mm

O.1O
septum lengths resonator lengths
0.09
d1=1.1119 mm L1=16.0482 mm
0.08 d2=5.9275 mm L2=1 6.4223 mm
0.07 d3=7.2654 mm L3=1 6.4472 mm
0.06

0.05

0.04

0.03

0.02 -

0.01

0.00 —
9.2

Frequency (GHz)

Figure 6.6 Calculated insertion loss after optimization.

100.00 -∣
Equation (6.8)
Design
80.00 -

©
60.00 -
.≡
40.00 -

20.00 -

0.00
8.00 9.00 10.00 11.00 12.00 13.00
Frequency (GHz)

Figure 6.7 Comparison of calculated insertion loss with (6.8).

the passband insertion loss, the TE30 coupling between E-plane septa was taken
into account. The recalculated insertion loss displayed no noticeable difference
from that shown in Figure 6.6.
The measured insertion loss of the fabricated design of a five-resonator
E-plane bandpass filter is shown in Figure 6.8. The measurement was made
using an H P 85 10C vector network analyzer. A full two-port calibration was
used. Waveguide standards calling for a short, offset short, sliding load, and
through were used during the calibration. The filter design was fabricated using
 Design of E-Plane Filters by Optimization 143

100.00-∣

Insertion Loss (dB)

80.00 - Theory
Measurement

60.00 -

40.00 -

20.00 -

0.00 -----
8.00 9.00 10.00 11.00 12.00 13.00
Frequency (GHz)

Figure 6.8 Measured (dashed line) and calculated (solid line) insertion loss.

a brass waveguide housing and a copper insert. The metal insert was realized
by using spark erosion. The measured insertion loss in the passband was less
than 0.45 dB. The discrepancies between the measured insertion loss and
theory within the passband are due to ohmic losses in metallic walls and metal
insert.
That the response of the fabricated filter is equal ripple can be seen from
the measured return loss, which is shown in Figure 6.9. Compared with the
theoretical prediction, the passband of the fabricated filter is shifted downward
slightly. Considering the inaccuracies involved in the practical realization of
E-plane filters [2,7,10], the agreement between theory and measurement is
very good. The comparison of the measured and simulated filter responses
shows slight disagreement (bandpass is shifted). That disagreement may be due
to spark erosion errors.
Figure 6.10 is a photograph of a five-resonator metal insert E-plane
bandpass filter 1 structure together with the corresponding waveguide housing
at X-band.
144 Generalized Filter Design by Computer Optimization

50.00 -

30.00 -

20.00 -

10.00-

0.00- -
Γ- -
1 1
I
9.20 9.60 10.00
Frequency (GHz)

Figure 6.9 Measured return loss.

Figure 6.10 Photograph of conventional E-plane bandpass filter.

 Design of E-Plane Filters by Optimization 145

References

[1] Lim, J. B., C. W. Lee, and T. Itoh, “An Accurate CAD Algorithm for E-Plane Type
Bandpass Filters Using a New Passband Correction Method Combined With the Synthesis
Procedures,” IEEE MTT-S Int. Microwave Symp. Dig., June 1990, pp. 1179-1182.
[2] Arndt, F., “The Status of Rigorous Design of Millimetre Wave Low Insertion Loss Fin-
Line and Metalic E-Plane Filters,” J. Instn. Electronics and Telecom. Engrs., Vol. 34,
No. 2, 1988, pp. 107-119.
[3] Rhodes, J. D., “Microwave Circuit Realizations,” in Microwave Solid State Devices and
Applications, D. V. Morgan and M. J. Howes, eds., England: Peregrinus, 1980,
pp. 49-57.
[4] Shih, Y. C., “Design of Waveguide E-Plane Filters With All Metal Inserts,” IEEE Trans.
Microwave Theory & Tech., Vol. MTT-32, July 1984, pp. 695-704.
[5] Hong, J. S., “Design of E-Plane Filters Made Easy,” IEE Proc., Vol. 136, Pt. H , 1989,
pp. 215-218.
[6] Bui, L. Q., D. Ball, and T. Itoh, “Broad-Band Millimeter-Wave E-Plane Bandpass
Filters,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-32, December 1984,
pp. 1655-1658.
[7] DBFILTER Reference Manual, Tesla Communications Ltd., London, England.
[8] Levy, R., “Theory of Direct Coupled Cavity Filters,” IEEE Trans. Microwave Theory &
Tech., Vol. MTT-15, June 1967, pp. 340-348.
[9] Rhodes, J. D., Theory of Electric Filters, New York: Wiley, 1976.
[10] Gololobov, V. P., and M. Yu. OmeΓyanenko, “Bandpass Filters Based on Planar Metal-
Dielectric Structures in the E-Plane of a Rectangular Waveguide (A Review),” Radioelec-
tronics and Communications Systems, Vol. 30, No. 1, 1987, pp. 1-15.
[11] Ortega, J. M., and W. C. Rheinboldt, Iterative Solution of Nonlinear Equations in Several
Variables, New York: Academic Press, 1970.

Selected Bibliography
Cohn, S. B., “Direct Coupled Resonator Filters,” Proc. IRE, Vol. 45, February 1957,
pp. 187-196.
Konishi, Y., and K. Uenakada, “The Design of Bandpass Filter With Inductive Strip-Planar
Circuit Mounted in Waveguide,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-22,
October 1974, pp. 869-873.
Levy, R., “Tables of Element Values for the Distributed Low-Pass Prototype Filter,” IEEE
Trans. Microwave Theory dr Tech., Vol. MTT-13, September 1965, pp. 514-536.
Matthaei, G., L. Young, and E. M. T. Jones, Microwave Filters, Impedance— Matching Networks
and Coupling Structures, Dedham, MA: Artech House, 1980.
Politi, M., et al., “An Equivalent Circuit for the Design of E-Plane Metal-Insert Filters in
Millimeter-Wave Applications,” Proc. 20th European Microwave Conf, Budapest, Hungary,
1990, pp. 1257-1262.
146 Generalized Filter Design by Computer Optimization

Rhodes, J. D., “Design Formulas for Stepped Impedance Distributed and Digital Wave Maxi-
mally Flat and Chebyshev Low-Pass Prototype Filters,” IEEE Trans. Circuits Syst.i Vol.
CAS-22, November 1975, pp. 866-874.
7
Design of Ridged Waveguide Filters by
Optimization

All-metal inserts placed in the E-plane of a rectangular waveguide along the

waveguide axis offer the potential of realizing low-cost, mass-producible, and
low-dissipation-loss mm-wave filters [l-3]. However, despite their favorable
characteristics, the attenuation in the second stopband (i.e., where the resonators
are about one wavelength long) often may be too low and too narrow for
many applications, such as for diplexers, in which frequency selectivity and
high stopband attenuation are considered to be important filtering properties.
That is due to the effect that is characteristic for the common single insert
design: beyond the cutoff frequency of the fundamental mode within the
septum section, which is determined by the distance between the septa and
the waveguide sidewalls, the power is increasingly transported directly by propa-
gating waves [4], causing degradation of the selective properties of resonators
based on two septa and the waveguide between them.
In recent years, much effort has been devoted to the study of E-plane
bandpass filters with improved stopband performance. Several solutions have
been proposed:

• Reduce the distance between the metal insert and the waveguide side-
walls by a thick metal insert [5] or use several metal inserts rather
than a single insert [5-7] . The first approach achieves good stopband
performance, as is indicated in [5], but it has the disadvantage of
high passband insertion loss. The latter approach also achieves good
stopband performance, but it requires greater effort in the mounting
and adjustment of the several metal inserts in the waveguide.

147
148 Generalized Filter Design by Computer Optimization

• Use narrower or wider waveguide for the filter section, depending on

the position of the filter passband within the waveguide single-mode
bandwidth [4]. The reduction of the waveguide housing width
decreases the distance between the metal insert and the waveguide
sidewalls so that the propagation of modes along the coupling subsec-
tions is suppressed up to higher frequencies. That results in an improved
stopband attenuation when the passband is not the high end of the
waveguide band. Increasing the waveguide housing width reduces the
guide wavelength of the filter resonators, resulting in an improved
stopband performance when the passband is close to the cutoff fre-
quency of the standard width guide.
• Use resonators of different cutoff frequency [8,9] . That is achieved by
employing sections of rectangular waveguide in which all the waveguide
sections are resonant at the same fundamental frequency. However,
due to different guide wavelengths in the different sections, the sections
are not all simultaneously resonant at any higher frequency, resulting
in an improved stopband attenuation.

This chapter investigates a new solution for improvement in the second stop-
band. Section 7.1 presents the proposed filter configuration. Section 7.2
describes the circuit representation and design procedure for the design of
E-plane bandpass filters with improved stopband performance (ridged wave-
guide filters). Section 7.3 discusses the efficient computer implementation of
equal-ripple optimization, while Section 7.4 considers design examples to con-
firm the improvement and presents experimental results to confirm the accuracy
of the design procedure.

7.1 Proposed Filter Configuration

Based on the existing idea to use different waveguide resonators with different
characteristic impedances and different cutoff frequencies [4] , improved stop-
band performance can be met by the E-plane filter configurations shown in
Figure 7.1.
That improved performed is due to the nonlinear relation between guide
wavelength and frequency, which can be influenced favorably by a suitable
reduction of the cutoff frequency of the fundamental mode within the filter
resonators. The proposed filter configuration is constructed of direct-coupled
ridged waveguide sections that have, in general, identical cutoff frequencies
and characteristic impedances, and reactive elements (metal septa) arranged in
 Design of Ridged Waveguide Filters by Optimization 149

Resonators (ridged waveguide) E-plane septa

Figure 7.1 Configuration of ridged waveguide filter structure

such a manner that each section is resonant at the same fundamental frequency.
The main features of the new structure are the use of conventional rectangular
waveguide housing and the use of a metal insert that, when mounted, introduces
ridges in the resonators. The improvement in the upper stopband associated
with the superior electrical performance of ridged waveguide, such as cutoff
frequency reduction, provides a convenient way to realize E-plane bandpass
filters with improved stopband performance. The structure is simple and com-
patible with the E-plane manufacturing process.

7.2 Circuit Representation and Design Procedure

7.2.1 Circuit Representation

The proposed filter structure in Figure 7.1 can be represented as shown in
Figure 7.2 by use of the asymmetrical impedance inverter illustrated in
Figure 7.3. The design of these filters usually is based on the design procedure
described in Section 7.2.2, with the septa being related to impedance inverters.
Figure 7.3 shows a rwo-port defined by its ABCD matrix. We assume that it
150 Generalized Filter Design by Computer Optimization

z0

Figure 7.2 Impedance inverter.

Zl K n.n+1

Figure 7.3 Distributed half-wave prototype using impedance inverters.

is connected to lines of characteristic impedances Z∖ and Zγ at its two-ports.

It is always possible to find reference planes P∖ and 7¾ a t electrical distances
φ∖ and 02 from the respective ports to convert the asymmetric two-port (metal
septum in rectangular waveguide between two different ridged waveguides) to
the symmetric impedance inverter K The equations giving the reference plane
locations are in terms of the ABCD matrix parameters, leading to the results
that follow.
The normalized element value of the impedance inverter (see Figure 7.3)
can be derived directly from the normalized ABCD matrix of the septum.
Normalization is with respect to the guide impedance of the rectangular wave-
guide. It is given by
 Design of Ridged Waveguide Filters by Optimization 151

⁄r=√zΓ¾(√Σ-√F7 T) (7.1)

with

A = 1 + ∣[( - d) 1 + ( b - e)2 ] (7.2)

where a, b> c, and d are normalized elements of the ABCD matrix given as

a = 21λS b= (7.2
Vi √ZfZ2
⁄γ
c - C∖∣Z↑ ∑2 d - (7.4)

and Z∖ and Z are normalized guide impedances of the ridge waveguide

resonator sections, which are frequency dependent.
The reference plane locations are given by the equations

tan(2 ≠1) - 2 ,2x /»2 2× (7.5)

∖a — a ) + {b — c )

tan(2φ2 ) - 2 2x ⁄,2 2λ
{d - a ) + {b - c )

In practice, electrical distances (<⁄>ι and φf) can be realized as negative values
in the adjacent positive line length, which therefore becomes shortened in the
final network. Mathematical details of the derivations (K, φ↑, and </>2) are
given in the appendix to this chapter. Although an impedance inverter has the
form of a normalized ABCD matrix, in the case of an E-plane septum, K, as
defined by (7.1), is not constant and has a nonlinear frequency dependence.

7.2.2 Design Procedure

A common approach to the design of the conventional E-plane bandpass filters

(as described in [10—12]), can be used with minor modifications for filter
structures with different impedances and cutoff frequencies such as E-plane
bandpass filters with improved stopband performance. Here, only the most
important steps in the design procedure, which should include the concept of
 152 Generalized Filter Design by Computer Optimization

impedance inverters and impedance scaling of the impedance levels of the

prototype filter, are presented. The design procedure for the E-plane bandpass
filters with improved stopband performance is to apply (7.1), (7.5), and (7.6)
at the center frequency of the specified passband to calculate ⁄ z and ⅛ which
correspond to the impedance inverters. This approximate treatment of the
frequency dependence of (7.1), (7.5), and (7.6) can result in a designed passband
that differs considerably from that which is specified, and optimization is then
required to tune the filter dimensions to satisfy the design specification.
For a given filter specification such as the two passband edge frequencies
yielding and Λ ⁄y, passband return loss (Lr), stopband attenuation (Z⁄),
the waveguide housing dimensions (a,l>), the metal septum thickness (zu), and
the ridged waveguide gap (∫), the modified design procedure is summarized as
follows:

1. Determine the modified scaling parameter a and midband wavelength

A ) from the following nonlinear equations, which can be readily
solved numerically.

⅞θ ∖ λgL ⁄

6“ ∙ ⁄ <5 ∖ v

a- — sin — ---- = —1 1
(7.7)
AgO ∖ ⁄

2. Determine the passband ripple level 6 from the minimum passband

return loss, which is defined as:

Lr = 10 log 1 0 l 1 + — (7.8)

3. Determine the number of resonators, n, from

3
Li = 10 log 1 0 < 1 + e 2 T 1n (7.9)

at the designated stopband frequency, f s.

4. Calculate the impedances of the distributed element and impedance

inverter values from
 Design of Ridged Waveguide Filters by Optimization 153

2 · 2↓ rτr
' + sin —
⁄ 2α . (2r - 1) 77 ∖n ]
Z7 r = — sin
y 2n 4 ay l . (2r + l)τr j
sin

2 . 2(r- i)w
1- y + sin ----------- -1
n
l . (2r-3)τr ,
Qin --------------- -

and

2 ∙ 2⁄ —w7γ
' + sin
n
(7.11)

where

jv
• √ -1 sinh
= sinh ∙ k-* -r (7.12)
∖n 6

5. Recognizing that the normalized impedances of the resonators of the

ridged waveguide filters (Z r , r = 1, . . . , ri) are not identical to 1, we
must scale the internal impedance levels of the distributed half-wave
prototype filter in Figure 7.4(a) as shown in Figure 7.4(b) , where

Z* = Z r 7 = 0 , 1, . . . , n + 1 (7.13)

and

=
⁄ Z r Z r+ ι
K r ,r,+∖ K r ,r,+↑ V yl yl r = 0, 1, . . . , n (7.14)
Ď r r+ j

with

Z o = Z n+ 1 = Z ,o = Z ln+∖ = 1
(7.15)

Z tr and are the impedance values of the distributed half-wave

prototype filter (see Figure 7.4) given by (7.10) and (7.11), and the
154 Generalized Filter Design by Computer Optimization

Z', 01 Z'⁄ «',2 Z'⁄ z', n,n+l

Z ' n+1
zl

K*n z* K -2 z* Z* κ*n.n+1
λ

Figure 7.4 Final network (a) using impedance inverters after impedance scaling and (b) of
the distributed half-wave prototype.

values of Z r are the normalized guide impedances of the resonator

sections.

6. Determine the zth septum length ⅛, shown in Figure 7.5(b), by solving

(7.1) so that the required impedance inverter is realized. Because those
functions are not available explicitly, we must implement a root-
seeking routine to find the value of width that is provided by the
required normalized impedance value K for each impedance inverter.
7. Finally, the length (⁄ z∙) of the zth resonator (see Figure 7.5b) formed
by the zth and (z ÷ l)th septa is given by

k = ⅛-
2 77
{φ2,i+ √>1,Z+ 1)] i = ∖, 2 , . . . , n (7.16)

where φ↑ and φ% are given by (7.5) and (7.6), and the electrical
distance θ corresponds to the physical distance Λ θ⁄2.

The main limitations of this approach are the frequency dependence of the
guide impedances and the frequency dependence of the impedance inverters
or the inaccurate approximation of the E-plane septa made in the derivation
of the design procedure. Once the dimensions of the filter have been found,
the frequency response of the overall filter at each frequency can be simulated
by cascading the ABCD matrices of the resonators and the septa. To illustrate
 Design of Ridged Waveguide Filters by Optimization 155

Figure 7.5 (a) Two-port representation of ridged waveguide filter using ABCD matrix for
metal septum and (b) symmetrical ridged waveguide bandpass filter.

the application of this procedure to the design of E-plane bandpass filters with
improved stopband performance, the design of an E-plane bandpass filter with
the specifications given in Section 7.4 is considered. Figure 7.6 shows the
calculated passband return loss (before optimization: the solid line) designed
using this procedure. As can be seen, the design specification is still unsatisfac-
tory, and optimization often is required in practice for the accurate design of
these filters.

7.3 Numerical Implementation of Equal-Ripple Optimization

To apply the equal-ripple optimization technique described in [1] to the design
of E-plane filters with improved stopband performance, it is necessary, for a
given set of insert dimensions, to be able to calculate the insertion loss on a
sample of frequency points within the specified passband. For a symmetrical
E-plane bandpass filter with improved stopband performance, the insertion
loss can be expressed in terms of an ABCD matrix. The matrix representation
of the whole filter (see Figure 7.5) is
156 Generalized Filter Design by Computer Optimization

80

70

60
Return loss (dB)

50

40

30

20

10

0—
9.0 9.2 9.4 9.6 9.8 10.0
Frequency (GHz)

Figure 7.6 Calculated return loss before (solid line) and after (dashed line) optimization (rwg
gap = 8 mm).

COS θ∖
JB∖ pl
=
D \jC }

COS θ∖

A2
jC 2
COS θ2

cos θ n sin θπ ∖
l I pn÷ι
I ' A 1
cos θn

(7.17)
 Design of Ridged Waveguide Filters by Optimization 157

with

2 π lt
(7.18)
⅛i

in which ⁄ z∙ is the ridged waveguide resonator length, is the wavelength in

the ridged waveguide resonator for each frequency, and f c is the cutoff frequency
in the ridged waveguide resonator. The overall filter response (insertion loss,
Lj) can be expressed in terms of elements of the total ABCD matrix of the
filter at each frequency (by directly combining the ABCD matrices of the
individual filter sections) as

⁄ A + B + C + Z)Ď
L1 = 20 log1 0 l --------- --------- J (7.19)

The elements of the ABCD matrices of the individual filter sections are calcu-
lated using the mode-matching method [13,14]. The propagation constants
of the eigenmodes in ridged waveguides are related to the cutoff frequencies,
which can be calculated according to [15]. The transcendental equation of the
eigenvalue of the zτth mode in ridged waveguide was solved numerically. How-
ever, due to the singular behavior of the magnetic field at the edges of the
septa, a large number of modes need to be included in the field expansions
to ensure good convergence. That is similar to the situation for the septum in
a rectangular waveguide and is due to the singular behavior of the magnetic
field at the edges of the septum.
Neither accurate numerically fitted closed-form expressions nor accurate
design tables for the electrical parameters of the E-plane septa in terms of
septum dimensions (length and thickness) and frequency are yet available. The
accurate design of ridged waveguide filters thus requires the direct calculation
of the electrical parameters of E-plane septa. That highlights the need in the
optimized design of those filters for optimization techniques that minimize
the number of calculations of the electrical parameters of E-plane septa. A
good approximate design of ridged waveguide filters can be obtained by the
procedure described in Section 7.2. This procedure tries to include implicitly
the actual frequency dependence of the E-plane septa and results in passbands
that nearly meet design specifications. It is, therefore, adopted in this chapter
as a means of generating a starting point for the optimization.
158 Generalized Filter Design by Computer Optimization

7.4 Numerical and Experimental Results

To illustrate the accuracy of the developed method for the design of symmetrical
and asymmetrical ridged waveguide filters, a few ridged waveguide bandpass
filters in WG16 were designed. Table 7.1 shows the specifications of symmetri-
cal ridged waveguide bandpass filter, with ridged waveguide gap (rwg gap) =
8 mm.
Figure 7.7 shows the calculated passband return loss (solid line) of ridged
waveguide filter (rwg gap = 8 mm) using the approximate method described
in Section 7.2 and in [16]. Mode matching with 100 modes is used in both
the design and the calculations. This approximate design was used as a starting
point for equal-ripple optimization. The passband return loss calculated using
the insert dimensions obtained on convergence also is shown in Figure 7.7
(dashed line). Table 7.2 lists the insert dimensions before and after optimization,
which took four iterations. Mode matching with 100 modes was used through-
out the optimization. Figure 7.6 showed the calculated insertion loss (solid
line) of the final design of a five-resonator ridged waveguide bandpass filter
(with rwg gap = 8 mm) over both the X-band (8.2 to 12.4 GHz) and the Ku-
band (12.4 to 18 GHz). Also included in Figure 7.6 is a plot of the measured
insertion loss (dashed line) of the fabricated design. The designed filter was
fabricated using a brass waveguide housing and a copper metal insert, which
was realized using spark erosion. Very good agreement between theory and
experiment was observed. The measurement over both the X-band and the
Ku-band was made using an H P 85 10C vector network analyzer in two steps.
Step 1 was the measurement of the frequency responses of the fabricated
filter between 8.2 and 12.4 GHz. A full two-port calibration with short, offset
short, sliding load, and through as waveguide standards was used.

Table 7.1
X-Band Five-Resonator Ridged Waveguide Bandpass Filter Specifications

Waveguide WG16 (WR90) internal dimensions: 22.86 mm by 10.16 mm

Midband frequency: 9.50 GHz
Passband: 9.25-9.75 GHz
TEjq cutoff frequency in rectangular waveguide: 6.556 GHz
T∈ιo cutoff frequency in ridged waveguide: 6.450 GHz
Passband return loss: 20.00 dB
Ripple level: 0.050 dB
Number of resonators: 5
Metal insert thickness: 0.100 mm
Ridged waveguide height: 8.000 mm
 Design of Ridged Waveguide Filters by Optimization 159

100.00

Insertion Loss (dB)

80.00 -

60.00 -

40.00 -

20.00 -

0.00 ------
8.00 10.00 12.00 14.00 16.00 18.00

Frequency (GHz)

Figure 7.7 Measured (dashed line) and calculated (solid line) insertion loss of X-band five-
resonator ridged waveguide bandpass filter (rwg gap = 8 mm).

Table 7.2
Symmetrical Ridged Waveguide Bandpass Filter
rwg gap = 8.00 mm
Insert thickness = 0.10 mm

Septum Lengths (mm) Resonator Lengths (mm)

Before After Before After
Optimization Optimization Optimization Optimization

d∖ = d6: 1.3528 1.1507 ∣↑ = ⅛ 15.8109 15.7789

d2 - ⅛ 6.4127 6.0724 ∣2 = /4: 16.0920 16.1070
⅛ = dι- 7.7719 7.4306 16.1032 16.1274

Step 2 was the measurement of the fabricated filter frequency response

between 12 and 18 GHz. Because the cutoff frequency of the TE20 mode is
about 13.20 GHz, that mode becomes a propagating mode for frequencies
above 13.20 GHz. By using stepped WG18⁄WG16 transitions, the generation
of the TE20 mode can be avoided. That is attributable to the fact that as a
result of the symmetry of the transition, the TE20, like all even modes, is not
actually generated in the resonator waveguides. A full two-port through-reflect-
line (TRL) calibration was used [17]. The only standards required for such
calibration are a reflect (short), 20-mm through line, and a 25-mm line.
160 Generalized Filter Design by Computer Optimization

Figure 7.8 is a photograph of a five-resonator ridged waveguide bandpass filter

(with rwg gap = 8 mm) together with the corresponding waveguide housing
at X-band.
To demonstrate the advantages of the ridged waveguide bandpass filters
over the conventional E-plane bandpass filters, a five-resonator X-band conven-
tional E-plane bandpass filter (Table 7.3) and two ridged waveguide bandpass
filters (in which the widths of the ridges are arbitrarily chosen) (Tables 7.4
and 7.5) have been designed.
Figure 7.9 shows the calculated passband return loss (solid line) of ridged
waveguide filter (rwg gap = 5 mm) using the approximate method described
in Section 7.2. Mode matching with 100 modes is used in both the design
and the calculations. The approximate design was used as a starting point for
equal-ripple optimization. The passband return loss calculated using the insert
dimensions obtained on convergence (after optimization) also is shown in
Figure 7.9 (dashed line). Table 7.6 lists the insert dimensions before and after
optimization.

Figure 7.8 Photograph of ridged waveguide bandpass filter.

 Design of Ridged Waveguide Filters by Optimization 161

Table 7.3
Conventional E-Plane Bandpass Filter

Waveguide WG16 (WR90) internal dimension: 22.86 mm by 10.16 mm

Midband frequency: 9.5 GHz
Passband: 9.25-9.75 GHz
Number of resonators: 5
Metal insert thickness: 0.10 mm
Passband return loss: 20 dB min
Filter characteristic: Chebyshev

Table 7.4
Symmetrical Ridged Waveguide Bandpass Filter 1

Waveguide WG16 (WR90) internal dimension: 22.86 mm by 10.16 mm

Midband frequency: 9.5 GHz
Passband: 9.25-9.75 GHz
Number of resonators: 5
Metal insert thickness: 0.10 mm
Ridged waveguide gap: 8.00 mm
Passband return loss: 20 dB min
Filter characteristic: Chebyshev

Table 7.5
Symmetrical Ridged Waveguide Bandpass Filter 2

Waveguide WG16 (WR90) internal dimension: 22.86 mm by 10.16 mm

Midband frequency: 9.5 GHz
Passband: 9.25-9.75 GHz
Number of resonators: 5
Metal insert thickness: 0.10 mm
Ridged waveguide gap: 5.00 mm
Passband return loss: 20 dB min
Filter characteristic: Chebyshev

This took four iterations. Mode matching with 100 modes was used
throughout the optimization. Figure 7. 10 shows the comparison between calcu-
lated insertion losses of the conventional (solid line) and the ridged waveguide
bandpass filter (rwg gap = 8 mm) (dashed line). Table 7.7 lists the insert
dimensions for conventional E-plane and ridged waveguide bandpass filters.
Figure 7.11 shows the comparison between calculated insertion losses of
the conventional (solid line) and the ridged waveguide bandpass filter
162 Generalized Filter Design by Computer Optimization

40-
Return loss (dB)

30-

20-

10-

0—
9.0 9.2 9.4 9.6 9.8 10.0
Frequency (GHz)

Figure 7.9 Calculated r e t u r n loss b e f o r e (solid line) a n d after (dashed line) optimization
( r w g g a p = 5 mm).

Table 7.6
Symmetrical Ridged W a v e g u i d e B a n d p a s s Filter
rwg g a p = 5.00 m m
I n s e r t t h i c k n e s s = 0.10 m m

Septum Lengths (mm) Resonator Lengths (mm)

Before After Before After
Optimization Optimization Optimization Optimization
-6

0.5615 0.4805 ∣] = ⅛ 15.4961 15.3539

II
τS

4.9200 4.6439 ∣2 = ⅛· 15.7558 15.6814

II
,
-<S,

6.2612 5.9867 15.7507 15.6829

II

(rwg gap = 5 mm) (dashed line). Table 7.8 lists the insert dimensions for
conventional E-plane and ridged waveguide bandpass filters. As can be seen,
by using ridged waveguide bandpass filters, stopband performance can be
improved.
 Design of Ridged Waveguide Filters by Optimization 163

100.00 ->

CQ
T5 80.00 -

o
60.00 -

s 40.00 -

20.00 -

0.00 -----
8.00 10.00 12.00 14.00 16.00 18.00
Frequency (GHz)

Figure 7.10 Comparison of c a l c u l a t e d insertion loss of X-band five-resonator E-plane b a n d -

p a s s filters. Solid line i s conventional E-plane bandpass filter; dashed line i s
r i d g e d w a v e g u i d e filter (rwg g a p = 8 mm).

Table 7.7
Symmetrical Ridged W a v e g u i d e Bandpass Filter
rwg g a p = 8.00 m m

Septum Lengths (mm) Resonator Lengths (mm)

Conventional Filter rwg Filter Conventional Filter rwg Filter

= 1.3334 = 1.1507 ⁄ 1 = 15.8258 ⁄i = 15.7789

⅛

= 6.2502 = 6.0724 ⁄ 2 = 16.1814 ⁄ 2 = 16.1070

<s>-

= 7.6076 = 7.4306 ⁄ 3 = 16.2054 ⁄ 3 = 16.1274

= 7.6076 = 7.4306 ⁄ 4 = 16.1814 ⁄ 4 = 16.1070
s- & &

= 6.0724 ⁄ 5 = 15.8258 ⁄ 5 = 15.7789

A two-resonator X-band asymmetrical ridged waveguide bandpass filter

in WG-16 (Figure 7.12), in which the widths of the ridges are arbitrarily
chosen has been designed with the specifications listed in Table 7.9.
Configuration of the asymmetrical ridged waveguide bandpass filter struc-
ture is shown in Figure 7.12. Figure 7.13 shows the calculated passband
insertion loss (solid line) of the ridged waveguide filters (rwg gap 1,
η = 9.00 mm; rwg gap 2, ¾ = 8.00 mm) using the approximate method
164 Generalized Filter Design by Computer Optimization

ιoo.oo -∣

7□ 80.00 -

o
c 60.00 -
#o
⅛5
Φ
C 40.00 -

20.00 -

0.00 -----
8.00 10.00 12.00 14.00 16.00 18.00
Frequency (GHz)

Figure 7.11 Comparison of c a l c u l a t e d insertion loss of X - b a n d five-resonator E-plane b a n d -

p a s s filters. Solid line i s conventional E-plane bandpass filter; d a s h e d line i s
r i d g e d w a v e g u i d e filter (rwg g a p = 5 mm).

Table 7.8
Symmetrical Ridged W a v e g u i d e B a n d p a s s Filter
rwg g a p = 5.00 m m

Septum Lengths (mm) Resonator Lengths (mm)

Conventional Filter rwg Filter Conventional Filter rwg Filter
II II II II II II
— » σ> **j -*j σ> —k
-S ∙

⁄ 1 = 15.8258 ⁄ 1 = 15.3539
CJ1 GO
GO
NO 4⅛

0.4805
ω k) σ> ⅛ nj ω

rS1~

4.6439 ⁄ 2 = 16.1814 ⁄ 2 = 15.6814

O
σ>

6.3867 ⅛ = 16.2054 ⅛ = 15.6829

O

⁄ 4 = 16.1814 ⁄ 4 = 15.6814
CD

5.9867
O

<□?* cj?*

⅛ = 15.8258 ⁄ 5 = 15.3539
-P⅛ NO
GO GΠ

4.6439
GO O
σ5**

0.4805

described in Section 7.2 and in [16]. The approximate design was used as a
starting point for equal-ripple optimization. Mode matching with 60 modes
was used in both the design and the calculations. Table 7.10 lists the insert
dimensions before and after optimization. The passband insertion loss calculated
using the insert dimensions obtained on convergence (after optimization) also is
shown in Figure 7.13 (dashed line). This took five Newton-Raphson iterations.
Mode matching with 60 modes was used throughout the optimization.
 Design of Ridged Waveguide Filters by Optimization 165

Figure 7.12 Configuration of the asymmetrical ridged waveguide bandpass filter structure.

Table 7.9
Asymmetrical Ridged Waveguide Bandpass Filter

Waveguide WG16 (WR90) internal dimension: 22.86 mm by 10.16 mm

Midband frequency: 9.5 GHz
Passband: 9.25-9.75 GHz
Number of resonators: 2
Metal insert thickness: 0.10 mm
Ridged waveguide gap 1: 9.00 mm
Ridged waveguide gap 2: 8.00 mm
Passband return loss: 20 dB min
Filter characteristic: Chebyshev
166 Generalized Filter Design by Computer Optimization

0.1-

0.08-
Insertion loss (dB)

0.06-

0.04-

0.02-

0.0 —
9.2 9.3 9.4 9.5 9.6 9.7 9.8
Frequency (GHz)

Figure 7.13 Calculated insertion loss before (solid line) and after (dashed line) optimization
of the asymmetrical ridged waveguide bandpass filter.

Table 7.10
Asymmetrical Ridged Waveguide Bandpass Filter

Septum Lengths (mm) Resonator Lengths (mm)

Before After Before After
Optimization Optimization Optimization Optimization

0.7458 0.8659 l↑·. 15.5748 15.5169

3.0016 3.2400 ∣2'. 15.3952 15.4829
0.6003 0.6132

References

[1] Postoyalko, V., and D. Budimir, “Design of Waveguide E-Plane Filters With All-Metal
Inserts by Equal-Ripple Optimization,” IEEE Trans. Microwave Theory & Tech., Vol.
MTT-42, February 1994, pp. 217-222.
[2] Arndt, F., “The Status of Rigorous Design of Millimeter Wave Low Insertion Loss Fin-
Line and Metallic E-Plane Filters,” J. Instn. Electronics and Telecom. Engrs., Vol. 34, No.
2, 1988, pp. 107-119.
 Design of Ridged Waveguide Filters by Optimization 167

[3] Gololobov, V. P., and M. Yu. OmeΓyanenko, “Bandpass Filters Based on Planar Metal-
Dielectric Structures in the E-Plane of a Rectangular Waveguide (A Review),” Radioelec.&
Comm. Sys., Vol. 30, No. 1, 1987, pp. 1-15.
[4] Vahldieck, R., and W. J. R. Hoefer, “Finline and Metal Insert Filters With Improved
Passband Separation and Increased Stopband Attenuation,” IEEE Trans. Microwave
Theory & Tech., Vol. MTT-33, December 1985, pp. 1333-1339.
[5] Arndt, F., et al., “E-Plane Integrated Circuit Filters With Improved Stopband Attenua-
tion,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-32, October 1984, pp. 1391—
1394.
[6] Gololobov, V. P., and M . Yu. OmeΓyanenko, “Filters Based on Multilayered Metallic
Structures in a Waveguide,” Soviet J. Commun. Technol. Electron., Vol. 33, Pt. 8, 1988,
pp. 69-74.
[7] Arndt, F., et al., “Stopband Optimized E-Plane Filters With Multiple Metal Inserts of
Variable Number per Coupling Elements,” IEE Proc., Vol. 133, Pt. H , June 1986,
pp. 169-174.
[8] Riblet, H . J., “Waveguide Filters Having Nonidentical Sections Resonant at Same Funda-
mental Frequency and Different Harmonic Frequencies,” U.S. Patent No. 3,153,208,
1964.
[9] Bornemann, J., and F. Arndt, “Metallic E-Plane Filter With Cavities of Different Cutoff
Frequency,” IEE Electronics Letters, Vol. 22, May 1986, pp. 524-525.
[10] Shih, Y. C., “Design of Waveguide E-Plane Filters With All Metal Inserts,” IEEE Trans.
Microwave Theory & Tech., Vol. MTT-32, July 1984, pp. 695-704.
[11] Bui, L. Q., D. Ball, and T. Itoh, “Broad-Band Millimeter-Wave E-Plane Bandpass
Filters,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-32, December 1984,
pp. 1655-1658.
[12] Lim, J. B., C. W. Lee, and T. Itoh, “An Accurate CAD Algorithm for E-Plane Type
Bandpass Filters Using a New Passband Correction Method Combined With the Synthesis
Procedures,” IEEE MTT-S Int. Microwave Symp. Dig., June 1990, pp. 1179-1182.
[13] Shih, Y. C., “The Mode-Matching Method,” in Numerical Techniques for Microwave
and Millimeter-Wave Passive Structures,T. Itoh, ed., New York: Wiley, 1989, pp. 592-621 .
[14] Wexler, A., “Solution of Waveguide Discontinuities by Modal Analysis,” IEEE Trans.
Microwave Theory & Tech., Vol. MTT-15, 1967, pp. 508-517.
[15] Montgomery, J. P., “ O n Complete Eigenvalue Solution of Ridged Waveguide,” IEEE
Trans. Microwave Theory & Tech., Vol. MTT-19, June 1971, pp. 547-555.
[16] D. Budimir, “Optimized E-Plane Bandpass Filters With Improved Stop Band Perfor-
mance,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-45, February 1997,
pp. 212-220.
[17] Hewlett-Packard Product Note 8510-8, “Applying the HP8510B TRL Calibration for
Non-Coaxial Measurements,” August 1987.

Selected Bibliography
Bornemann, J., and F. Arndt, “Waveguide E-Plane Triple-Insert Filter,” 15th Eur. Microwave
Conf. Dig., Paris, France, 1985, pp. 726-731.
168 Generalized Filter Design by Computer Optimization

DBFILTER Reference Manual, Tesla Communications Ltd., London, England.

Gupta, K. C., R. Gary, and R. Chadha, Computer-Aided Design of Microwave Circuits, Dedham,
MA: Artech House, 1981.
Levy, R., “The Generalized Design Technique for Practical Distributed Reciprocal Ladder
Networks,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-21, August 1973, pp.
519-526.
Levy, R., “Theory of Direct Coupled Cavity Filters,” IEEE Trans. Microwave Theory & Tech.,
Vol. MTT-15, June 1967, pp. 340-348.
Matthaei, G., L. Young, and E. M. T. Jones, “ Microwave Filters, Impedance— Matching Networks
and Coupling Structures,” Dedham, MA: Artech House, 1980.
Rhodes, J. D., Theory of Electric Filters, New York: Wiley, 1976.
Rhodes, J. D., “Microwave Circuit Realizations,” in Microwave Solid State Devices and Applica-
tions, D . V. Morgan and M. J. Howes, eds., England: Peregrinus, 1980, pp. 49-57.

APPENDIX 7A: Derivation of the Impedance Inverter

Equations

The ABCD matrix of the two-port network between two reference planes is

. . I cos θ∖ j Z∖ sin θ∖ ∖
⁄ √4' jtf∖ I sn ∣
σ Dγλ⁄ = I i i θ∖ I
⁄ ∖
C0S θ
U
Ď Ď A 7 ∖ j⁄

. χ ⁄ cos #2 7¾ s in ⅛ ∣
⁄ y4' jB'∖ I s i n 1
∙r - r>' ∣7 <% ft I (7A.1
Ď ⁄ Ď A ⁄

which, by definition, is an impedance inverter if A∣= D∣= 0, that is, if

cos
A cos Φ2 ~ ~ET sin Φ2 I Φ∖ ~
Al ⁄

Z∖ sin φ∖ ( C cos 02 + — sin 02 I = θ (7A.2)

Ď A ⁄

( D cos 02 “ C Z2 sin φf) cos φ↑ -

1 +
r ( 5 cos 02 A Z sin 02) = 0 (7A.3)
A∖
 Design of Ridged Waveguide Filters by Optimization 169

Defining

tan 0ι = t↑ tan φ2 - (7A.4)

and dividing (7A.2) and (7A.3) by cos j∖ cos fy gives

B DZ 1
0 (7A.5)
Z2

B - Z2 =
Z) — CZ2t2 ~ z z
θ (7A.6)
ι ι

Solving (7A.5) for t 2 and substituting in (7A.6) gives

B ∖ (A-CZ λt x ) Z2
d 1 (CZ2) 0 (7A.7)
~ z⁄ ) (B + DZ 1 r1 )

or

z
- ACZ 2 ] ( 1 - tj) +
2 ⁄

t∖ = 0 (7A.8)
Z x

Using the identity

2 tan φ
tan 2 0 = ----------2— (7A.9)
1 - tan 0

then (7A.8) can be rearranged to give

2
tan 2 0 1 = -----------2-----------------------T (7A.10)
A λZ 1 Bλ o D12
~y~
z ÷ zyy z ~ C Z 1 Z 2 - —y z
ι l 2 2

Using the normalization given in (3) and (4), (7A.10) is then recognized as
being equivalent to (5). Equation (6) is derived by interchanging A and D and
170 Generalized Filter Design by Computer Optimization

subscripts 1 and 2 in (7A.10). Equation (1) is derived by noting that the

insertion loss of the impedance inverter is given by

(7A.11)

which can be solved to give K⅛s a function of Z√, resulting in (7.1).

Design of Coplanar Waveguide Filters
by Optimization

In recent years, CPW has emerged as an alternative to microstrip line for the
design of microwave and mm-wave integrated circuits [1]. This is due to the
fact that CPWs offer several advantages over conventional microstrip lines
for hybrid and monolithic M I C applications. End-coupled half-wavelength
resonator filters employing capacitive coupling [2], broadside-coupled CPW
filters [3], direct-coupled CPW filters employing inductive coupling [4], ribbon-
of-brick-wall CPW bandpass filters [5], M M I C bandpass filters using parallel-
coupled CPW lines [6], multilayer CPW filters using an overlapping of the
lines [7], and edge-coupled CPW filters [8] have been reported in the literature.
General purpose optimization techniques based on least y>th objective
functions use general forms of error minimization algorithms [9], which simply
force the filter transfer characteristic to be within specified constraints, whereas
a filter must have a specified ripple characteristic, for example, the Chebyshev
function. Usually the response of an optimizable filter is sampled at a number
of equally spaced frequencies, and the error between that sampled response
and the desired response is computed at each frequency. The optimization
program, through an iterative process, reduces the error to a minimum, arriving
at a final filter design in terms of the optimized filter parameters. These
optimization techniques cannot be guaranteed to satisfy filter specifications
and may even converge to a local minimum.
The approach presented here requires less frequency sampling than previ-
ous methods. This method optimizes the passband of a filter with respect to
the Chebyshev (or minimax) criteria, which relates directly to the way filters
are fabricated in practice. This vector procedure has several advantages over

171
172 Generalized Filter Design by Computer Optimization

the general purpose optimization routines previously applied to the design of

CPW filters. Design of a combined edge-end coupled bandpass filter by the
use of a cascade of half-wave resonators with electromagnetic simulations driven
indirectly by an equal ripple based optimizer was used as example.
Section 8.1 describes an approximate synthesis-based design procedure
of CPW bandpass filters. Section 8.2 examines the numerical implementation
of equal-ripple optimization, in the context of the design of CPW bandpass
filters. Section 8.3 presents a design example.

8.1 An Approximate Synthesis-Based Design Procedure

8.1.1 Circuit Representation

The proposed filter structure in Figure 8.1 can be represented as shown in

Figure 8.2 as a cascade of impedance inverters and resonators. The design of
this type of filter usually is based on the design procedure described in
Section 8.2.1, with the edge-coupled and end-coupled sections being related
to impedance inverters. Figure 8.3 shows a two-port defined by its ABCD
matrix. We assume that it is connected to lines of characteristic impedances
Z o at its two ports.
The element value of the impedance inverter (see Figure 8.3) can be
derived directly from the ABCD matrix of the edge- or end-coupled section.
It is given by

K = y [ L - y ∣L - l (8.1)

where

L=1 + ±[(Λ - D)2 + ( B - C) 2 ] (8.2)

The reference plane locations are given by the equation

2 (AB - CD)
tan(2Φ) = (8.3)
( D - A 1 ) + ( B2 - C 2 )
2

In practice, the electrical distance (Φ) can be realized as a negative value in

the adjacent positive line length, which therefore becomes shortened in the
final network.
 Design of Coplanar Waveguide Filters by Optimization 173

d d
2 2

Edge coupled CPW coupling CPW coupling Edge coupled

CPW lines 8a P gap CPW lines

Feed CPW CPW CPW Feed

li e
line C 1 resonator C 2 resonator C 2 resonator C j "

Figure 8.1 Schematic circuit diagram and layout of CPW filter.

8.1.2 Design Procedure

A common approach to the design of the conventional CPW bandpass filters

[2] can be used for the filter structure described in this chapter. Here, only
the most important steps in the design procedure, which include the concept
of impedance inverters, are presented. The design procedure for the CPW
bandpass filters is to apply (8.1) and (8.3) at the center frequency of the
specified passband to calculate the θ t and ⅛ which correspond to the impedance
174 Generalized Filter Design by Computer Optimization

Figure 82 Equivalent circuit of CPW filter using impedance inverters.

K l2 K nt n + l

θl θl ..... On

Figure 8.3 Impedance inverter.

inverters. This approximate treatment of the frequency dependence of (8.1)

and (8.3) can result in a designed passband that differs considerably from that
which is specified, and optimization is then required to tune the filter dimen-
sions to satisfy the design specification.
For a given filter specification such as the two passband edge frequencies
yielding and fa passband return loss (Z/?); stopband attenuation (£/); the
waveguide dimensions (Figure 8.4), such as waveguide gap (y), width (w),
substrate thickness (A), the metal thickness (r); and the dielectric constant of
the substrate (βr ), the modified design procedure is summarized as follows:

1. Determine the passband ripple level, 6, from the minimum passband

return loss, which is defined as
 Design of Coplanar Waveguide Filters by Optimization 175

Figure 8.4 Coplanar waveguide.

Lr = 10 logi(⁄ 1 + — λ) (8.4)
Ď β ⁄

2. Determine the number of resonators, n, from:

L1 = 10 log 1 0 [ l + β 2 T 2n(ω')] (8.5)

where

T n (ω') = cosh[w Arcosh(α>, )], for ∣ω'∣ > 1 (8.6a)

T n (ω') = cos[w arccos(ω')], for 0 < ω ≤ 1 (8.6b)

with

i
.⁄ (f fi>∖ ⁄o 7 A
(
ω
■ Aa " ~f) ’

8= (8.8)
⁄o

and

fa = yJfiifL (8 ∙9)

at the designated stopband frequency, f.

3. Calculate the impedance inverter values for the first edge-coupling

structure:
176 Generalized Filter Design by Computer Optimization

⅞,1 ' πδ
2
(8.10)
Z o g∂g∖

for the intermediate end-coupling structures:

7r
⅞∙,r+ 1 _ <S
r = 1, 2, . . . , n — 1 (8.11)
Z 2
0 grgr+∖

and for the final edge-coupling structure:

K∏,n+∖ _ . ⁄
Z 2
(8.12)
0 V gngn+1

4. Determine the Ah edge- or end-coupling length, A) in Figure 8.5 (a),

by solving (8.1) so that the required impedance inverter corresponds
to the characteristic impedance of the prototype filter. The elements
of the ABCD matrix (∠4,, B∣,C{, D 1 ), shown in Figure 8.5(b),
corresponding to the zth edge- or end-coupling section

jB∣ JB

n+l

D n+l D n +I
D,

(a)
θi ' θ
d
1

d d
2 2

(b)

Figure 8.5 (a) CPW bandpass filter and (b) equivalent circuit of CPW filter using ABCD
matrix for filter discontinuities.
 Design of Coplanar Waveguide Filters by Optimization 177

(z = 1, 2, . . . , n + 1) are a function of the length of the edge-

or end-coupling section (⅛). Since those functions are not available
explicitly, we must implement a root-seeking routine to find the value
of width that is provided by the required impedance value K and the
angle Φ for each impedance inverter.
5. Finally, the length, θ l , of the zth resonator (see Figure 8.5b) formed
by the zth and (z + l)th gaps is given by

φ i
θ i = τ r - ( Φ 2 ,⁄ + l,⁄ + l) = 1> 2 , ∙ ∙ ∙ , n
(8.13)

are
Φ∖ and Φ 2 given by (8.3).

Once the dimensions of the filter have been found, the frequency response
of the overall filter at each frequency can be simulated by cascading the ABCD
matrices of the resonators and the edge- and end-coupling sections. To illustrate
the application of this procedure to the design of CPW bandpass filters, the
design of an CPW bandpass filter with the specifications given in Section 8.3
(Table 8.1) is considered. Figure 8.6 shows the calculated passband return loss
(before optimization: dashed line) designed using this procedure. As can be
seen, the design specification still is unsatisfactory (e.g., passband return loss
is -4 dB at 22 GHz and -5 dB at 24 GHz; the specified value is - 1 4 dB),
and optimization is often required in practice for the accurate design of these
filters.

8.2 Numerical Implementation of Equal-Ripple Optimization

To apply the equal-ripple optimization technique described in [10] to the design

of CPW bandpass filters, it is necessary, for a given set of filter dimensions, to
be able to calculate the insertion loss on a sample of frequency points within

Table 8.1
Coplanar Waveguide Filter Specifications

Passband: 22-24 GHz

Bandwidth: 2 GHz
Number of resonators: 3
Passband return loss: -14 dB
Connectors: CPW probes
Impedance: 50∩
178 Generalized Filter Design by Computer Optimization

0.00 - r

Return loss (dB)

I
10.00-

2O.OO-
I

-30.00- > '⁄

_ ------- After optim.

------- Before optim.
-40. 00 I ∣ ι ∣ Γ ∣ I ∣ I ∣
20.00 21.00 22.00 23.00 24.00 25.00
Frequency (GHz)

Figure 8.6 Calculated return loss before (dashed line) and after (solid line) optimization of
the filter.

the specified passband. For a CPW bandpass filter, the insertion loss and
the return loss can be expressed in terms of an ABCD matrix. The matrix
representation of the whole filter (see Figure 8.5) is

Λ jB∖ _ I dis∖ jBdis∖ Ď φ ⁄ ‰1 7‰ι

√C & J y7Qiri tyisi J y/Gwi

γ
⁄ disi j disiX . ⁄ resi j resi∖ . . ⁄ disn+l j disn+l\
I jCdisi Ddisi j ∖j -'resi resi I ∖j disn+1 disn+1 j

(8.14)

in which A⅞s, C⅞s, and D 6ns are elements of the ABCD matrix of the
CPW filter discontinuity such as edge- or end-coupling section and A res , Brep
C res, and D res are elements of the ABCD matrix of the CPW filter resonator.
The overall filter response (the insertion loss, Z√, and the return loss, Lp>) can
be expressed in terms of elements of the total ABCD matrix of the filter at
each frequency (by directly combining the ABCD matrices of the individual
filter sections) as

Z1 = 20 logιo (8.15)
 Design of Coplanar Waveguide Filters by Optimization 179

( A + B + C + Z)Ď
Lr = 20 lo g l 0 4+ c - p J (8.16)

The elements of the ABCD matrices of the individual filter sections are calcu-
lated using the Em software package by Sonnet Software [11] and the Touch-
stone circuit simulator [12]. Neither accurate numerically fitted closed-form
expressions nor accurate design tables for the electrical parameters of the edge-
coupling and end-coupling sections in terms of section dimensions (length)
and frequency are yet available. The accurate design of CPW filters thus requires
direct calculation of the electrical parameters of those sections. That highlights
the need in the optimized design of the filters for optimization techniques that
minimize the number of calculations of the electrical parameters of sections.
A good approximate design of a CPW filter can be obtained by that procedure,
which implicitly includes the frequency dependence of the coupling gap and
results in passbands that nearly meet design specifications. It is therefore adopted
in this chapter as a means of generating a starting point for the optimization.

8.3 Numerical and Experimental Results

To illustrate the new approach, a three- resonator combined edge- and end-
coupled CPW bandpass filter has been designed with the specifications listed
in Table 8.1.
The filter can be described by four parameters: gap (⅛) and lengths
(Zμ 4, ⅛), as marked in Figure 8.1. We used equal-ripple optimization with
4 , 4> i> and ⅛ as variables for filter, while Sp was fixed at 0.074 mm. The
geometry for 5θΩ CPW line on 0.635-mm substrate is width of 0.25 mm
and gap of 0. 127 mm. The optimization variables before and after optimization
are listed in Table 8.2.

Table 8.2
Coplanar Waveguide Bandpass Filter
Thickness of the substrate = 0.635 mm
Dielectric constant of the substrate = 9.9

Parameters Before Optimization After Optimization

d↑ (mm): 0.195 0.642

⁄ 1 (mm): 2.455 1.982
d2 (mm): 0.205 0.090
∣2 (mm): 2.610 2.571
180 Generalized Filter Design by Computer Optimization

Figure 8.6 shows the calculated passband return loss of both filters (dashed
line) using the approximate method. The approximate design was used as a
starting point for equal-ripple optimization. The passband return loss calculated
using the filter dimensions obtained on convergence are also shown in
Figure 8.6 (the solid line). As can be seen, the return loss at band-edge
frequencies (22 and 24 GHz) is higher than the specified value of —14 dB.
Thus, optimization is required to satisfy the filter specifications.

References

[1] Holder, P. A. R., “X-Band Microwave Integrated Circuits Using Slotlines and Coplanar
Waveguide,” The Radio and Electronic Engineer, Vol. 48, No. 1/2, January/ February
1978, pp. 38-42.
[2] Williams, D . F., and S. E. Schwarz, “Design and Performance of Coplanar Waveguide
Bandpass Filters,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-31, July 1983,
pp. 558-566.
[3] Nquyen, C., “Broadside-Coupled Coplanar Waveguides and Their End-Coupled Band-
Pass Filter Applications,” IEEE Trans. Microwave Theory Tech., Vol. MTT-40, No. 12,
December 1992, pp. 2181-2189.
[4] Everard, J. K. A., and K. K. M. Cheng, “High Performance Direct Coupled Bandpass
Filters on Coplanar Waveguide,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-41,
September 1993, pp. 1568-1573.
[5] Lin, F-L., C-W Chiu, and R-B Wu, “Coplanar Waveguide Bandpass Filters- A Ribbon-
of-Brick-Wall Design,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-43, July
1995, pp. 1589-1596.
[6] Mernyei, F., I. Aoki, and H . Matsuura, “MMIC Bandpass Filter Using Parallel-Coupled
CPW Lines,” IEE Electronics Letters, Vol. 30, No. 22, October 1994, pp. 1862-1863.
[7] Menzel, W., et al., “Compact Multilayer Filter Structures for Coplanar MMIC’s,” IEEE
Microwave & Guide Wave Lett., Vol. 2, December 1992, pp. 497-498.
[8] Karacaoglu, U., et al., “An Investigation of CPW Bandpass Filters Using End-Coupled
Resonators and Square Dual-Mode Rings,” 25th Eur. Microwave Conf., Bologna, Italy,
1995, pp. 519-523.
[9] Bandler, J. W., and S. H . Chen, “Circuit Optimization: The State of the Art,” IEEE
Trans. Microwave Theory & Tech., Vol. MTT-36, 1988, pp. 424-443.
[10] Postoyalko, V., and D. Budimir, “Design of Waveguide E-Plane Filters With All-Metal
Inserts by Equal-Ripple Optimization,” IEEE Trans. Microwave Theory dr Tech., Vol.
MTT-42, February 1994, pp. 217-222.
[11] Em User’s Manual, Vol. 1, Release 4.0, Sonnet Software Inc., Liverpool, NY, 1996.
[12] Touchstone Reference Manual, Version 3.0, EEsof Inc., Westlake Village, CA, 1991.
 Design of Coplanar Waveguide Filters by Optimization 181

Selected Bibliography
Cohn, S. B., “Direct-Coupled-Resonator Filters,” Proc. IRE, Vol. 45, February 1957,
pp. 187-196.
DBFILTER Reference Manual, Tesla Communications Ltd., London, England.
Gupta, K. C., et al., Microstrip Lines and Slotlines, 2nd ed., Norwood, MA: Artech House,
1996.
Gupta, K. C., R. Gary, and R. Chadha, Computer-Aided Design of Microwave Circuits, Dedham,
MA: Artech House, 1981.
Hasler, M., and J. Neiryuck, Electrical Filters, Dedham, MA: Artech House, 1986.
Kulke, R., and I. Wolff, “Design of Passive Coplanar Filters in V-Band,” IEEE MTT-S Int.
Microwave Symp. Dig, 1996, pp. 1647-1650.
LINMIC+ User Manual, Version 2.1, Jansen Microwave, Germany, 1989.
M/FILTER Reference Manual, Eagleware Corp., USA, 1993.
Matthaei, G., L. Young, E. M. T . Jones, Microwave Filters, Impedance— Matching Networks and
Coupling Structures, Dedham, MA: Artech House, 1980.
MDS Reference Manual, Release 6.0, Hewlett-Packard Co., Palo Alto, CA, 1994.
Menzel, W., W. Schwab, and G. Strauss, “Investigation of Coupling Structures for Coplanar
Bandpass Filters,” IEEE MTT-S Int. Microwave Symp. Dig, 1995, pp. 1407-1410.
OSA90⁄hope Reference Manual, Version 3.5, Optimization System Associates Inc., Canada, 1995.
Rayit, A. K., and N. J. McEwan, “Coplanar Waveguide Filters,” IEEE MTT-S Int. Microwave
Symp. Dig, 1993, pp. 1317-1320.
Schwab, W., F. Boegelsack, and W. Menzel, “Multilayer Suspended Stripline and Coplanar Line
Filters,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-42, July 1994, pp. 1403-1407.
Series IV/PC Reference Manual, Version 6.0, Hewlett-Packard Co., Palo Alto, CA, 1995.
Super-Compact User's Manual, Rev. 6.5, Compact Software Inc., Paterson, NJ, 1994.
Swanson, D . G., and R. J. Forse, “An HTS End Coupled CPW Filter at 35 GHz,” IEEE
MTT-S Int. Microwave Symp. Dig, 1994, pp. 199-202.
9
CAD Programs

Preceding chapters have described the electromagnetic simulation, optimiza-

tion, and filter design by computer optimization. All numerical results presented
in this book were obtained using the DBFILTER software package, which was
especially developed based on the theory outlined in this book. DBFILTER has
been implemented in the FORTRAN language. The current suite (DBFILTER)
includes software for synthesis and design of lumped-element lowpass filters
(LCFIL), synthesis and design of lumped-element lowpass filters by computer
optimization (LCFILTER), synthesis and design of CPW bandpass filters
(CPWFIL), synthesis and design of CPW bandpass filters by computer optimi-
zation (CPWFILTER), electromagnetic modeling of waveguide discontinuities
using mode matching (WGMODEL), synthesis of E-plane waveguide bandpass
filters (EPSYNFIL), analysis and design of E-plane waveguide bandpass filters
(EPFIL), design of E-plane waveguide bandpass filters by computer optimiza-
tion (EPFILTER), electromagnetic modeling of ridged waveguide (RWGMO-
DEL), synthesis of ridged waveguide bandpass filters (RWGSYNFIL), analysis
and design of ridged waveguide bandpass filters (RWGFIL), and design of
ridged waveguide bandpass filters by computer optimization (RWGFILTER).
Computer programs are supplied for use with PC-386⁄486⁄Pentium computers
and VMS and UNIX workstations. This chapter presents two computer pro-
gram examples (LCFIL and EPFILTER) that show the application of the
approach described in this book.

9.1 The LCFILTER Program

The LCFILTER is an optimization-oriented software package for the synthesis
and design of lumped-element lowpass filters (see Figure 5.10). A flow chart

183
184 Generalized Filter Design by Computer Optimization

for the LCFILTER program is shown in Figure 9.1. The six subroutines
incorporated into the main program are as follows:

• PROTYCH calculates element values for Chebyshev lumped prototype

filters. It is called by the main program.
• RESCL computes return loss and insertion loss for the whole filter.
• INDUC and CAPAC calculate the ABCD parameters for the inductors
and capacitors. These subroutines are called by the subroutine RESCL.
• MULT performs ABCD matrix multiplication. It is called by the
subroutine RESCL.
• EROPTIM optimizes a lumped-element lowpass filter having equal-
ripple insertion loss or return loss in the passband. This subroutine
is called by the main program.

The source code of the LCFIL program follows.

* LCFIL
* The main program for CAD of Lumped element lowpass filters

PARAMETER (Ml =200 ,Nl=200,M=5000,N=12 )

INTEGER II,II, I, 10,MO, NO,IA, IB,IJOB,IER,K0, J,JO
INTEGER NOPT, NTH, NFP,M00, N00,NM, IFIL, IRESP,ITECH, IFE
REAL*8 DET, SIGMA, F0 ,FS ,FOP ,FC ,AO ,BO ,CO ,RET,INL
REAL*8 DAVG ,SUMA ,DELTA ,TL I ,TH I ,Tn ,INLSB ,RATIO ,FSTOP
REAL*8 BW, RL, EPS, RIPPLE, FH,FL,TL, TH, W,S,H,T, ER,TAND
REAL*8 LM1 ,LM2 ,LM3 ,LM4 ,LM5 ,LM6 ,LR1 ,LR2 ,LR3 ,LR4 ,LR5 ,LR6
REAL*8 DI ,D2 ,D3 ,D4 ,D5 ,D6 ,D7 ,D8 ,D9 ,D10 ,Dll ,D12
REAL*8
AL(N+1),AE(N),AE1(N) ,AE2(N),AE3(N),AE4 (N),AE5(N),AE6(N)
REAL*8 AJ1(N),AJ2 (N) ,AJ3(N),AJ4 (N),AJ(N,N),KINV(N+1),CAP(N+1)
REAL*8 WA(N),TC(N) ,AIJ(N,N),UNIT(N,N),SG(N)
REAL*8 AJT(N,N) ,TRM(M),REF(M),FRE(M),ROM(M),DER(N)
REAL*8 TAC(N,N),TCN(N,N) ,TOC(N,N) ,AB(N),AJ5(N),AJ6 (N)
REAL*8 TRMS(M),RP(M) ,DIM(N+1) ,AERR(M),TN
REAL*8 AJ7 (N) ,AJ8 (N) ,AJ9(N),AJ10 (N),AJ11 (N),AJ12 (N)
REAL*8 AE7 (N),AE8 (N),AE9 (N),AE10 (N),AE11(N),AE12 (N)
REAL*8 X,Y,KK(20),LV(20) ,CV(20) ,GK(20) ,A0(20),BB(20),PI
REAL*8 FL,FH,Fo, Bw, Bwn, Lrip, Er,Zo, Zr,eps, ripple
REAL*8 Lr (20),Ca(20),Lbps(20) ,Lbpp(20),Cbps(20) ,Cbpp(20)
REAL*8 Lbss ( 20 ) ,Lbsp (20) ,Cbss (20) ,Cbsp (20)
REAL*8 Lhp(20) ,Chp(20) ,GG(20) ,KF(20)
CL0=299 .7925000D00
PI=4*DATAN(1.0D00)
 CAD Programs 185

FILTER SPECIFICATIONS

ARE FILTER PARAMETERS KNOWN ?

YES NO

SYNTHESIS

ANALYSIS OPTIMIZER
(EROPTIM)

NO
ARE SPECS MET ?

YES

FABRICATION

Figure 9.1 F l o w c h a r t f o r the LCFILTER program.

186 Generalized Filter Design by Computer Optimization

ZJ= (O z 1)
* OPEN (UNIT=1 z FILE= ' drfdim. in' )
* OPEN (UNIT=2 z FILE= ' drf.in' )
OPEN (UNIT=6 z FILE= , LCFILTER.DAT' )
print *,'Number of Elements'
read ( * z * )NTH
print * z 'CUTOFF FREQUENCY in GHz '
read(* z * )FC
print * z ' MIN. PASSBAND RETURN LOSS in dB: '
read ( * z * )RL
print * z ' MIN. STOPBAND INSERTION LOSS in dB'
read(* z *)INLSB
print * z ' STOPBAND FREQQUENCY in GHz'
read ( * z * )Fstop
print * z 'OPERATION FREQUENCY in GHz'
read(* z * )FOP
print * z ' FREQUENCY STEP in GHz'
read(* z *)FSTEP
print * z 'NUMBER OF FRQUENCY POINTS'
read ( * z *)NFP
print * z ' Source Impedance'
read (* z *)Zo
print * z 'Load Impedance'
read ( * z * )Zr
print *z' Do you know filter element values YES ( 1 ) ⁄
NO ( 0 )'
print *
print *
read ( * z * ) I
print *
IF (I.EQ.1) goto 3
goto 4
3 print * z 'Ll'
read ( * z * )1ml
print * z 'Cl '
read ( * z * )1R1
if(nth.eq.l) goto 50
print * z ' L2 '
read ( * z * ) lm2
print * z 'C2 '
read ( * z * )1R2
if(nth.eq.2) goto 50
print * z ' L3 '
read ( * z * )lm3
print * z 'C3 '
read ( * z *)1R3
if(nth.eq.3) goto 50
print * z ' L4 '
 CAD Programs 187

read ( * ,* ) lm4
print * , , C4 ,
read(* z *)1R4
if(nth.eq.4) goto 50
print * z ' L5 '
read ( * z * )lm5
print * z , C5'
read(* z *)lr5
if(nth.eq.5) goto 50
print * z ' L6 ,
read ( * z * )lm6
print * z 'C6 '
read ( * z * )1R6
if(nth.eq.6) goto 50
print * z 'L7 ,
read ( * z * )lm7
print * z z C7 '
read ( * z *)lr7
if(nth.eq.7) goto 50
print * z z L8 z
read ( * z * )lm8
print * z z C8 z
read ( * z *)1R8
if(nth.eq.8) goto 50
print * z z L9 z
read ( * z * )lm9
print * z 'C9 ,
read ( * z *)1R9
if(nth.eq.9) goto 50
print * z ' L10 '
read ( * z * )lmlθ
print * z , CIO '
read ( * z * )LR10
if(nth.eq. 10) goto 50
print * z z Lil z
read ( * z * )1ml1
print * z z Cll z
read ( * z * )lrll
if(nth.eq. 11) goto 50
goto 1000
4 Fo=Fc
BW=Fc
BWN=BW⁄Fo
EPS=DSQRT (10**(0.1*RL)-1.0)
EPS=1⁄EPS
RIPPLE= 10 *DLOG1 0 (1+EPS**2)
RATIO=FSTOP⁄FC
Zn=Zr⁄Zo
188 Generalized Filter Design by Computer Optimization

FS= (FL-FH) ⁄(NFP-1)

DO 1=1,100
EPS= (10**(0. 1*RIPPLE)-1.0)
IF (RATIO.LT. 1) GOTO 6
TN= ( I*DLOG (RATIO+DSQRT (RATIO* *2 -1.00 ) ) )
TN= (DEXP(TN) +DEXP(-TN) )/2.00
GOTO 7
6 TN=DCOS (I*DACOS (TN) )
7 INL=10*DLOG ( 1+EPS**2*TN* *2 )
NRES=I
IF (NTH. GT. 11) goto 1000
IF ( INL.GT. INLSB) goto 53
END DO
53 EPS=10** (0. 1*RIPPLE)-1.0
RL=10*DLOG10 (1.00+ (1.0⁄EPS))
CALL PRTYPCH (NTH ,Fc ,Zo ,RL,LV,CV, Lr ,Ca ,GK, KK )
goto 120
50 EPS=DSQRT ( 10* *(0.1*RL)-1.0)
EPS=1⁄EPS
RIPPLE=10*DLOG10 (1+EPS**2)
LR(1)=LM1
CA(1) =LR1
if(nres .eq .1 ) goto 120
LR(2) =LM2
CA(2) =LR2
if(nres . eq. 2 ) goto 120
LR(3)=LM3
CA(3) =LR3
if(nres . eq. 3 ) goto 120
LR(4) =LM4
CA(4) -LR4
if(nres . eq. 4 ) goto 120
LR(5)=LM5
CA(5)=LR5
if(nres. eq. 5 ) goto 120
LR(6)=LM6
CA(6)=LR6
if(nres.eq. 6 ) goto 120
CA(7)=LR7
LR(7)=LM7
if(nres .eq .7 ) goto 120
LR(8) =LM8
CA(8)=LR8
if(nres .eq .8 ) goto 120
CA(9) =LR9
LR(9) =LM9
if(nres . eq .9 ) goto 120
LR(10) =LM10
 CAD Programs 189

CA(10)=LR10
if(nres .eq. 10 ) goto 120
CA(11)=LR11
LR(11) =LM11
if(nres . le.11 ) goto 120
goto 1000

120 write(6 z * )
write(6,*) 'TYPICAL PERFORMANCE DATA FOR A Nth-DEGREE ,
write( 6 , ★)
write(6 z *) ' * *** **LUMPED ELEMENT LOW PASS FILTER***'
write(6 z * )
write( 6 z * )
write( 6 z * )
WRITE (6 z *) 'FILTER CHARACTERISTIC :CHEBYSHEV '
write(6 z * )
write(6 z *) 'NUMBER OF ELEMENTS ' z NTH
write(6 z * )
WRITE (6 z *) 'CUTOFF FREQUENCY in GHz :' z FC
WRITE (6 z *)
WRITE (6 z *) 'RIPPLE LEVEL IN dB' z RIPPLE
write(6 ,* )
write(6 z *) 'MIN. PASSBAND RETURN LOSS in dB:' z RL
WRITE(6 z *)
write(6 z *) 'MIN. STOPBAND INSERTION LOSS in dB:' z INLSB
WRITE (6 z *)
write(6 z *) 'STOPBAND FREQUENCY in GHz :' z FSTOP
write(6 z * )
write(6 z *) 'STARTING OPERATION FREQUENCY in GHz ' z FOP
write(6 z * )
write(6 z *) 'NUMBER OF FREQUENCY POINTS ' z NFP
write(6 z * )
write(6 z *) 'FREQUENCY STEP in GHz ' z FS
write(6 z * )
write(6 z * )
write(6 z *) 'FILTER ELEMENT VALUES'
write(6 z * )
write(6 z * )
write(6 z *) 'LOWPASS FILTER PROTOTYPE ELEMENT VALUES'
write(6 z *) DO I=l z NTH+2
write(β z *) 'Gv' z GK( I )
ENDDO
write(6 z * )
write(6 z * )
DO I=l z NTH
write(6 z * ) ' Lr=Zo*gv(l⁄fc)' z Lr(I)
ENDDO
write(6 z * )
190 Generalized Filter Design by Computer Optimization

DO I=1,NTH
write(6,*) ' Ca=gv⁄ ( Zo* fc )' ,Ca ( I )
ENDDO
write( 6 ,* )

FO=FOP
N00=NTH+l
M00=NFP
print *,'For Shunt Capacitor as 1st filter element (2)'
print *
DO J=l,M00
F0=F0+FS
CALL RESCL (NRES, M,LR,CA,AE,F0 ,Zo ,RET, INL)
TRM(J) =-INL
ROM (J)=-RET
FRE(J) =F0
ENDDO
write( 6 ,* )
WRITE(6,*) 'FREQUENCY, RETURN and INSERTION LOSS BEFORE
OPT. '
write( 6 ,* )
DO I=l,M00
RP( I ) =DELTA
WRITE(6,*) FRE(I),ROM(I),TRM(I)
ENDDO
1000 END

SUBROUTINE PRTYPCH (NRES ,Fc ,Zin,RL,LV,CV,LA,CAP,GK ,KK)

INTEGER Nres,Neven
REAL*8 Nodd,Ndif
REAL* 8 CL0 ,PI ,RL,RIPPLE, X0 ,Y0 ,LA(20 ) ,CAP(20),FC ,Zin,GG ( 20 )
REAL*8 X,Y,EPS,KK(20),LV ( 2 0 ) ,CV ( 20 ) ,GK ( 2 0 ) ,A0( 20 ) ,BB( 2 0 )

CL0=299 . 792500D00
PI=4*DATAN ( 1.0D00 )

EPS=DSQRT (10**(0.1*RL)-1.0)
EPS=1⁄EPS
X0=DLOG( (l⁄eps)+dsqrt (1+(l⁄eps)**2))
X=0.5*X0
Y0=X0⁄Nres
Y=0. 5*(DEXP (Y0 ) -DEXP ( -Y0) )
DO I=l,Nres
A0(I) =DSIN( (2*1-1)*PI⁄ (2*Nres) )
BB(I) =Y**2+(DSIN(I*PI⁄Nres))**2
ENDDO
GG(0)=1.0
 CAD Programs 191

GG(l)=2.00*A0(l)⁄Y
Do J=2,Nres
GG(J) =4. 00*A0 (J-l)*A0 (J) ⁄(BB(J-l )*GG (J-l) )
ENDDO
DO J=3,Nres+l
GK(J) =GG(J-1)
ENDDO
GK ( 2 ) =GG ( 1 )
GK ( 1 ) =GG( 0 )
Nodd=Nres⁄2.00
Neven=Nres⁄2
Ndi f =Nodd-Neven
IF(Ndif .GT.0.2) goto 1
* Nres even
GK(Nres+2) =( (DEXP(X) +DEXP(-X)) ⁄(DEXP(X) -DEXP(-X) ) ) **2
goto 2
* Nres odd 1
1 GK(Nres+2) =1.000
2 DO I=2,NRES+1
LV ( I -1 ) =GK ( I )
La(I-l)=LV(I-l)*Zin⁄ ( 2 .0*pi*Fc*l . E+9 )
ENDDO
DO I=1,NRES
CV(I)=GK(I+1)
Cap(I)=CV(I) ⁄(2 . 0*pi*Fc*l.E+9*Zin)
ENDDO
GOTO 100
3 DO I=1,NRES-1
KK(I) =DSQRT(Y**2+ (DSIN(I*PI⁄Nres) )**2) ⁄Y
ENDDO
DO I=1,NRES
LV(I) =(2 .00⁄Y) *DSIN( (2*1-1) *PI⁄ (2*NRES) )
La(I)=LV(I)*Zin⁄ (2. 0*pi*Fc*l. E+9)
ENDDO
DO I=1,NRES-1
CV(I)=LV(I) ⁄(KK(I)) **2
Cap (I)=CV(I) ⁄(2.0*pi*Fc*l.E+9*Zin)
ENDDO
1000 END

SUBROUTINE RESCL ( No ,M ,LR ,CAP,AER ,Freq ,Zo ,RL, I L )

INTEGER Fpts ,No, Nres ,M

COMPLEX* 16 Al ,Bl ,Cl ,DI ,A2 ,B2 ,C2 ,D2 ,ATi ,BTi ,CTi ,DTi
COMPLEX* 16 DET, VI ,Il ,Zin, ZJ, Detl ,Det2 ,At ,Bt ,Ct ,Dt ,Sil ,S2 1
COMPLEX* 16 Att ,Btt ,Ctt ,Dtt ,Atl ,Bt 1 ,Ctl ,Dtl ,At2 ,Bt2 ,Ct2 ,Dt2
COMPLEX* 16 At3 ,Bt3 ,Ct3 ,Dt3 ,At 4 ,Bt4 ,Ct4 ,Dt4 ,All ,Bll ,Cll ,Dll
COMPLEX* 16 A12 ,B12 ,C12 ,D12 ,A13 ,B13 ,C13 ,D13 ,A14 ,B14 ,C14 ,D14
192 Generalized Filter Design by Computer Optimization

COMPLEX* 16 Al 5 ,Bl 5 ,Cl 5 ,DI 5 ,At 5 ,Bt5,Ct 5 ,Dt5 ,A3 ,B3 ,C3 ,D3
REAL*8 Fstart ,Fstop, Er ,Qu ,Zo, Zn, Fo, Bw, Lrip,LP, IL,RL
REAL*8 I IM,VIM,Io,Zr ,Freq, CLO ,PI ,S21M,S11M,S11DB,S21DB
REAL*8 LR(20),CAP(20)

* For Nres=5 *
REAL*8 LR(5),CAP( 5 ) ,Results(3 ,10) ,G(7) ,K(6)
* For Nres=6
* REAL*8 LR(6),CAP(6) ,Results(3,10),G(8),K(7)
* For Nres=7
* REAL*8 LR(7) ,CAP(7) ,Results(3, 10),G(9),K(8)
* COMMON Er,Qu,Zo,Zn,CLO,PI,ZJ

CL0=299 .792500D00
PI=4*DATAN(1 . 0D00)
zσ= (0, i)

* cap ( 1 ) =2. 55E-12

* cap ( 3 )=2.10E- 12
* cap(5) =0.49E-12
* Lr(2)=6.25E-09
* Lr(4)=3.66E-09

CALL CAPAC (Cap ( 1 ) ,Freq, Zo, A2 ,B2 ,C2 ,D2 )

CALL INDUC ( Lr ( 2 ) ,Freq, Zo ,Al ,Bl ,Cl ,DI )
CALL MULT ( A2 ,B2 ,C2 ,D2 ,Al ,Bl ,Cl ,DI ,At ,Bt ,Ct ,Dt )
IF(No. eq. 2) GOTO 600
CALL CAPAC (Cap ( 3 ) ,Freq,Zo, A3 ,B3 ,C3 ,D3 )
CALL MULT (At ,Bt ,Ct ,Dt ,A3 ,B3 ,C3 ,D3 ,Atl ,Btl ,Ctl ,Dtl )
IF (No. eq. 3) GOTO 500
CALL INDUC ( Lr ( 4 ) ,Freq ,Zo ,All ,Bll ,Cll ,DI 1 )
CALL MULT ( Atl, Btl,Ctl,Dtl, All, Bll,Cll,Dll, At, Bt,Ct,Dt)
IF (No. eq. 4 ) GOTO 600
CALL CAPAC (Cap ( 5 ) ,Freq, Zo ,A2 ,B2 ,C2 ,D2 )
CALL MULT (At ,Bt ,Ct ,Dt ,A2 ,B2 ,C2 ,D2 ,At 1 ,Btl ,Ctl ,Dtl )
IF (No. eq. 5) GOTO 500
CALL INDUC ( Lr (6 ) ,Freq, Zo ,Al1 ,Bl1 ,Cl 1 ,DI 1 )
CALL MULT ( Al 1 ,Bl 1 ,Cl 1 ,DI 1 ,At 1 ,Bt 1 ,Ct 1 ,Dt 1 ,At ,Bt ,Ct ,Dt )
IF (No. eq. 6) GOTO 600
CALL CAPAC (Cap ( 7 ) ,Freq, Zo ,A2 ,B2 ,C2 ,D2 )
CALL MULT ( A2 ,B2 ,C2 ,D2 ,At ,Bt ,Ct ,Dt ,At 1 ,Btl ,Ctl ,Dtl )
IF (No. eq. 7) GOTO 500
CALL INDUC ( Lr ( 8 ) ,Freq,Zo ,Al1 ,Bl1 ,Cl 1 ,Dll )
CALL MULT ( Al 1 ,Bl1 ,Cl 1 ,DI 1 ,At1 ,Bt 1 ,Ct 1 ,Dt 1 ,At ,Bt ,Ct ,Dt )
IF (No. eq. 8) GOTO 600
CALL CAPAC (Cap ( 9 ),Freq, Zo ,A2 ,B2 ,C2 ,D2 )
CALL MULT ( A2 ,B2 ,C2 ,D2 ,At ,Bt ,Ct ,Dt ,Atl ,Btl ,Ctl ,Dt 1 )
IF(No.eq.9) GOTO 500
 CAD Programs 193

CALL INDUC ( Lr ( 10 ) ,Freq, Zo ,Al 1 ,Bl 1 ,C l 1 ,DI 1 )

CALL MULT ( All,Bll,Cll,Dll,Atl,Btl,Ctl,Dtl, At,Bt,Ct,D t )
IF(No.eq,10) GOTO 600
CALL CAPAC (Cap( 1 1 ) ,Freq,Zo, A2 ,B2 ,C2 ,D2 )
CALL MULT ( A2 , B2 ,C2 ,D2 ,At ,Bt ,Ct ,Dt ,At 1 , Bt 1 ,Ct 1 ,Dt 1 )
IF (No. eq. 11) GOTO 500
CALL INDUC (Lr( 12 ) ,Freq, Zo, All ,Bll ,Cll ,Dll )
CALL MULT ( All,Bl 1,C l 1,DI 1,Atl, Btl,Ct 1,Dtl, At,Bt,Ct,Dt)
IF (No. eq. 12) GOTO 600
CALL CAPAC (Cap( 13 ) ,Freq,Zo, A 2 , B2 ,C2 ,D2 )
CALL MULT ( A2 , B2 ,C 2 ,D2 ,At , Bt ,C t ,Dt ,Atl ,Btl ,Ct 1 ,Dtl )
IF(No.eq.l3) GOTO 500
CALL INDUC (Lr( 14 ) ,Freq,Zo, All ,Bll ,Cll ,Dll )
CALL MULT ( Al 1 , Bl 1 ,Cl 1 ,DI 1 ,At 1 ,Bt 1 ,Ct 1 ,Dt 1 ,At ,Bt ,C t ,Dt )
IF(No.eq.l4) GOTO 600
CALL CAPAC (Cap( 1 5 ) ,Freq ,Zo ,A2 , B2 ,C2 ,D2 )
CALL MULT (A2 , B2 ,C2 ,D2 ,At ,Bt ,Ct ,Dt ,Atl ,Btl ,Ctl ,Dtl )
IF(No.eq.l5) GOTO 500
CALL INDUC (Lr( 16 ) ,Freq, Zo, A2 ,B2 ,C2 ,D2 )
CALL MULT ( A2 ,B2 ,C2 ,D2 ,At 1 , Bt 1 ,Ct 1 ,Dt 1 ,At ,Bt ,Ct ,Dt )
IF(No.eq.l6) GOTO 600
GOTO 1000

600 S21=(2/(At+Bt+Ct+Dt) )
S21M=CDABS(S21)
Sll= ( (At+Bt-Ct-Dt) ⁄(At+Bt+Ct+Dt) )
S11M=CDABS(Sil)
S21DB=20*DLOG10 (S21M)
S11DB=20*DLOG10 (SUM)
IL=20*DLOG10 (1⁄S21M)
RL=20*DLOG10(1⁄S11M)
write(6 ,* )
write(6,*) 'Frequency (GHz ) ,Transm. ,Ref lec . '
write(6,*) Freq,S21,Sll
write(6,*) 'Frequency (GHz ) ,Transm. (dB) ,Ref lec . (dB)'
write(6,*) Freq, S21DB, S11DB
write(6,*) 'Frequency (GHz) ,Ins . loss (dB) ,Ref . loss (dB) '
write(6,*) Freq, IL,RL
write( 6 ,* )
goto 1000
500 S21=(2/(Atl+Btl+Ctl+Dtl) )
S21M=CDABS(S21)
Sll= ( (Atl+Btl-Ctl-Dtl) ⁄(Atl+Btl+Ctl+Dtl) )
S11M=CDABS(Sil)
S21DB=20*DLOG10 (S21M)
S11DB=20*DLOG10 (SUM)
IL=20*DLOG10(1⁄S21M)
RL=20*DLOG10 (1⁄S11M)
194 Generalized Filter Design by Computer Optimization

* write(6 z * )
write(6 z *) 'Frequency (GHz) z Transm. z Ref lec . '
* write(6 z *) Freq z S21 z Sll
write(6 z *) Freq z S21DB z S11DB
write(6 z *) 'Frequency (GHz ) z Ins .loss (dB) z Ref . loss (dB)'
write(6 z *) Freq z IL z RL
write(6 z * )
1000 END
*************************************************************
SUBROUTINE INDUC ( LI ,FREQ,Zo, Al ,Bl ,Cl ,DI )
COMPLEX* 16 GAMMA, Al ,Bl ,Cl ,DI ,A3 ,B3 ,C3 ,D3 ,ZJ
COMPLEX*16 GAMMA, A2 ,B2 ,C2 ,D2 ,A, B,C,D
REAL*8 Er,Qu, Zo, Zn, CL0 ,PI ,Freq, Bw, No, Lt ,LI ,L2 ,CAP,IL
COMPLEX* 16 IM, B1M,C1M,DIM, A3M,B3M,C3M,D3M,AM,DM, BM,CM
* COMMON Er,Qu, Zo, Zn,CL0, PI,ZJ
ZJ~ (0,1)
PI=4*DATAN ( 1 . 0D00 )
Al=(1,0)
B1=ZJ* ( 2 *PI*Freq*l . E+9*L1 ) ⁄Zo
Cl= (0,0)
Dl= (1,0)
1000 END
*************************************************************
SUBROUTINE CAPAC ( CAP,FREQ ,Zo ,A2 ,B2 ,C2 ,D2 )

COMPLEX* 1 6 GAMMA, Al ,Bl ,Cl ,DI ,A3 ,B3 ,C3 ,D3 ,ZJ
COMPLEX *1 6 GAMMA ,A2 ,B2 ,C2 ,D2 ,A ,B ,C ,D
REAL*8 Er ,Qu ,Zo ,Zn ,CL0 ,PI ,Freq ,Bw, No ,Lt ,LI ,L2 ,CAP ,IL
COMPLEX* 16 1M, B1M,C1M,DIM,A3M, B3M,C3M,D3M,AM,DM, BM,CM
* COMMON Er,Qu,Zo,Zn,CL0,PI,ZJ

ZJ= (0,1)
PI=4 *DATAN ( 1 .0D00 )

A2=(1,0)
B2=(0,0)
C2=ZJ*2*PI*FREQ*1 .E+9*CAP*Zo
D2=(1,0)
1000 END
*************************************************************
SUBROUTINE MULT (Al ,Bl ,Cl ,DI ,A2 ,B2 ,C2 ,D2 ,A3 ,B3 ,C3 ,D3 )

REAL*8 IL
COMPLEX* 16 A3 ,B3 ,C3 ,D3 ,A2 ,B2 ,C2 ,D2 ,Al ,Bl ,Cl ,DI
A3=A1*A2+B1*C2
B3=A1*B2+B1*D2
C3=C1*A2+D1*C2
D3=C1*B2+D1*D2
1000 END
 CAD Programs 195

9.2 The EPFILTER Program

In today’s increasingly competitive environment, waveguide filter designs are

continuously seeking to improve product reliability and decrease performance
variations without applying overly stringent manufacturing tolerances. With
EPFILTER, designers can predict performance parameters quickly and easily
before costly prototypes are built. EPFILTER is an optimization-oriented soft-
ware package for the design of E-plane waveguide filters (see Figure 6.1) tailored
to microwave and mm-wave applications. A flow chart for the EPFILTER
program is shown in Figure 9.2. The software supplies a full-wave solution
for filter discontinuities, using the well-known mode-matching method and
equal-ripple optimization technique for filter optimization. EPFILTER soft-
ware substantially reduces time to market and increases circuit performance.
No other software offers such a wide range of simulation, synthesis, and
optimization tools. The software is designed for ease of use.
The nine subroutines incorporated into the main program are as follows:

• PROTYCH calculates the inverters’ impedance value for Chebyshev

distributed prototype filters. It is called by the main program.
• STARV calculates starting filter parameters, such as lengths of the
metal septa and resonators.
• EROPTIM optimizes the E-plane metal insert bandpass filter having
equal-ripple insertion loss or return loss in the passband. It is called
by the main program.
• MODEL computes even and odd mode impedances of the metal
septum inside rectangular waveguide (waveguide bifurcation) for a
given frequency and dimension. It is called by the subroutines
EROPTIM and STARV.
• NVMAT performs real matrix inversion. This subroutine is called by
the subroutines EROPTIM and MODEL.
• RMULL performs real matrix multiplication. This subroutine is called
by the subroutine EROPTIM.
• CMULL performs multiplication of complex matrices. This subroutine
is called by the subroutine MODEL.
• RETINL computes return loss and insertion loss for the whole filter.
It is called by the main program, the subroutine EROPTIM, and the
subroutine STARV.
• ABCD performs ABCD matrix multiplication. It is called by the
subroutine RETINL.
196 Generalized Filter Design by Computer Optimization

FILTER SPECIFICATIONS

ARE FILTER PARAMETERS KNOWN ?

YES NO

SYNTHESIS

MODEL
(Mode-Matching
Method)

OPTIMIZER
ANALYSIS (EROPTIM)

NO
ARE SPECS MET ?

YES

FABRICATION

Figure 9.2 Flow chart for the EPFILTER program.

 CAD Programs 197

The source code of the subroutine PRTYPCH follows:

SUBROUTINE PRTYPCH (N ,Ml ,Nl ,AO ,CO ,FL,FH ,NFP ,LI ,DI S ,NP ,AL,
> RL,KP,Z,ZK, Z0 ,ROE, ROO, TR116L, TR126L, TR216L,
> KC1 ,KC,WK,INT, Ya,Yb, SUMO ,SUM,SK2 ,Aal ,TCI ,Abl ,Gl ,G,
> KZ1 ,KZ, TC10 ,TCNO, TCNE, H2E, H2O, ZCO, H3NM,ZBO, ZBE, ZCE,
> H4L ,H6L ,H60L ,H5LM,H3 5LM,H3 0NM, HIM,UNITC ,AlJC ,WA,RESS )

INTEGER I,II,NFM,J,P,M1,N1,10,IMPI ,NFP ,NP, NRES ,N

REAL*8 DO ,WA, KOI ,K12 ,K23 ,FC, FS,FSB,F02 ,TETA1 ,RESK01 ,Tn, INLSB
REAL*8 LAMDAGO ,LAMDAC ,LAMDAO ,FI1 ,FI2 ,TETA2 ,FI3 ,FI4 ,RESK12
REAL*8 LAMDAL ,LAMDAGL ,LAMDAH ,LAMDAGH ,LAM1 ,LAM2
REAL* 8 TETA3 ,TETA4 ,D5 ,D6 ,D7 ,D8 ,XS ,XP ,ALFA ,Y ,EPS ,DELTA ,CLO ,DLO
REAL*8 Lamdas, Lamdags, FOP, FSTEP,GW0
REAL*8 Z0(N) ,Z(N+2),AL(N+1),KP(N+1),ZK(N+1)
REAL*8 F0,A0,B0,C0,T

DATA MIO, EPSO/12 . 56637 IE-10,8 .85419E-15⁄

DATA CLO, ETA/299 .792500D00, 376 .70D00⁄
PI=4*DATAN(1 .00D00)

NR=N
NRES=N
I1=M1
II=N1
A00=2*C0
BW3DB=FH-FL
KCW=PI⁄A0

LAMDAC =2 *PI ⁄KCW

LAMDAL=CL0⁄FL
LAMDAH =CL0⁄FH
LAMDAGL=LAMDAL⁄DSQRT(1- (LAMDAL/LAMDAC) **2)
LAMDAGH=LAMDAH⁄DSQRT(1- ( LAMDAH ⁄LAMDAC ) **2)
LAMDAG0= (LAMDAGH+LAMDAGL) /2. 0
LAM1 =LAMDAGL *DCOS (PI*LAMDAGH/ (2*LAMDAGL))
LAM2= LAMDAGH *DCOS (PI*LAMDAGL⁄ (2*LAMDAGH) )
LAM3=DSIN(PI*LAMDAGH⁄ (2*LAMDAGL) )
LAM4=DSIN( PI *LAMDAGL⁄ (2*LAMDAGH) )
LAMDAGO =LAMDAG0+ ( (LAM1+LAM2 ) ⁄( (LAM3+LAM4)*PI))
ALFA=LAMDAG0⁄ ( LAMDAGL*DSIN ( PI *LAMDAG0 /LAMDAGL ))
LAMDAO =1+( LAMDAG 0 ⁄LAMDAC ) **2
LAMDAO =DSQRT ( LAMDAO )
LAMDAO =LAMDAGO ⁄LAMDAO
FC =CL0 /LAMDAO
K0=2*PI /LAMDAO
EPS=DSQRT ( 10 **(0.1*RL)-1.0)
EPS=1⁄EPS
198 Generalized Filter Design by Computer Optimization

RIPPLE=10*DLOG10 (1+EPS**2)
EPS=DSQRT ( 1O**( 0 . 1*RIPPLE)-1.0)
EPS=1⁄EPS
Y-DSINH( (1⁄NR)*DLOG ( EPS+DSQRT ( EPS* *2 +1. 00 )) )
EPS=1⁄EPS
DO I =1,N
Z0 (I)=2*ALFA*SIN(( 2*1-1 ) *PI ⁄( 2 *N) ) ⁄Y-(1/(4*Y*ALFA)
> *((Y**2+(SIN(I*PI⁄N) )**2) ⁄SIN((2*1+1)*PI⁄ (2*N) )
> +(Y**2+(SIN( (1-1)*PI⁄N)) **2) ⁄SIN((2*1-3 ) *PI⁄ (2*N) ) )
ENDDO
Z(l)=1.00
Z(N+2)=1.00
DO J=2,N+1
Z (J)=Z0 (J-l)
ENDDO
ZK(1) =1.000
DO I=2,N+1
IO=I-1
ZK(I)=DSQRT(1+ (SIN(IO*PI⁄N) ⁄Y)**2 )
ENDDO
DO I=1,N+1
KP(I)=ZK(I) ⁄DSQRT(Z(I) *Z(I +1))
ENDDO
1000 END

Selected Bibliography
Budimir, D., and V. Postoyalko, “EPFILTER: A CAD of Waveguide E-Plane Filters,” Microwave
Journal, August 1996.
Postoyalko, V., and D. Budimir, “Design of Waveguide E-Plane Filters With All-Metal Inserts
by Equal-Ripple Optimization,” IEEE Trans. Microwave Theory & Tech., Vol. MTT-42,
February 1994, pp. 217-222.
Budimir, D., “Design of E-Plane Filters With Improved Stopband Performance,” IEEE Trans.
Microwave Theory dr Tech., February 1997, pp. 212—220.
Appendix A Parameters

A.1 ABCD PARAMETERS

Figure A. 1 shows a two-port network. ABCD parameters ( ½ w, »zw dependent;

V0l3 out independent) for a such network are defined as

Vout
(A.1)
out _

For reciprocal networks:

AD - BC = 1 (A.2)

For symmetrical networks:

1 ,
Iin lout

+
a I
Two-port A
,
V in Network T V out
I

Port 1 (input) Port 2 (output)

Figure A.1 A two-port network with voltage and current defined.

199
200 Generalized Filter Design by Computer Optimization

A=D (A.3)

For a cascade of two or more networks (Figure A.2), the overall ABCD matrix
is given by

(A.4)
C]A2 + D∖ C2 C∖B2 + A 2

For a parallel of two networks (Figure A.3), the overall ABCD matrix is given
by

A↑B2 + B↑A2 BχB2

A b1 + b2 B↑ + B1
C D∖B2 + Bγ D 1
B2 + B2
(A.5)

ABCD parameters for some of the commonly used two-port networks are
shown in Figure A.4.

D D.

Figure A.2 A cascade connection of two-port networks.

 Appendix A Parameters 201

Figure A.3 A parallel connection of two-port networks.

A.2 Z Parameters

Z (impedance) parameters ( V‰ V 0ut dependent; 4 I out independent) for the

network shown in Figure A.1 are defined as

Vin Zu
(A.6)
Vout . z22J∣‰J

A.3 Y Parameters

Y (admittance) parameters (/,„, I0ut dependent; V out , V out independent) for the
network shown in Figure A. 1 are defined as

41 = Γr n 12 I Γ K
(A.7)
Iout 21 22 ½>at
202 Generalized Filter Design by Computer Optimization

Two-port Network ABCD matrix

Coshγl Z o Sinhyi

Z<> Z.?<
Sinhγl
Cosh)i
A transmission line

Z
1 z
z z
0 1
A series impedance

1 0

Y 1
A shunt admitance

2. W
Z,

1 Z
~Z
T-network

Y 1
Y

π-network

where W = Zι Z 2 + Z 2 Z 3 + Zι Z 3

Figure A.4 A two-port network.

A.4 S Parameters

In Figure A. 5, the two-port network is fed at port 1 (input) via transmission

line of characteristic impedance Zθι and at port 2 (output) via transmission
line of characteristic impedance Z . S parameters (Z⅛, bout dependent; a ιw
aout independent) for such a network are defined as

(A.8)

For a reciprocal network, the 5 matrix is symmetrical, that is,

S = St (A.9)

where t indicates the transpose of a matrix.

For a lossless network:

∣Sn∣2 + ∣⅞1∣2 = 1 (A.10)

A B
For a cascade of two networks with 5 matrices S and 5 (Figure A.6), the
submatrices of the overall 5 matrix are given by

S1 1 = sf 1 + S⅛S∏FS⅛ (A∙11)
5ι 2 = 5⅛(t⁄÷5fιAS⅛)5f 2 (A.12)
⅛ = S%i FS⅛ (A.13)

∙⅞2 = 22 + S2↑FS22S↑2 (A.14)

a in ------ Cl out ∣
Two-port I
Zoι Network
bin b out '

Port 1 (input) Port 2 (output)

Figure A.5 A two-port network with a and b terms defined.

204 Generalized Filter Design by Computer Optimization

Figure A.6 A cascade connection of two S matrices.

with
1
F = (⁄-⅛SfιΓ (A. 15)

where I is the unit matrix.

S parameters for some of the commonly used two-port networks (see
Figure A.4) are as follows:

• A transmission line

(Z 2 - Zp)sinh yl
(A. 16)
2ZZq cosh yl + (Z 2 + Zp)sinh yl
_____________2ZZ 0 _____________
(A. 17)
2ZZp cosh yl + (Z 2 + Zθ)sinh yl

_____________2ZZ 0 _____________
2
(A. 18)
2ZZθ cosh yl + (Z + Zθ)sinh yl

(Z 2 - Zp)sin hy
(A. 19)
2ZZp cosh yl + (Z 2 + Zp)sinh yl

A series impedance

Z- Z x + Z;
11 (A.20)
Z + Zj + Z'

512 (A.21)
Z;

(A.22)
Z,

522 (A.23)
z + z1 + z
 A shunt admittance

ς Ki - r 2 - r
511 = (A.24)
y + y1 + γ 2
γ2
Sn = (A.25)
γ + y1 + γ2

2yfγiY2
521 = (A.26)
γ + Y1+ ~ Y2

ς r2 - r1 - r
s22 (A.27)
= y ÷ y 1 ÷ y2

A 7⅛etwork

c ⅞-⅞ + -⅞-⅞ + ⅞Z 1 + 4)(4 - Z 2 ) ~ Zq (A.28)

J]∣= 7
Zq + Z x Z 2 + Z 2 Z3 + Z3Z1 + Zq(Zi + Z 2 + 2Z3)
c 2Z 0 Z 2
ji 2 = +
(A.29)
Zq + Z'}Z'⅛ + Z Z∖ + Zq(Z∖ + Z2 + 2Z 5 )

⅛ = ________________2ZqZ2 ________________ (A.30)

Zθ + Z]Z 2 + Z 2 Z 3 + Z3Z] + Zθ(Z∣+ Z 2 + 2Z3)

⅛ _ Z↑Z 2 + Z 2 Z 3 + Z3Z1 - Z 0 (Z1 - Z 2 ) - Zq

(A.31)
Zθ + Z↑Z 2 + Z 2 Z3 + Z3Z1 + Zθ(Z 1 + Z 2 + 2Z3)

A 77-network

* 2
r γ 0 - ( y1 y2 + y2 y3 + y3 y) - y0 ( y1 - y2 )
2
(A.32)
γ 0 + ( y1 y2 ÷ y2 y3 + y3 y) ÷ y0 ( y1 + y2 + 2γ5 )

s ____________________2 y0 y3 __________________
(A.33)
12
" y 02 ÷ ( Y i Y2 ÷ r 2 y3 + y3 y) + y0 ( y1 ÷ y2 + 2γ3 )

c 2 y 0 y3
⅛1 = 72 (A.34)
γ 0 + ( y1 y2 + y2 y3 + y3 y) + y0 ( y1 ÷ y2 ÷ 2 γ 3 )

y⅞ - ( y t y2 ÷ y2 y3 ÷ y3 y) ÷ y0 ( y i - y2 )
(A.35)
2 -
Yo ÷ ( r ι y 2 + y2 y3 + y3 y) ÷ y0 (yι ÷ y2 + 2 γ 5 )
206 Generalized Filter Design by Computer Optimization

A.5 T Parameters
T parameters (⁄zzz2, dependent; bout , aout independent) for the network shown
in Figure A.5 are defined as

ι
W = h (A.36)
L * y ιn J L 7
21 ⅞2 J L a out J

For reciprocal networks:

1
Tn T2 2 - Ti2 7⅞ι = (A.37)
For symmetrical networks:

7⅛ι = - Γ 12 (A.38)
For a cascade of two networks with T matrices, T λ and T B
(Figure A.7), the
overall T matrix is
-
_∣
λ7 -,r '~pA
1
r
r A -1 -r B

⅞ ⅞j [⅛
711 T↑2 11 7 12 7 11
7⅛]'L⅛ (A.39)

A.6 Relationships Between Z and Y Parameters

A.6.1 From Z to Y Parameters

*11
= _ _⅞ _ _-
(A.40)
11¾ -¾2-⅞l

*12 = ~-Z12
-
(A.41)
-¾l-⅞2 ⅞⅞

~⅞1
*21 (A.42)
11¾ - ⅞∑21

= _______?\\ _______
*22 (A.43)
Z∖∖Z22 ~ -¾2-⅞l

Figure A.7 A cascade connection of two T matrices.

 Appendix A Parameters 207

A.6.2 From Y to Z Parameters

7 *22
(A. 44)
11
*11*22- r 12 y21

r
z ~ '2 (A.45)
12 λ
~ * ii*z22 - r 1 2 y 2 1
7 — *21 (A.46)
21
" * l l * z2 2 - *12*21

r
z ∏ (A.47)
22
*11 *22 - *"12*21

A.7 Relationships Between Z and ABCD Parameters

A.7.1 From Z to ABCD Parameters

(A.48)
z2l
d ⅞1⅞2 ~ -¾2⅞
# = (A.49)
zz
21

c= (A.5θ)
z
21

∑22 (A.51)
z
21

A.7.2 From ABCD to Z Parameters

A
Z11 = (A.52)
c
A D - BC
Z12 = c
(A.53)

1
(A.54)
Z =
21 c

D
Zn = (A.55)
~c
208 Generalized Filter Design by Computer Optimization

A.8 Relationships Between Y and ABCD Parameters

A.8.1 From Y to ABCD Parameters

γ
λ ~ 22
a (A.56)
- ¾
(A.57)
Λ21

rι 2 r 2 ι- *11⅛
(A.58)

(A.59)
r2 ι

A.8.2 From ABCD to Y Parameters

y
K11 =
D (A.60)
B
BC-AD
*12= b
(A.61)

r21 = (A.62)
4

γ (A.63)
*22 = jβ

A.9 Relationships Between S and ABCD Parameters

A.9.1 From S to ABCD Parameters

Λ __ ⁄ 1 ÷S,,-⅛-ASĎ J⅛ (A 64 )
Ď 2
⅛1 ⁄ v z
02

B _ ⁄ 1 ÷ ⅝, 2 ÷ ⅛ ÷ Δ5Ď J ⅛z (A 65)
Ď ⅛1 ⁄ v 02

/ 1 - 511 - ⅛ + Δ5Ď ∣Z
C = -----------τ-z------------- A —0 l (A.66)
Ď 2d
21 ⁄ N z O2

Zq∖
 Appendix A Parameters 209

where Zθι and Zθ 2 are the normalized impedances at ports 1 (input) and
2 (output), respectively, and

Δ 5 = S 11 ⅞ 2 - ⅞ 1 5i2 (A.68)

A.9.2 From ABCD to S Parameters

Λ4Z()2 + B - CZq1 Z 0 2 ~ -OZqΔ

(A.69)
(aZo2 + B + CZq↑Zq2 + DZ 0 1 ⁄
⁄ 2(AD - BC)-∖JZq↑Zq2 ∖
(A.70)
(aZ 02 + B + CZq↑Zq2 + ∙C>Zoι )

⁄ 2λ⁄Z 0 i Z0 2 Ď
+ (A.71)
(⁄4Zo2 B + CZq↑Zq2 + DZq↑ J
⁄-AZq2 + B - CZoi Z θ 2 + DZq∖
(A.72)
Ď Zθ2 + B + CZq↑Zq2 + DZq↑

A.10 Relationships Between S and Z Parameters

A.10.1 From S to Z Parameters

7 _ Zqi [(1 + 5 u ) ( l - ⅛) + ⅛⅛]

11 - (A.73)
(1 5 1 1 )(1 - ‰) - ⅛⅛
y _________2Zq1 S1 2 _______
z, 2 = (A.74)
' ( i - s11)(i - s22) - s12s21
2Z0 2 ⅝i
Z21 = (A.75)
(1 - 5 1 1 )(1 - S2 2 ) - s 1 2 ⅛

y _ Z02[(l ~ 11)(1 + ⅛) + 12⅞1]

22 -
(A.76)
(1 2⅛ι ι)(l - ‰) - ⅛⅛
210 Generalized Filter Design by Computer Optimization

A.10.2 From Z to S Parameters

⅞2'⅞1
Z 0 lZ 02
(A.77)
-⅞-⅞l
Zθ]Zθ 2

Z 12
2 √z 2
02
(A.78)
⁄ Zη ⅞2⅞1
Ď⅞ι +
ZθlZ 0 2

______________ Zpi ______________

(A.79)
Zη + 1 W j⅞2 + A _ '⅞2⅞
Z0ι ⁄ĎZ 0 2 ⁄ Z 0 ι Z 02

⅞2⅞1
-⅞l-⅞2
(A.80)
⅞2⅞
Z 0 ι Z 02

A.11 Relationships Between S and Y Parameters

A.11.1 From S to Y Parameters

γ _ 0 1 [ ( l ~ ⅝)(1 + ⅞2) + ⅛¾1

11
(A.81)
( 1 + 5 1 1 ) ( 1 + 522) - S12⅛

v ___________~ 2 *oι⅛ _________

12 =
(A.82)
(1 + S 11 ) d + 5 22 ) - S 12 ⅛

γ ___________ ‰⅞1 _________

21
(A.83)
(1 + 5h)(1 + 522) - - 12⅛1
γ _ ‰[(1 ÷ ⅜1)(1 - ⅛) + ∙S12-⅞1]
22 + + -
(A.84)
(1 11)(l ⅛) ‰⅞ι
 Appendix A Parameters 211

A.11.2 From Y to S Parameters

⁄ 1 Jzι ∆ ⁄ 1 r 22 Ď r 12 r 21
+ r
V ‰Λ W 0 i r 02
(A.85)
11
Λ W 1 Xn∖ >12 ⅛
Ď +
>oι⁄Ď +
M r 0 ι‰

- 2 ‰
. ________________ ‰ _____________
(A.86)
12
⁄ 1 r 1 1 Ď⁄ r 22 Ď r 12 r 21
+ +
V r 0 ιΛ M r 01 y02

ς ______________ ‰ _____________
(A.87)
21
’ Λ Wι 722
Ď 712721

+ +
V r 0 ιΛ M ‰r 02

Λ 2W 1 ⅛Ď >12 ⅞
+ +
>01Λ W >bl>02
(A.88)
622
Λ W 1 >22Ď >12>21
+ +
I >bιΛ >02/ ⅞>02

A.12 Relationships Between S and T Parameters

A.12.1 From T to S Parameters

(A.89)

(A. 90)

(A.91)

(A.92)
212 Generalized Filter Design by Computer Optimization

A.12.2 From S to T Parameters

-
τ ⁄⅛⅛ Sn⅛Ď
11 (A.93)
■ (l ⅛ )
(A.94)

(A.95)

T⅛2 = (A.96)

A.13 Relationships Between ABCD and T Parameters

A.13.1 From ABCD to T Parameters

τ ⁄ 4Zθ2 + B + CZq↑Zq2 + DZ 0 ↑∖ (∖a-7∖

τ
''Λ 2 Zm Z m ) ' a ' 97 >
_ 7 ΛZ m - B t CZ m Z m - DZ m ∖
Ď 2√Z 01 Z 02 ⁄

τ 2 1 _ 7 ⁄l⅞ 8-g⅛-⅞,
Ď
l (A 99)
2ĎZ0iZ02 ⁄
1τ ~ B - CZq↑Zq2 + 7)Z Ď
Ď
01
22 1 - /z„ z7
2λ (A. 100)
√ 01 02 ⁄
A. 13.2 From T to ABCD Parameters

/ 7 1 1 + 7 i2 + 721 + 722Ď ⁄ Z 0 l
A = 1 - l1 ------- 2. ------- - -∣ A ⁄ -=2i (A. 101)
Ď ⁄ V z 02

(A.102)

γ
c. h÷⅞-⅞-⅞ (A|Q3)
2 ⁄Z0iZ02
d __ 7 r 11 - ⅞ - r 12 ÷ ⅞ĎJ⅛ (A io4)

Ď 2
⁄ z
V 01
 Appendix A Parameters 213

A.14 Relationships Between S Parameters and Z oe (Zo0 )

Input Impedance

Figure A.8(a) shows a two-port passive lossless linear network. By using Bartlett’s
bisection theorem, the S parameters for such network can be written as

11 - zoezoo
5ιι = ‰ = (A.105)
(1 + ⅞ )(1 +

⅞1 = 12 = (A.106)
(1 ÷ ⅞6>)(1 + Z oo }

where Z oe and Z oo are the even- and odd-mode impedances of a symmetrical

network. Z oe (Z 00 ) is the input impedance of the two identical one-ports formed
by placing a magnetic (electric) wall at the plane of symmetry, as shown in
Figure A.8(b) and Figure A.8(c), respectively, and is a reactance function.

Selected Bibliography
Bahl, I. J., and P. Bhartia, Microwave Solid State Circuit Design, New York: Wiley, 1988.
Gupta, K. C., R. Garg, and R. Chadha, Computer-Aided Design of Microwave Circuits, Norwood,
MA: Artech House, 1981.
Rizzi, P. A., Microwave Engineering Passive Circuits, Englewood Cliffs, NJ: Prentice Hall, 1988.
Saad, T. S., Microwave Engineers Handbook, Vol. 1, Dedham, MA: Artech House, 1971.
214 Generalized Filter Design by Computer Optimization

Port 1 (input) Port 2 (output)

I I

Two-port
Network
(N)

(a)

Magnetic wall
(open circuit)

N/2

Zoe

Electric wall
(short circuit)

N/2

Zoo

(C)

Figure A.8 (a) Two-port symmetrical network; (b) even-mode impedance; (c) odd-mode
impedance.
Appendix B

B.1 Physical Constants

Velocity of light in free space, c: 2.997925 × 108 m/s

Permittivity of free space, e 0 : 8.854 × 10- 1 2 § (l »36 2r) × 10 9
F/m
Permeability of free space, 0 Â. 4 Œ× 10 7 H/m
Free-space wave impedance, 0Â. 376.7 § 12022@
Boltzmann’s constant, kï 1.380 × 10 O1 J/K
Electron rest mass, 2w t 9.1095 × 10 kg
-27
Proton rest mass: 1.67252 × 10 kg
c7
—19
Electron charge magnitude, eï 1.6022 × 10 1 ' c

B.2 Resistivity and Conductivity of Some Common Metals

Material Resistivity p ( -cm) Conductivity <r(S »m)

Silver 1.6 6.3 × 10 7

Copper 1.7 5.9 × 10 7
Gold 2.4 4.2 × 10 7
Aluminum 2.9 3.4 × 10 7
Brass 6.7 1.5 × 10 7
Platinum 12 8.3 × 10 6
Chromium 13 7.7 × 10 6
Tantalum 14 7.1 × 10 6

215
216 Generalized Filter Design by Computer Optimization

B.3 Properties of Common Substrate Materials at

Microwave Frequencies

Material Relative Permittivity, er Loss Tangent, tan δ

Silicon 11.70-12.90 0.001—0.003

Alumina 9.60-10.10 0.0005-0.002
Germanium 16.00
Gallium arsenide 12.90 0.0005-0.001
Sapphire 9.40 0.0002
Beryllium oxide 6.70 0.001-0.002
PTFE/woven glass 2.84 0.001-0.002
PTFE/microfiberglass 2.26 0.0005-0.001
RT/Duroid 5880 2.26 0.0010
RT/Duroid 6006 6.36
Teflon 2.10
Epsilam 10 13.00
List of Principal Symbols and
Abbreviations

a Scaling parameter
β Phase constant
Δ(⁄) Passband correction factor
δ A step in the calculation of the Jacobian matrix
tan δ Loss tangent
€ Passband ripple level

Relative permittivity
Free-space wave impedance
y Propagation constant
Λf Cutoff wavelength
Guide wavelength
Guide wavelength at the lower bandedge frequency
Midband wavelength
Guide wavelength at the upper bandedge frequency
λ0 Free space wavelength
p0 Permeability of free space
μr Relative permeability
P Reflection coefficient
r Resistivity
s Conductivity
Electrical length of the impedance inverter

217
218 Generalized Filter Design by Computer Optimization

Angular frequency
A, B, C, D Elements of the ABCD matrix
a, b Waveguide housing dimensions
ABCD ABCD matrix
c Velocity of light in free space
Q Capacitance values of the lowpass prototype filter
dt Length of the ⅛h septum
E{xi ) Error vector
Et Tangential component of electric field
Ex x component of electric field
Ey y component of electric field
Ez z component of electric field
Operation frequency
fo Center frequency
fc Cutoff frequency
fL Lower bandedge frequency
fs Stopband frequency
fH Upper bandedge frequency
gi Normalized element values of the lowpass prototype filter
h Thickness of the substrate
Ht Tangential component of magnetic field
Hx x component of magnetic field
Hy y component of magnetic field
Hz z component of magnetic field
I Unity matrix
Jacobian matrix
kc Cutoff wave number
k0 Free-space wave number
Kr Characteristic impedance of rth impedance inverter
Li Insertion loss
Li Inductance values of the lowpass prototype filter
4 Length of the ⅛h resonator
Lr Return loss
Qj, Magnetic hertzian potential
Qu Unloaded quality factor
s Scattering matrix
 List of Principal Symbols and Abbreviations 219

5 ι ι , 512, 52i, Elements of the scattering matrix

⅛2
Gap of the ⅛h ridged waveguide
τ Transmission matrix
t Thickness of the metal insert and the ridged waveguide
h i , η 2 , Γ2 1 , Elements of the T matrix
T2 2
T
f Temperature coefficient
τ1 n ΛZth degree Chebyshev polynomial of the first kind
Xi Vector components
x
sb x pi Normalized reactances
Y Admittance matrix
hi, h 2, hl, Elements of the admittance matrix
r 22
h Guide admittance of the ⅛h mode
z Impedance matrix
Z 1 ι, Z 1 2, Z 2 ι, Elements of the impedance matrix
Z 22
ze Normalized even-mode impedance
z i
Guide impedance of the ⅛h mode
z
o Normalized odd-mode impedance
z r Characteristic impedance of rth unite element
BPF Bandpass filter
BSF Bandstop filter
CAD Computer-aided design
CAFD Computer-aided filter design
CPU Central processor unit
CPW Coplanar waveguide
CPWFIL Software for synthesis and design of coplanar waveguide
filters
CPWFILTER Software for synthesis and design of coplanar waveguide
filters by computer optimization
dB Decibel
det Determinant
DR Dielectric resonator
EM Electromagnetic modeling
EPFIL Software for analysis and design of E-plane waveguide
filters
220 Generalized Filter Design by Computer Optimization

EPFILTER Software for electromagnetic modelling of waveguide

WGMODEL discontinuities
EPSYNFIL Software for synthesis of E-plane waveguide filters
EW Electronic warfare
GHz Gigahertz
GPS Global positioning system
GSM Global system for mobile communications
HFSS High-frequency structure simulator
HPF Highpass filter
KCC Kimberley Communications Consultants
LCFIL Software for synthesis and design of lumped-element
filters
LCFILTER Software for synthesis and design of lumped-element
filters by computer optimization
log Common logarithm
LPF Lowpass filter
MDS Microwave design suite
MMIC Monolithic Microwave Integrated Circuit
PCN Personal communication network
PCS Personal communication system
Quasi-TEM Quasi-transverse electromagnetic
RAM Random access memory
RF Radio frequency
RISC Reduced instruction set computer processor
rwg Ridged waveguide
RWGFIL Software for analysis and design of ridged waveguide
filters
RWGFILTER Software for synthesis and design of ridged waveguide
filters by computer optimization
RWGMODEL Software for electromagnetic modeling of ridged wave-
guide discontinuities
RWGSYNFIL Software for synthesis of ridged waveguide filters
SUN SUN workstation
TE Transverse electric
TEM Transverse electromagnetic
TM Transverse magnetic
UMTS Universal mobile telecommunication system
 List of Principal Symbols and Abbreviations 221

WG-16 22.86-10.16 mm
WLAN Wireless local area network
WR90 22.86-10.16 mm
X-band 8.2-12.4 GHz
About the Author

Djuradj Budimir was born in Serbian Krajina, formerly Yugoslavia. He received

Dipl. Ing. and M.Sc. degrees, both in electronic engineering, from the Univer-
sity of Belgrade, Belgrade, Serbia, and a Ph.D. degree in electronic engineering
from the University of Leeds, Leeds, UK. In March 1994, he joined the
Department of Electronic and Electrical Engineering at King’s College, Univer-
sity of London. Since 1997, he has been with the Department of Electronic
Systems at the University of Westminster in London. His research interests
include analysis and design of hybrid and monolithic microwave integrated
circuits such as amplifiers and filters, dielectric resonator filters employing high-
temperature superconductors for communications systems, the application of
numerical methods to the electromagnetic field analysis of passive microwave
and millimeter-wave circuits, and the design of waveguide filters and multi-
plexing networks for microwave and millimeter-wave applications. Dr. Budimir
has published over 30 technical papers in the field of microwave CAD and
owns a consulting company.

223
Index

ABCD matrix configuration of, 163, 165

CPW filter with, 176, 178 insertion loss, 166
of edge-coupled section, 172 structure, 165
element calculation, 179 two-resonator X-band, 162
of end-coupled section, 172
of impedance inverters, 127, 151 Bandedge frequencies, 135
of two-port network, 168 Bandpass attenuation, bandstop filter, 113
ABCD parameters, 199-201 Bandpass filters
for cascade of networks, 200 center frequency, 82
for parallel networks, 200 coplanar waveguide, 5
for reciprocal networks, 1 99 CPW, 5, 171, 179
relationship with S parameters, 208-9 design procedure, 111-12
relationship with T parameters, 212 E-plane, 5, 47, 129-30, 141-42
relationship with Y parameters, 208 insertion loss response of, 82
relationship with Z parameters, 207 MMIC, 171
for symmetrical networks, 1 99 realization, 117
See also Parameters resonator ladder, 112
Admittance return loss, 111
input, 115 ridged waveguide, 5, 154, 159-60
inverters, 115, 117 stopband attenuation, 111
shunt, S parameters, 205 symmetrical, 92
Approximate methods, 92 Bandstop filters
Approximate synthesis-based bandpass attenuation, 113
design, 103-20, 172-77 design procedure, 113-14
Approximating function, 86 illustrated, 114
Asymmetrical filters return loss, 1 1 3
numerical optimization scheme for, 97 ripple level, 113
ridged waveguide, 162-66 Bandwidth
defined, 82

225
226 Generalized Filter Design by Computer Optimization

Bandwidth (continued) edge-coupled, 72-73

resonator, 29 illustrated, 175
Bifurcated waveguides, scattering modeling, 68-70
parameters, 78-80 multilayer, 22
Broadside-coupled CPW filters, 171 in Xgeom window, 71
See also CPW filters; Transmission lines
CAD programs, 183-98 Coupled microstrip lines, 14, 21
EPFILTER, 195-98 Coupled stripline, 14, 20
LCFILTER, 183-94 CPWFIL program, 183
Capacitors, 25 CPWFILTER program, 183
planar, configurations, 25 CPW filters
shunt, 109, 110, 120 with ABCD matrix for filter
Center frequency, 82 discontinuities, 176, 178
Chebyshev criteria, 6-7, 86-88
accurate design of, 179
Chebyshev function, 171 advantages of, 171
response, 84
approximate synthesis-based design
ripple characteristic, 88 procedure, 172-77
Circular waveguides, 17 bandpass, 5, 171, 179
cross section, 1 3 broadside-coupled, 171
TE mode, 17 circuit representation, 172-73
See also Transmission lines coupling gap, 40
Coaxial lines, 12-13 design of, 171-80
cross-section, 13 design procedure, 173-77
illustrated, 12
dielectric constant of substrate, 174
See also Transmission lines direct-coupled, 171
Coefficient vectors, 59 discontinuity analysis, 70—73
Computer-aided analysis, 4—5 equal-ripple optimization numerical
Computer-aided design (CAD), 1-8 implementation, 177-79
defined, 1
equivalent circuit of, 174, 176
procedure flow chan, 2 examining with Em software, 71
segments, 3-7 impedance inverter, 174
See also CAD programs
layout of, 173
Computer optimization, 5-7
metal thickness, 174
Convergence, 62-68 numerical/experimental results, 179-80
of dominant eigenvalue, 64, 66 resonator determination, 175
in equal-ripple optimization, 9 3
return loss, 174, 178, 180
of higher eigenvalue, 64, 66
ribbon-of-brick-wall, 171
of insertion loss, 63-64, 65, 68
schematic circuit diagram of, 173
of normalized even/odd-mode specifications, 177
impedances, 63, 65
stopband attenuation, 174
of reflection coefficients, 66-67
substrate thickness, 174
Coplanar strips, 14, 21
waveguide gap, 174
Coplanar waveguides, 21 See also Coplanar waveguides
coupling gap, 72 Cutoff frequency
cross section, 14
highpass filter, 109
defined, 21
lowpass filter, 108
discontinuities, 68-73 resonators and, 148
 Index 227

ridged waveguide filter, 147 E-modes, 16

Em software, 4, 70-73
DBFILTER software package, 183 CPW discontinuity analysis with, 70-73
components of, 183 defined, 70
EPFILTER, 128, 195-98 Method of Moments, 4, 70
LCFILTER, 183-94 See also Xgeom
Degree EPFIL program, 183
defined, 82 EPFILTER program, 128, 195-98
determining, 91 defined, 128, 183, 195
Design flow chart, 1 96
CAD, 1-8 function of, 195
CPW filter, 171-80 PRTYPCH subroutine source
E-plane filter, 125 44 code, 197-98
lumped-element filter, 103-22
subroutines, 195
ridged waveguide filter, 147-66 See also CAD programs; EPFILTER program
Diagonal matrix, 60
E-plane discontinuities, 40—48
Dielectric materials, 29-32 evanescent modes between adjacent, 47-48
Alpha/Trans-Tech, 31
frequency dependence, 135
Murata Manufacturing, 32 metal septum in rectangular
NTK Piezoelectric Ceramics, 29
waveguide, 42 46
properties, 29-32 E-plane filter design, 125-44
Tekelec Components, 30 equal-ripple optimization numerical
Dielectric resonators, 28-30 implementation, 138 41
illustrated, 28
with passband correction method, 137-38
materials, 29-30 range of validity for, 134
Dielectric waveguides, 1 9-20 septa electrical parameters, 139
cross sections, 1 9 solution of nonlinear equations, 139-41
image line, 13, 19-20 E-plane filters, 5
types of, 1 9 bandpass, 5, 47
See also Transmission lines
conventional, 161
Direct-coupled cavity filters, 128-34 five-resonator, 142
Direct-coupled CPW filters, 171
with improved stopband
Direct search optimization method, 87
performance, 155
Discontinuities, 39-73 with metal insert, 1 29-30
analysis by mode-matching method, 42-62 photograph, 144
CPW, 39-40, 42, 68-73
specification, 141
electrical behavior of, 39
structure, 126
E-plane, 40 48
circuits, 19, 127
impedance inverters as, 129
insertion loss, 136
in ridged waveguide filters, 48-62
after optimization, 142
Distribution elements, 23
before optimization, 138
Edge-coupled CPW lines, 72-73 calculation, 138-39, 141
Electromagnetic simulation, 3-4 comparison, 142
Electromagnetic simulators, 4, 69 of five-resonator, 142
3-D planar, 70 measured and calculated, 143
Maxwell-Strata, 69 resonator length in, 47
return loss, 131
228 Generalized Filter Design by Computer Optimization

E-plane filters (continued) Image line, 13, 19-20

ripple level, 1 32 Immittance inverters, 1 14—20
septa, 134 approximation to, 118-19
symmetrical, 140 use of, 117
EPSYNF IL program, 183 Impedance inverters, 115, 116
Equal-ripple optimization, 85, 90 ABCD matrix of, 127, 151
convergence criterion in, 93 after impedance scaling, 153
formulation of, 91 of characteristic impedances, 133
numerical implementation CPW filter, 174
of, 120-21, 138-41, 155-57 as discontinuities, 129
ripple maxima, 94, 98-100 for edge-coupled structure, 175-76
as vector procedure, 93 equation derivations, 168-70
equivalent circuit using, 174
Filters frequency dependence of, 155
circuit parameters, 81—83 half-wave prototype using, 150
classification of, 86
illustrated, 150
design flowchart, 90 for intermediate end-coupling structures, 176
design of, 83-85 values, calculating, 1 52
optimization of, 81, 85-91 See also Inverters
optimization tools, 6 Impedance scaling, 1 53
parameters, important, 85 Inductors, 24-25
realization media characteristics, 87 comparison of, 25
See also specific filters planar configurations, 24
Finite-length septum, 60-61
series, 109, 110, 120
Frequency band designations, 16 Input admittance, 115
Generalized Chebyshev, 6, 7 Input impedance, 115, 213
Gradient optimization method, 88-89 Insertion loss, 85, 92
Group delay, 83 after optimization, 142
Guide impedances asymmetrical ridged waveguide filter, 166
frequency dependence of, 155 before optimization, 138
ridged waveguide filter, 150, 154 calculation of, 67, 138-39
comparison, 142
Half-wave prototype convergence of, 63-64, 65, 68
distributed, 153, 154 defined, 82
with impedance inverters, 150 E-plane filter, 136, 141, 163-64
Highpass filters passband, 141, 163
cutoff frequency, 109 response of bandpass filter, 8 2
design procedure, 109-10 ridged waveguide filter, 163-64
illustrated, 110 waveguide bandpass filter, 135
return loss, 109 X-band five-resonator ridged waveguide
ripple, 109 filter, 159
series capacitors, 110 Inverters
shunt inductors, 1 10 admittance, 115, 117
stopband attenuation, 109 immittance, 114-20
H-modes, 14 impedance, 115-16, 127, 150-53,
Hurwitz polynomial, 104 168-70, 174-76
input admittance of, 115
 Index 229

input impedance of, 115 capacitors, 25

K, 135 inductors, 24-25

Jacobian matrix, 95-96 Maxwell-Strata electromagnetic simulator, 69

evaluation of, 96-97 Metal-pipe waveguides, 20
evaluation time, 140 cross section, 1 5
finite difference calculation of, 95, 140 defined, 20
m-by-m, inverse of, 140 See also Transmission lines
of nonlinear functions, 97, 101 Metals, resistivity/conductivity of, 215
Metal septum, 39
LCF IL program, 183
E-plane, 47-48
defined, 184 finite-length, 60-61
source code, 184-94 geometry of, 43
See also DB FILTER software package length determination, 1 54
LCFILTER program, 183-94 between rectangular/ridged
defined, 183 waveguides, 62, 63
flowchart, 185 in rectangular waveguides, 42-46, 53-61
subroutines, 184 in ridged waveguides, 62, 64
See also CAD programs; DBFILTER symmetrical, 44, 46, 48
software package T-equivalent for, 129-30
Lowpass filters thickness, 131, 152
Chebyshev, 104-8 Microshield lines, 22-23
cutoff frequency, 108 cross section, 1 5
design procedure, 108-9 defined, 23
equal ripple, 91 See also Transmission lines
lumped-element, 5, 109 Microstrip lines
prototype, 104-8 coupled, 21
return loss, 108
cross section, 14
series inductors, 109, 122 defined, 21
shunt capacitors, 109, 122 See also Transmission lines
stopband attenuation, 108 MMIC bandpass filters, 171
symmetrical, 91 Modal coefficients, 59
Lumped-element filter design, 83-85
Mode-matching method, 41-62, 163
approximate synthesis-based, 103-20
Multilayer coplanar waveguides, 15, 22
bandpass filter, 111-12
bandstop filter, 113-14 Newton-Raphson method, 95, 96, 132, 136
by optimization, 1 03-22 general form, 140
flowchart, 90 iterations, 141, 164
highpass filter, 109-10 Jacobian evaluation in, 139
immittance inverter, 1 14-20 Numerical optimization
lowpass, 108-9 for asymmetrical filters, 97
procedures, 108-20 equal-ripple, 120-22, 138-41, 155-57
Lumped-element filters, 103-22 for symmetrical filters, 92
approximate synthesis-based design, 103-20
Chebyshev lowpass, 104-8 Optimization, 85-91
lowpass, 5, 109 CPW filter design, 171-80
Lumped elements, 23-25 direct search method, flow chart, 87
E-plane filter design, 125-44
230 Generalized Filter Design by Computer Optimization

Optimization (continued) Reflection coefficients, convergence

equal-ripple, 85, 90, 120-21 of, 66-67
general-purpose techniques, 88, 171 Rejection, 8 3
gradient method, 88-89 Resonators, 25-32
implementing, 81 bandwidth, 29
lumped-element filter design, 103-22 CPW filters and, 175
numerical, 92, 97 cutoff frequency and, 148
ridged waveguide filter design, 147-66 defined, 26-27
Optimization algorithms dielectric, 28-30
asymmetrical case, 97-101 geometric shapes for, 27
description of, 91-101 illustrated, 27-28
direct search, 88 planar, 27
fundamental steps in, 8 8 Return loss
for solving system of nonlinear bandpass filter, 111
equations, 95-96 bandstop filter, 113
symmetrical case, 91-97 before and after, 156, 162, 178
Organization, this book, 7-8 calculated, 119, 158, 162
Orthogonality relation, 76 CPW filter, 174, 178, 180
defined, 8 3
Parameters, 199-214 E-plane, 131
ABCD, 199-201, 207, 208-9, 212 highpass filter, 109
S, 203-5,208-13 lowpass filter, 108
T, 206,211-12 measured, 144
Y, 201-2, 206-7, 208, 210-1 1
passband, 158, 160, 180
Z, 201,206-7, 209-10 ridged waveguide filter, 158, 162
See also Scattering parameters Rhodes synthesis procedure, 128, 134
Passband correction method, 134-38 Ribbon-of-brick-wall CPW filters, 171
CAD algorithm using, 134-37 Ridged waveguide filters
design procedure using, 137-38 advantages of, 160
factors, 135 approximate design of, 1 57
Physical constants, 215 asymmetrical, 162-65
Planar resonators, 27 bandpass, 159-61
Power handling, 8 3 circuit representation, 149-51
Program structure, 62 configuration of, 148-49
Rectangular dielectric waveguides, 1 9 cutoff frequency, 147
cross section, 1 3 design of, 147-66
defined, 19 design procedure, 15 1-55
See also Transmission lines direct-coupled, 148
Rectangular waveguides, 13-17 equal-ripple optimization numerical
illustrated, 1 3 implementation, 155-57
metal septum in, 42-46, 53-62 guide impedance, 150, 154
properties of, 18 insertion loss, 163-64
TE/TM mode support, 14-17 numerical/experimental results, 158-66
See also Transmission lines photograph, 160
Reference plane locations, 1 5 1 resonator length, 157
resonators, normalized impedances, 1 53
 Index 231

return loss before/after relationship with Z parameters, 209-10

optimization, 156, 162 for series impedance, 204
structure illustration, 149 for shunt admittance, 205
symmetrical, 154, 159, 161-64 for T-network, 205
two-port representation of, 154 for transmission line, 204
X-band five-resonator, 158 See also Parameters
Ridged waveguides, 5, 17-18 Stopband attenuation
analysis of, 49-53 bandpass filter, 111
cross section, 1 3 CPW filter, 174
discontinuities in, 48-62 highpass filter, 109
geometry, 49 lowpass filter, 108
metal septum between, 62, 64 Striplines, 20
mode cutoff frequencies, 49 coupled, 14, 20
single-mode operation, 18 cross sections, 14
TE fields in, 49 structure of, 20
See also Transmission lines suspended, 14, 20
Ripple See also Transmission lines
bandstop filter level, 113 Substrate materials, properties of, 216
defined, 8 3 Suspended stripline, 14, 20
E-plane filter level, 132 Symmetrical filters
highpass filter level, 109 bandpass, 92
level definition, 105-6 E-plane, 140
lowpass filter level, 108 lowpass, 91
passband, 106 numerical optimization scheme for, 92
Ritz-Galerkin method, 52 ridged waveguide, 154, 159, 161-64
RWGFIL program, 183
RWGFILTER program, 183 TE fields, 75-76
RWGMODEL program, 183 TE modes, 14-15
RWGSYNFIL program, 183 circular waveguides and, 17
electromagnetic field components, 14-15
Scaling parameter, 152 rectangular waveguide and, 1 7
Scattering matrix, 61, 81 in ridged waveguide, 49
Scattering parameters, 59-60 Temperature range, 83
for bifurcated waveguides, 78-80 Through-reflect-line (TRL) calibration, 159
representation, 61 TM fields, 76
See also Parameters TM modes, 16, 17
Selectivity, 8 3 T-network, S parameters, 205
Series impedance, S parameters, 204 T parameters, 206
Shunt admittance, S parameters, 205 relationship with ABCD parameters, 212
Space-Mapping optimization technique, 70 relationship with S parameters, 211-12
S parameters, 203-5 See also Parameters
for cascade of networks, 203-4 Transmission lines, 1 1-23
for reciprocal network, 203 applications of, 11-12
relationship with ABCD parameters, 208-9 circular waveguide, 17
relationship with input impedance, 213 coaxial line, 12-13
relationship with T parameters, 211-12 comparison of, 23
relationship with Y parameters, 210-11 coplanar line, 21
232 Generalized Filter Design by Computer Optimization

Transmission lines (continued) symmetrical, 214

dielectric waveguide, 19-20 with voltage and current defined, 199
E-plane circuit, 1 9
frequency ranges, 23 Unity matrix, 60
impedance ranges, 23 WGMODEL program, 183
metal-pipe waveguide, 20
microshield line, 22-23 Xgeom, 70-73
microstrip line, 21 CPW discontinuity analysis with, 70-73
multilayer coplanar waveguide, 22 defined, 70
rectangular dielectric-loaded waveguide, 19 See also Em software
rectangular waveguide, 13-17
Y parameters, 201-2
ridged waveguide, 17-18
relationship with ABCD parameters, 208
S parameters for, 204
relationship with S parameters, 210-11
stripline, 20
relationship with Z parameters, 206-7
structures of, 12-15
See also Parameters
unloaded Q-factors, 23
Transmission matrix, 60 Z parameters, 201
Two-port networks, 81 relationship with ABCD parameters, 207
cascade connection of, 200 relationship with S parameters, 209-10
illustrated, 202 relationship with Y parameters, 206-7
lossless, 104 See also Parameters
parallel connection of, 20 1
 The Artech House Microwave Library

Acoustic Charge Transport: Device Technology and Applications,

R. Miller, C Nothnick, and D. Bailey
Advanced Automated Smith Chart Software and User's Manual,
Version 2.0, Leonard M. Schwab
Algorithms for Computer-Aided Design of Linear Microwave Circuits,
Stanislaw Rosloniec
Analysis, Design, and Applications of Fin Lines, Bharathi Bhat and
Shiban K. Koul
Analysis Methods for Electromagnetic Wave Problems,
Eikichi Yamashita, editor
Automated Smith Chart Software and User's Manual,
Leonard M. Schwab
CINL2 for Windows: Linear and Nonlinear Microwave Circuit Analysis
and Optimization, Software and User's Manual, Stephen A. Maas
and Arthur Nichols
Capacitance, Inductance, and Crosstalk Analysis, Charles S. Walker
Computer-Aided Analysis, Modeling, and Design of Microwave
Networks: The Wave Approach, Janusz A. Dobrowolski
Computer-Aided Analysis of Nonlinear Microwave Circuits,
Paulo J. C. Rodrigues
Design of Impedance-Matching Networks for RF and Microwave
Amplifiers, Pieter L. D. Abrie
Designing Microwave Circuits by Exact Synthesis, Brian J. Minnis
Dielectric Materials and Applications, Arthur von Hippel, editor
Dielectrics and Waves, Arthur von Hippel
Digital Microwave Receivers, James B. Tsui
Electric Filters, Martin Hasler and Jacques Neirynck
Electrical and Thermal Characterization of MESFETs, HEMTs, and HBTsl
Robert Anholt
E-Plane Integrated Circuits, P. Bhartia and P. Pramanick, editors
Feedback Maximization, Boris J. Lurie
Filters with Helical and Folded Helical Resonators, Peter Vizmuller
FINCAD:Fin-Line Analysis and Synthesis for Millimeter-Wave
Application Software and User's Manual, S. Raghu Kumar,
Anita Sarraf, and R. Sathyavageeswaran
Frequency Synthesizer Design Toolkit Software and User's Manual,
Version 1.0, James A. Crawford
Fundamentals of Distributed Amplification, Thomas T. Y. Wong
GaAs MESFET Circuit Design, Robert A. Soares, editor
Generalized Filter Design by Computer Optimization, Djuradj Budimir
GASMAP: Gallium Arsenide Model Analysis Program, J. Michael Golio,
et al.
GSPICE for Windows, Sigcad Ltd.
Handbook for the Mechanical Tolerancing o f Waveguide Components,
W. B. W. Alison
HELENA: HEMT Electrical Properties and Noise Analysis Software and
User's Manual, Henri Happy and Alain Cappy
HEMTs and HBTs: Devices, Fabrication, and Circuits, Fazal Ali,
Aditya Gupta, and lnder Bahl, editors
High-Power Microwave Sources, Victor Granatstein and Igor Alexeff,
editors
High-Power GaAs FET Amplifiers, John Walker, editor
High-Power Microwaves, James Benford and John Swegle
Introduction to Microwaves, Fred E. Gardiol
Introduction to Computer Methods for Microwave Circuit Analysis and
Design, Janusz A. Dobrowolski
Introduction to the Uniform Geometrical Theory of Diffraction,
D. A. McNamara, C. W. I. Pistorius, and J. A. G. Malherbe
UNPAR for Windows: Matrix Parameters for Multiconductor
Transmission Lines, Software and User's Manual, Antonije Djordjevic,
Miodrag B. Bazdar, Tapan K. Sarkar, Roger F. Harrington
LOSLIN: Lossy Line Calculation Software and User's Manual,
Fred E. Gardiol
Lossy Transmission Lines, Fred E. Gardiol
Low-Angle Microwave Propagation: Physics and Modeling, Adolf Giger
Low Phase Noise Microwave Oscillator Design, Robert G. Rogers
MATCHNET: Microwave Matching Networks Synthesis,
Stephen V. Sussman-Fort
Matrix Parameters for Multiconductor Transmission Lines: Software
and User's Manual, A. R. Djordjevic, et al.
Microelectronic Reliability, Volume I: Reliability, Test, and Diagnostics,
Edward B. Hakim, editor
Microelectronic Reliability, Volume II: Integrity Assessment and
Assurance, Emiliano Pollino, editor
Microstrip Lines and Slotlines, Second Edition, K. C. Gupta,
Ramesh Garg, lnder Bahl, and Prakash Bhartia
Microwave and Millimeter-Wave Diode Frequency Multipliers,
Marek T. Faber, Jerzy Chamiec, Miroslaw E. Adamski
Microwave and RF Circuits: Analysis, Synthesis, and Design, Max Medley
Microwave and RF Component and Subsystem Manufacturing
Technology, Heriot-Watt University
Microwaves and Wireless Simplified, Thomas S. Laverghetta
Microwave Circulator Design, Douglas K. Linkhart
Microwave Engineers' Handbook, Two Volumes, Theodore Saad, editor
Microwave Materials and Fabrication Techniques, Second Edition,
Thomas S. Laverghetta
Microwave MESFETs and HEMTs, J. Michael Golio, et al.
Microwave and Millimeter Wave Heterostructure Transistors and
Applicatons, F. Ali, editor
Microwave and Millimeter Wave Phase Shifters, Volume I: Dielectric
and Ferrite Phase Shifters, S. Koul, and B. Bhat
Microwave and Millimeter Wave Phase Shifters, Volume II:
Semiconductor and Delay Line Phase Shifters, S. Koul and B. Bhat
Microwave Mixers, Second Edition, Stephen Maas
Microwave Transmission Design Data, Theodore Moreno
Microwave Transition Design, Jamal S. Izadian and Shahin M. Izadian
Microwave Transmission Line Couplers, J. A. G. Malherbe
Microwave Tubes, A. S. Gilmour, Jr.
Microwaves: Industrial, Scientific, and Medical Applications, J. Thuery
Microwaves Made Simple: Principles and Applicatons,
Stephen W. Cheung, Frederick H. Levien, et al.
MMIC Design: GaAs FETs and HEMTs, Peter H. Ladbrooke
Modern GaAs Processing Techniques, Ralph Williams
Modern Microwave Measurements and Techniques,
Thomas S. Laverghetta
Monolithic Microwave Integrated Circuits: Technology and Design,
Ravender Goyal, et al.
MULTLIN for Windows: Circuit-Analysis Models for Multiconductor
Transmssion Lines, Software and User's Manual, Antonije R.
Djordjevic, Darko D. Cvetkovic, Goran M. Cujic, Tapan K. Sarkar,
Miodrag B. Bazdar
Nonuniform Line Microstrip Directional Couplers, Sener Uysal
PC Filter: Electronic Filter Design Software and User's Guide,
Michael G. Ellis, Sr.
PLL: Linear Phase-Locked Loop Control Systems Analysis Software and
User's Manual, Eric L. Unruh
RF Design Guide: Systems, Circuits, and Equations, Peter Vizmuller
Scattering Parameters of Microwave Networks with Multiconductor
Transmission Lines: Software & User's Manual, A. R. Djordjevic, et al.
Solid-State Microwave Power Oscillator Design, Eric Holzman and
Ralston Robertson
Terrestrial Digital Microwave Communications, Ferdo lvanek, et al.
Transmission Line Design Handbook, Brian C. Waddell
TRAVIS Pro: Transmission Line Visualization Software and User's
Manual, Professional Version, Robert G. Kaires and
Barton T. Hickman
TRAVIS Student: Transmission Line Visualization Software and User's
Manual, Student Version, Robert G. Kaires and Barton T. Hickman
Yield and Reliability in Microwave Circuit and System Design,
Michael Meehan and John Purviance

For further information on these and other Artech House titles, includ-
ing previously considered out-of-print books now available through
our In-Print-Forever™ (I PF™) program, contact:

Artech House Artech House

685 Canton Street Portland House, Stag Place
Norwood, MA 02062 London SW1E 5XA England
781-769-9750 +44(0) 171-973-8077
Fax: 781-769-6334 Fax: +44(0) 171-630-0166
Telex: 951-659 Telex: 951-659
e-mail: [email protected] e-mail: [email protected]

Find us on the World Wide Web at: www.artech-house.com

 ISBN 0-85006-579-9
90000

9 780890 065792

Artech House Publishers BOSTON • LONDON