MICROWAVE AND RF DESIGN: MODULES Third Edition Volume 4

STEER
MICROWAVE
Microwave and RF Design: Modules focuses on the design of systems based on microwave modules. The
use of modules has become increasingly important in RF and microwave engineering for rapidly realizing
high performance microwave systems. When integration is ultimately to be used, building a system up using
modules provides a rapid means of prototyping and testing system concepts. A wide variety of RF modules

AND RF DESIGN
including amplifiers, local oscillators, switches, circulators, isolators, phase detectors, frequency multipliers
and dividers, phase-locked loops, and direct digital synthesizers are considered. Detailed design strategies
for synthesizing filters based on parallel coupled lines are presented. The reader will gain an appreciation
of design by synthesis. This book is suitable as both an undergraduate and graduate textbook, as well as a
career-long reference book.

MICROWAVE AND RF DESIGN: MODULES, Volume 4

MODULES
KEY FEATURES OTHER VOLUMES
• The fourth volume of a comprehensive series on Microwave and RF Design
microwave and RF design
Radio Systems
• Open access ebook editions are hosted by NC State
Volume 1
University Libraries at:
ISBN 978-1-4696-5690-8
https://repository.lib.ncsu.edu/handle/1840.20/36776
• 23 worked examples
• An average of 21 exercises per chapter Microwave and RF Design
• Answers to selected exercises Transmission Lines
• 6 case studies illustrating design procedures Volume 2
• Emphasis on synthesis as well as building a rich ISBN 978-1-4696-5692-2
library of microwave functions
• A companion book, Fundamentals of Microwave Microwave and RF Design
and RF Design, is suitable as a comprehensive Networks
undergraduate textbook on microwave engineering Volume 3
ISBN 978-1-4696-5694-6
ABOUT THE AUTHOR
Microwave and RF Design
Michael Steer is the Lampe Distinguished Professor
Amplfiers and Oscillators
of Electrical and Computer Engineering at North
Volume 5
Carolina State University. He received his B.E. and Ph.D.
degrees in Electrical Engineering from the University of ISBN 978-1-4696-5698-4
Queensland. He is a Fellow of the IEEE and is a former
editor-in-chief of IEEE Transactions on Microwave ALSO BY THE AUTHOR
Theory and Techniques. He has authored more than 500 Fundamentals of Microwave
publications including twelve books. In 2009 he received and RF Design
a US Army Medal, “The Commander’s Award for Public ISBN 978-1-4696-5688-5
Service.” He received the 2010 Microwave Prize and the
2011 Distinguished Educator Award, both from the IEEE
Microwave Theory and Techniques Society.

Published by NC State University Distributed by UNC Press

Michael Steer Third Edition

Microwave and RF Design
Modules
Volume 4
Third Edition
Michael Steer
Microwave and RF Design
Modules

Volume 4
Third Edition
Michael Steer

c 2019 by M.B. Steer

Copyright 

Citation: Steer, Michael. Microwave and RF Design: Modules. Volume 4. (Third

Edition), NC State University, 2019. doi: https//doi.org/10.5149/
9781469656977 Steer

This work is licensed under a Creative Commons Attribution-NonCommercial

4.0 International license (CC BY-NC 4.0). To view a copy of the license, visit
http://creativecommons.org/licenses.

ISBN 978-1-4696-5696-0 (paperback)

ISBN 978-1-4696-5697-7 (open access ebook)

Published by NC State University

Distributed by the University of North Carolina Press

www.uncpress.org

Printing: 1
To my Son- and Daughter-in-Laws

Claudio and Tracy

Preface
The book series Microwave and RF Design is a comprehensive treatment
of radio frequency (RF) and microwave design with a modern “systems-
first” approach. A strong emphasis on design permeates the series with
extensive case studies and design examples. Design is oriented towards
cellular communications and microstrip design so that lessons learned can
be applied to real-world design tasks. The books in the Microwave and RF
Design series are:

• Microwave and RF Design: Radio Systems, Volume 1

• Microwave and RF Design: Transmission Lines, Volume 2
• Microwave and RF Design: Networks, Volume 3
• Microwave and RF Design: Modules, Volume 4
• Microwave and RF Design: Amplifiers and Oscillators, Volume 5

The length and format of each is suitable for automatic printing and binding.
Rationale
The central philosophy behind this series’s popular approach is that the
student or practicing engineer will develop a full appreciation for RF and
microwave engineering and gain the practical skills to perform system-
level design decisions. Now more than ever companies need engineers with
an ingrained appreciation of systems and armed with the skills to make
system decisions. One of the greatest challenges facing RF and microwave
engineering is the increasing level of abstraction needed to create innovative
microwave and RF systems. This book series is organized in such a way that
the reader comes to understand the impact that system-level decisions have
on component and subsystem design. At the same time, the capabilities of
technologies, components, and subsystems impact system design. The book
series is meticulously crafted to intertwine these themes.
Audience
The book series was originally developed for three courses at North
Carolina State University. One is a final-year undergraduate class, another an
introductory graduate class, and the third an advanced graduate class. Books
in the series are used as supplementary texts in two other classes. There
are extensive case studies, examples, and end of chapter problems ranging
from straight-forward to in-depth problems requiring hours to solve. A
companion book, Fundamentals of Microwave and RF Design, is more suitable
for an undergraduate class yet there is a direct linkage between the material
in this book and the series which can then be used as a career-long reference
text. I believe it is completely understandable for senior-level students
where a microwave/RF engineering course is offered. The book series is a
comprehensive RF and microwave text and reference, with detailed index,
appendices, and cross-references throughout. Practicing engineers will find
the book series a valuable systems primer, a refresher as needed, and a
vi PREFACE

reference tool in the field. Additionally, it can serve as a valuable, accessible

resource for those outside RF circuit engineering who need to understand
how they can work with RF hardware engineers.
Organization
This book is a volume in a five volume series on RF and microwave
design. The first volume in the series, Microwave and RF Design: Radio
Systems, addresses radio systems mainly following the evolution of cellular
radio. A central aspect of microwave engineering is distributed effects
considered in the second volume of this book series, Microwave and RF
Design: Transmission Lines. Here transmission lines are treated as supporting
forward- and backward-traveling voltage and current waves and these
are related to electromagnetic effects. The third volume, Microwave and RF
Design: Networks, covers microwave network theory which is the theory
that describes power flow and can be used with transmission line effects.
Topics covered in Microwave and RF Design: Modules, focus on designing
microwave circuits and systems using modules introducing a large number
of different modules. Modules is just another term for a network but the
implication is that is is packaged and often available off-the-shelf. Other
topics that are important in system design using modules are considered
including noise, distortion, and dynamic range. Most microwave and RF
designers construct systems using modules developed by other engineers
who specialize in developing the modules. Examples are filter and amplifier
modules which once designed can be used in many different systems. Much
of microwave design is about maximizing dynamic range, minimizing noise,
and minimizing DC power consumption. The fifth volume in this series,
Microwave and RF Design: Amplifiers and Oscillators, considers amplifier and
oscillator design and develops the skills required to develop modules.
Volume 1: Microwave and RF Design: Radio Systems
The first book of the series covers RF systems. It describes system concepts
and provides comprehensive knowledge of RF and microwave systems.
The emphasis is on understanding how systems are crafted from many
different technologies and concepts. The reader gains valuable insight into
how different technologies can be traded off in meeting system requirements.
I do not believe this systems presentation is available anywhere else in such
a compact form.
Volume 2: Microwave and RF Design: Transmission Lines
This book begins with a chapter on transmission line theory and introduces
the concepts of forward- and backward-traveling waves. Many examples are
included of advanced techniques for analyzing and designing transmission
line networks. This is followed by a chapter on planar transmission lines
with microstrip lines primarily used in design examples. Design examples
illustrate some of the less quantifiable design decisions that must be made.
The next chapter describes frequency-dependent transmission line effects
and describes the design choices that must be taken to avoid multimoding.
The final chapter in this volume addresses coupled-lines. It is shown how to
design coupled-line networks that exploit this distributed effect to realize
novel circuit functionality and how to design networks that minimize
negative effects. The modern treatment of transmission lines in this volume
emphasizes planar circuit design and the practical aspects of designing
PREFACE vii

around unwanted effects. Detailed design of a directional coupler is used

to illustrate the use of coupled lines. Network equivalents of coupled lines
are introduced as fundamental building blocks that are used later in the
synthesis of coupled-line filters. The text, examples, and problems introduce
the often hidden design requirements of designing to mitigate parasitic
effects and unwanted modes of operation.
Volume 3: Microwave and RF Design: Networks
Volume 3 focuses on microwave networks with descriptions based on S
parameters and ABCD matrices, and the representation of reflection and
transmission information on polar plots called Smith charts. Microwave
measurement and calibration technology are examined. A sampling of
the wide variety of microwave elements based on transmission lines is
presented. It is shown how many of these have lumped-element equivalents
and how lumped elements and transmission lines can be combined as a
compromise between the high performance of transmission line structures
and the compactness of lumped elements. This volume concludes with an
in-depth treatment of matching for maximum power transfer. Both lumped-
element and distributed-element matching are presented.
Volume 4: Microwave and RF Design: Modules
Volume 4 focuses on the design of systems based on microwave modules.
The book considers the wide variety of RF modules including amplifiers,
local oscillators, switches, circulators, isolators, phase detectors, frequency
multipliers and dividers, phase-locked loops, and direct digital synthesizers.
The use of modules has become increasingly important in RF and microwave
engineering. A wide variety of passive and active modules are available
and high-performance systems can be realized cost effectively and with
stellar performance by using off-the-shelf modules interconnected using
planar transmission lines. Module vendors are encouraged by the market
to develop competitive modules that can be used in a wide variety of
applications. The great majority of RF and microwave engineers either
develop modules or use modules to realize RF systems. Systems must also
be concerned with noise and distortion, including distortion that originates
in supposedly linear elements. Something as simple as a termination
can produce distortion called passive intermodulation distortion. Design
techniques are presented for designing cascaded systems while managing
noise and distortion. Filters are also modules and general filter theory is
covered and the design of parallel coupled line filters is presented in detail.
Filter design is presented as a mixture of art and science. This mix, and the
thought processes involved, are emphasized through the design of a filter
integrated throughout this chapter.
Volume 5: Microwave and RF Design: Amplifiers and Oscillators
The fifth volume presents the design of amplifiers and oscillators in
a way that enables state-of-the-art designs to be developed. Detailed
strategies for amplifiers and voltage-controlled oscillators are presented.
Design of competitive microwave amplifiers and oscillators are particularly
challenging as many trade-offs are required in design, and the design
decisions cannot be reduced to a formulaic flow. Very detailed case studies
are presented and while some may seem quite complicated, they parallel the
level of sophistication required to develop competitive designs.
viii PREFACE

Case Studies

A key feature of this book series is the use of real world case studies
of leading edge designs. Some of the case studies are designs done in
my research group to demonstrate design techniques resulting in leading
performance. The case studies and the persons responsible for helping to
develop them are as follows.
1. Software defined radio transmitter.
2. High dynamic range down converter design. This case study was
developed with Alan Victor.
3. Design of a third-order Chebyshev combline filter. This case study was
developed with Wael Fathelbab.
4. Design of a bandstop filter. This case study was developed with Wael
Fathelbab.
5. Tunable Resonator with a varactor diode stack. This case study was
developed with Alan Victor.
6. Analysis of a 15 GHz Receiver. This case study was developed with
Alan Victor.
7. Transceiver Architecture. This case study was developed with Alan
Victor.
8. Narrowband linear amplifier design. This case study was developed
with Dane Collins and National Instruments Corporation.
9. Wideband Amplifier Design. This case study was developed with Dane
Collins and National Instruments Corporation.
10. Distributed biasing of differential amplifiers. This case study was
developed with Wael Fathelbab.
11. Analysis of a distributed amplifier. This case study was developed with
Ratan Bhatia, Jason Gerber, Tony Kwan, and Rowan Gilmore.
12. Design of a WiMAX power amplifier. This case study was developed
with Dane Collins and National Instruments Corporation.
13. Reflection oscillator. This case study was developed with Dane Collins
and National Instruments Corporation.
14. Design of a C-Band VCO. This case study was developed with Alan
Victor.
15. Oscillator phase noise analysis. This case study was developed with
Dane Collins and National Instruments Corporation.
Many of these case studies are available as captioned YouTube videos and
qualified instructors can request higher resolution videos from the author.

Course Structures

Based on the adoption of the first and second editions at universities,

several different university courses have been developed using various parts
of what was originally one very large book. The book supports teaching
two or three classes with courses varying by the selection of volumes
and chapters. A standard microwave class following the format of earlier
microwave texts can be taught using the second and third volumes. Such
a course will benefit from the strong practical design flavor and modern
treatment of measurement technology, Smith charts, and matching networks.
Transmission line propagation and design is presented in the context of
microstrip technology providing an immediately useful skill. The subtleties
of multimoding are also presented in the context of microstrip lines. In such
PREFACE ix

a class the first volume on microwave systems can be assigned for self-
learning.
Another approach is to teach a course that focuses on transmission line
effects including parallel coupled-line filters and module design. Such a class
would focus on Volumes 2, 3 and 4. A filter design course would focus
on using Volume 4 on module design. A course on amplifier and oscillator
design would use Volume 5. This course is supported by a large number of
case studies that present design concepts that would otherwise be difficult to
put into the flow of the textbook.
Another option suited to an undergraduate or introductory graduate class
is to teach a class that enables engineers to develop RF and microwave
systems. This class uses portions of Volumes 2, 3 and 4. This class then omits
detailed filter, amplifier, and oscillator design.
The fundamental philosophy behind the book series is that the broader
impact of the material should be presented first. Systems should be discussed
up front and not left as an afterthought for the final chapter of a textbook, the
last lecture of the semester, or the last course of a curriculum.
The book series is written so that all electrical engineers can gain an
appreciation of RF and microwave hardware engineering. The body of the
text can be covered without strong reliance on this electromagnetic theory,
but it is there for those who desire it for teaching or reader review. The book
is rich with detailed information and also serves as a technical reference.
The Systems Engineer
Systems are developed beginning with fuzzy requirements for components
and subsystems. Just as system requirements provide impetus to develop
new base technologies, the development of new technologies provides new
capabilities that drive innovation and new systems. The new capabilities
may arise from developments made in support of other systems. Sometimes
serendipity leads to the new capabilities. Creating innovative microwave
and RF systems that address market needs or provide for new opportunities
is the most exciting challenge in RF design. The engineers who can
conceptualize and architect new RF systems are in great demand. This book
began as an effort to train RF systems engineers and as an RF systems
resource for practicing engineers. Many RF systems engineers began their
careers when systems were simple. Today, appreciating a system requires
higher levels of abstraction than in the past, but it also requires detailed
knowledge or the ability to access detailed knowledge and expertise. So what
makes a systems engineer? There is not a simple answer, but many partial
answers. We know that system engineers have great technical confidence and
broad appreciation for technologies. They are both broad in their knowledge
of a large swath of technologies and also deep in knowledge of a few
areas, sometimes called the “T” model. One book or course will not make
a systems engineer. It is clear that there must be a diverse set of experiences.
This book series fulfills the role of fostering both high-level abstraction of
RF engineering and also detailed design skills to realize effective RF and
microwave modules. My hope is that this book will provide the necessary
background for the next generation of RF systems engineers by stressing
system principles immediately, followed by core RF technologies. Core
technologies are thereby covered within the context of the systems in which
they are used.
x PREFACE

Supplementary Materials
Supplementary materials available to qualified instructors adopting the book
include PowerPoint slides and solutions to the end-of-chapter problems.
Requests should be directed to the author. Access to downloads of the books,
additional material and YouTube videos of many case studies are available
at https://www.lib.ncsu.edu/do/open-education
Acknowledgments
Writing this book has been a large task and I am indebted to the many
people who helped along the way. First I want to thank the more than
1200 electrical engineering graduate students who used drafts and the
first two editions at NC State. I thank the many instructors and students
who have provided feedback. I particularly thank Dr. Wael Fathelbab, a
filter expert, who co-wrote an early version of the filter chapter. Professor
Andreas Cangellaris helped in developing the early structure of the book.
Many people have reviewed the book and provided suggestions. I thank
input on the structure of the manuscript: Professors Mark Wharton and
Nuno Carvalho of Universidade de Aveiro, Professors Ed Delp and Saul
Gelfand of Purdue University, Professor Lynn Carpenter of Pennsylvania
State University, Professor Grant Ellis of the Universiti Teknologi Petronas,
Professor Islam Eshrah of Cairo University, Professor Mohammad Essaaidi
and Dr. Otman Aghzout of Abdelmalek Essaadi Univeristy, Professor
Jianguo Ma of Guangdong University of Technology, Dr. Jayesh Nath of
Apple, Mr. Sony Rowland of the U.S. Navy, and Dr. Jonathan Wilkerson
of Lawrence Livermore National Laboratories, Dr. Josh Wetherington of
Vadum, Dr. Glen Garner of Vadum, and Mr. Justin Lowry who graduated
from North Carolina State University.
Many people helped in producing this book. In the first edition I was
assisted by Ms. Claire Sideri, Ms. Susan Manning, and Mr. Robert Lawless
who assisted in layout and production. The publisher, task master, and chief
coordinator, Mr. Dudley Kay, provided focus and tremendous assistance in
developing the first and second editions of the book, collecting feedback
from many instructors and reviewers. I thank the Institution of Engineering
and Technology, who acquired the original publisher, for returning the
copyright to me. This open access book was facilitated by John McLeod
and Samuel Dalzell of the University of North Carolina Press, and by Micah
Vandergrift and William Cross of NC State University Libraries. The open
access ebooks are host by NC State University Libraries.
The book was produced using LaTeX and open access fonts, line art was
drawn using xfig and inkscape, and images were edited in gimp. So thanks
to the many volunteers who developed these packages.
My family, Mary, Cormac, Fiona, and Killian, gracefully put up with my
absence for innumerable nights and weekends, many more than I could have
ever imagined. I truly thank them. I also thank my academic sponsor, Dr.
Ross Lampe, Jr., whose support of the university and its mission enabled me
to pursue high risk and high reward endeavors including this book.
Michael Steer
North Carolina State University
Raleigh, North Carolina
mbs@ncsu.edu
List of Trademarks

3GPP R
is a registered trademark of the European Telecommunications Stan-
dards Institute.
802R
is a registered trademark of the Institute of Electrical & Electronics En-
gineers .
APC-7 R
is a registered trademark of Amphenol Corporation.
R
AT&T is a registered trademark of AT&T Intellectual Property II, L.P.
AWR R
is a registered trademark of National Instruments Corporation.
AWRDE R
is a trademark of National Instruments Corporation.
Bluetooth R
is a registered trademark of the Bluetooth Special Interest Group.

R
GSM is a registered trademark of the GSM MOU Association.
Mathcad R
is a registered trademark of Parametric Technology Corporation.
MATLAB R
is a registered trademark of The MathWorks, Inc.

R
NEC is a registered trademark of NEC Corporation.
OFDMA R
is a registered trademark of Runcom Technologies Ltd.

Qualcomm R is a registered trademark of Qualcomm Inc.
Teflon R
is a registered trademark of E. I. du Pont de Nemours.
R
RFMD is a registered trademark of RF Micro Devices, Inc.
SONNET R
is a trademark of Sonnet Corporation.
Smith is a registered trademark of the Institute of Electrical and Electronics
Engineers.
Touchstone R
is a registered trademark of Agilent Corporation.

R
WiFi is a registered trademark of the Wi-Fi Alliance.
WiMAX R
is a registered trademark of the WiMAX Forum.

All other trademarks are the properties of their respective owners.

Contents

Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v

1 Introduction to RF and Microwave Modules . . . . . . . . . . . . . 1

1.1 Introduction to Microwave Modules . . . . . . . . . . . . . . . 1
1.2 A 15 GHz Receiver Subsystem . . . . . . . . . . . . . . . . . . 2
1.3 Book Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.4 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.A RF and Microwave Circuit Schematic Symbols . . . . . . . . . 5
1.A.1 Element and Circuit Symbols . . . . . . . . . . . . . . . 5
1.A.2 Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.A.3 Diodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.A.4 Bipolar Junction Transistor . . . . . . . . . . . . . . . . 8
1.A.5 Junction Field Effect Transistor . . . . . . . . . . . . . . 9
1.A.6 Metal-Oxide-Semiconductor Field Effect Transistor . . 9

2 Filters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.1.1 Filter Prototypes . . . . . . . . . . . . . . . . . . . . . . 12
2.1.2 Image Parameter Versus Insertion loss Methods . . . . 13
2.2 Singly and Doubly Terminated Networks . . . . . . . . . . . . 13
2.2.1 Doubly Terminated Networks . . . . . . . . . . . . . . 14
2.2.2 Lowpass Filter Response . . . . . . . . . . . . . . . . . 15
2.3 The Lowpass Filter Prototype . . . . . . . . . . . . . . . . . . . 17
2.4 The Maximally Flat (Butterworth) Lowpass Approximation . 18
2.4.1 Butterworth Filter Design . . . . . . . . . . . . . . . . . 18
2.4.2 Construction of the Transfer Function . . . . . . . . . . 20
2.4.3 nth-Order Reflection Approximation . . . . . . . . . . 21
2.4.4 Bandwidth Consideration . . . . . . . . . . . . . . . . . 21
2.5 The Chebyshev Lowpass Approximation . . . . . . . . . . . . 22
2.5.1 Chebyshev Filter Design . . . . . . . . . . . . . . . . . 22
2.5.2 Chebyshev Approximation and Recursion . . . . . . . 23
2.5.3 Bandwidth Consideration . . . . . . . . . . . . . . . . . 24
2.6 Element Extraction . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.6.1 Ladder Synthesis . . . . . . . . . . . . . . . . . . . . . . 24
2.6.2 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.7 Butterworth and Chebyshev Filters . . . . . . . . . . . . . . . 29
2.7.1 Butterworth Filter . . . . . . . . . . . . . . . . . . . . . 30
2.7.2 Chebyshev Filter . . . . . . . . . . . . . . . . . . . . . . 31
2.7.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.8 Impedance and Admittance Inverters . . . . . . . . . . . . . . 33
2.8.1 Properties of an Impedance Inverter . . . . . . . . . . . 34
2.8.2 Replacement of a Series Inductor by a Shunt Capacitor 34
2.8.3 Replacement of a Series Capacitor by a Shunt Inductor 36
2.8.4 Ladder Prototype with Impedance Inverters . . . . . . 36
xiv CONTENTS

2.8.5 Lumped-Element Realization of an Inverter . . . . . . 37

2.8.6 Narrowband Realization of an Inverter Using Trans-
mission Line Stubs . . . . . . . . . . . . . . . . . . . . 39
2.9 Filter Transformations . . . . . . . . . . . . . . . . . . . . . . . 40
2.9.1 Impedance Transformation . . . . . . . . . . . . . . . . 40
2.9.2 Frequency Transformation: Lowpass . . . . . . . . . . 41
2.9.3 Lowpass to Highpass Transformation . . . . . . . . . . 42
2.9.4 Lowpass to Bandpass Transformation . . . . . . . . . . 42
2.9.5 Lowpass to Bandstop Transformation . . . . . . . . . . 45
2.9.6 Transformed Ladder Prototypes . . . . . . . . . . . . . 46
2.10 Cascaded Line Realization of Filters . . . . . . . . . . . . . . . 46
2.11 Butterworth and Chebyshev Bandpass Filters . . . . . . . . . 48
2.12 Richards’s Transformation . . . . . . . . . . . . . . . . . . . . . 50
2.12.1 Richards’s Transformation and Transmission Lines . . 51
2.12.2 Richards’s Transformation and Stubs . . . . . . . . . . 52
2.12.3 Richards’s Transformation Applied to a Lowpass Filter 52
2.12.4 Richards’s Transformation Applied to a Highpass Filter 53
2.13 Kuroda’s and Norton’s Network Identities . . . . . . . . . . . 54
2.13.1 Kuroda’s Identities . . . . . . . . . . . . . . . . . . . . . 54
2.13.2 Norton’s Identities . . . . . . . . . . . . . . . . . . . . . 56
2.14 Inter-resonator Coupled Bandpass Filters . . . . . . . . . . . . 57
2.15 Bandpass Filter Topologies . . . . . . . . . . . . . . . . . . . . 61
2.16 Case Study: Design of a Bandstop Filter . . . . . . . . . . . . . 62
2.17 Active Filters . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
2.17.1 Radio Frequency Active Filters . . . . . . . . . . . . . . 66
2.17.2 Biquadratic Filters . . . . . . . . . . . . . . . . . . . . . 67
2.17.3 Distributed Active Filters . . . . . . . . . . . . . . . . . 68
2.18 Transient Response of a Bandpass Filter . . . . . . . . . . . . . 69
2.19 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
2.20 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
2.21 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
2.21.1 Exercises by Section . . . . . . . . . . . . . . . . . . . . 77
2.21.2 Answers to Selected Exercises . . . . . . . . . . . . . . 77

3 Parallel Coupled-Line Filters . . . . . . . . . . . . . . . . . . . . . . 79

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
3.2 Parallel Coupled Lines . . . . . . . . . . . . . . . . . . . . . . . 80
3.2.1 Coupled-Line Configurations . . . . . . . . . . . . . . 83
3.2.2 Coupled-Line Circuit Models . . . . . . . . . . . . . . . 87
3.3 Inverter Network Scaling . . . . . . . . . . . . . . . . . . . . . 91
3.4 Case Study: Third-Order Chebyshev Combline Filter Design . 93
3.4.1 Realization of the Input/Output Inverters . . . . . . . 101
3.4.2 Implementation . . . . . . . . . . . . . . . . . . . . . . 102
3.4.3 Alternative Combline Filter Layouts . . . . . . . . . . . 112
3.5 Parallel Coupled-Line Filters in an Inhomogeneous Medium . 114
3.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
3.7 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
3.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
3.8.1 Exercises by Section . . . . . . . . . . . . . . . . . . . . 120
3.8.2 Answers to Selected Exercises . . . . . . . . . . . . . . 120
CONTENTS xv

4 Noise, Distortion, and Dynamic Range . . . . . . . . . . . . . . . . 123

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
4.2 Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
4.2.1 Observations of Noise Spectra . . . . . . . . . . . . . . 124
4.2.2 Characterization of Thermal Noise . . . . . . . . . . . 126
4.2.3 Environmental Noise . . . . . . . . . . . . . . . . . . . 128
4.2.4 Thermal Noise and Capacitors . . . . . . . . . . . . . . 129
4.2.5 Physical Source of Shot Noise . . . . . . . . . . . . . . 129
4.2.6 Physical Source of Flicker Noise . . . . . . . . . . . . . 130
4.3 Noise Characterization . . . . . . . . . . . . . . . . . . . . . . . 130
4.3.1 Noise in a Cascaded System . . . . . . . . . . . . . . . 133
4.3.2 Measurement of Noise Figure . . . . . . . . . . . . . . 135
4.3.3 Measurement of Noise Temperature . . . . . . . . . . . 137
4.3.4 Measuring the Noise Figure of Low Noise Devices . . 137
4.3.5 Radiometer System . . . . . . . . . . . . . . . . . . . . 138
4.3.6 Noise Figure of a Two-Port Amplifier . . . . . . . . . . 139
4.4 Oscillator Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
4.4.1 Observations of Noise Spectra of Oscillators . . . . . . 141
4.5 Nonlinear Distortion . . . . . . . . . . . . . . . . . . . . . . . . 144
4.5.1 Amplitude and Phase Distortion . . . . . . . . . . . . . 144
4.5.2 Gain Compression . . . . . . . . . . . . . . . . . . . . . 145
4.5.3 Intermodulation Distortion . . . . . . . . . . . . . . . . 146
4.5.4 Third-Order Distortion . . . . . . . . . . . . . . . . . . 147
4.5.5 Spectral Regrowth . . . . . . . . . . . . . . . . . . . . . 149
4.5.6 Second-Order Distortion . . . . . . . . . . . . . . . . . 150
4.5.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . 150
4.6 Dynamic Range . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
4.7 Passive Intermodulation Distortion . . . . . . . . . . . . . . . 154
4.7.1 Sources of PIM . . . . . . . . . . . . . . . . . . . . . . . 154
4.7.2 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . 156
4.8 Breakdown . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
4.8.1 Multipactor Effect . . . . . . . . . . . . . . . . . . . . . 157
4.8.2 Corona Effect . . . . . . . . . . . . . . . . . . . . . . . . 157
4.8.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . 158
4.9 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
4.10 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
4.11 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
4.11.1 Exercises by Section . . . . . . . . . . . . . . . . . . . . 163
4.11.2 Answers to Selected Exercises . . . . . . . . . . . . . . 163

5 Passive Modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165

5.1 Introduction to Passive Modules . . . . . . . . . . . . . . . . . 165
5.2 Diodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
5.3 Case Study: Tunable Resonator with Varactor Diode Stack . . 170
5.4 Switch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
5.5 Ferrite Components: Circulators and Isolators . . . . . . . . . 177
5.5.1 Gyromagnetic Effect . . . . . . . . . . . . . . . . . . . . 177
5.5.2 Circulator . . . . . . . . . . . . . . . . . . . . . . . . . . 179
5.5.3 Circulator Isolation . . . . . . . . . . . . . . . . . . . . 179
5.5.4 Isolator . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
5.5.5 YIG-Tuned Bandpass Filter . . . . . . . . . . . . . . . . 181
xvi CONTENTS

5.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181

5.7 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
5.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
5.8.1 Exercises By Section . . . . . . . . . . . . . . . . . . . . 184
5.8.2 Answers to Selected Exercises . . . . . . . . . . . . . . 184

6 Mixer and Source Modules . . . . . . . . . . . . . . . . . . . . . . . 185

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
6.2 Mixer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
6.2.1 Mixer Analysis . . . . . . . . . . . . . . . . . . . . . . . 187
6.2.2 Mixer Performance Parameters . . . . . . . . . . . . . . 189
6.2.3 Mixer Waveforms . . . . . . . . . . . . . . . . . . . . . 191
6.2.4 Switching Mixer . . . . . . . . . . . . . . . . . . . . . . 194
6.2.5 Subharmonic Mixer . . . . . . . . . . . . . . . . . . . . 196
6.3 Single-Ended, Balanced, and Double Balanced Mixers . . . . . 197
6.3.1 RFIC Mixers . . . . . . . . . . . . . . . . . . . . . . . . 197
6.3.2 Gilbert Cell . . . . . . . . . . . . . . . . . . . . . . . . . 198
6.3.3 Integrated Gilbert Cell-Based Mixer . . . . . . . . . . . 198
6.3.4 Ring-Based Mixers . . . . . . . . . . . . . . . . . . . . . 200
6.3.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . 201
6.4 Local Oscillator . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
6.4.1 Phase Noise in Local Oscillators . . . . . . . . . . . . . 202
6.5 Voltage-Controlled Oscillator (VCO) . . . . . . . . . . . . . . . 204
6.6 Phase Detector . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
6.7 Frequency Multiplier . . . . . . . . . . . . . . . . . . . . . . . . 207
6.8 Frequency Divider . . . . . . . . . . . . . . . . . . . . . . . . . 208
6.9 Phase-Locked Loop (PLL) . . . . . . . . . . . . . . . . . . . . . 209
6.9.1 Operation . . . . . . . . . . . . . . . . . . . . . . . . . . 209
6.9.2 First-Order PLL . . . . . . . . . . . . . . . . . . . . . . . 210
6.9.3 Applications . . . . . . . . . . . . . . . . . . . . . . . . 211
6.10 Direct Digital Synthesizer . . . . . . . . . . . . . . . . . . . . . 213
6.11 Diode and Vacuum Sources . . . . . . . . . . . . . . . . . . . . 214
6.11.1 Two-Terminal Semiconductor Sources . . . . . . . . . . 214
6.11.2 Vacuum Devices . . . . . . . . . . . . . . . . . . . . . . 216
6.12 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
6.13 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
6.14 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
6.14.1 Exercises By Section . . . . . . . . . . . . . . . . . . . . 221
6.14.2 Answers to Selected Exercises . . . . . . . . . . . . . . 221

7 Cascade of Modules . . . . . . . . . . . . . . . . . . . . . . . . . . . 223

7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
7.2 Nonlinear Distortion in a Cascaded System . . . . . . . . . . . 225
7.2.1 Gain Compression in a Cascaded System . . . . . . . . 225
7.2.2 Intermodulation Distortion in a Cascaded System . . . 228
7.3 Cascaded Module Design Using the Budget Method . . . . . 231
7.4 Cascaded Module Design Using the Contribution Method . . 232
7.4.1 Noise Contribution . . . . . . . . . . . . . . . . . . . . 233
7.4.2 Intermodulation Contribution . . . . . . . . . . . . . . 234
7.4.3 Design Methodology for Maximizing Dynamic Range 235
7.4.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . 236
CONTENTS xvii

7.5 Case Study: High Dynamic Range Down-Converter Design . 236

7.5.1 Architecture . . . . . . . . . . . . . . . . . . . . . . . . . 237
7.5.2 Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
7.6 Case Study: Analysis of a 15 GHz Receiver . . . . . . . . . . . 238
7.7 Case Study: Frequency Planning of a Transceiver . . . . . . . 241
7.7.1 Transceiver Architecture . . . . . . . . . . . . . . . . . 242
7.7.2 Frequency Planning . . . . . . . . . . . . . . . . . . . . 244
7.7.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . 247
7.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
7.9 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
7.10 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
7.10.1 Exercises By Section . . . . . . . . . . . . . . . . . . . . 250
7.10.2 Answers to Selected Exercises . . . . . . . . . . . . . . 250

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
 CHAPTER 1
Introduction to RF and Microwave
Modules

1.1 Introduction to Microwave Modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 A 15 GHz Receiver Subsystem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Book Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.4 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.A Appendix: RF and Microwave Circuit Schematic Symbols . . . . . . . . 5

1.1 Introduction to Microwave Modules

Most microwave design develops microwave systems using modules such
amplifiers, integrated circuits (ICs), filters, frequency multipliers, and
passive components. Economics necessitate that since modules are expensive
to design, e.g RF integrated circuits, that they be developed for multiple
applications. In system design modules are chosen for their dynamic
range, noise performance, DC power consumption, and cost. Foremost
the system designer must have knowledge of available modules and be
prepared to design a module itself if this results in competitive performance
and manages cost. Modules are interconnected by transmission lines, bias
settings and matching networks must be designed, and a system frequency
plan must be developed that trades-off cost and performance while avoiding
interference. All these topics are addressed on this book which develops the
knowledge of modules, and the skills required to design microwave systems.
Most RF systems are composed of a cascade of modules each of which
is separately designed and characterized. They often have 50 Ω input and
output impedances so that modules can be freely interconnected. Just as
often matching networks must be design by the user as otherwise operating
frequencies are pre-determined by the module vendor as a matching
network usually has a narrower bandwidth than that of the functional
core of the module. In high volume applications, several modules could be
monolithically integrated, but even then, design is based on the concept of
modules. Many modules are available “off the shelf” and high-performance
RF systems can be constructed using them.
Examples of commercially available modules are shown in Figure 1-1.
They can range in complexity from the surface mount resistor of Figure 1-
1(a) and the transformer of Figure 1-1(d), up to the mixer and synthesizer
modules shown in Figures 1-1(b and c). Modules can have very good
performance as it is cost effective to put considerable design effort into a
module that can be used in many applications and thus design costs shared.
2 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 1-1: Modules in

surface-mount packages.
Copyright Synergy Mi-
crowave Corporation, used
with permission [1]. (a) Resistor (b) Mixer (c) Synthesizer (d) Transformer

1.2 A 15 GHz Receiver Subsystem

An example of a microwave subsystem designed using modules is the
15 GHz receiver subsystem shown in Figure 1-2 with details of the frequency
conversion section shown in Figure 1-3. This receiver subsystem is itself
a module used in building a complete transceiver system for a point-to-
point microwave link connecting two basestations. The modules of the
receiver include amplifiers, frequency multipliers, a mixers, an isolator, and a
waveguide adaptor. These modules are available as off-the-shelf components
from companies that specialize in developing them, and selling them to a
large user base. This receiver is the subject of a case study in Section 7.6 where
more details of the modules, frequency planning, and signal flow are given.

1.3 Book Outline

This book describes modules and tools required to design microwave
systems using modules. A large number of modules are presented with
the aim of building a ’module vocabulary.’ An engineer must know what
modules are available before a microwave system can be designed using
modules.
The first set of modules examined are filters in Chapter 2. These have
the important task of only letting wanted signals through a circuit while
blocking noise and unwanted signals. Then parallel coupled line (PCL) filters
are considered in Chapter 3. A PCL filter is the most common type of
microwave filter and is based on the inherent filtering properties of a pair
of coupled transmission lines.
Chapter 4 introduces metrics for noise, distortion, and dynamic range and
introduces the tools that are required to design systems of cascaded modules
that minimize noise and distortion, and maximize dynamic range.
Chapter 5 develops knowledge of passive modules, particular ones that
use diodes and switches. The final chapter, Chapter 6, considers mixer and
source modules.
This book is the fourth volume in a series on microwave and RF design.
The first volume in the series addresses radio systems [2] mainly following
the evolution of cellular radio. A central aspect of microwave engineering is
distributed effects considered in the second volume of the series [3]. Here
the transmission lines are treated as supporting forward- and backward-
traveling voltage and current waves and these are related to electromagnetic
effects. The third volume [4] covers microwave network theory which is
the theory that describes power flow in microwave circuits and can be
used to describe transmission line effects. Topics covered in this volume
include scattering parameters, Smith charts, and matching networks that
INTRODUCTION TO RF AND MICROWAVE MODULES 3

A SMA connector, Reference LO in H 50 Ω transmission line

B Attenuator I DC blocking capacitor
C DC blocking capacitor J Mitered bend
D Reference LO amplifier K 50 Ω transmission line
E Bias line L Waveguide-to-microstrip adaptor, RF in
F Radial stub M Interface to frequency conversion section
G High-impedance transmission line N Interface to IF section

Figure 1-2: A 14.4–15.35 GHz receiver module itself consisting of cascaded modules
interconnected by microstrip transmission lines. Surrounding the microwave circuit are DC
conditioning and control circuitry. RF in is 14.4 GHz to 15.35 GHz, LO in is 1600.625 MHz to
1741.875 MHz. The frequency of the IF is 70–1595 MHz. Detail of the frequency conversion
section mounted on the mat is shown in Figure 1-3. The silk-screened mat provides a strain-
relieving surface for mounting ceramic- and semiconductor-based modules.

enable maximum power transfer. The fifth volume in this series [5] considers
amplifier and oscillator design and develops the skills required to develop
modules.
The books in the Microwave and RF Design series are:
• Microwave and RF Design: Radio Systems
• Microwave and RF Design: Transmission Lines
• Microwave and RF Design: Networks
• Microwave and RF Design: Modules
• Microwave and RF Design: Amplifiers and Oscillators
4 STEER MICROWAVE AND RF DESIGN: MODULES

A Reference LO amplifier K RF amplifier

B Power supply decoupling capacitor L Power supply decoupling capacitor
C ×2 frequency multiplier M Power supply decoupling capacitor
D Edge-coupled PCL bandpass filter N Voltage variable attenuator
E Microstrip bend O Power supply decoupling capacitor
F Microstrip transmission line P Power supply decoupling capacitor
G ×2 frequency multiplier Q RF amplifier
H Power supply decoupling capacitor R Microstrip transmission line and bend
I stepped impedance lowpass filter S Edge-coupled PCL bandpass filter
J Microstrip transmission line T Subharmonic mixer

Figure 1-3: Frequency conversion section of the receiver module shown in Figure 1-2. The
reference LO is applied to the frequency conversion section at c, the RF is applied at h following
the isolator. The IF is output at j.

1.4 References
[1] http://www.synergymwave.com. [6] IEEE Standard 315-1975, Graphic Symbols
[2] M. Steer, Microwave and RF Design, Radio Sys- for Electrical and Electronics Diagrams
tems, 3rd ed. North Carolina State Univer- (Including Reference Designation Letters),
sity, 2019. Adopted Sept. 1975, Reaffirmed Dec. 1993.
[3] ——, Microwave and RF Design, Transmission Approved by American National Standards
Lines, 3rd ed. North Carolina State Univer- Institute, Jan. 1989. Approved adopted for
sity, 2019. mandatory use, Department of Defense,
[4] ——, Microwave and RF Design, Networks, United States of America, Oct. 1975. Ap-
3rd ed. North Carolina State University, proved by Canadian Standards Institute, Oct.
2019. 1975.
[5] ——, Microwave and RF Design, Amplifiers and [7] R. Baker, CMOS Circuit Design, Layout, and
Oscillators, 3rd ed. North Carolina State Uni- Simulation, 2nd ed. Wiley-Interscience, IEEE
versity, 2019. Press, 2008.
INTRODUCTION TO RF AND MICROWAVE MODULES 5

Appendix
1.A RF and Microwave Circuit Schematic Symbols
1.A.1 Element and Circuit Symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.A.2 Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.A.3 Diodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.A.4 Bipolar Junction Transistor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.A.5 Junction Field Effect Transistor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.A.6 Metal-Oxide-Semiconductor FET . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

This appendix lists the symbols commonly used with RF and microwave circuits. Symbols
are from IEEE Standard 315-1975 [6]. Up until the 1970s IEEE was active in establishing
standard symbols for all electrical engineering fields and in particular circuit schematic
symbols to be used with microwave circuits. Since then vendors of microwave computer-aided
design tools have developed their own symbols but very often a vendor tends to adopt symbols
similar to those used by other vendors. However there are differences and as a result there
has not been a consensus to adopt a more modern standard for microwave symbols. What is
presented in this chapter follows the earlier IEEE standard where possible and for components
that are not in the standard, an attempt has been made to select symbols that are in common
use in technical papers.

1.A.1 Element and Circuit Symbols

Table 1-1: IEEE standard qualifying properties added to schematic symbols to identify a
particular property.

Qualifying property Symbol

Adjustable

Adjustable, continuously adjustable

Adjustable, stepped

Linear

Nonlinear

Positive
Negative
6 STEER MICROWAVE AND RF DESIGN: MODULES

Table 1-2: Standard schematic symbols of RF and microwave components.

Component Symbol Alternate Component Symbol Alternate

Analog to digital Capacitor, variable
converter
Attenuator, fixed Capacitor, nonlinear

Attenuator, balanced Capacitor, shielded

Attenuator, Circulator
unbalanced
Coaxial cable
Attenuator, variable
Conductive path

Attenuator, continu- Connector, female

ously variable Connector, male

Attenuator, stepped Contact, fixed

Contact, closed

Amplifier Contact, open

Delay
Antenna, general
Digital-to-analog
converter (DAC)
Antenna, balanced
Element, linear (* to
be replaced by des-
Antenna, dipole ignation)

Antenna, loop Ground, general

Balun Ground, chassis

Balun with coaxial Coupler

line and dipole an-
tenna Filter, bandpass fil-
Capacitor, general ter (BPF)

Capacitor, polarized Filter, lowpass filter

(LPF)
INTRODUCTION TO RF AND MICROWAVE MODULES 7

Component Symbol Alternate Component Symbol Alternate

Filter, highpass fil- Shield
ter (HPF)
Filter, bandstop fil- Short, movable
ter (BSF)
Isolator Source, AC

Inductor, general Source, DC

Inductor with
Switch,
magnetic core
multiposition
Junction
Test, point
Junction of paths
Transformer
Network, linear
(* to be replaced by
Transformer with
designation)
magnetic core
Open
Transformer, cen-
Phase shifter ter tapped

Piezoelectric
resonator
Port Triax

Power divider
Twinax

Radio link Twinax with

shield showing
Radio link with
connection
antennas
Twinax with
Rectifier shield grounded
Resistor, general
Short
Resistor, variable Wire
Wires, connected
Resistor, nonlinear
Resistor with open
path Wires, uncon-
nected, crossing
Resistor with short
8 STEER MICROWAVE AND RF DESIGN: MODULES

1.A.2 Sources
Commonly used symbols for sources.

Component Symbol
Voltage source Component Symbol
Controlled current source
Current source
Voltage noise source
AC source
Current noise source
Controlled voltage source

1.A.3 Diodes
IEEE standard symbols for diodes and a rectifier [6]. (1 In the direction of anode (A) to cathode
(K). 2 Use symbol for general diode unless it is essential to show intrinsic region.)

Component Symbol
Component Symbol
Diode, general
(including Schottky)1 Rectifier 1
1
Gunn diode Tunnel diode1
IMPATT diode1
Varactor diode1 or
1,2
PIN diode
Zener diode1
Light emitting diode (LED)1

1.A.4 Bipolar Junction Transistor

IEEE standard schematic symbols for bipolar junction transistors (BJT and HBT) [6] and
commonly used symbols in layouts [7]. The letters indicate terminals: B (base), C (collector), E
(emitter).

IEEE Commonly used IEEE Commonly used

Transistor symbol symbol Transistor symbol symbol

BJT, pnp BJT, npn

INTRODUCTION TO RF AND MICROWAVE MODULES 9

1.A.5 Junction Field Effect Transistor

Table 1-3: IEEE standard schematic symbols for junction field effect transistors (MESFET,
HEMT, JFET) [6] and symbols more commonly used in schematics. The letters indicate
terminals: G (gate), D (drain), S (source).
IEEE Commonly used
Transistor symbol symbol

FET, pJFET

FET, nJFET, MESFET, HEMT

1.A.6 Metal-Oxide-Semiconductor Field Effect Transistor

Table 1-4: IEEE standard schematic symbols for MOSFET transistors [6] and symbols more
commonly used in schematics [7]. The MOSFET symbols are for enhancement- and depletion-
mode transistors. The letters indicate terminals: G (gate), D (drain), S (source), U (bulk). Four-
terminal and three-terminal common symbols are shown. The three-terminal common symbol
is most often used when the bulk is connected to the most negative connection in the circuit,
and the pMOSFET symbol is used when the bulk is tied to VDD (the most positive connection).
The bulk connection is often not shown, as it is assumed to be connected to the most negative
voltage point.
IEEE Commonly used Commonly used
Transistor symbol symbol (three-terminal) symbol (four-terminal)

FET, nMOS, depletion

FET, pMOS, depletion

FET, nMOS, enhancement

FET, pMOS, enhancement

 CHAPTER 2

Filters

With Wael Fathelbab

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 Singly and Doubly Terminated Networks . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3 The Lowpass Filter Prototype . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.4 The Maximally Flat (Butterworth) Lowpass Approximation . . . . . . 18
2.5 The Chebyshev Lowpass Approximation . . . . . . . . . . . . . . . . . . . . . . . 22
2.6 Element Extraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.7 Butterworth and Chebyshev Filters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.8 Impedance and Admittance Inverters . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.9 Filter Transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.10 Cascaded Line Realization of Filters . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.11 Butterworth and Chebyshev Bandpass Filters . . . . . . . . . . . . . . . . . . 48
2.12 Richards’s Transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
2.13 Kuroda’s and Norton’s Network Identities . . . . . . . . . . . . . . . . . . . . . . . 54
2.14 Inter-resonator Coupled Filters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
2.15 Bandpass Filter Topologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
2.16 Design of a Bandstop Filter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
2.17 Active Filters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
2.18 Transient Response of a Bandpass Filter . . . . . . . . . . . . . . . . . . . . . . . 69
2.19 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
2.20 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
2.21 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

2.1 Introduction
Filters are the most fundamental of signal processing circuits using energy
storage elements to obtain frequency-dependent characteristics. Some of
the important filter attributes are (1) controlling noise by not allowing
out-of-band noise to propagate in a circuit; (2) keeping signals outside
the transmit band, especially harmonics, from being transmitted; and (3)
presenting only signals in a specified band to active receive circuitry. At
microwave frequencies a filter can consist solely of lumped elements, solely
of distributed elements, or a mix of lumped and distributed elements. The
distributed realizations can be transmission line-based implementations of
12 STEER MICROWAVE AND RF DESIGN: MODULES

the components of lumped-element filter prototypes or, preferably, make

use of the particular frequency characteristics found with certain distributed
structures. Loss in lumped elements, particularly above a few gigahertz,
means that the performance of distributed filters nearly always exceeds
that of lumped-element filters. However, since the basic component of
a distributed filter is a one-quarter wavelength long transmission line,
distributed filters can be prohibitively large below a few gigahertz.
Only a few basic types of responses are required of most RF filters as
follows:

(a) Lowpass—providing maximum power transfer at frequencies below

the corner frequency, f0 . Above f0 , transmission is blocked. See Figure
2-1(a).

(b) Highpass—passing signals at frequencies above f0 . Below f0 , transmis-

sion is blocked. See Figure 2-1(b).

(c) Bandpass—passing signals at frequencies between lower and upper

corner frequencies (defining the passband) and blocking transmission
outside the band. This is the most common type of RF filter. See Figure
2-1(c).

(d) Bandstop (or notch)—which blocks signals between lower and upper
corner frequencies (defining the stopband). See Figure 2-1(d).

(e) Allpass—which equalizes a signal by adjusting the phase generally to

correct for phase distortion elsewhere. See Figure 2-1(e).

The above list is not comprehensive, as actual operating conditions may

mandate specific frequency profiles. For example, WiFi systems operating at
2.45 GHz are susceptible to potentially large signals being transmitted from
nearby cellular phones operating in the 1700–2300 MHz range. Thus a front-
end filter for a 2.45 GHz system must ensure very high levels of attenuation
over the 1700–2300 MHz range. Thus the optimum solution here, probably,
will have an asymmetrical frequency response with high rejection on one
side of the passband obtained by accepting lower rejection on the other side.

2.1.1 Filter Prototypes

In the 1960s an approach to RF filter design and synthesis was developed
and this is still followed. The approach is to translate the mathematical
response of a lowpass filter. A filter with the desired lowpass response is
then synthesized using lumped elements and the resulting filter is called a
lowpass prototype. The lowpass filter is then transformed so that the new
lumped-element filter has the desired RF response, such as highpass or
bandpass. In the case of a bandpass filter, each inductor and capacitor of the
lowpass prototype becomes a resonator that is coupled to another resonator.
In distributed form, this basic resonator is a one-quarter wavelength long
transmission line.
Filter synthesis is a systematic approach to realizing circuits with desired
frequency characteristics. A filter can also be thought of as a hardware
implementation of a differential equation producing a specific impulse,
FILTERS 13

(a) Lowpass (LPF) (b) Highpass (HPF) (c) Bandpass (BPF) (d) Bandstop (BSF)

(e) Allpass

Figure 2-1: Ideal filter transfer function, T (f ), responses.

step, or frequency response. It is not surprising that a filter can be

described this way as the lumped reactive elements, L and C, describe first-
order differential equations and their interconnection describes higher-order
differential equations.
The synthesis of a filter response begins by identifying the desired
differential equation response. The differential equations are specified using
the Laplace variable, s, so that in the frequency domain (with s = ω) an nth-
order differential equation becomes an nth-order polynomial expression in
s. If possible, and this is usually the case, a lowpass version of a filter can be
developed. The conversion to the final filter shape, say a bandpass response,
proceeds mathematically through a number of stages. Each stage has a circuit
form and each stage is called a filter prototype.

2.1.2 Image Parameter Versus Insertion loss Methods

Filter networks can be synthesized using the image parameter method or
the reflection coefficient (or insertion loss) method [1]. The image parameter
measured is based on cascading two-port networks with each two-port
having the essential desired filter response [2]. Several of these two-ports are
cascaded to achieve a periodic-like structure that has the final desired filter
response. This method is restrictive and arbitrary filter responses can not be
obtained. It is rarely used in RF and microwave filter design but is useful in
analyzing simple structures with cascades of identical elements [3]. A better
approach is to use the insertion loss method as used in this chapter.

2.2 Singly and Doubly Terminated Networks

Filters are generally two-port networks and they provide maximum power
transfer from a source to a load over a specified frequency range while
rejecting the transmission of signals at other frequencies. Two possible filter
networks are shown in Figure 2-2. The network in Figure 2-2(a) is known
as a doubly terminated network, as both ports are resistively terminated.
14 STEER MICROWAVE AND RF DESIGN: MODULES

(a) Doubly terminated network (b) Singly terminated network

Figure 2-2: Terminated networks.

The network in Figure 2-2(b) is called a singly terminated network, as only

one port is terminated in a resistor. The doubly terminated network is
much closer to the type of network required at RF where loads and source
impedances are finite. A singly terminated network is applicable in some RF
integrated circuit applications where very little RF power is involved and
often when feedback is used. In such cases the output of an RFIC amplifier
can approximate an ideal voltage source as the Thevenin equivalent source
impedance can be negligible. Most synthesis of singly terminated filter
networks uses an analogous procedure to that presented here for doubly
terminated filters.

2.2.1 Doubly Terminated Networks

The established filter synthesis procedure for doubly terminated filter
networks focuses on realizing the input reflection coefficient. For a lossless
filter the square of the reflection coefficient magnitude is one minus the
square of the insertion loss and this is the origin of this method being called
the insertion loss method.
The input reflection coefficient of the doubly terminated network in Figure
2-2(a) is
Zin,1 (s) − RS
Γ1 (s) = , (2.1)
Zin,1 (s) + RS

where the reference impedance is the source resistance, RS , and s is the

Laplace variable. In the passband of a filter the reflection coefficient is
approximately zero.
There are several other parameters used in filter design and these will be
introduced now. The transducer power ratio (TPR) is defined as

Maximum power available from the source

TPR =
Power absorbed by the load
 1 R V (s) 2
  
1 2
(Vg (s)/2) /RS L g
= 2 1 2 =  . (2.2)
 
V
2 2 (s)/R L  2 R V
S 2 (s) 

The transmission coefficient, T (s), of the network is


1 RS V2 (s)
T (s) =  =2 , (2.3)
TPR(s) RL Vg (s)

IF the filter is lossless, the insertion loss (IL) (or transducer function) is
2
IL(s) = TPR(s) = |1/T (s)| . (2.4)
FILTERS 15

The next step in development of the filter synthesis procedure is

introduction of the characteristic function, defined as the ratio of the
reflection and transmission coefficients:
Γ1 (s) N (s)
K (s) = = , (2.5)
T (s) D (s)
where N is the numerator function and D is the denominator function. Ide-
ally the filter network is lossless and so, from the principle of energy conser-
vation, the sum of the magnitudes squared of the transmission and reflection
coefficients must be unity:
2 2 2 2
|T (s)| + |Γ1 (s)| = 1. (2.6) That is, |T (s)| = 1 − |Γ1 (s)| . (2.7)
Dividing both sides of Equation (2.7) by |T (s)|2 results in
2
1 |Γ1 (s)| 2
1= 2 − 2 , (2.8) or 1 = |IL(s)| − |K (s)| . (2.9)
|T (s)| |T (s)|
Rearranging Equation (2.9) leads to
2
|IL(s)| = 1 + |K (s)| (2.10)
1 2
|K(s)|
and so |T (s)|2 = 2 (2.11) and |Γ1 (s)|2 = . (2.12)
1 + |K (s)| 1 + |K(s)|2
In the above it is seen that both the reflection and transmission coefficients
are functions of the characteristic function of the two-port network. A
milestone has been reached. In an RF filter, the frequency-dependent
insertion loss or transmission coefficient is of most importance, as these
are directly related to power flow. Equation (2.11) shows that this can be
expressed in terms of another function, K(s), which, from Equation (2.5),
can be expressed as the ratio of N (s) and D(s). For lumped-element circuits,
N (s) and D(s) are polynomials.

2.2.2 Lowpass Filter Response

As an example of the relationship of the various filter responses, consider the
lossless lowpass filter responses shown in Figure 2-3. This filter has a lowpass
response with a corner frequency expressed in radians as ωc = 1 rad/s.
Ideally T (s) for s ≤  (note s = ω) would be one, and for s > , T (s) would
be zero. It is not possible to realize such an ideal response, and the response
shown in Figure 2-3(a) is typical of what can be achieved. The reflection
coefficient is shown in Figure 2-3(b), with the characteristic function K(s)
shown in Figure 2-3(c). If |K(s)|2 is expressed as the ratio of two polynomials
(i.e, as N (s)/D(s)), then it can be seen from Figure 2-3(e and f) that the zeros
2
of the reflection coefficient, |Γ1 (s)| , are also the zeros of N (s). Also, it is
observed that the zeros of the transmission coefficient are also the zeros of
D(s), as shown in Figure 2-3(a and c).
The first objective in lumped-element filter design is development of the
transfer function of the network in the frequency domain or, equivalently, the
s domain.1 The input-output transfer function of the generic filter in Figure
2-4 is

1 Care is being taken in the use of terminology as s can be complex with some filter types, but
these filters are realized using digital signal processing.
16 STEER MICROWAVE AND RF DESIGN: MODULES

(a) (d)

(b) (e)

(c) (f)

Figure 2-3: An example of a lowpass filter in terms of various responses: (a) transmission
coefficient; (b) reflection coefficient response; and (c) characteristic function response. Detailed
responses are shown in (d), (e), and (f), respectively.
FILTERS 17

Figure 2-4: Generic filter.

T (s) = Y (s)/X(s) (2.13)

and the design procedure is to match this response to a response defined by

the ratio of two polynomials:

N (s) am sm + am−1 sm−1 + · · · + a1 s + a0

T (s) = = . (2.14)
D(s) sn + bn−1 sn−1 + · · · + b1 s + b0

Here N stands for numerator and D for denominator, and these are not the
same as those in Equation (2.5) (where they were just labels for numerator
and denominator). The filter response using a pole-zero description can
be synthesized so the design process begins by rewriting Equation (2.14)
explicitly in terms of zeros, zm , and poles, pn :

N (s) am (s + z1 )(s + z2 ) · · · (s + zm−1 )(s + zm )

T (s) = = . (2.15)
D(s) (s + p1 )(s + p2 ) · · · (s + pn−1 )(s + pn )

Since only the frequency response is of interest, s = ω, thus simplifying the
analysis for sinusoidal signals.
The poles and zeros can be complex numbers and can be plotted on the
complex s plane. Conditions imposed by realizable circuits require that D(s)
be a Hurwitz polynomial,2 which ensures that its poles are located in the
left-half plane. N (s) determines the location of the transmission zeros of the
filter, and the order of N (s) cannot be more than the order of D(s). That is,
n ≥ m so that the filter has finite or zero response at infinite frequency.
Two strategies can be employed in deriving the filter response. The first is
to derive the polynomials N (s) and D(s) in Equation (2.14). This seems like
an open-ended problem, but it was discovered in the 1950s and 1960s that
in normal situations there are only a few types of useful responses that are
described by a few polynomials, including Butterworth, Bessel, Chebyshev,
and Cauer polynomials.

2.3 The Lowpass Filter Prototype

Most filter design is based on the synthesis of a lowpass filter equivalent
called a lowpass prototype. Transformations are then used to correct for the
actual source and load impedances, frequency range, and the desired filter
type, such as highpass or bandpass. An ideal lowpass response is shown in
Figure 2-5(a), which shows a transmission coefficient of 1 up to a normalized
frequency of 1 rad/sec. This type of response defines what is called a brick-
wall filter, which unfortunately cannot be physically realized. Instead, the
response is approximated and specified in terms of a response template (see
Figure 2-5(b)).

2 A Hurwitz polynomial is a polynomial whose coefficients are positive real numbers and whose
zeros are located in the left-half of the complex s plane.
18 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 2-5: Lowpass

filter response. (a) Ideal brick-wall response (b) Response template

The template specifies a passband response that is between β1 and β2 at

frequencies below the corner radian frequency (at ω = 1), and below β3
above a radian frequency ω ′ > 1. The lowpass filter is harder to realize the
closer the specified response is to the ideal response shown in Figure 2-5(a),
that is, when
′
(β2 − β1 ) → 0, β3 → 0, and (ω − 1) → 0. (2.16)

Several polynomials have particularly interesting characteristics that

match the requirements of the response template. At first it may seem
surprising that polynomial functions could yield close to ideal filter
responses. However, as will be seen, there is something special about
these polynomials: they are natural solutions to extreme conditions. For
example, the nth-order Butterworth polynomial has the special property
that the first n derivatives at s = 0 are zero. The other polynomials
also have extreme properties, and filters that are synthesized using them
also have extreme properties. As well as the maximally flat property
resulting from Butterworth polynomials, filter responses obtained using
Chebyshev polynomials have the steepest skirts (i.e., fastest transitions from
the frequency region where signals are passed to the region where they are
blocked). Usually it is one of these extreme cases that is most desirable.
A streamlining of the filter synthesis procedure is obtained by first
focusing on the development of normalized lowpass filters having a 1 rad/s
corner frequency and 1 Ω reference impedance. Transformations transform
a lowpass filter prototype into another filter having the response desired.
These transformations give rise to symmetrical responses.

2.4 The Maximally Flat (Butterworth) Lowpass

Approximation
Butterworth filters have a maximally flat response (Figure 2-6) that in the
time domain corresponds to a critically damped system.

2.4.1 Butterworth Filter Design

Butterworth filters have the transfer function
N (s) k
T (s) = = n . (2.17)
D(s) s + bn−1 sn−1 + · · · + b1 s + b0

This is an all-pole response. The characteristic polynomial of the Butterworth

filter is
FILTERS 19

Figure 2-6: Maximally

flat, or Butterworth, low-
pass filter approximation
for various orders, n, of
the filter.

n Factors of Bn (s) Table 2-1: Factors

1 (s + 1) of the Bn (s)
2 (s2 + 1.4142s + 1) polynomial.
3 (s + 1)(s2 + s + 1)
4 (s2 + 0.7654s + 1)(s2 + 1.8478s + 1)
5 (s + 1)(s2 + 0.6180s + 1)(s2 + 1.6180s + 1)
6 (s2 + 0.5176s + 1)(s2 + 1.4142s + 1)(s2 + 1.9319s + 1)
7 (s + 1)(s2 + 0.4450s + 1)(s2 + 1.2470s + 1)(s2 + 1.8019s + 1)
8 (s2 + 0.3902 + 1)(s2 + 1.1111s + 1)(s2 + 1.6629s + 1)(s2 + 1.9616s + 1)

|K(s)|2 = �s2n � = ω 2n ,
� �
(2.18)

since s = ω and where n is the order of the function. Thus the transmission
coefficient is
1 1 1
|T (s)|2 = 2 = 2n
= , (2.19)
1 + |K (s)| 1 + |s | 1 + ω 2n

which is often written as

2 1 1
|T (s)| = 2 = . (2.20)
1 + |K (s)| Bn (s)Bn (−s)

In the filter community, Bn (s) is called the Butterworth polynomial and has
the general form
 � � � �
�n/2 2 2k + n − 1
s − 2s cos π + 1 for n even


 k=1 2n


Bn (s) = � � � � .

 �(n−1)/2 2 2k + n − 1
 (s + 1) k=1 s − 2s cos π +1 for n odd


2n
(2.21)

Bn (s) is given in factorized form in Table 2-1. Further factorization of the

second-order factors results in complex conjugate roots, so they are generally
left in the form shown.
20 STEER MICROWAVE AND RF DESIGN: MODULES

Insight is gained by examining the characteristic polynomial for a real

radian frequency, ω (where s = ω), at the following frequency points:
2 1
at ω = 0 → |T (0)| = =1 (2.22)
1+0
1 1
at ω = 1 → |T (1)|2 = = . (2.23)
1+1 2
Thus the transmission response of the filter is at half power at the corner
frequency ω = 1. Another observation is that the transmission coefficient
at ω = 0 and ω = 1 are independent of the degree of the characteristic
polynomial. This is shown in Figure 2-6, where the responses of Butterworth
filters are plotted for three different orders.

2.4.2 Construction of the Transfer Function

Often the characteristic function, K, is specified in terms of the variable ω,
and for synthesis purposes the response must be transformed back to the s
domain. As an example, consider the maximally flat function of third order
with a characteristic polynomial equal to ω 3 (from Equation (2.19)):
2 1
K(ω) = ω 3 and |T (ω)| = . (2.24)
1 + ω6
(Note that here the first six derivatives of |K(s)|2 with respect to s are zero at
s = 0.) So, from Equation (2.7),

2 1 ω6
|Γ1 (ω)| = 1 − = . (2.25)
1 + ω6 1 + ω6
At real frequencies (i.e., frequencies lying on the imaginary axis in the s
plane)
s = ω, so ω = s/ = −s. (2.26)
Thus, in the s domain, the reflection coefficient becomes
6
2 2 2 (−s)
|Γ1 (ω)| = |Γ1 ( (−s))| = |Γ1 (s)| = 6. (2.27)
1 + (−s)
2 −s6
That is, |Γ1 (s)| = = Γ1 (s) Γ1 (−s) . (2.28)
1 − s6
Factoring the denominator polynomial yields
1 − s6 = (1 − s) (1 + s) s2 + s + 1 s2 − s + 1 .
  
(2.29)
By choosing those factors with roots only in the left-half plane (required for
a realizable network), the zeros of the denominator are obtained. Now the
2
numerator of |Γ1 (s)| is easily factored, and since this is third order, it only
has three reflection zeros at DC. The resulting function is
s3
Γ1 (s) = , (2.30)
(s + 1) (s2 + s + 1)
1
and so T (s) = . (2.31)
(s + 1)(s2 + s + 1)
This example illustrates how Γ1 (s) and T (s) can be obtained from K(s). This
procedure is generalized in the next section.
FILTERS 21

2.4.3 nth-Order Reflection Approximation

Following the procedure outlined in the previous section, generalization
leads to the nth-order Butterworth response:
2
|Γ1 (s)| = Γ1 (s) · Γ1 (−s)
(−s2 )n (−s2 )n
= = n . (2.32)
1 + (−s2 )n  2n

(s − si ) . (s − sj )
i=1 j=n+1

2
So |Γ1 (s)| has n roots (these are the si s) lying in the left-half s plane, and n
roots (the sj s) lying in the right-half √
s plane. Note that j is used as an index
and  (, without the dot) represents −1. It is reasonable to group all of the
left-half plane roots together (a circuit could not be synthesized if there were
a right-half plane root) so that
n
(−s)
Γ1 (s) = 
n . (2.33)
(s − si )
i=1

Solving for the

 roots
n of the denominator of Equation (2.32) (i.e., finding the
roots of 1 + −s2 = 0) yields the following roots in the left-half s plane:
 π
si = exp  (2i − 1 + n) i = 1, 2, ..., n, (2.34)
2n
where exp (θ) = cos (θ) +  sin (θ).

EXAMPLE 2.1 Reflection Coefficient Derivation for Butterworth Filter

Develop the reflection coefficient from the roots of the third-order (n = 3) Butterworth
lowpass filter prototype.
Solution: From Equation (2.34) the three roots are
   
π 2
s1 = exp  (2 × 1 − 1 + 3) = exp  π (2.35)
2×3 3
 
π
s2 = exp  (2 × 2 − 1 + 3) = exp { π} (2.36)
2×3
   
π 4
s3 = exp  (2 × 3 − 1 + 3) = exp  π (2.37)
2×3 3

and the input reflection coefficient of the filter normalized to 1 Ω is

s3 s3
Γ1 (s) =  √ √  = . (2.38)
(s + 1) (s2 + s + 1)
 
1 3 1 3
s+ 2
− 2
(s + 1) s + 2
+ 2

This is the same as Equation (2.30).

2.4.4 Bandwidth Consideration

At the corner frequency of the Butterworth response, that is, at 1 rad/s, the
transmission response is 3 dB down from its maximum transfer response.
22 STEER MICROWAVE AND RF DESIGN: MODULES

Table 2-2: Radian frequen- Order, n

cies at which the response of n=3 4 5 6 7 8 9 10
a Butterworth filter is 1 dB 0.798 rad/s 0.844 0.873 0.901 0.908 0.919 0.928 0.935
down for a corner frequency
f0 = 0.1592 Hz (ω0 =
1 rad/s). At f0 the transmis-
sion response is down 3 dB.

(a) Seventh-order filter, n = 7 (b) Sixth-order filter, n = 6

Figure 2-7: Chebyshev lowpass filter responses. In decibels, the ripple is RdB = −10 log[1/(1 +
ε2 )] = 10 log(1 + ε2 ), ε is called the ripple factor.

Sometimes microwave filters are designed to have 1 dB bandwidths. The

radian frequencies at which the responses of various orders of Butterworth
filters are 1 dB down are given in Table 2-2. By frequency scaling the
Butterworth response the filter can be design for a specified 1 dB bandwidth.

2.5 The Chebyshev Lowpass Approximation

The maximally flat approximation to the ideal lowpass filter response is
best near the origin but not so good near the band edge. Chebyshev filters
have better responses near the band edge, with lower insertion loss near the
edges, but at the cost of ripples in the passband. Example reflection and
transmission responses are shown in Figure 2-7 for a seventh-order and a
sixth-order Chebyshev lowpass filter.

2.5.1 Chebyshev Filter Design

The general form of the Chebyshev transmission coefficient is
1
|T (s)|2 = , (2.39)
1+ ε2 |K(s)|2
where ε is the ripple factor and defines the passband ripple (PBR):

PBR = 1 + ε2 , or in decibels RdB = PBR |dB = 10 log 1 + ε2 .

 
(2.40)
FILTERS 23

The PBR can be seen in the transmission response, |T (s)|2 , in Figure 2-7.
In the passband the peaks of the lossless filter response have |T (s)|2 =  1
and the minimums of the ripple response all have |T (s)|2 = 1/ 1 + ε2 =
1/PBR. Consequently Chebyshev filters are also known as equiripple all-
pole lowpass filters. Also note that the corner radian frequency, ω = 1 for the
lowpass filter prototype,
 has a transmission response (i.e., insertion loss IL)
of |T (s)|2 = 1/ 1 + ε2 , whereas the Butterworth transmission response was


at half power at the corner frequency. For the Chebyshev filter, the insertion
loss at the corner frequency is the ripple:

IL = 1RdB = 10 log(1 + ε2 ). (2.41)

For the nth-order Chebyshev (lowpass filter) approximation, the square of

the characteristic function is
cos2 n cos−1 (ω) ,
  
2 −1 ≤ ω ≤ 1
|Kn (ω)| = 2
 −1
 , (2.42)
cosh n cosh (|ω|) , ω ≤ −1, ω ≥ 1

which can be expressed as a polynomial. For example, with n = 3,

K3 (ω) = 4ω 3 − 3ω, for all ω. (2.43)

(This equivalence was derived by Pafnuty Chebyshev.) It is surprising that

the trigonometric expression has such a simple polynomial equivalence.
From Equation (2.11) the transmission coefficient is (for −1 ≤ ω ≤ 1)
1
|T (ω)|2 = (2.44)
1+ ε2 cos2 [n cos−1 (ω)]
and the reflection coefficient is
ε2 cos2 n cos−1 (ω)
 
2
|Γ1 (ω)| = . (2.45)
1 + ε2 cos2 [n cos−1 (ω)]
Factorizing the denominator of either Equation (2.44) or Equation (2.45)
yields the following roots (of the denominators of Γ1 (s) and T (s)):
    
(2i − 1) π 1 −1 1
si = sin sinh sinh
2n n ε
    
(2i − 1) π 1 −1 1
+  cos cosh sinh i = 1, 2, . . . , n. (2.46)
2n n ε
The roots of the numerator of Γ1 (s) in the s plane are
(2k − 1) π
sk =  cos k = 1, 2, . . . , n. (2.47)
2n
Equations (2.46) and (2.47) can be used to obtain the reflection and
transmission coefficients directly in the s domain.

2.5.2 Chebyshev Approximation and Recursion

The characteristic function of the Chebyshev approximation can be obtained
from the recursion formula,

Kn (ω) = 2ωKn−1 (ω) − Kn−2 (ω) , (2.48)

24 STEER MICROWAVE AND RF DESIGN: MODULES

Table 2-3: Radian frequencies at which the transmission response of an nth Chebyshev filter is
down 1 dB and 3 dB for a corner frequency ω0 = 1 rad/s. (Note that ω0 is the radian frequency
at which the transmission response of a Chebyshev filter is down by the ripple, see Figure 2-7.)
Response 1 dB down Response 3 dB down
Ripple n=3 n=5 n=7 n=9 Ripple n=3 n=5 n=7 n=9
0.01 dB 1.564 1.192 1.097 1.058 0.01 dB 1.877 1.291 1.145 1.087
0.1 dB 1.202 1.071 1.036 1.022 0.1 dB 1.389 1.134 1.068 1.041
0.2 dB 1.127 1.045 1.023 1.014 0.2 dB 1.284 1.099 1.050 1.030
1 dB 1.000 1.000 1.000 1.000 1 dB 1.095 1.0338 1.017 1.010
3 dB – – – – 3 dB 1.000 1.000 1.000 1.000

with K1 (ω) = ω; K2 (ω) = 2ω 2 − 1. (2.49)

For example, with n = 3,

K3 (ω) = 2ωK3−1 (ω) − K3−2 (ω) (2.50)

= 2ω 2ω 2 − 1 − ω = 4ω 3 − 2ω − ω = 4ω 3 − 3ω.
 

2.5.3 Bandwidth Consideration

At the corner frequency of the Chebyshev filter the transmission response is
down by the amount of the ripple. This can be seen in Figure 2-7. However,
the bandwidth of a filter is usually specified in terms of its 1 dB or 3 dB
bandwidth at which the transmission response is down 1 dB or 3 dB,
respectively, from its maximum response. The radian frequencies at which
the responses of various orders of Chebyshev filters are 1 dB down and 3 dB
down are given in Table 2-3. By frequency scaling the Chebyshev response,
the filter can be designed for a specified 1 dB or 3 dB bandwidth.

2.6 Element Extraction

In the previous two sections the mathematical responses of Butterworth and
Chebyshev filters were derived for various orders. In this section it will be
shown how these filters can be implemented with inductors and capacitors
using what is called ladder synthesis [4].

2.6.1 Ladder Synthesis

To obtain the element values yielding the desired transfer function, an
impedance or admittance function must first be obtained. The impedance
or admittance function can be readily obtained from the input reflection
coefficient of a network, but for now the focus is on synthesizing a given
impedance function. A general impedance function can be expressed as

an s2 + ω12 s2 + ω32 s2 + ω52 ...

   
Z(s) = , (2.51)
bm s (s2 + ω22 ) (s2 + ω42 ) (s2 + ω62 ) ...

where an and bm are constants. This can be realized using L and C elements
in a network terminated by a resistor, provided that the degree of the
numerator and denominator differ by no more than unity (i.e., |m−n| ≤ 1). In
FILTERS 25

Figure 2-8: Extraction of a network X to reduce

an impedance Zin,i to a lower-order impedance
Zin,i+1 .

Impedance basis Admittance basis

Zin = ZC + Zrem Yin = YL + Yrem

1 1
ZC = YL =
sC0 sL0
(a) (b)

Zin = ZL + Zrem Yin = YC + Yrem

ZL = sL∞ YC = sC∞

(c) (d)

Zin = ZLC + Zrem Yin = YLC + Yrem

sLi sCi
ZLC = YLC =
s2 Li Ci + 1 s2 Li Ci + 1
(e) (f)

Figure 2-9: Synthesis of impedance and admittance functions. Starting with an impedance
function Z(s): (a) extraction of a series capacitor; (c) extraction of a series inductor; and (e)
extraction of a series parallel LC block i. Starting with an admittance function Y (s): (b)
extraction of a shunt inductor; (d) extraction of a shunt capacitor; and (f) extraction of a shunt
series LC block.

the case of a doubly terminated network, this resistor is the load. The element
extraction procedure, shown in Figure 2-8, involves extracting a network X
from Zin,i , leaving a reduced-order impedance Zin,i+1 .
The extraction of inductors and capacitors is illustrated in Figure 2-9. Thus,
following the extraction of an element or a pair of elements an impedance,
Zrem , or admittance, Yrem , remains that can be similarly simplified. For ex-
ample, and referring to Figure 2-9(a), Z(s) = 1/(sC) + Zrem . So a pole of Z(s)
at DC requires the extraction of a series capacitor
 of value (see Figure 2-9(a))
1 
C0 = , (2.52)
sZ(s)  s=0

while a pole at infinity requires the extraction of a series inductor of value

(see Figure 2-9(c)) 
Z(s) 
L∞ = . (2.53)
s s=∞
Another possibility is a pole at a finite frequency (call this ω0 ), which requires
the extraction of a series parallel LC block, as shown in Figure 2-9(e), with
26 STEER MICROWAVE AND RF DESIGN: MODULES

elements of value

s  1
Ci = 2 2
 and Li = . (2.54)
(s + ω0 ) Z(s) s=ωo ω02 .Ci

The extraction process can also be carried out on an admittance basis. First,

bm s s2 + ω22 s2 + ω42 s2 + ωs2 . . .

   
Y (s) = . (2.55)
an (s2 + ω12 ) (s2 + ω32 ) (s2 + ω52 ) . . .

Now a pole at zero requires the extraction of a shunt inductor of value (see
Figure 2-9(b))

1 
L0 = (2.56)
sY (s) s=0

and a pole at infinity requires the extraction of a shunt capacitor of value (see
Figure 2-9(d))

Y (s) 
C∞ = . (2.57)
s s=∞

A pole at a finite frequency requires the extraction of a shunt-series LC block

(as shown in Figure 2-9(f)) with values

s  1
Li = 2  and Ci = 2 . (2.58)
(s + ωo2 ) Y (s) s=jωo ωo .Li

Many aspects of filter synthesis can seem abstract when presented in full
generality. Consequently it is common to illustrate filter synthesis concepts
using examples. Following this time-honored tradition, an example is now
presented.

EXAMPLE 2.2 Element Extraction for a Third-Order Lowpass Filter

A third-order maximally flat filter has the reflection coefficient

s3
Γ1 (s) = . (2.59)
(s + 1) (s2 + s + 1)

Synthesize this filter as a doubly terminated network.

Solution:
The reflection coefficient function (Equation (2.59)) has all its poles located at infinity, so the
corresponding network realization must be made of simple L or C elements and terminated
in a resistor. Hence, referring to Figure 2-2 and considering a 1 Ω system,

1 + Γ1 (s) 2s3 + 2s2 + 2s + 1

Zin,1 (s) = = . (2.60)
1 − Γ1 (s) 2s2 + 2s + 1

Note that the input impedance approaches infinity as the frequency goes to infinity, hence a
series inductor must be extracted. The value of this inductor is

Zin,1 (s) 
L∞1 = = 1 H. (2.61)
s 
s=∞
FILTERS 27

The filter is developed by extracting one element at a time. Following the extraction of the
first element, the second-stage impedance is left. Now the impedance function is

2s3 + 2s2 + 2s + 1
Zin,2 (s) = Zin,1 (s) − sL∞1 = − sL∞1
2s2 + 2s + 1
2s3 + 2s2 + 2s + 1 − s 2s2 + 2s + 1
 
s+1
= = 2 .
2s2 + 2s + 1 2s + 2s + 1
Note that the stage impedance above, Zin,2 , approaches zero as the frequency goes to infinity.
There is not a single series element that would cause this. However, the stage admittance
function,

1 2s2 + 2s + 1
Yin,2 (s) = = , (2.62)
Zin,2 (s) s+1

goes to infinity as the frequency approaches infinity and so a shunt capacitor is extracted:
1
Yin,3 (s) = Yin,2 (s) − sC∞2 = , (2.63)
s+1
where

C∞2 = 2 F. (2.64)

So sometimes it is more convenient to consider extraction of an admittance and sometimes it

is better to consider extraction of an impedance.
By examining the remaining stage impedance, it is seen that a pole exists at infinity, and so a
series inductor, L∞3 , is extracted. The value of this inductor comes from
1
Zin,3 = =s+1Ω (2.65)
Yin,3

and so the inductor value is


s + 1 
L∞3 = = 1 H. (2.66)
s  S=∞

The final step is to extract a load of value 1 as follows:

Zin,4 = Zin,3 − sL∞3 = 1 Ω. (2.67)

This example synthesized a doubly terminated network. The resulting network, called a
ladder circuit, is shown in Figure 2-10. The left-most 1 Ω resistor is part of the source.
This circuit has a dual form consisting of two shunt capacitors separated by a series inductor.
The dual circuit derives from realizing the admittance function obtained from the reflection
coefficient. Other network extraction techniques are presented in Scanlan and Levy [4, 5] and
Matthaei et al. [1].

Figure 2-10: Synthesized maximally flat network with a third-

order lowpass reflection response.
28 STEER MICROWAVE AND RF DESIGN: MODULES

2.6.2 Summary
The input impedance function of a lumped-element circuit can always be
expressed as the ratio of two polynomials in s and the order of the numerator
and the denominator polynomials can differ by at most one [5]. If the orders
differ by one, then a single inductor or capacitor can always be extracted,
however, the remaining impedance function may not be realizable. This
indicates that a more complex LC (and possibly R) combination is required.
To be able to systematically extract arbitrarily complex circuits, a long list of
possible functions, such as those shown in Figure 2-9, is required. For most of
the circuits of interest the LC combinations shown in Figure 2-9 are sufficient.
The next example describes impedance function extraction that requires an
LC combination.

EXAMPLE 2.3 Element Extraction of an Impedance Function

4s3 + 4s2 + 2s + 2
Realize the impedance function Zw = .
4s2 + 2s + 1
Solution:
The order of the numerator is 1 greater than the order of the denominator and this indicates
that a series inductor is perhaps present. The series inductance is

Zw (s) 
L1 = = 1 H. (2.68)
(s) z=∞

The remaining impedance is

4s3 + 4s2 + 2s + 2 2s2 + s + 2
Zin,2 = Zw − sL1 = 2
−s= 2 . (2.69)
4s + 2s + 1 4s + 2s + 1
The numerator and denominator of Zin,2 have the same order. Therefore a simple L or C
element cannot be used to reduce the complexity of the impedance function. Thus an initial
series inductor was not the right choice and the extraction must backtrack.
Figure 2-9 shows several element combinations that can be used to reduce the complexity
of an impedance function. Insight into which alternative to choose comes from factoring zw ,
and note that real roots are required, thus

4s3 + 4s2 + 2s + 2 (2s2 + 1)(2s + 2)

Zw = 2
= . (2.70)
4s + 2s + 1 4s2 + 2s + 1
Examination of Figure 2-9 reveals that a ready fit to Zw is not found. Instead consider the
admittance function
1 4s2 + 2s + 1
Yw = = . (2.71)
Zw (2s2 + 1)(2s + 2)

So the reduction shown in Figure 2-9(f) looks like the right candidate. The general choice for
the element is
as
yx = 2 . (2.72)
bs + 1
Choosing b = 2 now reduces complexity (since part of the factored denominator of Yw now
occurs), so
4s2 + 2s + 1
 
as as
Yw = 2 + − (2.73)
2s + 1 (2s2 + 1)(2s + 2) 2s2 + 1
(4 − 2a)s2 + (2 − 2a)s + 1
 
as
= 2 + . (2.74)
2s + 1 (2s2 + 1)(2s + 2)
FILTERS 29

s 2s2 + 1 s 1
Choose a = 1, Yw = + = 2 + (2.75)
(2s2 + 1) (2s2 + 1)(2s + 2) 2s + 1 2s + 2
s
= 2 + Yin,2 . (2.76)
2s + 1
So C1 L1 = b = 1, C1 = a = 1 F, L1 = 2 H, and
1 1
Yin,2 = or Zin,2 = = 2s + 2. (2.77)
(2s + 2) Yin,2

The final network is

2.7 Butterworth and Chebyshev Filters

The nth-order lowpass filters constructed from the Butterworth and
Chebyshev polynomials have the ladder circuit forms of Figure 2-11(a or b).
Figure 2-11 uses several shorthand notations commonly used with filters.
First, note that there are two prototype forms designated Type 1 and Type
2, and these are referred to as duals of each other. The two prototype forms
have identical responses with the same numerical element values g1 , . . ., gn .
Consider the Type 1 prototype of Figure 2-11(a). The right-most element is
the resistive load, which is also known as the (n + 1)th element. The next
element to the left of this is either a shunt capacitor (of value gn ) if n is even,
or a series inductor (of value gn ) if n is odd. So for the Type 1 prototype,
the shunt capacitor next to the load does not exist if n is odd. The same
interpretation applies to the circuit in Figure 2-11(b).

(a) Type 1

(b) Type 2

Figure 2-11: Filter prototypes in the Cauer topology. Here n is the order of
the filter.
30 STEER MICROWAVE AND RF DESIGN: MODULES

Table 2-4: Order, n 2 3 4 5 6 7 8 9

Coefficients of g1 1.4142 1 0.7654 0.6180 0.5176 0.4450 0.3902 0.3473
the Butterworth g2 1.4142 2 1.8478 1.6180 1.4142 1.2470 1.1111 1
lowpass prototype g3 1 1 1.8478 2 1.9318 1.8019 1.6629 1.5321
filter normalized to g4 1 0.7654 1.6180 1.9318 2 1.9615 1.8794
a radian corner fre- g5 1 0.6180 1.4142 1.8019 1.9615 2
quency of 1 rad/s g6 1 0.5176 1.2470 1.6629 1.8794
and a 1 Ω system g7 1 0.4450 1.1111 1.5321
impedance (i.e., g8 1 0.3902 1
g9 1 0.3473
g0 = 1 = gn+1 ).
g10 1

2.7.1 Butterworth Filter

A generalization of the example of the previous section leads to a formula for
the element values of a ladder circuit implementing a Butterworth lowpass
filter. For a maximally flat or Butterworth response the element values of the
circuit in Figure 2-11(a and b) are
 π 
gr = 2 sin (2r − 1) r = 1, 2, 3, . . . , n (2.78)
2n
and g0 = 1 = gn+1 . Table 2-4 lists the coefficients of Butterworth lowpass
prototype filters up to ninth order.

EXAMPLE 2.4 Fourth-Order Butterworth Lowpass Filter

Derive the fourth-order Butterworth lowpass prototype of Type 1.

Solution:
From Equation (2.78),

g1 = 2 sin [π/(2 · 4)] = 0.765369 H (2.79)

g2 = 2 sin [3π/(2 · 4)] = 1.847759 F (2.80)
g3 = 2 sin [5π/(2 · 4)] = 1.847759 H (2.81)
g4 = 2 sin [7π/(2 · 4)] = 0.765369 F. (2.82)

Thus the fourth-order Butterworth lowpass prototype circuit with a corner frequency of
1 rad/s is as shown in Figure 2-12.

Figure 2-12: Fourth- L1 = 0.765369 H C2 = 1.847759 F

order Butterworth low- L3 = 1.847759 H C4 = 0.765369 F
pass filter prototype.
FILTERS 31

2.7.2 Chebyshev Filter

For a Chebyshev response, the element values of the lowpass prototype
shown in Figure 2-11 are found from the recursive formula [1, 6, 7]:
2a1
g0 = 1 g1 = (2.83)
γ
 
1, n odd
gn+1 = (2.84)
tanh2 (β/4), n even
4ak−1 ak
gk = , k = 2, 3, . . . , n (2.85)
bk−1 gk−1
 
(2k − 1)π
ak = sin , k = 1, 2, . . . , n , (2.86)
2n
 
β
where γ = sinh , (2.87)
2n
 
kπ
bk = γ 2 + sin2 k = 1, 2, . . . , n (2.88)
n
     
RdB RdB
β = ln coth = ln coth (2.89)
2 · 20 log(e) 17.3717793
RdB = 10 log 1 + ε2 ,
 
(2.90)
n is the order of the filter, and ε is the ripple factor and defines the level of the
ripple in absolute terms. RdB is the ripple expressed in decibels (the ripple is
generally specified in decibels).
An interesting point to note here is that the source resistor, the value of
which is given by g0 , and terminating resistor, the value of which is given
by gn+1 , are only equal for odd-order filters. For an even-order Chebyshev
filter the terminating resistor, gn+1 , will be different and a function of the
filter ripple. Because it is generally desirable to have identical source and
load impedances, Chebyshev filters are nearly always restricted to odd order.
Thus the odd-order Chebyshev prototypes are as shown in Figure 2-13.
Also, for an odd-degree function (n is odd) there is a perfect match at DC,
|T (0)|2 = 1, (2.91)

(a) Type 1

Figure 2-13: Odd-order Chebyshev low-

pass filter prototypes in the Cauer topol-
(b) Type 2 ogy. Here n is the order of the filter.
32 STEER MICROWAVE AND RF DESIGN: MODULES

Table 2-5: Coefficients of a Order n=3

Chebyshev lowpass proto- Ripple 0.01 dB 0.1 dB 0.2 dB 1.0 dB 3.0 dB ε = 0.1
type filter normalized to a g1 0.62918 1.03156 1.22754 2.02359 3.34874 0.85158
radian corner frequency of g2 0.97028 1.14740 1.15254 0.99410 0.71170 1.10316
ω0 = 1 rad/s and a 1 Ω sys- g3 0.62918 1.03156 1.22754 2.02359 3.34874 0.85158
tem impedance (i.e., g0 =
1 = gn+1 ). The ripple factor, Order n=5
ε, is related to the ripple in Ripple 0.01 dB 0.1 dB 0.2 dB 1.0 dB 3.0 dB ε = 0.1
decibels by Equation (2.90) g1 0.75633 1.14681 1.33944 2.13488 3.48129 0.97140
(e.g., ε = 0.1 is a ripple g2 1.30492 1.37121 1.33702 1.09111 0.76192 1.37208
of 0.0432 dB). (Note that ω0 g3 1.57731 1.97500 2.16605 3.00092 4.53755 1.80136
g4 1.30492 1.37121 1.33702 1.09111 0.76192 1.37208
is the radian frequency at
g5 0.75633 1.14681 1.33944 2.13488 3.48129 0.97140
which the transmission re-
sponse of a Chebyshev fil- Order n=7
ter is down by the ripple, Ripple 0.01 dB 0.1 dB 0.2 dB 1.0 dB 3.0 dB ε = 0.1
see Figure 2-7.)
g1 0.79694 1.18118 1.37226 2.16656 3.51852 1.00794
g2 1.39242 1.42281 1.37820 1.11151 0.77220 1.43678
g3 1.74813 2.09667 2.27566 3.09364 4.63898 1.93981
g4 1.63313 1.57340 1.50016 1.17352 0.80381 1.62196
g5 1.74813 2.09667 2.27566 3.09364 4.63898 1.93981
g6 1.39242 1.42281 1.37820 1.11151 0.77220 1.43678
g7 0.79694 1.18118 1.37226 2.16656 3.51852 1.00794

Order n=9
Ripple 0.01 dB 0.1 dB 0.2 dB 1.0 dB 3.0 dB ε = 0.1
g1 0.81446 1.19567 1.38603 2.17972 3.53394 1.02347
g2 1.42706 1.44260 1.39389 1.11918 0.76604 1.46186
g3 1.80436 2.13455 2.30932 3.12143 4.66906 1.98372
g4 1.71254 1.61672 1.53405 1.18967 0.81181 1.67776
g5 1.90579 2.20537 2.37280 3.17463 4.72701 2.06485
g6 1.71254 1.61672 1.53405 1.18967 0.81181 1.67776
g7 1.80436 2.13455 2.30932 3.12143 4.66906 1.98372
g8 1.42706 1.44260 1.39389 1.11918 0.76604 1.46186
g9 0.81446 1.19567 1.38603 2.17972 3.53394 1.02347

while for an even-degree function (i.e., n is even) a mismatch exists of value

2 4RL 1
|T (0)| = 2 = (2.92)
(RL + 1) 1 + ε2
  2
so that RL = gn+1 = ε + (1 + ε2 ) . (2.93)

Coefficients of several Chebyshev lowpass prototype filters with different

levels of ripple and odd orders up to ninth order are given in Table 2-5.

2.7.3 Summary
A Butterworth filter has a monotonic response without ripple, but a
relatively slow transition from the passband to the stopband. A Chebyshev
filter has a rapid transition but has ripple in either the stopband or passband.
Butterworth and Chebyshev filters are special cases of elliptical filters,
which are also called Cauer filters. In general, an elliptical filter has ripple in
both the stopband and the passband. The level of the ripple can be selected
FILTERS 33

Figure 2-14: Impedance inverter (of impedance K in

ohms): (a) represented as a two-port; and (b) the two-
port terminated in a load.

independently in each band. With zero ripple in the stopband, but ripple in
the passband, an elliptical filter becomes a Type I Chebyshev filter. With
zero ripple in the passband, but ripple in the stopband, an elliptical filter
becomes a Type II Chebyshev filter. With no ripple in either band the
elliptical filter becomes a Butterworth filter. With ripple in both the passband
and stopband, the transition between the passband and stopband can be
made more abrupt or alternatively the tolerance to component variations
increased.
Another type of filter is the Bessel filter which has maximally flat group
delay in the passband, which means that the phase response has maximum
linearity across the passband. The Legendre filter (also known as the
optimum “L” filter) has a high transition rate from passband to stopband
for a given filter order, and also has a monotonic frequency response (i.e.,
without ripple). It is a compromise between the Butterworth filter, with
monotonic frequency response but slower transition and the Chebyshev
filter, which has a faster transition but ripples in the frequency response.
More in-depth discussions of a large class of filters along with coefficient
tables and coefficient formulas are available in Matthaei et al. [1], Hunter [3],
Daniels [8], Lutovac et al. [9], and in most other books dedicated solely to
microwave filters.

2.8 Impedance and Admittance Inverters

Inverters are two-port networks used in many RF and microwave filters.
The input impedance of an inverter terminated in an impedance ZL is
1/ZL . Impedance and admittance inverters are the same network, with the
distinction being whether siemens or ohms are used to define them. An
inverter is sometimes called a unit element (UE). At frequencies of a few
hundred megahertz and below an inverter can be realized using operational
and transconductance amplifiers. At microwave frequencies the simplest
inverter is a one-quarter wavelength long line. In RF and microwave filter
design they are used to convert a series element into a shunt element. It
is much easier to realize shunt elements in distributed circuits than series
elements. Similar circuit transformations enable an inductor to be replaced
by a capacitor.
The schematic representation of an impedance inverter is shown in Figure,
2-14(a). The constitutive property of the inverter is that the input impedance
of the terminated impedance inverter in Figure 2-14(b) is
K2
Zin = . (2.94)
ZL

So the inverter both inverts the load impedance and scales it. Similarly, if
Port 1 is terminated in ZL the input impedance at Port 2 is Zin as defined
above.
An impedance inverter has the value K (in ohms), and sometimes K
is called the characteristic impedance of the inverter. Sometimes K is just
34 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 2-15: Inverter equivalence: (a) two-

port impedance inverter (of impedance K):
(b) a quarter-wave transmission line of
characteristic impedance Z0 = K; and (c)
a terminated one-quarter wavelength long
line. (a) (b) (c)

called the impedance of the inverter. For an admittance inverter J is used

and is called the characteristic admittance of the inverter, and sometimes
just the admittance of the inverter. They are related as J = 1/K. In Section
2.4.6 of [10] it is shown that a λ/4 long line with a load has an input
impedance that is the inverse of the load, normalized by the square of
the characteristic impedance of the line. So an inverter can be realized at
microwave frequencies using a one-quarter wavelength long transmission
line (see Figure 2-15(b)). For the configuration shown in Figure 2-15(c),

K2
Zin = . (2.95)
ZL

2.8.1 Properties of an Impedance Inverter

An impedance inverter has the ABCD matrix
 
0 K
T= , (2.96)
/K 0

where K is called the characteristic impedance of the inverter. With a load

impedance, ZL (at Port 2), the input impedance (at Port 1) is (as expected)

AZL + B K K2
Zin (s) = = = . (2.97)
CZL + D (/K) ZL ZL

Now the ABCD matrix of the transmission line of Figure 2-15(b) is

 
cos θ Z0 sin θ
, (2.98)
(/Z0 ) sin θ cos θ

which is identical to Equation (2.96) when the electrical length is θ = π/2

(i.e., when the line is λ/4 long). The inverter is shown in Figure 2-14(a) as
a two-port and its implementation as a λ/4 long line is shown in Figure 2-
14(c). The bandwidth over which the line realizes an impedance inverter is
limited, however, as it is an ideal inverter only at the frequency at which it is
λ/4 long.

2.8.2 Replacement of a Series Inductor by a Shunt Capacitor

A series inductor can be replaced by a shunt capacitor surrounded by a pair
of inverters followed by a negative unity transformer (i.e., an inverter with
K = 1). This equivalence is shown in Figure 2-16 and this will now be shown
mathematically.
FILTERS 35

(a) (b) (c)

Figure 2-16: Equivalent realizations of a series inductor: (a) as a two-port; (b) its realization using
a capacitor, inverters of characteristic impedance K, and a negative unity transformer; and (c)
an alternative realization. C = L/K 2 .

From Table 2-1 of [11], the ABCD matrix of the series inductor shown in
Figure 2-16(a) (which has an impedance of sL) is
 
1 sL
TL = (2.99)
0 1

and the ABCD matrix of the shunt capacitor (which has an admittance of
sC) is, from Table 2-1 of of [11],
 
1 0
T1 = . (2.100)
sC 1

The ABCD matrix of an inverter with K in ohms (generally the unit is

dropped and ohms is assumed) is
 
0 K
T2 = , (2.101)
/K 0

and finally, the ABCD matrix of a negative unity transformer, n = −1, is,
from Table 2-1 of [11],
 
−1 0
T3 = . (2.102)
0 −1

Then the ABCD matrix of the cascade shown in Figure 2-16(b) is

TC = T2 T1 T2 T3
    
0 K 1 0 0 K −1 0
=
/K 0 sC 1 /K 0 0 −1
   
0 K 1 0 0 −K
=
/K 0 sC 1 −/K 0
  
sCK K 0 −K
=
/K 0 −/K 0
2
 
1 sCK
= . (2.103)
0 1

Thus TC = TL if L = CK 2 (compare Equations (2.99) and (2.103)). Thus a

series inductor can be replaced by a shunt capacitor with an inverter before
and after it and with a negative unity transformer. The unity transformer
36 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 2-17: A series capacitor: (a) as a two-

port; (b) its realization using a shunt in-
ductor, inverters and negative unity trans- (a) (b)
former’ L = CK 2 .

may also be placed at the first port, as in Figure 2-16(c). Thus the two-ports
shown in Figure 2-16 are all electrically identical, with the limitation being
the frequency range over which the inverter can be realized. An interesting
and important observation is that as a result of the characteristic impedance
of the inverter (e.g., 50 Ω), a small shunt capacitor can be used to realize a
large series inductance value.

EXAMPLE 2.5 Inductor Synthesis Using an Inverter

Consider the network of Figure 2-16(c) with inverters having a characteristic impedance of
50 Ω. What value of inductance is realized using a 10 pF capacitor?
Solution:
K = 50, so L = CK 2 = 10−11 · 2500 = 25 nH.

2.8.3 Replacement of a Series Capacitor by a Shunt Inductor

A series capacitor can be replaced by a shunt inductor plus inverters and a
negative transformer (see Figure 2-17). The ABCD parameters of the series
capacitor in Figure 2-17(a) are
 
1 1/sC
T= , (2.104)
0 1

and here it is shown that the cascade in Figure 2-17(b) has the same ABCD
parameters. The cascade in Figure 2-17(b) has the ABCD parameters
    
0 K 1 0 0 K −1 0
T=
/K 0 1/sL 1 /K 0 0 −1
   
0 K 1 0 0 −K
=
/K 0 1/sL 1 −/K 0
  
K/sL K 0 −K
=
/K 0 −/K 0
2
 
1 K /sL
= . (2.105)
0 1

So the series capacitor, C, can be realized using a shunt inductor, L, inverters,

and a negative unity transformer, and C = L/K 2. It is unlikely that this
transformation would be exploited, as much better lower-loss capacitors can
be realized at RF than inductors.

2.8.4 Ladder Prototype with Impedance Inverters

The transformations discussed in the previous two sections can be used
to advantage in simplifying filters. In this section the transformations are
FILTERS 37

(a) Lumped-element prototype without inverters

(b) First stage in transformation using inverters

(c) Final prototype with inverters

Figure 2-18: Ladder prototype filters using impedance inverters: (a) lumped-element prototype;
(b) first stage in transformation using inverters; and (c) final stage.

Figure 2-19: Admittance in-

verter: (a) as a two-port; (b)
realized using lumped ele-
ments with B = −J; and
(c) lumped equivalent circuit
(the element values in (c) are
(a) (b) (c) impedances).

applied to the prototype lowpass ladder filter shown in Figure 2-18(a).

The inductors in the ladder circuit are a particular problem as they have
considerable resistance at microwave frequencies. The series inductors can
be replaced by a circuit with capacitors, inverters, and transformers, as
shown in Figure 2-18(b). This simplifies further to the realization shown in
Figure 2-18(c), as the negative unity transformers only affect the phase of the
transmission coefficient. So a lowpass ladder filter can be realized using just
capacitors and inverters.

2.8.5 Lumped-Element Realization of an Inverter

The admittance inverter is functionally the same as the impedance inverter
(see Figure 2-14(a)) and the schematic is the same (see Figure 2-19(a)). As
will be shown, an inverter can be realized using frequency-invariant lossless
elements (i.e., elements whose reactance or susceptance do not vary with
frequency) using the network of Figure 2-19(b). Recall that J is used to
identify an admittance inverter and K identifies an impedance inverter.
If not specified by the context, the inverter (with value specified by a
number) defaults to being an impedance inverter. Alternatively units can
be used to indicate which type of inverter is being used. The function of
38 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 2-20: Impedance inverter: (a) as a two-port; and

(b) its lumped equivalent circuit (the element values in
(b) are impedances). (a) (b)

the inverter is the same in any case; both can be realized by one-quarter
wavelength long lines, for example. For the remainder of this chapter it will
be more convenient, most of the time, to use the admittance inverter, as many
calculations will be in terms of admittances since most lumped elements in
filters synthesis will be in shunt.
Now it will be shown that the lumped-element network of Figure 2-19(b)
realizes an inverter. To do this the inverter and the lumped-element network
must have the same two-port parameters. First, the ABCD matrix of an
inverter of characteristic admittance J is
 
0 /J
TJ = . (2.106)
J 0
Referring to Table 2-1 of [11], the circuit of Figure 2-19(b) has the ABCD
matrix
   
1 0 1 1/(B) 1 0
T =
−B 1 0 1 −B 1
    
1 1/(B) 1 0 0 −/B
= = , (2.107)
−B 0 −B 1 −B 0
where B is the susceptance of the frequency-invariant elements. Equation
(2.107) is identical to Equation (2.106) if B = −J. More practical equivalents
of the circuit of Figure 2-19(b) can be derived, as shown later.
For completeness, the lumped-element equivalent of the impedance
inverter is shown in Figure 2-20 (derived from Figure 2-19 with J = 1/K).

EXAMPLE 2.6 Lumped Inverter Analysis

Demonstrate that Figure 2-19(b) is a lumped-element admittance inverter.

Solution:
Terminating the network in Figure 2-19(b) results in the network shown in Figure 2-21(a).
This is relabeled in Figure 2-21(b), where the elements are admittances. Then

yin = y3 //(y1 $(y2 + yL )), (2.108)

where // indicates “in parallel with” and $ indicates “in series with.” These are common
shorthand notations in circuit calculations. Continuing on from Equation (2.108),
  −1  −1
1 1 y2 + yL + y1
yin = y3 + + = y3 + . (2.109)
y1 y2 + yL y1 (y2 + yL )

Substituting y1 = B and y2 = y3 = −B, this becomes

B(yL − B) −ByL + ByL + B 2 B2

yin = −B + = = . (2.110)
yL yL yL

Thus the lumped-element circuit of Figure 2-19(b) is a lumped-element admittance inverter

of value B (in siemens).
FILTERS 39

Figure 2-21: Terminated lumped-

(a) (b) element admittance inverter.

Figure 2-22: Narrowband inverter

equivalents at frequency f0 : (a)
impedance inverter with characteristic
impedance K; (b) lumped-element
equivalent network; and (c) inverter
realized by short- and open-circuited
(a) (b) (c) stubs.

2.8.6 Narrowband Realization of an Inverter Using

Transmission Line Stubs
In this section it will be shown that an impedance inverter can be
implemented using short- and open-circuited stubs. The match is good
over a narrow band centered at frequency f0 . An impedance inverter is
shown in Figure 2-22(a) and its equivalent lumped-element network is
shown in Figure 2-22(b). A stub-based implementation is shown in Figure
2-22(c), where there are short- and open-circuited stubs of characteristic
impedance Z0 . The input impedance of the stubs is shown at the inputs of
the stubs. The stubs have an electrical length θ at f0 and the stubs are one-
quarter wavelength long (i.e., resonant) at what is called the commensurate
frequency, fr .
Now it will be shown that the network of Figure 2-22(c) is a good
representation of the inverter at f0 . This is done by matching ABCD
parameters. The ABCD parameter matrix of an inverter is
 
0 K
T = (2.111)
/K 0

and, at frequency f0 , the ABCD parameter matrix of the stub circuit of

Figure 2-22(c) is
   
1 0 1 Z0 tan (θ) 1 0
T =
−1/[Z0 tan (θ)] 1 0 1 −1/[Z0 tan (θ)] 1
  
1 Z0 tan (θ) 1 0
=
−1/[Z0 tan (θ)] 0 −1/[Z0 tan (θ)] 1
 
0 Z0 tan (θ)
= . (2.112)
/[Z0 tan (θ)] 0

Thus, equating Equations (2.111) and (2.112), the stub network is a good
representation of the inverter if

K = Z0 tan (θ) , (2.113)

and so the required characteristic impedance of each stub at frequency f0 is

K K
Z0 = =  . (2.114)
tan (θ) tan π2 ff0r
40 STEER MICROWAVE AND RF DESIGN: MODULES

Special Case, fr = 2f0

In most designs the stub resonant frequency, fr (also called the commen-
surate frequency), is chosen to be twice that of the center frequency of the
design, f0 . So with fr = 2f0 , then at f0

K K
Z0 =  = =K (2.115)
tan π f0 tan π/4
2 2f0

and the input impedance of the stub is K. So the characteristic impedance
of the transmission line stub is Z0 = K.

2.9 Filter Transformations

So far the discussion has centered around lowpass filters. Filter design
technology has developed so that the design of a corresponding lowpass
filter is the essential first step, and this is used as a prototype to derive a
filter with other characteristics. The three most important transformations
are listed here:
1. Impedance scaling: A lowpass filter prototype is referenced to a
standard impedance. Usually 1 Ω is used, so the reference source
and load resistances are also 1 Ω. To reference to a higher or lower
impedance, scaling of the impedance of all elements of the filter is
required.
2. Corner frequency scaling: The corner frequency of a lowpass filter
prototype is normalized to 1 rad/s. To reference to another frequency,
the values of the elements must be altered so that they have the same
impedance at the scaled frequency.
3. Filter type transformation: These transformations enable the circuit of
a lowpass filter to be converted to a circuit with a bandpass, bandstop,
or highpass response. The concept is that the response at DC is to be
replicated at infinite frequency for the highpass (so capacitors become
inductors, etc.); to be replicated at the center of the passband for a
bandpass filter (so capacitors become a shunt LC circuit); and to be
inverted at the center of a bandstop filter (so capacitors become a series
LC circuit).
The transformations can be performed in any order. The key point is that
the complexity of the lowpass circuit design is minimal since the number of
elements does not grow until the filter type transformation is performed, so
this step is often done last.

2.9.1 Impedance Transformation

The lowpass prototypes discussed up to now have been referred to a 1 Ω
system. The system impedance can be changed to any level by simply
scaling the impedances of all the circuit elements in a filter by the same
amount, as shown in Figure 2-23. The impedance transformation follows the
same procedure for all filter types (the procedure is the same for lowpass,
highpass, bandpass, and bandstop filters, for example). It should also be
noted that impedance scaling of a transmission line is simply multiplication
of the line’s characteristic impedance by the scale factor.
FILTERS 41

(a) Impedance transformation of an inductor

(b) Impedance transformation of a capacitor

Figure 2-23: Impedance transformations.

The impedances of the elements are
(c) Impedance transformation of an inverter increased by a factor Z0 .

(a) Filter normalized to a reference impedance of 1 Ω

Figure 2-24: Impedance trans-

(b) Filter impedance transformed to a reference impedance of 50 Ω formation of an example filter.

EXAMPLE 2.7 Lowpass Filter Design

Consider the lowpass filter, with an inverter, shown in Figure 2-24(a). This filter is
referenced to 1 Ω, as the source and load impedances are both 1 Ω. Redesign the
filter so that the same frequency response is obtained with 50 Ω source and load
impedances.
Solution:
It is necessary to impedance transform from 1 Ω to 50 Ω. The resulting filter is shown
in Figure 2-24(b). Each element has an impedance (resistance or reactance) that is
50 times larger than it had in the 1 Ω prototype.

2.9.2 Frequency Transformation: Lowpass

Frequency transformations differ depending on the type of the filter. So it
is normal to frequency transform the lowpass prototype before converting
the prototype to another form (such as bandpass). The lowpass prototypes
normally have a band-edge or cutoff frequency at an angular frequency of
unity, that is, 1 rad/s. The band-edge frequency can be transformed from
unity to an arbitrary angular frequency ωc , as shown in Figure 2-25, by
scaling the reactive elements, as shown in Figures 2-26 and 2-27(a), so that
42 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 2-25:
Frequency trans-
formation of a
lowpass filter re-
sponse from (a)
one normalized to
a corner frequency
of 1 rad/s, to (b)
one with a radian
corner frequency of
ωc . (a) (b)

(a) Frequency transformation of an inductor

Figure 2-26: Frequency transformations. The
impedance of the new (scaled) element at the new
(scaled) frequency is the same as it was at the
original frequency. (b) Frequency transformation of a capacitor

they have the same impedance at the transformed frequency as they did at
the original frequency. The inverter is unchanged, as are the source and load
resistances, since these are frequency independent.

2.9.3 Lowpass to Highpass Transformation

The transformation of a lowpass filter prototype to a highpass filter is shown
diagrammatically in Figure 2-28. Mathematically ω in the transfer function
of the lowpass prototype is replaced by −1/ω, that is

Thighpass (ω) = Tlowpass (−1/ω). (2.116)

In terms of a lumped element in the lowpass prototype circuit, if an

element has an impedance ωL and L is frequency independent, then the
corresponding element in the highpass prototype filter has an impedance
1/(ω 2 L). Thus reactive elements are transformed as shown in Figure 2-27(b),
where ω0 is the corner frequency of both the lowpass and highpass prototype
circuits. So inductors are transformed into capacitors and capacitors into
inductors. For example, the odd-order lowpass filter prototypes shown in
Figure 2-13 are transformed into the highpass filters shown in Figure 2-29.

2.9.4 Lowpass to Bandpass Transformation

Understanding the transformation of the lowpass filter into its correspond-
ing bandpass form requires that the lowpass filter be considered with both
its positive and negative frequency responses, as shown in Figure 2-30(a).
This response is shifted in frequency to obtain the bandpass response shown
in Figure 2-30(b). Mathematically the radian frequency, ω, in the response
FILTERS 43

(a) LOWPASS PROTOTYPE TO LOWPASS TRANSFORMATION

Lowpass Lowpass Reactance transformation
prototype element ω0 = corner frequency
C1 = C0 /ω0

L1 = L0 /ω0

(b) LOWPASS PROTOTYPE TO HIGHPASS TRANSFORMATION

Lowpass Highpass Reactance transformation
prototype element ω0 = corner frequency
L1 = 1/(ω0 C0 )

C1 = 1/(ω0 L0 )

(c) LOWPASS PROTOTYPE TO BANDPASS TRANSFORMATION

Lowpass Bandpass Reactance transformation
prototype element ω0 = center of passband

C1 = αC0 /ω0 L1 = 1/(αC0 ω0 )

Figure 2-27: Transformations

L1 = αL0 /ω0 C1 = 1/(αL0 ω0 ) of the elements of a pro-
totype lowpass filter to ob-
tain specific filter types. The
(d) LOWPASS PROTOTYPE TO BANDSTOP TRANSFORMATION
corner frequency of the low-
Lowpass Bandstop Reactance transformation
pass prototype is 1 rad/s. In
prototype element ω0 = center of stopband
the transformations to band-
L1 = α/(C0 ω0 ) C1 = C0 /(ω0 α) pass and√ bandstop filters,
√
ω0 = 1/ L1 C1 = ω1 ω2 ;
ω1 and ω2 are the band-
edge frequencies, and α is the
L1 = L0 /(αω0 ) C1 = α/(L0 ω0 ) transformation constant, α =
ω0 /(ω2 − ω1 ).

Figure 2-28:
Lowpass to highpass
(a) Lowpass response (b) Highpass response transformation.
44 STEER MICROWAVE AND RF DESIGN: MODULES

(a) Type 1 (b) Type 2

Figure 2-29: Odd-order Chebyshev highpass filter prototypes in the Cauer topology. Here n is
the order of the filter.

(a) Two-sided lowpass filter response (b) Bandpass filter approximation

Figure 2-30: Frequency responses in lowpass to bandpass transformation (s = ω).

function is replaced by its bandpass form,

 
ω ω0
ω→ − . (2.117)
ω0 ω
 
ω ω0
That is, Tbandpass (ω) = Tlowpass − , (2.118)
ω0 ω
and so the transfer function of the bandpass filter is derived from the transfer
function of the bandpass filter with ω replaced by (ω/ω0 − ω0 /ω). This
separately maps the −1 and +1 band-edge radian frequencies of the lowpass
response to the bandpass frequencies ω1 and ω2 :
   
ω1 ωo ω2 ωo
−1 → − and + 1 → − . (2.119)
ωo ω1 ωo ω2
Solving the above equations simultaneously yields the center frequency ω0
and the band-edge frequencies ω1 and ω2 with
√
ωo = ω1 ω2 (2.120)

and the so-called transformation constant

ωo
α= . (2.121)
ω2 − ω1
The resulting element conversions are given in Figure 2-27(c).
FILTERS 45

Figure 2-31: Lumped-element

odd-order (nth-order) Chebyshev
bandpass filter prototypes in the
Type II Cauer topology.

Figure 2-32:
Frequency
responses
in lowpass
to bandstop
(a) Two-sided lowpass filter response (b) Bandstop filter approximation transformation.

As an example, a lumped-element Type 2 Cauer bandpass filter is shown

in Figure 2-31. The shunt LC combination and the series LC combinations
are resonators resonant at the center frequency of the filter. Here the
filter is normalized to Z0 source and load impedances. As these are the
same, this filter topology only applies for an odd-order filter. Combining
transformations, the element values of a lumped bandpass filter with center
radian frequency ω0 = 2πf0 and radian bandwidth ωBW = 2π(f2 − f1 ) are as
follows (gr is from the lowpass prototype):

 gr 
ωBW Z0
r = odd 
 2 r = odd
 ωBW Z0  ω0 g r

 
Cr = and Lr = . (2.122)
ω
 2 BW  gr Z0
 
 r = even 

r = even
ω0 g r Z0 ωBW

2.9.5 Lowpass to Bandstop Transformation

Again, consider both the positive and negative frequency responses of the
lowpass filter prototype, as shown in Figure 2-32(a). This response is shifted
in frequency to obtain the bandstop response shown in Figure 2-32(b). Math-
ematically the frequency, ω, in the response function is replaced by its band-
stop form:
� � ��−1
ω ω0
ω→ α − . (2.123)
ω0 ω
� � �−1 �
1 ω ω0
That is, Tbandstop (ω) = Tlowpass − . (2.124)
α ω0 ω

The center frequency (corresponding to DC in the lowpass prototype re-

sponse) is
√
ωo = ω1 ω2 (2.125)
46 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 2-33: Lumped-element

odd-order (nth-order) Chebyshev
bandstop filter prototypes in the
type II Cauer topology.

and the transformation constant is

ωo
α= , (2.126)
ω2 − ω1

where ω1 and ω2 are the band-edge radian frequencies. The resulting element
conversions are given in Figure 2-27(d).
Combining transformations, the element values of a lumped bandstop
filter with center radian frequency ω0 = 2πf0 and radian bandwidth ωBW =
2π(f2 − f1 ) are as follows:
g ω
Z0

r BW

 2 r = odd  r = odd
 ω0 Z0  ωBW gr
 

Cr = and Lr = . (2.127)
 1  g ω Z0
r = even  r BW
 
  r = even
ω02

ωBW gr Z0

A lumped-element type II Cauer bandstop filter is shown in Figure 2-33.

The parallel LC combination and the series LC combinations are resonators
resonant at the center frequency of the filter. The parallel LC resonator
is an open circuit at the center frequency of the stopband and the series
LC resonators are short circuits. The LC resonators are implemented using
resonators, usually transmission line segments and not lumped components.

2.9.6 Transformed Ladder Prototypes

Combining the filter type transformations, and with appropriate use of
inverters, the original lowpass prototype ladder filter and its various filter
type transforms are shown in Figure 2-34.

2.10 Cascaded Line Realization of Filters

In this section, filters are presented that use cascaded sections of transmission
line, realizing series inductors and shunt capacitors. For example, a short
(less than one-quarter wavelength long) length of relatively high-impedance
line behaves predominantly as a series inductance. Also, a very short (much
less than one-quarter wavelength long) length of relatively low-impedance
line acts predominantly as a shunt capacitance. So a Pi network of lumped
elements can be realized with alternate sections of low- and high-impedance
microstrip lines. Such an inductive line is shown in Figure 2-35(a), which has
the two equivalent circuits shown in Figure 2-35(b and c). Basic transmission
line theory gives the input reactance of the line of length, ℓ (with a low-
impedance load),

XL = Z0 sin (2πℓ/λg ) , (2.128)

FILTERS 47

(a) Lowpass prototype ladder filter

(b) Highpass prototype ladder filter

(c) Bandstop prototype ladder filter

(d) Bandpass prototype ladder filter

Figure 2-34: Ladder prototype filters.

(a) (b) (c)

Figure 2-35: Inductive length of line with adjacent capacitive lines: (a) microstrip form; (b)
lumped equivalent circuit; and (c) lumped-distributed equivalent circuit.
48 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 2-36: Capacitive length of

line with adjacent inductive lines:
(a) microstrip; and (b) lumped
equivalent (the left-most and right-
most series inductors come from
the high-impedance lines). (a) (b)

Butterworth Chebyshev
C11 = C31 = 31.8310 pF C11 = C31 = 41.5369 pF
L11 = L31 = 795.775 pH L11 = L31 = 609.824 pH
C21 = 159.155 fF C21 = 188.299 fF
L21 = 159.155 nH L21 = 134.522 nH
ripple, ε = 0.1 = 0.043 dB

Figure 2-37: Lumped-element 3rd-order bandpass filters in a Z0 = 50 Ω system with center

frequency f0 = 1 GHz and a 3 dB bandwidth of 10%.

so that the length of this predominantly inductive line is

 
λg ωL
ℓ= sin−1 . (2.129)
2π Z0

Previously it was shown that a short length of line having a relatively

low characteristic impedance yields a capacitive element and this is shown,
together with its equivalent circuit, in Figure 2-36. The predominating shunt
capacitance is determined by first considering the susceptance
 
1 2πℓ λ
B= sin (2.130) so that ℓ = g sin−1 (ωCZ0 ) . (2.131)
Z0 λg 2π
Thus a cascade of low-impedance and high-impedance lines can realize
(approximately) an LC ladder network.
It is tempting to take filters synthesized in lumped-element form and
realize them in the above manner with transmission lines. The series element
realization concept presented in this section could be extended using shorted
and open stubs to better realize shunt elements. The problem with this
approach is that the resulting filters are narrowband and the response
outside the desired operating range is unpredictable. It is far better to employ
the Richards’s transformation, considered in Section 2.12, which is a much
broader bandwidth technique for realizing filters in distributed form.

2.11 Butterworth and Chebyshev Bandpass Filters

Designs of Butterworth and Chebyshev filters with center frequencies of
1 GHz and 3 dB bandwidths of 10% are shown in Figure 2-37. Their
transmission, S21 , and reflection, S11 , responses are shown in Figure 2-38.
The filter skirts, i.e. the transitions from the passband to the stopbands, is
steeper for the Chebyshev filter than for the Butterworth filter as expected.
The ripple of the Chebyshev filter is very small and not seen in this plot. In a
real filter there will be loss and low-level ripples in the magnitude response
FILTERS 49

Figure 2-38: Insertion loss

(S21 ) and return loss (S11 )
of the Butterworth and
Chebychev lumped-element
bandpass filters in a 50 Ω
system.

Figure 2-39: Phase of the

transmission response (S21 )
of the Butterworth and
Chebyshev lumped-element
filters. The discontinuities
seen in the phase from −180◦
to +180◦ are artifacts and the
phases of the filters reduce
monotonically with increas-
ing frequency. (For example,
the phase at 0.8 GHz is
−110◦ + 360◦ = 250◦.)

disappear. However even with loss the steep skirts of the Chebyshev
transmission response remain. Also, even with loss, the impact of the
Chebyshev ripples is clearly seen in the reflection (S11 ) response. In Figure
2-38 three distinct S11 zeros are seen and these correspond to the three poles
of the Chebyshev filter’s S21 response, but of course we cannot see these.
(In the Laplace transfer function there are 3 complex poles each pair being
transformed from one of the three poles of the lowpass prototype.) The
Butterworth filter also has three (complex) S11 zeros and these are all at the
center frequency of the bandpass filter, 1 GHz.
Another characteristic that differs between the Chebyshev and Butter-
worth responses is seen in their phase responses plotted in Figure 2-39. Each
pole in the S21 characteristic causes a 90◦ phase change. The three complex
poles (i.e. six actual poles) of S21 then result in six 90◦ phase changes in S21
for a total phase change of 450◦ . Small ripples are seen in the Chebyshev
phase responses in the pass band while the phase changes for the Butter-
worth filter are smooth. (The Chebyshev phase ripples remain even with
low-level loss.)
The magnitude and phase responses, Figures 2-38 and 2-39, do not provide
complete visualization of the filter characteristics. Additional insight is
provided in the S11 loci on the Smith chart, see Figure 2-40. Figure 2-40 shows
the S11 characteristics plotted on Smith charts. (In the passband the S21 locus
would be very close to the unit circle and little of value is observed in the
passband.) There is a wealth of information here. First consider the response
for the Chebyshev filter, Figure 2-40(a). As frequency increases from 0.8 GHz,
the locus of S11 is first close to the unit circle and then approaches the origin
50 STEER MICROWAVE AND RF DESIGN: MODULES

(a) Chebyshev (b) Butterworth

Figure 2-40: Smith chart plot of S11 of the Butterworth and Chebychev lumped-element filters.

of the polar plot as the frequency approaches the passband of the filter. A
special characteristic of the Chebyshev response is the looping which here
results in three passes of the locus through the origin. These are the three
zeros. Eventually the locus of S11 increases as frequency increases above the
passband. A loop is also seen in the Butterworth response, Figure 2-40(b).
This loop goes through the origin and while it seems that there is just one
zero there are actually three. What is happening is that as frequency increases
and as the locus of the Butterworth S11 response approaches the origin, the
movement of the locus with respect to frequency slows down. The best way
to be convinced that there are three zeros of S11 is to look at the transmission
phase ( S21 ) response in Figure 2-38.
Comprehensive visualization of the filter response requires the rectangular
plots of the magnitudes of S21 and S11 , Figure 2-38, of S21 phase, Figure 2-
39, and the Smith chart plot of S11 , Figure 2-40. The phase plot convinces
you of the number of zeros in your design which is important in interpreting
the Butterworth results. A physical implementation of the design will not be
exact and so tuning is required. Then the most important characterizations
are the rectangular magnitude and Smith chart responses. Here the loops on
the Smith chart, even if they do not go through the origin exactly, distinguish
the Butterworth and Chebyshev responses. Matching may be required to
shift the loops to the origin of the polar plot.

2.12 Richards’s Transformation

Richards’s transformation is a remarkable scheme that takes into account
the actual properties of transmission lines, yielding broadband transmission
line-based implementations of lumped-element filter prototypes [12–15].
FILTERS 51

2.12.1 Richards’s Transformation and Transmission Lines

Consider a section of transmission line of electrical length θ with ABCD pa-
rameters  
cos (θ) Z0 sin (θ)
T = . (2.132)
/Z0 sin (θ) cos (θ)

If this line is terminated in a load, ZL , then its input impedance is

cos(θ)ZL + Z0 sin(θ)
Zin (θ) = . (2.133)
j/Z0 sin(θ)ZL + cos(θ)
Now examine two extreme conditions. As the load impedance increases,
eventually becoming an open circuit, the input impedance of a line with
electrical length θ is defined in terms of a cotangent of the electrical length:
Z0
ZL → ∞ ⇒ Zin (θ) = cot(θ) (2.134)

Yin (θ) = Y0 tan(θ). (2.135)

As the load impedance reduces to become a short circuit, the input

impedance of a line with electrical length θ is defined in terms of a tangent
of the electrical length:

ZL → 0 ⇒ Zin (θ) = Z0 tan(θ). (2.136)

These results lead to the Richards’s transformation, which replaces the

Laplace variable, s, by Richards’s variable, S, where S = α tan(θ). This
transformation is written
s → S = α tan(θ). (2.137)

For now α and θ are constants that can be chosen as design variables. θ,
of course, is the electrical length of the line. Also, α must have the units of
impedance and it is the characteristic impedance of the transmission line.
Applying the Richards’s transformation to a capacitor, the admittance of
the element is transformed as follows:

y = sC → Y = SC = αC tan(θ), (2.138)

so that the capacitor is transformed into an open-circuited stub with

characteristic admittance

Y0 = αC . (2.139)

If a lumped-element capacitor with admittance y = sC is to be realized using

a transmission line, the admittance Y = SC = αC tan(θ) is instead realized.
There are two parameters to select to realize this admittance. The first, α,
is the characteristic admittance of the transmission line (and for any given
transmission line topology there is a minimum and maximum characteristic
admittance or impedance that can be realized), and θ is the electrical length
of the line.
Applying the transformation to an inductor, the impedance of the element
is transformed as follows:

Z = sL → Z = SL = αL tan(θ), (2.140)

52 STEER MICROWAVE AND RF DESIGN: MODULES

s, Laplace variable S, Richards’s variable

Schematic Equivalence

X = Z0
fr = 2f0

X = Z0
fr = 2f0

X = −Z0
Figure 2-41:
fr = 2f0
Equivalences result-
ing from Richards’s
transformation. With
fr = 2f0 the trans-
mission line stubs are X = −Z0
one-eighth wavelength fr = 2f0
long at f0 .

so that the inductor is transformed into a short-circuited stub with

characteristic impedance

Z0 = αL. (2.141)

Thus the Richards’s transform converts an inductor into a short-circuited

stub and a capacitor into an open-circuited stub.

2.12.2 Richards’s Transformation and Stubs

There is a duality between stubs and inductors and capacitors; they are
coupled by Richards’s transformation. One of the important quantities used
in the transformation is the commensurate frequency, fr , which most often
is chosen as twice the operating frequency, f0 . Considering stubs that are
one-quarter wavelength long at fr , the duality is as shown in Figure 2-41.

2.12.3 Richards’s Transformation Applied to a Lowpass Filter

In this section, Richards’s transformation is used to realize a lumped-element
filter in distributed form. The design example begins by considering a
Chebyshev lowpass filter with the transmission characteristic
1
|T (s)|2 = . (2.142)
1+ ε2 |K (s)|2
With the Richards’s transformation (s → ω → α tan (θ)) this becomes
1
|T (α tan θ)|2 = . (2.143)
1+ ε2 |K (α tan θ))|2
Thus the passband edge at ω = 1 is mapped to ω = θ1 as
FILTERS 53

(a)

Figure 2-42: Lowpass prototypes: (a) low-

pass prototype as a ladder filter with invert-
ers; and (b) lowpass distributed prototype
with open-circuited stubs (the impedance
(b) looking into the stub is indicated).

Figure 2-43:
Lowpass to dis-
tributed lowpass
transformation.
(a) Lowpass filter response (b) Distributed lowpass filter response s = ω.

Figure 2-44:
Highpass to dis-
tributed highpass
transformation.
(a) Highpass filter response (b) Distributed highpass filter response s = ω.

1
ω = 1 → α tan (θ1 ) (2.144) so that α= . (2.145)
tan (θ1 )
Recalling that a capacitor is transformed into an open-circuited stub (see
Equation (2.139)), Richards’s transformation applied to the lowpass filter
prototype results in a filter with transmission line elements only, as shown
Figure 2-42, provided that the inverters are realized using transmission lines.

The implementation of a lowpass filter in distributed form results in

passbands and stopbands that repeat in frequency, as shown in Figure 2-43.
This occurs because transmission lines are used to realize lumped elements
and the two-port parameters of a transmission line repeat every wavelength
(or one-half wavelength in some cases). For example, the input impedance of
a stub is the same whether it is one-half wavelength long or one wavelength
long.

2.12.4 Richards’s Transformation Applied to a Highpass Filter

With reference to Figure 2-44(a), the passband edge at ω = 1 is mapped to θ1 ,
as shown in Figure 2-44(b). This means that the passband at ω = 1 is mapped
54 STEER MICROWAVE AND RF DESIGN: MODULES

(a) Lowpass prototype

(b) Highpass prototype response

Figure 2-45: Ladder prototype transforma-

tion. (The input impedances of the stubs are
indicated in (c).) (c) Distributed highpass prototype

to θ1 as (note that θ1 is the electrical length at the band edge)

1
ω = 1 → α tan (θ1 ) (2.146) so that α= . (2.147)
tan (θ1 )
The sequence of steps transforming a lumped lowpass prototype to its
distributed highpass prototype are shown in Figure 2-45. As previously
discussed, all the inverters can be approximated by transmission lines of
length π/2 at the corner frequency of the filter.

2.13 Kuroda’s and Norton’s Network Identities

Kuroda’s and Norton’s network identities are a number of pairs of
equivalent networks that facilitate the transformation from one prototype
to another, particularly transformations that enable realizable transmission
line implementations.

2.13.1 Kuroda’s Identities

Kuroda’s identities embody a number of specific manipulations using
impedance or admittance inverters. They are particularly useful in
implementing Richards’s transformations as they physically separate
transmission line stubs, transform series stubs into shunt stubs, and can
change characteristic impedances that are either too small or too high to
practically realizable values. Kuroda’s identities are a number of equivalent
two-port networks, as shown in Figure 2-46. The proof is derived by
obtaining the ABCD parameters of the two ports similar to the technique
used throughout this chapter. The identities shown in Figure 2-46 are
narrowband. With stubs replacing the lumped elements, Kuroda’s identities
then have broader bandwidths. Kuroda’s identities with stubs are shown in
Figure 2-47. The major use of these identities is to transform designs with
series stubs (in addition to possible shunt stubs) into designs with shunt
FILTERS 55

Figure 2-46: Kuroda’s identities.

Here the inverters are impedance
inverters and the designation
refers to the impedance of the in-
verter. Recall that an inverter of
impedance Z1 can be realized by
a one-quarter wavelength long
transmission line of characteris-
tic impedance Z1 . (As usual, ele-
ment impedances are indicated.)

Figure 2-47: The stub form of

Kuroda’s identities with impedance
inverters. The stub impedances
shown are the input impedances of
the stubs.

stubs only.
To see how these identities are used, consider the identity shown in Figure
2-46(a). The network on the left has a series inductor that, using transmission
lines, is realized by a series stub. A series stub cannot be realized in
most transmission line technologies, including microstrip. Using the identity
shown on the right in Figure 2-46(a), the series stub is replaced by the shunt
stub used to realize the shunt capacitor. At the same time, impedance scaling
56 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 2-48:
Transformation of
a series inductor into
a shunt capacitor
between inverters.

can be used. If the impedance of the inverter in the network on the left is too
low, then it can be scaled by a factor n2 , where n2 = 1 + Z2 /Z1 .
One use of Kuroda’s transforms is to convert a series inductor into a
shunt capacitor. Consider the inductor transformation shown in Figure 2-48.
Figure 2-48(a) shows that two inverters in cascade is electrically equivalent
to an inverting transformer. Since each inverter is a unity inverter (either
a 1 Ω impedance inverter or a 1 S admittance inverter), the equivalence is
a unitary inverting transformer that corresponds to a 180◦ phase change.
So the transformation of a series inductor begins with the transformation
of the inductor (with value x) in Figure 2-48(b) into the network of Figure
2-48(c) in which the inverter cascade and the inverting transformer cancel
each other out. Using the Kuroda identity in Figure 2-47(a), the network
in Figure 2-48(c) converts to the electrically identical network in Figure 2-
48(d). Since the 180◦ rotation usually does not matter in circuits, the final
transformation shown in Figure 2-48(e) is usually acceptable. (The rotation
would only matter if there was another path between the input and output
as then phasing would affect the way signals on multiple paths combined.)
Now the capacitor in Figure 2-48(e) has the numerical value x, the same as
the numerical value of the original inductor. This is a result of using unitary
inverters. The key result here is that a series inductor is equivalent to a shunt
capacitor flanked by two inverters.

2.13.2 Norton’s Identities

Norton’s transformations enable the magnitude of lumped element values
to be scaled [16, 17]. Norton’s transformations are shown in Figure 2-49
where K is the scaling factor. Additional elements are introduced in these
transformations including possibly negatively valued elements. Generally
these elements can be combined with other elements so that the elements
to be realized are positive (e.g. positive capacitor) and the transformer can
also be replaced through subsequent transformations using, for example,
Kuroda’s identities.
FILTERS 57

(a) Type 1 Norton’s transformation

(b) Type 2 Norton’s transformation Figure 2-49: Norton’s transformations. .

Figure 2-50:
(a) Bandpass resonator (b) Coupled bandpass resonators Bandpass resonators.

Figure 2-51: The electrical design of a cross-coupled

filter. There are four resonators labeled 1, 2, 3, and
4. Thebandpass resonators are coupled by admittance
inverters.

2.14 Inter-resonator Coupled Bandpass Filters

The synthesis method that has been described so far leads to bandpass filters
that are a cascade of coupled bandpass resonators. A bandpass filter then
comprises pairs of resonators of the type shown in Figure 2-50(a) that are
coupled using what is called inter-resonator coupling (see Figure 2-50(b)). A
generalization of this architecture is that a bandpass filter comprises coupled
resonators that do not need to be in a linear cascade. Coupling the output of
one resonator latter in a cascade to the input of an earlier resonator is called
cross coupling, and a filter that uses this arrangement is said to be a cross-
coupled resonator filter or a cross-coupled filter. By reusing resonators in this
way, fewer resonators are required to achieve a specified bandpass response
and the filter can have higher performance or be smaller. An example of the
electrical design of a cross-coupled filter is shown in Figure 2-51.
Design of a cross-coupled bandpass filter is not as systematic as the
design of a ladder filter. The design concept is based on using a number
of resonators where each resonator on its own has the same resonant
58 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 2-52: Lowpass filter approximation with capacitors separated

by an admittance inverter.

Figure 2-53: Segment of a filter with two resonators. C1 =

C2 = 2.7171 pF, Z01 = Z02 = 67.7535 Ω, and Z012 =
442.3836 Ω. The stubs are resonant at fr = 2f0 where the filter
center frequency f0 = 1 GHz. Thus the stubs are λ/8 long at
f0 .

frequency f0 , where f0 is the center frequency of the filter. The desired

filter characteristics are obtained by adjusting the inter-resonator coupling,
that is, the coupling between pairs of resonators. Design is not a synthesis
process based on a series of mathematical transformations, but instead is
based on the understanding of the impact of changing the coupling. The
design process will now be described for a pair of resonators.

Coupling of a Pair of Resonators

A segment of a bandpass filter fundamentally comprises a pair of coupled
bandpass resonators. This is based on the lowpass prototype with a pair
of capacitors separated by an admittance inverter (see Figure 2-52). In
transforming this to a bandpass filter with a center frequency of f0 , each
capacitor is replaced by a resonator such as that shown in Figure 2-50(a).
This leads to a pair of resonators, each resonant at f0 , that are coupled by an
inverter (see Figure 2-50(b)). The value of the inverter determines the level
of coupling of the resonators.
Consider the circuit in Figure 2-53, which consists of a pair of coupled
resonators, with each resonator resonant at f0 = 1 GHz, and the resonators
are coupled by the series stub with characteristic impedance Z012 (i.e., the
Z012 stub). The circuit of Figure 2-53 is a segment of a Chebyshev 1 GHz
bandpass filter with a fractional bandwidth of 10%. This is an interim stage
of a bandpass filter design3 and the system impedance of this circuit segment
is 139.4 Ω. The magnitude and phase of the 139.4-Ω S21 and S11 parameters
of this circuit are shown in Figure 2-54(a), and the phase of S21 is shown
in Figure 2-54(b). Specific features to identify here are the two peaks in
the phase response at f1 and f2 . These are also the frequencies of peak
transmission response where |S21 | = 1. The bandwidth defined by f1 and
f2 is called the coupling bandwidth (= f2 − f1 ) or the bandwidth of the
filter (note that this is not the 3-dB bandwidth of the filter). The level of
coupling determines the coupling bandwidth. That is, changing the level of
coupling provided by the Z012 stub changes the position of f1 and f2 . An
interpretation of what is happening is that by coupling the two resonators,
the resonate frequency of the resonators on their own, f0 , has been split by
the coupling, resulting in two peaks at f1 and f2 . The higher the coupling, the
larger the coupling bandwidth, and the lower the coupling, the smaller the
coupling bandwidth. For example, a narrowband bandpass filter comprises

3 Specifically these resonators are the left-most resonators in a design developed in Chapter 3
(see Figure 3-31).
FILTERS 59

(a) Magnitude of transmission and reflection responses

Figure 2-54: Response

of the coupled
resonator network in
(b) Phase of transmission response Figure 2-53.

resonators that have low-level coupling.

So following the choice of an appropriate architecture, that is, the
number of resonators and the coupling arrangement, design and manual
optimization becomes one of changing the level of coupling between pairs
of resonators. Design is not as clean as for ladder filters, but the advantages
of a design with fewer resonators is a compelling justification. Invariably,
extensive (perhaps a day’s worth) manual adjustment of the coupling of the
final fabricated filter is required. Overall this is an approach worth adopting
for high-value filters such as those used in a basestation, where the high unit
costs resulting from manual tuning of individual filters can be justified. This
type of filter design is considered further in [3, 18–21].

Relationship of Ladder Synthesis and Inter-resonator Coupling

The relationship of design focused on interresonator coupling and design
based on the ladder synthesis approach can be demonstrated for a pair of
resonators as follows. Consider the lowpass prototype circuit of Figure 2-52.
The input impedance of this circuit is
 2
−1
Zin (s) = sC1 + J12 /(sC2 ) . (2.148)
With s = ω, the poles of Zin occur when
2
J12 − ω 2 C1 C2 = 0, (2.149)
thus the radian frequencies of the poles are
J12
ω = ±√ . (2.150)
C1 C2
60 STEER MICROWAVE AND RF DESIGN: MODULES

Applying the bandpass transformation (so that the capacitors are replaced
by the resonators of Figure 2-50(a)) results in the bandpass circuit of Figure
2-50(b). With this transformation the frequency is transformed as
 
ω ω0
ω→α − . (2.151)
ω0 ω
That is, the transmission coefficient of the bandpass filter is
  
ω ω0
TBPF (ω) = TLPF α − , (2.152)
ω0 ω
where TLPF (ω) is the transmission coefficient of the lowpass filter at the
radian frequency ω. The poles of the bandpass filter are at ω1 , where
 
ω1 ω0 J12
α − = +√ , (2.153)
ω0 ω1 C1 C2
 
ω2 ω0 J12
and ω2 , where α − = −√ . (2.154)
ωo ω2 C1 C2
Then the coupling bandwidth of the filter is
ω1 − ω2 J12 ω0
f1 − f2 = = √ . (2.155)
2π 2πα C1 C2
Thus the coupling bandwidth can be directly related to the steps in the
ladder-based synthesis of a bandpass ladder filter.

Group Delay
Group delay of a network, and in particular of a filter, is the delay to send
information through a network. It is the time required for a modulated
carrier signal to appear at the output of a filter after being applied to the
input of the filter. Phase delay is a similar measure of delay but does not
describe the time it takes to send information. It describes the delay to send
a particular phase of a single sinewave. Group delay is a steady-sate concept
and so only approximately captures the transient response of a filter.
If a two-port has the transmission coefficient
T (s) = S21 = a + b = t� ϕ, (2.156)
where a and b are its real and imaginary parts, t is its magnitude, and ϕ is its
phase. With s = ω, the phase of T is
 
b
ϕ (ω) = tan−1 . (2.157)
a
The group delay, τD , is the negative of the derivative of this phase as follows:
  
dϕ d b
Group delay = τg (ω) = − =− tan−1 . (2.158)
dω dω a
This compares to the phase delay:
 
ϕ b
Phase delay = τϕ (ω) = − = −tan−1 . (2.159)
ω a
FILTERS 61

Figure 2-55: Group

delay of a resonator.

Figure 2-56: Group delay of res-

onators in a filter. Additional delay is
(a) Single resonator (b) Coupled resonator pair introduced by the coupling inverter.

Figure 2-57: Microstrip layout of a third-order combline filter based

on coupled lines. Each of the vertical microstrip line forms a
resonator with the top gap capacitors. The input and output gap
capacitors provide impedance matching.

Figure 2-55(c) shows the group delay response of one of the resonators in
Figure 2-53, where the group delay peaks at the resonant frequency of the
resonator f0 . The peak of the group delay occurs at the resonant frequency
of the resonator without loading. From examination of the phase of S21 (see
Figure 2-54(b)), it is clear that the group delay peaks very close to f1 and f2 .
So the coupled resonator pair has two peaks in the group delay. This can be
seen in Figure 2-56, which plots the group delay of a single resonator and a
coupled pair of resonators.

2.15 Bandpass Filter Topologies

The essential structure of a bandpass filter comprises resonators that are
coupled to each other. A large number of filter architectures based on this
concept have been deployed. A parallel coupled line (PCL) filter called
a combline filter is shown in Figure 2-57. Here the microstrip lines form
the resonator and the coupling of parallel lines provides the interresonator
coupling. A variation on a filter with PCL resonators is shown in Figure 2-
58 [22]. This filter uses edge-coupled microstrip resonators that are λ/2 long
(at the center frequency). Figure 2-59 shows another distributed bandpass
filter topology utilizing end-coupled microstrip resonators. That is, the gaps
provide the interresonator coupling. All transmission line bandpass filters
have spurious passbands [23–25]. The root cause of the spurious responses
derives from the transformation of the parallel LC resonators into their
62 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 2-58: Microstrip layout of a

parallel coupled bandpass filter.

Figure 2-59: General microstrip layout

for an end-coupled bandpass filter (se-
ries coupling gaps between cascaded
straight resonator elements).

Figure 2-60: A microstrip

coupled dielectric res-
onator bandpass filter
configuration.

Figure 2-61: Bandpass filter

approximation. C1 = 0.85239 F = C3 C2 = 1.10422 F

transmission line form.

A dielectric-resonator-based bandpass filter is shown in Figure 2-60 [26].
Here the pucks are the bandpass resonators and they are coupled and the
evanescent fields outside the pucks provide the interresonator coupling. The
pucks shown are cylinders of high-permittivity material (typically having a
relative permittivity of 500–85 000) with an approximate magnetic wall at the
cylindrical surface of the puck. Thus the puck resonates when its diameter is
approximately λ/2.4
With all these filters the interresonator coupling functions as an inverter.
Thus the basic functional unit of the bandpass filters is the coupled resonator
structure shown in Figure 2-50(b).

2.16 Case Study: Design of a Bandstop Filter

The design of a bandstop filter begins with the lowpass filter prototype
shown in Figure 2-61. To the lowpass prototype, the highpass transformation
is applied to obtain the highpass prototype of Figure 2-62. Picking a center
frequency of approximately 1 GHz and corner frequencies of f1 = 950 MHz
and f2 = 1050 MHz, corresponding to a bandwidth of approximately 10%,
the bandstop transformation is now applied. This results in the prototype
of Figure 2-63. Finally, scaling the system impedance to 50 Ω leads to the
prototype of Figure 2-64. The impedance inverters must remain set at 50 Ω
in order to obtain a broad match. In the passband, energy will pass at all

4 A better estimate is developed from the zeros of Bessel functions, as the fields inside the pucks
have a Bessel function dependence (this is the form of the solution of the wave equation in
cylindrical coordinates).
FILTERS 63

Figure 2-62:
Bandstop filter
1 1 1
L1 = L3 = C1 = C3 = 1.1743 H L2 = C2 = 0.9064 H prototype.

Figure 2-63: Bandstop fil-

ter following transforma-
L1 = L3 = 1.868 nH C1 = C3 = 13.587 pF tion from the lowpass
L2 = 1.442 nH C2 = 17.601 pF prototype.

′ ′
L1 = L3 = 1.868 × 50 = 93.448 nH
′ ′
C1 = C3 = 13.587/50 = 0.2717 pF Figure 2-64: Bandstop
′
L2 = 1.442 × 50 = 72.1362 nH filter after impedance
′
C2 = 17.601/50 = 0.352 pF transformation.

Figure 2-65: Transformations of the resonators in the

bandstop filter to obtain realizable values. The series
(a) (b) LC resonator in (a) is transformed to the form in (b).

frequencies.
In Figure 2-64, the inductor values are relatively large and the capacitor
values are relatively small so that it will be difficult to realize the filter in
either lumped or distributed forms. These values must be scaled to obtain
realizable values. One possible transformation is shown in Figure 2-65. To
establish that the left-hand and right-hand networks are equivalent, at least
near one frequency, the impedances and derivatives must be matched. For
the circuit in Figure 2-65(a),

ω 2 LC − 1 dZ1 ω 2 LC + 1
Z1 =  (2.160) and = , (2.161)
ωC dω ω2C
and for the circuit in Figure 2-65(b),

ω 2 L1 C1 − 1 + ω 2 L1 C0
Z2 =  (2.162)
ωC0 (1 − ω 2 L1 C1 )
dZ2 ω 4 L21 C12 − 2ω 2 L1 C1 + ω 4 L21 C0 C1 + ω 2 L1 C0 + 1
and = 2 . (2.163)
dω ω 2 C0 (ω 2 L1 C1 − 1)

Equating the above enables C0 and C1 to be found for a chosen value of L1 .

Thus the series LC resonators in Figure 2-66(a and c) are replaced by the
64 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 2-66: Intermediate

bandstop filter
prototype.

Figure 2-67: Equivalence of shunt

bandpass resonator to shunt short-
circuited stub.

networks in Figure 2-66(b and d). This results in the filter of Figure 2-66(e).
At this stage the bandpass resonators are then equated to short-circuited
stubs by equating the admittance, Y1 , of the lumped circuit in Figure 2-67(a)
with the admittance, Y2 , of the stub in Figure 2-67(b). That is, by equating
 2 
ω CL − 1 1
Y1 =  and Y2 =  . (2.164)
ωL Z0 tan π2 ωωr

The characteristic impedance of the stub, Z0 , is selected so that the frequency,

ωr , is not too far above the upper band-edge frequency of the filter, in this
case 1.05 GHz. Choosing Z0 = 20 Ω results in the stub transformations
shown in Figure 2-68(a–d). The bandstop filter prototype with stubs is shown
in Figure 2-68(e). The final physical layout of the bandstop filter is shown
in Figure 2-69. The response of the final bandstop filter design is shown in
Figure 2-70.

2.17 Active Filters

With active circuits, losses in lumped-element components can be compen-
sated for by the gain of active devices resulting in filters with high Q. Invert-
ers can also be realized efficiently. The net result is that active filters provide
a performance per unit area advantage over passive element realizations. Of
course, active filters are limited to low-level signals (such as in the receive
circuitry, and prior to the power amplifier stages) as the full power of the
FILTERS 65

Figure 2-68: Bandstop filter

prototype using stub ap-
proximations. The stub in
(b) is the transmission line
approximation of the par-
allel resonant circuit in (a).
The stub in (d) approximates
the circuit in (c).

Figure 2-69: Physical layout of a bandstop

filter in microstrip.

Figure 2-70: Response of the band-

stop filter shown in Figure 2-69.

signal must be handled by the active devices in the active filter. With active
devices, additional noise is introduced and so the noise figure of an active
filter must be considered at RF where bandwidths are significant.
At microwave frequencies active filters based on traditional low-frequency
concepts have relatively low Q. Higher Q and narrower-band applications
require active inductors or distributed techniques. The smallest active filters
are the ones that use active inductors and the largest use distributed
transmission line elements.
66 STEER MICROWAVE AND RF DESIGN: MODULES

Sometimes the main role of a filter is to limit the dynamic range of signals
presented to active circuits, thereby limiting distortion. Clearly active filters
are not suitable when large out-of-band signals are a concern.

2.17.1 Radio Frequency Active Filters

There are several problems with lumped-element filters. High-order filters
require many poles and zeros that must be precisely placed, requiring small
tolerances. With active filters it is possible, to some degree, to achieve
higher-order filters with fewer reactive elements and so the tolerancing
problem is considerably reduced. Also, there are a few designs that can be
tuned electronically. Lumped-element bandpass filters and highpass filters
require inductors. Inductors are a major problem, as RF inductors are lossy
and also only relatively small values can be obtained. On-chip inductors
take considerable area and can thereby dominate the cost of RFICs. With
active filters, feedback and capacitors can be used to realize inductor-like
characteristics, defined as a positive imaginary impedance increasing with
frequency at least over a small frequency range. Many times, and especially
at lower radio frequencies, lumped inductors are no longer required. The
size of capacitors in filters is another issue, and using feedback and gain, the
effective value of capacitors can be magnified. All this comes with limits. The
principle limit is that the cutoff frequency of the transistors must be much
higher than the operating frequency of the filters. This is a cost incurred
whenever feedback is used. A rough rule of thumb is that the transistor cutoff
frequency should be about 10 times the operating frequency. So for a 5 GHz
bandpass filter, a 50 GHz transistor process is required. The other significant
cost is that active filters are only suitable for small signals. The efficiency, a
measure of the amount of DC power used in producing the RF output signal,
is then not a consideration. Active filters cannot be used where signal levels
are likely to be large (such as at the output of a power amplifier or at the
receiver input, due to likely out-of-band signals).
The essential aspect of low-RF active filters is the use of feedback, with
reactive feedback elements, to realize a frequency-domain transfer function
with poles and zeros. The aim is not to synthesize effective capacitors and
inductors, but to focus on the overall response. Active filters are generally
developed as stages, with each stage realizing a second- or higher-order
transfer function. Design is simulation driven, as this is the only way to
account for the active device parasitics.
In operational amplifier design, feedback is used to achieve gain stability,
but at the cost of reduced bandwidth. If reactive feedback elements are
used, high-order transfer functions can be obtained with just a few reactive
elements. In Figure 2-71, a twin-T notch network is connected in the
operational amplifier feedback path to obtain a bandpass filter [27]. Away
from the notch frequency the feedback path has a high impedance and
the overall amplifier gain falls off. As the signal frequency approaches the
notch frequency, the feedback path becomes effective and the amplifier gain
increases. The notch frequency, fn , is proportional to the RC product of the
feedback components and the Q is proportional to the amplifier gain. The
governing equations for this filter are

fn = 1/(2πRC) and Q = (1 + G)/4. (2.165)

FILTERS 67

Figure 2-71: Active twin-T bandpass filter:

center frequency f0 = πRC/4 and Q =
(G + 1)/4.

Figure 2-72:
Operational transcon-
ductance
amplifier with
transconductance GM .
(a) Schematic (b) RF biquad bandpass filter In (a) IO = GM VIN .

2.17.2 Biquadratic Filters

Biquadratic filters are an important class of filters and are the basic building
block used in analog and RF active filters. Biquad filters commonly use
operational transconductance amplifiers (OTAs), the schematic of which
is shown in Figure 2-72(a). Here IO = GM VIN , which is also the basic
characteristic of an individual transistor. These circuits have characteristics
typical of operational amplifiers, including gain stability and resistance to
component tolerance variations. OTA-based filters have the advantage of
easy tunability, with pole and zero frequencies electronically adjustable. An
RF bandpass biquad filter is shown in Figure 2-72(b). Stages such as this are
cascaded to realize high-order filters.
Biquadratic filters implement the transfer function
VOUT N (s) a 2 s 2 + a 1 s + a0 a2 (s + z1 )(s + z2 )
H(s) = = = 2 = . (2.166)
VIN D(s) s + b1 s + b0 (s + p1 )(s + p2 )
This is referred to as a biquadratic function and filters that implement this
function are called biquadratic filters or simply biquad filters. There are
several special types of biquad filters.
The form of the biquadratic function for a second-order lowpass filter is
a0 Kωp2
HLP (s) = = . (2.167)
s2 + b 1 s + b 0 s2 + (ωp /Qp )s + ωp2
HLP (s) has a double zero at s = ∞ and thus negligible response at very high
frequencies.
The highpass form of the biquadratic filter is described by
a2 s2 Ks2
HHP (s) = = . (2.168)
s2 + b 1 s + b 0 s2 + (ωp /Qp )s + ωp2
68 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 2-73: Active

coupled distributed
filter. After [30]. (a) Layout (b) Equivalent circuit model

The response of HHP (s) at high frequencies is K and the response at very
low frequencies goes to zero.
The bandpass form of the biquadratic filter is described by

a1 s K(ωp /Qp )s
HBP (s) = = 2 . (2.169)
s2 + b 1 s + b 0 s + (ωp /Qp )s + ωp2

The response of HBP (s) at high and low frequencies goes to zero, and only
at and near the center frequency, ω = ωp , is there a reasonable response.
The band-reject or notch form of the biquadratic filter is described by

a2 s2 + a0 K(s2 + ωz2 )
HBR (s) = = 2 . (2.170)
s2 + b1 s + b0 s + (ωp /Qp )s + ωp2

The response of HBR (s) at high and low frequencies is high, but there is a
double zero at the notch frequency, ω = ωz , where the response is very low.
With all of the biquadratic filters, the sharpness of the response is
determined by Qp . The edge frequency or center frequency is ωp for the
bandpass, lowpass, and highpass filters, and the notch frequency is ωz for
the bandstop filter.

2.17.3 Distributed Active Filters

At high microwave frequencies, good results can be obtained by combining
distributed elements with a gain stage. Commonly the active devices are
used as coupling devices. A negative resistance can be produced using
the active devices that can compensate for the losses associated with the
transmission line elements or lumped passive elements in the filter. The
concept uses a resonant tank circuit formed by a transmission line or
lumped-element resonator with a negative resistance coupled into the tank
circuit by the active devices [28–30]. This concept is illustrated in Figure
2-73. Here a one-half wavelength long resonator forms a resonant circuit
(commonly called a tank circuit in this context). A negative resistance is
coupled into the tank circuit to compensate for losses in the resonator and
the result is effectively a lossless tank circuit.
An active resonator circuit is shown in Figure 2-74. The design of the
lossless resonator is based on the negative resistance obtained when the
capacitor is connected at the source of the MESFET. This element in the
FILTERS 69

Figure 2-74: An active resonator cir-

cuit. After [31].

Figure 2-75: Lumped-

element 5th-order
Chebyshev filter.

Figure 2-76: A linear chirp.

source provides a feedback path between the output of the circuit, the
drain-to-ground voltage, and the input gate-to-source voltage. An increase
in the drain-source current leads to a voltage at the source that changes
the gate-source voltage. This induces a negative resistance that is adjusted
through the feedback capacitance, CF B , to compensate for inductor losses.
An additional inductor, LP , is added at the gate. The inductor resonates with
the series combination of CF B and CGS . CF B can be implemented using a
varactor diode to enable electronic tuning.

2.18 Transient Response of a Bandpass Filter

A bandpass filter is normally characterized by its center frequency and
bandwidth, and sinusoidal signals that are within the bandwidth are
transmitted by the filter and those outside the passband are reflected by
the filter. However, if the signal changes frequency quickly, the transient
response of the filter can be unexpected [32, 33].
Figure 2-75 is a fifth-order filter with each pair of shunt resonators coupled
by a series resonator. In response to an RF pulse (a pulse of sinusoidally
varying RF), the filter passes energy in-band when all of the resonators have
reached steady state. When the RF pulse is removed, the resonators will loose
energy and eventually the RF signals on the resonators disappear. There will
be a finite time to “charge” and “discharge” the filter’s resonators.
A classic test of the RF response over frequency is to use a linear RF chirp,
as shown in Figure 2-76. With a linear chirp, the frequency of the signal
in the pulse changes linearly and smoothly over time from one frequency
to another. Figure 2-77(a) presents the transient response at the output of
70 STEER MICROWAVE AND RF DESIGN: MODULES

(a) Slow chirp response (b) Fast chirp response

Figure 2-77: Output transient response of a 3rd-order Chebyshev filter with a center frequency
of 1 GHz and a 30 MHz bandwidth excited by a −20 dBm linear chirp from 950 MHz to
1050 MHz: (a) chirp rate = 20 MHz/µs, (b) chirp rate = 400 MHz/µs. At the slower chirp rate,
the filter response is approximately a superposition of steady-state responses as the frequency
changes (i.e., the response is quasi-stationary). At the higher chirp rate, the filter response can no
longer be approximated as a sum of steady-state responses (i.e., it is no longer quasi-stationary).
After [34].

a 30 MHz-bandwidth 1 GHz filter excited by a 1 µs-long linear chirp from

950 MHz to 1050 MHz with a chirp rate of 20 MHz/µs. This is a relatively
slow chirp. The chirp starts at 0 µs and stops at 5 µs. Thus the frequency
of the chirp is in-band at 1.75 µs and is out-of-band at 3.25 µs. The chirp is
sufficiently slow that the transient transmission response corresponds to the
frequency response of the filter. However, if the chirp is fast, the resonators
may not reach steady state, and then the transient response of the filter
cannot be simply extrapolated from its frequency-domain response. This is
seen in Figure 2-77(b), which is the response of the filter to a fast chirp with
a chirp rate of 400 MHz/µs and a chirp duration of 250 ns, and beginning
at 950 MHz and ending at 1050 MHz. The RF chirp begins out of band at
0 ns, becomes in-band at 0.0875 µs, and goes out-of-band again at 0.1625 µs
before turning off at 0.25 µs. Interestingly, there is an almost immediate
output response, around 0.03 µs when the RF chirp is out of band. Overall the
chirped-filter response does not follow the frequency response of the filter.
Another view of the same phenomenon is examined for a 7th-order
Chebyshev filter with a center frequency of 900 GHz and 34 MHz bandwidth
from 883 MHz to 917 MHz. The frequency-domain transmission response of
the filter is shown in Figure 2-78. The transmission responses to 0.25 µs-long
RF pulses of different frequencies are shown in Figure 2-79. The frequency of
the RF pulse is shown in each subplot. The top-left plot is the response when
the RF pulse is in-band. Here the RF pulse is applied at 0 µs and removed
at 0.25 µs. An initial delay in the transmission response and what is referred
to as a long-tail response is seen when the RF pulse is removed. The other
plots in Figure 2-79 show the filter response as the RF is backed off from the
passband. So, even when the RF is out of band, there can be an appreciable
output from the filter at the beginning and end of the RF pulse.
FILTERS 71

Figure 2-78: Frequency-domain transmission response

of a 7th-order Chebyshev filter with a center frequency
of 900 GHz and 34 MHz bandwidth from 883 MHz to
917 MHz.

Figure 2-79: Transient

response at the out-
put of the 900 MHz
7th-order Chebyshev
bandpass filter with
a passband from
883 MHz to 917 MHz.
The frequency re-
sponse is shown in
Figure 2-78. After [34].

2.19 Summary
Radio frequency and microwave filter design combines the mathematical
synthesis of a circuit with the required performance and the intuitive
realization that particular structures inherently have a desired frequency
selectivity. The art of filter design is to direct the mathematical synthesis
and companion circuit so that the circuit structures in the synthesized filter
prototype match the functionality of physical microwave structures. A very
important step in this process is the development of circuit equivalent
models of physical structures. Once a microwave filter with appropriate
topology has been designed and the close-to-final physical filter is designed,
the physical layout is optimized in a circuit simulator to account for
72 STEER MICROWAVE AND RF DESIGN: MODULES

parasitics and higher-order EM effects.

A similar design procedure applies to active filter designs where particular
circuit topologies, for example, the biquad network blocks, are identified and
exploited. The synthesis approach provides design insight and exploitation
of all parameters of a transfer function, leading to optimum network
topologies. Synthesis can be time consuming and specialized, but it is the
only way to develop filters with optimum performance. Insight gained
during design identifies yield issues and provides the basis for inventing
new filter topologies.

Further Reading
The design of microwave filters has a rich tradition. Many excellent books
and articles have been written about microwave filter design techniques.
Books and articles with extensive treatments are references [1, 3, 4, 14, 16–19,
35–46]. However the most important systematic approaches to microwave
filter design were presented in this chapter. Many journal and conference
papers present topologies in various technologies that can be used to
implement filters. The best way to locate these papers is to search using the
specifics of the technology of interest. For example “wideband microstrip
bandstop filters on glass substrates” will yield a list of papers on the
topic. There are many small companies that specialize in different types of
filter design. For very high performance filters, for example in basestation
applications, the number of designers and vendors is quite small and filter
designers follow the literature and patents closely. For designers building
microwave systems with small to medium volumes the best approach is to
use microstrip design. For bandpass microstrip design the most common
choice is to use filter design based on coupled microstrip design which is
considered in the next chapter.

2.20 References
[1] G. Matthaei, L. Young, and E. Jones, Mi- Hall, 2001.
crowave Filters, Impedance-Matching Networks [10] M. Steer, Microwave and RF Design, Transmis-
and Coupling Structures. McGraw-Hill, 1965, sion Lines, 3rd ed. North Carolina State Uni-
reprinted in 1980, Artech House. versity, 2019.
[2] D. M. Pozar, Microwave engineering, 4th ed. [11] ——, Microwave and RF Design, Networks,
John Wiley & Sons, 2012. 3rd ed. North Carolina State University,
[3] I. Hunter, Theory and Design of Microwave Fil- 2019.
ters. IEE Press, 2001. [12] P. Richards, “Resistor-transmission-line cir-
[4] J. Scanlan and R. Levy, Circuit Theory. Oliver cuits,” Proc. of the IRE, vol. 36, no. 2, pp. 217–
& Boyd, 1973, vol. 2. 220, Feb. 1948.
[5] ——, Circuit Theory. Oliver & Boyd, 1973, [13] ——, “General impedance-function theory,”
vol. 1. Quarterly Applied Mathematics, vol. 6, no. 6,
[6] V. Belevitch, “Techbyshev filters and ampli- pp. 21–29, 1948.
fier networks,” Wireless Engineer, pp. 106– [14] J. Rhodes, Theory of Electrical Filters. John
110, Apr. 1952. Wiley & Sons, 1976.
[7] H. Orchard, “Formulae for ladder filters,” [15] H. Baher, Synthesis of Electrical Networks.
Wireless Engineer, pp. 3–5, Jan. 1953. John Wiley & Sons, 1984.
[8] R. Daniels, Approximation Methods for Elec- [16] M. E. Van Valkenburg, Introduction to modern
tronic Filter Design. McGraw-Hill, 1974. network synthesis. John Wiley & Sons, 1960.
[9] M. Lutovac, D. Miroslav, D. Tosic, V. Dejan, [17] A. I. Zverev et al., Handbook of filter synthesis.
and B. Evans, Filter Design for Signal Process- John Wiley & Sons, 1967.
ing using MATLAB and Mathematica. Prentice [18] R. Cameron, R. Mansour, and C. Kudsia,
FILTERS 73

Microwave Filters for Communication Systems: 1990, pp. 499–502.

Fundamentals, Design and Applications. John [31] U. Karacaoglu and I. Robertson, “High se-
Wiley and Sons, 2007. lectivity varactor-tuned MMIC bandpass fil-
[19] R. Cameron, “Advanced coupling matrix ter using lossless active resonators,” in IEEE
synthesis techniques for microwave filters,” 1994 Microwave and Millimeter-Wave Mono-
IEEE Trans. on Microwave Theory and Tech- lithic Circuits Symp. Dig. of Papers, May 1994,
niques, vol. 51, no. 1, pp. 1–10, Jan. 2003. pp. 237–240.
[20] I. Hunter, J. Rhodes, and V. Dassonville, [32] G. Mazzaro, M. Steer, and K. Gard, “Inter-
“Dual-mode filters with conductor-loaded modulation distortion in narrowband ampli-
dielectric resonators,” IEEE Trans. on Mi- fier circuits,” IET Microwaves, Antennas Prop-
crowave Theory and Techniques, vol. 47, no. 12, agation, vol. 4, no. 9, pp. 1149–1156, Sep. 2010.
pp. 2304–2311, Dec. 1999. [33] G. Mazzaro, M. Steer, K. Gard, and A. Walker,
[21] W. Fathelbab and M. Steer, “A reconfig- “Response of RF networks to transient wave-
urable bandpass filter for RF/microwave forms: Interference in frequency-hopped
multifunctional systems,” IEEE Trans. on Mi- communications,” IEEE Trans. on Microwave
crowave Theory and Techniques, vol. 53, no. 3, Theory and Techniques, vol. 56, no. 12, pp.
pp. 1111–1116, Mar. 2005. 2808–2814, Dec. 2008.
[22] A. Nakamura, K. Hibino, F. Yamamoto, and [34] G. Mazzaro, “Time-frequency effects in wire-
S. Kamihashi, “Expert system for microwave less communication systems,” Ph.D. disser-
filter design,” in 1990 IEEE MTT-S Int. Mi- tation, North Carolina State University, 2009.
crowave Symp. Dig., May 1990, pp. 1183–1186. [35] J. Malherbe, Microwave Transmission Line Fil-
[23] W. Fathelbab and M. Steer, “Broadband net- ters. Artech House, 1979.
work design,” in Multifunctional Adaptive Mi- [36] J.-S. Hong and M. Lancaster, Microstrip Filters
crowave Circuits and Systems, M. Steer and for RF/Microwave Applications. John Wiley &
W. Palmer, Eds., 2008, ch. 8. Sons, 2001.
[24] ——, “Parallel-coupled line filters with en- [37] H. Howe, Stripline Circuit Design. Artech
hanced stopband performance,” IEEE Trans. House, 1974.
on Microwave Theory and Techniques, vol. 53, [38] R. E. Collin, Foundations for Microwave Engi-
no. 12, pp. 3774–3781, Dec. 2005. neering. John Wiley & Sons, 2007.
[25] P. Cheong, S.-W. Fok, and K.-W. Tam, “Minia- [39] D. G. Swanson and W. J. Hoefer, Microwave
turized parallel coupled-line bandpass filter circuit modeling using electromagnetic field sim-
with spurious-response suppression,” IEEE ulation. Artech House, 2003.
Trans. on Microwave Theory and Techniques, [40] W.-K. Chen, Theory and design of broadband
vol. 53, no. 5, pp. 1810–1816, May 2005. matching networks. Pergamon Press, 1976.
[26] I. Bahl and P. Bhartia, Microwave Solid State [41] ——, Passive and Active Filters: Theory and Im-
Circuit Design. John Wiley & Sons, 1988. plementations. John Wiley & Sons, 1986.
[27] F. Rosenbaum, R. Gregory, W. Richard, [42] H. J. Blinchikoff and A. I. Zverev, Filtering in
W. Ou, F. Kuhns, and T. Trimble, “An MMIC the time and frequency domains. Krieger Pub-
twin-tee active bandpass filter,” in 1993 IEEE lishing Co., 1986.
MTT-S Int. Microwave Symp. Dig., 1993, pp. [43] L. Zhu, S. Sun, and R. Li, Microwave Bandpass
361–364. Filters for Wideband Communications. John
[28] B. Kapilevich, “Active microwave filters,” Wiley & Sons, 2011.
Telecommunication Radio Engineering, vol. 4, [44] J.-S. Hong, Microstrip Filters for RF/Microwave
no. 2, pp. 51–58, Feb. 1985. Applications, 2nd ed. John Wiley & Sons,
[29] C. Chang and T. Itoh, “Narrowband planar 2011.
microwave active filter,” Electronics Letters, [45] N. Kinayman and M. Aksun, Modern mi-
vol. 25, no. 18, pp. 1228–1229, Aug. 1989. crowave circuits. Artech House, 2005.
[30] C.-Y. Chang and T. Itoh, “A varactor-tuned, [46] R. M. Foster, “Academic and theoretical as-
active microwave band-pass filter,” in 1990 pects of circuit theory,” Proc. IRE, vol. 50,
IEEE MTT-S Int. Microwave Symp. Dig., May no. 5, pp. 866–871, May 1962.
74 STEER MICROWAVE AND RF DESIGN: MODULES

2.21 Exercises
1. The characteristic function of a doubly termi- That is, develop the RLC circuit that realizes
nated network is K(s) = s2 . Zw . [Parallels Example 2.3]
(a) What is the magnitude-squared transmis- 8. Synthesize the impedance function
sion coefficient (|T (s)|2 )?
(b) What is the magnitude-squared reflection 4s2 + 2s + 1
Zx = .
coefficient (|Γ(s)|2 )? 4s2 + 1
(c) What is transmission coefficient T (s) for a
That is, develop the RLC circuit that realizes
circuit that can be realized using positive R,
Zx . [Parallels Example 2.2. You may also want
L, and C elements)?
to consult Figure 2-9.]
2. The characteristic function of a doubly ter-
minated network is K(s) = s4 . What is 9. Synthesize the impedance function
the magnitude-squared transmission coefficient 4s4 + 2s3 + 5s2 + 2s + 1
(|Γ(s)|2 )? Zx = .
4s4 + 4s3 + 7s2 + s + 1
3. Consider the design of a fourth-order lowpass
Butterworth filter. [This problem follows the de- That is, develop the RLC circuit that realizes
velopment in Section 2.4.] Zx . [Parallels Example 2.2. You may also want
(a) What is the magnitude-squared characteris- to consult Figure 2-9.]
tic polynomial, |K(s)|2 , of the Butterworth 10. Develop the lowpass prototype of a fifth-order
filter? Butterworth lowpass filter. There may be more
(b) What is the magnitude-squared transmis- than one solution. That is, draw the circuit of the
sion coefficient (or transfer function)? lowpass filter prototype with element values.
(c) What is the magnitude-squared reflection 11. Develop the lowpass prototype of a fifth-order
coefficient function? Chebyshev lowpass filter with 1 dB ripple and 1
(d) Derive the reflection coefficient function rad/s corner frequency.
(i.e., Γ(s)). Write down the reflection coeffi-
cient in factorized form using up to second- 12. Develop the lowpass prototype of a ninth-order
order factors. Chebyshev lowpass filter with 0.01 dB ripple
(e) What are the roots of the numerator polyno- and 1 rad/s corner frequency.
mial of the reflection coefficient function? 13. A 0.04 S admittance inverter is to be imple-
(f) What are the roots of the denominator poly- mented in microstrip using a single length of
nomial of the reflection coefficient function? transmission line. The effective permittivity of
(g) Identify the conjugate pole pairs in the fac- the line is 9 and the design center frequency is
torized reflection coefficient. 10 GHz.
(h) Plot the poles and zeros of the reflection co- (a) What is the characteristic impedance of the
efficient on the complex s plane. transmission line?
4. Derive the reflection coefficient poles of a (b) What is the wavelength in millimeters at the
second-order Butterworth filter and write out design center frequency in free space?
the reflection coefficient with nominator and de- (c) What is the wavelength in millimeters at the
nominator polynomials, that is not in factorized design center frequency in microstrip?
form. [Parallels Example 2.1] (d) What is electrical length of the microstrip
5. Derive the reflection coefficient poles and zeros transmission line in degrees at the design
of a fourth-order Chebyshev filter with a ripple center frequency?
factor, ε, of 0.1. (e) What is the length of the microstrip trans-
mission line in millimeters?
6. Synthesize the impedance function
14. In Section 2.8.2 it was seen that a series induc-
s3 + s2 + 2s + 1 tor can be replaced by a shunt capacitor with
Zx = .
s2 + s + 1 inverters and a negative unity transformer. If
That is, develop the RLC circuit that realizes Zx . the inverter is realized with a one-quarter wave-
[Parallels Example 2.2] length long transmission line of characteristic
7. Synthesize the impedance function impedance 50 Ω:

4s2 + 2s + 1 (a) Derive the ABCD parameters of the cas-

Zw = 3 2
. cade of Figure 2-16(c) with 50 Ω inverters.
8s + 8s + 2s + 1
FILTERS 75

(b) What is the value of the shunt capacitance 20. Design a third-order Type 1 Chebyshev high-
in the cascade required to realize a 1 nH pass filter with a corner frequency of 1 GHz, a
inductor? system impedance of 50 Ω, and 0.2 dB ripple.
There are a number of steps in the design, and
15. A series inductor of 10 pH must be realized
to demonstrate that you understand them you
by an equivalent circuit using shunt capacitors
are asked to complete the partial designs indi-
and sections of one-quarter wavelength long 1 Ω
cated below. A Cauer 1 lowpass filter prototype
transmission line. Design the equivalent circuit.
is shown below with ωc being the corner radian
[Hint: The one-quarter wavelength long lines
frequency, fc = ωc /(2π) being the corner fre-
are impedance inverters.]
quency, and Z0 being the system impedance.
16. A series inductor of 10 nH must be realized by
an equivalent circuit using shunt capacitors and
sections of one-quarter wavelength long 50 Ω
transmission line. [Hint: The one-quarter wave-
length long lines are impedance inverters.] De-
sign the equivalent circuit.
17. At 5 GHz, a series 5 nH inductor is to be realized (a) Design an LPF with ωc = 1 rad/s, Z0 = 1 Ω.
using one or more 75 Ω impedance inverters, a (b) Design a HPF with ωc = 1 rad/s, Z0 = 1 Ω.
unity transformer, and a capacitor. What is the (c) Design a HPF with fc = 1 GHz, Z0 = 1 Ω.
value of the capacitor? (d) Design a HPF with fc = 1 GHz, Z0 = 50 Ω.
18. In Section 2.8.3 it was seen that a series capacitor 21. This problem considers the design of a Butter-
can be replaced by a shunt inductor with invert- worth bandpass filter at 900 MHz.
ers and a negative unity transformer. Consider (a) Design an LC second-order Butterworth
that the inverters are realized with a one-quarter lowpass filter with a corner frequency of
wavelength long transmission line of character- 1 rad/s in a 1 Ω system.
istic impedance 100 Ω. (b) Using the above filter prototype, design a
(a) Derive the ABCD parameters of the cas- lowpass filter with a corner frequency of
cade of Figure 2-17 with the 100 Ω invert- 900 MHz.
ers. (c) Design a second-order Butterworth band-
(b) What is the value of the shunt inductance pass filter at 900 MHz using the lowpass fil-
in the cascade required to realize a 1 pH ter prototype in (a). Use a fractional band-
capacitor? width of 0.1 and a system impedance of
50 Ω.
19. A 50 Ω impedance inverter is to be realized us- (d) What is the 3 dB bandwidth of the filter in
ing three resonant stubs. The center frequency of (c)?
the design is f0 . The first resonant frequency of
the stubs is fr = 2f0 . 22. Design a third-order maximally flat bandpass
filter prototype in a 50 Ω system centered at 1
(a) Draw the circuit using stubs. On your di-
GHz with a 10% bandwidth. The lowpass pro-
agram indicate the input impedance and
totype of a third-order maximally flat filter is
characteristic impedance of each of the stubs
shown in Figure 2-10.
if fr = 2f0 .
(b) What is the input impedance of a shorted (a) Convert the prototype lowpass filter to a
one-eighth wavelength long transmission lowpass filter with inverters and capacitors
line if the characteristic impedance of the only; that is, remove the series inductors.
line is Z01 ? (b) Scale the filter to take the corner frequency
(c) What is the input impedance of an open- from 1 rad/s to 1 GHz.
ended one-eighth wavelength long trans- (c) Transform the lowpass filter into a bandpass
mission filter. That is, replace each shunt capacitor by
(d) What is the input impedance of a shorted a parallel LC network. This step will estab-
one-eighth wavelength long transmission if lish the bandwidth of the filter.
the characteristic impedance of the line is (d) Transform the system impedance of the fil-
Z02 ? ter from 1 to 50 Ω.
(e) What is the length of each of the stubs in the 23. The lowpass prototype of a fourth-order low-
inverter in terms of the wavelength at the pass Butterworth filter is shown below. The cor-
frequency f0 ? ner frequency is 1 rad/s.
76 STEER MICROWAVE AND RF DESIGN: MODULES

(d) Complete the third HPF column of the table

with fc = 1 GHz, Z0 = 50 Ω.
ELEMENT LPF
ωc = 1 rad/s, Z0 = 1 Ω
L or C Value (units)
L1 = 0.765369 H C2 = 1.847759 F X1
L3 = 1.847759 H C4 = 0.765369 F
X2
Based on this, develop a fourth-order Butter- X3
worth bandpass filter prototype centered at
109 rad/s with a fractional bandwidth of 5%. ELEMENT HPF
(a) Scale the lowpass prototype to have a corner ωc = 1 rad/s, Z0 = 1 Ω
frequency of 109 rad/s. Draw the prototype L or C Value (units)
with element values. X1
(b) Draw the schematic of the lumped-element X2
fourth-order Butterworth bandpass proto- X3
type based on the original lowpass filter pro-
totype. ELEMENT HPF
(c) Derive the element values of the lumped- fc = 1 GHz, Z0 = 1 Ω

element bandpass filter prototype in a 75 Ω L or C Value (units)

system. X1
24. Design a third-order Type 2 Chebyshev high- X2
pass filter with a corner frequency of 1 GHz, a X3
system impedance of 50 Ω, and 0.2 dB ripple. ELEMENT HPF
There are a number of steps in the design, and fc = 1 GHz, Z0 = 50 Ω
to demonstrate that you understand them you L or C Value (units)
are asked to complete the table below. For each
X1
stage of the filter synthesis you must indicate
whether the element is an inductance or a ca- X2
pacitance by writing L or C in the appropriate X3
cell. Other cells require a numeric value and you
must include units. The X element is identified
25. What is the input impedance of a shorted
in the prototype below. A Cauer 2 lowpass filter
λ/8 long transmission line with a characteristic
prototype is shown with ωc being the corner ra-
impedance of Z0 ?
dian frequency, fc = ωc /(2π) being the corner
frequency, and Z0 being the system impedance. 26. What is the input impedance of an open-
circuited λ/8 long transmission line with a char-
acteristic impedance of Z0 ?
27. Apply Richards’s transformation to a shunt in-
ductor with a reactance of 50 Ω. What is the elec-
trical length of the shorted stub if the stub has a
characteristic impedance of 50 Ω?
(a) Complete the LPF (lowpass filter) column of
the table with ωc = 1 rad/s, Z0 = 1 Ω. 28. Apply Richards’s transformation to a shunt ca-
(b) Complete the HPF (highpass filter) column pacitor with a reactance of −50 Ω. What is the
of the table with ωc = 1 rad/s, Z0 = 1 Ω. characteristic impedance of the open-circuited
(c) Complete the second HPF column of the ta- stub if the electrical length of the stub is one-
ble with fc = 1 GHz, Z0 = 1 Ω. quarter wavelength long?
FILTERS 77

2.21.1 Exercises by Section

†
challenging, ‡ very challenging
§2.2 1† , 2 §2.6 6† , 7† , 8† , 9† §2.9 20† , 21† , 2† 2, 23† , 24†
§2.4 3† , 4† †
§2.7 10, 11 , 12 †
§2.12 25, 26, 27† , 28†
§2.5 5† §2.8 13, 14† , 15† , 16† , 17† , 18† , 19†

2.21.2 Answers to Selected Exercises

s2 7 13(c) 9
1(c) Γ(s) = √
s2 + 2s + 1 17 889 fF
2 |T (ω)|2 = 1/(1 + ω 8 ) 18 10 nH
3(f) −0.38 + 0.92 19(c) −Z2 = −50 Ω
−0.38 − 0.92 24 1 GHz, 50 Ω,
1 sC(50)2
 
−0.92 + 0.38 HPF: X3 = 6.5 nH
14(a)
−0.38 − 0.92 0 1
 CHAPTER 3

Parallel Coupled-Line Filters

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
3.2 Coupled Line Configurations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
3.3 Inverter Network Scaling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
3.4 Case Study: Third-Order Chebyshev Combline Filter Design . . . . 93
3.5 Parallel Coupled Line Filters in an Inhomogeneous Medium . . . 114
3.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
3.7 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
3.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116

3.1 Introduction
Central to microwave filter design is the identification of a particular
distributed structure that inherently has the desired frequency-shaping
attributes. Then this structure is adjusted to match an ideal mathematical,
or perhaps lumped-element, representation of a filter. Filter design is
both science and art, matching synthesis with instinctive knowledge of
appropriate physical and circuit structures.
A large number of microwave filters are realized using coupled
transmission lines and the most important of these are parallel coupled
line (PCL) filters. These are derived from prototypes, with the development
following a strategy that enables the filter to be realized using one of several
coupled-line configurations. So the strategy is to first examine coupled-line
configurations and determine the types of circuit structures that can be
realized. Then the steps in synthesis are designed to go from an LC filter
prototype to a prototype that has the structures that can be realized by a
coupled-line configuration.
Since the beginnings of microwave circuit synthesis, it has become
common to present circuit concepts using examples, very often because the
steps in realizing a filter may be too difficult to specify algorithmically and
there are many steps that require intuition. This procedure is followed here.
This introduction concludes with an example that illustrates the art and
science of microwave engineering. It is seen that a simple pair of coupled
lines with appropriate matching has a bandpass filter response. The intrinsic
bandpass response is supported by coupled lines, so the synthesis procedure
is adapting the intrinsic response to specifications.
80 STEER MICROWAVE AND RF DESIGN: MODULES

EXAMPLE 3.1 Coupled Lines as a Basic Element of a Bandpass Filter

Design Environment Project File: RFDesign Coupled Shorted Microstrip Lines C.emp

A coupled pair of microstrip lines terminated in short circuits is shown in Figure 3-1. This
circuit has a bandpass response centered at 7.5 GHz, as shown in Figure 3-2. So while the
insertion loss reduces to about 3.5 dB at 7.5 GHz, the return loss is not very high. This is a
good indication that matching could be used to reduce the insertion loss. So the problem is
then one of determining the appropriate matching network and to do this an understanding
of the circuit response needs to be gained. Figure 3-3 plots the S11 and S21 responses on
a Smith chart. By looking at S11 it is seen that between 7 and 8 GHz the input impedance
is primarily inductive. This indicates that a series capacitance could be used in matching.
Another view is shown in Figure 3-4, which plots the magnitude and phase of the input
impedance at Port 1 of the coupled microstrip lines (this will be the same as the input
impedance at Port 2). So around 7 GHz, the input impedance is inductive since the phase
of the impedance is close to 90◦ and the magnitude of the impedance is increasing with
frequency. At 7 GHz the impedance is approximately 50 Ω, which would be resonated out by
a series capacitor of 0.45 pF. However it will not be as simple as this, as the tuning capacitor
will need to be placed at both Ports 1 and 2. However, this does serve to provide an initial
design point.
The schematic required to implement the design above is shown in Figure 3-5. In
this circuit schematic, the coupled line shown in Figure 3-1(b) is captured as a subcircuit
called ”CoupledLine.” The series capacitors at Ports 1 and 2 are ”C1” and ”C2,” that is,
C1 and C2 , respectively. C1 and C2 are established as tunable elements with the governing
variable being ”CC.” This capacitor is tuned to obtain the optimum bandpass response. With
C1 = C2 = 44.8 fF, the bandpass response shown in Figure 3-6 is obtained. This is an
almost ideal maximally flat bandpass filter response. The main passband is at 7 GHz and
there is a parasitic bandpass response at the third harmonic. This spurious passband at an
odd harmonic occurs often in transmission line designs. A small alteration of the tuning
capacitor can change the response to have ripples in the passband and sharper filter skirts
(see Figure 3-7), where the tuning capacitors are each 37.6 fF. (To convince yourself of the
sharper skirt consider the insertion loss at 100 MHz away from the passband. The insertion
loss of the filter with the maximally flat response is 23 dB there, and that of the filter with
the ripple response is 27 dB.) This filter has two passband poles and is simple enough to
design as done here. Higher-order filters (with more than two passband resonators) require
a more sophisticated design approach. Still this example demonstrates that the coupled-line
structure has a good passband response on its own provided that appropriate matching is
used.
The bandpass characteristic is a result of the phase velocities of the even and odd modes
being different. If they were the same, the transmission coefficient would be zero at 7 GHz.

3.2 Parallel Coupled Lines

Several of the key advantages of coupled-line filters over other distributed
filters is that the coupling between resonators can be strong if the pair of
coupled-line resonators are close to each other, or low if they are widely
separated. This enables filters to be realized with high (30–70%), medium
(10–30%), or low (2–10%) bandwidths. The low-bandwidth filters require
low coupling of the resonators, while the high bandwidth filters require high
coupling. Coupled-line filters are also compact and can be implemented in
microstrip configurations or most of the other transmission line technologies.
The different transmission line forms have different Qs. Planar transmission
lines have moderate Qs especially at high frequencies, due principally to
PARALLEL COUPLED-LINE FILTERS 81

(a) (b)

Figure 3-1: Coupled microstrip line layout: (a) schematic; and (b) layout in an EM simulator.
Dimensions of the coupled lines are w = 500 µm, s = 100 µm, ℓ = 1 cm, W = 6 mm, and L
= 12 mm. The metal is 6 µm thick gold (conductivity σ = 42.6 × 106 S/m) and the alumina
substrate height is 600 µm with relative permittivity εr = 9.8 and loss tangent of 0.001.

Figure 3-2: Insertion loss (S21 )

and return loss (S11 ) of the
coupled line of Figure 3-1 with
an s = 100 µm gap.

Figure 3-3: Insertion loss and return loss over a narrow

frequency range plotted on a Smith chart. The annotation
at the markers is the frequency in GHz.

radiated fields. Low Q means that the insertion loss of filters will be high and
the filter skirts will not be as steep as they would be if a high-Q transmission
line structure was used. For example, up to the early 2000s the RF filters
in the front end of cellular phones were mostly seventh-order Chebyshev
filters using coupled slablines as shown in Figure 3-8 (although here only
82 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 3-4: Input impedance at Port 1 of the

coupled microstrip lines in Figure 3-1. Port
2 is terminated in 50 Ω.

Figure 3-5: Capacitively

coupled microstrip lines:
(a) schematic; and (b) rep-
resentation in a microwave
CAD tool with a subcircuit
representing the coupled
lines. The layout of the
subcircuit is shown in
Figure 3-1. (a) (b)

(a) Wide frequency view (b) Detail

Figure 3-6: Return loss and insertion loss when the coupling capacitors are 44.8 fF. This results
′
in a Butterworth-like response. S21 is the response without C1 and C2 .

four resonators are shown for clarity). Typically the top and bottom plates
of the resonators were 2–3 mm apart and the area of the slabline filter was
1 cm×1 cm. This is too large for today’s thin smart phones. However, they
are very good filters and cheap to produce. The electrical design procedure
for parallel coupled slabline filters and for using other transmission line
structures is the same as for parallel coupled microstrip filters. The difference
is only in the final physical implementation.
PARALLEL COUPLED-LINE FILTERS 83

(a) Narrowband response (b) Detail showing ripples

Figure 3-7: Return loss and insertion loss when the coupling capacitors are 37.6 fF. This results
′
in a Chebyshev-like response. S21 is the response without C1 and C2 .

Figure 3-8: Slabline: (a) cross

section of a single slabline; and
(b) parallel coupled-line con-
figuration with four coupled
(a) (b) slablines.

3.2.1 Coupled-Line Configurations

The coupled-line configuration considered in Example 3.1 is called a
combline section. This is one of many that have desirable frequency-selective
responses. The characteristics of many parallel coupled-line configurations
having bandpass, all-pass, or all-stop characteristics are shown in Table 3-1.
The most important configurations are starred (⋆).
The PCL section of Filter (a) in Table 3-1 is called an interdigital section and
is a bandpass filter. The dual filter is Filter (b), with what is called a parallel
coupled section, and both can be used to realize narrow (2–3%) to wide (up
to 30%) bandwidth filters with close to symmetrical responses around the
center frequency f0 . For a wide bandwidth filter the lines are close and for
a narrow bandwidth filter the parallel lines are widely separated so that
there is low-level coupling. Filter (c) is also a bandpass filter, however, the
frequency selectivity is not as good as with Filters (a) and (b). Filters (d),
(e), and (f) are all-pass filters. These filters are used to adjust the phase of
a transmitted signal, usually to accommodate phase dispersion (e.g., phase
variation with respect to frequency) that was introduced somewhere else.
While this analog function was once important, it is less so now as DSP
techniques can usually equalize the phase sufficiently. Filter (g) is an all-
stop filter, meaning that negligible signal is transmitted. This filter has no
practical use. Filters (h) and (i) are inherently all-stop filters. An example of
the response of the combline section of Filter (h) was seen in Figure 3-2 of
Example 3.1, where the all-stop response is centered at 4 GHz. The response
is too broad to be useful as a bandstop filter (in any practical application).
84 STEER MICROWAVE AND RF DESIGN: MODULES

Table 3-1: Responses of the nine coupled-line configurations with narrow (2–3%), moderate (3–
10%), and wide (10%–30%) bandwidth. The most important configurations are starred (⋆).
Attributes Circuit Response
⋆ (a) Interdigital section.
Bandpass.
Narrow to wide bandwidth.
ℓ = λ/4 at f0 .
Symmetrical passband.
⋆ (b) Parallel coupled section.
Bandpass.
Narrow to wide bandwidth.
ℓ = λ/4 at f0 .
Symmetrical passband.

(c) Bandpass.
Moderate to wide bandwidth.
ℓ = λ/4 at f0 .

(d) All-pass.
0 < ℓ ≤ λ/2 at f0 .

(e) All-pass.
0 < ℓ ≤ λ/2 at f0 .

(f) All-pass.
0 < ℓ ≤ λ/2 at f0 .

(g) All-stop (no practical use).

⋆ (h) Combline section.

All-stop without matching.
Bandpass with matching.
Moderate to wide bandwidth.
ℓ = λ/4 at f0 . Compact.
Asymmetrical response.
(i) All-stop without matching.
Bandpass with matching.
Moderate to wide bandwidth.
ℓ = λ/4. Compact.
Asymmetrical response.

However, with minimal matching, such as a series capacitor at each of ports

(see Figure 3-5), the combline section becomes a very good bandpass filter
(see Figures 3-6 and 3-7). The combline bandpass filter, Filter (h), is more
compact than Filters (a) and (b). However, it has an asymmetrical bandpass
response (which can be seen in Figure 3-7(a)), whereas Filters (a) and (b)
have a symmetrical frequency response. Which type of response is preferred
depends on the application. Filter (i) is the dual of Filter (h), but in practice
PARALLEL COUPLED-LINE FILTERS 85

Table 3-2: Responses of coupled-line configurations having lowpass and bandstop responses.
Attributes Circuit Response
⋆ (j) Lowpass.
Narrow to moderate
bandwidth.
ℓ = λ/2 at f0

⋆ (k) Lowpass.
Narrow to moderate
bandwidth.
ℓ = λ/4 at f0

⋆ (l) Lowpass.
Narrow bandwidth.
ℓ = λ/4 at f0 .

⋆ (m) Bandstop.
Narrow bandwidth.
ℓ = λ/4 at f0 .

the combline configuration of Filter (h) is preferred.

Table 3-2 presents coupled-line sections having lowpass and bandstop
responses. All of these filters have desirable characteristics and so all are
starred (⋆). Filter (l) is not a PCL filter but it is used in trade-off among the
three lowpass filters (i.e., Filters (j), (l) and (m)). The usual application of a
bandstop filter is to notch out an undesired signal (e.g. to prevent an LO from
appearing where it is not desired). As such, the desired frequency response
is narrowband and Filter (m) is a good choice for a bandstop filter.
Filter design using PCL sections begins with the choice of a PCL
configuration having the essential desired frequency response. Each of the
PCL sections in Tables 3-1 and 3-2 have two resonators and can implement
bandpass, bandstop, and lowpass filters having second-order responses.1 To
meet specifications it is generally necessary to replicate the basic section. So
treating the two resonator configurations as unit cells, a multi-cell filter can
be realized. The multi-cell forms of the main bandpass configurations are
shown in Table 3-3. The multi-cell implementation of the interdigital, parallel

1 Recall that the order designation comes from the lowpass prototype so that here a second-
order bandpass filter actually has two resonators, each having an LC-like response.
86 STEER MICROWAVE AND RF DESIGN: MODULES

Table 3-3: Multi-cell forms of bandpass parallel coupled-line configurations. The wavelength, λ,
is at the center frequency of the filter.
Name Unit cell Multi-cell form
(a) Interdigital bandpass filter.
ℓ = λ/4.

(b) Parallel edge-coupled

bandpass filter.
ℓ = λ/4.

(c) Parallel edge-coupled

hairpin bandpass filter.
ℓ > λ/4.

(d) Combline bandpass filter.

ℓ = λ/4. Matching required.

(e) Combline bandpass filter

with extended stopband. ℓ = λ/4.
ℓ2 ≈ λ/8 (typically).
Matching required.

edge-coupled bandpass filters, (a) and (b) in Tables 3-3, is straightforward.

A variation is shown in Filter (c), which folds the parallel edge-coupled
sections to realize a compact multi-cell form called a hairpin filter [1–3].
One of the consequences of using transmission line segments to realize a
filter is that there are spurious passbands. This results because a λ/4 section
of line looks the same electrically as a 3λ/4 section. Consequently there
are normally spurious passbands at the odd-harmonic frequencies. If the
basic resonator is λ/2 long, then the spurious passbands would be at all
of the harmonic frequencies. A solution is to incorporate capacitors in the
PARALLEL COUPLED-LINE FILTERS 87

→ Figure 3-9: Network

model of a pair of
coupled lines.

resonators as shown in Filter (e) in Table 3-3 [2, 4, 5]. Thus each of the original
transmission line resonators now becomes a combination of a capacitor and
a shorter transmission line segment. If the new transmission line segment
is λ/8 long, the first spurious passband will now be at 5f0 rather than 3f0 .
The spurious passbands can be pushed further up in frequency by using
even shorter transmission line lengths, but the performance of the filter at
the passband frequency, f0 , will be compromised.
There are many variations on the PCL filter and several different design
techniques have been developed [6–28]. Refer to these citations to explore
alternative PCL configurations and alternative methods of design to that
presented here. This chapter presents one of the common approaches to
synthesizing PCL filters and the scheme accommodates the use of capacitive
loading used to extend the stopband of a passband filter.
Before launching into the synthesis procedure for a PCL filter, a comment
on synthesis versus optimization is warranted. With a simple topology such
as the second-order filter topologies shown in Table 3-1, an optimization
procedure could be used to design the widths and lengths of the lines and
the two or four variables describing the required external matching networks
that are not shown. A global optimization is certainly feasible. However, it is
not feasible to use global optimization of, for example, a seventh-order filter
required to meet typical cell phone specifications. An exception is if a design
for a very similar specification is available and the changes required are
small. Competitive filter design requires synthesis. This will lead to optimum
performance, and the insight gained can be used in topology modifications.

3.2.2 Coupled-Line Circuit Models

An approximate coupled-line model of a pair of coupled lines was presented
in Section 5.9.5 of [29]. This model is repeated in Figure 3-9. The parameters
of the network model are related to the model impedances as follows:
1 Z0e + Z0o Z0S
n= = (3.1) Z01 = √ , (3.3)
K Z0e − Z0o 1 − K2
√
 1 − K2
Z0S = (Z0e Z0o ), (3.2) Z02 = Z0S . (3.4)
K2
Various terminating arrangements of the coupled lines result in several
useful filter elements. One arrangement is shown in Figure 3-10. Also shown
in this figure is the development of the network model based on the model
in Figure 3-9. The final network model is a transmission line of characteristic
impedance Z01 in cascade with an open-circuited stub. Consider what
happens at the resonant frequency, fr (the frequency at which the lines are
one-quarter wavelength long). At lower frequencies, f ≪ fr , the Z02 line
88 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 3-10: Lowpass distributed network section derived from a pair of coupled lines with Port
1 open-circuited. The open circuit is indicated by a node (open circle) with a line through it. The
final network model is a transmission line of characteristic impedance Z01 and an open-circuited
stub of characteristic impedance Z02 . The lines and stubs are one-quarter wavelength long at the
corner frequency. (Thus with the stub fr = f0 , and the characteristic impedance of the stub is as
shown.)

↓
Figure 3-11: Parallel
coupled-line section
with Ports 1 and 3
open-circuited and net-
work models. (For the
stub, the characteristic ←
impedance of the stub
is shown and fr = f0 .)

is an open circuit and signals travel along Z01 . At resonance the Z02 stub
becomes a short circuit and signals do not pass. This is a crude verification
that this is a lowpass structure. The process is visual and is expected to be
self-explanatory. Other examples are shown in Figures 3-11 to 3-13.
The model of a combline section is shown in Figure 3-13. The final network
reduction is repeated in Figure 3-14, and it will be shown that the model in
Figure 3-14(b) is equivalent to the model in Figure 3-14(a). In the synthesis
of a combline filter, the network of Figure 3-14(b) is obtained and this can be
PARALLEL COUPLED-LINE FILTERS 89

↓
Figure 3-12:
Interdigital sec-
tion and network
models. (For the
stub, the character-
← istic impedance of
the stub is shown
and fr = f0 .)

←
Figure 3-13: Combline
↓ section and network
models. (For the stubs,
the characteristic
impedances of the
stubs are shown and
fr = f0 .)

Figure 3-14: Equivalent models of a section

of combline. (For the stubs, the characteristic
impedances of the stubs are shown and fr =
(a) (b) f0 .)

related back to the dimensions of the coupled line. The equivalence is done
using ABCD parameters and, as will be seen, the equivalence will not be
at just one frequency but will be broadband. The ABCD parameters of the
network in Figure 3-14(a) are obtained by cascading the ABCD parameters
of three two-ports (the ABCD parameters of which are given in Table 2-1
of [30]). The ABCD parameters of the network in Figure 3-14(a) are (from
90 STEER MICROWAVE AND RF DESIGN: MODULES

the multiplication of three ABCD parameter matrices)

TA = TTRANSFORMER TSERIES STUB TSHUNT STUB (3.5)
� �� �� �
1/n 0 1 Z02 tan θ 1 0
= (3.6)
0 n 0 1 −/(Z01 tan θ) 1
� �� �
1/n Z02 tan θ/n 1 0
= (3.7)
0 n −/(Z01 tan θ) 1
 � � 
1 Z02
1 + Z 02 tan θ/n
= n Z01 . (3.8)
−n/(Z01 tan θ) n
The ABCD parameters of the network in Figure 3-14(b) are
TB = TSHUNT STUB TSERIES STUB TSHUNT STUB (3.9)
� �� �� �
1 0 1 Z012 tan θ 1 0
=
−/(Z011 tan θ) 1 0 1 −/(Z022 tan θ) 1
(3.10)
� �� �
1 Z012 tan θ 1 0
= (3.11)
−/(Z011 tan θ) 1 + Z012 /Z011 −/(Z022 tan θ) 1
1+ Z
 
� Z012 tan θ
012
� Z022
=  − 1 1 Z012 Z012  . (3.12)
+ + 1+
tan θ Z011 Z022 Z011 Z022 Z011
Equating Equations (3.8) and (3.12) yields
� �
Z012 1 Z02
1+ = 1+ (3.13)
Z022 n Z01
Z012 = Z02 /n (3.14)
� �
Z011 + Z022 + Z012 n
= (3.15)
Z011 Z022 Z01
Z012
1+ = n, (3.16)
Z011
which have the solution
Z02
Z012 = (3.18)
n
Z012 Z02 Z01 Z02
Z011 = = (3.17) Z022 = . (3.19)
n−1 n(n − 1) Z02 − (n − 1)Z01
Rearranging these, expressions for Z01 , Z02 , and n can be obtained:
� �
Z012 nZ011 Z022
n=1+ , Z01 = , and Z02 = nZ02 Z012 . (3.20)
Z011 Z011 + Z022 + Z012
Using these, Equations (3.1)–(3.4), and the coupled-line analysis of Section
5.6 of [29], the geometric parameters of the combline coupled-line section
can be obtained corresponding to the stub circuit of Figure 3-14(b).
Thus the equivalent circuit of the combline section, the top left figure in
Figure 3-13, has the equivalent circuit shown in Figure 3-14. Filter synthesis
can be directed at developing circuit structures like that in Figure 3-14(b)
and from this electrical design, the physical design consisting of combline
sections can be developed.
PARALLEL COUPLED-LINE FILTERS 91

(a) (b)

(c) (d)

Figure 3-15: Inverter network: (a) network with two identical admittance inverters with an
inserted shunt element of admittance, y; (b) equivalent network using equivalence shown in
Figure 2-19; (c) scaled original network; and (d) scaled equivalent network. Element values are
impedances except for J, y and y1 , which are admittances. The use of the equivalent networks
in filter synthesis is illustrated in the case study in Section 3.4.

3.3 Inverter Network Scaling

A filter that uses transmission lines is nearly always synthesized from
a filter prototype that contains inverters. The synthesis of a filter begins
with a normalized prototype that is transformed to the desired frequency,
impedance, and type. It is the purpose of this section to show how to use such
scaling when there are inverters in the prototype. In particular, it is shown
that scaling the admittance of the network of Figure 3-15(a), a common
inverter subnetwork, by a factor x results in the network in Figure 3-15(c).
The nodal admittance matrix of the network in Figure 3-15(a) is
 
0 −J 0
Y =  −J y −J  . (3.21)
0 −J 0

Assigning nodal voltages to Terminals 1, 2, and 3, the nodal admittance

matrix equation is
    
0 −J 0 v1 J1
 −J y −J   v2  =  0  , (3.22)
0 −J 0 v3 J3

and by eliminating Node 2 using network condensation (see Section 1.A.16)

of [31], this reduces to
� 2
J /y J 2 /y
�� � � �
v1 J1
= , (3.23)
J 2 /y J 2 /y v3 J3

and this describes the external characteristics of the subnetwork in Figure

3-15(a).
92 STEER MICROWAVE AND RF DESIGN: MODULES

The nodal admittance matrix of the scaled network in Figure 3-15(c) is

 √ 
0√ −J x 0√
Y ′ =  −J x yx√ −J x  . (3.24)
0 −J x 0

That is,
 √    
0√ −J x 0√ v1 J1
 −J x yx√ −J x   v2  =  0  , (3.25)
0 −J x 0 v3 J3

and by eliminating Node 2 this reduces to

(J 2 x)/(yx) (J 2 x)/(yx)
� � 2
J 2 /y
� �� �� � � �
v1 J /y v1 J1
= = .
(J 2 x)/(yx) (J 2 x)/(yx) v3 J 2 /y J 2 /y v3 J3
(3.26)

Thus the original network shown in Figure 3-15(a) has the same external
electrical characteristics as the scaled network of Figure√ 3-15(c), with the
characteristic admittance of the inverters scaled by x and the shunt
admittance scaled by x.
A generalization of this result (which is useful when there are additional
connections between Nodes 1 and 3) is that multiplying a row and a column
of the nodal admittance matrix by the same factor results in identical external
characteristics. Note that the element sharing a row and column is multiplied
twice.

EXAMPLE 3.2 Inductor Synthesis

Lumped inductors have low Q. Fortunately, in microwave design they can be realized using
inverters and a shunt capacitor that have a high Q. Realize the series inductor in Figure 3-
16(a) with a shunt capacitor and 10 Ω impedance inverters.
Solution:
The series inductor in a 1 Ω system is transformed into a network with inverters and a
shunt capacitor using the transformation shown in Figure 2-48. Thus here the series inductor
can be realized by the circuit of Figure 3-16(b), where the shunt capacitor C has the same
numeric value as the series inductor. The inverters here can be either impedance inverters or
admittance inverters, as their value is equal to one. The design requires that these be realized
as 10 Ω inverters, but scaling is performed on admittance inverters, as shown in Figure 3-
√
16(c), where the value of the admittance inverters is 0.1 S = x. Thus x = 0.01 and the
admittance of the capacitor is scaled by 0.01, so C1 = C/100 = 10 pF. The final design is
shown in Figure 3-16(d).
PARALLEL COUPLED-LINE FILTERS 93

3.4 Case Study: Third-Order Chebyshev Combline Filter

Design
This section presents the design of a 1 GHz third-order Chebyshev combline
bandpass filter with 10% bandwidth. The combline filter is one of the most
commonly used transmission line-based filters and uses the combline section
shown in Figure 3-13. The synthesis that will be presented here combines
many of the concepts discussed previously in this chapter. There are also
design decisions that must be made in response to particular situations
encountered during synthesis. An example of a design decision is deciding
what to do if the characteristic impedance of a transmission line is too low.

Step 1: Choice of Lowpass Filter Prototype

Design begins with the choice of a prototype that is expected to achieve
the required specifications. If the choice is eventually found to be not
suitable, then another topology must be chosen and the synthesis restarted.
Consider a third-order Chebyshev lowpass filter prototype, shown in Figure
3-17, with a passband ripple factor of ε = 0.1 (or 0.043 dB). The element
values were calculated using Equations (2.83)–(2.90). Note that at the corner
frequency that defines the bandwidth of the lowpass filter, the square of
the transmission coefficient is the ripple level (see Figure 2-7). That is, at
frequency ω0 = 1 rad/s, the squared transmission coefficient is 1/(1 + ε2 )
(or in decibels the ripple is RdB = 10 log(1 + ε2 )).

Step 2: Replace Series Elements

The next step in the transformation of the lowpass prototype in Figure 3-
17 is replacing the series inductor by a shunt capacitor in cascade with
admittance inverters, as shown in Figure 3-18. This step uses the series
inductor transformation shown in Figure 2-16.

Figure 3-16: Realization of a series inductor as a shunt capacitor with inverters.

Figure 3-17: Step 1. A third-order

Chebyshev lowpass filter proto-
type.
C11 = 0.85158 F = C31 L21 = 1.10316 H g1 = g3 = 0.85158, g2 = 1.10316.
94 STEER MICROWAVE AND RF DESIGN: MODULES

(a)
Figure 3-18: Step 2. Prototype filter with induc-
tor replaced by a capacitor and admittance in-
verter combination: (a) with a unity inverting
transformer; and (b) with the transformer elim-
inated as it shifts the transmission phase by 180◦
and so has no effect on the filter response. Note (b)
that |C21 | = |L21 |. C11 = 0.85158 F = C31 C21 = 1.10316 F

Figure 3-19: Target combline filter

physical layout. (a) Three-resonator combline filter (b) Model

Step 3: Bandpass Filter Transformation

In synthesizing the bandpass filter the final physical topology must be
considered. The essential element of a third-order combline filter comprises
three coupled lines as shown in Figure 3-19(a). The three resonators can be
modeled as two pairs of coupled lines, as shown in Figure 3-19(b), with one
line shared by each pair. Each pair of coupled lines can be modeled by a Pi
network of stubs as shown in Figure 3-14(b). Thus the lowpass prototype of
Figure 3-18(b) needs to be converted to a bandpass filter and eventually this
needs to be transformed into two cascaded Pi networks of stubs.
In this step the prototype is converted to a bandpass circuit. The filter
has 10% bandwidth with approximate corner frequencies at f1 = 950 MHz
and f2 = 1050 MHz. Using the filter type transformations in Figure 2-27,
ω0 = 2π109 , and α = ω0 /(ω2 − ω1 ) = 1/(fractional bandwidth) = 10, the first
capacitor in Figure 3-18, C11 , is transformed to
αC11 1
C1′ = = 1355.33 pF in parallel with L′1 = = 0.0186894 nH.
ω0 αC11 ω0
(3.27)
The other capacitors are transformed similarly and the bandpass prototype
becomes that shown in Figure 3-20.
The lumped-element version of the bandpass filter is shown in Figure 3-
21 along with its simulated response. This response is the ideal electrical
response and provides a reference in design. Note that the ripples in the
transmission response, i.e. S21 , are not evident on this scale. Losses in the
implemented filter will smooth the ripples even further although the poles
in the S21 response are not evident, these correspond to the zeros in the S11
response and these are clearly visible. Figure 3-21 shows three zeros in the
S11 response, as expected for a third-order filter.
PARALLEL COUPLED-LINE FILTERS 95

Figure 3-20: Step 3. Bandpass filter pro-

C1′ = 1355.33 pF = C3′ L′1 = 0.0186894 nH = L′3 totype with inverters derived from the
C2′ = 1755.73 pF L′2 = 0.0144271 nH lowpass filter prototype of Figure 3-18.

Z0 = 1 Ω
C12 = 1.35533 nF = C32
L12 = 18.6894 pH = L32
C22 = 14.4271 pF
L22 = 1.75573 nH

Figure 3-21: A third-order lumped-element Chebyshev bandpass filter with 10% bandwidth,
ripple factor of ε = 0.1, and center frequency of 1 GHz. The S parameters are referenced to 1 Ω.

Figure 3-22: Step 4. Bandpass filter

C1♯ = 27.1066 pF = C3♯ L♯1 = 0.934468 nH = L♯3 approximation where the inverters
C2♯ = 35.1147 pF L♯2 = 0.721359 nH are now impedance inverters.

Step 4: Impedance Scaling

The system impedance is now scaled from 1 to 50 Ω, leading to the prototype
in Figure 3-22 (derived from Figure 3-20).

Step 5: Conversion of Lumped-Element Resonators

Each lumped-element resonator in Figure 3-22 comprises a capacitor and
an inductor. The inductors are particularly difficult to realize efficiently and
have high loss. The idea is to replace each lumped resonator (e.g., C1♯ and
L♯1 in Figure 3-22) with a network comprising a lumped capacitor and a
transmission line stub.

Equivalence of a Capacitively-Loaded Stub and an LC Resonator

Here it will be shown that a capacitively-loaded stub is a broadband
approximation of a lumped LC resonator. The resonator equivalence is
shown in Figure 3-23. The equivalence is first developed at frequency f0 =
ω0 /(2π) by making sure that the circuits have the same admittance at the
96 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 3-23: Resonator equivalence: (a) lumped resonator

resonant at radian frequency ω0 ; (b) mixed lumped-
distributed resonator containing a short-circuited stub
of characteristic impedance Z0 and resonant at radian
frequency ωr ; and (c) short-circuited transmission line
stub resonant at radian frequency ωr . In (b) and (c) Z01 is
the characteristic impedance of the transmission line and
Z1 is the input impedance of the shorted transmission
line. (a) (b) (c)

filter center frequency. The broadband approximation is obtained by also

equating the derivatives of the admittances at ω0 .
The input admittance of the parallel LC lumped-element resonator is
 ω 2 LC − 1
� �
Yin = , (3.28)
ωL
and has a radian frequency derivative of
 ω 2 LC + 1
� �
dYin
= . (3.29)
dω ω2L
Now consider a short-circuited stub that is resonant at radian frequency
ωr ; that is, the short-circuited stub is one-quarter wavelength long at ωr .
From Equation (2.100) of [29], the input impedance at ω is
� �
π ω
Z1 = Z01 tan . (3.30)
2 ωr
Thus for the lumped-distributed network in Figure 3-23(b), the input
admittance is
� � � �
ω
 ωCo Z01 tan π2 −1
ωr
Yin′ = � � , (3.31)
π ω
Z01 tan
2 ωr
and its radian frequency derivative is
� � � ��2 � � ��2 � 
π ω π ω
 2Co Z01 tan 2 ω ωωr + π + π tan
2 ωr
dYin′

1 r 
=  ��2 .
dω 2
� �
π ω
 
Z01 tan ωr

2 ωr
(3.32)
For the lumped-element and lumped-distributed resonators to be identical
at ω0 and approximating each other at nearby frequencies, Equations (3.28)
and (3.31) are equated,
′
Yin (ω0 ) = Yin (ω0 ), (3.33)
and Equations (3.29) and (3.32) are also equated,
� ′
�
dYin �� dYin ��
= . (3.34)
dω �ω=ω0 dω �
�
ω=ω0
PARALLEL COUPLED-LINE FILTERS 97

Figure 3-24: Short-circuited transmission line used as a stub

At ωr , resonator in a bandpass filter. The line is resonant at the
θ = 90◦ , Z1 = ∞ radian frequency ωr and so the electrical length of the line
At ω = ωr /2, is θ = 90◦ . Using the design choice here, the filter center
θ = 45◦ , Z1 = Z01 passband frequency ω = ωr /2, and so θ = 45◦ and so the
input impedance of the shorted stub is Z1 = Z01 .

Specific Design Choice, ωr = 2ω0

At this stage a design choice must be made regarding the relationship of ω0
and ωr ; that is, the resonant frequency of the stub in relation to the resonant
frequency of the resonator (which is also the center frequency of the filter
passband). Here a shorted stub is considered that is resonant when it is λ/4
long. The specific design choice made here is that the resonant frequency
of the stub, fr , is twice the center frequency of the filter passband, i.e.
ωr = 2ω0 . This is a common design choice. The consequence of this choice
is illustrated in Figure 3-24. The frequency fr is called the commensurate
frequency to avoid confusion resulting from the resonant frequency of the
filter resonators being different from the resonant frequency of stub2 . The
stub design proceeds as follows.
The equivalence expressed by Equations (3.33) and (3.34) is established at
the center of the passband (i.e., at ω0 ). Now ω = ω0 = 12 ωr , so Equations
(3.28), (3.31), and (3.33) become
      
ωr 2 ωr π ωr /2
 LC − 1  C0 Z01 tan −1
2 2 2 ωr
ωr =   . (3.35)
L π ωr /2
2 Z01 tan
2 ωr
ω 
r
 2
ωr LC − 4
 C 0 Z 0 tan(π/4) − 1
That is, = 2 . (3.36)
2ωr L Z01 tan(π/4)
 2  1 
π ωr LC − 4 ωr C0 Z01 − 1
Since tan = 1, = 2 , (3.37)
4 2ωr L Z01
4 2
and rearranging, C0 = C − 2 + . (3.38)
ωr L ωr Z01

Another relationship comes from equating derivatives. From Equations

(3.29), (3.32), and (3.34), and with ω = 12 ωr and tan( π2 ωωr ) = tan π4 = 1,

2C0 Z01 21 ωr2 + π + π

 2
C0 Z01 ωr2 + 2π
    
ωr LC + 4 1
= 2 = . (3.39)
ωr2 L Z01 21 ωr Z01 ωr

Substituting for C0 from Equation (3.38) and rearranging, the characteristic

impedance of the stub is

ωr L  π
Z01 = 1+ . (3.40)
4 2

2 Of course there are resonant frequencies at every multiple of one-quarter wavelength. The
first resonance occurs at fr , as then the input impedance of the short-circuited stub is an open
circuit. The characteristic that establishes resonance is that the input is either an open- or short-
circuit and energy is stored.
98 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 3-25: Step 5. Bandpass

combline filter with impedance
inverters. The transmission
line stubs present impedances
Z1 = Z01 , Z2 = Z02 , and
Z3 = Z03 since the resonant fre-
quencies of the stubs are twice that
′′ ′′ ′′
of the design center frequency, i.e. C1 = 21.0881 pF = C3 C2 = 27.3181 pF
fr = 2f0 . Z01 = 7.54713 Ω = Z03 Z02 = 5.82598 Ω

In the passband the input impedance of the stub is

Z1 = Z01 tan(βℓ) = Z01 tan(π/4) = Z01 . (3.41)

Thus the broadband (≈30%) approximation of the parallel LC resonant

circuit is a capacitively-loaded stub, see Figure 3-23(b), of length λ/8 at the
resonant frequency of the LC resonator. The resonant frequency of each stub,
i.e. the frequency fr at which they are λ/4 long, is twice the design center
frequency f0 , i.e. fr = 2f0 . So the concept used here is that the resonant
frequency of a stub does not need to be the same as the design center
frequency. It is common for all the stubs in a design to have the same resonant
frequency fr and this is specified in the design documentation. In general fr
can have any relationship to f0 but the most common situation is fr = 2f0 as
used here.

Returning to Step 5
For this filter, the center of the passband is 1 GHz, and so ω0 = 2π109 and
ωr = 2ω0 = 2π(2 · 109 ) defines the stub resonant frequency as 2 GHz. The
first resonator in Figure 3-22 with L = 0.934468 nH and C = 27.1066 pF can
be replaced by the capacitively-loaded stub in Figure 3-23(b). where, from
Equation (3.40),

Z01 = 7.54713 Ω, and from Equation (3.38), C0 = 21.0881 pF. (3.42)

This process is repeated for each lumped-element resonator in Figure 3-22,

leading to the prototype shown in Figure 3-25.

Steps 6 and 7: Equating Characteristic Impedances of Stubs

The stubs in Figure 3-25 can be physically realized using transmission
line segments. It would be ideal if the impedances of the stubs (i.e., the
impedances looking into the stubs) are identical. To achieve this, the method
of nodal admittance matrix scaling described
 in Section 3.3 is used with the
impedance inverters multiplied by Z01 /Z02 and the admittance, C2′′ //Z2 ,
divided by Z01 /Z02 (C2′′ is divided by Z01 /Z02 , and Z2 is multiplied by
Z01 /Z02 ). This step scales each inverter impedance to 56.9084 Ω and also sets
the characteristic impedance of all the shunt stubs to 7.54713 Ω. The combline
filter is now as shown in Figure 3-26.
In Section 2.8.6 it was shown that a good narrowband model of an inverter
is a Pi network of stubs. So the inverter in the current filter design, shown
in Figure 3-27(a), is modeled by the lumped-element circuit in Figure 3-27(b)
and the stub circuit shown in Figure 3-27(c).
PARALLEL COUPLED-LINE FILTERS 99

Figure 3-26: Step 6. Bandpass

combline filter with impedance
′
C1′′ = C3′′ = C2′′′ = 21.0881 pF and Z02 = 7.54713 Ω = Z01 = Z03 inverters with fr = 2f0 .

(a) (b) (c)

Figure 3-27: Inverter translation used with the combline filter design: (a) impedance inverter;
(b) realization as a lumped-element circuit; and (c) realization using stubs resonant at twice the
passband center frequency.

Stub Impedance When fr = 2f0

The most common design choice is to select the resonant (i.e., the
commensurate) frequency of the stubs to be twice the center frequency of
the design. With fr = 2f0 , Equation (2.114) becomes
1 1 1
Z0 =  =  π  = = K. (3.43)
π ω0 J tan J
J tan 4
2 2ω0
Equation (3.43) defines the characteristic impedance of the transmission
line stubs realizing an inverter. Then the elements of the subnetwork in
Figure 3-27(b) all have the admittance magnitude

J = 1/K = 1/(56.9084 Ω) = 17.5721 mS (3.44)

and the stubs in Figure 3-27(c) have the characteristic impedance




1 K K 
Z0 = = =   = 56.9084 Ω. (3.45)
J tan (θ) tan (θ) π ω 
tan
2 ωr ω=ω0

At this stage there are several pairs of stubs in parallel. Figure 3-28
illustrates how a pair of parallel stubs can be replaced by a single stub. Now
the prototype is as shown in Figure 3-29.

Step 8: Scaling Characteristic Impedances of Stubs

The filter will be realized using coupled microstrip lines and a general
statement can be made about how the stubs in the prototype in Figure
3-29 correspond to the physical structure of the coupled lines. The stubs
′ ′′ ′
with characteristic impedances Z01 , Z02 , and Z03 are largely realized by the
′
properties of the individual microstrip lines. A low Z01 , for example, would
100 STEER MICROWAVE AND RF DESIGN: MODULES

(a)

(b)

(c)

Figure 3-28: Step 6b. Procedure combining two parallel

stubs to obtain one stub: (a) two parallel stubs; (b)
represented as parallel impedances; (c) as one impedance; (d)
and (d) final representation as one stub. Note that the stubs
must have the same commensurate frequency, i.e. they must
have the same electrical length.

Figure 3-29: Step 7: Bandpass fil-

′
ter approximation where Z01 ,
′′ ′
Z02 , Z03 , Z012 , and Z023 are the
characteristic impedances of the ′′ ′′ ′′′ ′′
stubs with commensurate fre- C1 = 21.0881 pF = C3 = C2 Z02 = 10.2715 Ω
′ ′
quency fr = 2f0 . Z01 = 8.70106 Ω = Z03 Z012 = 56.9084 Ω = Z023

require a wide microstrip line. The stubs with characteristic impedances Z012
and Z023 each correspond to coupling between pairs of microstrip lines with
high impedances corresponding to low-level coupling and widely spaced
microstrip lines.
In Figure 3-29 the short-circuited shunt stubs have characteristic
impedances of 8.7 Ω and 10.3 Ω, which are too low, and will result in wide
microstrip lines.3 These need to be scaled to a higher impedance level to raise
the characteristic impedances of the short-circuited stubs to an acceptable
value. Note that the characteristic impedance of the middle shunt short-
circuited stub can be raised to 80 Ω if the system impedance is raised from
50 Ω to 389.426 Ω. Doing this leads to the element values shown in the
bandpass circuit of Figure 3-30.

Step 9: Scaling to a 50 Ω System Impedance

The previous step raised the system impedance (the required source and
load impedances) to an unreasonably high level (389.426 Ω). For the filter to
operate in a 50 Ω system, input and output impedance inverters are required
to scale the source and load back to 50 Ω. The resulting circuit is shown in

3 What is reasonable for the width of a microstrip line is subjective and based on experience.
With a substrate having εr = 10, the width required for a characteristic impedance, Z0 , of
30–80 Ω is reasonable. A line with Z0 < 30 Ω will be very wide, use too much area, and
potentially lead to multi-moding. A line with Z0 > 80 Ω will be narrow, close to the maximum
Z0 that can be realized in a technology, have manufacturing tolerance issues, and have large
radiation. The range can be extended to 20–100 Ω but typically with degraded performance.
Also see Example 3.4 of [29] for a microstrip design guideline.
PARALLEL COUPLED-LINE FILTERS 101

Figure 3-30:
Step 8:
Ct1 = 2.70759 pF = Ct3 = Ct2 Z0t2 = 80 Ω Bandpass filter
Z0t1 = 67.7683 Ω = Z0t3 Z0t12 = 443.232 Ω = Z0t23 approximation.

Figure 3-31:
Step 9:
Bandpass
Ct1 = 2.70759 pF = Ct3 = Ct2 Z0t2 = 80 Ω filter
Z0t1 = 67.7683 Ω = Z0t3 Z0t12 = 443.232 Ω = Z0t23 approximation.

Figure 3-32: Inverter equivalence: (a)

resistively-terminated impedance inverter;
and (b) its equivalent capacitor network. Here
Zin = 1/Yin = 389.426 Ω, RL = 50 Ω, and
(a) (b) K = 139.540 Ω.
√
Figure 3-31 with input and output inverters of impedance 50 × 389.426 =
139.540 Ω.

3.4.1 Realization of the Input/Output Inverters

At this stage the design is as shown in Figure 3-31, where the inverters at
the input and output remain to be synthesized. There are many possible
realizations of the inverters. Focusing on just the output inverter with a load
RL , as shown in Figure 3-32(a), it will be shown here how this can be realized
using the capacitive network of Figure 3-32(b).
The input admittance of the inverter plus load resistor of Figure 3-32(a) is

RL
Yin = , (3.46)
K2

and the input admittance of the circuit in Figure 3-32(b) is

1
Yin = sCa + . (3.47)
1/(sCb ) + RL

Thus, equating the real parts of Equations (3.46) and (3.47),

RL ω 2 Cb2 RL
ℜ (Yin ) = 2 2 2 = 2, (3.48)
RL ω Cb + 1 K
102 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 3-33: The external inverters of the proto-

type shown in Figure 3-31 replaced by capaci-
tive networks. Ca = −1.06484 pF Cb = 1.22170 pF

C1 = Ca + Ct1 = 1.64276 pF = C3 Ct2 = 2.70759 pF Cb = 1.22170 pF

Z0t1 = 67.7683 Ω = Z0t3 Z0t2 = 80 Ω Z0t12 = 443.232 Ω = Z0t23

Figure 3-34: Step 10: The bandpass filter approximation combining the capacitive equivalent of
the inverters with the first and last capacitors in the circuit of Figure 3-31.

and equating the imaginary parts yields

 
Cb
ℑ (Yin ) = ω Ca + = 0. (3.49)
1 + (ωCb RL )2
1 −Cb
Thus Cb =  , Ca = 2
, (3.50)
2
ωK 1 − RL /K 2 1 + (ωCb RL )

and the inverters shown in Figure 3-31 can be replaced by the capacitive
networks shown in Figure 3-33. This capacitive network is not a general
replacement of an inverter. It only works when the inverter is terminated
in a resistor. Note that Ca is negative, and this is absorbed in the first and
last resonators so that the final lumped-distributed realization is as shown in
Figure 3-34. This is the electrical design of the bandpass filter in a form that
can be realized using the combline connection of coupled lines.

3.4.2 Implementation
The filter designed in the previous section can be implemented using
three parallel microstrip lines, as shown in Figure 3-35(a), where the three
capacitors of the electrical design in Figure 3-34 appear directly in Figure 3-
35(a). The double Pi connection of stubs shown in Figure 3-34, and again in
Figure 3-36(a), becomes the three coupled lines shown in Figure 3-36(b).

Derivation of Physical Dimensions

The task now is to determine the dimensions of the three parallel microstrip
lines. The dimensions to be determined are shown in Figure 3-36(b) and are
the length of the lines, L, the widths of the three lines, w1 , w2 , and w3 , and
the gaps s1 and s2 . The approach is to consider that the three parallel lines
form two coupled-line pairs with each pair of coupled lines implementing a
Pi network of stubs. The physical design procedure is outlined in Figure 3-36
with one of the lines shared. This, of course, is an approximation but a fairly
PARALLEL COUPLED-LINE FILTERS 103

accurate one. The physical synthesis of a pair of coupled lines is outlined in

Figure 3-37 where the parameters of the Pi model, Figure 3-37(a), lead to the
parameters of the transmission lines in the combline model, Figure 3-37(b).
This then leads to the parameters defining the pair of coupled lines, Figure
3-37(c), in particular the even and odd-mode characteristic impedances, Z0e
and Z0o and the system impedance Z0S .
Using the quantities in Figure 3-37(a) and rearranging Equations (3.14)–
(3.16), the parameters of the coupled-line model in Figure 3-37(b) are
obtained as follows:
 
Z012 Z011 Z022
n=1+ , Z02 = nZ012 and Z01 = n . (3.51)
Z011 Z011 + Z022 + Z012

For the left pair (and also for the right pair because of symmetry)) of coupled
lines in Figure 3-36(d), Z012 = Z0t12 = 443.232 Ω, Z011 = Z0t1 = 67.7683 Ω,

(a) (b)

(c) (d)

Figure 3-35: Physical layout of the combline bandpass filter designed in Section 3.4.1: (a) with
discrete capacitors; (b) with a microstrip coupled transmission line element (MCLIN); (c) with
an EM subcircuit block; and (d) the EM subcircuit block with layout to be simulated in an
EM simulator. Details: 6 µm gold metalization, 635 µm thick alumina substrate with εr = 10,
400 µm × 400 µm tantalum vias, and the EM enclosure has perfectly conducting walls with
XDIM = 22 mm, YDIM = 20 mm, and height = 5.635 mm.
104 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 3-36: Physical design of the third-order combline bandpass filter. The double Pi network
in (a) is partitioned into two Pi networks in (c) and (e). Similarly the three coupled microstrip
lines in (b) are partitioned into two pairs of coupled lines in (d) and (f). Then physical design
of the filter involves transforming the electrical design in (c) into the physical design in (d).
Similarly the circuit in (e) is transformed into the layout in (f). With a shared central stub, the
electrical circuit in (a) becomes the three coupled-line realization in (b).

Figure 3-37: Equivalent

circuits
for a combline section Z0e , Z0o , Z0S
(from Figure 3-13). (a) (b) (c)

Z022 = Z0t2 = 80 Ω and so4

Z012
n=1+ = 7.540, Z02 = 3342 Ω, and Z01 = 69.17 Ω. (3.52)
Z011

4 Note that the precision has been reduced. It was necessary to retain high precision during the
electrical synthesis but the physical dimensions do not need the same precision.
PARALLEL COUPLED-LINE FILTERS 105

The physical parameters of the coupled line pair in Figure 3-37(c) are
derived using the results in Section 5.9.6 of [29]. Thus

K = 1/n = 0.1326. (3.53)

From Equations (5.197) and (5.198) of [29] two estimates of the system
impedance, Z0S , are obtained:

Z0S,(3.54) = Z01 1 − K 2 = 68.56 Ω (3.54)
K2
Z0S,(3.55) = Z02 √ = 59.30 Ω. (3.55)
1 − K2
These two values of Z0S are different and this is because the Pi arrangement
of stubs is not symmetrical (i.e., in Figure 3-37(a) (Z011 = 67.7 Ω) �= Z
( 022 =
80 Ω)).5 The assumption behind the development of Equations (3.3) and (3.4)
are that the pair of coupled lines is symmetrical. The only reasonable choice
is to use the geometric mean of the two values. So

Z0S = Z0S,mean = Z0S,(3.54) Z0S,(3.55) = 63.77 Ω. (3.56)

Then, from Equation (5.202) of [29],


n−1 Z2 Z0e
Z0o = Z0S = 55.80 Ω, Z0e = 0S = 72.87 Ω, and = 1.306.
n+1 Z0o Z0o
(3.57)

The next step is to choose a substrate and realize the physical dimensions
of the coupled lines. It is important that there be dimensional stability
and that there be no multimoding. The best substrate for dimensional
stability is a hard substrate like alumina, sapphire, or glass. Soft substrates
like FR4 or teflon are too variable to be useful for filters. Alumina is
particularly attractive as it is very hard and dimensions are reproducible.
Alumina shrinks when it is processed but the amount of shrinkage is
well controlled and reproducible. So once a design has been physically
tuned, typically by drilling out part of the dielectric to change the effective
permittivity, filters can be reliably reproduced and require only a small
amount of adjustment. Alumina is polished to a well defined thickness
and with a relative permittivity of around 10 a 50-Ω microstrip line has a
width approximately equal to the thickness of the substrate. Having equal
dimensions is a good thing to have for manufacturability. The thickness of
the substrate must be decided on at this stage in design. A thicker substrate
is less likely to crack during handling however a thicker substrate is more
likely to support multimoding. Here a (h =) 635 µm-thick alumina substrate
with εr = 10 is chosen and calculations (not shown here) indicate that there
will not be multimoding at the filter’s operating frequencies.
Physical dimensions could be derived by iteratively solving for Z0e and
Z0o using the formulas in Section 5.6 of [29]. As an approximation, here the

5 Thus error is being introduced at this step. In reality the final manufactured filter will need
to be tuned anyway as the filter will require tolerances of about 0.1% or better and the
best manufactured dimensional tolerance is usually about 1%. Also material parameters are
usually not known to better than 1%. So this is an error that must be accepted.
106 STEER MICROWAVE AND RF DESIGN: MODULES

physical dimensions are derived from Table 5-3 of [29]. Table 5-3 of [29] is
for a system impedance, Z0S , of 50 Ω and not the 63.77 Ω found here. Thus
there will be an error, but it can be expected that there will be error in any
case and these can be resolved using optimization in a simulator. From Table
5-3 of [29], the initial physical dimensions of the pairs of lines are
u = 0.93, w = uh = 591 µm, thus with rounding w1 = w2 = w3 = 600 µm
(3.58)
g = 1.0, s = gh = 635 µm, thus with rounding s1 = s2 = 650 µm (3.59)
The rounding has been done as for EM simulation the lines will be laid out
on a regular grid and a 50 µm grid is reasonable.

Scaling of Line Width

At this stage the error made with the characteristic impedance of the
microstrip line should be considered. While any error can be corrected when
optimizing the final design, making some correction at this stage will reduce
the final adjustments needed. Three system impedances have been derived,
Z0S,(3.54) = 68.56 Ω, Z0S,(3.55) = 59.30 Ω, and the geometric mean Z0S,mean =
63.77 Ω. This divergence is one source of error. Another source of error is
that a table for a 50 Ω system impedance was used to determine the coupled-
line parameters. Note that the electrical design is exact and the tuning of the
physical design will attempt to match the response of the physical circuit to
that of the electrical design. However some initial corrections will reduce the
amount of tuning that needs to be done.
Scaling for the higher actual system impedance (63.77 Ω versus 50 Ω)
means using higher impedance microstrip lines (which will be narrower)
and higher coupling impedance (requiring a greater separation of the lines).
There are three possible characteristic impedance to scale to and a design
decision needs to be made. Should the lines be scaled from 50 Ω to 54.2 Ω,
68.6 Ω, or 61.9 Ω? The most conservative approach is to scale from 50 Ω
to 54.2 Ω, a factor of 1.08. Here only adjustment of the line widths will be
considered and two correction approaches will be described. The first is
based on the rule of thumb described in Example 3.4 of [29] which indicated
that the characteristic
√ impedance of the microstrip lines is approximately
proportional to 1/ w. So the line width should be reduced by a factor of
1.082 = 1.17 to shift the characteristic impedance. Thus the line width should
be reduced from 600 µm to about 600 µm/1.17 = 512 µm or 500 µm after
rounding.
A second approach to correction is based on Table 3-3 of [29]. For ε = 10
and using the more conservative system impedance (i.e. 55.8 Ω which is
closest to 50 Ω), a better choice for u is to replace u = 0.93 by u = 0.78
and, after rounding, w1 = w2 = w3 = 500 µm. This cannot be expected to
completely correct for the error, but it should be closer than the choice of
600 µm.
Scaling the system impedance should also require that s be increased but
there is not a simple guideline to follow and so this will be skipped. Another
reason for needing to increase s is that there is more coupling in our coupled
line system than that predicted by considering two pairs of coupled lines.
For example, there is coupling from lines 1 directly to line 3. To reduce the
coupling s will need to be increased, perhaps significantly. Another source
PARALLEL COUPLED-LINE FILTERS 107

Figure 3-38: Lumped-element bandpass filter in

a 50 Ω system. (Derived by multiplying the
impedances in Figure 3-21 by 50.)
Z0 = 50 Ω,
C13 = C33 = 27.107 pF,
L13 = L33 = 934.47 pH,
C23 = 288.54 fF, L23 = 87.787 nH.

Figure 3-39: Return

loss and insertion
loss for the filter
modeled using the
lumped-element
model.

of increased coupling is that the phase velocities of the even and odd modes
differ. This tends to increase coupling. The adjustment of s is left to tuning of
the physical design in an EM simulator.

Line Lengths
The stubs are one-eighth wavelength long at the center frequency of the filter.
Thus the coupled lines are also one-eighth wavelength long. From Table 5-
3 of [29], the effective relative permittivity of the even mode is εee = 7.24
and for the odd mode is εeo = 5.95. These are quite different so the best
that can be done is to use the geometric mean. Thus the effective relative
√
permittivity is εe = εee εeo√= 6.56. Thus at 1 GHz the line length L =
√
λ8 /8 = λ0 / εe = (30 cm)/(8 6.56) = 14.64 mm (14.65 mm with rounding).
The use of a single effective permittivity in determining the lengths of the
lines is perhaps the largest error and will result in the resonant frequency of
each of the resonators, consisting of a capacitor (either C1 , Ct2 , or C3 ) and the
λ/8-long line, being shifted from 1 GHz. The resonators can be re-centered
by adjusting the capacitors. Adjusting the capacitance values also corrects
for error in the widths of the lines since the main effect of an error in the
width of a line is an error in its characteristic impedance and hence the input
impedance of the resonator stub.

Microwave Circuit Simulation

Before simulating the combline filter the lumped-element bandpass filter will
be considered to provide a benchmark. The 50-Ω lumped-element bandpass
filter is shown in Figure 3-38. The responses of this filter is shown in Figures
3-39, 3-40, and 3-41. These responses provide a reference as the coupled-line
108 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 3-40: Wideband inser-

tion loss and return loss for
the filter modeled using the
lumped-element model.

Figure 3-41: Plot of S11 on a Smith

chart for the lumped-element band-
pass filter. The zeros of the S11 re-
sponse, and hence the poles of the
S21 response, are at 0.96, 1.00, and
1.04 GHz. The looping near the origin
is characteristic of Chebyshev filters.

filter design is optimized.

Figure 3-35(a) shows the combline filter to be modeled in a microwave
circuit simulator. In a microwave simulator the filter can be modeled using
a coupled microstrip element (known as the MCLIN element) or using EM
simulation of an actual layout. The first set of results that will be presented
uses the coupled microstrip element line model, the MCLIN element, and
the filter model is as shown in Figure 3-35(b). The results of the circuit
simulation are shown in Figure 3-42. The curves (a) are responses for a gap of
s = 650 µm. The passband is too wide and this is reduced by increasing s to
1150 µm yielding curves (b). The effect of loss is seen with S21 less than 0 dB
in the passband. The S21 response has a slight slope down in the passband
and the S11 response does not show the three zeros seen in the lumped-
element response, see Figure 3-39.
Plotting the S11 response of the MCLIN-based filter on a Smith chart, see
Figure 3-43, provides more insight into what is happening. This should be
contrasted to the S11 response of the lumped-element filter shown in Figure
PARALLEL COUPLED-LINE FILTERS 109

Figure 3-42: Insertion loss

and return loss for the filter
modeled using the MCLIN
element with a line length
of L = 14,650 µm and line
widths w1 = w2 = w3 =
500 µm. C1 = C3 = 1.6428 pF,
Ct2 = 2.7076 pF, and
Cb = 1.2217 pF. For curves
(a) s1 = s2 = 650 µm and for
curves (b) s1 = s2 = 1150 µm.

Figure 3-43: Plot of S11 on a Smith chart

for filter modeled using the MCLIN
element and gaps of s1 = s2 = 1150 µm
and a line length of L = 14,550 µm and
line widths w1 = w2 = w3 = 500 µm.
C1 = C3 = 1.6428 pF, Ct2 = 2.7076 pF,
and Cb = 1.2217 pF.

3-41. Recall that the lumped-element filter is the ideal electrical design.
The looping of S11 at the origin in Figure 3-41 indicates that the ideal
electrical response has three zeros as the locus of S11 passes through the
origin three times. This behavior corresponds to the poles of the transmission
response but these poles cannot be seen in the S21 response. Consequently in
optimizing a filter the designer focuses on the S11 response.
Again consider the S11 responses of the two filters plotted on Smith charts
(i.e. Figures 3-41 and 3-43). The S11 response of the lumped-element filter
has two small loops corresponding to peaks in the in-band |S11 | response
seen in the rectangular plot of Figure 3-39. In contrast, the locus of S11 of
the MCLIN-based filter does not go through the origin, see Figure 3-43, and
there is one loop although the hint of a second loop is seen at 1.06 GHz.
The performance is close to that required but optimization is necessary. In
manually optimizing, also called tuning, the design of the MCLIN-based
filter, both the rectangular and Smith chart plots should be overlaid with
the lumped-element response. By adjusting C1 , Ct2 , and C3 (by 6%, 16%,
110 STEER MICROWAVE AND RF DESIGN: MODULES

(a) Narrowband response

Figure 3-44: Insertion

loss and return loss
for the filter modeled
using the MCLIN el-
ement and gaps of
s1 = s2 = 1150 µm
and a line length
of L = 14,550 µm
and line widths
w1 = w2 = w3 = 500 µm.
C1 = C3 = 1.9076 pF,
Ct2 = 2.8748 pF, and
Cb = 1.2217 pF. (b) Wideband response

and 6% respectively), the resonant frequencies of the resonators are centered

at 1 GHz and the filter response is also centered at 1 GHz6 . The result of
this tuning is shown in Figures 3-44 and 3-45 indicating that the required
performance has been obtained.
In Figure 3-44(a) the MCLIN-based filter response is compared to the
lumped-element response. The bandwidths of the MCLIN and lumped-
element implementations are closely matched. The MCLIN implementation
has the steep skirts resulting from the Chebyshev prototype. The Smith chart
plot of S11 of the MCLIN-based filter is shown in Figure 3-45. The loops seen
with the lumped-element response, in Figure 3-41, are seen but now there is
an additional rotation with respect to frequency. This is because the MCLIN-
based filter has additional transmission-line delays and hence frequency-
dependent phase shift. Also the S11 response in Figure 3-45 is rotated by
180◦ from the S11 locus for the lumped-element filter (see Figure 3-41). This is
because of the additional inverter at the input of the distributed filter design.
Some gross features are seen in the combline response that are not seen
in the ideal electrical response. Referring to Figure 3-44(a), these include the
additional zero seen in the S21 response at 1.2 GHz, the steeper skirt above

6
√
Approximately, the resonant frequency of a resonator is inversely proportional to C. So
increasing the capacitances will reduce the center frequency of the filter.
PARALLEL COUPLED-LINE FILTERS 111

Figure 3-45: Plot of S11 on a Smith chart for filter

modeled using the MCLIN element and gaps of
s1 = s2 = 1150 µm and a line length of L =
14,550 µm and line widths w1 = w2 = w3 =
500 µm. C1 = C3 = 1.9076 pF, C2 = 2.8748 pF,
and Cb = 1.2217 pF.

the 1 GHz passband, and the less-steep skirt below the passband. These are
related. These features are common to PCL designs that use a combline. The
origin of this behavior is the additional transmission path directly from the
first line to the last line of the filter, the 1-3 coupling. During design the only
transmission path considered was from line 1 to line 2, and then from line 2
to line 3, the 1-2-3 coupling. Apparently, at 1.2 GHz the 1-3 and 1-2-3 signals
have the same magnitude but opposite phase and so cancel producing the
S21 zero. Above the passband the signals on the two paths destructively
interfere causing the steeper skirt. Below the passband the 1-3 and 1-2-3
signals constructively interfere and reduce the steepness of the filter skirt.
The additional zero can be exploited to significantly reduce the level of
a particular signal such as a local oscillator or a signal in a neighboring
communication band. If the reduced filter skirt below the passband is not
acceptable, then another arrangement of coupled lines should be used.
Figure 3-44(b) is a wideband response of the MCLIN-based filter and
shows the spurious passbands that occur with most transmission-line based
networks. In this case the properties of the lines are the same at λ/8, 5λ/8,
and 9λ/8. So, by just considering the transmission lines alone it would be
expected that there would be spurious passbands at 5 GHz and 9 GHz. The
spurious passbands are lower because the impedances of the capacitances
reduce at the higher frequencies, resulting in the passbands at four times and
seven times the center frequency of the filter. In practice, a simple lowpass
filter eliminates the spurious passbands. Often parasitics in the system do
this without specific circuitry. Also, appropriate choice of matching networks
can provide adequate elimination of the spurious passbands.
Before finalizing the filter layout, a more accurate EM simulation is
required instead of using the MCLIN element. The EM simulation captures
more subtle effects than can be included in the MCLIN model. The layout of
the parallel coupled lines is shown in Figure 3-35(d) and the connection to
incorporate these in a complete circuit simulation is shown in Figure 3-35(c).
The results of the EM-based simulation are shown in Figures 3-46 and 3-47. It
can be seen that the bandwidth of the filter reduces. Additional optimization
would result in the EM-based filter response more closely matching the ideal
electrical response.
112 STEER MICROWAVE AND RF DESIGN: MODULES

(a) Narrowband response

Figure 3-46: Return

and insertion loss
for the filter mod-
eled using EM mod-
eling and gaps of
s1 = s2 = 1150 µm
and a line length
of L = 14,550 µm
and line widths
w1 = w2 = w3 =
500 µm. C1 =
C3 = 1.9076 pF,
Ct2 = 2.8748 pF,
and Cb = 1.2217 pF. (b) Wideband response

3.4.3 Alternative Combline Filter Layouts

Up to a few gigahertz surface-mount capacitors can be used in the imple-
mentation of the capacitively-loaded stubs. Above that other arrangements
should be considered. There are several layout variations for the combline
filter. The main variations relate to replacing the lumped-element capacitors
by gap capacitors and alternative realizations of the input and output invert-
ers. Several variations are shown in Figure 3-48. The layout in Figure 3-48(a)
implements the capacitors using gap capacitors. At 1 GHz it is difficult to
realize the required capacitance using gap capacitors, then the interdigitated
capacitors shown in Figure 3-48(b) can be used to obtain higher capacitance
values. Alternatively a coupled-line section can be used to implement the
input and output inverters as shown in Figure 3-48(c).
PARALLEL COUPLED-LINE FILTERS 113

Figure 3-47: Plot of S11 on a Smith

chart for filter modeled using EM
modeling and gaps of s1 = s2 =
1150 µm, line length L = 14,550 µm,
line widths w1 = w3 = w2 = 500 µm.
C1 = C3 = 1.9076 pF, C2 = 2.8748 pF,
and Cb = 1.2217 pF.

(a) (b)

(c)

Figure 3-48: Physical layouts of the combline bandpass filter: (a) with gap capacitors having
gaps of g1 , g1 , . . ., g5 ,; (b) with interdigitated capacitors to obtain higher capacitance; and (c)
with coupled lines (with separations s4 and s5 ) realizing the input and output inverters.
114 STEER MICROWAVE AND RF DESIGN: MODULES

3.5 Parallel Coupled-Line Filters in an Inhomogeneous

Medium
A complexity arises when a parallel coupled-line filter is realized in an
inhomogeneous transmission medium due to the inequality of the even- and
odd-mode phase velocities. This effect can be investigated by utilizing the
full ABCD matrix (see Equation (5.184) of [29].) of each PCL section in a filter
for a known even- to odd-mode velocity ratio. The inhomogeneous medium
can result in additional spurious passbands centered approximately at even
multiples of f0 [4, 32–45]. This provides all the more reason to perform an
accurate full EM simulation following filter synthesis.

3.6 Summary
Filters using parallel coupled-line (PCL) segments are an important class of
microwave filters. There are various ways a pair of coupled lines, a four-
port, can be configured using shorts and opens to realize a two-port network.
Many of these PCL configurations have desirable frequency selectivity
characteristics. That is, they inherently have the bandpass, lowpass,
highpass, or bandstop characteristics desired of a filter. In particular, the
various configurations often have very sharp responses. One drawback,
however, is that they have spurious passbands that derive from a line and
its counterpart that is one-half wavelength longer having the same input
reflection coefficient. These spurious passbands can often be pushed up in
frequency by setting, in the case of a bandpass filter, the resonant frequency
of the transmission line resonators above the center frequency of the filter
being developed.
The synthesis of PCL filters follows the general microwave design
philosophy of identifying a transmission line structure that inherently has
the desired response. The combline filter, a type of PCL filter, considered
in this chapter, for example, exploits the bandpass properties of coupled
microstrip transmission lines that are all shorted at the same end. Once a
filter with appropriate topology has been designed, the physical realization
is optimized in a circuit simulator to account for parasitics and higher-order
EM effects.
While this chapter specifically addressed the design of PCL filters, the
principles can be used with all distributed filter design. Bandpass filters
are the most important microwave filter type and it was seen that a
bandpass filter comprises coupled resonators. So other physical structures
that present coupled resonators can also become components of a bandpass
filter. The synthesis approach provides design insight and exploitation of all
parameters of a transfer function. This leads to optimum network topologies.
Synthesis can be time consuming and specialized, but it is the only way to
develop filters with optimum performance.
PARALLEL COUPLED-LINE FILTERS 115

3.7 References
[1] E. Cristal and S. Frankel, “Hairpin-line [18] E. Cristal, “Tapped-line coupled transmis-
and hybrid hairpin-line/half-wave parallel- sion lines with applications to interdigital
coupled-line filters,” IEEE Trans. on Mi- and combline filters,” IEEE Trans. on Mi-
crowave Theory and Techniques, vol. 20, no. 11, crowave Theory and Techniques, vol. 23, no. 12,
pp. 719–728, Nov. 1972. pp. 1007–1012, Dec. 1975.
[2] B. Minnis, “A capacitively coupled bpf de- [19] W. Fathelbab and M. Steer, “Parallel-coupled
sign using a suspended stripline,” Microwave line filters with enhanced stopband perfor-
Journal, vol. 26, pp. 71–74, Nov. 1993. mance,” IEEE Trans. on Microwave Theory and
[3] U. H. Gysel, “New theory and design for Techniques, vol. 53, no. 12, pp. 3774–3781, Dec.
hairpin-line filters,” IEEE Trans. on Microwave 2005.
Theory and Techniques, vol. 22, no. 5, pp. 523– [20] M. C. Horton and R. Wenzel, “General theory
531, May 1974. and design of optimum quarter-wave tem fil-
[4] W. Fathelbab and M. Steer, “Parallel-coupled ters,” IEEE Trans. on Microwave Theory and
line filters with enhanced stopband perfor- Techniques, vol. 13, no. 3, pp. 316–327, May
mance,” IEEE Trans. on Microwave Theory and 1965.
Techniques, vol. 53, no. 12, pp. 3774–3781, Dec. [21] A. Feldshtein and L. Yavich, Synthesis of Mi-
2005. crowave Two-Ports and Four-Ports. Syaz, 1965.
[5] R. Wenzel, “Synthesis of combline and capac- [22] A. Riddle, “Reviewing the basics of sus-
itively loaded interdigital bandpass filters pended stripline,” Microwave Journal, pp. 82–
of arbitrary bandwidth,” IEEE Trans. on Mi- 90, Oct. 2002.
crowave Theory and Techniques, vol. 19, no. 8, [23] ——, “High performance parallel coupled
pp. 678–686, Aug. 1971. microstrip filters,” in 1988 IEEE MTT-S In-
[6] G. Matthaei, L. Young, and E. Jones, Mi- ternational Microwave Symposium Digest, May
crowave Filters, Impedance-Matching Networks 1988, pp. 427–430.
and Coupling Structures. McGraw-Hill, 1965, [24] L. Maloratsky, “Improve bpf performance
reprinted in 1980, Artech House. with wiggly coupled lines,” Microwave & RF,
[7] J.-S. Hong and M. Lancaster, Microstrip Filters vol. 7, pp. 53–62, Apr. 2002.
for RF/Microwave Applications. John Wiley & [25] R. Rhea, HF Filter Design and Computer Simu-
Sons, 2001. lation. Noble Publishing, 1994.
[8] R. Mongia, I. Bahl, and P. Bhartia, RF and Mi- [26] G. Matthaei, “Comb-line band-pass filter of
crowave Coupled-line Circuits. Artech House, narrow or moderate bandwidth,” Microwave
1999. Journal, vol. 7, pp. 428–439, Aug. 1964.
[9] I. Hunter, Theory and Design of Microwave Fil- [27] R. Wenzel, “Application of exact synthesis
ters. IEE Press, 2001. methods to multichannel filter design,” IEEE
[10] S. B. Cohn, “Parallel-coupled transmission- Trans. on Microwave Theory and Techniques,
line-resonator filters,” IRE Trans. on Mi- vol. 13, no. 1, pp. 5–15, Jan. 1965.
crowave Theory and Techniques, vol. 6, no. 2, [28] W. Fathelbab and M. Steer, “Design of band-
pp. 223–231, Apr. 1958. stop filters utilising circuit prototypes,” Mi-
[11] A. Zverev, Foundations of Oscillator Circuit De- crowaves, Antennas & Propagation, IET, vol. 1,
sign. John Wiley & Sons, 1967. no. 2, pp. 523–526, 2007.
[12] L. Young, Microstrip Filters Using Parallel Cou- [29] M. Steer, Microwave and RF Design, Transmis-
pled Lines. Artech House, 1972. sion Lines, 3rd ed. North Carolina State Uni-
[13] J. Rhodes, Theory and Design of Microwave Fil- versity, 2019.
ters. Institute of Electrical Engineers, 2001. [30] ——, Microwave and RF Design, Networks,
[14] B. Minnis, “Printed circuit coupled-line fil- 3rd ed. North Carolina State University,
ters for bandwidths up to and greater than an 2019.
octave,” IEEE Trans. on Microwave Theory and [31] ——, Microwave and RF Design, Radio Systems,
Techniques, vol. 29, no. 3, pp. 215–222, Mar. 3rd ed. North Carolina State University,
1981. 2019.
[15] H. Ozaki and J. Ishii, “Synthesis of a class of [32] I. Bahl, “Capacitively compensated high per-
strip-line filters,” IRE Trans. on Circuit Theory, formance parallel coupled microstrip filters,”
vol. 5, no. 2, pp. 104–109, Jun. 1958. in 1989 IEEE MTT-S Int. Microwave Symp.
[16] B. Minnis, Designing Microwave Circuits by Dig., Jun. 1989, pp. 679–682.
Exact Synthesis. Artech House, 1996. [33] S.-M. Wang, C.-H. Chi, M.-Y. Hsieh, and
[17] H. Carlin and P. Civalleri, Wideband Circuit C.-Y. Chang, “Miniaturized spurious pass-
Design. CRC Press, 1998. band suppression microstrip filter using me-
116 STEER MICROWAVE AND RF DESIGN: MODULES

andered parallel coupled lines,” IEEE Trans. on Microwave Theory and Techniques, vol. 49,
on Microwave Theory and Techniques, vol. 53, no. 9, pp. 1593–1598, Sep. 2001.
no. 2, pp. 747–753, Feb. 2005. [40] T. Lopetegi, M. Laso, F. Falcone, F. Mar-
[34] M. Velazquez-Ahumada, J. Martel, and tin, J. Bonache, J. Garcia, L. Perez-Cuevas,
F. Medina, “Parallel coupled microstrip fil- M. Sorolla, and M. Guglielmi, “Microstrip
ters with ground-plane aperture for spuri- ”wiggly-line” bandpass filters with mul-
ous band suppression and enhanced cou- tispurious rejection,” IEEE Microwave and
pling,” IEEE Trans. on Microwave Theory and Wireless Components Letters, vol. 14, no. 11,
Techniques, vol. 52, no. 3, pp. 1082–1086, Mar. pp. 531–533, Nov. 2004.
2004. [41] B. Kim, J. Lee, and M. Song, “An im-
[35] ——, “Parallel coupled microstrip filters plementation of harmonic-suppression mi-
with floating ground-plane conductor for crostrip filters with periodic grooves,” IEEE
spurious-band suppression,” IEEE Trans. on Microwave and Wireless Components Letters,
Microwave Theory and Techniques, vol. 53, vol. 14, no. 9, pp. 413–415, Sep. 2004.
no. 5, pp. 1823–1828, May 2005. [42] J.-T. Kuo, W.-H. Hsu, and W.-T. Huang, “Par-
[36] L. Zhu, H. Bu, and K. Wu, “Broadband and allel coupled microstrip filters with suppres-
compact multi-pole microstrip bandpass fil- sion of harmonic response,” IEEE Microwave
ters using ground plane aperture technique,” and Wireless Components Letters, vol. 12,
IEE Proc., Microwaves, Antennas and Propaga- no. 10, pp. 383–385, Oct. 2002.
tion, vol. 149, no. 1, pp. 71–77, Feb. 2002. [43] S.-F. Chang, Y.-H. Jeng, and J.-L. Chen,
[37] J.-T. Kuo, M. Jiang, and H.-J. Chang, “Design “Tapped wiggly-coupled technique applied
of parallel-coupled microstrip filters with to microstrip bandpass filters for multi-
suppression of spurious resonances using octave spurious suppression,” Electronics Let-
substrate suspension,” IEEE Trans. on Mi- ters, vol. 40, no. 1, pp. 46–47, Jan. 2004.
crowave Theory and Techniques, vol. 52, no. 1, [44] J. Garcia-Garcia, F. Martin, F. Falcone,
pp. 83–89, Jan. 2004. J. Bonache, I. Gil, T. Lopetegi, M. Laso,
[38] J.-T. Kuo and M. Jiang, “Enhanced microstrip M. Sorolla, and R. Marques, “Spurious pass-
filter design with a uniform dielectric over- band suppression in microstrip coupled line
lay for suppressing the second harmonic re- band pass filters by means of split ring res-
sponse,” IEEE Microwave and Wireless Compo- onators,” IEEE Microwave and Wireless Com-
nents Letters, vol. 14, no. 9, pp. 419–421, Sep. ponents Letters, vol. 14, no. 9, pp. 416–418,
2004. Sep. 2004.
[39] T. Lopetegi, M. Laso, J. Hernandez, M. Ba- [45] M. Makimoto and S. Yamashita, “Band-
caicoa, D. Benito, M. Garde, M. Sorolla, pass filters using parallel coupled stripline
and M. Guglielmi, “New microstrip stepped impedance resonators,” IEEE Trans.
ldquo;wiggly-line rdquo; filters with spu- on Microwave Theory and Techniques, vol. 28,
rious passband suppression,” IEEE Trans. no. 12, pp. 1413–1417, Dec. 1980.

3.8 Exercises
1. A parallel coupled-line section is shown in Fig- is λ/4 long at the center frequency, f0 , of the sec-
ure 3-11. The even-mode impedance of the cou- tion and Port 4 is terminated in 50 Ω. [Hint: Con-
pled line is 60 Ω and the odd-mode impedance sider the Z01 and Z02 lines. They are λ/4 long at
is 40 Ω. The PCL section is λ/4 long at the center f0 . Perhaps more information has been provided
frequency of the section. than you need.]
(a) What is the system impedance Z0S ? (a) What is the input impedance at f0 looking
(b) What is Z01 ? into Port 2?
(c) What is Z02 ? (b) What is the input impedance at 3f0 looking
(d) What is n? into Port 2?
(e) What is the input impedance at Port 2 if Port (c) What is the input impedance at DC looking
4 is terminated in 50 Ω? Calculate this at the into Port 2?
frequency at which the section is λ/4 long.
3. An interdigital coupled-line section of a fil-
2. A parallel coupled-line section of a filter ter is shown in Figure 3-12. The even-mode
is shown in Figure 3-11. The even-mode impedance of the coupled line is 65 Ω and the
impedance of the coupled line is 65 Ω and the odd-mode impedance is 45 Ω. The PCL section
odd-mode impedance is 35 Ω. The PCL section is λ/4 long at the center frequency of the filter.
PARALLEL COUPLED-LINE FILTERS 117

(a) What is the system impedance Z0S ? 7. Draw and derive the values of the lowpass fil-
(b) What is Z01 ? ter prototype of a third-order Butterworth low-
(c) What is Z02 ? pass prototype filter with a corner frequency of
(d) What is n? 1 rad/s and in a 1 Ω system. Show and describe
(e) What is the input impedance at Port 2 if Port your working. [Parallels Step 1 in Section 3.4.]
4 is terminated in 50 Ω? Calculate this at the 8. The lowpass prototype of a third-order Butter-
center frequency of the filter. worth filter with a corner frequency of 1 rad/s
4. An interdigital coupled-line section of a fil- and in a 1 Ω system is shown in Figure 3-
ter is shown in Figure 3-12. The even-mode 49(a), where C1 = 1 F, L2 = 2 H, and C3 =
impedance of the coupled line is 65 Ω and the 1 F. Transform this circuit to a lumped-element
odd-mode impedance is 35 Ω. The PCL section bandpass filter with corner frequencies of f1 =
is λ/4 long at the center frequency, f0 , of the fil- 878.42 MHz and f2 = 1.1384 GHz, and in a 50 Ω
ter and Port 4 is terminated in 50 Ω. [Hint: Con- system. Draw and derive the values of the band-
sider the Z01 and Z02 lines. They are λ/4 long at pass filter.
f0 . Perhaps more information has been provided 9. The lowpass prototype of a third-order Butter-
than you need.] worth filter with a corner frequency of 1 rad/s
(a) What is the input impedance at f0 looking and in a 1 Ω system is shown in Figure 3-49(a),
into Port 2? where C1 = 1 F, L2 = 2 H, and C3 = 1 F. Re-
(b) What is the input impedance at 3f0 looking place the series element using a shunt element
into Port 2? and inverters. This should have electrical prop-
(c) What is the input impedance at 2f0 looking erties identical to those of the circuit in Figure
into Port 2? 3-49(a). [Parallels Step 2 in Section 3.4.]
10. The lowpass prototype of a third-order Butter-
5. A combline section of a filter is shown in Fig-
worth filter with a corner frequency of 1 rad/s
ure 3-13. The even-mode impedance of the cou-
and in a 1 Ω system is shown in Figure 3-49(a),
pled line is 60 Ω and the odd-mode impedance
where C1 = 1 F, L2 = 2 H, and C3 = 1 F. Replace
is 40 Ω. The PCL section is λ/4 long at the center
the series element using a shunt element and
frequency of the filter.
inverter(s) and draw and derive the values of
(a) What is the system impedance Z0S ? the new prototype now with a corner frequency
(b) What is Z01 ? of 1 Hz. This should have electrical properties
(c) What is Z02 ? identical to those of the circuit in Figure 3-49(a).
(d) What is n? [Parallels Step 2 in Section 3.4.]
(e) What is the input impedance at Port 1 if Port
11. Consider eliminating the transformer in Figure
2 is terminated in 50 Ω?
3-49(b).
6. A combline section of a filter is shown in Fig- (a) What effect does this have on the properties
ure 3-13. The even-mode impedance of the cou- of the lowpass filter?
pled line is 55 Ω and the odd-mode impedance (a) Draw and derive the values of the new pro-
is 35 Ω. The PCL section is λ/4 long at the center totype without the transformer.
frequency, f0 , of the filter and Port 2 is termi- 12. Derive the bandpass filter prototype of a third-
nated in 50 Ω. [Hint: Consider the Z01 and Z02 order bandpass filter with corner frequencies of
lines. They are λ/4 long at f0 . Perhaps more in- f1 = 878.42 MHz and f2 = 1.1384 GHz. Base
formation has been provided than you need.] this on the 1 Ω third-order Butterworth lowpass
(a) What is the input impedance at f0 looking prototype shown in Figure 3-49(c) with C1 =
into Port 1? C3 = 1 F and C2 = 2 F.
(b) What is the input impedance at 3f0 looking (a) What is the center frequency of the bandpass
into Port 1? filter?
(c) What is the input impedance at 2f0 looking (b) What is the fractional bandwidth?
into Port 1? (c) First consider a 1 Ω system (i.e., the source
Exercises 7 to 30 cover the design of a third-order and load resistances are 1 Ω). Use two 1 Ω
Butterworth combline filter. Most exercises begin impedance inverters, but use no stubs. Draw
from the solution of the previous exercise. Generally and derive the values of the new prototype.
your answer for each exercise should correspond to [Parallels Step 3 in Section 3.4.]
the starting point provided for the next step in the (d) Transform the prototype developed in (a) to
design. You must show your detailed working. a 50 Ω system (i.e., the source and load resis-
118 STEER MICROWAVE AND RF DESIGN: MODULES

tances are 50 Ω). Draw and derive the val- resonator and the derivative of the admittance
ues of the new prototype. [Parallels Step 4 in must be the same at f0 for both resonators. Draw
Section 3.4.] and derive the values of the new resonator. [Par-
13. Consider a shorted stub that is resonant at fre- allels Step 5 in Section 3.4.]
quency fr . 19. The prototype of a 50 Ω third-order Butterworth
(a) What is the input impedance of the stub at bandpass filter is shown in Figure 3-49(e). The
resonance (this should be taken as the first filter has a center frequency of f0 = 1 GHz,
resonance)? where C1♯ = C3♯ = 12.2427 pF, L♯1 = L♯3 =
(b) In terms of wavelengths, how long is the 2.06901 nH, C2♯ = 24.4854 pF, and L♯2 =
stub at fr ? 1.03451 nH. Replace each of the three lumped-
(c) In terms of wavelengths, how long is the element resonators by a resonator consisting of
stub at 12 fr ? a single capacitor and a shorted stub. Each of the
new resonators must have the same admittance
14. Consider an open-circuited stub that is resonant
at frequency fr . at f0 as the original resonator it replaced and
the derivative with respect to frequency of the
(a) What is the input impedance of the stub at
admittances of each of the original and replace-
resonance (this should be taken as the first
ment resonators must be the same at f0 . Draw
resonance)?
and derive the values of the new filter prototype.
(b) In terms of wavelengths, how long is the
[Parallels Step 5 in Section 3.4.]
stub at fr ?
(c) In terms of wavelengths, how long is the 20. A shorted stub has a characteristic impedance
stub at 12 fr ? of Z01 = 75 Ω and is resonant at the frequency
fr = 2f0 .
15. Consider a capacitor C.
(a) What is the input impedance of the stub at
(a) What is the admittance with respect to ra-
fr ?
dian frequency of the capacitor at frequency
(b) What is the input impedance of the stub at
f0 = ω0 /2π?
f0 ?
(b) What is the derivative of the admittance
of the capacitor with respect to radian fre- 21. The prototype of a 50 Ω third-order Butterworth
quency of the capacitor of the resonator at bandpass filter is shown in Figure 3-49(f). The
f0 ? filter has a center frequency of f0 = 1 GHz,
16. A resonator comprising a capacitor C in parallel where C1′′ = C3′′ = 9.52443 pF, Z01 = Z03 =
with an inductor L is resonant at a frequency f0 . 16.7102 Ω, C2′′ = 19.0489 pF, Z02 = 8.35509 Ω.
The resonant frequency of each of the stubs is
(a) What is the admittance of the resonator at
fr = 2f0 . [Parallels Step 6 in Section 3.4.]
f0 ?
(b) What is the derivative with respect to radian (a) What is the input impedance at f0 of the
frequency of the admittance of the resonator stub with characteristic impedance Z01 ?
at f0 ? (b) Transform the prototype so that each stub
has the characteristic impedance Z01 .
17. Consider a shorted stub that is resonant at fre-
quency fr = 2f0 . 22. An impedance inverter has a characteristic
impedance of 50 Ω. Develop the lumped-
(a) In terms of wavelengths, how long is the
element equivalent circuit of the inverter. The
stub at fr ?
equivalent circuit should have three lumped
(b) In terms of wavelengths, how long is the
impedances. Draw and derive the values of the
stub at f0 ?
equivalent circuit.
(c) What is the admittance of the stub at f0 ?
(d) What is the derivative with respect to radian 23. An impedance inverter has a characteristic
frequency of the admittance of the stub at impedance of 60 Ω. Develop the lumped-
f0 ? element equivalent circuit of the inverter at
18. A lumped-element resonator consists of a par- 1 GHz. The equivalent circuit should have three
♯
allel capacitor C1 = 12.2427 pF and inductor lumped elements. Draw and derive the values of
♯
L1 = 2.06901 nH. Derive an equivalent res- the equivalent circuit with inductor and capaci-
onator comprising a capacitor in parallel with tor values.
a shorted stub that is resonant at the frequency 24. The prototype of a 50 Ω third-order Butterworth
fr = 2f0 . The new resonator must have the bandpass filter is shown in Figure 3-49(g). The
same admittance at f0 = 1 GHz as the original filter has a center frequency of f0 = 1 GHz,
PARALLEL COUPLED-LINE FILTERS 119

where C1′′ = C2′′′ = C3′′ = 9.52443 pF, Z01 = where Cb = 2.57780 pF, C1 = C3 = 2.21562 pF,
′ ′′
Z02 = Z03 = 16.7102 Ω, and Zy = 70.7107 Ω. Z0t1 = Z0t3 = 55.2444 Ω, Z0t2 = 80 Ω, Z0t12 =
The resonant frequency of the stubs is fr = Z0t23 = 178.5275 Ω, Ct2 = 3.77241 pF. Sketch
2f0 . Draw and derive the values of the filter the physical layout of the circuit assuming that
prototype with the inverters replaced by short- lumped-element capacitors will be used. You
circuited stubs resonant at fr . [Parallels Step 7 need not develop the dimensions of the mi-
in Section 3.4.] crostrip lines. [Parallels Step 9 in Section 3.4.]
25. The prototype of a 50 Ω third-order Butterworth 29. Design a third-order maximally flat bandpass
bandpass filter is shown in Figure 3-50(h). The filter prototype in a 50 Ω system centered at 1
filter has a center frequency of f0 = 1 GHz, GHz with a 10% bandwidth. The lowpass pro-
where C1′′ = C3′′ = 9.52443 pF, Z01 ′
= Z03′
= totype of a third-order maximally flat filter is
21.8811 Ω, C2′′′ = 9.52443 pF, Z02 ′′
= 31.6862 Ω, shown in Figure 2-10. The problem will parallel
and Z012 = Z023 = 70.7107 Ω. The resonant fre- the development in Section 3.4 and the end re-
quency of the stubs is fr = 2f0 . The characteris- sult of this development will be a bandpass pro-
tic impedances Z01 ′
, Z02′′ ′
, and Z03 are too low to totype filter with the form of that in Figure 3-29.
be realized in microstrip and so need to be scaled Note that there will be differences as the filter is
to a more reasonable microstrip impedance (say a different type.
between 30 Ω and 80 Ω). Scale the middle stub, (a) Convert the prototype lowpass filter to a
′′
Z02 to 80 Ω. This will shift the source and load lowpass filter with inverters and capacitors
impedances away from 50 Ω, but this can be ac- only; that is, remove the series inductors.
commodated in a latter step. [Parallels Step 8 in (b) Scale the filter to take the corner frequency
Section 3.4.] from 1 rad/s to 1 GHz.
(c) Transform the lowpass filter into a bandpass
26. The prototype of a third-order Butterworth
filter. That is, replace each shunt capacitor by
bandpass filter is shown in Figure 3-50(i). The
a parallel LC network. This step will estab-
filter has a center frequency of f0 = 1 GHz,
lish the bandwidth of the filter.
where RS = RL = 126.238 Ω, Ct1 = Ct2 =
(d) Transform the system impedance of the fil-
Ct3 = 3.77241 pF, Z0t1 = Z0t3 = 55.2444 Ω,
ter from 1 to 50 Ω.
Z0t2 = 80 Ω, and Z0t12 = Z0t23 = 178.528 Ω.
(e) Replace the parallel LC circuits by short-
The resonant frequency of the stubs is fr = 2f0 .
circuited stubs in parallel with lumped ca-
Incorporate an inverter at the input and at the
pacitors. (The circuit will now be in a form
output of the filter so that it interfaces with 50 Ω
similar to that in Figure 3-25.)
source and load impedances. Draw and derive
(f) For each inverter, derive the three-lumped-
the values of the new prototype. [Parallels Step
element equivalent circuit as in Figure 2-20.
9 in Section 3.4.]
Do not update the filter prototype yet, but
27. The prototype of a third-order Butterworth instead draw and label the lumped-element
bandpass filter is shown in Figure 3-50(j). equivalent circuits of each inverter.
The filter has a center frequency of f0 = 30. A two-port consisting of a three shorted stubs
1 GHz, where the impedance inverters have in a Pi structure with the shunt stubs having a
impedances Z1 = Z2 = 79.4475 Ω, Ct1 = Ct2 = characteristic impedance of 55.2444 Ω and the
′′
Ct3 = 3.77241 pF, Z0t2 = 80 Ω, Z0t1 = Z0t3 = series stub has a characteristic impedance of
55.244 Ω, and Z0t12 = Z0t23 = 178.5275 Ω. The 178.5275 Ω. The operating frequency is f0 =
resonant frequency of the stubs is fr = 2f0 . In- 1 GHz and the resonant frequency of the stubs
corporate an inverter at the input and and at the is fr = 2f0 . The substrate has a thickness of
output of the filter so that it interfaces with 50 Ω 500 µm and a relative permittivity εr of 10.
source and load impedances. Draw and derive
(a) Draw the inverter.
the values of the new prototype. Realize that
(b) Draw the pair of coupled lines in combline
each of the impedance inverters using a two-
configuration and draw two equivalent cir-
capacitor network noting that one side of the in-
cuit models of the combline. Note that the
verters is a resistance. Note that one of the ca-
coupled lines are a λ/8 long at f0 . Calculate
pacitor values in each network will be negative.
the element values in your equivalent mod-
[Parallels Step 9 in Section 3.4.]
els.
28. The prototype of a third-order Butterworth (c) Calculate two values for the system
bandpass filter is shown in Figure 3-50(k). The impedance. Take the system impedance as
filter has a center frequency of f0 = 1 GHz the geometric mean of these two values.
120 STEER MICROWAVE AND RF DESIGN: MODULES

(d) Calculate the odd-mode impedance of the the geometric mean of the even- and odd-
coupled lines. (Use 50 Ω if you were not able mode values.
to solve Part (c).) 31. The prototype of a third-order Butterworth
(e) Calculate the even-mode impedance of the bandpass filter is shown in Figure 3-50(k). The
coupled lines. (Use 50 Ω if you were not able filter has a center frequency of f0 = 1 GHz,
to solve Part (c).) where Cb = 2.57780 pF, C1 = C3 = 2.21562 pF,
(f) Calculate the width and separation of the Z0t1 = Z0t3 = 55.2444 Ω, Z0t2 ′′
= 80 Ω, Z0t12 =
coupled lines. Use may want to use Fig- Z0t23 = 178.5275 Ω, Ct2 = 3.77241 pF. Sketch
ures 5-10 to 5-13 and/or Tables 5-2 and 5-3 the physical layout of the circuit assuming that
of [29]. Also, you may need to interpolate. lumped-element capacitors will be used. Calcu-
Use the tables or figures even if your system late the widths and lengths of the microstrip
impedance is not 50 Ω. lines if the thickness of the microstrip substrate
(g) What is the effective permittivity of the even is 500 µm and the relative permittivity of the
mode? substrate is 10. Use can use, with some error,
(h) What is the effective permittivity of the odd Figures 5-10–5-13 and/or Tables 5-2–5-3 of [29]
mode? in calculating the widths and separations of the
(i) Determine the length of the coupled lines coupled lines even if your system impedance is
taking the average effective permittivity as not 50 Ω.

3.8.1 Exercises by Section

†
challenging, ‡ very challenging
§3.2 1† , 2† , 3† , 4† , 5† , 6† 17, 18† , 19† , 20, 21† , 22, 23,
§3.4 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 24† , 25† , 26† , 27† , 28† , 29‡ , 30‡ , 31‡

3.8.2 Answers to Selected Exercises

1(d) 2Ω 14 λ/8 30(d) 1.48 cm
3 5.5 16(b) [C +1/(ω02 L)] 31 39.4 Ω
5(d) 5 18 9.5 pF
6(c) 0Ω 20(a) ∞
PARALLEL COUPLED-LINE FILTERS 121

(a) (b)

(c) (d)

(e)

(f)

(g)

Figure 3-49: Early prototypes in the development of a third-order Butterworth combline filter.
122 STEER MICROWAVE AND RF DESIGN: MODULES

(h)

(i)

(j)

(k)

Figure 3-50: Late-stage prototypes in the development of a third-order Butterworth combline

filter.
 CHAPTER 4
Noise, Distortion, and Dynamic
Range

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123

4.2 Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
4.3 Noise Characterization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
4.4 Oscillator Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
4.5 Nonlinear Distortion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
4.6 Dynamic Range . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
4.7 Passive Intermodulation Distortion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
4.8 Breakdown . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
4.9 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
4.10 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
4.11 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161

4.1 Introduction
Measures of noise, distortion, and dynamic range are system metrics
describing the sensitivity of active systems and modules. In this chapter the
limitations described by these metrics will be discussed in a form that is
applicable to both active device design and system design using modules.
Noise and nonlinear distortion set the bounds on the range of signals that
can be processed by an RF circuit. Noise (i.e., random fluctuations of voltage
and current) establishes the minimum detectable signal, while nonlinear
distortion sets the level of the largest signal from which information can
be reliably extracted or which can be transmitted without affecting other
systems.

4.2 Noise
Fluctuations of voltage and current arise from several different physical
processes yielding noise with various statistical properties. Various types of
noise are important in electronic circuits and these types range from noise,
such as thermal noise, that is very well understood, to noise that has been
observed and seriously affect the performance of RF circuits but not well
understood. An example of the latter is phase noise on oscillators which
manifests itself as random fluctuations of the phase or frequency of an
oscillation signal. One of the problems in understanding noise is that it can be
difficult to describe a particular type of noise in both the time and frequency
124 STEER MICROWAVE AND RF DESIGN: MODULES

domains. Not all types of noise can be described in a straightforward way

and the sources of some types of noise are not understood. The impact of
noise on an RF system is described in the frequency domain whereas the
physical origins of noise must necessarily be in the real world (i.e., the time
domain). The impact of noise is also in the time domain resulting in bit errors
of a received, demodulated, and processed digitally modulated signal. Noise
originating from a noise source is shaped by the characteristics of a circuit
before it is observed externally and thus the true nature of noise is further
obscured.

4.2.1 Observations of Noise Spectra

Noise in electronics is attributed to the random movement of carriers and
most types of noise have a power spectral density that is flat with respect to
frequency. Such noise is called white noise and if it is filtered, lowpass filtered
or bandpass filtered, it is called additive white Gaussian noise (AWGN)
as the noise then has an asumed Gaussian statistical distribution. The best
understood noise is thermal noise and is attributed to the random movement
of electrons due to the random vibration of the lattice of a conducting
material. The theory of thermal noise is based on the fluctuation-dissipation
theorem [1] which can be used for most materials in thermal equilibrium.
This theorem applies both to classical and quantum mechanical systems and
describes the classical noise encountered up to several terahertz as well as
the quantum-mechanical effects that shape noise above a few terahertz at
room temperature or at much lower frequencies as temperature approaches
absolute zero [2].
The fluctuation-dissipation theorem relates thermally-induced fluctua-
tions in a material to the resistance of the material. The physical origin of
thermal noise is the net effect of the rapidly-fluctuating currents resulting
from thermal fluctuations of free electrons in a resistive (or conductive) ma-
terial. The fluctuations of the electrons result from the vibrations of the atoms
in a material’s lattice. These vibrations relate directly to temperature and
temperature is regarded as a direct measure of the state of entropy or ran-
dom vibration of the lattice. So the fluctuation-dissipation theorem describes
how vibration (i.e. temperature) of the lattice induces small current fluctua-
tions in a material and thus describes how heat energy is converted into the
electrical energy known as Johnson noise. This is the opposite of the effect of
resistance which converts electrical energy into heat energy when the move-
ment of electrons (as current) causes the lattice to vibrate more. So it is not
surprising that the noise current is directly related to resistance.
Key results of the fluctuation-dissipation theorem are that the available
noise power from a resistor or group of resistors is linearly proportional to
temperature, the available noise power is independent of the resistor value,
and that the power spectral density (in watts per hertz) is independent of
frequency, i.e. it is white, up to a few terahertz at room temperature.
Consider the block of resistive material in Figure 4-1(a). Lattice vibrations
cause random movement of electrons and thus there is a myriad of tiny
current sources. The block of resistive material has terminals and at those
terminals a resistance can be measured. Application of the fluctuation-
dissipation theorem determines that the net effect of the little noise-current
sources in the material is the same as that of a current source in parallel
NOISE, DISTORTION, AND DYNAMIC RANGE 125

(b) Noisy resistor and network

(a) Noisy block (c) Loaded resistor (d) Filtered noisy resistor

Figure 4-1: Noisy resistive networks

with the effective resistance as shown in Figure 4-1(a). The noise current
source, in (t), has statistical properties such that the available noise power
is proportional to the bandwidth over which the noise power is measured
and to the temperature of the material in kelvin. If the resistive material is at
a temperature T0 , then the noise temperature of the material T0 . Going one
step further, it can be imagined that the resistive material can be described by
a network of resistors as shown in Figure 4-1(b). If RS and each resistor in the
network is at the same temperature T0 , then the whole network, including
RS , is equivalent both in terms of resistance and noise to a single noisy
resistance Rx , that is at T0 . So the noise temperature of the resistive network
is equal to that of Rx both being T0 . This is true provided that there are no
non-thermal sources of noise inside the two-port network. For example, if
there were transistors then an additional source of noise is shot noise. Then
the noise temperature looking into port 2 of the network would be greater
than the noise temperature of the source resistance at port 1.
In Figure 4-1(c) a noisy resistance RS has an available noise power and
all of that available noise power is delivered to the load RL provided that
RL = RS . (All of the available noise power from RL will also be delivered
to RS under matched conditions.) Even if a resistor is not loaded, so RS is
on its own, there will be noise power in the resistor which will be constantly
generated and reabsorbed (through resistive heating) so that the resistor is in
thermal equilibrium.
A noisy resistor is equivalent both electrically and from a noise perspective
to a noise-free resistor with a shunt noise current source as in Figure 4-
1(a) or equivalently as the same value of resistance with a noise voltage
source (which will be introduced latter). The noise voltage and current
sources are random and if the noise is lowpass or bandpass filtered as in
Figure 4-1(d), the resulting noise voltage vnL (t) across the noise-free load
will have Gaussian statistics. Since it is inevitable that noise will be filtered
in a circuit, e.g. there will be at least parasitic capacitances, thermal noise is
often treated as being additive white Gaussian noise (AWGN) as its statistical
properties will be Gaussian. This is indeed fortuitous as it is possible to
greatly simplify the treatment of noise if it can be consider to be random with
Gaussian statistics. This is exploited in the development of the mathematics
of random processes in Appendix 1.A of [3] as it applies to both noise and
126 STEER MICROWAVE AND RF DESIGN: MODULES

(a) Noisy resistor network (b) Noise-free resistor network (c) Noise-free resistor

Figure 4-2: Attenuator example.

digitally modulated signals. The key result is that noise can be described in
the frequency domain and this understanding and characterization can be
translated to the real world, i.e. the time domain.
As an example, it is possible to undertake a frequency domain analysis
of the circuit shown in Figure 4-2(a) where a source RS is connected to a
resistive attenuator and then to a noise-free load RL . As usual the matched
resistances external to the two-port are the same here so that RS = RL . The
network of Figure 4-2(a) can be replaced by a network of noise-free resistors
with noise voltage sources vnS , vn1 , vn2 , and vn3 . The noise voltage sources
are random and independent and so are uncorrelated. Thus to evaluate the
noise current inL in RL the powers of the individual contributions to inL
need to be summed first. This can be shown to be identical to calculating inL
in Figure 4-2(c) where the equivalent resistance Rx of the network is found
and then the noise voltage source, vnx for that resistor used. Of course since
the two-port is an attenuator Rx = RS = RL .

4.2.2 Characterization of Thermal Noise

While the impact of noise on RF circuits is measured and categorized in
the frequency domain, the physical sources of noise are in the real world.
Noise in a conductor is manifested as random fluctuations in time of voltage
and current. While random, the noise can have different statistics depending
on how it originates. The three major physical sources of noise affecting
electronic circuits are thermal, shot, and flicker.
Thermal noise is more formally known as Johnson-Nyquist noise and
is also referred to as Johnson noise or Nyquist noise. The noise is due to
random fluctuations of charge carriers inside a conductor occurring with
or without applied voltage or current. A resistor at room temperature (290–
298 K or 19–25◦C) has an available noise power of −174 dBm in 1 hertz of
bandwidth. The noise is uncorrelated so that the noise power in a second
hertz of bandwidth will add. So in 2 hertz of bandwidth the available noise
power is (−174 + 3) dBm = −171 dBm.
The extent of fluctuation is linearly proportional to absolute temperature.
Also the noise power generated is independent of the resistance of the
conductor. The original derivation of thermal noise is due to Nyquist [4],
who showed that the power spectral density (PSD) of the available noise
power from a resistor (of any value) is

St (f ) = kT, (4.1)
NOISE, DISTORTION, AND DYNAMIC RANGE 127

where k is the Boltzmann constant and T is the temperature in kelvin. The

subscript t here is used to indicate thermal noise. The SI units of St are watts
per hertz but is more commonly expressed as dBm per hertz. The available
noise power in a bandwidth B is (in units of watts)

Pt (f ) = kT B. (4.2)

When quantum effects are important, Equation (4.1) is modified and the
thermal noise PSD is
hf
St (f ) = , (4.3)
ehf /kT − 1
where h is Plank’s constant. In Equation (4.3) hf is the energy of a photon
of frequency f and kT is the average thermal energy (i.e. vibrational
kinetic energy) of the material. To determine the frequency at which the
simpler form of St can be used consider the following. At low to moderate
frequencies Equation (4.3) can be expanded as

hf kT
St (f ) = 1 2
≈ 1 . (4.4)
1 + (hf /kT ) + 2 (hf /kT ) + ... − 1 1+ 2 (hf /kT )

Thus the thermal noise power available will drop off as frequency increases,
and is at one-half its low-frequency value at a critical frequency fc = 2kT /h.
At room temperature this is approximately 12 THz. So quantum effects on
thermal noise are not of concern at room temperature at frequencies below a
few terahertz.
Extensive modern treatments of thermal noise are available in [5] and [6].

EXAMPLE 4.1 Available Noise Power

What is the available noise power from a resistor in a 50 MHz bandwidth and at 20◦ C.
Solution:
The PSD is, from Equation (4.1),

St (f ) = (1.381 · 10−23 J/K) · ((273 + 20) K)

= 4.046 · 10−21 J = 4.046 zJ = 4.046 zW/Hz = −173.9 dBm/Hz. (4.5)

That is, 4.046 zJ (zepto joules). The thermal noise at room temperature is usually taken as
−174 dBm/Hz. This noise power is equally divided between amplitude noise and phase
noise (each is −177 dBm/Hz) [7]. St (f ) is multiplied by the bandwidth to obtain the total
available thermal noise power, so for a 50 MHz bandwidth, the thermal noise power is

Pt (f ) = (4.046 · 10−21 J) × (50 · 106 Hz)

= 2.023 · 10−13 J · s−1 = 202.3 fW = −100.1 dBm. (4.6)

This is an appreciable power given that cell phones can operate with receive signals smaller
than −90 dBm. So the lesson here is to use the smallest bandwidth possible in designs.
128 STEER MICROWAVE AND RF DESIGN: MODULES

4.2.3 Environmental Noise

Noise in RF and microwave systems includes noise from the environment
as well as noise generated within the circuitry itself. Noise from the
environment can have galactic origins, when it is known as cosmic
background noise, from black-body radiation, or can be artificially generated
noise. In cellular communication systems the major source of interference is
from other phones and base stations in the cellular system. Provided this
is uniformly random over the communication band, it can be treated as
random noise. Uniformly random noise (i.e., white noise) can be modeled
by a resistor held at what is called the noise temperature.
A noisy resistor generates white noise that has a flat PSD, i.e. noise spectral
density is independent of frequency. A noisy resistor can be modeled by a
noise-free resistor and a random voltage or current source denoted by vn
and in , respectively (see Figure 4-3). The sources vn and in are random and
their spectral densities, Svn and Sin respectively, are related by
Svn (f ) = R2 Sin (f ). (4.7)
The noise voltage spectral density of a resistor R at temperature T is [8]
vn2
Svn (f ) = = 4kT R, (4.8)
B
so that vn2 = 4kT BR. (4.9)
Here k is the Boltzmann constant, T is the temperature in kelvins, and B is
the bandwidth in hertz.
It is possible for the noise temperature looking into a two-port to be
less than the ambient temperature and then the effective noise temperature
of the structure is used as a measure of available noise power. A typical
situation is specifying the noise presented by an antenna to a receiver.
The noise captured by an antenna is from the environment with usually
only a small portion of the noise coming from the antenna itself. In the
absence of antenna loss, an antenna pointed into space will have a noise
temperature corresponding to the cosmic microwave background radiation
with an effective noise temperature of about 3 K.
Two noise voltage sources, vn1 and vn2 , in series can be partly correlated.
Then the two noise sources can be replaced by a single source vn , where
vn2 = vn1
2 2
+ vn2 + 2Cn1,n2 vn1 vn2 . (4.10)
Here Cn1,n2 is the correlation coefficient, and −1 ≤ Cn1,n2 ≤ 1. If the sources
are uncorrelated, as they would be for two resistors, Cn1,n2 = 0 and
vn2 = vn1
2 2
+ vn2 . (4.11)

Figure 4-3: Thermal noise

equivalent circuits: (a) noisy
resistor modeled as a noise-free
resistor in series with a random
noise voltage source vn ; and
(b) noisy resistor modeled as
a noise free resistor in parallel
with a random noise current
source in . (a) (b)
NOISE, DISTORTION, AND DYNAMIC RANGE 129

Correlation of noise sources is important to modeling noise in transistors, as

there can be a common physical origin for noise that is modeled as two noise
sources in a circuit model.

4.2.4 Thermal Noise and Capacitors

Thermal noise is also seen with reactive components such as a capacitor
where it is known as kT /C (read as k-T-C) noise. The series resistance, R,
of a capacitor, C, contributes the thermal noise but the RC combination also
filters the noise. This noise can be treated the same way as the thermal
noise analysis above, but there is a short-hand way of looking at the
noise. Sarpeshkar et al. [9] showed that the mean-square noise voltage on
a capacitor of value C within the noise bandwidth of the RC circuit (derived
as 1/(4RC) in hertz) is

kT
v̄n2 = . (4.12)
C
temperature (i.e., T = 20◦ C = 293 K), the root mean square noise
At room 
voltage, v̄n2 , on a 10 pF capacitor with a 1 Ω series resistance is 20 µV in a
25 GHz bandwidth.

4.2.5 Physical Source of Shot Noise

Shot noise is due to current being carried by discrete charge carriers. It is
important when there is a region that is scarce of free carriers. Shot noise is
particularly important with semiconductor devices but was first observed by
Schottky in 1926 in vacuum tubes. In a semiconductor the charge carriers,
electrons and holes, are discrete and independent. As such the current
fluctuates as the number of carriers varies in discrete steps. On average
there is a net velocity of carriers passing a point per time interval. For shot
noise to be observed above thermal noise, the carriers should be constrained
to pass in just one direction. This is the situation in many semiconductor
devices where the depletion region formed at the interface of pn junctions
forces the current to flow in just one direction. Shot noise is more significant
when the number of charge carriers is small, a situation that also exists
in semiconductors. However, even in a semiconductor, there are enough
carriers for shot noise to have a Gaussian distribution so that statistically
it looks like thermal noise [10]. The shorter the critical time scale, and for
microwave and RF circuits this is the period of the waveform, the fewer
the number of carriers that will pass a point and the greater the fractional
contributions of fluctuations in carrier numbers.
The RMS current fluctuation due to shot noise is
√
σi = 2eIB, (4.13)

where e is the elementary charge, B is the bandwidth in hertz, and I is the

current. For a DC current of 1 mA and in a 1 Hz bandwidth, the RMS current
fluctuation due to shot noise is σi = 18 pA. If this current flows through a
resistance R, the spectral noise density in the resistor is

Ps = 2e|I|. (4.14)
130 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 4-4: Noise

and two-ports: (a)
amplifier; (b) am-
plifier with excess
noise; and (c) noisy
two-port network. (a) (b) (c)

Note that Ps is independent of temperature and frequency, although

eventually quantum effects become important [11]. In contrast, thermal
noise is proportional to temperature. The current in Equation (4.14) is the
instantaneous current. So shot noise varies during an RF cycle as the current
flow varies. If the DC current is much larger than the time-varying current,
then the noise that is created by the separable pulses has a flat frequency
spectrum and can be modeled by a white noise source.
So the lesson here is that in active circuit designs, bias currents should
be minimized. Also note that the noise power is a function of the resistance
value but active circuits with high resistance values also tend to have low
current levels.

4.2.6 Physical Source of Flicker Noise

The third type of noise that is of concern with RF and microwave circuits
is flicker noise, sometimes called 1/f (one-over-f) noise because of its
power spectral density shape. Flicker noise is due to diffusion, traps in a
semiconductor, and surface traps. A free carrier is immobilized or trapped
when it falls into a trap, that is, a recombination center. When several such
carriers are trapped, it means that they are not available for conduction
and as a result, the resistance of the semiconductor is modulated. These
fluctuations have multiple relaxation times. Flicker noise is considered again
in Section 6.4 in regards to the characterization of local oscillator modules. A
complete physical understanding of flicker noise is not available, and flicker
noise is a major concern with oscillators.

4.3 Noise Characterization

Amplifiers, filters, and mixers in an RF front end process (e.g., amplify, filter,
and mix) input noise the same way as an input signal. In addition, these
components contribute excess noise of their own. Without loss of generality,
the following discussion considers noise with respect to the amplifier shown
in Figure 4-4(a), where vs is the input signal. The development here applies
only to white noise (i.e., thermal and shot noise). The noise signal, with
source designated by vn , is uncorrelated and random, and described as an
RMS voltage or by its noise power.
The most important noise-related metric is the SNR. Denoting the noise
power input to the amplifier as Ni , and denoting the signal power input to
the amplifier as Si , the input signal-to-noise power ratio is SNRi = Si /Ni .
If the amplifier is noise free, then the input noise and signal powers are
amplified by the power gain of the amplifier, G. Thus the output noise power
is No = GNi , the output signal power is So = GSi , and the output SNR is
SNRo = So /No = SNRi .
In practice, an amplifier is noisy, with the addition of excess noise,
NOISE, DISTORTION, AND DYNAMIC RANGE 131

Ne , indicated in Figure 4-4(b). The excess noise originates in different

components in the amplifier and is either referenced to the input or to the
output of the amplifier. Most commonly it is referenced to the output so that
the total output noise power is No = GNi + Ne . In the absence of a qualifier,
the excess noise should be assumed to be referred to the output. Ne is not
measured directly. Instead, the ratio of the SNR at the input to that at the
output is the noise factor:
SNRi
F = , (4.15)
SNRo
and this is the way noise performance is normally measured. If the circuit is
noise free, then SNRo = SNRi and F = 1. If the circuit is not noise free, then
SNRo < SNRi and F > 1. F can be related to the excess noise produced in
the circuit. With the excess noise, Ne , referred to the circuit output,

SNRi SNRi 1 Si N o Si GNi + Ne

F = = = =
SNRo 1 SNRo N i So Ni GSi
Ne
=1+ . (4.16)
GNi
One of the conclusions that can be drawn from this is that the noise factor,
F , depends on the available noise power at the input of the circuit. As a
standard reference, the available noise power, NR , from a resistor at standard
temperature, T0 (290 K) [12], and over a bandwidth, B (in Hz), is used,

Ni = NR = kT0 B, (4.17)

where k (= 1.381 × 10−23 J/K) is the Boltzmann constant. If the input of an

amplifier is connected to this resistor and all of the noise power is delivered
to the amplifier, then

Ne Ne
F =1+ =1+ . (4.18)
GNi GkT0 B
Several random physical processes inside a circuit contribute to excess
noise, and not all of these processes vary linearly with temperature.
Consequently F is a function of temperature, although usually a weak
one. It is also a function of bandwidth, and there is a problem in using F
with cascaded systems in which bandwidths vary for different subsystems.
Even with all these problems, F is the most important measure used
to characterize noise performance. It can be used to determine the noise
performance of a cascade, when the noise factors and gains of the subsystem
constituents are known. F is the ratio of powers, and when expressed in
decibels, the noise figure (NF) is used:

NF = 10 log10 F
= SNRi |dB − SNRo |dB , (4.19)

where the SNR is expressed in decibels.

Consider the amplifier in Figure 4-4. If the excess noise contribution of an
amplifier is ignored, the output noise power will be

No = GkT0 B. (4.20)
132 STEER MICROWAVE AND RF DESIGN: MODULES

With the amplifier’s excess noise, Ne , included, the output noise power is

No = GkT0 B + Ne = GkT0 B(1 + Ne /(GkT0 B))

= F GkT0 B. (4.21)

Rearranging this equation, the excess noise power can be written as

Ne = (F − 1)GkT0 B. (4.22)

The output noise of a system can be expressed in terms of its noise figure.
From Equation (4.18), the output noise is

No = GNi + Ne = F GNi . (4.23)

That is, No |dBm = NF + G|dB + Ni |dBm . (4.24)

So if the input noise of an amplifier with a gain of 20 dB and a noise

figure of 3 dB is −90 dBm, then the noise at the output of the amplifier is
−67 dBm. This analysis of output noise is only correct if the input source is
the equivalent of a resistor held at standard temperature, T0 . So the analysis
would not apply if the input to the system was an antenna pointed into the
sky (away from cosmic objects such as the sun and moon) which would have
a noise temperature of 4 K (due to the cosmic background radiation) plus
antenna and cable noise.

EXAMPLE 4.2 Effective Noise Temperature

An antenna with a noise temperature of 50 K is connected to an amplifier with a bandwidth

of 20 MHz, a noise figure of 3 dB, and a gain of 10 dB. What is the effective noise temperature
at the output of the amplifier?
Solution:
The output effective noise temperature, Teff , is the temperature of a resistor that would
produce the same available noise power, No , as the output available noise power of the
amplifier. The amplifier adds excess noise, Ne , to the input noise, Ni , that is amplified by
the amplifier. So to determine No , Ne must be determined first. This is obtained from the
noise figure, NF, which is defined in terms of noise power of a resistor at the input with a
temperature of T0 = 290 K.
The noise factor F = 10NF = 1.995 and from Equation (4.22)

Ne = (F − 1)GkT0 B = (1.995 − 1) · 1010/10 · (1.3807 ·−23 J/K) · (290 K) · (20 · 106 Hz)
= 7.969 · 10−13 W. (4.25)

The input noise of a resistor at 50 K is

Ni = kT B = (1.3807 ·−23 J/K) · (50 K) · (20 · 106 Hz) = 1.381 · 10−14 W. (4.26)

So the total output noise power is

No = GNi + Ne = 1010/10 · 1.381 · 10−14 + 7.969 · 10−13 = 9.350 · 10−13 W. (4.27)

Thus the effective noise temperature at the output of the amplifier is

No 9.350 · 10−13 W
Teff = = = 3386 K. (4.28)
kB (1.3807 ·−23 J/K) · (20 · 106 Hz)
NOISE, DISTORTION, AND DYNAMIC RANGE 133

EXAMPLE 4.3 Noise Figure of an Attenuator

What is the noise figure of a 20 dB attenuator in a 50 Ω system?

Solution:
Denoting the attenuator as being in a 50 Ω system indicates that an appropriate circuit
model to use in the analysis consists of the attenuator driven by a generator with a 50 Ω
source impedance, and the attenuator drives a 50 Ω load. Also, the input impedance of the
terminated attenuator is 50 Ω, as is the impedance looking into the output of the attenuator
when it is connected to the source. The key point is that the noise coming from the source
is the noise thermally generated in the 50 Ω source impedance, and this noise is equal to the
noise that is delivered to the load, as the impedance presented to the load is also 50 Ω. So the
input noise, Ni , is equal to the output noise:

No = Ni . (4.29)

The input signal is attenuated by 20 dB (= 100), so

So = Si /100, (4.30)

and thus the noise factor is

SN Ri Si No Si Ni
F = = = = 100 (4.31)
SN Ro Ni So Ni Si /100

and the noise figure is

NF = 20 dB. (4.32)
THus the noise figure of an attenuator (or a filter) is just the loss of the component. This is not
true for amplifiers of course, as there are other sources of noise, and the output impedance
of a transistor is not a thermal resistance.

4.3.1 Noise in a Cascaded System

Section 4.3 developed the noise factor and noise figure measures for a two-
port. This result can be generalized for a system. Considering the second
stage of the cascade in Figure 4-5, the excess noise at the output of the second
stage, due solely to the noise generated internally in the second stage, is

N2e = (F2 − 1)kT0 BG2 . (4.33)

Then the total noise power at the output of a two-stage cascade is

N2o = (F2 − 1)kT0 BG2 + No,1 G2

= (F2 − 1)kT0 BG2 + F1 kT0 BG1 G2 . (4.34)

This relies on the reasonable assumption that the excess noise added in one
stage is uncorrelated to the excess noise from other stages as well as being
uncorrelated to the input noise. Thus powers can be added. The second term

Figure 4-5: Cascaded noisy two-

ports.
134 STEER MICROWAVE AND RF DESIGN: MODULES

in Equation (4.34) is the noise output from the first stage amplified by the
second stage with gain G2 .
Generalizing the above result yields the total noise power at the output of
the mth stage:
 
�m �n �m
Nmo = (Fn − 1) kT0 B Gj  + F1 kT0 B Gn . (4.35)
n=2 j=2 n=1

Thus��an m-stage
� cascade has a total cascaded system noise factor of F T =
T T
Nmo G N1i , with G being the total cascaded available gain and N1i
being the noise power input to the first stage. In terms of the parameters
of individual stages, the total system noise factor is

F2 − 1 F3 − 1 F4 − 1
F T = F1 + + + + ··· , (4.36)
G1 G1 G2 G1 G2 G3
that is,
m
F −1
�n n
�
T
The similarly named F = F1 + . (4.37)
Friis transmission for- n=2 i=2 Gi−1
mula refers to antenna
systems. This equation is known as Friis’s formula [13].

EXAMPLE 4.4 Noise Figure of Cascaded Stages

Consider the cascade of a differential amplifier and a filter shown in Figure 4-6.
(a) What is the midband gain of the filter in decibels? Note that IL is insertion loss.
(b) What is the midband noise figure of the filter?
(c) The amplifier has a gain G1 = 20 dB and a noise figure of 2 dB. What is the overall gain
of the cascade system in the middle of the band? Express your answer in decibels.
(d) What is the noise factor of the cascade system?
(e) What is the noise figure of the cascade system?
Solution:
(a) G2 = 1/IL, thus G2 = −3 dB.
(b) For a passive element, NF2 = IL = 3 dB.
(c) G1 = 20 dB and G2 = −3 dB, so GTOTAL = G1 |dB + G2 |dB = 17 dB.
(d) F1 = 10NF1 /10 = 102/10 = 1.585, F2 = 10NF2 /10 = 103/10 = 1.995, G1 = 1020/10 = 100,
and G2 = 10−3/10 = 0.5. Using Friis’s formula

F2 − 1 1.995 − 1
FTOTAL = F1 + = 1.585 + = 1.594. (4.38)
G1 100
(e) NFTOTAL = 10 log10 (FTOTAL ) = 10 log10 (1.594) = 2.03 dB.

Figure 4-6: Differential amplifier

followed by a differential filter.
NOISE, DISTORTION, AND DYNAMIC RANGE 135

EXAMPLE 4.5 Noise Figure of a Two-Stage Amplifier

Consider a room-temperature (20◦ C) two-stage amplifier where the first stage has a gain of
10 dB and the second stage has a gain of 20 dB. The noise figure of the first stage is 3 dB and
the second stage is 6 dB. The amplifier has a bandwidth of 10 MHz.
(a) What is the noise power presented to the amplifier in 10 MHz?
(b) What is the total gain of the amplifier?
(c) What is the total noise factor of the amplifier?
(d) What is the total noise figure of the amplifier?
(e) What is the noise power at the output of the amplifier in 10 MHz?
Solution:
(a) Noise power of a resistor at room temperature is −174 dBm/Hz (or more precisely
−173.86 dBm/Hz at 293 K). In 10 MHz the input noise power is
Ni = −173.86 dBm + 10 log(107 ) = −173.86 + 70 dBm = −103.86 dBm.
(b) Total gain GT = G1 G2 = 10 dB + 20 dB = 30 dB = 1000.
(c) F1 = 10NF1 /10 = 103/10 = 1.995, F2 = 10NF2 /10 = 106/10 = 3.981. Using Friis’s formula,
F2 − 1 3.981 − 1
the total noise figure is F T = F1 + = 1.995 + = 2.393.
G1 10
(d) The total noise figure is NFT = 10 log10 (FT ) = 10 log10 (2.393) = 3.79 dB.
(e) Output noise power in 10 MHz bandwidth is No = F T kT0 BGT = (2.393) · (1.3807 ·
10−23 · J · K−1 ) · (293 K) · (107 · s−1 )(1000) = 9.846 · 10−11 W = −70.07 dBm.
Alternatively, No |dBm = Ni |dBm + G dB + NFT = −103.86 dBm + 30 dB + 3.79 dB =
T


−70.07 dBm.

4.3.2 Measurement of Noise Figure

The Y -factor method is the basis of modern microwave automatic noise
figure measurement systems. The technique involves measuring the noise
power at the output of a device under test (DUT) when two different noise
sources are attached to the input of the DUT [14]. The manual form of
the Y -factor method is used at millimeter-wave frequencies. The method
is dependent on the accuracy of gain measurement, the ability to generate
precise levels of excess noise power, and the sensitivity of noise power
measurement. Gain and noise power measurement are subtly different, with
gain generally a coherent measurement while the noise power measurement
is necessarily incoherent. In both automatic and manual systems, the
measurement setup is a cascade system in which the DUT is the first stage
and the test set is the last and usually second set. The noise contribution of
the test set is only negligible if the gain of the DUT is high.
Under matched conditions, the available power gain of the DUT is

So
G= , (4.39)
Si

so signal power can be eliminated from the expression for the noise factor by
combining Equations (4.15) and (4.39) :

SNRi Si N o No
F = = = . (4.40)
SNRo N i So Ni G

The output noise is larger than the amplified input noise because of the noise
inserted by the DUT. Denoting the component of the output noise power due
136 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 4-7: Noise figure test setup for

determining the noise figure of a DUT.

solely to the DUT by ND , the output noise power is

No = Ni G + ND . (4.41)

The final component of the development is noting that the input noise
power is related to the temperature of the input match so that Ni = kT B
where k is the Boltzmann constant and B is the measurement bandwidth.
Conventionally F is referenced to standard temperature T0 = 290K
(specifically the input noise temperature is T0 ) and so

ND = kT0 BG(F − 1). (4.42)

In the Y -factor method, two noise sources with known noise temperatures
T1 and T2 (with T2 > T1 ) are applied to the input of the DUT and the
corresponding output noise powers N1 and N2 measured (see Figure 4-7).
In the laboratory a calibrated noise source, such as a reverse-biased diode in
avalanche, is used to produce the noise source at T2 . The other temperature,
T1 , is often obtained by turning the noise source off so that T1 = T0 , the
ambient temperature. This leads to the Y factor, which is defined as

Y = N2 /N1 . (4.43)

In the off state, that is when T1 = T0 , the output noise power in the off state
is called the “off power”:

N1 = kT0 BG + ND = kT0 BG + kT0 BG(F − 1). (4.44)

The second noise source, with noise temperature T2 , produces calibrated

excess noise and the noise power under these conditions is called the “on
power”:

N2 = kT2 BG + ND = kT2 BG + kT0 BG(F − 1). (4.45)

Combining Equations (4.42)–(4.45) yields

T2 − T0
F = . (4.46)
T0 (Y − 1)

Expressing Equation (4.47) in decibels and integrating (perhaps through

measurement) over the system bandwidth yields the noise figure,

NF = 10 log(F ) = ENRdB –10 log(Y − 1), (4.47)

where ENRdB = 10 log[(T2 − T0 )/T0 ] is the excess noise ratio in decibels of

the calibrated noise source.
NOISE, DISTORTION, AND DYNAMIC RANGE 137

(a) Forward cascade (b) Reverse cascade

Figure 4-8: Y -factor test set actively incorporating the second-stage contribution effect.

4.3.3 Measurement of Noise Temperature

If the noise figure of an amplifier replacing the DUT is known, then the noise
temperature of a source can be determined. From Equation (4.46),

T2 = F T0 (Y − 1) + T0 . (4.48)

In a laboratory application the amplifier could be cryogenically cooled and

then there is negligible excess noise contribution by the amplifier so that
F ≈ 1 and T2 = Y T0 F .

4.3.4 Measuring the Noise Figure of Low Noise Devices

One of the factors that affects the accuracy of noise figure determination
is noise originating in the measurement test set. This leads to an error
sometimes referred to as the second-stage contribution effect [15]. This is a
particularly important issue when measuring the noise figure of low-gain
devices, as then the noise contribution of the second stage can be significant.
One approach to minimizing this error is the insertion of a high-gain low-
noise amplifier, referred to as the instrumentation amplifier, between the
DUT and the test set. The noise figure of the instrumentation amplifier
should be known precisely if the error it introduces is to be removed from
the raw noise figure measurement.
The Y -factor method relies on the instrumentation amplifier having a
much lower noise figure than the DUT. When this cannot be achieved,
the best that can be done is to use two (almost) identical amplifiers with
the same gain and noise characteristics. The extended Y -factor technique
described in this section utilizes two DUTs in a two-stage cascade first with
one arrangement of the DUTs and then with the reverse cascade [16]. The
technique makes use of the cascaded noise factor operation twice. Figure 4-8
illustrates the test setup with the two possible arrangements of the cascaded
DUTs. In Figure 4-8 the power meter is configured to measure noise power
(normalized to a 1 Hz bandwidth).
The individual noise factors of the DUTs for a cascaded system with DUT
A followed by DUT B, the forward cascade, are denoted by F1A and F2B ,
and those for the reverse cascade with DUT B followed by DUT A are F1B
and F2A . The corresponding gains of the stages are G1A , G2B for the forward
cascade, and G1B , G2A for the reverse cascade. Here the first subscript refers
to the position in the cascade (either first or second stage), and the second
subscript identifies the particular DUT (either A or B). Finally, the total noise
factors of the two-cascaded systems are denoted FAT and FBT according to
138 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 4-9: Radiometer: (a) hetero-

dyne architecture showing calibra-
tion switch; and (b) satellite ra-
diometer that switches between two
beams, one oriented at the region
being monitored and the other at
much colder space that serves as
calibration. (a) (b)

whether DUT A or DUT B is the first stage. Using Friis’s formula for the
forward cascade, Equation (4.37) can be written as,

FAT = F1A + (F2B − 1)/G1A , (4.49)

and for the reverse cascade

FBT = F1B + (F2A − 1)/G1B . (4.50)

The technique presumes that the parameters of the DUTs are invariant of
their position in the cascade so that F1A = F2A = FA and F1B = F2B = FB as
well as G1A = G2A = GA and G1B = G2B = GB . Equations (4.49) and (4.50)
can now be solved simultaneously for the unknown noise factors of the two
stages:

FB = FBT GA GB –GA (1 − FAT ) − 1 /(GA GB )

 
(4.51)
FA = FAT GA –FB + 1 /(GA )
 
and (4.52)

from the measured FAT , FBT , GA , and GB . So with the gains of the two
stages measured independently, the noise factors of the two stages can be
determined from the measured noise factors of the forward and reverse
cascades.
In the special situation of matched DUTs where the noise and gain of the
two stages are identical (so that FA = FB = F and GA = GB = F ) and
FAT = FBT = F T , then the calculations simplify to yield the noise factor of
each stage:

GF T
F = . (4.53)
G2 + 1
One of the assumptions of the extended Y -factor method is that the gain
and noise of the stages are invariant with respect to the order of the stages in
the cascade. Any departure will result in an error. A technique that reduces
the sensitivity to placement is to use small attenuators at the inputs and
outputs of the stages. The noise contributions can then be removed from the
final measured result.

4.3.5 Radiometer System

A radiometer, as shown in Figure 4-9(a), measures the power in EM radiation
predominantly at microwave frequencies and most commonly measures
noise. Radiometers are used in remote sensing and radio astronomy,
NOISE, DISTORTION, AND DYNAMIC RANGE 139

(a) Noisy active device. (b) Noise-free active device.

Figure 4-10: Amplifier model for noise factor calculation: (a) with noisy active device or amplifier
where IS is the source generator and YS is the Norton equivalent admittance of the source; and
(b) with the noisy active device or amplifier replaced by the noise voltage, en , and noise current,
in , (the noise sources), and a lossless active device or amplifier.

especially by satellites and aircraft. Understanding the physical process that

creates uncorrelated radiation at different frequencies enables vegetation,
air and sea temperatures, ice coverage, ocean salinity, and other surface
and atmospheric sources to be identified from the spectrum captured by
a radiometer. Radiometers monitor discrete windows of the spectrum,
particularly at frequencies corresponding to molecular resonances. A
radiometer includes a mechanism for rapid calibration, such as a Dicke
switch, quickly switching between the object being observed and another
object serving as a calibrated noise source. In aircraft and on land, the
calibration sources are resistors, often held at low temperatures. A satellite-
based radiometer is shown in Figure 4-9(b), and instead of a Dicke switch, the
calibration is obtained by switching the antenna beam between the observed
region (Beam A) and an empty area of space (Beam B). With a Dicke switch
or antenna beam switching, it is possible to achieve better than 0.01 K of
resolution.

4.3.6 Noise Figure of a Two-Port Amplifier

The parameters that define the noise figure of two-port amplifiers were
set forth by Haus et al. [12] in 1960 and are used by microwave transistor
manufacturers in their datasheets. The amplifier model used in noise factor
computation is shown in Figure 4-10.
Figure 4-10 has the active device or amplifier, which is noisy, represented
as a two-port with a Norton equivalent source, which will have its own noise,
and a load. The noisy active device or amplifier can be replaced by a noiseless
two-port with a voltage noise source, en , and a current noise source, in , at the
front of the two-port. Any linear noisy two-port can be represented by a noise
free two-port with noise sources. With an active device or amplifier there
will be many internal noise sources but it is the noise sources at the input,
prior to gain, which are most significant. For an active device or amplifier
the voltage and current sources will be at least partially correlated and so the
source admittance is important in determining how they combine. A worst
case would be when they combine to present the maximum possible noise
at the input of the active device or amplifier. It is also possible to adjust the
source admittance so that the noise sources combine to present the minimum
noise to the active device or amplifier. This effect is captured in the noise
140 STEER MICROWAVE AND RF DESIGN: MODULES

factor of the amplifier in Figure 4-10. The noise factor of this amplifier is [12]
rn  2
F = Fmin + ys − yopt  , (4.54)
gs

where rn = (Rn /Z0 ) is called the equivalent noise resistance of the two-port
and Fmin is the minimum noise factor obtained by adjusting tuners at the
input of the amplifier to present all possible values of YS to the input of the
amplifier. The normalized admittance presented by the tuners at Fmin is yopt .
With ys (= YS /Z0 ) and gs = ℜ{ys } being the actual normalized admittance
and conductance, respectively, Equation (4.54) enables the noise factor to be
calculated for an actual design. The parameters in Equation (4.54) describe
the effect of internal amplifier noise sources and how they are correlated.
More commonly the noise parameters are reported in terms of reflection
coefficients rather than admittance. The source reflection coefficient, Γs ,
comes from
1 − Γs
ys = (4.55)
1 + Γs
and the optimum source reflection coefficient, Γopt , comes from

1 − Γopt
yopt = . (4.56)
1 + Γopt

Substituting these into Equation (4.54) results in

 2
4rn Γs − Γopt 
F = Fmin +   2 . (4.57)
2 
1 − |Γs | 1 + Γopt 

Together Fmin , rn , and Γopt are called the noise parameters of a device and
must be measured. The noise parameters of a pHEMT transistor are given in
Table 4-1.
In general, the design for best noise performance does not yield the best
gain. The reduction in gain is usually small however. Designing the input
and output matching networks of an amplifier to be conjugately matched
yields maximum amplifier gain. For best noise performance, however, the
input matching network is not conjugately matched and instead the input
reflection coefficient looking into the matching network from the active
device is Γopt . As a result, maximum gain is not obtained. The interpretation
of why this is necessary is that a particular mismatch minimizes the
combined noise contributions of partially correlated internal active device
noise sources. In modern RF and microwave design, however, if noise
performance is of concern, two or more amplifier stages are used, with the
first stage designed for optimum noise performance and subsequent stages
designed to obtain the required gain.
NOISE, DISTORTION, AND DYNAMIC RANGE 141

4.4 Oscillator Noise

Noise on oscillators has a particular characteristic that affects the use of
oscillators in RF systems. When oscillating, the frequency of oscillation varies
slightly and this variation is captured as phase noise.

4.4.1 Observations of Noise Spectra of Oscillators

In the frequency domain, noise is characterized by its power spectral density,
S, which is the noise power in a specified bandwidth. It is typical to use 1 Hz
as the reference bandwidth and then S is expressed in units of watts per hertz
or, more commonly, as dBm per hertz. Some types of noise are proportional
to the RF signal or carrier level and are expressed relative to the level of the
carrier as dBc/Hz (10 log of the ratio of the noise power per hertz relative to
the power of the carrier). Noise can be decomposed into a phase component,
Sϕ , and an amplitude component, Sa . Then the observed phase fluctuations,
as classified in [18], are as shown in Figure 4-11. There are also amplitude
fluctuations that have a similar characteristic. For example, flicker amplitude
noise will also have an f −1 dependence. However, the phase component
generated by electronic circuits gets the most attention as small amplitude
variations are quenched by device nonlinearities. In Figure 4-11, the log of

Frequency Fmin NFmin Γn  Γn rn /50 Table 4-1: Noise parameters of an

(GHz) (dB) degrees (Ω) enhancement-mode pHEMT transistor
0.90 1.07 0.29 0.747 15.70 0.165 model FPD6836P70 [17]. Fmin is the
1.80 1.09 0.38 0.623 24.95 0.176 minimum noise factor, NFmin is the
2.40 1.11 0.44 0.795 37.45 0.158 minimum noise figure, and Γn = Γopt is
2.60 1.11 0.47 0.640 47.15 0.159 the source reflection coefficient yielding
2.80 1.12 0.49 0.670 47.90 0.160 Fmin .
3.20 1.13 0.53 0.617 51.20 0.156
4.00 1.15 0.61 0.542 68.70 0.141
5.00 1.18 0.72 0.465 85.00 0.120
5.50 1.19 0.77 0.431 91.10 0.114
6.00 1.21 0.83 0.366 101.15 0.107
7.00 1.24 0.93 0.262 122.10 0.096
8.00 1.27 1.04 0.188 153.60 0.100
9.00 1.30 1.14 0.135 −165.60 0.121
10.00 1.33 1.25 0.162 −126.80 0.138
11.00 1.37 1.35 0.183 −85.95 0.187
12.00 1.40 1.46 0.270 −68.40 0.239
13.00 1.43 1.57 0.343 −50.25 0.355
14.00 1.47 1.67 0.431 −43.95 0.461
15.00 1.51 1.78 0.573 −25.80 0.604

Figure 4-11: Wideband noise showing only

phase and frequency noise. Sϕ is the
spectral density of the phase noise and f is
frequency.
142 STEER MICROWAVE AND RF DESIGN: MODULES

the phase noise power spectral density relative to a reference level is plotted
as a function of the log of frequency. The graph does not go all the way down
to DC because the frequency axis is logarithmic. This does not imply that
the noise power goes to infinity1 as, of course, the bandwidth is becoming
smaller and smaller at low log(f ).
Figure 4-11 is the usual way to plot noise, as straight-line regions are
clearly observed experimentally and in these regions the noise power
spectral density varies with frequency as 1/f n , where n is a positive
integer. Sometimes one or more of these regions will be missing, and this
is interpreted as the crossover frequencies, the fc,−n s in Figure 4-11, being
out of order.
The observed noise characteristic (Figure 4-11) is intriguing. Why are there
straight line regions and what are the physical processes that produce noise
with such characteristics? The straight-line regions are in contrast to what
could, perhaps, be expected to be a smooth continuous transition from a low
spectral density region to a high spectral density region.
Names have been given to the regions [18], with the first being white phase
noise, noise with no frequency dependence (i.e., a frequency dependence
of f 0 ). This noise has a mean and a variance but no higher (i.e., third and
above) moment. In the time domain this noise has (actually assumed to
have) a Gaussian distribution. This is the form of noise that can be translated
between the time and frequency domains using integer calculus.
The next noise region seen in Figure 4-11 is called the flicker phase noise
region with noise spectral density dependent on 1/f (or f −1 ). Traditionally
this has been a puzzling noise characteristic and is possibly due to chaotic
behavior. The 1/f characteristic indicates that there is long-term memory,
which is the same as saying that fluctuations (i.e., noise) are correlated to
fluctuations at past times. In RF and microwave systems, flicker phase noise
is often the most significant noise and sets the noise-related performance
limits receiver circuits in particular. Energy consumption is also affected
as flicker noise is usually suppressed by increasing oscillator bias currents.
Following the 1/f region are the white frequency, f −2 , flicker frequency, f −3 ,
and random walk frequency regions, f −4 . These regions also indicate long-
term correlation of fluctuations. It has been shown that at least some of the
noise in the f −2 region is die to up-converted white noise from near DC [22]
and from white noise near harmonics of the oscillating signal [23].
At RF the noise of greatest interest is the noise superimposed on a signal
that is amplified, or the noise that is generated by an oscillator. This noise
has a white noise component and one or more components with 1/(∆f )n
dependence, where n is a positive integer and ∆f is the offset from the center
signal frequency. The noise with 1/(∆f )n dependence has been found to

1 More importantly, the integration, using integer calculus, is not valid. The noise that
has the 1/f -like spectrum is believed to be fractal (i.e., of chaotic origin). Fractals are
formally irregular, rough, and nondifferentiable. That is, fractal processes are inaccessible to
treatment by integer calculus and thus they do not have a power spectrum [19]. However,
it has been shown that such a process when passed through an ideal bandpass filter will
become stationary and then have a power spectrum [20, 21]. What this means is that since
measurement equipment is band limited, the measured spectrum may appear to approach
infinity as the offset goes to zero, but the underlying physical noise process will not. As well,
the real bandpass filter of measurement equipment has loss and so a measurement artifact is
that noise measurements will level off for smaller and smaller frequency offsets.
NOISE, DISTORTION, AND DYNAMIC RANGE 143

(a) Double-sideband noise (b) Single-sideband noise

Figure 4-12: Narrowband noise around the carrier frequency, f0 , of an oscillator showing only
phase and frequency noise with ∆f = f − f0 .

be at least partly related to the noise observed at low frequencies. (This is

another indication of our lack of a full understanding of noise.)
The phase noise observed on the signal produced by an oscillator is
shown in Figure 4-12(a). While the noise appears both above and below
the oscillating frequency, so-called double-sideband noise, usually only the
upper sideband is plotted, as shown in Figure 4-12(b). The designations
of the straight-line regions [18] (see Figure 4-12(b)) correspond to the
designations for low-frequency noise (see Figure 4-11). The noise observed
on an oscillation signal or on the signal at the output of an amplifier must,
necessarily, come from processing of a physical noise source.
The noise on the signal at the output of an oscillator is almost entirely
phase fluctuations, as saturation quenches amplitude fluctuations. Phase
fluctuations in the frequency domain are characterized by spectral density
of the phase fluctuations. If the signal at the output of the oscillator is
v(t) = [V0 + ǫ(t)] sin [2πf0 t + φ(t)] , (4.58)
where ǫ(t) is the amplitude fluctuation and φ(t) is the phase fluctuation, the
spectral density of phase fluctuations is
E φ2 (t)
 
Sφ (∆f ) = Sφ (∆ω) = PSD [φ(t)] = , (4.59)
B
where ∆f = ∆ω/(2π) is the frequency offset from the carrier at frequency f0 ,
PSD refers to power spectral density, E[ ] refers to the estimated value (here
the mean of the squared phase), and B is the bandwidth over which the
estimate is made (i.e., the bandwidth of the measurement). With bandwidth
having the units of hertz, then the units of Sφ are radians2 /Hz. Equation (4.59)
includes contributions from both the upper and lower sidebands and so it
is a double-sideband measure of phase noise. The preferred measure is the
single-sideband phase noise spectral density L (read as script-L) so that2
L(f ) = 12 Sφ (f ). (4.60)

2 In the past, L(f ) was defined as the single-sideband noise power in a 1 Hz bandwidth divided
by the carrier power. This definition has been superseded by Equation (4.60) because of
ambiguities in applying the old definition when both amplitude and phase fluctuations are
significant.
144 STEER MICROWAVE AND RF DESIGN: MODULES

4.5 Nonlinear Distortion

Distortion imposes a fundamental limit to the efficiency that can be realized
in an RF system [24–27]. There must be enough DC power to ensure that
signals are processed with no more than the maximum acceptable distortion.
Nonlinear distortion originates when the output signal from an amplifier, for
example, approaches the extremes of the load line so that the output is not
an exact amplified replication of the input signal.

4.5.1 Amplitude and Phase Distortion

For a one-tone signal, the amplitude gain of the signal rolls off as the
input power increases, as shown in Figure 4-13(a). This figure plots the
power of the output sinewave against the power of the input sinewave. The
plot is called the AM-to-AM (AM-AM) response (the amplitude-dependent
amplitude response) of the amplifier. The AM-AM characteristic is linear
at low input powers, but eventually the gain reduces (compresses) and the
output power drops below the linear extension of the small-signal response.
In Figure 4-13(a), the 1 dB gain compression point is the point where the
difference between the extrapolated linear response exceeds the actual gain
by 1 dB. P1dB is the output power at the 1 dB gain compression point and is
the single most important metric of distortion, and amplifier designers use
P1dB as a point of reference.
A gain compression of 1 dB corresponds to a signal voltage amplitude
reduction of 11% so P1dB indicates a small but significant reduction in
the linear gain of an amplifier. For high orders of modulation this 11%
error could mean that the sampled received signal does not match the
actual transmitted constellation point. The range of power of a modulated
input signal is usually chosen so that the peak power of the envelope is
equal to or less than the 1 dB gain compression power. Thus for nearly all
modulation formats (a notable exception is FM) the average signal power
is backed-off from the 1 dB gain compression power by the peak-to-mean
envelope power ratio (PMEPR) of the modulated signal. With higher-order
modulation having a higher PMEPR a greater back-off is required.
At large powers, the parasitic capacitances of the transistors in the
amplifier vary the phase of the output signal, and hence phase distortion
occurs. Figure 4-13(b) shows what is called the AM-to-PM (AM-PM)
characteristic which is a measure of the amplitude-dependent phase
distortion. For a digitally modulated signal, phase distortion would be

Figure 4-13: Nonlinear

effects introduced by RF
hardware: (a) amplitude-
dependent amplitude
(AM-AM) distortion; and
(b) amplitude-dependent
phase (AM-PM) distor-
tion. (a) AM-AM (b) AM-PM
NOISE, DISTORTION, AND DYNAMIC RANGE 145

significant if it caused the wrong constellation point to be selected in a

receiver. So for 8-PSK modulation a 22.5◦ phase shift would be significant,
less if noise is considered. The AM-AM distortion generally occurs before
the output phase varies appreciably. This is because the nonlinearity of
a transistor is predominantly a current-voltage nonlinearity with phase
distortion coming mostly from the nonlinearity of parasitic capacitances.
While Figure 4-13 presents the distortion characteristics for a single
sinewave, it has proved to be a reasonable indicator of performance with
modulated signals. In particular, with the maximum peak envelope output
power being P1dB , in-band distortion is usually acceptable while maximizing
high amplifier efficiency [28].
The AM-AM and AM-PM characterizations describe distortion with an
amplifier but they are also used with other types of nonlinear subsystems
such as mixers.

EXAMPLE 4.6 Amplifier Back-Off

An amplifier has an output power of 10 dBm when the gain of a single tone is compressed
by 1 dB. What is the maximum output power of an undistorted QPSK signal with a PMEPR
of 3.6 dB?
Solution:
The maximum undistorted output QPSK signal is generally accepted as being when the peak
envelope power is equal to the 1 dB gain compression power. In dBm the peak envelope
power is greater than the mean signal power by the peak-to-mean envelope power ratio.
Thus the QPSK signal power is said to be backed-off from the 1 dB gain compression point
by PMEPR. Thus the maximum undistorted output power of the QPSK signal is

Po,QPSK = P1dB |dBm − PMEPR|dB = 10 dBm − 3.6 dB = 6.4 dBm. (4.61)

4.5.2 Gain Compression

The voltage vo (t) at the output of an amplifier that is biased in the middle
of the output current-voltage characteristics (i.e., a Class A amplifier) can be
modeled by the first few terms of a Taylor series
vo (t) = ao + a1 vi (t) + a2 vi2 (t) + a3 vi3 (t) + . . . , (4.62)
where vi (t) is the input voltage. In a Class A amplifier, the even-order terms
are small because distortion at the extremes of the voltage waveform are
largely symmetrical, and it is found that the coefficients of the odd-order
terms of the Taylor series rapidly decrease so that |a1 | ≫ |a3 | ≫ |a5 | . . . .
Also it is found that a3 is negative. With a single sinusoidal input voltage at
frequency f1 = ω1 /(2π), vi (t) = V1 cos(ω1 t) and the output voltage is
vo (t) =a0 + a1 Vi cos(ω1 t) + a2 V12 cos2 (ω1 t) + a3 Vi3 cos3 (ω1 t) + . . .
= a0 + 12 a2 Vi2 + a1 Vi + 34 a3 Vi3 cos(ω1 t)
   

+ 12 a2 Vi2 cos(2ω1 t) + 14 a3 Vi3 cos(3ω1 t) + . . . . (4.63)

RF and microwave amplifiers are used with bandpass filters and the
distortion corresponding to harmonics is easily filtered out. Thus the output,
after filtering and removing the DC component is
vo (t) = a1 Vi + 34 a3 Vi3 cos(ω1 t) = Vo cos(ω1 t),
 
(4.64)
146 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 4-14: A two-

tone signal: (a)
a two-tone input
waveform; and (b)
distorted output
showing compression
(dashed waveform is
undistorted). (a) (b)

Figure 4-15: Spectrum at the output of a nonlinear

amplifier with a two-tone input signal. This is the
spectrum of the waveform in Figure 4-14. Numerically
f3 = 2f1 − f2 and f4 = 2f2 − f1 .

where Vo = a1 Vi + 34 a3 Vi3 is the magnitude of the voltage at the input

frequency. So the voltage gain is
V Vo
G= = a1 + 34 a3 Vi2 (4.65)
Vi
and the power gain (assuming the input and output impedances are the
same) is 2
G = VG . (4.66)
Since a3 is negative, the gain for very low input voltages, small V1 , is linear
with V G = a1 and the gain reduces for larger input signals, resulting in
the gain compression, the reduction in the slope of Pout versus Pin , seen in
Figure 4-13(a). For other amplifiers, and particularly for switching amplifiers,
many more terms must be considered in the power series expansion, and
it is possible that the gain could increase before dropping and eventually
saturating. This temporary increase in gain is called gain expansion.

4.5.3 Intermodulation Distortion

A two-tone signal consisting of two sinusoidal signals is a better
representation of system performance with modulated signals. A signal
that is the linear combination of two sinusoidal signals of equal amplitude
is shown in Figure 4-14(a). When this signal is large and is input to an
amplifier in what is called a two-tone test, the extremes of the output signal
are compressed. This results in the saturated output waveform shown in
Figure 4-14(b), where the dashed curve is the undistorted waveform. In
the frequency domain this distortion produces additional tones so that this
distortion is said to produce intermodulation products (IMPs), as shown
in Figure 4-15. Here the f1 and f2 components have the frequencies of the
original two-tone input signal. The extra tones in the output, f3 and f4 , are
the intermodulation tones. The tone at f3 = 2f1 − f2 is known as the lower
IM3 (or lower third-order intermod) tone and the tone at f4 = 2f2 − f1 is
known as the upper IM3 tone.
NOISE, DISTORTION, AND DYNAMIC RANGE 147

The simplest way to view intermodulation distortion is to consider a two-

tone input signal,

vi (t) = Vi [cos(ω1 t) + cos(ω2 t)] , (4.67)

where the tones at frequencies f1 = ω1 /(2π) and f2 = ω2 /(2π) have equal

amplitude Vi . Substituting this signal into the Taylor series expansion in
Equation (4.62) leads to the output signal
2
vo (t) = a0 + a1 Vi cos(ω1 t) + a1 Vi cos(ω2 t) + 21 a2 Vi2 [cos(ω1 t) + cos(ω2 t)]
+ a3 Vi3 [cos(ω1 t) + cos(ω2 t)]3 + . . .
= a0 + a1 Vi cos(ω1 t) + a1 Vi cos(ω2 t) + 12 a2 Vi2 [1 + cos(2ω1 t)]
+ 21 a2 Vi2 [1 + a1 cos(2ω2 t)] + a2 Vi2 cos(ω1 − ω2 )t + a2 Vi2 cos(ω1 + ω2 )t
+ a3 Vi3 34 cos(ω1 t) + 14 cos(3ω1 t) + a3 Vi3 34 cos(ω2 t) + 14 cos(3ω2 t)
   

+ a3 Vi3 32 cos(3ω1 t) + 34 cos(2ω2 − ω1 )t + 34 cos(2ω2 + ω1 )t

 

+ a3 Vi3 32 cos(3ω2 t) + 34 cos(2ω1 − ω2 )t + 34 cos(2ω1 + ω2 )t + . . . .

 

(4.68)

The component of the output at the first fundamental is

Vo,1 = a1 Vi + 34 a3 Vi3 (4.69)

and the component of the output at the second fundamental is

Vo,2 = a1 Vi + 34 a3 Vi3 . (4.70)

Note that a1 is the linear voltage gain of the amplifier:

a1 = Vo,1 /Vi = Vo,2 /Vi . (4.71)

4.5.4 Third-Order Distortion

RF and microwave amplifiers are typically used with bandpass or lowpass
inter-stage matching networks or filters, and so the distortion corresponding
to harmonics is filtered out. However, the tones at frequency f3 = 2f1 − f2 =
ω3 /(2π) and f4 = 2f1 − f2 = ω4 /(2π) will be within the passband of the
amplifier if f1 and f2 are close. The appearence of these tones indicates third-
order distortion as the distortion is the result of third-order terms in the
power series expansion of the two-tone input-output characteristic as given
in Equation (4.68). That is, the output after filtering and removing the DC
component is

vo (t) = Vo,1 cos(ω1 t) + Vo,2 cos(ω2 t) + Vo,3 cos(ω3 t) + Vo,4 cos(ω4 t), (4.72)

where the component of the output

at the first fundamental, f1 , is Vo,1 = a1 Vi + 43 a3 Vi3 , (4.73)
3 3
at the second fundamental, f2 , is Vo,2 = a1 Vi + 4 a3 Vi , (4.74)
3 3
at the lower intermodulation frequency, f3 , is Vo,3 = 4 a3 Vi , (4.75)
3 3
at the upper intermodulation frequency, f4 , is Vo,4 = 4 a3 Vi . (4.76)
148 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 4-16: Output power versus input power

of an amplifier plotted on a logarithmic scale.
The IM3 response is a result of two-tone
intermodulation, and the input power is the
power of each of the two signals that have
equal amplitude. Extrapolations of the 1:1 linear
fundamental response and the 3:1 third-order
intermodulation response intersect at the IP3
point.

Thus the level of the intermodulation tones, the IMD level, increases as the
third power of the level of the two-tone input signal. Since the IMD levels
vary as the third power of the input tone level (Vi ), it is usual to refer to the
tones at f3 and f4 as third-order intermods, or IM3 tones. The one-tone and
IM3 responses are plotted in Figure 4-16 up until saturation, where higher-
order terms in the Taylor expansion become important.
The ratio of the amplitude of the intermodulation tones to the amplitude
of the input tones (recall that they have equal amplitude) enables the third-
order power series coefficient to be calculated. That is

4 Vo,3 4 Vo,4
a3 = 3 = . (4.77)
3 Vi 3 Vi

The gain and IM3 responses shown in Figure 4-16 are plotted on a
logarithmic scale. First consider the gain response, which is plotted for a
single sinewave input. At low input power levels the amplifier has linear
gain and the output power, Po1 , increases in proportion to the input power,
Pi1 , so the gain response has a 1:1 slope. As the input power further increases,
the output power saturates primarily because the waveform at the output
is constrained by the limits set by the supply and ground rails, but other
nonlinearities of the transistor impact the linearity of the response before
saturation is reached. The IM3 response (either the level of the f3 tone or
the f4 tone) in Figure 4-16 is when the levels of two discrete tones are the
same. Then typically the responses of the upper and lower IM3 tones are
the same unless capacitive effects become important [29–32]. At low levels
of the two input tones, each with power Pi3 , the output power, Po3 , at one
of the IM3 tones increases as the third power of Pi3 . Thus on a logarithmic
scale the slope of Po3 versus Pi3 is 3:1. As the input power increases the
IP3 response eventually saturates. The two simplest characterizations of the
nonlinear response of an amplifier are the 1 dB gain compression power,
as discussed before, and the intercept of the gain and IP3 responses. This
intersection is called the third-order intercept point, or IP3 point (see Figure
4-16). The input power at IP3 is called the input third-order intercept power,
or IIP3, and the output is called the output third-order intercept power, or
OIP3. If G is the linear power gain, OIP3 = G · IIP3. IIIP3 is mostly used
with receivers and OIP3 is used with transmitters.
NOISE, DISTORTION, AND DYNAMIC RANGE 149

4.5.5 Spectral Regrowth

Distortion with digitally modulated signals consists of in-band and out-of-
band distortion. The generation of in-band (within the bandwidth of the
digitally modulated signal) intermodulation distortion in an amplifier, or
any other nonlinear device such as a mixer, affects the ability to resolve
the constellation of a received signal. Thus distortion generated in-band
affects the ability to interpret the constellation of the signal and hence
causes bit errors. However, the phase of the intermodulation products is of
little concern except in cascaded systems where the phase affects how IMD
distortion from different stages combines.
Out-of-band distortion is represented in Figure 4-17, where the spectra
at the input and output of a nonlinear system are shown. The process that
results in increased signal levels in the adjacent sidebands is called spectral
regrowth. This distortion is similar to the intermodulation distortion with a
two-tone signal. The generation of signals in the adjacent channel affects the
function of other radios and the allowable level of these signals is contained
in system specifications.

(a) Input spectrum

Figure 4-17: Spectra at the input and

output of an amplifier with of a
(b) Output spectrum digitally modulated signal.
150 STEER MICROWAVE AND RF DESIGN: MODULES

4.5.6 Second-Order Distortion

The previous subsection discussed intermodulation distortion and intro-
duced the third-order intercept point, IP3, defined either by the input-
referred power of IP3, IIP3, or the output-referred power of IP3, OIP3, to
characterize third-order nonlinear performance. The same type of analysis
can be used to characterize the second-order nonlinear performance of a mi-
crowave module. Second-order distortion leads to the second harmonic of
the input tones and to the difference frequency of the input tones. The com-
ponents of the output response of the amplifier in Equation (4.68) at the sec-
ond harmonic and difference frequency are:

vo (t) = Vo,2nd,1 cos(2ω1 t) + Vo,2nd,2 cos(2ω2 t) + Vo,diff cos(ω2 − ω1 t). (4.78)

where the amplitude at the output at the

second harmonic of f1 is Vo,2nd,1 = 21 a2 Vi2

second harmonic of f2 is Vo,2nd,2 = 12 a2 Vi2
difference frequency f2 − f1 is Vo,diff = 12 a2 Vi2 . (4.79)

Thus the coefficient of the second-order term in the defining polynomial can
be obtained from the amplitude of either of the second harmonics or of the
difference tone.
Vo,2nd,1 Vo,2nd,2 Vo,niff
a2 = 2 =2 =2 (4.80)
Vi2 Vi2 Vi2

The gain and IM2 responses shown in Figure 4-18 are plotted on a
logarithmic scale. At low input power levels the amplifier has a linear gain
and initially the gain response has a 1:1 slope. The IM2 response (either
the levels of the second harmonic tones or of the difference tone) in Figure
4-18 is when the levels of two discrete tones are the same. At low levels
of the two input tones, each with power Pi2 , the output power, Po2 , at
one of the harmonic or difference tones increases as the quadratic of Pi32 .
Thus on a logarithmic scale the slope of Po2 versus Pi2 is 2:1. As the
input power increases the IP2 response eventually saturates. The simplest
characterizations of the second-order nonlinear response of an amplifier is
the IP2 response and the IP2 intercept point, the second-order intercept point.
The input power at IP2 is called the input second-order intercept power, or
IIP2, and the output is called the output second-order intercept power, or
OIP2.

4.5.7 Summary
The three simplest characterizations of the performance of an amplifier are
the linear gain, the input or output power at the IM2 intercept point, IP2,
and the input or output power at the IM3 intercept point, IP3, see Figures
4-18 and 4-19. From these the first three coefficients of a polynomial model
of an amplifier can be derived. The first, a1 , comes from the linear gain, see.
The second, a2 , comes from the level of the harmonics or difference tones
for a two-tone input signal, see Equation (4.80). The third, a3 comes from the
level of third-order intermodulation tones, see Equation (4.77).
NOISE, DISTORTION, AND DYNAMIC RANGE 151

4.6 Dynamic Range

While modern communication and radar systems use digitally modulated
signals, two-tone signals are used to characterize nonlinearity (usually) and
also in manual calculations. At low powers before compression becomes a
factor, the fundamental response (i.e., output power versus input power)
initially has a 1:1 slope with respect to the input, as shown in Figure 4-19. The
third-order intermodulation, IM3, response varies as the cube of the level of
input tones when both tones vary by the same amount, as is common in
a two-tone test. Thus the IM3 response initially has a 3:1 logarithmic slope
with respect to the input. Since the relations are linear in a log-log sense,
it is possible to describe the nonlinear performance of an amplifier by a
quantity called the dynamic range (DR) or by the similar spurious free
dynamic range (SFDR). SFDR describes the difference between the level
at which a signal is distorted and the level of noise (i.e., the noise floor).
DR is similar to SFDR except that the level of the minimum discernible

Figure 4-18: Output power versus input power

of an amplifier plotted on a logarithmic scale.
The IM2 response is a result of two-tone
intermodulation (or of harmonic generation),
and the input power is the power of each
of the two signals that have equal amplitude.
Extrapolations of the 1:1 linear fundamental
response and the 2:1 second-order response
intersect at the IP2 point.

Figure 4-19: Output power versus input power

of a stage or system plotted as output power
in decibels versus input power in decibels.
The IM3 response is a result of two-tone
intermodulation, and the input power is the
combined power of the two signals that have
equal amplitude. Extrapolations of the 1:1
linear fundamental response and the 3:1 third-
order intermodulation response intersect at the
IP3 point.
152 STEER MICROWAVE AND RF DESIGN: MODULES

signal (MDS) (also called the minimum detectable signal) is used instead
of the noise floor. MDS is higher than the level of the noise floor by the
minimum acceptable SNR (SNRMIN ). SNRMIN is dependent on the type of
modulation, on hardware inadequacies (captured by the implementation
margin), on processing gain, and on the error correction coding used in a
particular communications protocol.
It is interesting to note that the 1 dB gain compression level has no effect
on the dynamic range of microwave circuits and systems. This is because
dynamic range is concerned with the ability to detect a signal when it
is possible for the desired signal to be masked by spurious signals. Gain
compression on its own does not introduce spurious signals so it does not
interfere with the ability to detect a small signal.
The dynamic range does not relate to how accurately the sampled received
signal matches the constellation diagram of the transmitted digitally
modulated signal. Distortion of the constellation diagram is determined by
noise, gain compression, and intermodulation distortion.
In the following, an expression for SFDR is developed in terms of
input-referenced quantities, and this form of the SFDR is called the input
referred SFDR (SFDRi ). A similarly referenced dynamic range (DRi ) is also
developed.
Figure 4-19 graphically defines the dynamic ranges. The point of
intersection of the extrapolated linear and IM3 responses is called the third-
order intercept point (IP3 intercept). The point is identified by the output-
referred intercept power (OIP3) or by the input-referred IP3 intercept power
(IIP3) and these are key parameters in describing the linearity of nonlinear
subsystems.
In the linear gain region, the output power Po versus the input power Pi
has a slope of 1:1, so that

PdBm,i = PdBm,o − GdB , (4.81)

where GdB is the power gain in decibels. Po is used here as the output power,
with PdBm,o indicating the output power in dBm. Pi and PdBm,i are similarly
defined. In terms of input quantities

OIP3
IIP3 = = IIP3dBm = OIP3dBm − GdB , (4.82)
G

where again the dBm subscript indicates that the quantity is expressed in
decibels referred to 1 mW.
The nonlinearity of RF active components results in harmonics and inter-
modulation components. With the narrowband amplifiers of communication
and radar systems, output filters conveniently filter out harmonics. How-
ever, intermodulation distortion cannot be filtered out, as these components
are within the main passband. The intermodulation components are there-
fore spurious tones. Generally just one of these defines the maximum spuri-
ous tone and nearly always it is one of the third-order intermodulation tones
resulting from a two-tone input. Consideration of the maximum spurious
tone and the noise floor defines the SFDR.
Examining Figure 4-19 leads to the following inequality describing the
NOISE, DISTORTION, AND DYNAMIC RANGE 153

linear gain of the amplifier:

OIP3dBm − PdBm,o OIP3dBm − PdBm,o
=
IIP3dBm − PdBm,i (OIP3dBm − GdB ) − PdBm,i
= 1. (4.83)

The IM3 response is characterized by first introducing an equivalent input

power, PdBm,i3 (Pi3 expressed in dBm), defined as the average power of the
two-tone signal that generates an IM3 of power PdBm,o3 . Noting that PdBm,o3
varies with a 3:1 logarithmic slope with respect to PdBm,i3 , then
OIP3dBm − PdBm,o3
=3 (4.84)
(OIP3dBm − GdB ) − PdBm,i3
or
1
PdBm,i3 = 3 (2 · OIP3dBm + PdBm,o3 − 3GdB ) . (4.85)
The SFDR can now be defined when the third-order intermodulation product
of two-tone excitation is the dominant spurious tone. The SFDR is defined as
the difference between Pi3 and Pi when they produce IM3 and linear output,
respectively, that are both equal to the output noise power, No (see Figure
4-19); that is, when Po = Po3 = No . Replacing PdBm,o in Equation (4.85) with
No gives
PdBm,i3 = 31 (2 · OIP3dBm + NdBm,o − 3GdB ) (4.86)
and
PdBm,i = NdBm,o − GdB . (4.87)
Note that the difference between the linear output and the third-order
intermodulation reduces as the input power increases above Pi3 . Thus the
output-referred SFDR is

SFDRdB,o = PdBm,i3 − PdBm,i

2 1
= 3 OIP3dBm + 3 NdBm,o − GdB − NdBm,o + GdB
2
= 3 (OIP3dBm − NdBm,o ) . (4.88)

So SFDR is 2/3rds of the range in decibels from the noise intercept to the
third order intercept point. A similar development defines the identical
input-referred SFDR:
2
SFDRdB,i = 3 (IIP3dBm − NdBm,i ) . (4.89)

Note that Ni is the input-referred noise and includes noise applied to the
module as well as the noise produced internally in the module and referred
to the input. The SFDR provides a combined measure of distortion and noise.
However, for usable dynamic range the minimum acceptable SNR must be
considered. The minimum SNR (SNRMIN ) required is determined by the
communication or radar modulation format, error coding, and acceptable
BER. So in defining DR, the input power of the desired signal must increase
sufficiently to produce an SNR of at least SNRMIN . Since the desired spurious
level is still at the noise floor, this implies a direct subtraction in decibels of
the desired SNR. Therefore the input-referred third-order dynamic range,
preferred for receivers, is
2
DRi = 3 (IIP3dBm − NdBm,i − SNRdB,MIN ) (4.90)
154 STEER MICROWAVE AND RF DESIGN: MODULES

and the output-referred dynamic range, preferred for transmitters, is

2
DRo = 3 (OIP3dBm − NdBm,o − SNRMIN ) . (4.91)

The minimum discernible signal at the output is

MDSdBm,o = NdBm,o + SNRdB,MIN , (4.92)

so the output-referred dynamic range can be written as

2
DRo = 3 (OIP3dBm − MDSdBm,o ) . (4.93)

So DR is 2/3rds of the range in decibels from the level of the minimum

detectable signal to the third order intercept point. The minimum discernible
signal at the input is

MDSdBm,i = NdBm,i + SNRdB,MIN (4.94)

and the input-referred dynamic range (equal to DRo ) can be written as

2
DRi = 3 (IIP3dBm − MDSdBm,i ) . (4.95)

4.7 Passive Intermodulation Distortion

Passive intermodulation distortion (PIM) is the term used to describe
nonlinear distortion that occurs when passive components are supposedly
linear. Just as with nonlinear components such as amplifiers, the generation
of spurious tones can swamp small received signals. The typical test of
PIM is a two-tone test in which two large equal-amplitude sinusoids are
applied to the device under test (DUT) and the level of the intermodulation
signals measured. Except for the DUT generating the distortion being
passive, distortion appears as described in Section 4.5. PIM is observed in
a wide number of situations including coaxial cables [33, 34], microstrip
transmission lines [35], attenuators [36], terminations [37, 38], rectangular
waveguides [39, 40], antennas, and filters [41, 42].

4.7.1 Sources of PIM

Several causes of PIM have been identified or suggested. These include self-
heating and nonlinear junction effects when currents flow from one metal
into a dissimilar metal. The concept being that the dissimilar metals form a
weak diode-like current-voltage characteristic [39].
The most convenient method for measuring small levels of PIM uses
a filter to separate a small distorted intermodulation signal from the two
driving tones. The finite filter response requires that the RF tones be at least
1 MHz apart. However, distortion that is less than 1 MHz from a signal is
of concern in RF systems. Several measurements have shown that the level
of distortion increases rapidly as the distortion component gets closer in
frequency to the driving signal [37].
Figure 4-20 shows the third-order PIM of terminations measured in a two-
tone test [43]. The PIM level is plotted against the frequency separation of
the two tones. Similar results are seen with other components including
transmission lines and antennas [35–38]. The low-PIM termination, resulting
in Curve (a) in Figure 4-20, is the lowest PIM termination available
NOISE, DISTORTION, AND DYNAMIC RANGE 155

Figure 4-20: Measured

third-order passive in-
termodulation distortion
(lower IM3) of common
laboratory high-power,
finned N-type terminations
using a two-tone test at
460 MHz: (a) low-PIM
cable-based termination;
(b) part PE6097, component
A; (c) part PE6097, com-
ponent B; (d) part PE6035,
component A; and (e) part
PE6035, component A.
Measurement taken with
26 dBm input power for
each tone. After [43].

comprising, here, a terminated 100 m long slightly lossy cable with a loss of
0.1 dB/m. The distributed low-level loss spreads the heat generated out over
a considerable distance and so there is a negligible electrothermal PIM effect.
Curves (b) and (c) plot the measured PIM for two otherwise identical finned
N-type connectors. Curves (d) and (e) plot the measured PIM for another
pair of connectors. The reason for the discrepancy is unknown. The rapid
rise in PIM below 200 Hz tone separation is due to electrothermal effects.
The rapid increase in the level of PIM as the tone separation reduces is due
to electrothermal effects. When close in frequency, the two tones produce a
beating waveform that periodically heats a conductor, changing its resistance
through the thermal coefficient of resistance of the conductor. This change in
resistance results in a periodic variation of the current-voltage relation ( i.e.,
Ohm’s law) which results in the generation of intermodulation tones. It has
been shown that if dissimilar metals and magnetic materials are avoided, the
PIM generated is entirely due to electrothermal effects [37].
With tone spacings exceeding a few hundred kilohertz, or in some cases
a few megahertz, the PIM is being generated through another mechanism.
Several sources of this PIM have been postulated, and while all could be
physical sources of PIM, which is most important is not clear. PIM has been
shown to be almost entirely due to current density effects so reducing current
density is an effective means to reduce PIM [44]. Plausible mechanisms
generating such PIM include metal-insulator-metal and metal-metal contact
nonlinearities [45], particularly due to surface topography at contacts. For
example, it is known that high force applied across contacts reduces PIM. In
cellular base stations where the power of signals is very high, it is known to
be important to have very tight connections of cables carrying high-power
signals.
Tunneling at a metal-insulator-metal contact could produce PIM. This
is supposed to be a particular problem with aluminum-oxide-aluminum
contacts, as the oxide thickness may be just right, 2 nm, for tunneling to
occur [40, 45]. However, this is expected to be a minor contributor to PIM.
156 STEER MICROWAVE AND RF DESIGN: MODULES

Another possible source of PIM is thermionic emission [46]. Thermionic

emission is secondary to tunneling and will result in a small increase in
tunneling current due to other sources [46].
The metal-to-metal contact of dissimilar metals acts as a weak diode due to
the difference in the work functions of the metals. Alignment of Fermi levels
requires charge to transfer from the high work function metal to the low
work function metal. Since charge transfer has occurred between the metals,
a field therefore exists at the interface and so there is a contact potential. It is
known that copper, silver, and gold contacts form low PIM contacts and these
metals have very similar work functions, supporting the view above. Also at
metal-to-metal contacts there is a constriction of current flow due to rough
surface topology that exaggerates PIM due to increased current density.
Ferromagnetic materials such as iron, steel, cobalt, and nickel produce
high significant PIM [47]. This is also true for ferroelectric and piezoelectric
materials.

4.7.2 Summary
There are many possible sources of PIM, and provided that care is used
in avoiding dissimilar metals and avoiding ferromagnetic materials, the
only confirmed source of RF PIM is the electrothermal self-heating effect
[35, 37]. However, it is not always possible to build such components as
desirable mechanical, electrical, and packaging requirements necessitate the
use of dissimilar elements. Ferromagnetic metals such as nickel, a desirable
processing material, produce significant PIM. With dissimilar nonmagnetic
materials there is clearly another (unknown) source of PIM at high tone
spacings. Even with carefully designed components, PIM exists at large tone
spacings, and the main source of PIM in these circumstances has not yet
been identified. There is a reasonable confidence that such PIM is a current
nonlinearity and not a voltage nonlinearity [44]. So a strategy for reducing
PIM is to avoid high current densities, and for electrothermal PIM, to provide
rapid dissipation of heat as close to the source of heat as possible. Other
sources of PIM-like responses, that is, RF distortion where it is not expected,
have been found due to vibration [48, 49] and the transient response of a
filter [41, 42].

4.8 Breakdown
Breakdown occurs when the electric field is strong enough that a stray
electron strips electrons from atoms or molecules resulting a cascading effect
producing a conducting plasma of electrons. There are two types of effects
that cause breakdown. One of these is the multipactor effect which occurs at
the interface of a metal or dielectric with a vacuum or low pressure gas where
electrons are stripped from the metal or dielectric. The other is the corona
effect which occurs in a gas, even at fairly low pressures, in which electrons
are stripped from atoms or molecules in the gas. The plasma of electrons
shorts conducting surfaces at different potential, and in the worst case causes
destruction of components. But even before a sufficiently dense plasma is
established, effects include noise generation, increase in ohmic losses, and
nonlinear signal distortion since breakdown is strongly nonlinear effect the
number of free electrons created is a strong nonlinear function of electric field
NOISE, DISTORTION, AND DYNAMIC RANGE 157

and thus voltage.

4.8.1 Multipactor Effect

The multipactor effect is due to the secondary emission of electrons when a
fast moving electron in a vacuum or low pressure gas impacts a metal or a
dielectric. If the energy of the free electron is above a first critical field Ec1 , an
electron in the material will be excited and sometimes will be emitted from
the material. The initial electron will penetrate the material by several lattice
constants but follow a zig-zag path. If the energy of this electron is below
a second critical value Ec2 , then it will only penetrate one or a few lattice
constants and will itself be emitted back into the vacuum (or low pressure
gas). If the electron has an initial energy above Ec2 then it will penetrate
the material a considerable distance and will become trapped. The actual
material and the surface finish have a significant effect on electron emission.
The number of electrons that leave the surface for each impacting electron
is called the secondary electron emission yield (SEY) and this includes the
initial electron which may or may not be emitted. If SEY is greater than one
then the number of free electron will avalanche and could eventually lead
to device destruction. Before that occurs, there is a balancing effect which
limits growth in free electron numbers. Principle among these is the field
that is produced by the free electrons themselves which counters the field
created by an externally applied voltage potential. The reference geometry
used in measuring the multipactor effect is a parallel plate structure with
a voltage V between the plates that are separated by a distance d. This
geometry generally has the lowest thresholds for the multipactor effect. For
other geometries, i.e. non-parallel plate geometries, the high field region
where the secondary electrons are created may not be where the plasma,
i.e. the region where the free electrons accumulate, is located. Thus the
growth in the number of free electrons is curtailed. A further limiting effect
is the frequency of the signal applied to the plates where a high frequency
signal could reverse limiting the build-up of free electrons. Thus the applied
voltage, Vmultipactor , required to initiate the multipactor effect is a function
of d, material, material finish, and frequency f . If there is a free electron
between the plates then the electron will accelerate and acquire velocity and
thus energy. The voltage required to initiate impact ionization. It is found
experimentally that there is a minimum voltage Vmin required for the effect
to be initiated. Then there is a linear relationship between Vmultipactor and
the product f · d. Materials and material finish also have an impact on
ionization For example cleaned mirror-finish silver has a first critical energy,
Ec1 = 130 eV whereas silver exposed to the air has Ec1 = 20 eV. Guidelines
for the thresholds for multipactor ionization for various materials are given
in [50].

4.8.2 Corona Effect

Multipaction describes ionization of materials in a vacuum or in when the
environment has a low pressure gas. As the pressure of the gas increases the
molecules in the gas will breakdown and this is called the corona effect. If
e is an electron and M is a molecule then the corona ionization process is
158 STEER MICROWAVE AND RF DESIGN: MODULES

described by

e + M → e + e + M+ (4.96)

and the molecule takes on a positive charge after an electron has been
stripped off. The competing process which limits the number of free electrons
available is when a free electron combines with a molecule and in air these
interactions principally involve nitrogen N and oxygen O:

e + N+
2 →N+N and e + O2 → O− + O (4.97)

The minimum energy required for the corona effect is heavily dependent
on gas pressure. At very low pressures ionization is solely due to
multipaction. Above a critical but low pressure multipaction is replaced by
the corona effect as as the density of molecules increases it becomes easier to
ionize more molecules. The minimum energy required for the corona effect
occurs when the pressure in multiples of standard atmospheric pressure is
approximately equal to the frequency in gigahertz. Further increase in the
pressure reduces the free main path of an electron and fewer free electrons
are able to acquire the requisition energy. [50].

4.8.3 Summary
Breakdown can lead to device destruction but even at very low levels it
can cause intermodulation products, i.e. passive intermodulation distortion,
that obscure small received signals. Breakdown is a particular problem when
there are high fields, especially in filters with high energy densities because
of resonance. Breakdown can be a problem in basestations, radars, and
satellites as these often operate with EM powers of kilowatts to megawatts,
and even higher powers when operated with low duty cycles.

4.9 Summary
The use of modules has become increasingly important in microwave en-
gineering. A wide variety of passive and active modules are available and
high-performance systems can be realized enabling many microwave sys-
tems of low-to-medium volumes to be realized cost effectively and with
stellar performance. Module vendors are encouraged by the market to de-
velop competitive modules that can be used in a wide variety of applications.
The challenge is to develop a module with high performance and adapt-
ability. Many modules have matched 50 Ω input and output impedances so
that modules can be freely interconnected. Many integrated systems are first
prototyped using vendor-supplied modules that are, perhaps, gradually re-
placed by higher-performance monolithically integrated implementations.
Doing so is only justified when unit volumes are very high. The availabil-
ity of evaluation boards with coaxial connectors for many modules makes it
easy to do early design trade-offs. The great majority of RF and microwave
engineers either develop modules or use modules to realize RF systems.
The most important part of a system design is achieving the desired
functionality largely arrived at by choosing a suitable topology, usually
cascaded subsystems, and appropriate functional units (i.e., modules).
Then the major concern in design is managing noise, distortion, DC
power consumption, cost, and size. The noise and distortion concerns
NOISE, DISTORTION, AND DYNAMIC RANGE 159

can be combined together as dynamic range and then the objective is

maximizing dynamic range while controlling DC power consumption and
using affordable modules. Systems must also be concerned with distortion
that originates in supposedly linear elements. Something as simple as a
termination can produce PIM through heating and thus resistance changes.

4.10 References
[1] H. Nyquist, “Thermal agitation of elec- cation Note 57-2.
tric charge in conductors,” Physical review, [16] A. Victor and M. Steer, “Improved y factor
vol. 32, no. 1, p. 110, 1928. noise measurement using the second stage
[2] D. Chandler, Introduction to modern statistical contribution to advantage,” in 65th ARFTG
mechanics. Oxford University Press, 1987. Conf. Digest, Spring 2005, Jun. 2005, p. 3.
[3] M. Steer, Microwave and RF Design, Transmis- [17] QORVO, “FPD6836P70 data sheet, low
sion Lines, 3rd ed. North Carolina State Uni- noise high frequency packaged en-
versity, 2019. hancement mode phemt transistor,”
[4] H. Nyquist, “Thermal agitation of electric http://www.qorvo.com.
charge in conductors,” Phys. Rev., vol. 32, p. [18] “IEEE Standard 1139-2008, Standard Defini-
110, 1928. tions of Physical Quantities for Fundamental
[5] D. Middleton, An Introduction to Statistical Frequency and Time Metrology—Random
Communication Theory. IEEE Press, 1996. Instabilities,” IEEE, Feb. 2009.
[6] M. Gupta, “Thermal noise in nonlinear resis- [19] B. Mandelbrot and J. Van Ness, “Fractional
tive devices and its circuit representation,” brownian motions, fractional noises and ap-
Proc. of the IEEE, vol. 70, no. 8, pp. 788–804, plications,” SIAM Review, vol. 10, no. 4, pp.
Aug. 1982. 422–437, 1968.
[7] A. Hati, D. Howe, F. Walls, and D. Walker, [20] G. Wornell, “Wavelet-based representations
“Merits of pm noise measurement over noise for the 1/f family of fractal processes,” Proc.
figure: a study at microwave frequencies,” of the IEEE, vol. 81, no. 10, pp. 1428–1450, Oct.
IEEE Trans. on Ultrasonics, Ferroelectrics and 1993.
Frequency Control, vol. 53, no. 10, pp. 1889– [21] G. Wornell and A. Oppenheim, Signal Pro-
1894, Oct. 2006. cessing with Fractals: A Wavelet-Based Ap-
[8] H. Hartnagel, R. Katilius, and A. Matulio- proach. Prentice Hall, 1996.
nis, Microwave Noise in Semiconductor Devices. [22] D. Leeson, “A simple model of feedback os-
Wiley, 2001. cillator noise spectrum,” Proc. of the IEEE,
[9] R. Sarpeshkar, T. Delbruck, and C. Mead, vol. 54, no. 2, pp. 329–330, Feb. 1966.
“White noise in mos transistors and resis- [23] A. Hajimiri and T. Lee, “A general theory of
tors,” IEEE Circuits and Devices Magazine, phase noise in electrical oscillators,” IEEE J.
vol. 9, no. 6, pp. 23–29, Nov. 1993. of Solid-State Circuits, vol. 33, no. 2, pp. 179–
[10] A. Papoulis and S. Pillai, Probability, Random 194, Feb. 1998.
Variables and Stochastic Processes. McGraw- [24] W. Pastori, “Effects of DUT mismatch on
Hill, 1994. the noise figure characterization: a compara-
[11] Y. M. Blanter and M. Büttiker, “Shot noise in tive analysis of two Y-factor techniques,” Mi-
mesoscopic conductors,” Physics Reports, vol. crowave Journal, vol. 26, no. 4, pp. 50–60, Apr.
336, no. 1/2, pp. 1–166, Sep. 2000. 1983.
[12] H. Haus, W. Atkinson, G. Branch, W. Dav- [25] J.-M. Collantes, R. D. Pollard, and M. Sayed,
enport, W. Fonger, W. Harris, S. Harrison, “Effects of dut mismatch on the noise figure
W. McLeod, E. Stodola, and T. Talpey, “Rep- characterization: a comparative analysis of
resentation of noise in linear twoports,” Proc. two Y-factor techniques,” IEEE Trans. on In-
of the IRE, vol. 48, no. 1, pp. 69–74, Jan. 1960. strumentation and Measurement, vol. 51, no. 6,
[13] H. Friis, “Noise figures of radio receivers,” pp. 1150–1156, Dec. 2002.
Proc. of the IRE, vol. 32, no. 7, pp. 419–422, Jul. [26] U. Rohde and J. Whitake, Communications Re-
1944. ceivers. McGraw Hill, 2001.
[14] W. Mumford and E. Scheibe, Noise Per- [27] J. Pedro and C. N.B., Intermodulation Distor-
formance Factors in Communication Systems. tion in Microwave and Wireless Circuits. Nor-
Horizon House, 1968. wood, MA, USA: Artech House, Inc., 2003.
[15] “Noise figure measurement accuracy-the y- [28] J. Sevic and M. Steer, “On the signif-
factor method,” Agilent Technologies, appli- icance of envelope peak-to-average ratio
160 STEER MICROWAVE AND RF DESIGN: MODULES

for estimating the spectral regrowth of on Microwave Theory and Techniques, vol. 53,
an RF/microwave power amplifier,” IEEE no. 8, pp. 2515–2525, Aug. 2005.
Trans. on Microwave Theory and Techniques, [40] C. Vicente, D. Wolk, H. Hartnagel, and D. Ra-
vol. 48, no. 6, pp. 1068–1071, 2000. boso, “An experimental investigation on pas-
[29] N. de Carvalho and J. Pedro, “A compre- sive intermodulation at rectangular wave-
hensive explanation of distortion sideband guide interfaces,” in 2006 IEEE MTT-S Int.
asymmetries,” IEEE Trans. on Microwave The- Microwave Symp. Dig., Jun. 2006, pp. 242–245.
ory and Techniques, vol. 50, no. 9, pp. 2090– [41] G. Mazzaro, M. Steer, and K. Gard, “Inter-
2101, Sep. 2002. modulation distortion in narrowband ampli-
[30] A. Walker, M. Steer, and K. Gard, “Captur- fier circuits,” IET Microwaves, Antennas Prop-
ing asymmetry in distortion of an RF sys- agation, vol. 4, no. 9, pp. 1149–1156, Sep. 2010.
tem using a multislice behavioral model,” [42] G. Mazzaro, M. Steer, K. Gard, and A. Walker,
IEEE Microwave and Wireless Components Let- “Response of RF networks to transient wave-
ters, vol. 16, no. 4, pp. 212–214, 2006. forms: Interference in frequency-hopped
[31] W. Jang, A. Walker, K. Gard, and M. Steer, communications,” IEEE Trans. on Microwave
“Capturing asymmetrical spectral regrowth Theory and Techniques, vol. 56, no. 12, pp.
in RF systems using a multislice behav- 2808–2814, Dec. 2008.
ioral model and enhanced envelop transient [43] J. Wetherington and M. Steer, “Robust ana-
analysis,” Int. Journal of RF and Microwave log canceller for high-dynamic-range radio
Computer-Aided Engineering, vol. 16, no. 4, pp. frequency measurement,” IEEE Trans. on Mi-
400–407, 2006. crowave Theory and Techniques, vol. 60, no. 6,
[32] W. Jang, N. Kriplani, and M. Steer, “Be- pp. 1709–1719, Jun. 2012.
havioural modelling of amplifier asymmetry [44] A. Christianson and W. Chappell, “Measure-
in the time domain,” Int. Journal of Numeri- ment of ultra low passive intermodulation
cal Modelling: Electronic Networks, Devices and with ability to separate current/voltage in-
Fields, 2012. duced nonlinearities,” in 2009 IEEE MTT-
[33] M. Amin and F. Benson, “Coaxial cables as S Int. Microwave Symp. Dig., Jun. 2009, pp.
sources of intermodulation interference at 1301–1304.
microwave frequencies,” IEEE Trans. on Elec- [45] J. Russer, A. Ramachandran, A. Cangellaris,
tromagnetic Compatibility, vol. EMC-20, no. 3, and P. Russer, “Phenomenological modeling
pp. 376–384, Aug. 1978. of passive intermodulation (PIM) due to elec-
[34] M. Amin and I. Benson, “Nonlinear effects tron tunneling at metallic contacts,” in 2006
in coaxial cables at microwave frequencies,” IEEE MTT-S Int. Microwave Symp. Dig., Jun.
Electronics Letters, vol. 13, no. 25, pp. 768–770, 2006, pp. 1129–1132.
8 1977. [46] J. Simmons, “Generalized thermal j-v charac-
[35] J. Wilkerson, P. Lam, K. Gard, and M. Steer, teristic for the electric tunnel effect,” Applied
“Distributed passive intermodulation dis- Physics, vol. 35, no. 9, pp. 2655–2658, Sep.
tortion on transmission lines,” IEEE Trans. 1964.
on Microwave Theory and Techniques, vol. 59, [47] J. Henrie, A. Christianson, and W. Chap-
no. 5, pp. 1190–1205, May 2011. pell, “Engineered passive nonlinearities for
[36] J. Wilkerson, K. Gard, and M. Steer, “Electro- broadband passive intermodulation distor-
thermal passive intermodulation distortion tion mitigation,” IEEE Microwave and Wireless
in microwave attenuators,” in 36th European Components Letters, vol. 19, no. 10, pp. 614–
Microwave Conf., Sep. 2006, pp. 157–160. 616, Oct. 2009.
[37] J. Wilkerson, K. Gard, A. Schuchinsky, and [48] J. Wetherington and M. Steer, “Standoff
M. Steer, “Electro-thermal theory of inter- acoustic modulation of radio frequency sig-
modulation distortion in lossy microwave nals in a log-periodic dipole array antenna,”
components,” IEEE Trans. on Microwave The- IEEE Antennas and Wireless Propagation Let-
ory and Techniques, vol. 56, no. 12, pp. 2717– ters, vol. 11, no. 8, p. 1, Aug. 2012.
2725, Dec. 2008. [49] A. Hati, C. Nelson, and D. Howe, “Effect
[38] J. Wilkerson, K. Gard, and M. Steer, “Auto- of vibration on P and AM noise of oscilla-
mated broadband high-dynamic-range non- tory and non-oscillatory components at 10
linear distortion measurement system,” IEEE GHz,” in 2009 IEEE Int. Frequency Control
Trans. on Microwave Theory and Techniques, Symp., Joint with the 22nd European Frequency
vol. 58, no. 5, pp. 1273–1282, may 2010. and Time Forum, Apr. 2009, pp. 524–529.
[39] C. Vicente and H. Hartnagel, “Passive- [50] “E-20-01A: ECCS Standard on Multipaction
intermodulation analysis between rough and Test,” http://www.ecss.nl.
rectangular waveguide flanges,” IEEE Trans.
NOISE, DISTORTION, AND DYNAMIC RANGE 161

4.11 Exercises
1. An amplifier consists of three cascaded stages 10. A receive amplifier with a gain of 30 dB, a noise
with the following characteristics: figure of 2 dB, and bandwidth of 5 MHz is con-
Stage 1 Stage 2 Stage 3 nected to an antenna which has a noise temper-
ature of 20 K. [Parallels Example 4.2]
Gain (dB) −3 15 5
NF (dB) 3 2 2 (a) What is the available noise power presented
to the input of the amplifier in the 5 MHz
(a) What is the overall gain of the amplifier?
bandwidth (recall that the antenna noise
(b) What is the overall noise figure of the ampli-
temperature is 20 K?
fier?
(b) If instead the input of the amplifier is con-
2. Briefly describe the effect of a lossy filter on nected to a resistor held at standard temper-
SNR. Consider signals at the input and output ature, what is the available noise power pre-
of the filter. sented to the input of the amplifier in the
3. What is the available noise power of a 50 Ω re- 5 MHz bandwidth?
(c) What is the noise factor of the amplifier?
sistor in a 10 MHz bandwidth. The resistor is at
standard temperature. (d) What is the excess noise power of the ampli-
fier referred to the its output?
4. A 50 Ω resistor a 20 Ω resistor are in shunt. If (e) What is the effective noise temperature of
both resistors have a temperature of 300 K, what the amplifier when the amplifier is con-
is the total available noise power spectral den- nected to the antenna with a noise temper-
sity of the shunt resistors? ature of 20 K. That is, what is the effec-
5. The thermal noise power at the output of a sys- tive noise temperature of the resistor in the
tem is 1 fW and the shot noise power is 1 fW. Thevenin equivalent circuit of the amplifier
What is the available white noise power? output?

6. A 2 GHz amplifier in a 50 Ω system has a band- 11. A receive amplifier has a bandwidth of 5 MHz,
width of 10 MHz, a gain of 40 dB, and a noise a 1 dB noise figure, a linear gain of 20 dB. The
figure of 3 dB. The amplifier is driven by a circuit minimum acceptable SNR is 10 dB.
with a Thevenin equivalent resistance of 50 Ω (a) What is the output noise power in dBm?
held at 290 K (standard temperature). What is (b) What is the minimum detectable output sig-
the available noise power at the output of the nal in dBm?
amplifier? (c) What is the minimum detectable input sig-
7. A 30 dB attenuator is terminated at Port 2 in a nal in dBm?
matched resistor and both are at 290 K. What is 12. A 75 Ω attenuator has a loss of 16 dB and is be-
the noise temperature at Port 1 of the attenuator? tween a source with a Thevenin impedance of
8. A 20 dB attenuator is terminated in a matched 75 Ω and a load of 75 Ω.
resistor and both are held at 30◦ C. What is the (a) What is the noise power, Ni , available from
noise temperature at the input of the attenuator the 75 Ω source resistor at standard temper-
in kelvin? ature (290 K) in a 1 MHz bandwidth?
(b) Now consider that the source is connected
9. The effective noise temperature at the coaxial
to the attenuator which is also connected to
output of an antenna is 100 K. The antenna is
the load. If the source generates a modulated
connected to a bandpass filter with a bandwidth
signal that is 1 MHz wide and has an avail-
of 20 MHz and an insertion loss of 1 dB. [Paral-
able power, Si , of 10 fW, what is SNRi at the
lels Examples 4.2 and 4.3]
input to the attenuator at standard tempera-
(a) What is the available noise power in a ture?
20 MHz bandwidth at the output of the an- (c) With the attenuator connected to the source,
tenna? what is the Thevenin equivalent impedance
(b) What is the noise figure of the bandpass fil- looking into the output of the attenuator?
ter (consider only the passband)? (d) Calculate the noise power, No , available
(c) What is the excess noise power at the output from the attenuator with the source attached
of the filter? (Consider only the passband). at standard temperature (290 K) in a 1 MHz
(d) What is the total available noise power in the bandwidth?
passband at the output of the filter? (e) What is the signal power, So , delivered to
162 STEER MICROWAVE AND RF DESIGN: MODULES

the load? (b) If the amplifier has sufficient bandwidth to

(f) What is the SNR at the load, SNRo ? amplify both the RF and image frequency,
(g) What is the noise factor of the attenuator? what is the noise figure of the cascade?
(h) What is the noise figure of the attenuator? Note that the overall noise figure is a single-
13. The system shown below is a receiver with sideband noise figure.
bandpass filters, amplifiers, and a mixer. [Paral- 17. The first stage of a two-stage amplifier has a lin-
lels Example 4.4] ear gain of 40 dB and a noise figure if 3 dB. The
second stage has a gain of 10 dB and a noise fig-
ure of 5 dB.
(a) What is the overall gain of the amplifier?
(b) What is the overall noise figure of the ampli-
fier?
(a) What is the total gain of the system? 18. A subsystem consists of a matched filter with an
(b) What is the noise factor of the first filter? insertion loss of 2 dB then an amplifier with a
(c) What is the system noise factor? gain of 20 dB and a noise figure, NF, of 3 dB.
(d) What is the system noise figure? (a) What is the overall gain of the subsystem?
(b) What is NF of the filter?
14. An amplifier consists of three cascaded stages
(c) What is NF of the subsystem?
with the following characteristics:
19. A subsystem consists of a matched amplifier
Stage 1 Stage 2 Stage 3 with a gain of 20 dB and a noise figure of 2 dB,
Gain 10 dB 15 dB 30 dB followed by a 2 dB attenuator, and then another
NF 0.8 dB 2 dB 2 dB amplifier with a gain of 10 dB and NF of 3 dB.
(a) What is the overall gain of the subsystem?
What is the noise figure (NF) and gain of the cas-
(b) What is NF of the attenuator?
cade amplifier?
(c) What is NF of the subsystem?
15. The front end of a receiver for a cellular phone
20. Consider a digitally modulated signal and
has a bandpass filter with a 25 MHz passband
briefly describe the impact of a nonlinear am-
and a loss in the passband of 2 dB and is fol-
plifier on the signal. You must include several
lowed by two amplifier stages. The first stage
negative effects. Use one or more diagrams.
has a gain of 20 dB and a noise figure of 0.5 dB
and the second stage has a gain of 60 dB and a 21. An amplifier has a linear gain of 30 dB and an
noise figure of 2 dB. output-referred 1 dB gain compression point of
13 dBm. What is the input-referred 1 dB gain
(a) Sketch the system block diagram.
compression point of the amplifier?
(b) What is the gain of the system?
(c) What is the noise figure of the filter? 22. An amplifier has a linear gain of 30 dB and an
(d) What is the noise figure of the system? input-referred 1 dB gain compression point of
(e) The system is now connected to an antenna −30 dBm. What is the output-referred 1 dB gain
with an effective noise temperature of 30 K compression point of the amplifier?
that delivers a signal of 10 pW to the band- 23. An amplifier has an output power of 10 dBm
pass filter. Determine the noise temperature when the gain of a single tone is compressed by
at the output of the system and hence the 1 dB. What is the maximum output power of an
output noise power in the 25 MHz band- undistorted 64-QAM signal? (A 64-QAM signal
width. First calculate the excess noise tem- which has a PMEPR of 7.8 dB. [Parallels Exam-
perature added by the system to the output. ple 4.6]
1. Determine the signal-to-noise ratio at the 24. The input-referred 1 dB gain compression point
output of the front-end system. of an amplifier with a linear gain of 30 dB is
16. The RF front end of a communications unit con- 0 dBm.
sists of an amplifier followed by a mixer. The (a) What is the gain of the amplifier at 1 dB gain
amplifier has a gain of 20 dB and a noise figure compression?
of 4 dB. The mixer has a conversion loss of 6 dB (b) What is the output power at 1 dB gain com-
and a double-sideband noise figure of 8 dB. pression?
(a) Why is the double-sideband noise figure (c) Consider amplifying an 8-PSK signal with a
sometimes used with a mixer but not with PMEPR of 3.3 dB. What is the maximum out-
an amplifier? put power of the undistorted 8-PSK signal?
NOISE, DISTORTION, AND DYNAMIC RANGE 163

Maximum acceptable distortion is when the the output of the amplifier is −80 dBm. What is
envelope peak is compressed by 1 dB gain. the spurious free dynamic range of the receiver?
25. The gain of an amplifier at the 1-dB gain com- 31. A power amplifier has a linear gain of 20 dB,
pression point is 40 dB and the input power is an output 1 dB gain compression point of
−7 dBm. 30 dBm, and an output-referred third-order in-
(a) What is the power of the amplifier’s output tercept point OIP3 = 60 dBm. The noise floor at
signal? the output of the amplifier is −70 dBm. What
(b) If the power input to the amplifier is re- is the dynamic range of the amplifier if the re-
duced to −20 dBm, what is the amplifier’s quired minimum SNR at the output is 6 dB?
output power now?
32. A room-temperature two-stage amplifier in a re-
26. Briefly describe intermodulation distortion with ceiver has a bandwidth of 100 MHz, a noise fig-
a a two-tone signal. Use a diagram. ure of 3 dB, a linear gain of 32 dB, and an output-
27. Briefly describe what is meant by 1 dB gain com- referred third-order intercept point, OIP3, of
pression. Use a diagram. 27 dBm. The minimum SNR of the receiver sys-
29. A single-stage amplifier has a linear gain of tem is 16 dB.
16 dB, an output 1 dB gain compression point (a) What is the output noise power in dBm?
of 10 dBm, and an output-referred third-order (b) What is the difference between the input-
intercept point OIP3 = 30 dBm. The noise floor and output-referred spurious free dynamic
at the output of the amplifier is −60 dBm. The ranges?
communication protocol has a minimum SNR, (c) What is the SFDR in dB?
SNRMIN , of 6 dB. (d) What is the difference between the input-
(a) What is the dynamic range of the amplifier? and output-referred dynamic ranges?
(b) What is the SFDR of the amplifier? (e) What is the minimum detectable output sig-
30. A receiver system comprising a filter and two nal in dBm?
cascaded amplifiers has an overall linear gain of (f) What is the output-referred DR in dB?
80 dB, an output 1 dB gain compression point of 33. When determining the dynamic range of an am-
−10 dBm, and an output-referred third-order in- plifier the gain compression level is not used.
tercept point, OIP3 = 10 dBm. The noise floor at Briefly discuss why.

4.11.1 Exercises by Section

†
challenging, ‡ very challenging
§4.2 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 §4.5 20, 21, 22, 23, 24, 25, 26, 27, 28
§4.3 12† , 13, 14, 15† , 16† , 17, 18, 19 §4.6 29, 30, 31, 32, 33

4.11.2 Answers to Selected Exercises

19(b) 60 dB 12(g) 39.8
13(d) 5.17 dB 17(b) 3 dB
 CHAPTER 5

Passive Modules

5.1 Introduction to Passive Modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165

5.2 Diodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
5.3 Case Study: Tunable Resonator with Varactor Diode Stack . . . . 170
5.4 Switch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
5.5 Ferrite Components: Circulators and Isolators . . . . . . . . . . . . . . . . . 177
5.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
5.7 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
5.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182

5.1 Introduction to Passive Modules

Most RF systems are composed of a cascade of modules each of which is
separately designed and characterized. They usually have matched 50 Ω
input and output impedances so that modules can be freely interconnected.
In high volume applications, several modules could be monolithically
integrated, but even then, design is based on the concept of modules. Many
modules are available “off the shelf” and high-performance RF systems can
be constructed using commercially available modules, see Figure 5-1.
Many modules are available mounted on evaluation boards complete
with biasing components and SMA connectors so that RF modules can be
interconnected using coaxial cables. For example, an evaluation board is
shown in Figure 5-2 for a power amplifier. The engineer needs to attach
supply, Vs , ground, and control voltage Vctl , and the evaluation board
provides a working amplifier. Most RF and microwave engineers work
at the circuit board level and begin system design using modules. Some
companies develop some of their own proprietary modules, thus providing
a competitive advantage, but still use many modules developed by others.
An estimate sometimes made is that the value of RF and microwave circuit-
board-based module-using engineering is 30 times larger than the value of
module and RF integrated circuit engineering.
The modules that comprise a receiver are shown in Figure 5-3. Beginning
with the bandpass filter after the antenna, each module contributes noise and
nonlinear distortion. The system design objectives are generally to maximize
dynamic range, the region between the signal being sufficiently above the
noise level to be detected but before nonlinear distortion introduces spurious
166 STEER MICROWAVE AND RF DESIGN: MODULES

(a) Resistor (b) Mixer (c) Synthesizer (d) Transformer

Figure 5-1: Modules in surface-mount packages. Copyright Synergy Microwave Corporation,

used with permission [1].

(a) Evaluation board

Figure 5-2: Evaluation board for

the HMC414MS8G GaAs InGaP
HBT MMIC power amplifier
module operating between 2.2
and 2.8 GHz. The amplifier pro-
vides 20 dB of gain and +30 dBm
of saturated power at 32% PAE
from a +5V supply voltage. The
amplifier can also operate with
a 3.6 V supply selectable by the
resistors R1 and R2 . Copyright
Hittite Microwave Corporation,
used with permission [2]. (b) Schematic
RF AND MICROWAVE MODULES 167

Figure 5-3: Receiver as a cas-

cade of modules.

Component Symbol Table 5-1: IEEE standard symbols for

1 diodes and a rectifier [3]. (1 In the
Diode, general (including Schottky) direction of anode (A) to cathode (K).
2
IMPATT diode1 Use symbol for general diode unless
it is essential to show the intrinsic
Gunn diode region.)
PIN diode1,2
Light emitting diode (LED)1
Rectifier1
Tunnel diode1

Varactor diode1 or
Zener diode1

signals that limit the detectability of signals. At the same time, it is usual
to try and minimize the power consumed. This is particularly important in
mobile applications.
This chapter introduces many passive modules;modules and microwave
functional elements that do not require external supplies with he exception
of perhaps control voltages..

5.2 Diodes
Diodes are two-terminal devices that have nonlinear current-voltage
characteristics. The most common diodes used in microwave engineering
are listed in Table 5-1 along with their standard symbols. Ideally a rectifier
allows current to flow in one direction and not in the other. A general diode,
usually a junction diode or a Schottky diode, is one type of rectifier, but
rectifiers are more general, and, for example, they can be realized using
vacuum devices.

Junction and Schottky Diodes

Junction diodes are two-terminal devices that derive their characteristics
from the barrier effect that occurs at the junction of two different types
of semiconductor (one with excess free holes and one with excess free
electrons), or at the interface of a metal and a semiconductor. The result is
an asymmetric current-voltage characteristic, as shown in Figure 5-4(a). This
is not an ideal rectifier characteristic, as it requires a threshold voltage to be
reached before there is appreciable current flow. Semiconductor-based diode
168 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 5-4:
Characteristics of a
pn junction diode
or a Schottky diode:
(a) current-voltage
characteristic; (b)
capacitance-voltage
characteristic; and (c)
diode model. (a) (b) (c)

characteristics are described by [4–6]

   
qV
I = I0 exp −1 , (5.1)
nkT

where V is the voltage across the junction, q(= −e) is the absolute value of
the charge of an electron, k is the Boltzmann constant (1.37 · 10−23 J/K), and
T is the absolute temperature (in kelvin). I0 is the reverse saturation current
and is small, with values ranging from 1 pA to 1 nA. The quantity n is the
diode ideality factor, with n = 2 for graded-junction pn junction diodes and
n = 1.0 for step-junction diodes where the interface between p-type and n-
type semiconductor materials is abrupt. The abrupt junction is most closely
realized by a Schottky diode, where a metal forms one side of the interface
(typically) replacing the p-type semiconductor. Carriers recombine quickly
in the metal, much faster than they would in a semiconductor. Hence a
Schottky diode operates at higher frequencies than does a pn junction diode.
A unity ideality factor, n, of 1 is the best that can be achieved, resulting in
the strongest nonlinearity. When the applied voltage is sufficiently positive
to cause a large current to flow, the diode is said to be forward biased. When
the voltage is negative, the current flow is negligible and the diode is said
to be reverse biased. At sufficiently large reverse bias, electrons are ripped
from the valence bands of the semiconductor atoms and the current rapidly
increases in a process called impact ionization or avalanche. The voltage at
which this occurs is called the reverse breakdown voltage.
In a semiconductor diode, charge is separated over distance and so a diode
has appreciable capacitance, called the junction capacitance, mathematically
modeled as
Cj0
Cj (V ) = γ, (5.2)
(1 − (V /φ))
where φ is the built-in potential difference across the diode. This capacitance
profile is shown in Figure 5-4(b). The built-in potential is typically 0.6 V for
silicon diodes and 0.75 V for GaAs diodes. The doping profile can be adjusted
so that γ can be less than the ideal 12 of an abrupt junction diode.
Current must flow through bulk semiconductor before reaching the active
region of the semiconductor diode, and so there will be a resistive voltage
drop. Combining effects leads to the equivalent circuit of a pn junction or
Schottky diode shown in Figure 5-4(c).

Varactor Diode
A varactor diode is a pn junction diode operated in reverse bias and
optimized for good performance as a tunable capacitor. Ideally it has low
RF AND MICROWAVE MODULES 169

reverse saturation current, high reverse breakdown voltage, and a specific

capacitance profile designed for a particular application.
A common application of a varactor diode is as the tunable element
in a voltage-controlled oscillator (VCO) where the varactor, with voltage-
dependent capacitance, C, is part of a resonant circuit (often called a tank
circuit) with a lumped√inductor, L. The resonant frequency of the tank circuit
is proportional to 1/ LC, so by applying a voltage to a varactor diode,
C changes, the resonant frequency of the tank circuit is tuned, and the
oscillation frequency is changed. The capacitance versus voltage is described
by Equation (5.2) and voltage dependence is shown in Figure 5-4(b).

PIN Diode
A PIN diode is a variation on a pn junction diode with a region of intrinsic
semiconductor (the I in PIN) between the p-type and n-type semiconductor
regions. The properties of the PIN diode depend on whether there are
carriers in the intrinsic region. The PIN diode has the current-voltage
characteristics of a pn junction diode at low frequencies; however, at high
frequencies it looks like a linear resistor, as carriers in the intrinsic region
move slowly. When a forward DC voltage is applied to the PIN diode, the
intrinsic region floods with carriers, and at microwave frequencies the PIN
diode is then modeled as a low-value resistor. At high frequencies there is
not enough time to remove the carriers in the intrinsic region, so even if
the total voltage (DC plus RF) across the PIN diode is negative, there are
carriers in the intrinsic region throughout the RF cycle. If the DC voltage is
negative, carriers are removed from the intrinsic region and the diode looks
like a large-value resistor at RF. The PIN diode can be used as a microwave
switch controlled by a DC voltage.

Zener Diode
Zener diodes are pn junction or Schottky diodes that have been specially
designed to have sharp reverse breakdown characteristics [5]. They can
be used to establish a voltage reference or, used as a limiter diode, to
provide protection of more sensitive circuitry. As a limiter, they are found in
communication devices in a back-to-back configuration to limit the voltages
that can be applied to sensitive RF circuitry.

LED Diode
In semiconductor diodes, the recombination of holes and electrons can result
in photons being generated. This effect is enhanced to form light-emitting
diodes (LEDs) [4, 5, 7, 8].
170 STEER MICROWAVE AND RF DESIGN: MODULES

5.3 Case Study: Tunable Resonator with Varactor Diode

Stack
Varactors provide a tunable capacitance with the bias voltage across the
varactor, a reverse biased diode, changing the capacitance of the diode.
The applied bias voltage also changes the series resistance of the diode,
see Figure 5-4. The characteristics of a microwave varactor are shown in
Figure 5-5. At a reverse bias voltage VR = 1 V the junction capacitance
Cj = 7 pF and the series resistance is 1.8 Ω. At 5 GHz the diode impedance
ZD (1 V ) = 1.8 − 4.54 Ω and its Q(1 V ) = 2.5. At a reverse bias voltage
VR = 10 V, the junction capacitance iCj = 1 pF, and the series resistance
is 0.4 Ω. At 5 GHz the diode impedance ZD (1 V ) = 0.4 − 31.8 Ω and its
Q(1 V ) = 80.
Generally a design proceeds based on a chosen system impedance which is
often 50 Ω so that the design space includes realizable impedances above and
below a system impedance. One way of effectively increasing the impedance
of a varactor diode is to stack diodes. This is shown in Figure 5-6 where four

(a)

Figure 5-5: Characteristics of a varactor

diode (Toshiba model JDV2S71E1, a sil-
icon epitaxial planar diode) as a function
of reverse bias voltage VR : (a) capacitance
Cj ; and (b) series resistance RS . (b)

Figure 5-6: A stub-loaded diode stack

resonator designed to produce a variable
capacitance in the 4–6 GHz range. The
element labeled TL1 is a low impedance
microstrip line. The choke inductor,
LCHOKE = 8.2 nH, presents an RF open
circuit and is part of the bias circuit.
Each varactor diode (D1 –D4 ) is model
JDS2S71E.
RF AND MICROWAVE MODULES 171

Figure 5-7: Equivalent circuits of the stub-

loaded diode stack resonator of Figure 5-
(a) (a) (a) 6.

varactor diodes D1 , . . . , D4 are shown together with the biasing arrangement

so that the each of the diodes can be adjusted by the single tuning voltage
Vtune applied through the 3.6 nH and 8.2 nH choke inductors. The choke
inductors provide a DC short and an RF open (actually an impedance
magnitude of 258 Ω for the 8.2 nH inductor at 5 GHz). Figure 5-6 is the
schematic of a tunable resonator, used between 4 and 5.3 GHz to tune an
oscillator, see [9] and Section 5.6 of [10].
The two 0.5 pF capacitors in Figure 5-6 scale the capacitance of the
diode stack and increase the effective Q of the variable capacitance. The
shorted transmission line presents an inductance impedance in shunt with
the augmented diode stack. A lumped inductor is not used here as it would
have too high a loss. The equivalent circuit of the resonator is shown in
Figure 5-7. The resonator network is designed to be resonant at a frequency
above the operating frequency. At the oscillation frequency the resonator
network must present an inductance to resonate with the capacitance of the
active device in an oscillator. In addition, for oscillator stability, the effective
inductance must have a particular derivative of admittance with respect to
frequency.
The two 0.5pF capacitors between TL1 and the varactor stack combine
to provide an effective capacitance of 0.25 pF which has an impedance of
127 Ω at 5 GHz. So it is not a DC blocking capacitor which would have a
very low impedance at RF, but it does block DC so that Vtune is not shorted
to ground through the transmission line. The RF function of the two series
0.5 pF capacitors is to reduce the value of the variable capacitance, increase
the operating reactance, and increase the Q of the effective capacitance.
Each varactor diode has a 0 V capacitance of 11 pF and a capacitance at
9 V of 2.15 pF. With the four varactors in series this becomes 2.75 pF total
capacitance at 0 V and 0.54 pF at 9 V. The 0.25 pF capacitance in series reduces
this to 0.229 pF at 0 V tuning voltage and 0.171 pF at 9 V tuning voltage for a
capacitance tuning ratio of 0.229 pF/0.171 pF = 1.34. The capacitance tuning
range has been reduced from the original ratio of 2.75 pF/ 0.54 pf = 5.05.
The tuning range of the resonant frequency of a parallel LC resonator is the
square
√ root of this, i.e. 1.16 (which comes from the resonant frequency being
1/ LC). If this is not enough the two 0.5 pF capacitors can be changed to
achieve the required overall capacitance tuning ratio.
The 10 pF capacitor between the Vtune terminal and ground is there to
ensure an RF short circuit. The magnitude if the impedance of this capacitor
at 5 GHz is 3.2 Ω, a factor of ten away from the 50 Ω system impedance.
While at RF the terminal at Vtune is ideally a short circuit to ground, there
will always be some wiring inductance.
The characteristics of the resonator can be understood by looking at the
plot of the simulated resonator reflection coefficient, Γr , on a Smith chart,
see Figure 5-8. Γr is shown for two tuning voltages as well as the device
locus of the inverse of the input reflection coefficient of a transistor circuit
172 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 5-8: Simulated re-

flection coefficient of the
resonator, Γr on a 50 Ω
Smith chart for two tun-
ing voltages (1 V and 3 V).
The drive locus shown is
the inverse of the input re-
flection coefficient of an ac-
tive network that would be
connected at the same port
as for Γr . The intersection
of Γr and the drive lo-
cus determines the operat-
ing frequency.

connected to the resonator. The combination of the resonator and the active
device forms an oscillator. The intersection of the active device and resonator
loci selects the oscillation frequency of the oscillator. The oscillator presents a
capacitance to the resonator and so the resonator must present an inductance
to the active device. The frequency of the oscillator is varied by changing Γr
of the resonator.
A photograph of the resonator is shown in Figure 5-9. The measured Γr is
shown in Figure 5-10 and this corresponds to the simulated plot in Figure 5-8.
The two plots are roughly in agreement but there is a discrepancy because it
is very hard to do these measurements at 5 GHz, and the models of the device
components are not exact. However the main reason for the discrepancy is
that this is a measurement of a resonant circuit and small discontinuities can
have a dramatic effect. What is presented here is typical of microwave circuit
design and why tremendous intuition is required to design at microwave
frequencies and to interpret measurements.

5.4 Switch
Microwave switches are commonly used to alternately connect an antenna
to a transmitter or a receiver. In some communication systems, such as
GSM and some modes of 4G and 5G, a phone does not transmit and
receive simultaneously. Consequently a switch can be used to separate the
transmitted and received signals. In multiband phones, a switch is used
to connect the correct transmitter and receiver, which are band specific,
to the antenna. In radar systems, switches are used to steer an antenna
beam by changing the phase of the microwave signal delivered to each
antenna in an array of antennas. An ideal microwave switch is shown in
RF AND MICROWAVE MODULES 173

Figure 5-9:
Photograph of
the microstrip
resonator net-
work with surface
mount compo-
nents. The shorting
bar is a 0 Ω surface
mount resistor and
is used to adjust
the effective length
of the transmission
line resonator, TL1 .

Figure 5-10: Measured reflec-

tion coefficient of the res-
onator network, Γr . Curve a,
is the characteristic of the res-
onator network for a tuning
voltage, Vtune , of 0 V. Curve
g is for Vtune = 9 V. Vtune is
equally spaced for Curves a–
g. The inverse of the reflec-
tion coefficient of an active de-
vice connected atthe Γr point
is shown and indicates that
the active device oresents a
negative conductance. The in-
tersection of the device locus
and Γr determines the oscilla-
tion frequency of an oscillator.
Thus adjusting Vtune changes
the oscilaltion frequency.
174 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 5-11(a), where an input port, RFIN , and an output port, RFOUT , are
shown. For maximum power transfer between the ports the switch should
have little loss and thus have low on resistance. At microwave frequencies,
realistic switches must be modeled with parasitics and with finite on and
off resistances. A realistic model applicable to many switch types is shown
in Figure 5-11(b). The capacitive parasitics, the CP s, limit the frequency of
operation of the switches and the on resistance, RON , impacts the switch
loss. Ideally the off resistance, ROFF , is very large, however, the parasitic
shunt capacitance, COFF , is nearly always more significant. The result is that
at high frequencies there is an alternative capacitive connection between the
input and output through COFF . The on resistance of the switch introduces
voltage division that can be seen by comparing the ideal connection shown
in Figure 5-11(c) and the more realistic connection shown in Figure 5-11(d).
From the voltage division ratio, the loss of the switch can be calculated.
Switches are configured to provide connections from one or more inputs to
one or more outputs. The configuration of a switch is indicated by poles and
throws, and several configurations are shown in Figure 5-12. In microwave
applications, single-pole switches are most commonly used and the input is
connected to an antenna. For example, the throws would be connected to
different bands of a multiband phone .

Figure 5-11: Microwave switches: (a) ideal switch connecting RFIN and RFOUT ports; (b) model
of a microwave switch; (c) ideal circuit model with switch on and with source and load; (d)
realistic low-frequency circuit model with the switch on; (e) switch realized using a PIN diode;
(f) switch realized using an FET; and (g) switch realized using a MEMS switch.
RF AND MICROWAVE MODULES 175

Figure 5-12: Switch configurations.

Table 5-2: Typical properties of small microwave switches. (Sources: 1 Radant MEMS, 2 RF Micro
Devices, 3 Tyco Electronics,and 2 pSemiTM Corporation.)
Switch Configuration Power Maximum Operating Actuation Response
type handling insertion loss frequency voltage time
MEMS1 SPDT 0.5 W 0.5 dB to 10 GHz 90 V 10 µs
MEMS1 SPST 4W 0.8 dB to 35 GHz 110 V 10 µs
FET (pHEMT)2 SPDT 10 W 0.3 dB to 6.5 GHz 5V 0.5 µs
FET (pHEMT)2 SPDT 0.3 W 1.1 dB to 25 GHz 5V 0.5 µs
PIN3 SPDT 13 W 0.35 dB to 2 GHz 12 V 0.5 µs
PIN3 SPDT 10 W 0.4 dB to 6 GHz 12 V 0.5 µs
FET (SOI)4 SPDT 2W 2.7 dB 60 GHz 3.6 V 8 ns

There are four main types of microwave switches: mechanical, PIN diode,
FET, and microelectromechanical system (MEMS) switches. Mechanical
switches are nearly lossless but tend to be large, relatively expensive, slow to
operate, and are mostly used in laboratory settings. The other switches are
of most interest for use in systems. The PIN diode, FET, and MEMS switches
are shown in Figures 5-11(e–g), respectively. With these technologies, most
higher-order switches are based on interconnections of SPST switches. The
attributes of these switches are summarized in Table 5-2 for switches that are
suitable for cell phone applications. PIN diode switches are the most robust,
handling the most RF power, and operating at higher frequencies than either
FET- or MEMS-based switches. However, this comes at a price. The PIN
diode used is similar to a pn junction diode with the addition of an intrinsic
layer between the p- and n-type materials. With applied forward bias the
diode has low RF series resistance. In reverse bias, the diode RF resistance
is large. Forward bias requires DC current and voltage, so control power is
consumed when a PIN diode switch is on. The circuit configuration for an
SPST PIN diode switch is shown in Figure 5-11(e). Series bias-decoupling
capacitors are required at the RF ports.
A FET makes a good electronic switch; with the correct bias applied to
the gate, the drain-source connection looks like a small resistance. Changing
the bias to the other extreme removes free carriers from the channel between
the drain and source, and a large resistance is the result. Both Si and GaAs
176 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 5-13: RF MEMS

switch: (a) the RFin
line in contact with the
RFout line; and (b) the
cantlever beam electro-
statically attracted to
the pedestal and there
is no RF connection. (a) On (b) Off

Figure 5-14: Model used in calculating the

loss of a switch in a 50 Ω system.

switches are used in cellular handsets, with GaAs switches operating at

extended frequencies approaching 6 GHz. The operation of a FET can be
described as a variable drain-source resistance with the gate-source voltage
controlling the cross section of the channel. The circuit for a FET-based SPST
switch is shown in Figure 5-11(f). Series bias-blocking capacitors are required
at the RF ports. Control power is only required to change the state of the
switch; negligible power is required to maintain the switch state.
A MEMS switch is fabricated using photolithographic techniques similar
to those used in semiconductor manufacturing [11–13]. They are essentially
miniature mechanical switches with a voltage used to control the position of
a shorting arm, which is usually a cantilever or a membrane, see Figure 5-13.
As there is no direct connection between the RF signal path and the control
circuitry, MEMS switches have inherently high operating frequencies. Power
is required to change the switch, but once switching has been accomplished,
negligible DC power is required to maintain the connection. Two of the
major problems are creating the contact with enough force to reduce the
series resistance, and maintaining the structural integrity of the contacts after
hundreds of millions of cycles. Generally it is not necessary to make a DC
contact, and creating a structure that has a large difference between the on
capacitance and off capacitance is sufficient for applications such as tunable
matching networks [11].

EXAMPLE 5.1 Insertion Loss of a Switch

What is the insertion loss of a switch with a 1 Ω on resistance when it used in a 50 Ω system?
Solution:
The model to be used for evaluating the insertion loss of the switch is shown in Figure 5-14.
The insertion loss is found by first determining the available power from the source and
then the actual power delivered to the load. The available input power is calculated by first
ignoring the 1 Ω switch resistance. Then there is maximum power transfer from the source
to the load. The available input power is
 1 2
1 2E E2
PAi = = , (5.3)
2 50 400
RF AND MICROWAVE MODULES 177

where E is the peak RF voltage at the Thevenin equivalent source generator. The power
delivered to the 50 Ω load is found after first determining the peak load voltage:
50 50
VL = E= E. (5.4)
50 + 1 + 50 101
Thus the power delivered to the load is
2
1 VL2

1 50
PD = = E2. (5.5)
2 50 100 101
The insertion loss is
2
E 2 100

PAi 101
IL = = = 1.020 = 0.086 dB. (5.6)
PD 400 E 2 50

5.5 Ferrite Components: Circulators and Isolators

Circulators and isolators are nonreciprocal devices that preferentially route
RF signals [14–16]. The essential element of a circulator is a slab or disc of
ferrite which, when magnetized, becomes nonreciprocal, with a preferred
direction of propagation resulting from what is called the gyromagnetic
effect.

5.5.1 Gyromagnetic Effect

An electron has an intrinsic magnetic field described by its magnetic moment
m (see Figure 5-15(a)) [17]. This is due to a quantum mechanical property
called spin bu the electron does not actually spin. In most materials electron
spin occurs in pairs with the magnetic field produced by one electron’s spin
being canceled by the spin of the other electron. However, in some materials
the spin does not occur in pairs and there are regions in which the small
magnetic fields produced at the atomic level line up to produce organized
regions of directed magnetic field. These regions are called domains, or
sometimes magnetic domains, and the combined magnetic field from the
aligned electron spins is called a magnetic moment.
The most common magnetic material is ferrite, which is a ceramic
containing iron(III) oxide (Fe2 O3 ). Very strong magnetic effects are obtained
with materials containing rare earth elements. The rare earth elements,
the lanthanoid elements, can align the spins of many electrons of each
atom and can produce particularly strong magnetic domains and magnetic
moments. The most common rare earth elements used in magnetic materials
are samarium and neodymium used in samarium-cobalt and neodymium-
iron-boron magnets, respectively.
The domains can be aligned by a strong externally generated magnetic

Figure 5-15: Magnetic moments: (a) electron withe

magnetic moment m; (b), magnetic domains; and
(a) (b) (c) (c) progression with a magnetic field bias.
178 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 5-16: Gyromagnetic effect on the

propagation of EM waves in a magnetic
material (here a ferrite) with an externally
applied magnetic bias field.

field (e.g., using an electromagnet), and once the domains are aligned they
can stay that way for years even when the external field is removed. This
situation is shown in Figure 5-15(b), where the individual domains have
been almost completely aligned. Each crystal of the material will usually
have many magnetic domains and a domain can grow in size (while another
reduces or disappears), but generally a domain will not cross a crystal
boundary. The domains can also be partially rotated by a moderate applied
magnetic field and when the applied field is removed the domains return
to their original alignment. The result is that magnetic energy is stored in
much the same way as energy is stored in a spring. The amount of energy
that is stored depends on the orientation of the applied magnetic field to
the domains. The energy storage capability is described by the permeability
of the material. In the absence of a constant biasing magnetic field, the
permeability would have three values, one for each of the x, y, and z
directions.
A most interesting microwave property occurs when the magnetic
material is biased by a strong DC magnetic field. This situation is depicted
in Figure 5-15(c). When a time-varying magnetic field is alsoapplied (e.g.,
the magnetic component of an EM field), the magnetic moment vector will
tend to rotate around the DC magnetic field as shown. This is called the
gyromagnetic effect. When the frequency of the applied field corresponds
to the characteristic frequency of rotation of the magnetic moment then
the effect is called gyromagnetic resonance and there is sustained low loss
rotation.
Even without resonance (because the magnetic bias field is too small)
the gyromagnetic effect affects the way an RF field propagates and this is
described by a nine element permeability called a tensor (or a dyadic or a
dyadic tensor) that relates each of the three H field components (in the x-, y-,
and z-directions) to each of the three B field components. The permeability
of a magnetically biased magnetic material is:
   
µxx µxy µxz µ0 0 0
[µ] =  µyx µyy µyz  =  0 µ κ  . (5.7)
µzx µzy µzz 0 −κ µ

This tensor can take other forms depending on the orientation of the time-
varying magnetic field to the DC biasing magnetic field.
The effect on propagation of an EM field is shown in Figure 5-16. When the
EM field is in the magnetized magnetic material, the wave does not travel in
a straight line and instead curves, in this case, to the right. Thus forward-
and backward-traveling waves diverge from each other and propagation is
not reciprocal. This can be used to separate forward- and backward-traveling
waves. This is called field displacement.
RF AND MICROWAVE MODULES 179

(a) Circulator schematic (b) Microstrip circulator (c) Microstrip isolator (d) YIG-tuned BPF

Figure 5-17: Ferrite components [21].

5.5.2 Circulator
A circulator exploits the gyromagnetic effect through field displacement
[18–20]. The schematic of a circulator is shown in Figure 5-17(a), where the
arrows indicate that the signal that enters Port 1 of the circulator leaves the
circulator at Port 2 and not at Port 3. Similarly power that enters at Port 2 is
routed to Port 3, and power entering at Port 3 is routed to Port 1. In terms of
S parameters, an ideal circulator has the scattering matrix

   
0 0 S13 0 0 1
S =  S21 0 0 = 1 0 0 . (5.8)
0 S32 0 0 1 0

A microstrip circulator is shown in Figure 5-17(b), where a disc of

magnetized ferrite can be placed on top of a microstrip Y junction to realize
a preferential direction of propagation of the EM fields [22]. Design requires
choice of the size of the ferrite disc and design of the appropriate magnetic
biasing field. In the absence of the biasing magnetic field, the circulation
function does not occur.
In addition to the insertion and return losses, the performance of a
circulator is described by its isolation, which is its insertion loss in the
undesired direction.

5.5.3 Circulator Isolation

The isolation of a circulator is the insertion loss from what is the output port
to the input port, i.e. in the reverse direction. Referring to the circulator in
Figure 5-17(a), if port 1 is the input port there are two output ports and so
there are two isolations equal to the return loss from port 3 to port 2, and
from port 2 to port 1. The smaller of these is the isolation quoted if only one
value is given. If this was an ideal circulator and port 2 is perfectly matched,
then the isolation would be infinite. If port 2 is not perfectly matched then
there will be finite isolation from port 3 to port 1. However the most common
source of limited isolation is when the circulator is not perfectly matched.
Consider that the reflection coefficient looking into each of the circulator’s
ports is Γ and if the circulator is well designed Γ ≪ 1. Then the S parameters
180 STEER MICROWAVE AND RF DESIGN: MODULES

of the circulator are

 
Γ α T
S= T Γ α  (5.9)
α T Γ

where T is the transmission factor and is close to 1 for a good circulator and
the leakage, α, is small. If the circulator is lossless then the unitary conditions,
from Equations (2.148) and (2.149) of [23], are

|T |2 + |α| + |Γ|2 = 1 (5.10)

Γα∗ + T Γ∗ + αT ∗ = 0 −→ α∗ + T (Γ∗ /Γ) + T ∗ (α/Γ) = 0 (5.11)

Since α is small αΓ is negligibly small, so Equation (5.11) indicates that

|α| = |Γ|. Then from Equation (5.10), |T |2 = 1 − 2|Γ|2 , and the S parameters
of the lossless slightly-mismatched circulator can be written
�
1 − 2|Γ|2
 
� Γ Γ
S= 1 − 2|Γ|2 � Γ Γ . (5.12)
Γ 1 − 2|Γ| 2 Γ

Thus even with ideal external matching at port 2, the isolation is 1/Γ. The
quality of the match of a microwave component is typically specified by its
VSWR at the ports. Thus the VSWR of a circulator indicates the isolation
that can be expected. Actual circulators have a small amount of loss so the
isolation and VSWR quoted for a circulator will not conform exactly to the
lossless situation considered here.

5.5.4 Isolator
Isolators are devices that allow power flow in only one direction. There
are two types, one based on field displacement and the other based on
gyromagnetic resonance. Ferrite isolators exploiting field displacement are
based on a three-port circulator with one of the ports terminated in a
matched load. Figure 5-17(c) shows a microstrip isolator based on a three-
port circulator. The puck at the center is a magnetic material such as ferrite
that when magnetized by a permanent magnet or electromagnet (which is
not shown) preferentially supports a counter-clockwise rotating EM wave.
So power entering Port 1 as a traveling wave is transferred to the ferrite and
emerges at Port 2. Virtually none of the power emerges at Port 3. A traveling
wave signal applied at Port 2 appears at Port 3, where it is absorbed in a
termination created by resistive material placed on top of the microstrip. The
resistive material forms a lossy transmission line and, provided that the lossy
line section is long enough, no power is reflected. Thus power can travel from
Port 1 to Port 2, but not in the reverse direction. An isolator is commonly
used to protect the output of equipment from high reflected signals. A four-
port version can implement a duplexer in radar systems and to separate the
received and transmitted signals in a transceiver.
A high-frequency microstrip isolator derived from a circulator is shown
in Figure 5-18. The input and output lines are at the top and are redirected
under the puck. The puck is epoxied to the microstrip substrate and a large
biasing magnet is attached on top of the puck. The third port is at the bottom
of the figure.
RF AND MICROWAVE MODULES 181

(a) Layout with carrier (b) Image

Figure 5-18: A microstrip isolator operating from 29 to 31.5 GHz. Isolator in (b) has the
dimensions 5 mm × 6 mm and is 6 mm high. The isolator supports 2 W of forward and reverse
power with an isolation of 18 dB and insertion loss of 1 dB. Renaissance 2W9 series, copyright
Renaissance Electronics Corporation, used with permission.

Isolators can also exploit gyromagnetic resonance and are then called
resonance isolators. A suitable rotation of the RF magnetic field relative to
the DC magnetic bias results for one of the directions of propagation in a
rectangular waveguide. At the gyromagnetic resonance frequency, RF energy
is coupled into the lattice and RF power is absorbed for one direction of
propagation and in the other direction the RF signal is little affected.

5.5.5 YIG-Tuned Bandpass Filter

The gyromagnetic effect also enables very sharp variable bandpass filters.
A Yttrium-Iron-Garnet (YIG)-tuned filter is shown in Figure 5-17(d) and
the result is that Ports 1 and 2 are only coupled at a precise frequency
determined by the state of magnetization of the YIG sphere. With an
electromagnet providing a DC magnetic field, the magnetic field can be
varied and the resonance frequency, and hence the bandpass frequency,
electronically tuned. Microwave spectrum analyzers often use a YIG-tuned
filter at the input of the analyzer.

5.6 Summary
The use of modules has become increasingly important in microwave
engineering. A wide variety of passive modules were introduced but there
are many more. Some microwave modules use unusual effects such as the
response of magnetic materials to magnetic fields. Many others use the
characteristics of propagating EM fields. The next chapter completes the
discussion of modules by covering active modules. A successful microwave
designer must have a good knowledge of the types and limitations of
available modules.
182 STEER MICROWAVE AND RF DESIGN: MODULES

5.7 References
[1] http://www.synergymwave.com. [12] ——, “RF MEMS for reconfigurable circuits
[2] http://www.hittite.com. and antennas,” in Multifunctional Adaptive
[3] IEEE Standard 315-1975, Graphic Symbols Microwave Circuits and Systems, M. Steer and
for Electrical and Electronics Diagrams W. Palmer, Eds. SciTech Publishing, 2008.
(Including Reference Designation Letters), [13] G. Rebeiz, RF MEMS: Theory, Design, and
Adopted Sept. 1975, Reaffirmed Dec. 1993. Technology. Wiley, 2003.
Approved by American National Standards [14] C. Fay and R. Comstock, “Operation of the
Institute, Jan. 1989. Approved adopted for ferrite junction circulator,” IEEE Trans. on Mi-
mandatory use, Department of Defense, crowave Theory and Techniques, vol. 13, no. 1,
United States of America, Oct. 1975. Ap- pp. 15–27, Jan. 1965.
proved by Canadian Standards Institute, Oct. [15] C. K. Queck and L. Davis, “Microstrip
1975. and stripline ferrite-coupled-line (fcl) circu-
[4] B. B. Streetman and S. Banerjee, Solid State lator’s,” IEEE Trans. on Microwave Theory and
Electronic Devices, 6th ed. Prentice Hall, Techniques, vol. 50, no. 12, pp. 2910–2917, Dec.
2006. 2002.
[5] S. Sze and K. Ng, Physics of Semiconductor De- [16] A. Borjak and L. Davis, “More compact fer-
vices, 3rd ed. John Wiley & Sons, 2007. rite circulator junctions with predicted per-
[6] D. Schroder, Semiconductor Material and De- formance,” IEEE Trans. on Microwave Theory
vice Characterization. IEEE Press and Wiley, and Techniques, vol. 40, no. 12, pp. 2352–2358,
2006. Dec. 1992.
[7] K. Ng, Complete Guide to Semiconductor De- [17] B. Lax and K. J. Button, Microwave ferrites and
vices. IEEE Press and Wiley Interscience, ferrimagnetics. McGraw-Hill, 1962.
2002. [18] D. K. Linkhart, Microwave circulator design.
[8] A. Bergh and P. Dean, “Light-emitting Artech House, 2014.
diodes,” Proc. of the IEEE, vol. 60, no. 2, pp. [19] V. G. Harris, “Modern microwave ferrites,”
156–223, Feb. 1972. IEEE Trans. on Magnetics, vol. 48, no. 3, pp.
[9] A. Victor and M. Steer, “Reflection coefficient 1075–1104, Mar. 2012.
shaping of a 5-GHz voltage-tuned oscilla- [20] A. Goldman, Modern ferrite technology.
tor for improved tuning,” IEEE Trans. on Mi- Springer Science & Business Media, 2006.
crowave Theory and Techniques, vol. 55, no. 12, [21] J. Helszajn, YIG Resonators and Filters. Wiley,
pp. 2488–2494, Dec. 2007. 1985.
[10] M. Steer, Microwave and RF Design, Amplifiers [22] U. Milano, J. Saunders, and L. Davis, “A y-
and Oscillators, 3rd ed. North Carolina State junction strip-line circulator,” Microwave The-
University, 2019. ory and Techniques, IRE Trans. on, vol. 8, no. 3,
[11] L. Katehi and D. Peroulis, “Rf mems compo- pp. 346–351, may 1960.
nents: switches and varactors,” in Multifunc- [23] M. Steer, Microwave and RF Design, Networks,
tional Adaptive Microwave Circuits and Sys- 3rd ed. North Carolina State University,
tems, M. Steer and W. Palmer, Eds. SciTech 2019.
Publishing, 2008.

5.8 Exercises
1. A connector used in a 50 Ω system introduces a and has a 5 Ω on resistance. The reactive para-
series resistance of 0.5 Ω. What is the insertion sitics of the switch are negligible.
loss of the connector? (a) What is the insertion loss of the switch in
2. A microwave switch is used in a 75 Ω system the on state?
and has a 5 Ω on resistance. The reactive para- (b) If the available power of the source is
sitics of the switch are negligible. 50 W, what is the power dissipated by the
switch?
(a) What is the insertion loss of the switch in
4. A microwave switch is used at 1 GHz in a 50 Ω
the on state?
system and it has a 2 Ω on resistance and a
(b) What is the return loss of the switch in the
2 kΩ off resistance. The reactive parasitics of the
on state?
switch are negligible.
3. A microwave switch is used in a 50 Ω system (a) What is the insertion loss of the switch?
RF AND MICROWAVE MODULES 183

(b) What is the isolation of the switch (i.e., 10. A three-port circulator has the S parameters
what is the insertion loss of the switch  
when it is in the off state)? 0 0 1
 1 0 0 .
5. A microwave switch is used at 1 GHz in a 50 Ω 0 1 0
system. The switch has a 0.5 Ω on resistance, a
If port 3 is terminated in a matched load to cre-
2 kΩ off resistance, and a parasitic capacitance
ate a two-port network
between the input and output of the switch of
1 pF when the switch is in the off state. Other (a) What is the name given to this network?
reactive parasitics of the switch are negligible. (b) Write down the S parameters of the two-
port?
(a) What is the insertion loss of the switch? 11. A three-port circulator has the S parameters
(b) What is the isolation of the switch ignor-  
ing the capacitance of the switch in the off 0 0 0.5
state (i.e., what is the insertion loss of the  20.5 0 0 .
switch when it is in the off state)? 0 0.5 0
(c) What is the isolation of the switch with the Port 3 is terminated in a matched load creating a
off capacitance included? two-port network.
(a) Find the S parameters of the two-port.
6. Consider a microwave switch used in a 50 Ω sys-
(b) What is the return loss in dB at Port 1 if Port
tem. The switch has a 0.5 Ω on resistance, a 1 kΩ
2 is terminated in a matched load?
off resistance, and a 0.1 pF off capacitance. Other
(c) What is the insertion loss in dB for a signal
reactive parasitics of the switch are negligible.
applied at Port 1 and leaving at Port 2 with
(a) What is the insertion loss of the switch at matched source and load impedances?
1 GHz? (d) What is the insertion loss in dB for a signal
(b) What is the isolation of the switch at 1 GHz applied at Port 2 and leaving at Port 1 with
(i.e., what is the insertion loss of the switch matched source and load impedances?
when it is in the off state)? 12. A three-port circulator in a 75-Ω system has the
(c) What is the insertion loss at 10 GHz? S parameters  
0.1 0 0.707
(d) What is the isolation at 10 GHz?  0.707 0.1 0 .
(e) What is the return loss of the switch at
0 0.707 0.1
1 GHz when the switch is in the on state?
Port 3 is terminated in 75-Ω creating a two-port
7. A microwave switch has a 5 Ω on resistance, and network.
an off resistance of 500 Ω. The reactive parasitics (a) Find the S parameters of the two-port.
of the switch are negligible. (b) What is the return loss in dB at Port 1 if Port
2 is terminated in 75-Ω?
(a) What is the insertion loss of the switch
(c) What is the insertion loss in dB for a signal
when it is used in a 50 Ω system?
applied at Port 1 and leaving at Port 2 with
(b) What is the insertion loss of the switch
75-Ω source and load impedances?
when it is used in a 100 Ω system?
(d) What is the insertion loss in dB for a signal
(c) What is the isolation of the switch when it
applied at Port 2 and leaving at Port 1 with
is used in a 50 Ω system? (That is, what is
75-Ω source and load impedances?
the insertion loss of the switch when it is in
the off state?) 13. A three-port circulator has the S parameters
(d) What is the isolation of the switch when it  
0 1 0
is used in a 100 Ω system?  0 0 1 .
8. The RF front end of a communications unit con- 1 0 0
sists of a switch, then an amplifier, and then a (a) In no more than 50 words describe the func-
mixer. The switch has a loss of 0.5 dB, the am- tion of a circulator?
plifier has a gain of 20 dB, and the mixer has a (b) Describe power flow for the circulator with
conversion gain of 3 dB. What is the overall gain the given S parameters. When power en-
of the cascade? ters one of ports, at which port does power
9. Using a schematic, show how an isolator can be leave? Consider power entering Port 1, then
obtained from a circulator. Port 2, then Port 3.
184 STEER MICROWAVE AND RF DESIGN: MODULES

14. An ideal three-port circulator in a 50 Ω system (a) Find the S parameters of the two-port.
is configured as an isolator from Port 1 to Port (b) What is the return loss in dB at Port 1 if Port
2. The termination at the third port of the circu- 2 is terminated in 50-Ω?
lator has a value of 45 Ω. What is the isolation (c) What is the insertion loss in dB for a signal
of the isolator? (This is the insertion loss in the applied at Port 2 and leaving at Port 1 with
reverse direction.) [Hint: Use signal flow graph 50-Ω source and load impedances?
analysis.] (d) What is the insertion loss in dB for a signal
applied at Port 1 and leaving at Port 2 with
15. Two isolators are used in cascade. Each isola-
50-Ω source and load impedances?
tor has an isolation of 20 dB. The isolators are
(e) What is is the name of this network?
matched so that their input and output reflection
coefficients are zero. Determine the isolation of 17. Write down the two-port S parameters of the
the cascaded isolator system? microstrip isolator shown below.
16. A three-port circulator in a 50-Ω system has the
S parameters  
0.1 0.01 0.5
 0.5 0.1 0.01  .
0.01 0.5 0.1
If port 3 is terminated in a matched load to cre-
ate a two-port network

5.8.1 Exercises By Section

†
challenging, ‡ very challenging
§5.4 1, 2† , 3† , 4† , 5† , 6† , 7† §5.5 10, 11, 12, 13, 14† , 15, 16, 17

5.8.2 Answers to Selected Exercises

3(a) 0.424 dB 5(c) 5.76 dB 8 22.5 dB
2(b) 29.8 dB 7 0.424 dB 15 40 dB
 CHAPTER 6

Mixer and Source Modules

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185

6.2 Mixer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
6.3 Single-Ended, Balanced, and Double Balanced Reject Mixers . 197
6.4 Local Oscillator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
6.5 Voltage-Controlled Oscillator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
6.6 Phase Detector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
6.7 Frequency Multiplier . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
6.8 Frequency Divider . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
6.9 Phase Locked Loop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
6.10 Direct Digital Synthesizer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
6.11 Diode and Vacuum Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
6.12 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
6.13 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
6.14 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220

6.1 Introduction
This chapter considers frequency generation and frequency translation
modules. In a receiver the basic frequency translation module is a mixer that
is used in a receiver to take a signal at a high frequency to a low frequency,
the baseband signal, where it can be more easily sampled by an analog-to-
digital converter (ADC) and then digitally processed. In a transmitter the
signal flow is reversed and a low frequency baseband signal is translated
to a high-frequency signal where it can be more easily transmitted. A local
oscillator (LO) is input to the mixer to provide the translation mechanism
by varying the impedance of an element at the LO frequency. A voltage-
controlled oscillator (VCO) allows for agility of the oscillator. A VCO with
a phase detector embedded in a phase-locked loop (PLL) provides a well-
defined variable frequency oscillation signal. Special RF signal sources are
based either on special characteristics of certain semiconductor devices or on
vacuum devices in which a beam of electrons interacts with an RF field in a
controlled way. These sources produce very high RF powers up to several
megawatts or substantial power at very high frequencies, up to several
terahertz.
186 STEER MICROWAVE AND RF DESIGN: MODULES

6.2 Mixer
Frequency conversion, mixing or heterodyning, is the process of converting
information at one frequency (present in the form of a modulated carrier)
to another frequency. The second frequency is either higher, in the case
of frequency up-conversion, where it is more easily transmitted, or lower,
when mixing is called frequency down-conversion, where it is more easily
captured. The mixer types are shown in Figure 6-1. Capture of the down-
converted signal is nearly always by an ADC. Frequency conversion can
occur with any nonlinear element.
Consider the information signal flow in the down-converter in Figure 6-
1(a). (A similar discussion applies to the up-converter in Figure 6-1(b).) From
the left, a modulated RF signal centered at fRF is presented to a mixer that
is pumped by a large LO signal at fLO . The intended function of the mixer
is to convert the information on the modulated RF to a lower intermediate
frequency (IF) centered at fIF = |fRF − fLO |. The spectrum of the mixer,
shown on the right on Figure 6-1(a), has another tone, fIM called the image
tone. The image at fIM is an interferer, as it is also down-converted to the IF
since fIF = |fIM − fLO |. Noise at the image is also down-converted to the IF.
For the up-converter the high-power noise coming from the power amplifier
at the image would be transmitted. Mixer design therefore must consider
how image and noise are handled, as well as the efficiency of the conversion
process.
In Figure 6-2(a) a nonlinear device is driven by two signals at ωM and
ωC . The larger signal, the LO, is also called the pump and the other signal
is called the RF. The spectrum of the signals present in the circuit is
shown in Figure 6-2(c). In this mixer the aim is to produce a signal at the
difference frequency (or IF) with the same modulation, and hence the same
information, as the original RF signal. The transistor mixer shown in Figure
6-3 uses filtering to separate the RF, LO, and IF components.

(a) Down-conversion

Figure 6-1: Frequency

conversion using a mixer. (b) Up-conversion
MIXER AND SOURCE MODULES 187

Figure 6-2: Diode

(a) (b) mixer: (a) cir-
cuit; (b) diode
current-voltage
characteristic; (c)
spectrum across
the nonlinear
device; and (d)
schematic symbol
(c) (d) for a mixer.

Figure 6-3: Single-ended FET mixer with LO,

RF, and IF bandpass filters.

6.2.1 Mixer Analysis

A mixer can be designed around any nonlinear device [1]. Using an
operational amplifier, an ideal multiplier can be designed that will multiply
two signals, each described as cosinusoids. If the LO cosinusoid is cos(ω1 t)
and the input RF cosinusoid is cos(ω2 t), the multiplication of the two signals
will produce an output (using the trigonometric identity cos(A) cos(B) =
1
2 [cos(A − B) + cos(A + B)]):

y(t) = [cos(ω1 t)] · [cos(ω2 t)]

1
= 2 {cos[(ω1 − ω2 )t] + cos[(ω1 + ω2 )t]} . (6.1)
This has two components: one at the radian frequency (ω1 − ω2 ) and the
other at (ω1 + ω2 ). If the LO and RF signal are close, then the component
at (ω1 − ω2 ) will be at a frequency much lower than either the LO or
the RF, and the component at (ω1 + ω2 ) will be at almost twice the input
frequencies. Appropriate filtering will choose one of these components
depending on whether the application is up-conversion or down-conversion.
At microwave frequencies, circuits that do not realize ideal multiplication
must be used. This section considers what happens when two components
are applied to an arbitrary nonlinearity described by a low-order polynomial.
The result is that a large number of tones will be generated in the mixing
process. Balanced circuit designs can significantly reduce many of these
tones, greatly reducing the filtering required to select a particular output.
A two-tone input
x(t) = |X1 | cos(ω1 t + φ1 ) + |X2 | cos(ω2 t + φ2 )
188 STEER MICROWAVE AND RF DESIGN: MODULES

can be written using complex notation as

x(t) = 21 X1 eω1 t + X1∗ e−ω1 t + X2 eω2 t + X2∗ e−ω2 t .

 

Note that the coefficient of the positive exponential frequency component is

one-half that of the phasor. Thus the phasor of the ω1 component is X1 =
|X1 |e(φ1 ) and the phasor of the ω2 component is X2 = |X2 |e(φ2 ) . The first
three powers of x can be easily expanded manually; for example, expanding
x2 gives
 2  2 2ω1 t
x2 (t) = 12 X1 e + 2X1 X1∗ + 2X1 X2 e(ω1 +ω2 )t + 2X1 X2∗ e(ω1 −ω2 )t
+ (X1∗ )2 e−2ω1 t + 2X1∗ X2 e(ω2 −ω1 )t + 2X1∗ X2∗ e−(ω1 +ω2 )t + X22 e2ω2 t

+ 2X2 X2∗ + (X2∗ )2 e−2ω2 t , (6.2)

and similarly, expanding x3 yields

 3  3 3ω1 t
x3 (t) = 21 X1 e + 3X12 X1∗ eω1 t + 3X12 X2 e(2ω1 +ω2 )t
+ 3X12 X2∗ e(2ω1 −ω2 )t + 3X1 (X1∗ )2 e−ω1 t + 6X1 X1∗ X2 eω2 t
+ 6X1 X1∗ X2∗ e−ω2 t + 3X1 X22 e(ω1 +2ω2 )t
+ 6X1 X2 X2∗ eω1 t + 3X1 (X2∗ )2 e(ω1 −2ω2 )t + (X1∗ )3 e−3ω1 t
+ 3(X1∗ )2 X2 e(ω2 −2ω1 )t + 3(X1∗ )2 X2∗ e−(2ω1 +ω2 )t + 3X1∗ X22 e(2ω2 −ω1 )t
+ 3X1∗ (X2∗ )2 e−(ω1 +2ω2 )t + X23 e3ω2 t + 6X1∗ X2∗ X2 e−ω1 t

+3X22 X2∗ eω2 t + 3X2 (X2∗ )2 e−ω2 t + (X2∗ )3 e−3ω2 t , (6.3)

so that the output of the cubic equation,

y(t) = a0 + a1 x(t) + a2 x2 (t) + a3 x3 (t),

can be calculated for a two-tone input. Table 6-1 lists these phasors and
groups them by frequency. A more general approach is described in [2].
The phasors of the various intermodulation products resulting from x, x2 ,
and x3 can be taken as the coefficients of the positive exponential frequency
components after the factor of two correction required for terms other than
DC [2]. Terms of the same frequency are summed to obtain the output at a
particular frequency. For example, the phasor output at (2ω1 − ω2 ) is given
by the sum of three intermodulation products:

Y(2ω1 −ω2 ) = a3 34 X12 X2∗ .

 
(6.4)

So the level of the output of a mixer, here Y(2ω1 −ω2 ) at radian frequency
(2ω1 − ω2 ), is related directly to the strength of the LO signal (with amplitude
|X1 |), the strength of the nonlinearity (captured by the an coefficients),
and the level of the input signal |X2 |. Unfortunately, low-order power
series analysis as used here is not sufficient to model practical mixers, and
computer-aided modeling tools are necessary. However, the manual analysis
enables operation to be understood and architectures to be developed that
intrinsically have the desired characteristics.
MIXER AND SOURCE MODULES 189

Intermodulation product Frequency Order Table 6-1: The intermodulation prod-

1 ∗ ucts resulting from x, x2 , and x3 ,
2 X1 X1 0 2
where x is a two-tone signal, show-
1 ∗
2 X2 X2 0 2 ing only the positive frequencies. The
2 12 X1 = X1 ω1 1 first column gives the complex ampli-
tudes (phasors) of the frequency com-
2( 12 )3 3X12 X1∗ = 34 X12 X1∗ ω1 3 ponents. (The order is the power of x.)
2( 21 )3 6X1 X2 X2∗ = 32 X1 X2 X2∗ ω1 3
2 12 X2 = X2 ω2 1
2( 12 )3 3X22 X2∗ = 34 X22 X2∗ ω2 3
2( 21 )3 6X1 X1∗ X2 = 32 X1 X1∗ X2 ω2 3
2( 21 )2 X12 = 12 2X12 2ω1 2
2( 12 )2 X22 = 1 2
2 X2 2ω2 2
2( 12 )3 X13 = 1 3
4 X1 3ω1 3
2( 12 )3 X23 = 1 3
4 X2 3ω2 3
2 12 X1 X2 = X1 X2 ω1 + ω2 2
2 21 X1 X2∗ = X1 X2∗ ω1 − ω2 2
2( 21 )3 3X12 X2 = 34 X12 X2 2ω1 + ω2 3
2( 12 )3 3X12 X2∗ = 34 X12 X2∗ 2ω1 − ω2 3
2( 21 )3 3X1 X22 = 3 2
4 X1 X2 ω1 + 2ω2 3
2( 12 )3 3X1∗ X22 = 3 ∗ 2
4 X1 X2 2ω2 − ω1 3

6.2.2 Mixer Performance Parameters

The main characteristics that define the performance of a mixer are the
conversion gain or loss, and the noise figure [3, 4]. Mixing results from a
nonlinear process that generates many tones and not just the ones of interest.
Consequently, additional parameters are used to describe the performance
of a mixer, and these derive from the generation of the additional tones. The
mixer performance parameters are as follows:
Conversion loss: This is the ratio of the available power of the input signal
to that of the output signal after mixing. It is usually expressed in decibels.
In the diode mixer shown in Figure 6-5, the conversion loss is

Pin (RF)
LC = . (6.5)
Pout (IF)

In decibels, the conversion loss is

 
Pin (RF)
LC |dB = 10 log10 . (6.6)
Pout (IF)

Noise figure (NF): The NF is 10 times the log of the noise factor F . The noise
factor is the ratio of the SNR at the RF input to the SNR at the IF output (using
the input noise generated by a resistor at standard temperature, 290 K).
190 STEER MICROWAVE AND RF DESIGN: MODULES

However, there are qualifications for mixers. The first is that two inputs (at
the RF and image frequencies) can produce noise and signal power at the IF.
The double-sideband (DSB) NF includes signal and noise contributions from
both the RF and the image frequencies. This is the situation with radiometry
and astronomy, where the signal is at both sidebands. Single-sideband (SSB)
NF includes the input signal at the RF only, but includes noise originating at
both the RF and at the image frequencies. This is the situation for mixers in
communications where the signal is only at one sideband.
Mixers can have substantial excess noise, as noise that is offset in frequency
from the LO and its harmonics by the magnitude of the IF frequency will be
down-converted to the IF for a down-converter, or converted to the RF in an
up-converter. This process is sometimes called noise folding.
In summary, two noise figures are used with mixers, SSB NF and DSB NF.
Which to use depends on the system in which the mixer is embedded.
Image rejection: If a mixer has an LO of 1.1 GHz and an RF of 1.5 GHz, then
the IF will be at 400 MHz. This IF can also be generated by the image signal at
700 MHz. Numerically the image is the reflection of the RF in the LO. Image
rejection refers to the ability of a mixer to reject the image signal. This can be
achieved, for example, by using an input RF bandpass filter.
If the applied image and intended signal powers are the same and the level
of the output signal (at the IF) produced by the intended RF signal is Pout ,
and that produced by the image signal is Pout,image , then the image rejection
ratio (IRR) is
Pout
IRR = . (6.7)
Pout,image

This is typically expressed in decibels and

 
Pout
IRR|dB = 10 log10 . (6.8)
Pout,image

EXAMPLE 6.1 Mixer Calculations

A mixer has an LO of 10 GHz. The mixer is used to convert a signal at 10.1 GHz to an IF at
100 MHz, and has a conversion loss, Lc of 3 dB and an image rejection of 20 dB. Two signals
are presented to the mixer, one at 10.1 GHz with a power of 100 nW and the other at 9.9 GHz
with a power of 1 µW.
(a) What is the output power of the (intended) signal at the IF?
(b) What is the signal-to-interference ratio at the IF (ignoring noise)?
Solution:
(a) Lc = 3 dB = 2 and from Equation (6.5) the output power at IF of the intended signal is

Pout = Pin (RF)/Lc = 100 nW/2 = 50 nW = −43 dBm. (6.9)

(b) Interference at the IF comes from the down-converted image signal. IRR = 20 dB = 100.
If the applied powers of the intended signal and the image signal are the same, from
Equation (6.7),
Pout Pout
IRR = i.e. Pout,image = (6.10)
Pout,image IRR
MIXER AND SOURCE MODULES 191

This must be modified to account for the difference in the applied power levels and

Pout Pin (RF, image) (50 nW) · (1 µW) (50 nW) · (1000 nW)
Pout,image = = = = 5 nW.
IRR Pin (RF) 100 · (100 nW) 100 · (100 nW)

The signal-to-interference ratio is

Pout 50 nw
SIR = = = 10 = 10 dB. (6.11)
Pout,image 5 nw

6.2.3 Mixer Waveforms

This section presents the waveforms of diode mixers. Diode mixers, such
as the balanced mixer in Figure 6-4, are bilateral. In a communication
transceiver such a mixer can be used for both the receive and transmit
functions, greatly simplifying filtering requirements. Most mixers in cell
phones, however, are based on transistors, but the underlying architecture
corresponds to a single-diode mixer or, more commonly, the diode ring
double-balanced mixer.
The first mixer to be considered is the single-ended mixer shown in Figure
6-5(a). Filters are not included thus revealing the full complexity of the
spectra. The transient waveform across the diode (measured at the test point)
is shown in Figure 6-5(b). The turn-on transient is due to both the turn-on
ramp of the LO and RF sources, and to the internal capacitance of the diode.
The spectra of the signal at the test point is shown in Figure 6-5(c and d).
The desired signal here is the IF, which is at the difference frequency of the
LO and RF (i.e., at 400 MHz). Extracting just the IF in the response shown in
Figure 6-5(d) would require significant filtering. The level of the IF tone here
is 11.5 mVpeak . If a lossless 400 MHz filter is attached to the test point and
the IF load (on the other side of the filter) is 50 Ω, the IF power delivered to
the load is Pout = 12 (11.5 mV)2 /(50 Ω) = 1.322 µW. The available RF power
is Pin = 12 (100 mV)2 /(50 Ω) = 100 µW. So the conversion loss is
100 µW
LC = = 75.08 = 18.8 dB. (6.12)
1.332 µW
Now consider the diode ring mixer shown in Figure 6-6. Waveforms and
spectra at the LO, IF, and RF test points are shown in Figure 6-7. Note how
relatively clean the spectra are at each of the test points. The reduction in
clutter is the result of the balanced circuit. The circuit in Figure 6-6 is known
as a double-balanced mixer, and the same concept can be used with transistor
mixers.
Focusing on just the spectra at the IF test point (Figure 6-7(e)), it can be
seen that the LO signal and its harmonics are nonexistent. The components
other than the 400 MHz IF are other sum and difference products of the RF
and LO. However, compared to the spectra of the single-ended diode mixer

Figure 6-4: Diode ring double-

balanced mixer.
192 STEER MICROWAVE AND RF DESIGN: MODULES

(a) Schematic (b) Waveform at the test point

Figure 6-5:
Single-
ended
diode
mixer. (c) Spectrum at the test point (d) Expanded spectrum at the test point

Figure 6-6: Diode

double-balanced
mixer schematic.

in Figure 6-5(c and d), many of these are eliminated as well. This is a very
attractive circuit, as it significantly reduces the specifications required for an
output filter. This is one of the special characteristics of RF and microwave
design. Much can be gained by being creative—a designer gets better with
experience. The level of the IF signal at the RF test point is 26.7 mVpeak . The IF
power delivered to the 50 Ω load is Pout = 12 (26.7 mV)2 /(50 Ω) = 7.126 µW.
The available RF power is Pin = 21 (100 mV)2 /(50 Ω) = 100 µW. So the
conversion loss is
100 µW
LC = = 14.03 = 11.47 dB. (6.13)
7.126 µW

The conversion loss is much lower than was obtained with the single-ended
mixer. In part this is because power was not dissipated in a large number of
spurious tones.
MIXER AND SOURCE MODULES 193

LO

(a) Spectrum at the LO test point (b) Waveform at the LO test point

RF

(c) Spectrum at the RF test point (d) Waveform at the RF test point

IF

(e) Spectrum at the IF test point (f) Waveform at the IF test point

Figure 6-7: Waveforms and spectra of the double-balanced diode ring mixer of Figure 6-6.
194 STEER MICROWAVE AND RF DESIGN: MODULES

6.2.4 Switching Mixer

The analysis of a mixer in Section 6.2.1 modeled the mixing element using
a polynomial. With a large LO, an alternative view of a mixer is to consider
it as a switching device in which the conductance of the mixing element is
periodically switched by the LO.
A simple switching diode mixer is shown in Figure 6-8(a). If the LO level
is large, then the mixer can be modeled by the equivalent circuit shown
in Figure 6-8(b) with a periodically varying conductance, gLO (t). This is
a square wave with the same frequency and period as the applied LO
sinewave. The diode effectively looks like a switch with the conductance
alternating between a maximum value gm and zero (see Figure 6-8(c)). The
first few terms of the Fourier series expansion of gLO (t) are
 
1 1 2
gLO (t) = gm + cos(ωLO t) + cos(3ωLO t) + . . . . (6.14)
2 π 3π

The small RF signal interacts with the switching conductance so that the
output current is

i(t) = gLO (t) [vRF − ix (t)R] . (6.15)

Ignoring R, then the IF current, where ωIF = ωRF − ωLO , at the point x is
gm
iIF (t) = gLO (t)vRF (t) = cos(ωLO t)vRF cos(ωRF t)
π
gm vRF
= {cos[(ωLO − ωRF )t] + cos[(ωLO + ωRF )t]} . (6.16)
2π
Typically the IF signal at (ωLO −ωRF ) would be extracted through a bandpass
or lowpass filter that allows only the IF component to pass and an IF voltage
is realized as the current passes through a load resistor.
One of the advantages of the switching mixer is that performance is
relatively insensitive to the level of the LO. The LO could be a sinewave and
still the variation of the conductance would be close to being a square wave.
So then the design of the mixer specifically is to develop a square variation
of the conductance.
Switching mixers can be realized using other circuits. One of these is the
diode ring mixer shown in Figure 6-9. Here the center-tapped transformers
produce differential LO and RF signals. The large LO is transformed by the
tapped transformer to produce a large differential signal that turns pairs
of diodes on in sequence. During the positive half of the LO cycle, the

(a) Diode mixer (b) Equivalent model (c) Conductance waveform

Figure 6-8: Switching diode mixer; gLO is a conductance.

MIXER AND SOURCE MODULES 195

(a) + LO current path (b) + LO equivalent model

(c) − LO current path (d) − LO equivalent model (e) Conductance waveform

Figure 6-9: Diode ring mixer as a switching mixer. gm is a conductance.

Figure 6-10: Diode ring mixer implemented using a

hybrid to produce a pseudo-balanced signal.

two right-hand diodes in Figure 6-9(a) pass current limited by the source
impedance of the LO source and each diode has a conductance 2gm , as
seen in Figure 6-9(b). During the negative half-cycle of the LO, the left-
hand diodes pass current and each has conductance 2gm , see Figure 6-9(c
and d). When not forward biased, the diodes appear as open circuits. This
process results in the conductance waveform shown in Figure 6-9(e). An
important characteristic of this waveform is that it is symmetrical even when
nonidealities are considered.
An RF implementation of the diode mixer is shown in Figure 6-10. The
180◦ hybrid replaces the transformer to distribute oppositely phased RF
signals to the diode ring. The IF is now taken from the center tap of the LO
transformer and is passed through a lowpass filter to remove all signals other
than the IF. This is a down-converting mixer. If the mixer is an up-converting
mixer, then the lowpass filter would be replaced by a bandpass filter.
Transistor-based mixer implementations using the ring mixer concept are
shown in Figure 6-11. These circuits are used in monolithic ICs with the
differential signals available from preceding stages and the IF output is also
differential. The transistors operate as switches that are controlled by the LO
signal and the conductance waveform is as for the diode mixer (i.e., as in
Figure 6-9(e)). Another switching mixer is the transistor commutating mixer
shown in Figure 6-12.
Traditionally the the biggest issue with switching mixers was is the
substantial LO power required and the limited dynamic range of the mixer in
196 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 6-11: Transistor ring

mixers. (a) MOS ring mixer (b) CMOS H-bridge ring mixer

(a) Commutating mixer (b) + LO model (c) − LO model (d) Conductance

Figure 6-12: Commutating transistor mixer.

converting RF to IF (for a down-converter) or IF to RF (for an up-converter).

This situation changes with the 32 nm and lower CMOS processing nodes.
This has resulted in CMOS switching mixers being important elements of
RFICs and these have superior dynamic range.

6.2.5 Subharmonic Mixer

If the nonlinear element of a mixer is particularly strong, or perhaps the drive
level is very high, mixing will occur when the frequency of the LO drive of
the mixer is a subharmonic of the the effective LO frequency. Consider the
switching mixers discussed in Section 6.2.4 and the switching conductance,
Figure 6-12(c), of the commutating mixer in Figure 6-12(a). With this mixer
the RF signal effectively mixes with an harmonic of the applied LO. Here
these harmonics are odd harmonics and so a commutating mixer driven by
an LO signal at 5 GHz will down-convert and RF signal at 15.1 MHz to an
IF frequency of 100 MHz. One of the advantages of a subharmoinic mixer
is that there is not a large 15 GHz signal which could be a problem as it
could leak into the front of a receiver and possibly radiate unintentionally.
Another advantage is that with a conventional down-conversion mixing
approach where the LO and RF are close in frequency, an LO at 15 GHz, in
previous example, would need to be generated from the 5 GHz signal. This
would consume considerable power. So it could be more power efficient to
use a subharmonic mixer rather than a frequency tripler and a conventional
mixer. A subharmonic mixer could be the preferred design option especially
for down-converting millimeter-wave signals. At high millimeter-wave
frequencies (e.g. > 200 GHz) the mixer is often a passive nonlinear element.
MIXER AND SOURCE MODULES 197

6.3 Single-Ended, Balanced, and Double Balanced Mixers

Single-ended, balanced, and double-balanced diode mixers are shown in
Figure 6-13. The simplest type of mixer that has been considered is the
single-ended or unbalanced diode mixer of Figure 6-2(a). This is shown as
a down-converting mixer where the RF signal at fRF mixes with the LO at
fLO to produce a down-converted IF signal at fLO − fRF . The image signal
at fLO + (fLO − fRF ) = 2fLO − fRF will be down-converted to the same
IF signal. It is of course not desirable to down-convert the image signal
and it could be prevented by using filtering as shown, but the use of filters
makes the filter quite large. There are many spurious frequencies with the
single-ended mixer as seen in Figure 6-5. Most of these can be removed with
filtering but using filters is undesirable and certainly not compatible with on-
chip implementations. The solution is to use symmetry to cancel out many
of the spurious tones. Figure 6-13(b and c) shows two types of symmetry,
balanced and double balanced mixers. Another level of symmetry could be
used (applying it to the IF) in the double balanced mixer and then it could be
called a triple balanced mixer. The triple balanced mixers is nearly always
referred to as a double-balanced mixer. One cannot say whether a mixer is
balanced, or even double balanced without tracing through the cancellation
process. With each increasing degree of symmetry greater linearity of the
conversion process is obtained. The double balanced mixer in FET form
is particularly attractive as the FETs do not need to be biased. Except at
very high millimeter-wave frequencies and terahertz frequencies, all mixers
today are of the double or triple balanced form. However there can be
unintentional mixing so the unbalanced mixer should be studied so that
unwanted mixing can be identified.

6.3.1 RFIC Mixers

RFIC mixers below millimeter-wave frequencies utilize the transistor as a
switch either with finite conductance when the switch closed and virtually
zero conductance when the switch is open. The first mixer to be considered
is based on the commutating mixer shown in Figure 6-12.

(a) Unbalanced (b) Balanced (c) Double balanced

Figure 6-13: Unbalanced (also known as single-ended), balanced, and double-balanced down-
conversion diode mixersthat mix an LO with an RF to produce a lower frequency IF. bandpass
filters.
198 STEER MICROWAVE AND RF DESIGN: MODULES

(a) Gilbert cell (b) + LO model (c) − LO model

Figure 6-14: Model of the Gilbert cell and equivalent circuit models. The circuit is used as a mixer
where vy is the LO and vx is the RF. (This is referred to as a double balanced mixer but strictly
it is a triple balanced mixer as there are RF, IF, and LO symmetries resulting in a high level of
cancellation of spurious tones.)

6.3.2 Gilbert Cell

The Gilbert cell builds on the commutating mixer circuit of Figure 6-12. The
commutating mixer conducts current in the IF leg (with RIF during only half
of the LO waveform. An immediate improvement is obtained by modifying
the circuit to provide switched conductance during both halves of the LO
waveform. Also, the RF and IF in the commutating mixer are single-ended
quantities whereas with a silicon IC differential signals are preferred. The
simplest Gilbert cell is shown in Figure 6-14(a), which has all of the desired
properties. M2 and M3 are designed as a differential amplifier pair with M4
and M5 functioning as time-varying loads for M2 , and M6 and M7 function as
time-varying loads for M3 . This leads to the equivalent model of the Gilbert
cell shown in Figures 6-14(b and c).
The Gilbert cell is not always used as a mixer. If both vy and vx are small
signals, then the Gilbert cell becomes a good analog multiplier with the
output vo = vx × vy . When used as a mixer, vy is a large LO and the Gilbert
cell is referred to as a Gilbert mixer. The Gilbert mixer is the functional core
of virtually every integrated mixer, especially in CMOS technology where
differential operation is preferred.

6.3.3 Integrated Gilbert Cell-Based Mixer

The performance of a Gilbert mixer is improved by replacing the source
resistor, RS , of M2 and M3 in the basic Gilbert cell of Figure 6-12 with
a current source, as shown in Figure 6-15(a). This is the most common
variation of the Gilbert mixer used in RFICs. A double-balanced Gilbert cell
mixer is shown in Figure 6-15(a). When the LO is small, the Gilbert mixer
has an almost ideal multiplier response with performance limited by how
closely the transistors can be matched [5]. In RFICs the transistors can be
MIXER AND SOURCE MODULES 199

(a) Gilbert mixer (b) Folded cascode implementation

Figure 6-15: Gilbert cell double-balanced mixer.

(a) Class A amplifier (b) Loadline (c) Tanh-like

characteristic

Figure 6-16: Simple transistor circuit (biasing not shown).

matched very well. When the LO is sufficiently large to produce a switching

conductance, there is much greater tolerance to imperfect matching. This
circuit has good performance and rejects RF and LO feed-through to the
output, and the degree of rejection is dependent on the quality of the
transistor match.
Before continuing the discussion the gross characteristics of a transistor
circuit will be introduced. The simplest transistor circuit is the amplifier
shown in Figure 6-16(a). With an RF input signal, the large capacitor at
200 STEER MICROWAVE AND RF DESIGN: MODULES

the gate of the transistor becomes a short circuit and the trajectory of the
transistor’s voltages and currents follows a load line as shown in Figure 6-
16(b). In the form of a transfer function, the characteristic of the amplifier is
as shown in Figure 6-16(c). This type of characteristic is typical of transistor
circuits with the saturation levels of the output voltage set by the supply
voltage and ground. The characteristic shown in Figure 6-16(c) closely
resembles the trigonometric tanh function and so the characteristic of a
transistor circuit is often said to be tanh-like. A tanh-like response is also
obtained with a Gilbert mixer circuit.
One problem with the Gilbert mixer circuit of Figure 6-15(a) is the reduced
voltage swing resulting from three drain-source voltage drops between the
supply rails. The classic technique for solving this problem is to use a folded
cascode design. The folded Gilbert cell double-balanced mixer then becomes
the circuit of Figure 6-15(b) [6, 7]. This merges a cascode amplifier design
with the double-balanced Gilbert cell mixer in Figure 6-15(a). The result is a
mixer that can have a larger voltage swing.
Gilbert mixers, as enhanced so far, can have poor linearity in the sense
that the IF output is linear only for small RF input signals. This problem is
exasperated by the trend to reduce the supply voltage of integrated circuits.
The solution used is to replace each of the amplifying transistors (M2 and
M3 in Figures 6-15(a and b)) by multiple transistors with progressively
offset biasing that staggers the tanh-like transfer characteristic of a transistor
circuit to realize an effective transistor with more linear characteristics [6, 8].
This design is used in the up-converting mixer shown in Figure 6-17, of
a WCDMA transmitter RFIC. and the variation in biasing of the multiple
amplifying transistors is achieved using sizing of each of the current source
transistors in the bottom row. Alternatively, separate bias could be applied to
each current source transistor. Also, the double-balanced Gilbert cell mixer
in Figure 6-15(a) is duplicated to realize a quadrature mixer.
The CMOS mixer has evolved from the core concept based on the diode
ring mixer, first replacing the diodes by transistors, and then introducing
transistor structures that cope with limited voltage supply and transistor
nonidealities. This process is the essence of RF integrated circuit design.

6.3.4 Ring-Based Mixers

Diode ring mixers were discussed in Section 6.2.4 and circuit implementa-
tions were shown in Figures 6-9 and 6-10. Transistor-based mixer implemen-
tations using the ring mixer concept are shown in Figure 6-18.
MIXER AND SOURCE MODULES 201

Figure 6-17: Quadrature up-converting mixer of a WCDMA transmitter. BB is the low-frequency

baseband signal with (I) and (Q) components. BB is the inverse (negative) of BB. LO and LO
are the differential local oscillator signal, and RF is the single-ended output signal. The LO is at
2 GHz, the baseband input frequency is 5 ± 1.72 MHz, and the output frequency is 1.95 MHz.
After Yang and Gard [6] and Yang [7]. Copyright K. Gard and X. Yang, used with permission.

6.3.5 Summary

Mixer type Feature Disadvatages

Unbalanced Very simple. Suitable for high millimeter-wave and Requires bulky filters to separate
mixer terahertz frequencies. signals. Large spurious signals.
Single Suppression of many spurious tones. Suppression Requires high LO drive level
balanced of amplitude-modulated noise on LO. Increased
mixer linearity compared to unbalanced mixer.
Double Increased linearity compared to single-balanced Relies on symmetry and matched
balanced mixer. All ports of the mixer are inherently isolated diodes and/or transistors. On RFIC
mixer from each other. Can be implemented with passive calibration is used to balance mixer.
mixing elements not requiring biasing. Inherently
broadband.
Triple Provides increased linearity. The type most com- Increasingly complex circuitry.
balanced monly used in RFICs although commonly referred
mixer to as a double balanced mixer.
202 STEER MICROWAVE AND RF DESIGN: MODULES

(a) MOS ring mixer

(b) CMOS H-bridge ring mixer (c) CMOS 4-phase mixer

Figure 6-18: Transistor ring-type mixers.

6.4 Local Oscillator

In this section the local oscillator that is used to drive a mixer will be treated
as a black box. The detailed design of oscillators is considered in Chapter 5
of [9].

6.4.1 Phase Noise in Local Oscillators

This section builds on the discussion of noise in Section 4.2. The performance
of most RF and microwave systems is limited by oscillator noise. In an
oscillator the noise close to the carrier, or oscillation frequency, is called
flicker noise and it significantly affects system performance. (In RF amplifiers
flicker noise is generated but it is of much less concern than it is with
oscillators.) Noise close to the oscillation center frequency (tens of hertz to
a few megahertz away for RF and microwave oscillators) manifests itself as
random fluctuations of amplitude and phase of the carrier. The amplitude
fluctuations are quenched by saturation in the oscillator and so are not of
concern. Thus the important close-in noise is just phase noise. This phase
noise increases the smaller frequency offset from the oscillation frequency.
The main use of an LO is to drive a mixer. Then phase noise on the LO
is added to the phase of the signal converted by the mixer and so the LO
phase noise becomes phase noise on the converted signal. In a digitally
modulated communication system the components of the phase noise at the
symbol rate and down to a few of its subharmonics have the most impact
on the ability to demodulate signals. Phase noise outside the bandwidth of
MIXER AND SOURCE MODULES 203

(a) Open loop

(b) Closed loop (c) Closed loop

Figure 6-19: Log-log plot of oscillator noise spectra: (a) open-loop noise showing flicker noise
(1/fm) and white noise regions; (b) closed-loop noise with low-Q loop; and (c) closed-loop noise
with high-Q loop.

the communication signal can be filtered out. Also phase noise slower than
a few subharmonics below the symbol rate are compensated for in signal
processing (e.g., using error correction codes).
Most RF oscillators comprise free-running oscillators whose oscillation
frequency can be controlled by an applied DC or low-frequency voltage.
Superior performance is obtained by comparing a scaled down version of
an oscillator output to a high-precision reference oscillator such as a crystal
oscillator. Without the feedback the oscillator is said to be an open-loop
oscillator, and with feedback it is said to be a closed-loop oscillator.
For a free-running oscillator (i.e., an open-loop oscillator), the phase noise
close to the carrier (i.e. the average oscillation frequency) is dominated by
flicker phase noise, as shown in Figure 6-19(a). This describes the intrinsic
noise property of the active device (and surrounding circuitry) and the white
noise and flicker noise responses are clearly seen. When the loop is closed,
the loop transfer characteristic shapes the noise response, producing noise
3
that has regions close to the carrier that has a 1/fm shape, and further
2
from the carrier it varies as 1/fm if the Q of the loop is low [10]. The
3 2
switch from 1/fm to 1/fm dependence is at what is called the transition
or crossover frequency, fc (see Figure 6-19(b)). If the Q of the loop is high,
3
the phase noise profile will transition from the 1/fm regime directly to the
204 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 6-20: Voltage-

controlled oscillator. (a) Schematic (b) Tuning curve

1/fm regime, and again the transition frequency is fc (see Figure 6-19(c)).
The transition frequency is usually around a few kilohertz to hundreds
of kilohertz offset from the carrier for microwave transistors. A feedback
3
analysis that describes how 1/f and white noise are converted to the 1/fm ,
2
1/fm , or 1/fm characteristics was developed by Leeson in 1966 [10] and
others [11]. An advanced discussion of oscillator phase noise is given in
Section 5.8 of [9] and the discussion here is limited but sufficient when
using oscillators as modules. Here the traditional but approximate approach
developed by Leeson is followed.
Phase noise was formally defined in Equation (4.60) and roughly it is the
ratio of the phase noise power in a 1 Hz bandwidth of a single sideband (SSB)
to the total signal power. This is measured at a frequency fm offset from the
carrier and denoted L(fm ) with the units of dBc/Hz (i.e., decibels relative to
the carrier power per hertz).
The phase noise that is important in RF and microwave oscillators (having
2
relatively low Q) is usually dominated by a 1/fm shape. Then the phase noise
at 1 MHz (a common frequency for comparing the phase noise performance
of different oscillators) is related to the phase noise measured at fm by
 2
1 MHz
L(1 MHz) = L(fm ) − 10 log . (6.17)
fm

Another commonly used quantitative assessment of oscillator perfor-

mance is provided by the oscillator figure of merit, FOM1 , which accounts
for the DC power consumed [12]:
 2  
1 MHz PDC
FOM1 = L(fm ) − 10 log + 10 log , (6.18)
fm Pref

where Pref is conventionally taken as 1 mW and FOM1 is referenced to the

phase noise at 1 MHz.

6.5 Voltage-Controlled Oscillator (VCO)

As its name implies, a controlling voltage sets the frequency of the output of a
VCO. The block diagram of a VCO is shown in Figure 6-20(a) where a slowly
varying input signal, vi , determines the frequency, f0 , of the signal produced
by an oscillator. This is usually achieved by varying the capacitance of a
varactor diode in a resonator.
One of the important characteristics of a VCO is the tuning curve, which
plots the output frequency f0 against the applied tuning voltage vi as shown
in Figure 6-20(b). Ideally the tuning curve is a straight line, but in practice it is
shaped and the actual range over which the VCO is used is less than the full
range supported. The tuning property is described by the tuning constant,
MIXER AND SOURCE MODULES 205

Table 6-2: Comparison of RF VCOs. Phase noise is worst case over tuning range; RF output
power is the minimum. All oscillators are hybrids unless indicated by IC, denoting an integrated
circuit. If fm is not 1 MHz, then a 1/f 2 dependence is assumed for the phase noise to calculate
the phase noise at 1 MHz. The CMOS VCOs are quadrature VCOs producing two outputs 90◦
apart. After [13] with corrected FOM1 . (Pref = 1 mW, fref = 1 MHz.)
f0 fBW PRF PDC fm L(fm ) L(1 MHz) FOM1 FOM2 Reference
GHz MHz dBm mW MHz dBm dBm dBm dBm
/Hz /Hz /Hz /Hz
4.92 770 0 150 1 −128 −106 −157 SiGe HBT hybrid [13]
5.05 500 0 150 1 −130 −106 −155 SiGe HBT hybrid [13]
5.16 229 −0.43 24 1 −111 −98 −136 InGaP/GaAs HBT [14]
11.5 550 9 0.1 −91 −111 −138 GaAs MESFET [15]
9.33 440 3.3 30.5 1 −102 −87 −128 GaN HEMT [16]
6.40 150 5.5 173 0.1 −105 −125 −85 −127 SiGe HBT [17]
5.94 166 −4.0 8.1 1 −110 −94 −134 CMOS IC [18]
4.87 70 −4.0 4.8 1 −131 −124 −149 GaInP/GaAs HBT [19]
5.38 120 −4.0 12.8 1 −127 −108 −148 GaInP/GaAs HBT [20]
5.29 270 −5.5 14 1 −106 −94 −130 SiGe HBT [21]
2.17 385 11.2 1.9 0.6 −120 −125 −122 −150 CMOS IC [22]
1.72 262 −11.5 75 1 −129 −111 −153 InGaP/GaAs HBT [23]
4.80 1200 4.8 36 1 −111 −95 −141 SiGe BiCMOS IC [24]
9.35 2500 18.3 570 1 −110 −82 −144 GaN/SiC pHEMT [25]
1.72 261 −10.3 55 1 −120 −103 −144 InGaP/GaAs HBT [26]
4.17 70 −6.1 102 1 −116 −96 −134 GaInP/GaAs HBT [27]
2.09 360 20.8 3 −140 −130 −117 CMOS VCO [28]
1.53 330 21.2 0.6 −133.5 −138 −125 CMOS VCO [29]
4.89 650 22 1 −124 −111 CMOS VCO [30]
1.85 280 20 3 −143 −133 −120 CMOS VCO [30]

which is also known as the tuning gain, K0 . This is the change in oscillation
frequency for a change in control voltage. For the VCO in Figure 6-20,

∆f0
K0 = . (6.19)
∆vi
The performance of a microwave VCO is one of the most competitive
aspects of RF design, as every decibel reduction in phase noise greatly
increases overall system performance. A high-performance VCO also relaxes
demands on other system components. While FOM1 (see Equation (6.18))
serves as a useful metric to compare VCOs, another FOM with bandwidth
weighting provides a better comparison of the performance of different
VCOs. This second figure of merit is [13]
 2  
1 MHz fBW
FOM2 = L(fm ) − 10 log − 10 log , (6.20)
fm fref

where fBW is the tuning bandwidth and fref is the reference bandwidth, taken
here as 1 MHz. Again the phase noise is referenced to 1 MHz. A number
of high-performance microwave oscillators are compared in Table 6-2. The
best phase noise that can typically be achieved by VCOs operating in the 1–
10 GHz range is −130 dBc/Hz at 1 MHz. This compares to the phase noise
component of white noise at standard temperature, which was shown in
Section 4.2.2 to be −177 dBc/Hz.
206 STEER MICROWAVE AND RF DESIGN: MODULES

(a) (b) (c)

Figure 6-21: A phase detector: (a) block diagram; (b) a digital phase detector using an XOR gate;
and (c) analog phase detector using a multiplier and filtering (not shown).

6.6 Phase Detector

A phase detector, also called a phase comparator, compares two waveforms
and the output of the phase detector is a representation of the phase
difference of the signals. There are two basic types of phase detectors:
sinusoidal phase detectors and square signal phase detectors, which operate
either in binary mode or by detecting zero crossings. The block diagram of a
phase detector is shown in Figure 6-21(a) with the output y(t) related to the
difference of the phase of the input signals x(t) and w(t)
A square wave detector is based on a logic circuit producing a signal that
is averaged (or integrated) over time. An example is the XOR gate shown
in Figure 6-21(b), which compares two digital signals that here have the
same frequency but are shifted in phase. The output of the XOR gate, y(t),
is also a pulse train, and the average value of y(t) is proportional to the
phase difference of x(t) and w(t). This average value can be obtained by
passing y(t) through a lowpass filter to obtain a DC value. If x(t) and w(t)
have different but close frequencies, the output of the lowpass filter will
be a slowly varying signal. This circuit could be used to detect the phase
difference of analog signals, but the signals first need to be converted to
digital signals. This can be obtained through saturating amplification of the
signals or by circuitry that responds to zero-crossings and produces a binary
signal. Other digital phase comparators can be realized using charge-pumps,
flip-flops, and sample-and-hold circuits.
Sinusoidal phase detectors can use a mixer or an analog multiplier, as
shown in Figure 6-21(c). The inputs could be two sinusoids with one of them
adjusted in phase either externally or within the detector itself. So consider
the inputs to the phase detector to be

x(t) = Ax sin(ωx t + φx ) and w(t) = Aw cos(ωw t + φw ). (6.21)

Then the output of the multiplier is

y(t) = Ax sin(ωx t + φx )Aw cos(ωw t + φw )

= 12 Ax Aw [sin(ωx t + ωw t + φx + φw ) + sin(ωx t − ωw t + φx − φw )] .
(6.22)

Typically the frequencies of the two input signals are close so that ωx =
ω + 12 ∆ω and ωw = ω − 12 ∆ω, where ∆ω is small and in most phase detector
applications either ∆ω = 0 or a feedback loop attempts to set ∆ω to zero so
there is little frequency error. So, following lowpass filtering, the output of
MIXER AND SOURCE MODULES 207

(a) Nonlinear frequency multiplier

Figure 6-22:
Frequency
(b) Resistive multiplier (c) Tuned multiplier multipliers.

the phase detector is

y(t) = 12 Ax Aw sin(∆ωt + φx − φw ). (6.23)

Now ∆ωt can be considered to be a phase error and incorporated in effective

phases of the input signals, ψx (t) = 12 ∆ωt + φx and ψw (t) = − 21 ∆ωt + φw .
Then

y(t) = 12 Ax Aw sin[ψx (t) − ψw (t)]. (6.24)

This is a functional phase detector provided that the difference in the phases
of the input signals is between −π/2 and π/2.
A variation of the sinusoidal detector is called a phase-frequency detector
(PFD) which has an extended range and the effective phase difference of the
input signals can have a magnitude greater than π/2 [31, 32].
The output signal, y(t), is proportional to the amplitudes of the input
signals and to their phase difference. In practice the amplitudes of both input
signals are scaled to a constant amplitude so that the output only depends
on the phase difference.
When a mixer is used in a phase detector, it operates much the same way
as a multiplier. With a switching mixer one of the input signals becomes
the drive of the switching part of the mixer and the other signal is the
sinusoidal drive. Now only the second input need be scaled to have a
constant amplitude.

6.7 Frequency Multiplier

There are three main types of microwave frequency multipliers. One uses
a nonlinear resistive element to generate harmonics and a bandpass filter
selects the appropriate harmonic for the output (see Figure 6-22(a)) [33]. A
second type uses a reactive element but is inherently narrow band. A third
type uses a mixer.
An example is the diode frequency multiplier shown in Figure 6-22(b).
If unbiased, the diode conducts current only during the positive half-cycle
of the input signal, creating a voltage across the diode that is rich in
harmonics. If no external bias is applied, some of the positive voltage swing
is required to turn the diode on. This type of multiplier is referred to as a
resistive frequency multiplier. The transistor circuit in Figure 6-22(c) uses
the resistive multiplier principle to achieve frequency multiplication. Now
208 STEER MICROWAVE AND RF DESIGN: MODULES

the bandpass filter is the tuned circuit in the collector leg of the amplifier.
The transistor is biased to realize a strong nonlinearity and the base-emitter
junction has an exponential current-voltage characteristic.
If the input signal is

x(t) = A cos(ωt + φ) (6.25)

then the output at the n harmonic is

y(t) = An cos(nωt + nφ). (6.26)

One of the issues with all frequency multipliers is the multiplication of the
phase noise on the input signal. If φ represents the phase noise on the original
signal, then the phase noise of the output signal will be increased by a factor
n. That is, even if the frequency multiplier introduces no noise of its own, the
signal-to-noise (i.e., phase noise) ratio of a signal will be reduced by a factor
n. The amplitude of the output signal, An , will usually be much less than
that of the input signal unless the nonlinear circuit incorporates an amplifier.
The increase in phase noise and power loss (without amplification) are the
major drawbacks of using a resistive nonlinear element to produce frequency
multiplication.
The bandwidth of a resistive frequency multiplier is limited by the
maximum bandwidth of a filter that will select just one harmonic at the
output. That is, if the input signal is at 10 GHz and with 10× multiplication
the output will be at 100 GHz. The useful fractional input bandwidth is
1
1/n = 10 of the input signal, otherwise two harmonics could appear
simultaneously in the output.
Another type of frequency multiplier uses a reactive element such as
the nonlinear capacitance of a reverse-biased semiconductor diode (i.e., a
varactor diode) [34]. This mixer is called a reactive frequency multiplier or
a parametric frequency multiplier. The nonlinear reactive element is part of
a resonant input circuit at the input frequency and also part of a resonant
output circuit at the output frequency. The efficiency of this type of mixer is
higher than for a resistive mixer but the bandwidth is much lower.
A third type of frequency multiplier uses a mixer. Some, but not all, mixers
can be used to realize frequency multiplication by applying the same signal
to the two input ports. The suitable mixer circuits are the balanced mixers
but not the mixers that rely on filtering to separate LO, RF, and IF signals.
A further type of frequency multiplier uses a phase-locked loop.
Frequency division in the feedback loop results in frequency multiplication
of the input signal. Yet another type uses flip-flops working on a microwave
frequency binary clock signal and this is the type often used in modern RFICs
to provide a square wave drive to a switching mixer.

6.8 Frequency Divider

A frequency divider is a module that reduces the frequency of a signal.
There are three main types of frequency dividers: those that work with
square waves and those that work with sinusoidal signals. The square wave
dividers are much simpler. A divide-by-2 square wave divider is shown in
Figure 6-23. The square wave input can be produced from a sinusoidal signal
using zero-crossing detection or a high-gain circuit with saturating levels.
MIXER AND SOURCE MODULES 209

Figure 6-23: Digital fre-

quency divider with the
waveforms shown for a
divide-by-2 divider.

Figure 6-24: Regenerative fre-

quency divider.

Figure 6-25: A phase-locked

loop with a phase-detector
and a frequency divider indi-
cated by (1/N ).

One type of analog frequency divider is the regenerative frequency divider

shown in Figure 6-24. The key element of this circuit is the mixer, which here
produces an output at the difference frequency of the input at frequency fi
and the signal is fed back at frequency fx . The output of the mixer is the
lowpass filtered difference of fi and fx . The loop stabilizes to produce the
divided frequency at the output.
Another analog frequency divider, called a locked-oscillator frequency
divider, uses injection locking of an oscillator [35]. It is relatively easy to
lock many oscillators by injecting a signal near the nth harmonic of the free-
running oscillation frequency. Then the oscillation frequency shifts and the
output has a frequency 1/nth that of the input signal.
Yet another type uses flip-flops dividing the frequency of a microwave-
frequency binary clock signal.

6.9 Phase-Locked Loop (PLL)

A phase-locked loop (PLL) is a feedback system in which the frequency and
phase of an output signal is related to the frequency and phase of an input
signal. The block diagram of a PLL is shown in Figure 6-25. An input signal
x(t) is compared to a feedback signal z(t). The frequency of y(t) will be
the average frequency of x(t). The way the loop achieves this is that the
output of the phase detector is proportional to the phase difference of x(t)
and z(t). This is then filtered by the block F (s) (generally a lowpass filter)
to produce a DC-like signal that drives a VCO. The filtering block, F (s),
removes undesired components from the phase detector and also sets the
dynamic response of the PLL.

6.9.1 Operation
The operation of the PLL in Figure 6-25 can be modeled as a linear system
with the assumption that x(t) and z(t) (= y(t) here) are nearly periodic
signals. Thus, approximately,

x(t) = Ax cos(ωx t + φx ) and z(t) = Aw cos(ωz t + φz ). (6.27)

210 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 6-26: Linearized model of an ana-

log phase-locked loop.

PLLs require that the radian frequency ωx be close to ωz , which is near the
free-running frequency of the VCO. (In practice, ωx is usually much less than
ωz and a frequency divider is used to divide the signal at ωz to obtain a signal
with a frequency close to ωx . The analysis is similar to that presented here.)
This defines the capture range of the PLL. Therefore Equation (6.27) can be
written
x(t) = Ax cos(Θx (t)) and z(t) = Aw cos(Θ0 (t)). (6.28)
The phases Θx and Θ0 incorporate the original phases φx and φw ,
respectively, and the effective time-dependent phase difference due to the
small time-dependent difference of the frequencies of x(t) and z(t). Analysis
of the PLL in Figure 6-25 begins with the voltage at the output of the phase
detector, v(t), which is proportional to the phase difference of the two input
signals and independent of their amplitude:
v(t) = Kd (Θx − Θ0 ), (6.29)
where Kd is the phase detector gain factor. The output of the phase detector
is filtered by the block with transfer function F (s). Usually this block is a
lowpass filter, but there are applications where it could be a bandpass filter
or have some other characteristic.
The output of the VCO is controlled by the voltage w(t) producing a signal
with frequency
f0 = fc + ∆f = fc + K0 v(t), (6.30)
where fc is the frequency of oscillation when the control voltage is zero and
K0 is the VCO gain factor.
Equations (6.29) and (6.30) describe the linear system shown in Figure 6-
26. The transfer function of this system is
θ0 Kd K0 F (s)/s G(s)
= = , (6.31)
θx 1 + Kd K0 F (s)/s 1 + G(s)
where G(s) = Kd K0 F (s). The phase error function is
 
Kd K0 F (s) sθx (s)
ǫ(s) = θx (s) − θ0 = θx 1 − = . (6.32)
s + Kd K0 F (s) s + Kd K0 F (s)
So the phase error function is directly related to the phase of the input signal
x(t). In the next section a particular choice of the transfer function F (s) is
used, leading to the identification of a particular PLL application.

6.9.2 First-Order PLL

Without a filter (i.e., F (s) = 1) the PLL is referred to as a first-order PLL or
first-order loop and its transfer function is
θ0 (s) Kd K0 /s Kd K0
= = . (6.33)
θx (0) 1 + Kd K0 /s s + Kd K0
MIXER AND SOURCE MODULES 211

This seems to be the transfer function of a lowpass filter, but the input to the
system is phase and the output is the phase deviation from the free-running
frequency of the VCO. The phase error function of the first-order PLL is
 
Kd K0 sθx (s)
ǫ(s) = θx (s) − θ0 = θx (s) 1 − = . (6.34)
s + Kd K0 s + Kd K0

The steady-state response of the first-order PLL to specific input signals

will now be derived. The long-term (steady-state) behavior is developed
using the final-value theorem. A function H(s) with all poles in the left-half
plane has a time-domain response h(t) with long-term behavior given by

lim h(t) = lim sH(s). (6.35)

t→∞ s→0

Consider now the response to a step change, ∆θx , of the phase of the input
signal. The Laplace transform of a step function of magnitude ∆θx is ∆θx /s
and this is what is used in Equation (6.34). Thus the steady-state error
function, ǫss = ǫ(t) as t → ∞, is
 2   2   
s θx (s) s ∆θx /s s∆θx
ǫss = lim = lim = lim
s→0 s + Kd K0 s→0 s + Kd K0 s→0 s + Kd K0

= 0. (6.36)

Equation (6.36) indicates that a first-order PLL will eventually track a phase
change of the input signal. However, it will not respond immediately as the
loop has a lowpass characteristic. In effect the PLL is acting as a high-Q
bandpass filter.
Now consider a step change in the frequency of the input signal. If the
step change in the radian frequency of x(t) is ∆ωx , then the resulting phase
change will be a ramp so that, in the Laplace domain, θ(s) = ∆ωx /s2 . Then
the steady-state error is
 2
s ∆ωx /s2
  
∆ωx
ǫss = lim = lim
s→0 s + Kd K0 s→0 s + Kd K0

= ∆ωx /(Kd K0 ). (6.37)

Equation (6.37) indicates that a first-order PLL will follow a change in

frequency with a phase error that is proportional to the magnitude of the
frequency change. So if the input signal x(t) is a frequency modulated
signal, the first-order PLL will demodulate the signal and recover the original
modulation as the error signal. Thus a first-order PLL demodulates an FM
signal, that is, it is a frequency demodulator.

6.9.3 Applications
Three applications of the PLL are shown in Figure 6-27. A frequency
synthesizer is used to create a signal locked to a fixed-frequency and very
accurate reference oscillator but at another frequency (see Figure 6-27(a)).
Typically the reference is a precision low-frequency reference oscillator, such
as a quartz crystal oscillator [36]. The output frequency is not at the same
frequency as the reference oscillator. This is accomplished by including a
frequency divider in the PLL. Normally the division factor, N , is an integer
212 STEER MICROWAVE AND RF DESIGN: MODULES

(a) Frequency synthesizer

(b) Phase modulator

Figure 6-27: Applications

using phase-locked loops. (c) Coherent amplitude demodulator

so that the output signal’s frequency will be an integer multiple of the

input frequency. A fractional-N frequency synthesizer is achieved using
two integer dividers with division factors N and M . A controller alternates
between the two division factors so that the VCO tends to output first
one frequency and then the next. The VCO stabilizes at a frequency that
is the weighted average of the two division factors. If the frequency is
divided by M for a time τM , then by N for time τN , and repeated, then
the effective division factor is (τM M + τN N )/(τM + τN ). The loop dynamics
(determined by the filter F (s)) is chosen so that the VCO can only change
more slowly than the toggling of the division factors. The division factors,
and the toggling between its two values, are chosen to synthesize an output
signal with the desired frequency.
The second application of the PLL is a phase modulator, see Figure 6-27(b).
The phase modulating signal m(t) effectively is adding to the phase error
generated by the phase detector. This results in a change in the phase of the
VCO output compensating for the inserted phase change.
The third application of the PLL is the coherent amplitude demodulator
shown in Figure 6-27(c). The input is an amplitude modulated signal

x(t) = A [1 + m(t)] sin(ω0 t), (6.38)

which is demodulated to recover m(t) by multiplying x(t) by an LO signal

with the same carrier frequency. The characteristic of F (s) is chosen so
that the PLL generates the LO signal as the loop establishes the oscillation
MIXER AND SOURCE MODULES 213

Figure 6-28: Direct digital synthesis (DDS)

module.

frequency as the central frequency of the modulated input. The loop filter
F (s) ensures that the VCO frequency cannot change too quickly and the
output frequency of the VCO approximates a sinusoidal signal having a
radian frequency ω0 with a phase offset. Thus
w(t) = x(t) sin(ω0 t + θ) = A [1 + m(t)] 21 [cos θ − cos(2ω0 tθ )] . (6.39)
Following lowpass filtering the output is
y(t) = A [1 + m(t)] cos θ. (6.40)
With loop dynamics chosen so that θ is small, the original modulation, m(t),
is recovered. This AM demodulator performs well when the input SNR is
low as demodulation is coherent.

6.10 Direct Digital Synthesizer

A direct digital synthesizer (DDS) uses a DSP unit, commonly called
a numerically controlled oscillator, to create the digital version of a
modulated waveform (see Figure 6-28). The digital output of the DSP drives
a DAC that is filtered by a lowpass filter producing the analog output. A DDS
module enables a fairly complex RF system to be realized using just a few
components, as frequency generation and frequency synthesis are embodied
in one unit. The accuracy of the frequency of the analog signal is established
by the reference oscillator that acts as a precision clock for the DSP and DAC.
The numerically controlled oscillator has an accumulator, often called a
phase accumulator, that counts out the phase of the analog signal to be
synthesized. If there is modulation, then it is incorporated at this stage. Both
modulation and the center frequency of the modulated output are changed
by changing the accumulator’s step size. The accumulator drives a sine look-
up table that then produces the digital words presented to the DAC.
A DDS has the advantage that the output frequency can be changed
almost instantaneously which is in contrast to a PLL in which the rate of
frequency change of the output is limited by the bandwidth of the loop filter.
A major disadvantage of a DDS is the number of spurious signals generated
by quantization issues, and this is mostly related to the bit length of the
accumulator. There are also aliasing or image signals generated as images
of the intended signal relative to the clock frequency. However, the spurious
signals from two DDS units with the same reference oscillator are correlated
and this can be exploited in some system architectures to remove the effect of
spurious signals. Most of the negative characteristics of a DDS are reduced by
using a longer accumulator and hence a DAC with a larger number of input
bits. This, however, leads to increased power consumption. These drawbacks
are sometimes insignificant when compared to the frequency agility and
system simplicity achieved.
214 STEER MICROWAVE AND RF DESIGN: MODULES

(a) Tunnel diode (b) IMPATT diode

Figure 6-29: Two ter-

minal semiconductor
sources (c) Gunn diode

6.11 Diode and Vacuum Sources

There are many types of sources based on special properties of diodes
or of vacuum devices. This section describes some of these sources.
While sources based on transistor circuits are preferred for integration and
manufacturability, diode and vacuum sources are in widespread use because
of either the high power that can be efficiently generated or because they
operate at high frequencies beyond the reach of transistor circuits.

6.11.1 Two-Terminal Semiconductor Sources

Two-terminal semiconductor sources are based on exploiting the negative
resistance presented at the terminals of several types of devices [37–40]. They
either exploit a dynamic negative resistance created in a high field region of
a semiconductor due to a charge imbalance, or they exploit the finite time
for carriers to cross a semiconductor region and thus produce an RF current
that is out of phase with the applied voltage, thus presenting a negative RF
resistance. Three representative devices are the tunnel diode, the IMPATT
diode, and the Gunn diode.

Tunnel Diode
A tunnel diode is a pn junction diode in which the conduction band states
on the n side are filled with electrons and these line up with empty valance
band states (i.e. holes) on the p side [37–46]. This results in a very narrow pn
junction barrier. For negative voltages and small and large applied voltages
the diode acts as a conventional pn junction diode with an exponential
current-voltage characteristic. However, as the applied voltages increase
above zero, the conduction and valance bands become more misaligned
and the voltage eventually drops before increasing again. This effect is
due to quantum mechanical tunnelling. This is seen in the current-voltage
characteristic of the tunnel diode shown in Figure 6-29(a), where the drop in
MIXER AND SOURCE MODULES 215

current creates a negative dynamic resistance. Embedded in an appropriate

circuit, even as simple as a parallel RLC circuit, the tunnel diode is the active
element of an RF oscillator. Usually a Gunn diode has a very high dopant
concentration so that the reverse breakdown voltage is very low.

IMPATT Diode
An IMPATT diode (IMPact ionization Avalanche Transit-Time) produces
high power and throughout the history of its use it has produced the highest
power levels at the highest frequency, up to 1000 GHz, of any semiconductor
device [37–40, 47–51].
The IMPATT diode is the most important of the transit time semiconductor
diodes, see Figure 6-29(b). In transit time devices the generation of charge
carriers is concentrated in one narrow region of the diode. In an IMPATT
diode, a high field at the boundary between a highly doped n region, the
n+ region, and a lightly doped n region leads to avalanche multiplication
producing holes and electrons. The holes are quickly collected by an adjacent
metal contact and the electrons transit through a drift region with usually
intrinsic doping and a constant field. If the drifting electrons are sufficiently
delayed, then the RF current through the device will be out of phase with
the applied RF voltage (superimposed on a biasing DC voltage) and a
negative RF resistance is presented at the device terminals. The roles of the
holes and electrons can also be exchanged. Thus the IMPATT diode can be
used as the active component of an oscillator [52]. They can be used in an
amplifier as a reflection device having a reflection coefficient greater than
one [53]. However, the oscillating signal produced has high noise due to the
underlying avalanche process.
Other effects can produce charges that eventually drift and produce a
negative RF resistance. An example is a TUNNETT diode that injects charges
through tunneling [54, 55]. Another device is the TRAPATT diode (trapped
plasma avalanche transit time diode), a pn junction diode, where the carrier
injection results from a trapped space-charge plasma formed within the
junction region [42, 56].

Gunn Diode
A Gunn diode is also called a transferred electron device or a Gunn effect
device. While strictly not a diode as there is not a junction, the name Gunn
diode has become common usage because there are two electrodes. The
structure and current-voltage characteristic of a Gunn diode are shown
in Figure 6-29(c) [37–40, 57–61]. The device has three n-type regions: two
heavily doped n+ regions at each contact, separated by a lightly doped n−
region. When a voltage is applied to the device, most of the voltage is across
the n− region and the device acts like a resistor with the current through
the Gunn diode proportional to the voltage across it. At higher voltages the
conductivity of the n− region drops and the current drops, so that there is a
region of negative dynamic resistance.
216 STEER MICROWAVE AND RF DESIGN: MODULES

6.11.2 Vacuum Devices

Some microwave systems use vacuum tubes to obtain high RF powers; the
most common vacuum tube devices are reviewed here.

Magnetron
The magnetron is the original device used for generating microwave power
and was invented during World War II for use in radar equipment. It is
most commonly used in microwave ovens, where it is the most efficient
means of producing microwave power at 2.4 GHz. It is used in military
systems today to produce megawatts of pulsed RF power and is used to
generate substantial power up to a few terahertz [62–65]. In a magnetron a
circular chamber, containing the cathode, is surrounded by and connected to
a number of resonant cavities. The walls of the chamber are the anode. The
cavity dimensions determine the frequency of the output signal. A strong
constant magnetic field in the chamber causes electrons that want to flow
from the cathode to the anode to rotate. As the electrons pass the entrance
of the circular cavities, the electrons interact with the EM field in the cavity,
enhancing the field at a characteristic frequency, which is the frequency of
oscillation of the magnetron.

Klystron
The klystron is a long, narrow vacuum tube with an electron gun (the
cathode) at one end and an anode at the other [66–69]. In between is a series
of doughnut-shaped resonant cavities aligned so that the electron beam from
the cathode passes through the hole. As the beam passes the cavities, small
changes in the electron beam affect the EM field in the cavities. The EM fields
in the cavities begin to oscillate, which in turn affect the passing electron
beam. A feedback effect results, and when the last and first resonant cavities
are connected, a large oscillating microwave signal is produced.

Traveling Wave Tube and Backward Wave Oscillator

A traveling wave tube (TWT) is a long vacuum tube with an electron gun
(the cathode) at one end and a collector (the anode) at the other. The electron
beam produced by the gun travels down the center of a long wire helix
[70–75]. An excitation coupling probe, an antenna, placed near the cathode
introduces a microwave signal. This modulates the electron beam, which in
turn induces an RF current in the helix. The helix is designed so that the EM
signal supported by the helix travels at the same speed as the electron beam
so that the electron beam and EM signal guided by the helix continuously
interact. Thus the RF current in the helix as well as the RF field grow in
strength along the tube. An output probe at the anode end of the tube couples
to the RF field and delivers an amplified version of the input RF signal. As
an amplifier the TWT is called a traveling wave tube amplifier (TWTA).
In the TWT the helix slows the RF signal down to match the speed of
the electrons and so the helix is called a slow-wave structure. Another tube
device that uses a slow-wave structure is the backward wave oscillator
(BWO). However, in the BWO the beam is directed against the traveling
wave supported by the helix.
MIXER AND SOURCE MODULES 217

6.12 Summary
This chapter described a number of complex mixer and oscillator modules
that are available to assemble high-performance RF and microwave systems.
These modules and the other modules used in this book are often
available as off-the-shelf items from companies that specialize in designing
modules. This is a cost effective way of developing low and medium
volume microwave systems such as basestation hardware. For a competitive
advantage a company selling a system or subsystem may develop some of
the modules themselves. For a high volume market, say for a cell phone,
many of the modules would be integrated on-chip. Design is then an
expensive proposition but overall performance and size would be optimized.

6.13 References
[1] S. Maas, Microwave Mixers. Artech House, Oscillators. Kluwer, 1999.
1986. [13] A. Victor and M. Steer, “Reflection coefficient
[2] M. Steer and P. Khan, “An algebraic formula shaping of a 5-GHz voltage-tuned oscilla-
for the output of a system with large-signal, tor for improved tuning,” IEEE Trans. on Mi-
multifrequency excitation,” Proc. of the IEEE, crowave Theory and Techniques, vol. 55, no. 12,
vol. 71, no. 1, pp. 177–179, 1983. pp. 2488–2494, Dec. 2007.
[3] R. Pettai, Noise in Receiving Systems. John [14] S.-S. Myoung and J.-G. Yook, “Low-phase-
Wiley & Sons, 1996. noise high-efficiency MMIC VCO based on
[4] N. Carvalho, J. Pedro, W. Jang, and M. Steer, InGaP/GaAs HBT with the LC filter,” Mi-
“Nonlinear simulation of mixers for assess- crowave and Optical Technology Letters, vol. 44,
ing system-level performance,” Int. J. Mi- no. 1, pp. 123–126, Jan. 2005.
crowave Millimeter Wave Computer Aided Engi- [15] C.-H. Lee, S. Han, B. Matinpour, and
neering, vol. 15, no. 4, pp. 350–361, Jul. 2005. J. Laskar, “A low phase noise X-band MMIC
[5] B. Gilbert, “A precise four-quadrant multi- GaAs MESFET VCO,” IEEE Microwave and
plier with subnanosecond response,” IEEE J. Guided Wave Letters, vol. 10, no. 8, pp. 325–
Solid-State Circuits, vol. 3, no. 4, pp. 365–373, 327, Aug. 2000.
Dec. 1968. [16] Z. Cheng, Y. Cai, J. Liu, Y. Zhou,
[6] X. Yang, A. Davierwalla, D. Mann, and K. Lau, and K. Chen, “A low phase-
K. Gard, “A 90nm CMOS direct conversion noise X-band MMIC VCO using high-
transmitter for WCDMA,” in 2007 IEEE Ra- linearity and low-noise composite-channel
dio Frequency Integrated Circuits (RFIC) Symp., Al0.3 Ga0.7 N/Al0.05 Ga0.95 N/GaN hemts,”
Jun. 2007, pp. 17–20. IEEE Trans. on Microwave Theory and Tech-
[7] X. Yang, “90nm cmos transmitter design for niques, vol. 55, no. 1, pp. 23–29, Jan. 2007.
WCDMA,” Ph.D. dissertation, North Car- [17] H. Zirath, R. Kozhuharov, and M. Ferndahl,
olina State University, 2009. “Balanced colpitt oscillator mmics designed
[8] M. Ding, K. Gard, and M. Steer, “A highly for ultra-low phase noise,” IEEE J. of Solid-
linear and efficient CMOS RF power ampli- State Circuits, vol. 40, no. 10, pp. 2077–2086,
fier with a new circuit synthesis technique,” Oct. 2005.
IEEE Trans. on Microwave Theory and Tech- [18] Y.-K. Chu and H.-R. Chuang, “A fully inte-
niques, vol. 60, no. 8, pp. 1–2, Nov. 2012. grated 5.8-GHz U-NII band 0.18–µ;m CMOS
[9] M. Steer, Microwave and RF Design, Amplifiers VCO,” IEEE Microwave and Wireless Compo-
and Oscillators, 3rd ed. North Carolina State nents Letters, vol. 13, no. 7, pp. 287–289, Jul.
University, 2019. 2003.
[10] D. Leeson, “A simple model of feedback os- [19] C. Meng, Y. Chang, and S. Tseng, “4.9-
cillator noise spectrum,” Proc. of the IEEE, GHz low-phase-noise transformer-based
vol. 54, no. 2, pp. 329–330, Feb. 1966. superharmonic-coupled GaInP/GaAs HBT
[11] A. Hajimiri and T. Lee, “Design issues in QVCO,” IEEE Microwave and Wireless Compo-
cmos differential lc oscillators,” IEEE J. of nents Letters, vol. 16, no. 6, pp. 339–341, Jun.
Solid-State Circuits, vol. 34, no. 5, pp. 717–724, 2006.
May 1999. [20] C. Meng, C. Chen, Y. Chang, and G. Huang,
[12] P. Kinget, Integrated GHz Voltage Controlled “5.4 ghz-127 dbc/hz at 1 MHz GaInP/GaAs
218 STEER MICROWAVE AND RF DESIGN: MODULES

HBT quadrature VCO using stacked trans- 2002, pp. 539–542.

formers,” Electronics Letters, vol. 41, no. 16, [31] J. Savoj and B. Razavi, “A 10-gb/s cmos clock
pp. 33–34, Aug. 2005. and data recovery circuit with a half-rate bi-
[21] T. Hancock and G. Rebeiz, “A novel super- nary phase/frequency detector,” IEEE J. of
harmonic coupling topology for quadrature Solid-State Circuits, vol. 38, no. 1, pp. 13–21,
oscillator design at 6 ghz,” in 2004 IEEE Ra- Jan. 2003.
dio Frequency Integrated Circuits (RFIC) Symp. [32] H. Johansson, “A simple precharged cmos
Digest of Papers, Jun. 2004, pp. 285–288. phase frequency detector,” IEEE J. of Solid-
[22] S.-W. Yoon, S. Pinel, and J. Laskar, “A 0.35- State Circuits, vol. 33, no. 2, pp. 295–299, Feb.
µm CMOS 2-GHz VCO in wafer-level pack- 1998.
age,” IEEE Microwave and Wireless Compo- [33] A. Raisanen, “Frequency multipliers for mil-
nents Letters, vol. 15, no. 4, pp. 229–231, Apr. limeter and submillimeter wavelengths,”
2005. Proc. of the IEEE, vol. 80, no. 11, pp. 1842–
[23] J.-H. Yoon, S.-H. Lee, A.-R. Koh, G. Kennedy, 1852, Nov. 1992.
P. Gary, and N.-Y. Kim, “Optimized phase [34] D. Holcomb, “Parametric frequency multi-
noise of LC VCO using an asymmetric- plier,” US Patent US Patent 3 076 133, Jan. 29,
inductance tank in InGaP/GaAs HBT tech- 1963.
nology,” Microwave and Optical Technology [35] M. Tiebout, “A cmos direct injection-locked
Letters, vol. 48, no. 6, pp. 1035–1040, Jun. oscillator topology as high-frequency low-
2006. power frequency divider,” IEEE J. of Solid-
[24] O. Esame, I. Tekin, A. Bozkurt, and Y. Gur- State Circuits, vol. 39, no. 7, pp. 1170–1174,
buz, “Design of a 4.2–5.4 GHz differential LC Jul. 2004.
VCO using 0.35 µm SiGe BiCMOS technol- [36] M. Odyniec, Ed., RF and Microwave Oscillator
ogy for ieee 802.11a applications,” Int. J. of Design. Artech House, 2002.
RF and Microwave Computer-Aided Engineer- [37] M. Shur, GaAs Devices and Circuits. Springer,
ing, vol. 17, no. 2, pp. 243–251, 2007. 1987.
[25] A. Maas and F. van Vliet, “A low-noise X- [38] K. Ng, Complete Guide to Semiconductor De-
band microstrip VCO with 2.5 GHz tuning vices. Wiley, 2002.
range using a GaN-on-SiC p-HEMT,” in 2005 [39] U. Mishra and J. Singh, Semiconductor De-
European Gallium Arsenide and Other Semicon- vices: Physics and Technology. John Wiley &
ductor Application Symp., (EGAAS 2005), Oct. Sons, 2008.
2005, pp. 1805–1808. [40] ——, Semiconductor Device Physics and Design.
[26] J.-H. Yoon, S.-H. Lee, A.-R. Koh, B. Shrestha, Springer, 2007.
S.-H. Cheon, G. Kennedy, and N.-Y. Kim, [41] B. B. Streetman and S. Banerjee, Solid State
“A novel harmonic noise frequency filtering Electronic Devices, 6th ed. Prentice Hall,
VCO for optimizing phase noise,” in 2006 2006.
IEEE MTT-S Int. Microwave Symp. Dig., Jun. [42] S. Sze and K. Ng, Physics of Semiconductor De-
2006, pp. 1805–1808. vices, 3rd ed. John Wiley & Sons, 2007.
[27] C. Meng, S. Tseng, Y. Chang, J. Su, [43] W. Chow, Principles of Tunnel Diode Circuits.
and G. Huang, “4-GHz low-phase-noise Wiley, 1964.
transformer-based top-series GaInP/GaAs [44] L. Esaki, “Discovery of the tunnel diode,”
HBT QVCO,” in 2006 IEEE MTT-S Int. Mi- IEEE Trans. on Electron Devices, vol. 23, no. 7,
crowave Symp. Dig., Jun. 2006, pp. 1809–1812. pp. 644–647, 1976.
[28] P. Andreani and X. Wang, “On the phase- [45] K. Ng, Complete Guide to Semiconductor De-
noise and phase-error performances of mul- vices. Wiley, 2002, ch. Tunnel diode.
tiphase lc cmos vcos,” IEEE J. of Solid-State [46] D. Schroder, Semiconductor Material and De-
Circuits, vol. 39, no. 11, pp. 1883–1893, Nov. vice Characterization. IEEE Press and Wiley,
2004. 2006.
[29] P. Vancorenland and M. Steyaert, “A 1.57 [47] T. Midford and R. Bernick, “Millimeter-wave
GHz fully integrated very low phase noise cw impatt diodes and oscillators,” IEEE
quadrature vco,” in 2001 Symp. on VLSI Cir- Trans. on Microwave Theory and Techniques,
cuits Dig. of Technical Papers, 2001, pp. 111– vol. 27, no. 5, pp. 483–492, May 1979.
114. [48] H. Xu, M. Morschbach, J. Werner, and
[30] S. Gierkink, S. Levantino, R. Frye, and V. Boc- E. Kasper, “Monolithic integrated oscillator
cuzzi, “A low-phase-noise 5 GHz quadrature with silicon impatt diode for automotive
CMOS VCO using common-mode inductive radar applications,” in European Microwave
coupling,” in Proc. of the 28th European Solid- Integrated Circuit Conference, 2008, Oct. 2008,
State Circuits Conf., (ESSCIRC 2002), Sep. pp. 20–23.
MIXER AND SOURCE MODULES 219

[49] P. Mukherjee and B. Gupta, “Terahertz (thz) org/54287276.

frequency sources and antennas—a brief re- [62] N. Alekseev, D. Malairov, and I. Bensen,
view,” Int. J. Infrared and Millimeter Waves, “Generation of high-power oscillations with
vol. 29, no. 12, pp. 1091–1102, 2008. a magnetron in the centimeter band,” Proc. of
[50] M.-S. Gupta, R. Lomax, and G. Haddad, the IRE, vol. 32, no. 3, pp. 136–139, Mar. 1944.
“Noise considerations in self-mixing impatt- [63] J. Osepchuk, “The magnetron and the mi-
diode oscillators for short-range doppler crowave oven: A unique and lasting relation-
radar applications,” IEEE Trans. on Microwave ship,” in 2010 Int. Conf. on the Origins and
Theory and Techniques, vol. 22, no. 1, pp. 37– Evolution of the Cavity Magnetron (CAVMAG),
43, Jan. 1974. Apr. 2010, pp. 46–51.
[51] W. Evans and G. Haddad, “A large signal [64] Gil Wong Choi, Hae Jin Kim, Hyoung-Jong
analysis of impatt diodes,” IEEE Trans. on Kim, and Jin-Joo Choi, “The self-injection-
Electron Devices, vol. 15, no. 10, pp. 708–717, locked magnetron,” in 2008 IEEE Interna-
Oct. 1968. tional Vacuum Electronics Conference (IVEC
[52] K. Kurokawa and F. Magalhaes, “An X-band 2008), Apr. 2008, pp. 445–446.
10-watt multiple-impatt oscillator,” Proc. of [65] T. Fleming, M. Lambrecht, and P. Mardahl,
the IEEE, vol. 59, no. 1, pp. 102–103, Jan. 1971. “Design and simulation of a mega-watt class
[53] M.-S. Gupta, “Large-signal equivalent circuit nonrelativistic magnetron,” IEEE Trans. on
for impatt-diode characterization and its ap- Plasma Science, vol. 40, no. 6, pp. 1563–1568,
plication to amplifiers,” IEEE Trans. on Mi- Jun. 2012.
crowave Theory and Techniques, vol. 21, no. 11, [66] G. Caryotakis, “The klystron: A microwave
pp. 689–694, Nov. 1973. source of surprising range and endurance,”
[54] H. Eisele, A. Rydberg, and G. Haddad, “Re- Physics of Plasmas, vol. 5, no. 5, pp. 1590–1598,
cent advances in the performance of InP Apr. 1998.
Gunn devices and GaAs TUNNETT diodes [67] A. M. Sessler and S. S. Yu, “Relativistic
for the 100-300-GHz frequency range and klystron two-beam accelerator,” Phys. Rev.
above,” IEEE Trans. on Microwave Theory and Lett., vol. 58, pp. 2439–2442, Jun. 1987.
Techniques, vol. 48, no. 4, pp. 626–631, Apr. [68] K. Hendricks, P. Coleman, R. Lemke, M. Ar-
2000. man, and L. Bowers, “Extraction of 1 gw
[55] J.-I. Nishizawa, T. Kurabayashi, Y. Miura, of RF power from an injection locked rela-
T. Sawai, P. Plotka, and M. Watenabe, “De- tivistic klystron oscillator,” Phys. Rev. Lett.,
velopment of sub-thz tunnett diode for vol. 76, pp. 154–157, Jan. 1996.
biomedical imaging,” in Infrared and Millime- [69] C. Beard, “Review of available power
ter Waves, 2007 and the 2007 15th International sources,” Nuclear Instruments and Methods in
Conference on Terahertz Electronics. IRMMW- Physics Research Section A: Accelerators, Spec-
THz. Joint 32nd International Conference on, trometers, Detectors and Associated Equipment,
Sep. 2007, pp. 150–151. vol. 557, no. 1, pp. 276–279, Feb. 2006.
[56] J. DeLoach, B.C. and D. Scharfetter, “Device [70] J. Pierce, “Theory of the beam-type traveling-
physics of trapatt oscillators,” IEEE Trans. on wave tube,” Proc. of the IRE, vol. 35, no. 2, pp.
Electron Devices, vol. 17, no. 1, pp. 9–21, Jan. 111–123, Feb. 1947.
1970. [71] Young-Min Shin, L. Barnett, and N. Luh-
[57] B. Bosch and R. Engelmann, The Gunn Effect. mann, “Phase-shifted traveling-wave-tube
Clarendon Press, 1974. circuit for ultrawideband high-power
[58] G. Hobson, The Gunn Effect. Clarendon submillimeter-wave generation,” IEEE Trans.
Press, 1974. on Electron Devices, vol. 56, no. 5, pp. 706–712,
[59] H. Eisele, “480 GHz oscillator with an InP Mar. 2009.
Gunn device,” Electronics Letters, vol. 46, [72] S. Bhattacharjee, J. Booske, C. Kory,
no. 6, pp. 422–423, Mar. 2010. D. van der Weide, S. Limbach, S. Gal-
[60] F. Amir, C. Mitchell, N. Farrington, and lagher, J. Welter, M. R. Lopez, R. Gilgenbach,
M. Missous, “Advanced Gunn diode as high R. L. Ives, M. E. Read, R. Divan, and
power terahertz source for a millimetre wave D. Mancini, “Folded waveguide traveling-
high power multiplier,” Proc. SPIE, vol. 7485, wave tube sources for terahertz radiation,”
pp. 74 850I–74 850I–11, Sep. 2009. IEEE Trans. on Plasma Science, vol. 32, no. 3,
[61] E. Alekseev, A. Eisenbach, D. Pavlidis, pp. 1002–1014, Jun. 2004.
S. Hubbard, and W. Sutton, “Development [73] W. Menninger, N. Robbins, D. Dibb, and
of GaN-based Gunn-effect millimeter-wave D. Lewis, “Power flexible ka-band traveling
sources,” Scientific Common, 2010. [Online]. wave tube amplifiers of up to 250-W RF for
Available: http://www.scientificcommons. space communications,” IEEE Trans. on Elec-
220 STEER MICROWAVE AND RF DESIGN: MODULES

tron Devices, vol. 54, no. 2, pp. 181–187, Feb. Science (ICOPS 2002), 2002, p. 171.
2007. [75] R. Grow, J. Baird, K. Bunch, and R. C.
[74] J. Neilson, R. Ives, M. Caplan, M. Mizuhara, Freudenberger, “Backward-wave oscillators
D. Marsden, and C. Kory, “High efficiency, for the frequency range from 300 GHz to 1
terahertz, backward wave oscillators,” in The THz,” in 2000 Int. Vacuum Electronics Confer-
29th IEEE International Conference on Plasma ence,, 2000.

6.14 Exercises
1. A mixer has an LO at 28.2 GHz. The mixer is (c) What is the signal-to-interference ratio (in
used to convert a signal at 28.1 GHz to an IF decibels) at the IF?
at 100 MHz, and has a conversion loss of 13 dB
8. A mixer has an LO at 100 GHz. The mixer is used
and an image rejection of 40 dB. Two signals are
to convert a signal at 110 GHz to an IF at 10 GHz.
presented to the mixer, one at 28.1 GHz with a
Two signals are presented to the mixer, one at
power of 1 pW and the other at 28.3 GHz with a
110 GHz with a power of 10 nW and an interfer-
power of 10 µW. [Parallels Example 6.1]
ing signal at 90 GHz with a power of 5 nW. If the
(a) What is the power of the (intended) signal at image rejection is 40 dB and the conversion loss
the IF in dBm? is 20 dB, what is the signal-to-interference ratio
(b) What is the signal-to-interference ratio at the (in decibels) at the IF?
IF (ignoring noise)?
2. Consider the single-ended diode mixer in Figure 9. A mixer has an LO at 18 GHz. The mixer is
6-5(a). used to convert a signal at 18.5 GHz to an IF at
500 MHz. The RF signal at 18.5 GHz has a power
1. Develop a symbolic expression for the volt-
of 100 pW. In addition, noise with a power
age at the test point. The diode is modeled
2 3 of 1 pW is applied to the mixer at 18.5 GHz
by iD = a1 vD + a2 vD + a3 vD .
and 17.5 GHz (1 pW at 18.5 GHz and 1 pW at
2. What are the frequencies and amplitudes of
17.5 GHz). If the image rejection is 6 dB and the
the components of the spectrum at the test
conversion loss is 10 dB. Ignore noise contribu-
point?
tions from the mixer. What is the SNR (in deci-
3. A mixer in a receiver has a conversion loss of bels) at the IF?
16 dB. If the applied RF signal has an available
power of 100 µW, what is the available power of 10. A mixer in a communication system has an
the IF at the output of the mixer? LO at 5.5 GHz. The mixer is used to convert a
10 MHz bandwidth signal at 5.6 GHz to an IF at
4. The RF signal applied to the input of a mixer has 100 MHz. The RF signal at 5.6 GHz has a power
a power of 1 nW and the output of the mixer at of 100 pW. The image rejection is ideal and the
the IF has a power level of 100 pW. What is the conversion loss is 10 dB. The mixer has a single-
conversion loss of the mixer in decibels? sideband noise figure of 6 dB.
5. A mixer in a receiver has a conversion gain of (a) What is the noise power at the input if
10 dB. If the applied RF signal has a power of the source is held at standard temperature
100 µW, what is the available power of the IF at (290 K)?
the output of the mixer? (b) What is the input SNR (in decibels)?
6. A mixer in a receiver has a conversion loss of (c) What is the SNR (in decibels) at the IF?
6 dB. If the applied RF signal has a power of
11. The double-balanced ring diode mixer shown
1 µW, what is the available power of the IF at
below has the special characteristic that the LO
the output of the mixer?
and RF tones are suppressed at the IF output
7. A mixer has an LO at 18 GHz. The mixer is port. Develop a symbolic expression for the volt-
used to convert a signal at 18.5 GHz to an IF age at the IF port. The diodes are matched and
at 500 MHz. Two signals are presented to the are modeled by iD = a1 vD + a2 vD 2 3
+ a3 vD .
mixer, one at 18.5 GHz with a power of 100 nW The LO voltage, at the LO terminal, is vLO =
and an interfering signal at 17.5 GHz with a A cos(ωLO t) and the RF voltage, at the RF termi-
power of 10 nW. If the image rejection is 20 dB nal, is vRF = B cos(ωRF t). Consider a 1:1 wind-
and the conversion loss is 10 dB. ing ratio. That is, the number of windings on the
(a) What is the signal power at the IF? secondary on each side of the center tap is equal
(b) What is the interference power at the IF? to the number of windings on the primary.
MIXER AND SOURCE MODULES 221

malized phase noise at 1 MHz offset, assuming

that the phase noise power varies as the square
of the inverse of frequency?
16. The phase noise of an oscillator was measured
as −125 dBc/Hz at 100 kHz offset. What is the
normalized phase noise at 1 MHz offset, assum-
ing that the phase noise power varies inversely
with frequency offset?

12. A diode double-balanced mixer has an LO at 17. When 0 V is applied to a VCO, the output fre-
100 GHz and has an input RF signal of 101 GHz. quency is 1 GHz. When the input to the VCO is
What will be the frequencies of the main signals 10 mV, the sinusoidal output of the VCO has a
at the IF? frequency of 1.01 GHz. What is the tuning gain
of the VCO?
13. The phase noise of an oscillator was measured as
−120 dBc/Hz at 10 kHz offset. What is the nor- 18. If a sinusoidal voltage is applied to the input of
malized phase noise at 1 MHz offset, assuming an analog VCO, describe the signal at the output
that the phase noise power varies as the inverse of the VCO.
of frequency? 19. Describe the design of a times-two frequency di-
14. The phase noise of an oscillator was measured as vider using a frequency multiplier based on a
−130 dBc/Hz at 10 kHz offset. What is the nor- diode and one or more bandpass filters. That is,
malized phase noise at 1 MHz offset, assuming sketch the circuit at the block diagram level.
that the phase noise power varies as the inverse 20. Describe the design of a times-three frequency
of frequency? divider using a frequency multiplier based on a
15. The phase noise of an oscillator was measured as diode and one or more bandpass filters. That is,
−125 dBc/Hz at 100 kHz offset. What is the nor- sketch the circuit at the block diagram level.

6.14.1 Exercises By Section

†
challenging, ‡ very challenging
§6.2 1, 2† , 3, 4, 5, 6, 7† , 8† , 9† , 10† §6.4 13† , 14† , 15† , 16 §6.9 19, 20
§6.3 11‡ , 12 §6.5 17, 18

6.14.2 Answers to Selected Exercises

3 −26 dBm 9 19.0 dB 17 1 GHz/V
5 0 dBm 10(b) 28 dB
7(b) 10 pW 12 One signal at 1 GHz
 CHAPTER 7

Cascade of Modules

7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223

7.2 Nonlinear Distortion of a Cascaded System . . . . . . . . . . . . . . . . . . . 225
7.3 Cascaded Module Design Using the Budget Method . . . . . . . . . . 231
7.4 Cascaded Module Design Using the Contribution Method . . . . . 232
7.5 Case Study: High Dynamic Range Down-Converter Design . . . 236
7.6 Case Study: Analysis of a 15 GHz Receiver . . . . . . . . . . . . . . . . . . . 238
7.7 Case Study: Frequency Planning of a Transceiver . . . . . . . . . . . . . 241
7.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
7.9 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
7.10 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248

7.1 Introduction
Design of a receiver or transmitter circuit requires the design of a cascade
of modules that achieves optimum dynamic range while managing DC
power consumption. As an example the frequency conversion stage of a
receiver is shown in Figure 7-1. The microstrip filters and transmission lines
are fabricated on alumina substrates. The semiconductor dies and the chip
decoupling capacitors as well as the alumina modules are epoxied to a
mat that was screen-printed on a brass housing. The mat provides a stress-
relieving (allowing for differences in thermal coefficients of expansion) and
conducting interface between the alumina and semiconductor substrates and
the brass housing. The modules and dies are interconnected by bond wires
arranged as two or more bond wires in parallel to reduce inductance. The
bond wires attach to the pads on the decoupling capacitors. These modules,
with the exception of filters, are described in this chapter.
The chapter begins with investigation of a cascade of modules given
the distortion characteristics of the individual modules. Coupled with an
earlier treatment of noise of a cascaded system this defines dynamic range.
A microwave systems comprising a cascade of mostly two-port modules
and must be designed to simultaneously minimize noise, distortion, DC
power consumption, spurious emissions, and maximize dynamic range.
There is necessarily a trade-off of these performance parameters and this
trade-off is at the heart of system design. What makes this particularly
challenging is that the two-port networks are designed separately and there
224 STEER MICROWAVE AND RF DESIGN: MODULES

Modules Modules
A Driver amplifier N Variable attenuator
B,H Bias decoupling capacitor Q Low noise amplifier
C ×2 frequency multiplier R Transmission line
D Edge-coupled bandpass filter S Edge-coupled bandpass filter
E Mitered bend T Subharmonic mixer
F Transmission line
G ×2 frequency multiplier Signal Path
I Lowpass filter e Frequency multiplied LO
J Transmission line h RF input
K Low noise amplifier i Reference LO input
L,M,O,P Bias decoupling capacitors j IF output

Figure 7-1: The frequency conversion portion of the 15 GHz receiver shown in Figure 7-12. The
RF input at h is centered at 15 GHz and is amplified and filtered before being presented to
athe subharmonic mixer at T. The reference LO at c is tunable from 1601 to 1742 MHz and is
frequency multiplied twice, at Cand G, and then presented as the LO to the harmonic mixer at
at T. The LO ranges from 6403 to 6968 MHz. The IF output at j ranges from 465 to 1595 MHz.

is incomplete characterization of performance. Indeed the only accurate way

of determining overall system performance would be to have the circuit-
level schematics of each module and then perform a whole system time-
domain circuit-level siulation using the actual modulated signals. Even if
all the details were available the simulation time would be prohibitive and
a time-domain simulation is unlikely to have the necessary fidelity. This is
true even if a large part of a system is implemented on the same chip. Thus
system design must use a methodical process using incomplete module-level
parameters such as the 1 dB gain compression, noise figure, and third-order
intercept characterizations. System design is often conservative designing for
worst case. Sometimes this is too conservative and allowances must be made.
CASCADE OF MODULES 225

Figure 7-2: Receiver as a

cascade of modules.

Testing and adjustment of fabricated prototypes is essential.

This section describes two approaches to cascaded module design; the
budget method and the contribution method. In the budget method gain and
noise performance metrics are initially assigned to each stage in a cascade
and then these are adjusted in a system simulator to iterate towards an
optimum solution. In the contribution method the focus is on dynamic range
and each critical module is assigned the same dynamic range and from this
an initial assignment of noise figure, gain, and required linearity is made for
each module. The optimization of system performance proceeds as with the
budget method. Generally this optimization must be performed manually
using experience as often it is difficult to quantify what an optimum is. Also
some design decisions require the acquisition of higher cost modules and if a
module is low cost then it makes sense to assign high dynamic range to such
a module to reduce the requirements on other higher cost modules. Perhaps a
module must be redesigned or designed in-house instead of being externally
sourced. These are not simple decisions. In the end, the system with cascaded
modules must be cost and performance competitive, and reducing time to
market is a competitive advantage.

7.2 Nonlinear Distortion in a Cascaded System

This section builds on the distortion analysis of two-port networks in Section
4.5 and addresses the determination of the nonlinear metrics of a cascaded
system such the receiver shown in Figure 7-2. The filters and interstage
matching networks eliminate harmonics from the system but, unfortunately,
allow in-band and close-in out-of-band distortion components to pass
through the system. The main metrics that describe nonlinear performance
are the power levels at the 1 dB gain compression point and at the third-order
intercept (IP3) point. In some systems, such as direct-conversion receivers,
the second-order intercept point (IP2) is also important. These metrics relate
to discrete tones and the correlated distortion generated in the different
stages. This contrasts with the calculation of noise of a cascaded system
where the noise added by stages is uncorrelated.

7.2.1 Gain Compression in a Cascaded System

When two amplifier stages are cascaded it is necessary to determine the 1 dB
gain compression point of the cascade. Consider the cascade in Figure 7-3.
The two stages have linear power gains G1 and G2 , and 1 dB compression
points P1,1dB and P2,1dB , respectively. The total linear power gain of the
system is GT = G1 · G2 . If the Taylor series expansion of the input-output
characteristics of the first stage in the cascade is

2 3
v1o (t) = ao + a1 v1i (t) + a2 v1i (t) + a3 v1i (t) + . . . , (7.1)
226 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 7-3: Cascade of two

stages used in determin-
ing the total 1 dB compres-
sion point of the system.

then at the interface of the two stages the amplitude of the tone at f1 due to
a single-tone input, vi (t) = Vi cos(ω1 t), is v1o (t) = V1o cos(ω1 t), where

V1o = a1 Vi + 43 a3 Vi3 . (7.2)

The harmonics are ignored, as design of the interstage matching network

would preferably be lowpass to eliminate the passage of harmonics. The
input of the second stage is v2i (t) = v1o (t), with the Taylor series expansion
of the second stage being
2 3
v2o (t) = bo + b1 v2i (t) + b2 v2i (t) + b3 v1i (t) + . . . . (7.3)

Thus the component of the final output at f1 will be

Vo = a1 b1 Vi + 43 b1 a3 Vi3 + 34 a1 b3 Vi3 + . . . . (7.4)

That is, the input signal is multiplied by the overall linear gain of the
amplifier; this is the first term on the right in Equation (7.4), and the last
two terms result in gain compression. If the stages are identical (i.e., a1 = b1
and a3 = b3 ) the two gain compression terms will be identical. Thus the two
stages will contribute equally to gain compression.
Note that in Equation (7.4) voltages add to yield the overall gain
compression. In general the two stages will not be identical, they will
have different gain and 1 dB compression points, however, this observation
(i.e., that the voltages add) does enable an approximate expression to be
developed for the 1 dB compression point of a cascaded system.
Examination of Equation (7.4) leads to a general formula for the total 1 dB
T
compression point, P1dB , of the two-stage cascade in Figure 7-3:
 T
− 1 1
−2
1
−2
P1dB 2 ≈ (G2 P1,1dB ) + (P2,1dB ) . (7.5)

Note that here the gains and the powers are absolute quantities and not
T
in decibels. For example, P1dB is the power in watts at the 1 dB gain
compression point.
Equation (7.5) is a very conservative estimate of the 1 dB gain compression
point of the amplifier and flags the lowest power levels at which gain
compression of the cascade could occur. The situation could be better
depending on the phasing of the distortion components but conservatism
is very important in system design. The worst case is what matters in
specifying system performance. Getting the effect of phasing right in system
design requires a circuit-level simulation but when working with modules
CASCADE OF MODULES 227

such information is rarely available. Equation (7.5) cannot be directly derived

from the voltage expansion as the power gain of microwave amplifiers
is due to difference of input and output impedance levels in addition to
voltage gain. It should be noted that a microwave amplifier deviates from
linearity following a sharper, tanh-like, response than that described by a
low-order polynomial model of linearity. Thus simply cascading low-order
polynomials describing each stage is not a viable option.
If required and if a more realistic estimate is required and can be justified,
then a less conservative formula that can be used to estimate the gain
compression level of two cascaded stages is
 T −1 −1 −1
P1dB ≈ (G2 P1,1dB ) + (P2,1dB ) . (7.6)

It is usually better to use the more conservative estimate of compression in

Equation (7.5).

EXAMPLE 7.1 Gain Compression of a Two-Stage Amplifier

The first stage of a two-stage amplifier in a transmitter has a gain G1 = 20 dB and an output
1 dB gain compression power P1,1dB = 10 dBm. The second stage has a gain G2 = 6 dB and
an output 1 dB gain compression power P2,1dB = 20 dBm.
(a) What is the linear gain of the two-stage amplifier?
(b) What is the gain of the two-stage amplifier at the 1 dB gain compression power?
(c) What is the 1 dB gain compression power of the cascaded system?
Solution:
(a) When the gain of an amplifier stage is given without qualification it should be assumed
to be the linear gain, that is, the gain at small signal levels. So the total linear power gain
of the two-stage amplifier is

GT = G1 G2 = 20 dB + 6 dB = 26 dB.

(b) The compressed gain will be 1 dB less, that is, GT = 25 dB.

T
(c) The overall gain compression power, P1dB is, approximately obtained using Equation
(7.5):
 − 1 1 1  − 1  − 1
T 2 2 2
P1dB = (G2 P1,1dB )− 2 + (P2,1dB )− 2 = 10(6/10) 10(10/10) + 10(20/10)
1 −2
 
1
T − −
P1dB = (3.981 · 10) 2 + (100) 2 mW = 14.97 mW = 11.8 dBm. (7.7)

A quick check is as follows.

If Stage 1 dominates gain compression, the output power at the 1 dB compression is
P1,1 dB = 10 dBm multiplied by the linear power gain of the second stage, that is,
T
P1dB = G2 P1,1 dB = 6 dB + 10 dBm = 16 dBm.
If Stage 2 dominates compression, the output power at the 1 dB compression of the two-
T
stage amplifier is just that of Stage 2: P1dB = 20 dBm.
228 STEER MICROWAVE AND RF DESIGN: MODULES

EXAMPLE 7.2 Second Example of Gain Compression of a Two-Stage Amplifier

The first stage of a two-stage amplifier in a transmitter has a gain G1 = 13 dB and an output
1 dB gain compression power P1,1dB = 10 dBm. The second stage gain is G2 = 10 dB and the
output 1 dB gain compression power is P2,1dB = 20 dBm. What is the 1 dB gain compression
power of the two-stage amplifier system?
Solution:
T
The total gain compression power, P1dB , is approximately obtained using Equation (7.5):

 − 1 1 1  10 10 − 1  20 − 1 1
T 2 2 2
P1dB = (G2 P1,1dB )− 2 + (P2,1dB )− 2 = 10 10 10 10 + 10 10 (mW)− 2
1 −2
 
1
T
P1dB = (100)− 2 + (100)− 2 mW = 25.00 mW = 13.98 dBm. (7.8)

A quick check is as follows.

If Stage 1 dominates compression, the output power at 1 dB compression of the two-
stage amplifier is P1,1dB multiplied by the linear power gain of the second stage, taht is
T
P1dB = G2 P1,1 dB = 10 dBm + 10 dB = 20 dBm.
T
If Stage 2 dominates compression, P1dB = P2,1dB = 20 dBm. Thus neither stage dominates
compression.

A final note is that the calculation of the 1 dB compression point when

both stages in a two-stage system contribute equally to gain compression is
approximate as the actual compression characteristic is complex and a third-
order Taylor series does not capture the total nonlinear response [1, 2].
The response of a multistage system can be extrapolated from the treat-
ment here for two stages. For an m-stage cascade,
 T − 1 −
1  − 1 −
1
P1dB 2 = (Gm . . . G2 P1,1dB ) 2 + . . . + G2 P(m−1),1dB 2 + (Pm,1dB ) 2 .
(7.9)

7.2.2 Intermodulation Distortion in a Cascaded System

The two-stage system shown in Figure 7-4 will be used here to determine
the total intermodulation response as described by the third-order input
and output intercept points, IIP3 and OIP3, respectively. The development
is based on the analysis of intermodulation distortion in Section 4.5.3 and
is called the cascade intercept method. One version of the method is called
the organized cascade intercept method, in which the worst-case situation

Figure 7-4: Cascade of two

stages used in determin-
ing the third-order intercept
point of a cascaded system.
CASCADE OF MODULES 229

Figure 7-5: Signal flow

for calculating the in-
termodulation distor-
tion in the cascade of
two stages.

is assumed, in that the IM3 produces from each stage in the cascade add
in phase. The other method is the unorganized cascade intercept method in
which the phases of IM3 produce by each stage are unknown, and assumes
that the most likely overall IM3 distortion will be obtained by assuming a
random phase relationship of the IM3 contributions.
The third-order intercept point, IP3, is determined by extrapolating the
small signal gain and IM3 responses. Since IM3 is small, the signal• flow
in the two-stage system is as shown in Figure 7-5. One path in the flow
linearly amplifies the input two-tone signal in Stages 1 and 2. The third-
order nonlinearity of Stage 1 creates low-level IM3 that is linearly amplified
by the second stage. A second set of IM3 signals is produced by the IM3 of
the second stage operating on the two-tone signal amplified by Stage 1. The
voltages of the two IM3 signals add constructively. A complication is that the
phases of the two IM3 signals, at x and y, are not known. This leads to two
approaches to estimating of the overall distortion.

Organized Cascade Intercept Method

The worst-case situation is that the IM3 voltages at x and y are in phase so
that the voltages add. Then the total IM3 distortion is obtained by adding
the IM3 voltages generated from each stage. Determining total OIP3, OIP3T
is the inverse problem. This cannot be solved precisely, but a good estimate
for OIP3T is obtained from [3, 4]
− 21 1 1
OIP3T ≈ (G2 OIP31 )− 2 + (OIP32 )− 2 .

(7.10)

Note that OIP3 is a power and G is a power gain. Now the total gain of the
cascade GT = G1 G2 , and since OIP3T = GT IIP3T , where OIP3T , the input
intercept of the cascade, can be written
− 1 −
1
−
1
IIP3T 2 ≈ (IIP31 ) 2 + (IIP32 /G1 ) 2 .

(7.11)

Generalizing these results, for an m-stage cascade the overall OIP3 is

− 21 1
−2 − 1
OIP3T
 
≈ (Gm . . . G2 OIP31 ) + . . . Gm OIP3(m−1) 2
1
−2
+ (OIP3m ) . (7.12)
230 STEER MICROWAVE AND RF DESIGN: MODULES

Since IIP3 = OIP3/G, the general result for multiple cascaded stages can be
written
1 1 1 1
G(m−2) . . . G1 G(m−1) . . . G1
   
1 2 1 2 2 2
≈ + ... + .
IIP3T IIP31 IIP3(m−1) IIP3m
(7.13)

The organized cascade intercept method often provides an overly

conservative (i.e., far too low) estimate of OIP3T and IIP3T distortion.

Unorganized Cascade Intercept Method

If the phases of the stages are random (or perhaps unknown), then
it is reasonable to add the powers of the distortion terms. This is an
approximation as the IM3 signals are still correlated, but this approach has
been found to provide a useful measure in design, then
−1
OIP3T ≈ (G2 OIP31 )−1 + (OIP32 )−1 ,

(7.14)

and again OIP3 is a power and G is a power gain. For an m-stage cascade
with random IM3 phase the total OIP3, OIP3T , is obtained from
−1 −1 −1 −1
OIP3T
 
≈ (Gm . . . G2 OIP31 ) + . . . + G2 OIP3(m−1) + (OIP3m )
(7.15)

and the total IIP3, IIP3T , is obtained from (since IIP3 = OIP3/G)

G(m−1) . . . G1
       
1 1 G1
≈ + + ···+ . (7.16)
IIP3T IIP31 IIP32 IIP3m

Summary
The cascade intercept method has led to two sets of results for IM3
distortion. The first, from the organized cascade intercept method, is the
worst-case situation in which the IM3 distortion of each stage combines
in the worst possible way. This yielded the overall IIP3 and OIP3 results
of Equations (7.12) and (7.13). The second, from the unorganized cascade
intercept method, assumes that the phases of the IM3 from each stage are
randomly related, perhaps the best estimate that can be made without a
circuit simulation. This yielded the overall IIP3 and OIP3 results of Equations
(7.15) and (7.16).
It is interesting to speculate what would happen if the stages were
designed so that the IM3 contributions at x and y (referring to Figure 7-
5) of the first and second stages were 180◦ out of phase but with the same
magnitude. If this could be done the IM3 contributions at the output would
be canceled. Indeed, it is possible to do this to a limited extent. The phase of
the IM3 signals depends on the signal level and so changes over the signal
range. Careful design, and only when there is complete control over the
design and integration of the stages, enables the IMD contributions to be
partially canceled over a range of signal levels, as shown in Figure 7-6. This
extends the dynamic range of the cascaded system, and the simple OIP3 and
IIP3 metrics are not sufficient to capture this complexity.
CASCADE OF MODULES 231

Figure 7-6: The gain and IM3 responses of two cascaded

stages showing the partial cancellation of the IM3 response
over a range of input signal levels.

EXAMPLE 7.3 Intermodulation Distortion of a Two-Stage Amplifier

The first stage of a two-stage amplifier in a transmitter has a gain of G1 = 20 dB and

an output third-order intercept point, OIP31 , of 30 dBm. The second stage has a gain of
G2 = 20 dB and an OIP3 of 40 dBm. Assume that the IMD contributions of each stage are in
phase. What is the OIP3 of the cascaded system?
Solution:
Using the organized cascade intercept method, the total OIP3, OIP3T , in milliwatts is
obtained using Equation (7.10)
 − 1  − 1  − 1  − 1  − 1
2 2 2
OIP3T ≈ 10(20/10) 10(30/10) + 10(40/10) = 105 2 + 104 2 . (7.17)
T
Thus OIP3 = 5772 mW = 37.6 dBm. (7.18)

This is the worst-case situation. It is worth comparing this to a calculation where the phases
of the IMD contributions are unknown. Then, using Equation (7.14),
 −1  −1  −1
OIP3T ≈ 10(20/10) 10(30/10) + 10(40/10) = 105 + 104 . (7.19)
T
Thus OIP3 = 9091 mW = 39.6 dBm. (7.20)

7.3 Cascaded Module Design Using the Budget Method

The budget method is used to design RF cascade systems, particularly the
design of receiver and transmitter cascaded systems, for specified gain,
noise, and distortion. In the budget method, initial assignments of gain and
noise performance are made to each stage in a cascade. This approach is
based on the calculation of the total noise figure and gain of a system from
the parameters of individual subsystem stages (see Section 4.3.1). This is
coupled with the calculation of nonlinear distortion in cascaded stages (in
Section 7.2), to calculate the spurious free dynamic range (SFDR) and make
assignments for noise and distortion contributions of each stage [5]. The
SFDR is related to the gain compression, noise, and nonlinear distortion
metrics in Figure 7-7. The system optimization goal is, in general, to choose
and adjust modules to maximize overall SFDR subject to the constraints of
cost and availability of modules with the required attributes. For example,
the gain of most amplifier modules can be adjusted and so change the
232 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 7-7: Output power versus input power

of a stage or system. Extrapolations of the
1:1 linear response and the 3:1 third-order
intermodulation response intersect at the IP3
point. DR = dynamic range, SFDR = spurious
free dynamic range, SNR = required minimum
signal-to-noise ratio.

Figure 7-8: Cascade system for receive down converter or transmit up converter.

nonlinear distortion performance of amplifier stages. In the budget method

the noise and nonlinear distortion metrics of each stage are set based on
experience.
Considering the stages in the cascade shown in Figure 7-8, generally
maximum gain is assigned to the first active stage, and so the first stage
contributes the most to the system noise figure. However, as a consequence
active devices in latter stages nearly always require significantly higher
linearity performance than may be necessary to meet the system SFDR
objective. Still the budget method is a systematic way to design a cascaded
system. Experience can also be used in the budget assignments. For example,
in transmitter cascade design the intent might be to provide as high an
intercept as possible in the lower-frequency up-conversion stages. This
approach is based on devices operating at lower frequencies having very
good linearity and therefore being most cost effective in achieving high IP3.

7.4 Cascaded Module Design Using the Contribution

Method
The contribution method is an approach to the design of cascade RF systems
for maximum SFDR [6] rather than separate treatment of noise and nonlinear
CASCADE OF MODULES 233

distortion. The contribution method provides a good initial assignment of

the noise figure, gain, and required linearity to individual stages and enables
informed assessment of trade-offs during system design.
In the budget method, noise performance is largely determined by the
first stage and the nonlinear performance by latter stages. This often results
in suboptimum system design. One of the costs is often the use of more
expensive stages than necessary or higher linearity requirements of some
stages leading to increased supply power.
The contribution method assigns to each active stage a percentage
contribution to the overall SFDR. Contributions of the stages are balanced as
no single stage dominates noise or distortion, and all contributions to SFDR
are equally weighted. In realizing a cascaded system, typically one or more
offending stages either dominate the SNR or significantly degrade overall
distortion. In some cases a redistribution of stage parameters (such as gain)
reduces the degradation. In other cases a change in architecture is required.
The mathematics behind the contribution approach is based on an
extension of the noise and distortion analyses of cascaded stages. The
noise performance of a stage is characterized by its noise factor F =
No /(GNi ), where G is the available power gain of the stage, No is the
output noise power, and Ni is the input noise power of a resistor held at
standard temperature (290 K). Distortion is characterized by the amount
of intermodulation distortion (IMD) produced in a two-tone test and
specifically the third-order intercept point (IP3) (see Figure 7-7). In the
following development a passive linear stage such as a filter or attenuator
is combined with the following active stage.

7.4.1 Noise Contribution

Friis’s formula yields the noise factor of a cascaded system and was derived
in Section 4.3.1. This section presents an alternative development of Friis’s
formula with interim steps that can be used in calculating the dynamic range
contributions of each stage. The analysis here was originally presented in [6].
First consider the contribution of individual stages to system noise, and
assume that all stages are matched. The nth stage, with all ports having equal
bandwidth B, has output noise power (from Equation (4.24)
Nno = Fn Gn kT0 B, (7.21)
where k is the Boltzmann constant, and Fn and Gn are the noise factor and
available power gain of the nth stage, respectively. The excess noise power of
successive stages is additional to that of the first. So, while the output noise
power contribution of the first stage is
N1o = F1 G1 kT0 B, (7.22)
the excess output noise power of the second stage is
N2o = (F2 − 1)kT0 BG2 . (7.23)
Note that the output noise of the first stage includes the noise contribution
of the source resistance held at the standard temperature. However, this is
not included in the noise contributed by the second or latter stages. The total
noise power at the output of a two-stage cascade is
T
N2o = (F2 − 1)kT0 BG2 + N1o G2 . (7.24)
234 STEER MICROWAVE AND RF DESIGN: MODULES

Then the total noise power at the output of the mth stage is
m n m
 
  
T
Nmo = (Fn − 1)kT0 B Gi + F1 kT0 B Gn . (7.25)
n=2 i=2 n=1

Thus an m-stage cascade has total cascaded system noise factor F T =
Nmo / G N1i , with GT being the total cascaded available gain and N1i the
T T


noise power input to the first stage. In terms of the parameters of individual
stages
m
T
 Fn − 1
F = F1 + m
 . (7.26)
n=2 Gi−1
i=n

The link between Equations (7.25) and (7.26) enables the noise contribution
of each stage to be determined. Treated separately, Equation (7.25) provides
the output noise power of a cascade such as a receiver. Although the
extension of Equation (7.25) to Equation (7.26) is normally associated with
the derivation of the receiver noise factor, it can also be used in transmitter
noise analysis.
Now the noise contribution of a stage can be defined. The gain
accumulated at the jth stage (the total cascade gain up to and including the
jth stage) is
j

GA
j = Gn (7.27)
n=1

(and so GT = GA m for an m-stage cascade). The fractional noise contribution

of the jth stage to the total output noise is then defined as
 
N Fj − 1
Cj = stage noise contribution ≡  A  . (7.28)
Gj /Gj F T

This is one component of the SFDR contribution of a stage.

7.4.2 Intermodulation Contribution

The intermodulation distortion of a cascade is assessed using the cascade
intercept method (see Section 7.2.2). The usual design approach of
establishing the SNR in the early stages of the cascade (this is the budget
method) results in stages further along the cascade needing to have higher
intercept values. Consequently these stages will generally have additional
power consumption, as this is usually required to increase IIIP3 (or OIP3).
Trade-offs of the contributions of individual stages to distortion and noise
will lead to constrained total power consumption while still achieving the
required system SFDR.
For receiver systems the contribution to nonlinear distortion is captured by
IIP3 and for all stages this will be referred to the input of the cascaded system.
Keeping with the nomenclature of the previous section, the accumulated IP3
of the jth stage referred to the system input is

IIP3A A
j = IIP3j /Gj−1 (7.29)
CASCADE OF MODULES 235

and is IIP3AdBm,j when expressed in dBm. Note that all prior gain and loss
up to the jth stage modifies the intercept when it is referred to the input of
the cascaded stages. Combining the contribution of individual stages using
Equation (7.16) yields the total system IIP3:
 m
−1
 1
IIP3T = . (7.30)
n=1
IIP3A
n

This leads to the definition of the fractional contribution of the jth stage to
the system IIP3:

IIP3T
CjIIP3 = IIP3 contribution ≡ . (7.31)
IIP3A
j

This is the second and final component of the SFDR contribution of a stage.

7.4.3 Design Methodology for Maximizing Dynamic Range

In this section a design methodology is developed for trading off the
performance of each stage in maximizing the SFDR of the RF down-
converter module shown in Figure 7-8. A central component of the
methodology is the use of a stage contribution graph that indicates
performance of each stage. This graph enables visualization of the
contribution of each stage to system dynamic range, identifying which stage
or stages dominate performance. For example, the initial contributions for
the example used in the case study that follows are shown in Table 7-1.
The noise and distortion contributions indicate the stage that tends to have
the most impact on noise factor or distortion. Redistribution of gain, noise,
or distortion alters this relationship and leads to changes in system SFDR.
Passive stages generally do not introduce distortion, but they do contribute
noise. So, to simplify the following discussion, passive stages immediately
preceding an active stage and the active stage itself will be considered as a
single stage.
First of all, consider the traditional budget method approach in which the
noise figure of a receiver cascade is established by maximizing the gain of
the first stage, and thus the first stage establishes the SNR of the system since
the noise contributions of subsequent stages are assumed to be negligible.
With the system SNR fixed, a target system SFDR determines the overall
IP3 performance described by the total system IIP3, IIP3T . Also, in the
budget method, the first stage has negligible impact on overall IIP3. Then a
reasonable choice in the design process is to select stages following the first as
contributing equally to the reduction in IIP3T . Thus, in the budget method,
the minimum acceptable IIP3 (in dBm) of the jth stage (i.e., IIP3dBm,j ) in
an m-stage cascade that is required to meet the target total IIP3 in dBm

Stage (i) Gain (dB) Gain Gi NFi (dB) Fi IIP3i (dBm) Table 7-1: Stage assign-
1 filter −3.98 0.4 3.98 2.5 – ments based on balanced
2 LNA 1.76 1.5 3.01 2.0 -0.969 contributions to the SFDR
3 filter −3.98 0.4 3.98 2.5 – of the cascade. Initial
4 mixer – – 2.04 1.6 -3.18 assignment, i = 1.
236 STEER MICROWAVE AND RF DESIGN: MODULES

T
(IIP3dBm ) is obtained from Equations (7.29) and (7.30) as

IIP3dBm,j = IIP3A A
dBm,j + GdB,(j−1) , (7.32)
T
where IIP3A
dBm,j = IIP3dBm + 10 log m. (7.33)
T
(Note that the overline in IIP3dBm identifies the target IIP3T dBm and not the
actual IIP3.) In addition, the gain or loss of preceding stages will modify this
IIP3 as indicated in Equation (7.29). That is, with this assignment each stage
makes an equal contribution to the overall intermodulation distortion. That
is, each stage has the same IP3 referred to the system input. For example, in
a cascade of three stages, each stage would need to have a minimum system
T
IIP3 of 4.8 dB (= 10 log 3) in excess of the target system IIP3 (i.e., IIP3dBm ).
If instead, using the contribution method, the cascade system was
designed so that each stage contributed equally to the overall noise (CjN
being the same for all stages), and each stage had the same gain (Gj being
the same for all stages), then the required individual IIP3 values tend to
be minimum. Maximizing system dynamic range becomes an exercise in
maintaining the lowest noise power and highest IIP3 value throughout the
cascade. The link between these parameters values is the distribution of gain
and loss [7, 8].
For a transmitter, the SFDR is most commonly referred to the output and
(repeating Equation (4.88))
2
SFDRdB,o = 3 (OIP3dBm − NdBm,o ) , (7.34)

and SFDRi = SFDRo . In the above, Ni and No are the total input and output
noise powers and are assigned a noise floor value dependent on the cascade
noise factor target and linear system gain.

7.4.4 Summary
The contribution method for designing systems of cascaded modules focuses
on assigning the same dynamic range to each module. Of course linear
modules such as a filter easily meet any system dynamic range requirement
so the dynamic range assignments should focus on the nonlinear modules.
Very often the dynamic range of an active module can be increased by either
increasing the biasing of the module or switching to an alternative module
but with high power requirements. Sometimes modules with different
technology could provide increased dynamic range without increasing
power consumption but such modules could be more expensive. The
contribution method and the previously considered budget method only
provide an initial starting point for system design. The system must still
be optimized and often manually optimized as there are too many hard to
quantify design goals. For example minimizing time-to-market and design
cost are goals that cannot be parameterized.

7.5 Case Study: High Dynamic Range Down-Converter

Design
In this section a down-converter in a receiver is designed. With the
architecture and modules chosen in the RF cascade, the only quantities
CASCADE OF MODULES 237

that can be selected by the user are the gains of the amplifier stages. If
adjusting the gains is not sufficient to meet system objectives, then it would
be necessary to choose other modules or change the architecture.

7.5.1 Architecture
A down-converter with the architecture of Figure 7-8 is designed here
considering the first four stages only and using the balanced contribution
method for the initial stage assignment. The system operates at 1.5 GHz with
T
a first IF of 250 MHz and an IIP3 design target of 0 dBm (i.e., IIP3dBm =
T
0 dBm) with a noise figure target of 10 dB (i.e., NF = 10 dB). The initial
stage assignments for balanced noise and IMD contributions are shown in
Table 7-1.
The stage assignments shown in Table 7-1 meet the target specifications
exactly, with no single stage contributing any more than the required
noise factor or nonlinearity (distortion) to meet the target. The LNA gain
is low, permitting a significant reduction in required mixer IP3. In an
RF system usually the mixer has the limiting distortion performance so
anything that can be done to reduce the linearity required for a mixer
is advantageous. The noise figure of a mixer can also be high. With this
in mind, the process of system cascade trade-off and selection of actual
element parameters continues in the following sections. That is, assignment
of balanced contributions to each stage is a good initial starting point. This is
followed up with further optimization and trade-offs.

7.5.2 Design
Design proceeds with a selection of available modules for stages and
continuous but limited variation of stage parameters (through bias control,
for example). The following stages were chosen for the stages (see Figure 7-
8): an MMIC amplifier for Stage 2 (NEC part UPC2745) and a MMIC mixer
for Stage 4 (MCL part SYM-2500). Bandpass dielectric resonator filters were
chosen for Stages 1 and 3. In particular, a tuned lumped-element Chebychev
bandpass filter with 0.1 dB ripple at 1.5 GHz was used to provide settable
loss in Stage 3 and this proved to be important in establishing balanced
stage contributions and improved dynamic range. The continuous control
parameters include controlling the mixer IMD contribution by changing the
LNA gain (through bias control), changing the mixer IIP3 (by varying the
LO drive level), and retuning the Stage 3 filter. The noise and distortion
contributions of stages with the initial stage assignment (from Table 7-
1), initial measured performance, and optimized design are shown in
Figures 7-9 and 7-10. Figure 7-11 depicts the dynamic range. Based on the
balanced contribution method, the SFDR is 109.5 dB normalized to a 1 Hz
bandwidth. The initial value provided by the devices selected is 108 dB.
After appropriate trade-off and adjustable loss in Stage 3, the dynamic range
achieved is 111 dB. Now the same system design could be achieved using
the budget assignment initially, but with much greater design effort.
238 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 7-9: Noise figure contribu-

tion: (a) initial calculation of bal-
anced noise figure contribution;
(b) measured contributions of each
stage using the initial assignments;
and (c) final measured contribu-
tions after altering bias, filter loss,
and mixer LO drive levels for im-
proved dynamic range.

Figure 7-10: Input intercept (IP3) contri-

bution: (a) initial calculation; (b) original
measured contributions; and (c) final per-
formance after alteration of bias, LO drive,
and filter loss in Stage 3. Final settings
in (c) led to the highest system dynamic
range.

Figure 7-11: SFDR by stage: (a) initial as-

signment; (b) initial system performance
after selection of modules but before op-
timization; and (c) final performance af-
ter setting bias, LO drive, and filter loss in
Stage 3.

7.6 Case Study: Analysis of a 15 GHz Receiver

Most RF and microwave systems convert information at one frequency
to information at another frequency that can either be more conveniently
processed in the case of a receiver or more conveniently radiated in the case
of a transmitter. Figure 7-12 is a 15 GHz receiver module which itself consists
of interconnected sub modules. Detail of the frequency conversion section is
given in Figure 7-13.
These sub modules could be called modules as well so that modules are
constructed by interconnecting other modules many of which are available
”off-the-shelf” from high-level modules such as a receiver. The receiver
module is part of a wireless service used in point-to-point microwave
link such as in a cellular systems to provide communication between
CASCADE OF MODULES 239

Modules Signal Path

A SMA connector, reference LO in a reference LO signal
B Attenuator b Reference LO
C Coupling capacitor c Reference LO
D Reference LO Amplifier d LO with bandpass filter
E Bias line e LO with lowpass filter
F Radial stub f Radio frequency (RF) input signal (∼15 GHz)
G RF coupling capacitor (DC block) g RF in
H Die attach mat h RF
I Frequency multiplier i RF path with bandpass filter
J Frequency multiplier j IF path with resistive antenuator
K Waveguide-to-microstrip adaptor k IF path
L Isolator l Intermediate frequency (IF) output
M variable RF amplifier block
N Mixer
O IF amplifier
P Voltage variable attenuator
Q IF out, SMC connector

Figure 7-12: A 14.4–15.35 GHz receiver module itself consisting of cascaded modules
interconnected by microstrip transmission lines. Surrounding the microwave circuit are DC
conditioning and control circuitry. Detail of the frequency conversion section mounted on the
mat is shown in Figure 7-13.
240 STEER MICROWAVE AND RF DESIGN: MODULES

A Reference LO amplifier K RF amplifier

B Power supply decoupling capacitor L Power supply decoupling capacitor
C ×2 frequency multiplier M Power supply decoupling capacitor
D Edge-coupled PCL bandpass filter N Voltage variable attenuator
E Microstrip bend O Power supply decoupling capacitor
F Microstrip transmission line P Power supply decoupling capacitor
G ×2 frequency multiplier Q RF amplifier
H Power supply decoupling capacitor R Microstrip transmission line and bend
I stepped impedance lowpass filter S Edge-coupled PCL bandpass filter
J Microstrip transmission line T Subharmonic mixer

Figure 7-13: Frequency conversion section of the receiver module shown in Figure 7-12. The
reference LO is applied to the frequency conversion section at c, the RF is applied at h following
the isolator. The IF is output at j.

basestations. The subsystem modules such as the amplifiers, frequency

multipliers, mixers, isolator, and waveguide adapter are available as off-
the-shelf components from companies that specialize in developing such
modules and selling them to a large user base. The block diagram of the
receiver module is shown in Figure 7-14.
The RF module engineer must interconnect these modules and mange
interference and dynamic range. The RF module designer either designs
the various filters or specifies them for design and fabrication by specialty
microwave filter companies. The conductive die attach mat H is material silk-
screened on to the brass back-plate and enables the alumina interconnects,
and the amplifier and mixer die, to be epoxy-attached. It is not wise to
directly attach die and ceramic substrates directly to the brass housing as
the difference in thermal coefficients of expansion and the rigidity of the
two systems means that the attachment could fail or the ceramic and die
crack. The mat is said to be ‘compliant’ providing stress relief when the solid
CASCADE OF MODULES 241

Figure 7-14: Block diagram of the receiver module shown in Figure 7-12.

objects expand at different rates. The ceramic substrates, usually alumina,

have repeatable and stable dimensions for realization of filters which have a
response that strongly depends on dimensions.

7.7 Case Study: Frequency Planning of a Transceiver

This case study develops the architecture and frequency plan of a transceiver
for the licensed fixed wireless service in the 15 GHz band that provides
channels supporting up to 140 Mbit/s. This is the service that provides
two-way point-to-point communications most commonly to connect pairs
of cellular basestations to each other. A central basestation in a cluster of
basestations then connects traffic into a backbone fiber optic-based network.
Frequency planning begins with the specifications provided by national
regulatory authorities which generally adopt recommendations from the
International Telecommunications Union. (See [9] for the recommendations
concerning the 15 GHz fixed wireless service.)
The 15 GHz fixed wireless band extends from 14.4 GHz to 15.35 GHz
supporting a number of channels with various channel spacing. The spacing
used will be based on local regulations and licensing. A universal transceiver
must support channel spacings of 2.5, 3.5, 7, 14, 28, and 56 MHz [9]. Figure
7-15(a) shows the channel assignment where fCH is the channel spacing. No
242 STEER MICROWAVE AND RF DESIGN: MODULES

(a) Fixed service channels

(b) Transmit/receive channel assignment

Figure 7-15: Channel assignment in the 15 GHz fixed wireless service band.
.
guard bands between the channels are specified but typically a few 10’s of
kilohertz are sufficient. Thus channel spacing is only slightly larger than
the modulation bandwidth and so there is a narrow guard band. Such a
relatively tight guard band is possible as significant digital processing power
can be used to resolve adjacent channels. At the low and high ends of the
15 GHz band there is a guard band that is about half of a channel spacing so
that other wireless services are not affected.
A transceiver in the fixed wireless service supports one or more transmit
and receive channel pairs. As specification in the ITU regulations the
separation of transmit and receive channel pairings can be 315, 322, 420, 490,
640, 644, 728, 840 MHz. That is, if a receive (or transmit) channel is centered
at 14.500 GHz, then the paired transmit(or receive) channel is centered at
14.815, 14.822, 14.920, 14.950, 15.140, 15.144, 15.228, or 15.340 MHz. Since a
two-way link is established, the center frequency of the received channel at
one end of the link is the center frequency of the transmitted channel at the
other end of the link.

7.7.1 Transceiver Architecture

A transceiver must simultaneously transmit and receive channels and
usually through the same antenna. Even if separate transmit and receive
antennas are used, there will be significant coupling of the transmitted and
received signal if they are located near each other. In practice transceivers use
a diplex filter, a diplexer, to separate the 15 GHz band into two sub bands of
equal bandwidth. In principal a diplex filter can be two filters combined to
provide a lowpass function for the upper sub band and a highpass filter for
the lower sub band. For fixed wireless service operation the diplex filter is
designed as two adjacent bandpass filters, one passing the lower sub band
and the other passing the upper sub band. Although the filter skirts can be
designed to be very steep, they necessarily have finite frequency extent so
CASCADE OF MODULES 243

Figure 7-16:
Transceiver
architecture.

that the middle of the channel is no longer available to this transceiver.

The high-level architecture of a transceiver for the 15 GHz fixed wireless
service is shown in Figure 7-16. This architecture provides high isolation of
the transmit and receive paths and supports a modular design approach. The
transmit section comprises the three modules shown on the top of the figure.
The first module is the transmit (Tx) digital baseband module that prepares
the digitally modulated signal for one (or more) transmit channels. Either
Reed-Solomon or trellis coded modulation (TCM) forward error correction
schemes are used. Five modulation formats are typically supported: QPSK,
QAM16, QAM32, QAM64 or QAM128 with the modulation scheme adjusted
to maintain a maximum bit error rate of 10−6 . The signal in digital form
is based in the transmit digital baseband module. The transmit analog
baseband module modulates the signal on a carrier at an intermediate
frequency of around 1 GHz. This is then passed to the transmit RF
module which generates the RF signal in the 15 GHz band. Typically
the transmitted power is 26 dBm using QPSK modulation, 23 dBm using
QAM16, 22 dBm using QAM32, 20 dBM using QAM64, and 19 dBm
using QAM128 modulation. The signal then passes into one port of the
diplexer, here a waveguide diplexer, before passing to the antenna. The
receiver sensitivity is typically −93 dBm for QPSK, −86.5 dBm for QAM16,
−75 dBm for QAM32, −73.5 dBm for QAM64 modulation, and −68.5 dBm
for QAM128 modulation, again for a BER of better than 10−6 .
The waveguide diplexer is shown again in Figure 7-17. High performance
diplexers can be designed in waveguide as there is very little loss and the Qs
of the filter resonators are high [10, 11]. Here the waveguide diplexer has a
bandpass response that passes the lower sub band, and a second bandpass
response that passes the upper sub band. The sub band responses are shown
in Figure 7-15(b). The two bandpass responses are designed together so
that the diplexer performance is better than if two independently designed
bandpass filters were used. Either sub band output of the diplexer could be
244 STEER MICROWAVE AND RF DESIGN: MODULES

Figure 7-17: Waveguide diplexer with two bandpass

filters inside the waveguide sections. One allows
the lower RF sub band to pass and the other lets
the upper sub band pass. The waveguide filter
transition can be very sharp so that the edges of
the upper and lower sub bands can be very close
eliminating only one channel.

connected to the transmitter (and the other to the receiver) depending on

whether the channel transmitted is in the lower or upper sub band.
The bottom three modules shown in Figure 7-16 comprise the receiver.
The RF signal enters the Receiver (Rx) RF module at RFIN where it is
down converted to an intermediate frequency signal around 1 GHz. The
intermediate signal passes to the Receive Analog Baseband module where
the desired channel (or perhaps more than one) is extracted and presented to
the receive digital baseband module at a frequency less than 100 MHz.
A common LO source is generated in the Reference LO module with one
output connected by cables to the Receive RF module and another version
of the LO, shifted by the appropriate receive/transmit frequency spacing,
routed to the Transmit RF module.
The transmit and receiver RF and analog baseband modules must
be mounted on the communication tower but the digital baseband and
reference LO modules can be located in an air conditioned room. Then
coaxial cables with signals up to 2 GHz route the RF signals and the reference
LO signals between the ground unit and the tower-mounted unit.

7.7.2 Frequency Planning

Basic frequency planning began with definition of the basic transceiver
architecture with the choice of an intermediate frequency after first
conversion (or before the final conversion in the case of the transmitter) that
is around 1 GHz. This choice was made as the transceiver must be able to
transmit across the entire 950 MHz bandwidth of the 15 GHz band. Once the
orientation of the diplexer has been fixed before delivery to the customer,
or in field installation, the transmit band is from one sub band to the other
sub band, a narrower but still significant 475 MHz bandwidth. However
the unit must still be designed to support operation over the whole band.
Thus, for the receiver, the design decision has been made to down convert an
entire sub band to a frequency range centered around 1 GHz. Such a circuit
must have essentially a lowpass response but with signals at DC and up
to a hundred megahertz or so above blocked by inductors and capacitors
providing DC connections and/or RF isolation. The fractional bandwidth
is too much for the design of a bandpass filter. The frequency conversion
description is similar for the transmitter.
Frequency planning has a major impact on the design cost of the
transceiver and in the unit cost, especially if extensive tuning is required
during manufacture. Frequency planning requires considerable experience
with knowledge of design cost drivers, manufacturability, which modules
and sub modules are available from vendors, and what one’s own
organization can do to provide a competitive advantage. To be competitive
CASCADE OF MODULES 245

(a) Basic signal flow in the RF receiver

(b) Extended signal flow in the RF receiver

Figure 7-18: Signal paths in the receive RF module in Figure 7-16.

the transceiver must be tunable across the entire 15 GHz band.

Cost is substantially reduced by sharing an LO between the transmitter
and receiver, with either the transmit or receiver versions of the LO shifted
by the appropriate transmit/receive spacing. The LO will be locked to a
low frequency reference using a phase-locked loop and then it must be
routed by coaxial cable to the receive and transmit RF modules. This means
that LO should ideally be below 2 GHz. This LO will now be called the
reference LO as it is not the ultimate frequency that is required by the mixer
in the frequency conversion section of the receiver RF module. Ultimately
the effective LO at the mixer must be tunable across the entire 15 GHz band
and so needs to tune at least by 950 MHz. Tuning can be accomplished using
phase-locked loops but realistically the tuning range that can be conveniently
achieved is 200 MHz or 15%, whichever is less up to 2 GHz.
The basic signal flow in the receiver RF module is shown in Figure 7-
18(a). The basic concept is that the received RF signal is mixed with an LO to
produce a low frequency version as the intermediate frequency signal, the IF.
In the signal flow, a typically low-level RF signal is presented to the receiver
at the port indicated by RFIN . The RF signal is amplified by a low noise
amplifier so that circuit noise in the following RF signal path has no impact.
Since the level of the received signal is not known, the amplitude of the signal
must be leveled. The reference LO is at a frequency below that required to
pump the mixer so that it can be more easily generated and routed in the
transceiver. Generally it would be preferable to keep the signals to be routed
below 2 GHz as then the quality of the cabling and connections is less critical.
As a result the frequency of the reference LO signal must be increased before
it is presented to the mixer.
Knowledge of which sub modules are readily available is also critical in
frequency planning. The use of frequency multipliers, essentially nonlinear
harmonic generators followed by filters, can increase the frequency by a
factor of 8 or so quite reliably. With frequency multiplication there is an
increase in the noise levels of the oscillator that impacts performance and
246 STEER MICROWAVE AND RF DESIGN: MODULES

Receive frequency plan Transmit frequency plan

A IF band A IF band
B Rx reference LO H Reference LO
C 2nd harmonic of Rx reference LO I 2nd harmonic of Tx reference LO
D 4th harmonic of Rx reference LO J 4th harmonic of Tx reference LO
E 8th harmonic of Rx reference LO K 8th harmonic of Rx reference LO
F RF band F RF band
G Lower RF sub band L Upper RF sub band

Figure 7-19: Transceiver frequency plan. The receive and transmit frequency plans are
exchanged if the RF received is in the upper RF band (and then the RF transmitted is in the
lower RF sub band. Also shown is the IF circuit bandwidth extending from 70 MHz to 2 GHz.

this means that the noise performance of the reference LO needs to be that
much better. Furthermore there is a need to adjust signal levels to avoid
distortion. One of the essential characteristics of most module design is that
modules can be freely interconnected without concern that one module will
affect the operation of another module. To achieve this isolation is required.
Another design choice is to use a subharmonic mixer in which the the pump
frequency is about half of the RF frequency. The subharmonic mixer also has
the property that very little of the second harmonic of the LO signal passes
into the RF circuitry. It is very important that the range of frequencies in the
LO path, the RF path, and the IF path do not overlap. With this knowledge
the signal flow path in the receiver RF module can be extended as shown in
Figure 7-18(b).
Using the above considerations and also considering the transmit
requirements leads to the detailed frequency plan shown in Figure 7-19.
This plan is arrived at iteratively considering design and unit cost, as well
as the required time to market. While Figure 7-19 shows the receiver using
the lower sub band and the transmitter using the upper side bands, this sub
band allocation could be reversed and this is achieved by simply switching
CASCADE OF MODULES 247

the diplexer sub band ports, i.e. by rotating the waveguide diplexer.

7.7.3 Summary
Once the frequency plan has been established the design of the individual
modules can proceed independently. The final receiver RF module was
shown in Figures 7-12 and 7-13 with the block diagram of the module shown
in Figure 7-14.
Frequency planning may seem rather simple but it is more than drawing
a spectral diagram such as that in Figure 7-19. The architecture of the
transceiver is defined along with the frequency plan. Thus frequency
planning is done by experienced engineers with a broad appreciation
for RF system performance, knowledge of available off-the-shelf modules,
appreciation for cost, and appreciation for what it takes to develop a
competitive product. Frequency planning has a tremendous impact on
system cost. The final system design has a value far exceeding the cost of
the component sub modules.

7.8 Summary
A microwave system or subsystem is generally constructed as a cascade
of two-port modules. Typically such modules are interconnected on a
circuit board using microstrip structures with the system designer providing
matching networks, transmission line networks, system architecture, and
frequency plan. One or two decades ago crafting such a system using
vendor-supplied modules would have required a significant performance
compromise. As a result microwave companies necessarily developed large
portions of a design in-house that resulted in long design cycles. With
programmability and user-defined adjustments, perhaps by setting bias
levels, the gain, bandwidth, and distortion performance of off-the-shelf
modules can be adjusted. Active modules necessarily introduce distortion
and noise, and so managing dynamic range while managing DC power
consumption is a considerable concern.
Design and analysis of RF and microwave systems is complicated by the
complex signals, i.e., modulated signals and not discrete sinewaves, used
in RF and microwave systems and by the excessively long simulation times
required to analyze microwave circuits with these signals. Design of linear
RF and microwave subsystems such as an amplifier or matching network
can proceed very well by simulating the structure one frequency at a time
and then stepping the signal frequency over the range of interest.
The analysis of nonlinear microwave subsystems such as amplifiers is
more complicated but often it is sufficient to consider two-tones, signals
with two sinusoidal components and then there are efficient simulation
techniques such as harmonic balance analysis that enables subsystem
performance to be efficiently but approximately evaluated. However, neither
of these strategies is sufficient when trying to determine the performance of
RF and microwave systems. In this case complex signals such as digitally
modulated signals must be considered, and there are usually signals at
widely different frequencies to be incorporated in any analysis. There is
however one special property of microwave signals that is exploited, and
this is that nearly every modulated microwave signal has a relatively small
248 STEER MICROWAVE AND RF DESIGN: MODULES

bandwidth compared to the center frequency of the signal. Thus system

simulation strategies have been developed that require only the smaller
bandwidth signal to be fully considered in analysis and bookkeeping used
to track the center frequency of the signal.
In this chapter metrics and design techniques were introduced that enable
competitive designs to be realized quickly. Often a company may decide
that it is to their competitive advantage to design one or more of their
own modules rather than using those from module vendors. Even then,
the large number of modules available enables design concepts to be tested
for feasibility early in the design cycle. Module vendors are challenged
with designing modules with close to ultimate performance, but with
enough adaptability that the module is suited to a wide variety of system
applications. The greater customer base for modules justifies the higher
module design costs leading to acceptable unit costs and high performance.

7.9 References
[1] H. Gutierrez, K. Gard, and M. Steer, “Non- system analysis and design via a contribu-
linear gain compression in microwave am- tion method,” Int. J. on RF and Microwave
plifiers using generalized power-series anal- Computer Aided Engineering, vol. 16, no. 4, pp.
ysis and transformation of input statistics,” 338–345, Jul. 2006.
IEEE Trans. on Microwave Theory and Tech- [7] R. Pettai, Noise in Receiving Systems. John
niques, vol. 48, no. 10, pp. 1774–1777, 2000. Wiley & Sons, 1984.
[2] K. Gharaibeh, K. Gard, H. Gutierrez, and [8] W. Sabin and E. O. Schoenike, Single-Sideband
M. Steer, “The importance of nonlinear order Systems and Circuits. McGraw-Hill, 1987.
in modeling intermodulation distortion and [9] “Itu recommendation f.636 f.636 : Radio-
spectral regrowth,” in 2002 IEEE Radio and frequency channel arrangements for fixed
Wireless Conf., 2002 (RAWCON 2002), 2002, wireless systems operating in the 14.4-15.35
pp. 161–164. ghz band,” Mar. 2012.
[3] S. Maas, “Third-order intermodulation dis- [10] A. Kirilenko, S. Senkevich, V. Tkachenko,
tortion in cascaded stages,” IEEE Microwave and B. Tysik, “Waveguide diplexer and mul-
and Guided Wave Letters, vol. 5, no. 6, pp. 189– tiplexer design,” IEEE Trans. on Microwave
191, Jun. 1995. Theory and Techniques, vol. 42, no. 7, pp. 1393–
[4] ——, Nonlinear Microwave and RF Circuits, 1396, Jul. 1994.
2nd ed. Artech House, 2003. [11] S. Shin and S. Kanamaluru, “Diplexer design
[5] R. Sagers, “Intercept point and undesired re- using EM and circuit simulation techniques,”
sponses,” IEEE Trans. on Vehicular Technology, IEEE Microwave Magazine, vol. 8, no. 2, pp.
vol. 32, no. 1, pp. 121–133, Feb. 1983. 77–82, Feb. 2007.
[6] A. Victor and M. Steer, “Transceiver cascade

7.10 Exercises
1. The first stage of a two-stage amplifier has a lin- in Section 7.2.1.
T
ear gain of 16 dB and output 1 dB gain compres- (e) Compare Po,1dB calculated in (c) and (d) and
sion, P1o,1dB = −20 dBm. For the second stage briefly discuss any discrepancy.
the linear gain is 30 dB and P2o,1dB = 0 dBm. 2. An amplifier has two cascaded stages with lin-
(a) Determine the input-referred gain compres- ear gains of G1 = 20 dB and G2 = 30 dB,
sion, P2i,1dB of stage 2. and output-referred third-order intercepts of
(b) Compare P2i,1dB and P1o,1dB . Which stage OIP31 = 0 dBm and OIP32 = 20 dBm, respec-
dominates gain compression? tively. What is IIP3 of the amplifier? Use the or-
(c) What is the amplifier’s output gain com- ganized cascade intercept method.
T
pression level, Po,1dB considering only com- 3. A single-stage amplifier has a linear gain of
pression from the dominant stage. 16 dB and an output 1 dB gain compression
T
(d) Calculate Po,1dB using the method described point of 10 dBm. A communication signal with
CASCADE OF MODULES 249

a PMEPR of 6 dB is used. What is the maximum 10. The first stage of a two-stage amplifier has a
average power of the input signal before the out- linear power gain of 26 dB, an output 1 dB
put suffers significant compression? This is de- gain compression power of 10 dBm, and an
fined at the point at which the peak signal is output-referred third-order intercept point OIP3
compressed by 1 dB. = 26 dBm. The second stage has a linear power
4. The first stage of a two-stage amplifier has a gain of 10 dB, an output 1 dB gain compression
linear gain G1 = 30 dB and an output 1 dB point of 13 dBm, and an output-referred third-
gain compression point P1o,1dB = −10 dBm. order intercept point OIP3 = 33 dBm.
The second stage has a linear gain G2 = 20 dB (a) What is the linear power gain of the two-
and an output 1 dB gain compression point stage amplifier?
P2o,1dB = 10 dBm. What is the output-referred (b) What is the output 1-dB gain compression
1 dB gain compression point of the cascade am- power of the two-stage amplifier for a sinu-
plifier? [Parallels Example 7.1] soidal RF input signal?
5. An amplifier consists of two cascaded stages. (c) What is the OIP3 of the two-stage amplifier?
The first stage has a linear gain G1 = 30 dB (d) What is the input-referred third-order inter-
and an output 1 dB gain compression point cept point, IIP3?
P1o,1dB = 0.1 dBm. The second stage has a lin- 11. The final RF output of a cell phone has a driver
ear gain G2 = 20 dB and an output 1 dB gain amplifier followed by a power amplifier. The
compression point P2o,1dB = 1 dBm. What is driver amplifier has a linear gain of 30 dB and
the input-referred 1 dB gain compression point an output-referred third-order intercept point,
of the cascade amplifier? [Parallels Example 7.1] OIP3, of 50 dBm. The power amplifier has a lin-
6. The stages of a two-stage amplifier have linear ear gain of 12 dB and an output-referred third-
gains of G1 and G2 , and output 1 dB gain com- order intercept point, OIP3, of 55 dBm. What is
pression powers of P1o,1dB and P2o,1dB , respec- the OIP3 of the driver-power amplifier cascade?
tively. Develop a symbolic expression for the 12. A two-stage amplifier has a linear power gain of
input-referred 1 dB gain compression point of 20 dB, an output 1 dB gain compression point of
the cascade amplifier. 30 dBm, and an output-referred third-order in-
7. An amplifier has two stages with linear gains of tercept point OIP3 = 53 dBm.
G1 = 20 dB and G2 = 30 dB, and output 1 dB (a) What is the power of the maximum input
gain compression powers of P1o,1dB = 0.1 dB- signal when the gain of the amplifier is com-
mand P2o,1dB = 1 dBm, respectively. What is the pressed by 1 dB?
input-referred 1 dB gain compression power of (b) What is the input-referred third-order inter-
the amplifier? cept point, IIP3?
8. The first stage of a two-stage amplifier has a lin- 13. The first stage of a two-stage amplifier has a
ear power gain of 26 dB and an output 1 dB gain linear power gain of 23 dB, an output 1 dB
compression power of 10 dBm. The correspond- gain compression power of 1 dBm, and an
ing parameters of the second stage are 10 dB and output-referred third-order intercept point OIP3
13 dBm. = 20 dBm. The second stage has a linear power
(a) What is the linear power gain of the two- gain of 10 dB, an output 1-dB gain compression
stage amplifier? point of 10 dBm, and an output-referred third-
(b) What is the output 1-dB gain compression order intercept point, OIP3 = 30 dBm. The signal
power of the amplifier for a sinusoidal RF applied to the amplifier uses QPSK modulation
input signal? with a PMEPR of 3 dB.
(c) What is the maximum average output RF (a) What is the linear power gain in decibels of
power of the 64-QAM-modulated signal the two-stage amplifier?
(with a PMEPR of 7.8 dB) for an undistorted (b) What is the output 1-dB gain compression
output (as defined by 1-dB gain compres- power, in dBm, of the two-stage amplifier?
sion)? (c) What is the OIP3, in dBm, of the two-stage
9. The stages of a three-stage amplifier have lin- amplifier?
ear gains of 10 dB, 20 dB, and 20 dB respec- (d) What is the input-referred third-order inter-
tively, and 1 dB output gain compression levels cept point, IIP3m?
of −60 dBm, −40 dBm, and −20 dBm respec- (e) What is the single-tone output power at 1 dB
tively. What is the output power when the gain gain compression?
of the amplifier is compressed by 1 dB? (f) What is the maximum output RF power of
250 STEER MICROWAVE AND RF DESIGN: MODULES

the QPSK-modulated signal for an undis- noise figure of the second stage is 6 dB. The min-
torted output? imum acceptable SNR, SNRmin , at the output of
14. An amplifier has two cascaded stages. The the amplifier is 16 dB.
stages have linear gains of G1 and G2 , and (a) What is the linear power gain of the two-
output-referred third-order intercepts of OIP31 stage amplifier?
and OIP3 dBm, respectively. What is IIP3 of the (b) What is the output 1-dB gain compression
amplifier? power of the two-stage amplifier for a sinu-
soidal RF input signal?
15. An amplifier has two cascaded stages with lin- (c) What is the OIP3 of the two-stage amplifier?
ear gains of G1 = 20 dB and G2 = 30 dB, (d) What is the noise figure of the two-stage am-
and output-referred third-order intercepts of plifier?
OIP31 = 0 dBm and OIP32 = 20 dBm, respec- (e) What is the noise, in dBm, applied to the in-
tively. What is IIP3 of the amplifier? Use the un- put of the two-stage amplifier in a 100 MHz
organized cascade intercept method. bandwidth is the source has a Thevenin re-
16. The first stage of a room-temperature two-stage sistor at room temperature?
amplifier with a 100 MHz bandwidth has a (f) What is the power of the noise, in dBm, in
linear power gain of 26 dB, an output 1 dB a 100 MHz bandwidth at the output of the
gain compression power of 10 dBm, and an two-stage amplifier?
output-referred third-order intercept point OIP3 (g) What is the output-referred spurious free
= 26 dBm. The second stage has a linear power dynamic range of the two-stage amplifier in
gain of 6 dB, an output 1 dB gain compression decibels?
point of 13 dBm, and an OIP3 of 33 dBm. The (h) What is the output-referred dynamic range
noise figure of the first stage is 3 dB and the of the two-stage amplifier in decibels?

7.10.1 Exercises By Section

†
challenging, ‡ very challenging
§7.2 1† , 2, 3, 4, 5† , 6† , 7† , 8, 9, 10, 11† , 12† , 13† , 14, 15 §7.7 16†

7.10.2 Answers to Selected Exercises

3 −11 dBm 10(e) 5.2 dBm
Index
A/D converter, 186 IIP3, 228, 234 digital-to-analog converter, equiripple, 23
accumulator, 213 intercept method, 228 213 excess noise, 131
active noise, 133, 233 diode, 167
filter, 64, 66 OIP3, 228, 234 Gunn, 215
F , 131
ADC, 186 organized intercept IMPATT, 215
Fathelbab, 11
admittance method, 229 junction, 167
ferrite, 177
inverter, 33, 37 unorganized intercept light-emitting, 169
FET, 175
all pole, 18 method, 230 mixer, 187
switch, 175
allpass, 12 case study PIN, 169
field
AM-AM, 144 bandstop filter, 62 Schottky, 167
effect transistor, 175
AM-PM, 144 cascade design, 236 transferred electron
figure of merit
amplifier combline filter, 93 device, 215
oscillator, 204, 205
back-off, 145 down converter, 236 transit time device, 215
filter, 11, 79
operational, 67 Cauer, 17, 29, 31, 44–46 tunnel, 214
design using image
transconductance, 67 characteristic TUNNETT, 215
parameter method, 13
amplitude function, 15 varactor, 168
design using insertion
demodulator, 212 polynomial, 20 Zenner, 169
loss method, 13, 14
distortion, 144 Chebyshev, 17, 23 direct
design using reflection
analog filter, 22 digital synthesizer, 213
coefficient method, 13,
-to-digital converter, 186 chirp, 69 discernible signal,
14
multiplier, 198, 206 circulator, 177, 179 minimum, 152
active, 64, 66, 67
distortion, 144, 225
coherent distributed, 68
AM-AM, 144
B, 131 amplitude demodulator, allpass, 12
AM-PM, 144
back-off 212 bandpass, 12, 61
amplitude, 144
gain, 144 combline bandstop, 12, 62
passive intermodulation,
backward-wave oscillator, filter, 93 Bessel, 33
154
216 case study, 93 biquad, 67
phase, 144
balanced commensurate frequency, biquadratic, 67
PIM, 154
mixer, 197 40, 97, 99 Butterworth, 30, 32
divider
bandpass filter, 12 commutating mixer, 195, prototypes, 29
frequency, 208
transient response, 69 197 Cauer, 32
double
bandstop filter, 12 compression Chebyshev, 22, 31, 32
balanced
BER, 153 gain, 145 prototypes, 29
mixer, 197
Bessel, 17 contribution method, 232 Type I, 33
sideband, 190
biquad filter, 67 conversion, 186 Type II, 33
doubly terminated
biquadratic filter, 67 down, 186 combline, 93
network, 13, 14
BJT loss, 189 Chebyshev, 93
down-conversion, 186
schematic symbol, 8 up-conversion, 186 corner frequency, 40
DPDT switch, 175
Boltzmann constant, 131 corner frequency, 12 dielectric resonator, 62
DPST switch, 175
breakdown, 156 corona effect, 157 elliptical, 32
DR, 151–153
brick-wall filter, 17 correlation end-coupled, 61
duplexer
budget method, 231, 232 noise, 128 equiripple, 23
radar, 180
Butterworth, 17 frequency
DUT, 154
filter, 18 scaling, 40, 41
DAC, 213 dyadic, 178
transformation, 41
BWO, 216 dynamic
DDS, 213 group delay, 60
demodulator, amplitude, range, 151
highpass, 12
212 cascaded system, 235
capture range, PLL, 210 impedance scaling, 40
detectable signal spurious free, 151
carrier impedance
oscillator, 203 minimum, 152 transformation, 40
cascade Dicke switch, 139 effective noise inverter
budget method, 231 dielectric temperature, 128 admittance, 37, 38
contribution method, 232 resonator, 62 electron spin, 177 lumped-element, 101
252 INDEX

scaling, 91 compression, 145 lumped-element, 101 bilateral, 191

ladder, 46 back-off, 144 scaling, 91 commutating, 195, 197
Legendre, 33 cascaded system, 225 stub realization, 39 conversion loss, 189
lowpass, 12 Gilbert IP2, 225 diode, 187, 192
to bandpass, 42, 43 cell, 198 IP3, 225 ring, 191–193
to bandstop, 43, 45 mixer, 198 intercept, 152 single ended, 191
to highpass, 43 group IRR, 190 double balanced, 197
optimum “L”, 33 delay, 60 isolator, 177, 180 triple balanced, 197
parallel coupled line, 61, filter, 60 Gilbert, 198
79 Gunn image, 190
, 21
PCL, 61, 79 diode, 215 rejection ratio, 190
junction diode, 167
RFIC, 66 effect device, 215 image reject, 197
Richards’ transformation gyromagnetic noise, 190
lowpass, 52 effect, 177, 181 k, 131 figure, 189
Richards’s resonance, 178 klystron, 216 performance parameters,
transformation, 50, 53 Kuroda’s identities, 54 189
highpass, 53 HBT RFIC, 198
synthesis, 18 schematic symbol, 8 LC , 189 subharmonic, 196
tank, 68 heterodyne, 186 lanthanoid elements, 177 switching, 194
transformations, 40 heterodyning, 186 LC resonator terms, 189
transient response, 69 highpass, 12 stub equivalence, 95 triple balanced, 197
type transformation, 40 Hurwitz polynomial, 17 LED, 169 unbalanced, 197
YIG, 179 Leeson, 204 waveforms, 191
YIG tuned, 181 Levy, 27 module, 1, 165
final-value theorem, 211 IF, 186 design
light-emitting diode, 169
first IIP3, 152 budget method, 231
linear chirp, 69
order loop, 210 cascaded system, 228, 234 contribution method,
LO, 186, 202
flicker noise, 202 IL, 14 232
phase noise, 202
FOM IM2, 151 multipactor effect, 157
local oscillator, 202
oscillator, 204, 205 IM3, 146, 148, 153 multiplier, 206
phase noise, 202
FPD6836P70 image frequency, 207
long
noise parameters, 141 parameter method, 13 parametric, 208
tail response, 70
frequency reject mixer, 197 reactive, 208
lowpass filter, 12, 15
commensurate, 97 rejection, 190 resistive, 207
conversion, see ratio, 190
conversion IMP, 146 magnetic
IMPATT diode, 215 domain, 177 neodymium, 177
demodulator, PLL, 211
impedance material, 177, 178 network identity
divider, 208
inverter, 33, 34 gyromagnetic effect, 181 Kuroda’s, 54
locked-oscillator, 209
scaling, 40 moment, 177 Norton’s, 54, 56
down-conversion, see
transformation, 40 magnetron, 216 NF, 131
down-conversion
implementation margin, Matthaei, 27 mixer, 189
heterodyne, see
152 maximally flat, 18 noise, 123, 128, 129
heterodyne
inhomogeneous MDS, 152 1/f , 202
multiplier, 207
medium MEMS, 172, 175 temperature
parametric, 208
parallel coupled-line switch, 176 effective, 128
reactive, 208
filter, 114 microstrip cascade, 233
resistive, 207
PCL filter, 114 circulator, 179 cascaded stages, 134
parametric
insertion loss, 14 filter, 79 cascaded system, 133
multiplier, 208
intercept isolator, 179, 180 correlated, 128
reactive multiplier, 208
IP3, 152 microwave correlation coefficient,
resistive multiplier, 207
method, 152, 228–230 modules, 1, 165 128
scaling (filters), 41
intermod, 146 vacuum environmental, 128
synthesizer, 211
intermodulation device, 216 excess, 130, 131
transformation (filters),
distortion tube, 216 factor, 131, 134, 189
41
cascaded system, 228, minimum system, 133
up-conversion, see
234 detectable signal, 152 figure, 131
up-conversion
product, 146 discernible signal, 152 amplifier model, 139
Friis’s formula, 134, 233
inverter, 33 mixer, 186 measurement, 135
admittance, 33, 37, 38 analysis, 187 measurement, low
gain impedance, 33, 34, 36 balanced, 191–193, 197 noise, 137
INDEX 253

measurement, Y-factor FOM, 204, 205 noise, 128 spin, 177

method, 135 voltage-controlled, 204 prototype, 17 SPST switch, 175
mixer, 189 OTA, 67 PSD spurious free dynamic
system, 133 output noise, 128 range, 151
two-ports, 139 intercept point, 152 pump, 186 spurious passband, 114
flicker, 202 noise, 132 standard
physical origin, 130 temperature, 131
radar
folding, 190 stub
parallel duplexer, 180
Friis’s formula, 134, 233 capacitively-loaded, 95
coupled-line radiometer, 138
measures, 130 lumped resonator
filter inhomogeneous rare earth elements, 177
mixer, 190 equivalence, 95
medium, 114 RdB , 22, 31
observation subharmonic mixer, 196
parametric frequency reactive frequency
frequency domain, 141 switch, 172, 175
multiplier, 208 multiplier, 207, 208
output of a two-port, 132 diode, 175
passband ripple, 22, 31 rectifier, 167
phase, 202 FET, 175
passive remote sensing, 138
VCO, 205 MEMS, 172, 175, 176
intermodulation resistor
PSD, 128 pHEMT, 172, 175
distortion, 154 noise at room
room temperature PIN, 175
PBR, 22 temperature, 126
resistor, 126 PIN diode, 172, 175
PCL resonator
shot, 129 switching
filter in inhomogeneous LC-stub equivalence, 95
physical origin, 129 mixer, 194
medium, 114 RF, 186
single sideband, 204 synthesis
PCL filter, 79 modules, 1, 165
source, 136 ladder, 24, 26
peak-to-mean envelope RFIC
temperature, 128 synthesizer, 211
power ratio, 144 mixer, 198
measurement, 137 direct digital, 213
PFD Richards’
thermal transformation, 53
phase-frequency
capacitor, 129 highpass, 53 T0 , 131
detector, 207
physical origin, 126 Richards’ transformation, tank circuit, 68
phase
value at room 50 temperature
accumulator, 213
temperature, 205 lowpass, 52 noise, 128
comparator, 206
uncorrelated, 128 ring mixer tensor, 178
detector, 206
white, 128 transistor, 195, 200 third
gain factor, 210
Y-factor, 135 ripple, 22, 31 -order intercept, 146
distortion, 144
definition, 136 factor, 22, 31 TPR, 14
frequency detector, 207
nonlinear room temperature transconductance
locked loop, 209
distortion, 144, 225 noise of a resistor, 126 amplifier, 67
modulator, 212
Norton’s transducer
noise, 202
identity, 56 function, 14
oscillator, 202 samarium, 177
transformation, 56 power ratio, 14
VCO, 205 satellite, 138
numerically controlled transferred electron device,
PIM, 154 Scanlan, 27
oscillator, 213 215
PIN diode, 169, 175 schematic symbols, 5
switch, 175 transit time device, 215
Schottky
OIP3, 152, 229 PLL, 208, 209 transmission
diode, 167
cascaded system, 228, 234 amplitude demodulator, coefficient, 14
secondary electron
operational amplifier, 67 212 TRAPATT, 215
emission, 157
organized cascade applications, 211 traveling wave tube, 216
SEY, 157
intercept method, 229 triple balanced mixer, 197
capture range, 210 SFDR, 151–153
tube
oscillator first-order, 210 shot noise, 129
backward-wave
carrier, 203 frequency demodulator, signal
oscillator, 216
figure of merit, 204 211 -to-noise ratio, see SNR
travelling wave, 216
local, 202 phase detector two-tone, 146
tuning
phase noise, 202 gain factor, 210 singly terminated network,
constant, 204
noise, 202 phase modulator, 212 13
gain, 205
numerically controlled, synthesizer, 211 SNR, 130, 152, 153
tunnel diode, 214
213 VCO, 210 SNRMIN , 152
TUNNETT diode, 215
phase noise, 202 PMEPR, 144 sources, 214
two-tone signal, 146
quadrature VCO, 205 back-off, 144 SP4T switch, 175
TWT, 216
QVCO, 205 pole-zero description, 17 SPDT switch, 175
VCO, 205 power spectrum
figure of merit, 204, 205 spectral density analyzer, 181 UE, 33
254 INDEX

unbalanced oscillator, 216 VCO, 204 Y-factor

mixer, 197 klystron, 216 phase noise, 205 definition, 136
unit magnetron, 216 PLL, 210 method, 135
element, 33 travelling wave tube, YIG, 181
voltage
unorganized cascade 216 filter, 179
controlled oscillator, 204
intercept method, 230 TWT, 216 tuned filter, 181
up-conversion, 186 in phase-locked loop, yttrium-iron-garnet, 181
TWTA, 216
210
tube, 216
vacuum TWT, 216 Zener diode, 169
device, 216 TWTA, 216 white noise, 128
backward-wave varactor diode, 168 WiFi, 12
MICROWAVE AND RF DESIGN: MODULES Third Edition Volume 4

STEER
MICROWAVE
Microwave and RF Design: Modules focuses on the design of systems based on microwave modules. The
use of modules has become increasingly important in RF and microwave engineering for rapidly realizing
high performance microwave systems. When integration is ultimately to be used, building a system up using
modules provides a rapid means of prototyping and testing system concepts. A wide variety of RF modules

AND RF DESIGN
including amplifiers, local oscillators, switches, circulators, isolators, phase detectors, frequency multipliers
and dividers, phase-locked loops, and direct digital synthesizers are considered. Detailed design strategies
for synthesizing filters based on parallel coupled lines are presented. The reader will gain an appreciation
of design by synthesis. This book is suitable as both an undergraduate and graduate textbook, as well as a
career-long reference book.

MICROWAVE AND RF DESIGN: MODULES, Volume 4

MODULES
KEY FEATURES OTHER VOLUMES
• The fourth volume of a comprehensive series on Microwave and RF Design
microwave and RF design
Radio Systems
• Open access ebook editions are hosted by NC State
Volume 1
University Libraries at:
ISBN 978-1-4696-5690-8
https://repository.lib.ncsu.edu/handle/1840.20/36776
• 23 worked examples
• An average of 21 exercises per chapter Microwave and RF Design
• Answers to selected exercises Transmission Lines
• 6 case studies illustrating design procedures Volume 2
• Emphasis on synthesis as well as building a rich ISBN 978-1-4696-5692-2
library of microwave functions
• A companion book, Fundamentals of Microwave Microwave and RF Design
and RF Design, is suitable as a comprehensive Networks
undergraduate textbook on microwave engineering Volume 3
ISBN 978-1-4696-5694-6
ABOUT THE AUTHOR
Microwave and RF Design
Michael Steer is the Lampe Distinguished Professor
Amplfiers and Oscillators
of Electrical and Computer Engineering at North
Volume 5
Carolina State University. He received his B.E. and Ph.D.
degrees in Electrical Engineering from the University of ISBN 978-1-4696-5698-4
Queensland. He is a Fellow of the IEEE and is a former
editor-in-chief of IEEE Transactions on Microwave ALSO BY THE AUTHOR
Theory and Techniques. He has authored more than 500 Fundamentals of Microwave
publications including twelve books. In 2009 he received and RF Design
a US Army Medal, “The Commander’s Award for Public ISBN 978-1-4696-5688-5
Service.” He received the 2010 Microwave Prize and the
2011 Distinguished Educator Award, both from the IEEE
Microwave Theory and Techniques Society.

Published by NC State University Distributed by UNC Press

Michael Steer Third Edition