Load-Pull Techniques with Applications to Power

Amplifier Design
The Springer Series in Advanced Microelectronics provides systematic information
on all the topics relevant for the design, processing, and manufacturing of micro-
electronic devices. The books, each prepared by leading researchers or engineers in
their fields, cover the basic and advanced aspects of topics such as wafer processing,
materials, device design, device technologies, circuit design, VLSI implementation,
and subsystem technology. The series forms a bridge between physics and engineer-
ing and the volumes will appeal to practicing engineers as well as research scientists.

Series Editors:
Dr. Kiyoo Itoh
Hitachi Ltd., Central Research Laboratory, 1-280 Higashi-Koigakubo,
Kokubunji-shi, Tokyo 185-8601, Japan

Professor Thomas H. Lee

Department of Electrical Engineering, Stanford University, 420 Via Palou Mall,
CIS-205 Stanford, CA 94305-4070, USA

Professor Takayasu Sakurai

Center for Collaborative Research, University of Tokyo, 7-22-1 Roppongi,
Minato-ku, Tokyo 106-8558, Japan

Professor Willy Sansen

ESAT-MICAS, Katholieke Universiteit Leuven, Kasteelpark Arenberg 10,
3001 Leuven, Belgium

Professor Doris Schmitt-Landsiedel

Lehrstuhl für Technische Elektronik, Technische Universität München,
Theresienstrasse 90, Gebäude N3, 80290 Munich, Germany

For further volumes:

www.springer.com/series/4076
Fadhel M. Ghannouchi r Mohammad S. Hashmi

Load-Pull
Techniques
with Applications
to Power Amplifier
Design
Fadhel M. Ghannouchi Mohammad S. Hashmi
Electrical and Computer Engineering, Electrical and Computer Engineering,
Intelligent RF Radio Laboratory Intelligent RF Radio Laboratory
University of Calgary University of Calgary
Calgary, Alberta Calgary, Alberta
Canada Canada

ISSN 1437-0387 Springer Series in Advanced Microelectronics

ISBN 978-94-007-4460-8 ISBN 978-94-007-4461-5 (eBook)
DOI 10.1007/978-94-007-4461-5
Springer Dordrecht Heidelberg New York London

Library of Congress Control Number: 2012939475

© Springer Science+Business Media Dordrecht 2013

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed. Exempted from this legal reservation are brief excerpts in connection
with reviews or scholarly analysis or material supplied specifically for the purpose of being entered
and executed on a computer system, for exclusive use by the purchaser of the work. Duplication of
this publication or parts thereof is permitted only under the provisions of the Copyright Law of the
Publisher’s location, in its current version, and permission for use must always be obtained from Springer.
Permissions for use may be obtained through RightsLink at the Copyright Clearance Center. Violations
are liable to prosecution under the respective Copyright Law.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
While the advice and information in this book are believed to be true and accurate at the date of pub-
lication, neither the authors nor the editors nor the publisher can accept any legal responsibility for any
errors or omissions that may be made. The publisher makes no warranty, express or implied, with respect
to the material contained herein.

Printed on acid-free paper

Springer is part of Springer Science+Business Media (www.springer.com)

Fadhel M. Ghannouchi dedicates this book to
his wife Asma, and daughters Layla and
Nadia
Mohammad S. Hashmi dedicates this book to
his wife Rabeya, and son Jafar
Preface

For the purpose of identifying the large-signal behavior of the transistor devices, the
use of linear S-parameter is often inadequate. Large-signal characterization, there-
fore, is essential for the estimation and determination of the device performance in
the nonlinear domain. The load-pull approach is one recommended approach for
the large-signal characterization, optimization, and design of transistor devices and
radio frequency (RF), microwave and mm-wave power amplifiers (PAs).
The load-pull technique was first reported almost four decades ago. This pioneer-
ing work brought a paradigm shift in the characterization, measurement, and opti-
mization of transistor devices and PAs. The first load-pull setup can be considered
rudimentary but has definitely helped in advancing the state-of-the-art in load-pull
techniques.
This book describes the principles of operation, calibration, design and real-
ization approaches and application of load-pull techniques in the context of PAs.
It explores the topic from the basic principles of load-pull techniques through to
their many interesting advancements, including passive and active techniques, high
power load-pull and envelope load-pull setups with applications to amplifier, mixer
and noise measurements. In addition, the book also covers waveform engineering
systems, their calibration techniques and applications.
The book can be used by graduate students, researchers and design engineers in
microwave and wireless design areas. It is assumed that the readers have already ac-
quired a basic knowledge of RF and microwave circuit design. A solid background
in transmission line theory and basic communication concepts is required. The book
may also be used as a textbook for a graduate course on large signal device mea-
surement and characterization.
Chapter 1 is a brief reminder of the basic concepts related to PA characteristics,
figures of merits of PA, power amplifier classes of operation, PA design methodolo-
gies, and introduction to load-pull systems along with their important features.
Chapter 2 is dedicated to passive load-pull techniques. It explains the funda-
mentals of passive load tuning techniques and elaborates on the two most com-
mon techniques, namely electronic tuner (ETS) and electromechanical tuner (EMT),
employed to achieve impedance tuning using passing approach. Measurement and

vii
viii Preface

calibration procedure applied in a load-pull measurement setup is then discussed

in detail. The chapter also provides extensive details on various passive harmonic
load-pull architectures along with their respective advantages and limitations. Sub-
sequently, common techniques used to enhance the tuning range of passive load-pull
setups are discussed.
Chapter 3 provides extensive details on active load-pull techniques and systems.
It starts with closed-loop active load-pull technique and its realization methods. Ad-
equate details have been included for the design of application specific closed-loop
load-pull system. It then covers feed-forward active load-pull and various methods
to develop hybrid setup, for enhancing the tuning range and achieving highly reflec-
tive load-pull systems. Active open-loop load-pull requires iterative operation of the
system for converging on optimal impedance solution. The last section is dedicated
to an algorithm for high speed convergence in active open-loop load-pull systems.
Chapter 4 presents the theory, techniques and principles behind using six-port
reflectometer in reverse and forward configurations to characterize transistors oper-
ated in large signal conditions, and the issues related to the implementation of these
techniques are discussed. Source-pull characterization using six-port reflectometer
for transistor noise measurement, mixer testing and design, as well as oscillator de-
vice line measurement purposes are explained and discussed. AM/AM and AM/PM
distortion measurement and passive and active load-pull large signal characteriza-
tion of transistors using the six-port reflectometer technique are also presented and
discussed.
Chapter 5 deals with the issues involved in the characterization of high power mi-
crowave transistor devices. There are multiple aspects that need to be addressed in
order to overcome those issues. All these have been discussed in detail in this chap-
ter. The techniques adopted in customizing the load-pull setup for high power de-
vice measurements and characterization applications have been elaborated and ex-
plained in detail. Finally, emerging solutions catering to large periphery high power
microwave devices are presented.
Chapter 6 presents the theory of active envelope load pull (ELP) and the associ-
ated design and calibration techniques of active envelope load-pull. Thereafter har-
monic envelope load-pull is explained in detail which is followed by some unique
measurement applications of envelope load-pull system.
Chapter 7 is dedicated to theory and calibration approaches adopted in develop-
ing error corrected nonlinear time-domain waveform measurements systems. Sub-
sequently the concept of waveform engineering is presented. Finally, a number of
applications of waveform engineering system are discussed.
Chapter 8 presents some advanced applications and configurations of load-pull
setups. The first part of this chapter primarily discusses the concept of load-pull sys-
tems for multi-tone and modulated excitations. It experimentally demonstrates that
such systems are extremely useful for real life practical applications. Then the use of
load-pull and source-pull systems in noise characterization is described in detail. Fi-
nally application of load-pull systems in mixer characterization and measurements
in presented.
Acknowledgements

We would like to gratefully acknowledge the help and support received from friends,
colleagues, support staff and students, both past and present at iRadio Laboratory,
University of Calgary, Calgary; Poly-grames Research Center, Ecole Polytechnique,
Montreal; and Cardiff University, UK. We are grateful to our great students and re-
searchers; this book could not have been completed without their fruitful research.
In particular, we would like to thank Dr. R.G. Bosisio and Dr. P.J. Tasker for their
useful comments, discussions, collaboration and for their help in producing many
of the results presented in this book over the years. In addition, we would like to
thank C. Heys for her help in proofreading and formatting the manuscript and Ivana
d’Adamo for her administrative support. The authors would also like to thank IEEE
and Focus Microwaves for their permission and courtesy to reproduce several fig-
ures and illustrations published in their journals and/or application notes.
Dr. M. Hashmi acknowledges Alberta Innovates Technology Futures (AITF), Al-
berta, Canada for the financial contribution to support the post-doctoral fellowship
at iRadio Laboratory, University of Calgary, which helped the completion of this
work. Dr Ghannouchi acknowledges the main sponsors and financial supporters of
the Intelligent Radio Laboratory (iRadio Lab), Alberta Innovates Technology Fu-
tures (AITF), Alberta, Canada, the Canada Research Chairs (CRC) program and
Natural Science and Engineering council of Canada (NSERC).
Finally we would like to profoundly thank our respective spouses Asma and
Rabeya, and children Layla Ghannouchi, Nadia Ghannouchi and Jafar Talal Hashmi
for their understanding and patience throughout the many evenings and weekends
taken to prepare this book. We are also thankful to our respective patents for their
encouragement and valuable support in our early professional years as graduate stu-
dents and young researchers.

ix
Contents

1 Fundamentals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 RF Power Amplifier Characteristics . . . . . . . . . . . . . . . . . 2
1.3 Figures of Merit . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3.1 Drain Efficiency and Power Added Efficiency . . . . . . . 5
1.3.2 Intermodulation and Harmonic Distortions . . . . . . . . . 6
1.3.3 Adjacent Channel Power Ratio . . . . . . . . . . . . . . . 8
1.3.4 Error Vector Magnitude . . . . . . . . . . . . . . . . . . . 9
1.4 Power Amplifier . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.5 Power Amplifier Design Methodologies . . . . . . . . . . . . . . 14
1.5.1 CAD-Based Design Methods . . . . . . . . . . . . . . . . 14
1.5.2 Measurement-Based Design Methods . . . . . . . . . . . . 15
1.6 Nonlinear Microwave Measurement System . . . . . . . . . . . . 16
1.6.1 What Is Load-Pull? . . . . . . . . . . . . . . . . . . . . . 17
1.6.2 Why Load-Pull? . . . . . . . . . . . . . . . . . . . . . . . 17
1.7 Important Load-Pull Features . . . . . . . . . . . . . . . . . . . . 18
1.7.1 Repeatability of Reflection Coefficients . . . . . . . . . . . 19
1.7.2 Tuning Range and Its Distribution . . . . . . . . . . . . . . 19
1.7.3 Tuning Speed . . . . . . . . . . . . . . . . . . . . . . . . 20
1.7.4 Power Handling Capability . . . . . . . . . . . . . . . . . 20
1.7.5 Tuner Resolution . . . . . . . . . . . . . . . . . . . . . . . 20
1.7.6 Tuner Bandwidth . . . . . . . . . . . . . . . . . . . . . . 21
1.7.7 Tuner Size . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.8 Common Load-Pull Systems . . . . . . . . . . . . . . . . . . . . 22
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2 Passive Load-Pull Systems . . . . . . . . . . . . . . . . . . . . . . . . 29
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.2 Passive Load-Pull System . . . . . . . . . . . . . . . . . . . . . . 30
2.2.1 Electromechanical Tuner (EMT) . . . . . . . . . . . . . . 30
2.2.2 Electronic Tuner (ETS) . . . . . . . . . . . . . . . . . . . 33

xi
xii Contents

2.2.3 ETS and EMT Comparisons . . . . . . . . . . . . . . . . . 34

2.3 Load-Pull Measurement . . . . . . . . . . . . . . . . . . . . . . . 35
2.3.1 Load-Pull Setup . . . . . . . . . . . . . . . . . . . . . . . 36
2.3.2 System Calibration . . . . . . . . . . . . . . . . . . . . . . 38
2.4 Harmonic Load-Pull System . . . . . . . . . . . . . . . . . . . . . 42
2.4.1 Triplexer Based Harmonic Load-Pull Setup . . . . . . . . . 44
2.4.2 Harmonic Rejection Tuner Based Harmonic Load-Pull Setup 45
2.4.3 Single Tuner Harmonic Load-Pull Setup . . . . . . . . . . 46
2.4.4 Harmonic Load-Pull Comparisons . . . . . . . . . . . . . 47
2.5 Tuning Range Enhancement . . . . . . . . . . . . . . . . . . . . . 49
2.5.1 Enhanced Loop Architecture . . . . . . . . . . . . . . . . 50
2.5.2 Cascaded Tuner . . . . . . . . . . . . . . . . . . . . . . . 51
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3 Active Load-Pull Systems . . . . . . . . . . . . . . . . . . . . . . . . 55
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.2 Closed-Loop Load-Pull System . . . . . . . . . . . . . . . . . . . 56
3.2.1 System Realization . . . . . . . . . . . . . . . . . . . . . 56
3.2.2 Analysis of Closed-Loop System . . . . . . . . . . . . . . 57
3.3 Closed-Loop Load-Pull Architectures . . . . . . . . . . . . . . . . 62
3.4 Optimized Closed-Loop Load-Pull System . . . . . . . . . . . . . 64
3.5 Feed-Forward Load-Pull System . . . . . . . . . . . . . . . . . . 68
3.6 Optimized Feed-Forward Load-Pull System . . . . . . . . . . . . 71
3.7 Harmonic Feed-Forward Load-Pull System . . . . . . . . . . . . . 74
3.8 Open-Loop Load-Pull System . . . . . . . . . . . . . . . . . . . . 76
3.9 Convergence Algorithm for Open-Loop and Feed-Forward
Load-Pull Techniques . . . . . . . . . . . . . . . . . . . . . . . . 78
3.10 Comparison of Active Load-Pull Techniques . . . . . . . . . . . . 83
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4 Six-Port Based Load-Pull System . . . . . . . . . . . . . . . . . . . . 87
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
4.2 Impedance and Power Flow Measurement . . . . . . . . . . . . . 88
4.3 SP in Reverse Configuration . . . . . . . . . . . . . . . . . . . . . 89
4.3.1 SP Calibration in Reverse Configuration . . . . . . . . . . 89
4.3.2 Error Box Calculation . . . . . . . . . . . . . . . . . . . . 93
4.3.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . 94
4.4 SP Based Source-Pull Configuration . . . . . . . . . . . . . . . . 95
4.5 SP Based Load-Pull Configuration . . . . . . . . . . . . . . . . . 96
4.5.1 Passive Load-Pull System . . . . . . . . . . . . . . . . . . 96
4.5.2 Active Branch Load-Pull System . . . . . . . . . . . . . . 97
4.5.3 Active Loop Load-Pull System . . . . . . . . . . . . . . . 99
4.6 On-Wafer Load-Pull Measurements . . . . . . . . . . . . . . . . . 99
4.7 Applications of Source-Pull Setup . . . . . . . . . . . . . . . . . . 101
4.7.1 Low Noise Amplifier Characterization . . . . . . . . . . . 102
4.7.2 Mixer Characterization . . . . . . . . . . . . . . . . . . . 103
Contents xiii

4.7.3 Power Amplifier Characterization . . . . . . . . . . . . . . 104

4.8 Oscillator Measurements . . . . . . . . . . . . . . . . . . . . . . 104
4.9 AM/AM and AM/PM Measurements . . . . . . . . . . . . . . . . 106
4.9.1 Principles of Operation . . . . . . . . . . . . . . . . . . . 107
4.9.2 Measurement Procedure . . . . . . . . . . . . . . . . . . . 110
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
5 High-Power Load-Pull Systems . . . . . . . . . . . . . . . . . . . . . 113
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
5.2 Limitations of Existing Load-Pull Systems . . . . . . . . . . . . . 113
5.2.1 Problems Due to High Standing Waves . . . . . . . . . . . 114
5.2.2 Problem of Large Load-Pull Power . . . . . . . . . . . . . 118
5.3 High-Power Load-Pull . . . . . . . . . . . . . . . . . . . . . . . . 119
5.3.1 Pre-matching Technique . . . . . . . . . . . . . . . . . . . 120
5.3.2 Enhanced Loop Architecture . . . . . . . . . . . . . . . . 122
5.3.3 Quarter Wave Transformer Technique . . . . . . . . . . . . 124
5.3.4 Broadband Impedance Transformer Technique . . . . . . . 126
5.4 Impact of a Transformation Network on PLP and VSWR . . . . . . 127
5.5 Hybrid Load-Pull System . . . . . . . . . . . . . . . . . . . . . . 130
5.6 Calibration and Data Extraction . . . . . . . . . . . . . . . . . . . 133
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
6 Envelope Load-Pull System . . . . . . . . . . . . . . . . . . . . . . . 139
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
6.2 Envelope Load-Pull Concept . . . . . . . . . . . . . . . . . . . . 140
6.2.1 Mathematical Formulation . . . . . . . . . . . . . . . . . . 140
6.3 Practical Realization . . . . . . . . . . . . . . . . . . . . . . . . . 142
6.3.1 Design of Control Unit . . . . . . . . . . . . . . . . . . . 142
6.4 ELP Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
6.4.1 Error Flow Model Formulation . . . . . . . . . . . . . . . 145
6.4.2 Simplification of the Error Flow Model . . . . . . . . . . . 146
6.4.3 Calibration Technique . . . . . . . . . . . . . . . . . . . . 148
6.4.4 Evaluation of the Calibration Technique . . . . . . . . . . 150
6.5 Stability Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 153
6.6 Features of the Envelope Load-Pull System . . . . . . . . . . . . . 154
6.7 Harmonic Envelope Load-Pull System . . . . . . . . . . . . . . . 155
6.8 Unique Measurement Applications . . . . . . . . . . . . . . . . . 157
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
7 Waveform Measurement and Engineering . . . . . . . . . . . . . . . 163
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
7.2 Theoretical Formulation . . . . . . . . . . . . . . . . . . . . . . . 164
7.3 Historical Perspectives . . . . . . . . . . . . . . . . . . . . . . . . 165
7.4 Practical Waveform Measurement System . . . . . . . . . . . . . 169
7.5 System Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . 170
7.5.1 First Step: Power Flow Calibration . . . . . . . . . . . . . 171
xiv Contents

7.5.2 Second Step: S-Parameter Calibration . . . . . . . . . . . . 172

7.5.3 Third Step: Enhanced Calibration . . . . . . . . . . . . . . 174
7.5.4 Calibration Evaluation . . . . . . . . . . . . . . . . . . . . 175
7.6 Six-Port Based Waveform Measurement System . . . . . . . . . . 177
7.6.1 Multi-harmonic Reference Generator . . . . . . . . . . . . 178
7.6.2 SP Reflectometer Principle . . . . . . . . . . . . . . . . . 178
7.6.3 Multi-harmonic SP Reflectometer Architecture . . . . . . . 179
7.6.4 Multi-harmonic SP Reflectometer Calibration . . . . . . . 181
7.6.5 Calibration Verification . . . . . . . . . . . . . . . . . . . 182
7.7 Waveform Engineering . . . . . . . . . . . . . . . . . . . . . . . 183
7.8 Applications of Waveform Engineering . . . . . . . . . . . . . . . 184
7.8.1 Transistor Characterization . . . . . . . . . . . . . . . . . 184
7.8.2 CAD Incorporation . . . . . . . . . . . . . . . . . . . . . 185
7.8.3 Power Amplifier Design . . . . . . . . . . . . . . . . . . . 186
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
8 Advanced Configurations and Applications . . . . . . . . . . . . . . 191
8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
8.2 Multi-tone Load-Pull Technique . . . . . . . . . . . . . . . . . . . 191
8.3 Real-Time Multi-harmonic Load-Pull Technique . . . . . . . . . . 197
8.4 Modulated Signal Load-Pull Technique . . . . . . . . . . . . . . . 201
8.5 Multi-tone Envelope Load-Pull Technique . . . . . . . . . . . . . 204
8.6 Wideband Load-Pull Technique . . . . . . . . . . . . . . . . . . . 208
8.6.1 Wideband Load-Pull Approach . . . . . . . . . . . . . . . 209
8.6.2 Setup Description . . . . . . . . . . . . . . . . . . . . . . 210
8.7 Noise Characterization . . . . . . . . . . . . . . . . . . . . . . . . 212
8.7.1 Noise Parameter Measurement . . . . . . . . . . . . . . . 212
8.7.2 Noise Parameter Test Setup . . . . . . . . . . . . . . . . . 215
8.8 Mixer Characterization . . . . . . . . . . . . . . . . . . . . . . . 217
8.8.1 Measurement Setup . . . . . . . . . . . . . . . . . . . . . 217
8.8.2 Experimental Procedure . . . . . . . . . . . . . . . . . . . 220
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
Authors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
About the Book . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
Chapter 1
Fundamentals

This chapter presents the fundamentals of three aspects, namely radio frequency
(RF) power amplifiers, approaches adopted for the optimal design of these ampli-
fiers, and load-pull measurement systems.

1.1 Introduction
The key component in any wireless communication system is the RF power am-
plifier (RFPA). RFPAs convert the DC power into RF power, which enables the
transmission of the RF signal containing digital information from the transmitter to
the receiver via wireless environments. The quality of RFPAs, in terms of efficiency
and linearity, has a significant impact on the cost, reliability, size and performance
of wireless communication systems.
Achievement of both high efficiency and linearity in RFPAs is a complex process,
as the power amplifier (PA) has to be designed to operate closer to saturation. Opera-
tion of the PA at saturation increases the efficiency, but also increases the distortion.
Therefore, the RFPA design process requires a trade-off between two important met-
rics, namely efficiency and linearity. Here, efficiency refers to the capability of the
PA in converting DC power into RF power, which is extremely important as any PA
exhibiting low efficiency will consume much more DC power than an efficient one
when delivering the same amount of RF power.
For a handheld application that uses a battery as the primary power source, poor
efficiency means shorter talk and standby times. This can seriously limit the com-
petency of the final product in the market. Moreover, poor efficiency leads to a large
amount of heat dissipation, which requires additional effort and money for high-
capacity cooling facilities.
In the context of the PA, linearity refers to a measure of how faithfully a PA
can amplify the input signal. PAs always demonstrate certain levels of nonlineari-
ties. The tolerance for nonlinearity really depends on the applications. For example,
CDMA (code division multiple access) and WCDMA (wideband code division mul-
tiple access) based wireless systems require, apart from the usual high efficiency,

F.M. Ghannouchi, M.S. Hashmi, Load-Pull Techniques with Applications to Power 1

Amplifier Design, Springer Series in Advanced Microelectronics 32
DOI 10.1007/978-94-007-4461-5_1, © Springer Science+Business Media Dordrecht 2013
2 1 Fundamentals

Fig. 1.1 Generic block diagram of a single-stage RF amplifier

a high linearity over a large dynamic power range. In contrast, the linearity is a
relatively less serious issue for PAs based on the GSM (global system for mobile
applications) standards, due to the constant envelope characteristics of the GSM
signal.
It is not possible for the efficiency and linearity to reach their optimal levels
simultaneously on a stand-alone power device. Various types of advanced PA ar-
chitectures, such as Doherty, Kahn and LINC (linear amplification using nonlinear
components), have been proposed to achieve good efficiency and linearity at the
same time [1–10]. These architectures provide promising results, but inevitably in-
crease the complexity of the communication systems, thereby contributing to higher
costs with decreasing reliability. Therefore, it is always a challenge for PA design-
ers to adopt an appropriate optimization solution with which a good compromise in
specifications can be reached.

1.2 RF Power Amplifier Characteristics

Before getting into the description of RFPAs, it is prudent to understand the charac-
teristics of RF amplifiers. For this purpose, a generic block diagram of a single-stage
RF amplifier is depicted in Fig. 1.1. It consists of input and output matching net-
works, and a transistor device. Ideally, this configuration is generic both for small-
or large-signal power amplifiers.
Matching networks are application specific and are appropriately designed. For
example, matching networks can be customized for amplifying extremely small sig-
nals with very low noise at the output (i.e., a low noise amplifier), can be designed
such that maximum gain is realized, or can be intended for maximum output power
(i.e., a power amplifier). Irrespective of the applications, the matching networks are
realized using lumped components, distributed elements (transmission lines) or a
combination of both. They are basically characterized by intrinsic linear elements
and, therefore, are considered linear time-invariant networks.
An active device can, however, be thought as either a linear or nonlinear network,
depending on the electrical power range of the signals involved in the process [11].
If the power level of the input signal is so small that the output is just the amplified
version of input, as depicted in Fig. 1.2, the device is considered to be operating
in linear mode. On the contrary, if the power level of the input is large enough to
1.2 RF Power Amplifier Characteristics 3

Fig. 1.2 Depiction of linear

and nonlinear behavior of an
amplifier

Fig. 1.3 Schematic of an

amplifier for definition of
load line

generate harmonics in the output signal, as also depicted in Fig. 1.2, the device is
said to be operating in nonlinear mode.
The linear mode of operation is modeled in terms of scattering parameters (S-
parameters). The S-parameters are frequency dependent, described for a specific
biasing condition and independent of the power level of the stimulus. These param-
eters are inadequate for describing the characteristics of devices operating in non-
linear mode. In order to address this problem, there have been proposals of complex
large-signal models [12–15] to describe the nonlinear characteristics of devices.
In an alternative context, it is a common practice to link the performance char-
acteristics of a power amplifier to the load line, as it identifies the capabilities of
devices for maximum output power application. The load line represents the trajec-
tory of all the instantaneous values of current (iDS ) and voltage (vDS ) of a device
when it is operated under a specific load and at a given bias point.
For the purposes of the explanation of a load line, let us consider the typical
schematic of an amplifier, as given in Fig. 1.3. The amplifier is biased through an
RF choke inductor, LC , with a bias voltage, VDD , and drain bias current, IDD . The
DC blocking capacitor, C0 , is selected to be large enough to keep a steady-state
voltage, VDD , during the entire RF cycle. During a steady state, this schematic can
4 1 Fundamentals

Fig. 1.4 Typical load line of

an amplifier (dashed straight
line: resistive load, solid
elliptic line: complex load)

be solved to determine the expression between the instantaneous current (iDS ) and
instantaneous voltage (vDS ) given in Eq. (1.1).
vDS − VDD
iDS = IDD − (1.1)
ZL
Equation (1.1), known as the load line equation, defines the trajectory of all the
combination of values (iDS , vDS ) of the device under specific operating conditions
of a device for specific quiescent points (IDD , VDD ) and load impedance (ZL ).
For the real value of ZL , the load line equation represents a straight line; whereas a
complex ZL converts the load line equation into the equation of a shifted and rotated
ellipse [78], as shown in Fig. 1.4.
In theory, vDS could be any value and, therefore, can produce any value of iDS .
In practice, however, this is not the case, as the load line is regulated by the device
DCIV characteristic, as can be seen in Fig. 1.4. It is evident that the trajectory is lim-
ited by the knee voltage (Vmin or Vknee ), the breakdown voltage (Vmax or Vbreakdown ),
the device’s maximum current (Imax ), and the zero value current (Imin ).
The load line is considered a very effective tool, as it can provide information
about the load impedance (real or complex) just by visual inspection. However, it
is imperative to understand that if an RF transistor is loaded with 50  impedance
without a matching circuit, the load line is an ellipse instead of a straight line. It is
due to the presence of intrinsic device output capacitance that combines with the real
50  impedance to form a complex load. In addition, load line trajectories also help
in the design of optimal matching circuit by determining the condition when the
device output capacitance is canceled out by the matching circuit i.e., the situation
when the device sees only real load impedance at its terminals.

1.3 Figures of Merit

Evaluation of an RF power amplifier for any particular application is carried out
using established figures of merit, such as drain efficiency (ηD ), power-added ef-
ficiency (PAE), harmonic distortion, intermodulation distortion, Adjacent Channel
1.3 Figures of Merit 5

Fig. 1.5 Power flow depiction in a power amplifier

Power Ratio (ACPR), and Error Vector Magnitude (EVM). The values of these fig-
ures of merit are the reference values for comparison with other amplifiers using
other design techniques or technologies.

1.3.1 Drain Efficiency and Power Added Efficiency

Several forms of the definition of efficiency have been adopted previously, but the
most commonly used are drain efficiency (ηD ) and the power-added efficiency
(PAE). For defining the efficiency metrics, let us look to the power flow in a power
amplifier, as depicted in Fig. 1.5. Power Pin is the power flowing into the amplifier
input over a specified frequency range; whereas Pout is the power flowing out of
the amplifier over a specified frequency. If Pin is only contained in one harmonic
component (i.e., the fundamental frequency), then Pout is the corresponding power
at the fundamental frequency and (Pout )H is the output power of the harmonic com-
ponents generated by nonlinear characteristics of the power amplifier.
The dc power, PDC , is supplied to the active device for the operation of the power
amplifier. The active device also receives the input ac power, Pin , and dissipates a
portion of the combined PDC and Pin as heat, while it converts the remaining into
ac power, PDRF , which gets delivered to the output matching network. The output
network dissipates a portion of PDRF and then delivers the remainder in the form of
Pout and (Pout )H . The optimal design of a power amplifier requires minimization of
these power losses at various stages and maximization of power transfer to the load
with minimal distortion.
The drain efficiency (ηD ) is defined as the ratio between the output power (Pout )
and the dc power (PDC ).
Pout
ηD = (1.2)
PDC
The drain efficiency is good for amplifiers that possess high gain or where the
input power comes at no cost. It is a very useful metric in the evaluation of con-
duction loss independent from the input power dissipation. As the drain efficiency
6 1 Fundamentals

ignores the effect of input power, it helps measure the effectiveness of the amplifier
in avoiding the dissipation on the controlled resistance of the output port [79].
An alternative efficiency metric called total efficiency (ηT ), although rarely used
but more physically significant, is defined as the ratio of the output power and the
sum of all powers fed to the amplifier. From Fig. 1.5, the expression of ηT can be
deduced as:
Pout
ηT = (1.3)
PDC + Pin
The total efficiency can be used to measure the effectiveness of an amplifier in
reducing the need for heat removal, as can be seen in Eq. (1.4), which relates the
total dissipated power, (Pdiss )T , and the total efficiency, ηT [79].
 
1
(Pdiss )T = PDC + Pin − Pout = − 1 Pout (1.4)
ηT
The most commonly used efficiency metric is called PAE which is defined as the
(RF) power added by the amplifier, (Pout − Pin ), divided by the dc power consump-
tion, PDC .
   
Pout − Pin Pout Pin 1
PAE = = 1− = ηD 1 − (1.5)
PDC PDC Pout G
The PAE carries more information than the drain efficiency, as it also depends
on the gain of the amplifier (G). While the drain efficiency increases monotonically
with the input power, the PAE reaches a maximum and then its value decreases until
reaching zero and could even have a negative value.

1.3.2 Intermodulation and Harmonic Distortions

A power amplifier is a nonlinear system that generates harmonic components in ad-

dition to the frequency corresponding to the excitation signal. Intermodulation and
harmonic distortions quantify the impact of distortion associated with the harmonic
components and provide a measure of linearity performance of a power amplifier.
Harmonic distortion is measured when the amplifier is excited with a single-tone
test signal and harmonic distortion components are generated at the output, as de-
picted in Fig. 1.2. Usually the second and third harmonics carry most of the energy;
therefore, the harmonic distortions are defined for these harmonic components.
 
Pout (2f0 )
HD2,dBc = 10 log (1.6)
Pout (f0 )
 
Pout (3f0 )
HD3,dBc = 10 log (1.7)
Pout (f0 )
The harmonic distortions are expressed in dBc relative to the fundamental frequency
power. The power level of the harmonics changes with the input power; therefore,
the corresponding harmonic distortion figures in dBc change as well.
1.3 Figures of Merit 7

Fig. 1.6 Illustration of the

intermodulation products and
the associated frequencies
with two-tone excitation

Another harmonic distortion term that is very commonly used is called total har-
monic distortion (THD) which is given in Eq. (1.8). It includes all the harmonic
distortion components in one figure of merit.
 
n≥2 Pout (nf0 )
THDdBc = 10 log (1.8)
Pout (f0 )

Intermodulation distortion (IMD) is more realistic for the actual wireless com-
munication system. It is the result of two or more signals interacting in a power
amplifier to produce additional unwanted signals. For example, the additional un-
wanted signals (intermodulation products) for two input signals occur at the sum
and difference of integer multiples of the original frequencies given by Eq. (1.17)
and as depicted in Fig. 1.6.

(IMD)products = mf1 ± nf2 (1.9)

where m and n are integers and define the order of intermodulation products as a
sum of m + n.
It is evident from Fig. 1.6 that the two-tone third-order components (2f1 − f2 and
2f2 − f1 ) are most relevant, as they are very close to the fundamental components
and, therefore, cannot be easily filtered. Higher order intermodulation products gen-
erally do not affect the performance of PAs significantly, as these components either
possess very low amplitudes or are far from the fundamental components [11].
Equation (1.10) corresponds to third-order intermodulation product (IMD3)
when the fundamental components f1 and f2 are very close.
   
Pout (2f2 − f1 ) Pout (2f1 − f2 )
IMD3,dBc = 10 log10 ≈ 10 log10 (1.10)
Pout (f2 ) Pout (f1 )

Another metric to describe the linearity performance of a power amplifier is

known as the intercept point. For third-order products, it is known as the third-order
intercept point (IP3), as shown in Fig. 1.7. IP3 is an important parameter and helps
in the estimation of spurious free dynamic range (DRSF ) [80].
It is important to understand that THD, IMD3, and IP3 are good metrics for
describing the performance of PAs exhibiting weak memory effects [16]. However,
these are insufficient for the situations when the PAs exhibit strong nonlinearity.
8 1 Fundamentals

Fig. 1.7 Representation of

third-order intercepts point

Fig. 1.8 Ideal representation

of main and adjacent channel
power spectra and their
respective frequency band
definitions [81], © IEEE 2006

1.3.3 Adjacent Channel Power Ratio

For amplifiers exhibiting strong nonlinearity or for digitally modulated excitations,

adjacent channel power ratio (ACPR) is more relevant considering that harmonic
distortion is applicable for single tone excitation and the usefulness of intermodula-
tion distortion is limited to excitations with specified number of tones (usually 2).
ACPR describes the level of spectral regrowth and is often expressed in dB below
the main carrier power (dBc) as depicted in Fig. 1.8 and is expressed as the ratio of
the power leaking into the adjacent channel to the power in the main channel given
by Eq. (1.11) [17].
  F3 
F2 P (F )dF
ACPRdBc = 10 log  F (1.11)
2
F1 P (F )dF
1.3 Figures of Merit 9

Fig. 1.9 Illustration of error

vector magnitude (EVM) and
its components

The plot shows a representative power spectrum showing the driving signal (be-
tween frequencies F2 and F3 ) and the adjacent channel power (ACP) resulting from
third order interactions only. In an ideal system, ACPR should be as high as possible
as it conveys that the leakage from the main channel to the side channel is low.
Alternatively, a measure of ACPR is the distortion level at a given frequency
offset Foff from the lowest desired frequency (F2 ) [81].

1.3.4 Error Vector Magnitude

According to 3GPP standards, EVM is a measure of the difference between the

reference waveform and the measured waveform [82]. The difference is called error
vector, illustrated in Fig. 1.9, and the EVM, usually mentioned in percentage, is
defined as the square root of the ratio of the mean error vector power to the mean
reference power expressed in Eq. (1.12).

k∈K |Sk − Rk |
2
EVM RMS =  (1.12)
k∈K |Rk |
2

where Sk is the received (measured) vector, RK is the reference symbol vector, and
K is the total number of symbols.
The ACPR provides information about the out-of-band distortion, the error vector
magnitude (EVM) estimates the in-band distortion caused by the amplifier nonlin-
earities. EVM possesses a direct relation with the signal to noise ratio and can be
used to determine the physical error introduced at different stages of communica-
tion system and thus serves as an easily tool for designers in troubleshooting specific
problems. One of the advantages of EVM is the simplicity in their measurement as
it does not require an entire communication system, instead EVM can be calculated
from the measured down-converted digitally modulated radio signal [17].
10 1 Fundamentals

Fig. 1.10 A generic topology

of transconductance amplifier

1.4 Power Amplifier

Transconductance amplifiers, in which the transistor device is operated as a current
source and whose current is dependent on the voltage presented to its input, are the
most common types of power amplifiers. In these amplifiers, the transistor devices
drive a controlled current into a load network which consists of a harmonic filter
and resistive load as depicted in Fig. 1.10. The harmonic filters are designed in such
a way that it possesses high impedance at fundamental frequency and extremely
low impedance at the harmonics. This ensures that the impedance presented to the
device at the fundamental frequency is simply the load resistance, RL , whereas the
harmonics are presented with virtually open circuit. As a result, the voltage across
the device output port remains sinusoidal for any current driven by the device.
It is imperative to understand that performance, such as efficiency, of transcon-
ductance amplifiers can be varied by varying the shape of current waveform con-
sidering that these amplifiers are utilize transistor devices as current source. The
modification or shaping of current waveforms leads to operating classes of power
amplifiers namely class-A, class-B, class-AB, class-C, and class-F.
In class-A, the transistor device is biased in such a way that it conducts current
all the time. In theory, the current through the device should exactly replicate the
shape of the input voltage signal; and, the dc bias current should be sufficient to
ensure that the device remains in conductance region at all times [79]. To satisfy
this requirement and to obtain the optimal performance of the amplifier in class-
A mode, the biasing point (IDD , VDD ) shown in Fig. 1.11 is chosen according to
Eqs. (1.13) and (1.14). This biasing provides plenty of room for the output signal
swing, and hence the chances of reaching cut-off and saturation regions are small.
It is primarily due to this reason that class-A amplifier exhibits very low distortion
in its output.
Imax
IDD = (1.13)
2
Vmax − Vknee Vmax + Vknee
VDD = Vknee + = (1.14)
2 2
1.4 Power Amplifier 11

Fig. 1.11 Load lines for

different loading conditions

The bias point defined by Eqs. (1.13) and (1.14) are also applicable for small-
signal amplifiers. However there is significant difference in the operation of large-
signal (power) amplifier and the small-signal amplifier. In the case of small-signal,
the appropriate matching involves the termination of device output with its complex
conjugate. Under this type of matching the device load line, plotted using Eq. (1.1),
is represented by A1 in Fig. 1.11 [83]. It is apparent that in such a situation the
current swing is less than the maximum device current (Imax ). Similarly for load
impedance with a smaller resistance than the device output resistance, the load line
is the one indicated as A3 in Fig. 1.11 and this leads to a condition where the voltage
swing is smaller as compared to the maximum voltage rating (Vmax ).
In the cases of both A1 and A3, the maximum achievable output power is not
reached, as these load lines do not utilize the full current and voltage swings. How-
ever, for a power amplifier, the goal is to obtain maximum output power from the
device; therefore, the load line must correspond to the one marked as A2 in Fig. 1.11.
For load line A2, the real part of load impedance given by Eq. (1.15) is dependent on
the maximum current and voltage ratings of the device, while the imaginary part of
load impedance is selected in such a way that it cancels the output reactance of the
device. However, this optimal value of load impedance for maximum output power
creates a mismatch between the device and load impedance and has the potential to
cause a high voltage standing wave ratio (VSWR) at the output of the device [18].
Vmax − Vknee
Re(ZL )opt-classA = (Ropt )classA = (1.15)
Imax
The voltage and current at the device output in the class-A operation mode are sinu-
soidal and swing between the minimum and maximum values, which are dependent
on the device’s technology. In such a situation, expressions for the dc power deliv-
ered to the device, the power at the output of device, and the drain efficiency are
given as:
  
Imax Vmax + Vknee
PDC = VDD IDD = (1.16)
2 2
12 1 Fundamentals
 
2 (R )
Iout opt classA (Imax /2)2 Vmax − Vknee
(Pout )classA = =
2 2 Imax
Imax (Vmax − Vknee ) Imax (VDD − Vknee )
= = (1.17)
8 4
(Pout )classA Vmax − Vknee VDD − Vknee
(ηD )classA = = = (1.18)
PDC 2(Vmax + Vknee ) 2VDD
It is apparent from Eqs. (1.16), (1.17), (1.18) that output power and drain efficiency
of class-A PA increases with the square of the output current Iout ; whereas the con-
sumed dc power is constant. For class-A, the optimum values of Pout and efficiency
occurs for the situation when the current Iout reaches its maximum value (Imax /2).
The theoretical maximum drain efficiency of class-A PA reaches 50 % when the
knee voltage (Vknee ) is very small. This low value is contributed by the continuous
consumption of dc power. In order to overcome this issue, and to improve the ef-
ficiency of PAs, bias current (IDD ) is reduced while keeping the same maximum
voltage and current swing. The reduction in the bias current is achieved by reducing
the conduction angle, which forms the basis of reduced conductions angle power
amplifiers [83].
In order to enhance the efficiency of transconductance amplifier, it is essential to
have at least one of the waveforms to be non-sinusoidal. For the replication of this
condition, the voltage waveform is kept same as in class-A while the current wave-
forms are altered such that there are periods in which the device does not conduct.
The PAs in such a situation are called reduced conduction angle power amplifiers
with the assigned terminology of classes AB, B, and C; as summarized in Table 1.1.
The conduction angle (α) is determined by the quiescent gate voltage (VGSQ ), which
is a function of the device pinch-off voltage (Vp ) as well as the device built-in volt-
age (Vbi ).
It is also important to understand that the linearity of PA is affected by the reduc-
tion in the conduction angle. It is essentially due to the fact that the output current
and voltage has smaller swing in the linear region when the conduction angle is re-
duced. The most linear amplifier is obtained for class-A operation and the linearity
worsens while moving from class-A to class-C. An amplifier operating in class-A or
class-B exhibits similar linearity performance if the transconductance of the device
(gm ) remains constant [21]. However, in practical situations this is not normally the
case and as a consequence class-A power amplifier exhibits better linearity.
Even with all the advantages, the reduced conduction angle power amplifiers
with shorted harmonics have some limitations. The main problem of such power
amplifiers lies in the fact that it involves increase in the input RF signal with the
reduction in the conduction angle if the peak current is to be maintained constant. As
a consequence, the gains of such amplifiers get reduced which limit the usefulness
of these techniques in the design of amplifiers with devices possessing high gain.
An alternative solution to overcome the problems caused by harmonics present in
the output of the PA is to provide either an open circuit condition at all harmonics ex-
cept the fundamental frequency for achieving class-E and class-D operation modes
[19, 20, 22, 23] or drain current and voltage shaping for achieving class-F, class-F−1 ,
1.4 Power Amplifier 13

Table 1.1 PA classes based on conduction angle

Class Conduction angle (α) Bias voltage (VGSQ ) Bias current (IQ )

A 2π Vp + 0.5 ∗ (Vbi − Vp ) 0.5 ∗ Imax

AB π–2π Vp + (0 → 0.5) ∗ (Vbi − Vp ) (0 → 0.5) ∗ Imax
B π Vp 0
C 0–π < Vp 0

class-J, and class-J* [23–27] operation modes, in order to obtain increased output
power and drain efficiency.
The concepts of class-F amplifiers and class-B amplifiers are the same with the
only difference being the addition of odd harmonics in the class-F voltage wave-
form. The addition of odd harmonics pushes the class-F voltage waveforms closer
to square shapes. In this mode of operation, the odd harmonics are open-circuited
which allows the harmonic voltages to exist without harmonic currents. The result
of such an exercise is that the presence of odd harmonics flattens the bottom of the
voltage waveform due to abrupt change in the incremental conductivity between the
triode and transconductance regions [79]. In addition, the flattening of the bottom
of the voltage waveforms also flattens the top of the voltage waveforms due to the
symmetry of the odd harmonics [79].
The main advantage of the class-F amplifiers lies in the fact that, in theory, they
can achieve efficiency up to 100 % for the case when all the odd harmonics are open-
circuited to obtain perfect square voltage waveform. In such a situation the amplifier
works like a switch and for this reason class-F amplifiers are termed as switched
mode power amplifiers. Another advantage of class-F is the increased output power,
by almost 27 %, without increasing the peak voltage or peak current considering
that the amplitude of square wave is 4/π times that of the sinusoidal wave [59, 79].
Complexity of class-F power amplifiers is the major limitation due to the fact
that the number of resonators required increases with each harmonics. Keeping this
in the perspective it is rare to see design of class-F amplifiers with more than three
harmonics terminations considering the complexity involved in their designs even
with the availability of synthesized load-pull data. Nevertheless even only three har-
monic tuning is sufficient to obtain high efficiency Class-F and its dual class-F−1
power amplifiers [22, 84].
Switched-mode power amplifiers, such as class-F, operate on the principle that
the transistor device remains out of the transconductance region of simultaneous
high voltage and high current. However, there is always some overlap between the
current and voltage, because only a limited number of harmonics are taken into ac-
count. Switching amplifiers overcome this issue by employing the transistor device
as a switch in such a way that it is either in the high impedance region or in the
high conductivity triode region. The operation of switch (device) in one direction
or the other with load impedance or inverted load impedance regulates the power
amplifiers, namely class-D, inverse class-D, class-E, and inverse class-E [85, 86].
14 1 Fundamentals

As there is no overlap between the current and voltage waveforms in the switching
amplifiers, the theoretical maximum achievable drain efficiency is 100 %.

1.5 Power Amplifier Design Methodologies

Power amplifier (PA) design techniques, which are driven by the availability of mea-
surement systems, modeling approaches and choice of simulation programs, are
commonly categorized into two types. The first category is computer-aided design
(CAD) based methods, which predict the behavior of the PA through the use of de-
vice models; and, the other type is measurement-based methods, where the device
is characterized a priori for relevant design parameters, such as gain, output power
and power-added efficiency, which are then used in the PA design process.

1.5.1 CAD-Based Design Methods

The design of linear microwave passive and active circuits/systems using CAD tools
has become standard design process; whereas, the design of nonlinear microwave
components, such as PAs, are still evolving. Nonetheless, CAD-based PA design is
gaining popularity due to the availability of powerful CAD tools, such as Advanced
Design System (ADS) and Microwave Office (MWO) [28, 29].
In this type of technique, as shown in Fig. 1.12, a model for the specified device
is developed in theory; and, the model extraction procedure is initially identified
[30, 31]. Subsequently, experimental characterization of the specified device is car-
ried out, using large-signal measurement techniques, such as vectorial large-signal
measurement [32], in order to obtain the necessary measurement data required for
model extraction.
By following the selected extraction procedure, the model is extracted and prefer-
ably implemented in a CAD environment [33–36]. The accuracy and reliability of
the model is tested, for which an iterative process is usually needed. The model can
be relied upon once the requisite accuracy in the desired performance is achieved.
PA designers can start CAD-based design by using the model; and, the designed PA
is finalized, if all the design specifications are met at the PA performance testing
stage.
The main advantage of this approach lies in the fact that nonlinear circuits and
systems can be built via models using CAD tools. The CAD tools assist the de-
signers in tackling complex design problems at the outset. For example, nonlinear
simulation algorithms, such as harmonic balance, allow the designers to design the
necessary circuitries around the nonlinear model and test the overall performance of
the system prior to the manufacturing of expensive prototypes.
Furthermore, CAD tools also give the designers a quick, easy and inexpensive
option to go back and restart the design, if a big design flaw is spotted during the
simulation. The use of CAD tools in design and testing stages, therefore, signifi-
cantly increase the chance of a first-pass design.
1.5 Power Amplifier Design Methodologies 15

Fig. 1.12 Flow-chart of

CAD-based PA design
methods

1.5.2 Measurement-Based Design Methods

In these methods, the specified transistor device is first characterized, in order to

obtain key design parameters of the device, such as the gain, efficiency and load-pull
contours. A generic design flow for measurement-based PA is given in Fig. 1.13.
In this type of technique, the measured data forms the primary basis of design;
and, these data are used to build PA prototypes along with the auxiliary circuitries,
such as matching and biasing circuits. The design process is finalized once the de-
signed PA successfully passes the test. The major advantage of this approach is
the time saved by discarding the time-consuming development of device modeling.
This type of design method relies on the accuracy of the measurement system in
generating the data.
This technique used to be extremely risky, but the emergence of accurate large-
signal measurement systems [37–43] has greatly improved the effectiveness of this
design strategy. These measurement systems allow for the capture of detailed key
16 1 Fundamentals

Fig. 1.13 Flow-chart of

measurement-based PA
design methods

characteristics of transistor devices with improved accuracy when the device under
test is driven in realistic working conditions. For example, the large-signal waveform
measurement systems provide the ability to monitor, control and alter the current
and voltage waveforms at device ports, in order to collect the required measurement
data for the design of high-efficiency power amplifiers [44, 45]. Such measurement
systems, therefore, enable first-pass PA design considering that the design is based
on measured data under required operating conditions.

1.6 Nonlinear Microwave Measurement System

The increased demand for enhanced performance from the transistor devices re-
quires them to be commonly operated nearer to compression, firmly in their nonlin-
ear regions. The measurement of devices operating in nonlinear operation requires
large-signal measurement systems (also called nonlinear measurement systems). For
PA applications, large-signal measurement systems [37–43], along with load-pull
test benches [46–52], are essential for accurate measurements of device character-
istics under realistic operating conditions.
1.6 Nonlinear Microwave Measurement System 17

Fig. 1.14 Block diagram

defining load reflection
coefficient, ΓL , generated
using an output tuner at the
load plane [46], diagram ©
IEEE 2011

1.6.1 What Is Load-Pull?

In generic terms, load-pull refers specifically to presenting a priori impedance to a

device under test (DUT) in a precisely controlled fashion, in order to extract the op-
timal performance from the DUT [47]. In load-pull measurements, optimal perfor-
mance is found by varying terminal impedances, along with frequency and bias, so
as to rapidly and accurately establish conditions under which optimal performances
from the specified DUT can be obtained.
For PAs, the optimal loading conditions primarily depend on the nonlinearities
exhibited by transistors. These are significantly different from linear cases, where
the optimal loading conditions are directly identified from S-parameters [18, 19].
In the nonlinear case, load-pull systems determine the appropriate impedance val-
ues experimentally through impedance tuners, while physically the changing load
reflection coefficient, ΓL , as shown in Fig. 1.14, for the extraction of desired design
parameters from transistor devices.
The desired matching impedance, ZL , the incident and reflected traveling waves,
a2 and b2 , at the output port and reflection coefficient, ΓL , are related by expressions
in Eqs. (1.19) and (1.20).
a2
ΓL = (1.19)
b2
ZL − Z0
ΓL = (1.20)
ZL + Z0
where Z0 is the characteristic impedance of the system in which the DUT is going
to be used (normally 50 ).
In summary, a load-pull system includes an active or passive impedance-tuner
[46–52], the controlling mechanism to precisely set the tuner impedance to achieve
desired impedance, and equipment and test set to measure the traveling waves at the
DUT ports.

1.6.2 Why Load-Pull?

There are various ways to identify the reasons for using load-pull based measure-
ments and characterization of high-frequency power transistors. The high-frequency
18 1 Fundamentals

active and passive design requires accurate measurements. Passive elements are as-
sumed linear and, therefore, can be fully defined through frequency dependent S-
parameters [18]; however, the active elements, such as power transistors, are usually
considered nonlinear and cannot be accurately defined through linear S-parameters.
For example, the output current and voltages are nonlinearly related in the case
of active components. The nonlinear dependence gives rise to unwanted spectral
components, harmonics or intermodulation products that were not present in the
excitations.
In the design of PAs, the appropriate terminations at the input and output are ex-
tremely important, as they help in the estimation and determination of the amplifier
behavior and performances. For this reason, it is important to assess the impact of
input and output terminations in the design of PAs. The correct choice of input and
output terminations can optimize the performance of PAs in terms of parameters,
such as the output power, PAE and gain.
Load-pull systems enable rapid, accurate and reliable determination of perfor-
mance parameters of the PAs, as they allow precise and controlled setting of the
terminal impedances. In addition, these systems allow for measurements and char-
acterization of transistor devices, in order to test the linearity of PAs under varying
loading conditions, while subjected to modulated excitation signals [53].
There is significant benefit in using load-pull systems for the determination of op-
timal loading conditions of a microwave device operating in nonlinear mode, such
as mixers [54] and oscillators [55, 56]. In many applications, the terminal loading
conditions at harmonic frequencies may also significantly affect the device perfor-
mance, as proven by both theory [57–59] and experiments [60–70]. In such situa-
tions, harmonic load-pull systems are very effective, as they allow for changing ΓL
values at a discrete set of frequencies (typically two or three) [47, 48].
Another significant application of load-pull systems is in the generation of mea-
surement data that can be utilized in CAD platforms for the development and vali-
dation of nonlinear device models [33–36], which is extremely important for right-
the-first-time PA design.

1.7 Important Load-Pull Features

How does one choose between various types of load-pull tuners [46–52]? What is
most significant factor in choosing a load-pull tuner? The answer is simple, while
tuner repeatability and resolution are important characteristics for all load-pull sys-
tems, the priority of other features depends on how specific tuner characteristics
meet the needs of a desired measurement application.
For example, a passive solid-state load-pull system [71] may be a better choice
for high-speed on-wafer device characterization because of the absence of mechani-
cal vibrations; whereas a passive mechanical load-pull system [72] can be used both
for noise and power characterization of a packaged device, and active load-pull sys-
tems [49–51] can be more appropriate for high-power applications considering their
1.7 Important Load-Pull Features 19

ability to synthesize high reflection coefficients. Therefore, several aspects need to

be looked into before deciding the appropriate tuner for a specified application.

1.7.1 Repeatability of Reflection Coefficients

Measurements using an automated tuner-based load-pull system require pre-

calibrated tuners, and, therefore ability of the tuners to repeat impedance states
is extremely important. Any discrepancy in the synthesized impedance translates to
discrepancy in synthesized matching network or developed nonlinear device model.
The repeatability of the tuner ensures that data obtained from automated load-pull
system could be fully trusted and therefore utilized in the development of accurate
application specific device models and matching circuit design.
Repeatability of the impedance tuner is the difference in the reflection coeffi-
cient between two or more repeated settings of the tuner to the same horizontal and
vertical positions and is measured by a calibrated vector network analyzer (VNA).
The repeatability of impedance can be tested by setting the tuner to a number of
positions around the Smith Chart twice: once to save the measured data in the VNA
memory; and, repeated to retrieve the data from the VNA and compare it with the
memory data. The error in repeatability is computed using the following expression
[73]:
 
(S11 )repeatability = 20. log(S11 )measured − (S11 )memory  (1.21)
where (S11 )measured stands for the last measured S-parameter, and (S11 )memory stands
for the S-parameter in memory.
The repeatability is mentioned in terms of dB. For example, commercially avail-
able passive-mechanical tuners and passive solid-state tuners typically exhibit better
than −60 dB and −70 dB repeatability, up to 50 GHz, respectively [74]. Active load-
pull systems [49–51] generally do not exhibit good repeatability with the exception
of the active envelope load-pull system, which possesses good repeatability around
−55 dB [75].

1.7.2 Tuning Range and Its Distribution

The tuning range of the tuner, often called the ability to establish a reflection envi-
ronment, refers to the range of reflection factors that a tuner can present. In principle,
this aspect is more relevant for the characterization of high-power devices, where
extremely low output impedances are often encountered. Usually, passive tuners are
very good for low- or medium-power devices where the desired reflection factor
is closer to the center of the Smith chart. However, passive tuners possess limited
ability to synthesize reflection coefficients close to the border of the Smith chart
[47, 48] and, therefore, find limited usefulness in applications, such as high-power
device characterization or harmonic load-pull measurements.
20 1 Fundamentals

On the other hand, active load tuners do not suffer from tuning range limita-
tions and, thus, are more appropriate for the measurement and characterization of
high-power devices and harmonic load-pull applications [46]. Active tuners can also
synthesize reflection coefficients outside the Smith chart boundary and, as a result,
find applications in the design of oscillators.
The tuning range distribution refers to coverage of the generated reflection fac-
tors over the Smith chart and the control of these reflection points in a precise man-
ner within the desired region of the Smith chart [74]. The pre-matched load-tuners
[47, 48] and active envelope load-pull [75] systems can be precisely controlled and
employed in applications where an accurate reflection environment is required in a
particular region of the Smith chart.

1.7.3 Tuning Speed

Tuning speed is the time taken by a tuner to move from one impedance state to
the next. This feature is very important for applications where high measurement
throughput is required. In addition to the tuner’s speed, the total measurement time
also includes time taken by the measurement equipments at each impedance position
while the measurements are being made [74]. Passive tuners are good for high yield
and fast measurement throughput applications.

1.7.4 Power Handling Capability

Power handling is an extremely important feature of any load-pull setup. It is de-

scribed as the maximum root mean square (rms) or peak power that can be delivered
to a tuner without any appreciable change in tuner impedance or any damage to the
tuner [74]. In practical applications, a high tuner insertion loss causes heating of
tuner elements which can lead to variation in the calibrated impedance seen by the
DUT; whereas an improperly designed tuner can exhibit corona discharge which has
the potential to damage the tuner or the DUT [76]. A passive tuner with high power
handling capability overcomes these problems [73, 74], and carefully configured
active load-pull setups with high power handling ability can often meat application
requirements [76].

1.7.5 Tuner Resolution

Tuner resolution, which refers to the resolution of impedance points synthesizable

by the tuner, is an important parameter for power transistor characterization as pa-
rameters such as PAE and output power of these devices are highly sensitive to even
1.7 Important Load-Pull Features 21

Fig. 1.15 A single probe

tuner, photo © Focus
Microwaves Inc.

slight variation in impedance states. Therefore a tuner exhibiting high resolution is

considered extremely important in precision measurements. Any standard passive-
mechanical tuners typically synthesize 10,000 points, which can be enhanced to
millions of points using cascaded tuners or interpolation [77]. However, in such
cases, the tuning accuracy is fully dependent on the accuracy of interpolation [77].
Furthermore, high resolution tuners increase the measurement time considering the
time required in covering all the impedance states.

1.7.6 Tuner Bandwidth

Tuner bandwidth is often interchangeably referred to both the frequency range of the
tuner and the instantaneous bandwidth of the tuner. The bandwidth of the tuner is the
frequency range over which the tuner is able to synthesize the specified impedances,
whereas the instantaneous bandwidth, or the modulation bandwidth, of the tuner is
the frequency range over which the group delay is constant. The standard passive
tuners are able to operate successfully from low MHz to 40 GHz, but are unable
to synthesize constant reflection coefficients under modulated excitations. Active
load-pull systems have potential to provide constant reflection coefficients over a
modulation bandwidth.

1.7.7 Tuner Size

Tuner size is extremely important in on-wafer high-speed device characterization

where in situ calibration is used; smaller size in such situations are required in order
to avoid acoustic vibrations and mechanical vibrations [72]. It is a standard practice
to perform a vibration test on the load-pull systems, even if utilizing small sized
tuners, in such on-wafer high-speed environments.
22 1 Fundamentals

Fig. 1.16 Setup diagram of a

fully functional passive
load-pull system, photo ©
Focus Microwaves Inc.

Fig. 1.17 An active load-pull setup (ALPS) for high frequency device characterization, photo ©
Focus Microwaves Inc.

1.8 Common Load-Pull Systems

Over the years, the configurations of load-pull systems have evolved; however, they
still fall into two categories, namely passive and active load-pull systems. They are
categorized and distinguished based on the fashion in which they synthesize the load
impedance.
In passive techniques, the desired impedance is synthesized by varying the re-
flection coefficient of the impedance controlling element, such as the single probe
tuner as depicted in Fig. 1.15; and, the measurement is carried out using a typical
setup, as shown in Fig. 1.16.
As mentioned earlier, in the passive technique, the reflection coefficient is
changed through the setting of the tuner, whereas the active load-pull approach syn-
thesizes the impedance environment by injecting a signal at the DUT port. A typical
setup of an active load-pull setup is depicted in Fig. 1.17. Once again, it is impor-
tant to mention that both active and passive techniques have unique benefits and,
therefore, have advantages in specific measurement applications. Passive load-pull
techniques are mostly needed for swept drive, frequency and bias applications; and,
active load-pull systems are more appropriate for applications requiring highly re-
flective environments.
References 23

References

1. H. Chireix, High power outphasing modulation. Proc. Inst. Radio Eng. 23(II), 1370–1392
(1935)
2. W.H. Doherty, A new high efficiency power amplifier for modulated waves. Proc. Inst. Radio
Eng. 24, 1163–1182 (1936)
3. L. Kahn, Single sideband transmission by envelope elimination and restoration. Proc. Inst.
Radio Eng. 40, 803–806 (1952)
4. D.C. Cox, Linear amplification with nonlinear components, IEEE Trans. Commun. 22, 1942–
1945 (1974)
5. F.H. Raab, Efficiency of Doherty RF power amplifier system. IEEE Trans. Broadcast. BC-
33(3), 77–83 (1983)
6. S. Bousnina, F.M. Ghannouchi, Analysis and experimental study of an L-band new topology
Doherty amplifier. in 2001 IEEE Microwave Theory and Techniques Society’s International
Microwave Symposium Digest, Phoenix, USA, vol. 2 (May 2001), pp. 935–938
7. B. Kim, J. Kim, I. Kim, J. Cha, The Doherty amplifier. IEEE Microw. Mag. 7(5), 42–50
(2006)
8. M. Helaoui, F.M. Ghannouchi, Linearization of power amplifiers using the reverse Mm-linc
technique. IEEE Trans. Circuits Syst. II, Express Briefs 57(1), 6–10 (2010)
9. S.C. Jung, O. Hammi, F.M. Ghannouchi, Design optimization and DPD linearization of GaN
based unsymmetrical Doherty power amplifiers for 3G multi-carrier applications. IEEE Trans.
Microw. Theory Tech. 57(9), 2105–2113 (2009)
10. R. Darraji, F.M. Ghannouchi, O. Hammi, A dual-input digitally driven Doherty amplifier ar-
chitecture for performance enhancement of Doherty transmitters. IEEE Trans. Microw. Theory
Tech. 59(5), 1284–1293 (2011)
11. S.A. Maas, Nonlinear Microwave and RF Circuits, 2nd edn. (Artech House, Norwood, 2003)
12. J. Wood, D.E. Root (eds.), Fundamentals of Nonlinear Behavioural Modelling for RF and
Microwave Design (Artech House, Norwood, 2005)
13. D.E. Root, J. Verspecht, D. Sharritt, J. Wood, A. Cognata, Broad-band poly-harmonic dis-
tortion behavioral models from fast automated simulations and large-signal vectorial network
measurements. IEEE Trans. Microw. Theory Tech. 53(11), 3656–3664 (2005)
14. J. Verspecht, D.E. Root, Poly-harmonic distortion modeling. IEEE Microw. Mag. 7(3), 44–57
(2006)
15. J. Verspecht, M. Vanden Bossche, F. Verbeyst, Characterizing components under large signal
excitation: defining sensible ‘large signal S-parameters, in 49th IEEE ARFTG Conference,
Denver, USA (June 1997), pp. 109–117
16. S.C. Cripps, RF Power Amplifiers for Wireless Communications, 2nd edn. (Artech House,
Norwood, 2006)
17. R. Gilmore, L. Besser, Practical RF Circuit Design for Modern Wireless Systems, vol. II
(Artech House, Norwood, 2003)
18. D.M. Pozar, Microwave Engineering, 3rd edn. (Wiley, New York, 2005). ISBN 0-471-17096-
8
19. G. Gonzalez, Microwave Amplifier Design, 2nd edn. (Prentice Hall, New York, 1996)
20. L.J. Kushner, Output performance of idealized microwave power amplifiers. Microw. J.
32(10), 103–116 (1989)
21. P. Colantonio, J.A. Garcia, F. Giannini, E. Limiti, E. Malaver, J.C. Pedro, High linearity and
efficiency microwave PAs, in European Gallium Arsenide and Other Semiconductor Applica-
tion Symposium (2004), pp. 183–186
22. F.H. Raab, Class E, class C, and class F power amplifiers based upon a finite number of
harmonics. IEEE Trans. Microw. Theory Tech. 49(8), 1462–1468 (2001)
23. J.H. Jeong, H.H. Seong, J.H. Yi, G.H. Cho, A class D switching power amplifier with high effi-
ciency and wide bandwidth by dual feedback loops, in International Conference on Consumer
Electronics, Rosemont, USA (June 1995), pp. 428–429
24 1 Fundamentals

24. J. Staudinger, Multiharmonic load termination effects on GaAs power amplifiers. Microw. J.,
60–77 (1996)
25. F.H. Raab, Maximum efficiency and output of class-F power amplifiers. IEEE Trans. Microw.
Theory Tech. 49(6), 1162–1166 (2001)
26. H. Kim, G. Choi, J. Choi, A high efficiency inverse class-F power amplifier using GaN HEMT.
Microw. Opt. Technol. Lett. 50(9), 2420–2422 (2008)
27. A. Ramadan, T. Reveyrand, A. Martin, J.-M. Nebus, P. Bouysse, L. Lapierre, J.-F. Villemazet,
S. Forestier, Two-stage GaN HEMT amplifier with gate-source voltage shaping for efficiency
versus bandwidth enhancements. IEEE Trans. Microw. Theory Tech. 59(3), 699–706 (2011)
28. Advanced Design System (ADS), Version 2009, Agilent Technologies: http://www.agilent.
com
29. Microwave Wave Office (MWO), Version 2010, AWR Corp: http://www.awrcorp.com
30. Y. Tsividis, Operation and Modeling of the MOS Transistor (McGraw-Hill, New York, 1987)
31. C.M. Snowden, Nonlinear modelling of power FETs and HBTs, in 3rd International Workshop
on Integrated Nonlinear Microwave and Millimeterwave Circuits, Duisberg, Germany (Oct.
1994), pp. 11–25
32. J. Verspecht, P. Van Esch, Accurately characterizing hard nonlinear behavior of microwave
components with the nonlinear network measurement system: introducing ‘nonlinear scatter-
ing functions’, in 5th International Workshop on Integrated Nonlinear Microwave and Mil-
limeterwave Circuits, Duisberg, Germany (Oct. 1998), pp. 17–26
33. D. Schreurs, J. Verspecht, S. Vandenberghe, G. Carchon, K. van der Zanden, B. Nauwelaers,
Easy and accurate empirical transistor model parameter estimation from vectorial large-signal
measurements, in IEEE Microwave Theory and Techniques Society’s International Microwave
Symposium Digest, Anaheim, USA (June 1999), pp. 753–756
34. A. Issaoun, R. Barrak, A.B. Kouki, F.M. Ghannouchi, C. Akyel, Enhanced empirical large-
signal model for HBTs with performance comparable with physics-based models. IEE Proc.
Sci. Meas. Technol. 151(3), 142–150 (2004)
35. M. Prigent, J.C. Nallatamby, M. Camiade, J.M. Nebur, E. Ngoya, R. Quere, J. Obregon, Com-
prehensive approach to the nonlinear design and modelling of microwave circuits, in IEEE
Signals, Systems, and Electronics Conference, Pisa, Italy (Oct. 1998), pp. 450–455
36. H. Qi, J. Benedikt, P. Tasker, A novel approach for effective import of nonlinear device char-
acteristics into CAD for large signal power amplifier design, in IEEE Microwave Theory and
Techniques Society’s International Microwave Symposium Digest, San Francisco, USA (June
2006), pp. 477–480
37. J.G. Lecky, A.D. Patterson, J.A.C. Stewart, A vector corrected waveform and load line mea-
surement system for large signal transistor characterization, in IEEE Microwave Theory and
Techniques Society’s International Microwave Symposium Digest, Orlando, USA (May 1995),
pp. 1243–1246
38. D.J. Williams, P.J. Tasker, An automated active source- and load-pull measurement system, in
6th IEEE High Frequency Student Colloquium, Cardiff, UK (Sept. 2001), pp. 7–12
39. S. Bensmida, P. Poire, F.M. Ghannouchi, Source-pull/load-pull measurement system based on
RF and baseband coherent active branches using broadband six-port reflectometers, in 37th
European Microwave Conference, Munich, Germany (Oct. 2007), pp. 953–956
40. W.S. El-Deeb, M.S. Hashmi, N. Boulejfen, F.M. Ghannouchi, Small-signal, complex dis-
tortion and waveform measurement system for multiport microwave devices. IEEE Instrum.
Meas. Mag. 14(3), 28–33 (2011)
41. J. Benedikt, R. Gaddi, P.J. Tasker, M. Goss, High-power time-domain measurement system
with active harmonic load-pull for high-efficiency base-station amplifier design. IEEE Trans.
Microw. Theory Tech. 48(12), 2617–2624 (2000)
42. D. Barataud, C. Arnaud, B. Thibaud, M. Campovecchio, J.-M. Nebus, J.P. Villotte, Measure-
ments of time-domain voltage/current waveforms at RF and microwave frequencies based on
the use of a vector network analyzer for the characterization of nonlinear devices – appli-
cation to high-efficiency power amplifiers and frequency multipliers. IEEE Trans. Instrum.
Meas. 47(5), 1259–1264 (1998)
References 25

43. F. van Raay, G. Kompa, A new on-wafer large signal waveform measurement system with
40 GHz harmonic bandwidth, in IEEE Microwave Theory and Techniques Society’s Interna-
tional Microwave Symposium Digest, Albuquerque, USA (June 1992), pp. 1435–1438
44. H.M. Nemati, A.L. Clarke, S.C. Cripps, J. Benedikt, P.J. Tasker, C. Fager, J. Grahn, H. Zirath,
Evaluation of a GaN HEMT transistor for load- and supply-modulation applications using
intrinsic waveform measurements, in IEEE Microwave Theory and Techniques Society’s In-
ternational Microwave Symposium Digest, Anaheim, USA (May 2010), pp. 509–512
45. Y.Y. Woo, Y. Yang, B. Kim, Analysis and experiments for high efficiency class-F and inverse
class-F power amplifiers. IEEE Trans. Microw. Theory Tech. 54(5), 1969–1974 (2006)
46. M.S. Hashmi, F.M. Ghannouchi, P.J. Tasker, K. Rawat, Highly reflective load-pull. IEEE Mi-
crow. Mag. 11(4), 96–107 (2011)
47. Maury Microwave Corporation, Device characterization with harmonic load and source pull,
Application Note: 5C-044, Dec. 2000
48. Focus Microwave, Load pull measurements on transistors with harmonic impedance control,
Technical Note, Aug. 1999
49. G.P. Bava, U. Pisani, V. Pozzolo, Active load technique for load-pull characterization at mi-
crowave frequencies. IEE Electron. Lett. 18(4), 178–180 (1982)
50. Y. Takayama, A new load pull characterization method for microwave power transistors, in
IEEE Microwave Theory and Techniques Society’s International Microwave Symposium Di-
gest, New Jersey, USA (June 1976), pp. 218–220
51. F.M. Ghannouchi, F. Beauregard, A.B. Kouki, Power added efficiency and gain improvement
in MESFETs amplifiers using an active harmonic loading technique. Microw. Opt. Technol.
Lett. 7(13), 625–627 (1994)
52. F.M. Ghannouchi, M.S. Hashmi, S. Bensmida, M. Helaoui, Enhanced loop passive source-
and load-pull architecture for high reflection factor synthesis. IEEE Trans. Microw. Theory
Tech. 58(11), 2952–2959 (2010)
53. B. Noori, P. Hart, J. Wood, P.H. Aaen, M. Guyonnet, M. Lefevre, J.A. Pla, J. Jones, Load-pull
measurements using modulated signals, in 36th European Microwave Conference, Manch-
ester, UK (Sept. 2006), pp. 1594–1597
54. D.-L. Le, F.M. Ghannouchi, Multitone characterization and design of fet resistive mixers
based on combined active source-pull/load-pull techniques. IEEE Trans. Microw. Theory
Tech. 46(9), 1201–1208 (1998)
55. F.M. Ghannouchi, Device line simulation of high-speed oscillators using harmonic-balance
techniques. Arab. J. Sci. Eng. 19(4B), 805–812 (1994)
56. P. Berini, M. Desgagne, F.M. Ghannouchi, R.G. Bosisio, An experimental study of the effects
of harmonic loading on microwave mesfet oscillators and amplifiers. IEEE Trans. Microw.
Theory Tech. 42(6), 943–950 (1994)
57. D.M. Snider, A theoretical analysis and experimental confirmation of the optimally loaded
and overdriven RF power amplifier. IEEE Trans. Electron Devices 14(12), 851–857 (1967)
58. J.D. Rhodes, Output universality in maximum efficiency linear power amplifiers. Int. J. Circuit
Theory Appl. 31, 385–405 (2003)
59. F.H. Raab, Class-F power amplifiers with maximally flat waveforms. IEEE Trans. Microw.
Theory Tech. 31(11), 2007–2012 (1997)
60. R. Negra, F.M. Ghannouchi, W. Bachtold, Analysis and evaluation of harmonic suppression
of lumped-element networks suitable for high-frequency class-E operation condition approx-
imation. IET Microw. Antennas Propag. 2(8), 794–800 (2008)
61. M. Helaoui, F.M. Ghannouchi, Optimizing losses in distributed multiharmonic matching net-
works applied to the design of an RF GaN power amplifier with higher than 80 % power-added
efficiency. IEEE Trans. Microw. Theory Tech. 57(2), 314–322 (2009)
62. F.M. Ghannouchi, M.S. Hashmi, Experimental investigation of the uncontrolled higher har-
monic impedances effect on the performance of high-power microwave devices. Microw. Opt.
Technol. Lett. 52(11), 2480–2482 (2010)
26 1 Fundamentals

63. G. Zhao, S. El-Rabaie, F.M. Ghannouchi, The effects of biasing and harmonic loading on
MESFET tripler performance. Microw. Opt. Technol. Lett. 9(4), 189–194 (1995)
64. F.M. Ghannouchi, F. Beauregard, A.B. Kouki, Power added efficiency and gain improvement
in MESFETs amplifiers using an active harmonic loading technique. Microw. Opt. Technol.
Lett. 7(13), 625–627 (1994)
65. P. Wright, J. Lees, J. Benedikt, P.J. Tasker, S.C. Cripps, A methodology for realizing high
efficiency class-J in a linear and broadband PA. IEEE Trans. Microw. Theory Tech. 57(12),
3196–3204 (2009)
66. M.S. Hashmi, A.L. Clarke, S.P. Woodington, J. Lees, J. Benedikt, P.J. Tasker, Electronic multi-
harmonic load-pull system for experimentally driven power amplifier design optimization,
in IEEE Microwave Theory and Techniques Society’s International Microwave Symposium
Digest, Boston, USA, vols. 1–3 (June 2009), pp. 1549–1552
67. E. Bergeault, O. Gibrat, S. Bensmida, B. Huyart, Multiharmonic source-pull/load-pull active
setup based on six-port reflectometers: influence of the second harmonic source impedance on
RF performances of power transistors. IEEE Trans. Microw. Theory Tech. 52(4), 1118–1124
(2004)
68. G.R. Simpson, M. Vassar, Importance of 2nd harmonic tuning for power amplifier design, in
48th Automatic Radio Frequency Techniques Group Conference, Florida, USA (Dec. 1996),
pp. 1–6
69. P. Butterworth, S. Gao, S.F. Ooi, A. Sambell, High-efficiency class-F power amplifier with
broadband performance. Microw. Opt. Technol. Lett. 44(3), 243–247 (2005)
70. G. Collinson, C.W. Suckling, Effects of harmonic terminations on power and efficiency of
GaAs HBT power amplifiers at 900 MHz, in IEE Colloquium on Solid-State Power Amplifiers
(Dec. 1991), pp. 12/1–12/5
71. Maury Microwave Corporation, LP series electronic tuner system, Technical Data, 4T-081,
2002
72. Focus Microwave, Mechanical vibrations of CCMT tuners used in on-wafer load-pull testing,
Application Note AN-46, Oct. 2001
73. Focus Microwave, Comparing tuner repeatability, Application Note AN-49, March 2002
74. Maury Microwave Corporation, Introduction to tuner-based measurement and characteriza-
tion, Technical Data, 5C-054, Aug. 2004
75. M.S. Hashmi, A.L. Clarke, S.P. Wooding, J. Lees, J. Benedikt, P.J. Tasker, An accurate
calibrate-able multi-harmonic active load-pull system based on the envelope load-pull con-
cept. IEEE Trans. Microw. Theory Tech. 58(3), 656–664 (2010)
76. Z. Aboush, J. Lees, J. Benedikt, P. Tasker, Active harmonic load-pull system for characterizing
highly mismatched high power transistors, in IEEE Microwave Theory and Techniques Soci-
ety’s International Microwave Symposium Digest, Long Beach, USA (June 2005), pp. 1311–
1314
77. Focus Microwave, High resolution tuners eliminate load-pull performance errors, Application
Note AN-15, Jan. 1995
78. M. Salib, D. Dawson, H. Hahn, Load-line analysis in the frequency domain with distributed
amplifier design examples, in IEEE Microwave Theory and Techniques Society’s International
Microwave Symposium Digest, Palo Alto, USA (June 1987), pp. 575–578
79. S. Kee, The class E/F family of harmonic-tuned switching power amplifiers, PhD Thesis,
California Institute of Technology, 2002
80. Anritsu, Intermodulation distortion measurements using the 37300 series vector network ana-
lyzer, Application Note-11410-00213, Sept. 2000
81. R.H. Caverly, J.C.P. Jones, Contributions to adjacent channel power in microwave and wireless
systems by PIN diodes, in IEEE Microwave Theory and Techniques Society’s International
Microwave Symposium Digest, San Francisco, USA (June 2006), pp. 910–913
82. 3GPP Technical Specification 25141, Base Station Conformance Testing (FDD)
83. M.A.Y. Medina, RF power amplifiers for wireless communications, PhD Thesis, Katholieke
Universiteit Leuven, 2008
References 27

84. B. Ingruber, J. Baumgartner, D. Smely, M. Wachutka, G. Magerl, F.A. Petz, Rectangularly

driven class-A harmonic-control amplifier. IEEE Trans. Microw. Theory Tech. 46(11), 1667–
1672 (1998)
85. S. El-Hamamsy, Design of high efficiency RF class-D power amplifier. IEEE Trans. Power
Electron. 9(3), 297–308 (1994)
86. M.J. Chudobiak, The use of parasitic nonlinear capacitors in class-E amplifiers. IEEE Trans.
Circuits Syst. I, Fundam. Theory Appl. 14(12), 941–944 (1994)
Chapter 2
Passive Load-Pull Systems

In general, a passive load-pull system is built around a passive tuner. The tuner is
used in combination with peripheral equipment and components, such as a vector
network analyzer (VNA), signal generators, power meters, bias tees, isolators, for
achieving the load-pull functionality. Primarily two types of passive tuners, namely
the electromechanical tuner (EMT) and the electronic tuner (ETS), are employed in
load-pull systems. It is imperative to understand the limitations of both these tuners
so that the best solution can be employed for any specified application. The initial
sections of this chapter are focused on the explanation and comparison of EMT and
ETS based passive load-pull systems.
The major challenge in setting up a passive load-pull setup involves calibration,
which is carried out to remove errors arising from mismatches, dispersions and im-
perfections in the cables and system components. A second aspect of calibration
involves shifting of the measurement planes, considering that the actual measure-
ment takes place away from the device under test (DUT) ports. Later sections of
this chapter address the calibration aspect in detail.

2.1 Introduction
An impedance tuner is the main engine of any passive load-pull system [1–4]. The
tuner is placed between the DUT and power meter, as shown in Fig. 2.1, for the
search of optimal matching conditions for given targeted performance. The synthe-
sis of optimal impedance involves changing of the tuner setting and then measuring
subsequently the output power, Pout , using the power meter. The optimal impedance
synthesized by the impedance tuner corresponds to the maximum reading of Pout by
the power meter, according to the maximum power transfer theorem [5]. A separate
power meter measures the power injected, Pin , by the source at the input of DUT.
Overall, setting of the impedance tuner directly provides the matching reflection
coefficient, as given in Eq. (2.1), while readings of the power meters at the input and
output relates the incident and reflected traveling waves, as given in Eqs. (2.2) and
(2.3), at the respective DUT ports.

F.M. Ghannouchi, M.S. Hashmi, Load-Pull Techniques with Applications to Power 29

Amplifier Design, Springer Series in Advanced Microelectronics 32
DOI 10.1007/978-94-007-4461-5_2, © Springer Science+Business Media Dordrecht 2013
30 2 Passive Load-Pull Systems

Fig. 2.1 Typical

representation of a passive
load-pull system

a2
Γout = (2.1)
b2
 
Pin = |a1 |2 − |b1 |2 = |a1 |2 1 − |Γin |2 (2.2)
 
Pout = |b2 |2 − |a2 |2 = |b2 |2 1 − |Γout |2 (2.3)
where, Γout and Γin are the reflection coefficients at the output and input ports of the
DUT, respectively; and, a and b with subscripts represent the incident and reflected
waves at the respective ports.
The optimal values of Eqs. (2.1), (2.2), (2.3) enable the determination of design
parameters such as gain, efficiency and power-added efficiency (PAE). The determi-
nation of the optimal values and, subsequently, the design parameters is an iterative
process and requires numerous changes in the configuration of the load-pull tuner.

2.2 Passive Load-Pull System

A passive load-pull system typically employs either electromechanical tuners
(EMTs) that rely on horizontal and vertical movement of probes along a transmis-
sion line [6] or electronic tuners (ETSs), which rely on the appropriate electronic
circuits with electronically changing matching properties [7]. EMTs are stub, slug
or slide screw type [8, 9], whereas ETSs are mostly solid state and employ PIN
diodes [10, 11].

2.2.1 Electromechanical Tuner (EMT)

These tuners allow precise positioning of probes/stubs/slugs in a slotted transmis-

sion line, in order to generate repeatable complex microwave reflection factors.
Movement of the probes changes the impedance by changing the parallel suscep-
tance. The probes are called mismatched elements and introduce mismatch through
their movement in the horizontal and vertical directions. The movement of a probe in
the vertical direction alters mainly the magnitude of the mismatch, while its move-
ment in the horizontal direction alters mainly the phase of the mismatch.
A pictorial presentation of working principle of single-stub EMT is depicted in
Fig. 2.2 [12]. In Fig. 2.2(a), the stub is outside the electric fields of the traveling wave
propagating in the central conductor (transmission line) and, therefore, does not
affect the impedance at the reference plane (usually set at 50 ). However, as soon
as the probe moves vertically downward towards the central conductor (transmission
line) the magnitude of the reflection factor at the reference plane increases, as can
2.2 Passive Load-Pull System 31

Fig. 2.2 (a) A generic representation of an EMT in a matched state, (b) the movement of a
stub/probe/slug in the vertical direction changes the magnitude of the reflection factor, (c) the
movement of a stub/probe/slug in the horizontal direction alters the phase of the reflection factor
[12]; © IEEE 2011

be seen in Fig. 2.2(b). The movement of the probe in the horizontal direction along
the central conductor varies the phase of the reflection coefficient, as evident in
Fig. 2.2(c).
In principle, in any EMT, the magnitude of the impedance mismatch at the ref-
erence plane is determined by the probe position (depth); and, the phase of the
impedance mismatch at the reference plane is determined by the carriage position
across the length of the tuner length. An EMT is considered good if the resolution
32 2 Passive Load-Pull Systems

Fig. 2.3 A single probe

tuner, photo © Focus
Microwaves Inc.

of the movements of the stub/slug/probe in the horizontal and vertical directions is

incrementally small in range of tens of microns [13]. A good EMT ensures that the
desired reflection coefficient can be synthesized on a dense grid. An example of a
commercial EMT (in this case, a single probe tuner) is given in Fig. 2.3.
The slide screw type of EMT consists of a 50  coax- or slab-line with a probe
with a nominal minimum length of λ/4 at its lowest frequency of operation and
has two degrees of freedom [14, 15]. One movement is up and down towards the
center conductor. This way the probe forms an adjustable shunt discontinuity with
the 50  transmission line. The other movement is the traveling of the probe along
the 50  mainline.
In stub tuners, two or more parallel sliding shorts (stubs) are placed on a 50 
line. When one is placed at the DUT reference plane and another λ/8 away from it,
they independently control the real and imaginary parts of the admittance [6]. For a
reflection factor (ΓL ) setting of 0, each stub has to be set to precisely nλ/8. A very
high voltage standing wave ratio (VSWR) can be reached with a stub tuner.
The slug tuner is a construction with a 50  coax- or slab-line, where two mov-
able slugs can slide over the center conductor. The slug reduces the characteristic
impedance locally to typically 10–15 , while an electrical length of λ/8 for the
transmission line gives maximum tuning range. In slug tuners, setting maximum
ΓL (|ΓL |) is easier, and losses tend to be constant over the tuning range, compared
to other EMT types [6].
Overall, in a coaxial environment, the frequencies at which EMT can be used are
around 200 MHz to 50 GHz in different multi-octave bands and tuner types. EMTs
that use waveguides cover standard WR bands from 26.5 to 110 GHz [13].
In any automatic load-pull system based around an EMT, the movements of the
probes in the horizontal and vertical directions are achieved through stepper motors;
and, the actual positioning of the probe are monitored and controlled by timing belts.
The use of timing belt to control axis positioning minimizes the vibrations translated
from the stepper motors to the axis and then the probes [3] and, thus, makes these
systems useful for on-wafer applications. In most cases, EMTs use the same vertical
anti-backlash mechanism with a resolution of 0.75 or 1.5 µm per motor step; and,
2.2 Passive Load-Pull System 33

Fig. 2.4 A generic diagram

of a 6-diode based electronic
tuner [10], © IEEE 1982

the horizontal step size varies between 1.25 and 25 µm, depending on the frequency
of operation, in order to obtain an optimal tuning speed/resolution ratio [13].

2.2.2 Electronic Tuner (ETS)

An ETS is an electronic circuit with the ability to electronically change matching

properties. These electronic circuits can be based on either varactor or PIN diodes.
The latter is more common for load-pull applications, due to the ability to handle
higher power compared to varactor-based ETS. The diode-based ETS synthesize
reflection coefficient by varying the impedance state of a number of PIN diodes
connected in parallel and placed in a precise manner along a transmission line, an
example of which is shown in Fig. 2.4.
In such a configuration, each diode is capable of generating a continuous discon-
tinuity as a spoke of a wheel of increasing |ΓL |, while other diodes can subsequently
change the angle of ΓL discretely. The ETS in Fig. 2.4 consists of directional cou-
plers, six PIN diodes, and external loads. This six-diode ETS can produce a total
of sixty-four (26 ) impedance states, corresponding to two states of each diode in-
dependently. Depending on the design, the diode impedance state can be varied
continuously or toggled discretely between on and off states [10].
Essentially, a single ETS is limited in tuning resolution; therefore, a number of
ETSs are cascaded together for applications requiring higher tuning resolution. The
cascaded configuration is transparent to the user and, therefore, behaves as one tuner
unit. Passive ETSs are generally quite lossy and, therefore, can usually achieve
VSWRs up to 10 [7]. However, their miniature size and light weight make them
appropriate for on-wafer measurements.
34 2 Passive Load-Pull Systems

Table 2.1 Comparison between ETS and EMT [16]

Characteristics ETS EMT

Reflection Factor, Γ Noise: O ++

Load Pull: -
Number of Impedances O ++
Insertion Loss - +/ + +
Tuning Resolution Noise: O ++
Load Pull: –
Maximum Power O/-/– ++
Depending on DUT
Frequency Bandwidth O ++
Spurious Oscillations O/- ++
On Wafer Operation ++ O
Tuner Size ++ On Wafer: -
Test Fixture: ++
Tuner Speed ++ -
Test Total Speed + +
Tuner Linearity O/- ++
Depending on DUT
DSB Noise Measurement – +
Temperature Drift - ++

Legend: ++ Excellent; + Good; O Acceptable; - Poor; – Unacceptable for certain tasks

2.2.3 ETS and EMT Comparisons

Load-pull systems are employed in diverse applications, such as the design of power
amplifiers and oscillators and noise measurements,. As a consequence it is imper-
ative to understand the features and limitations of ETSs and EMTs. Table 2.1 pro-
vides a comparison between EMT and ETS based on the most common load-pull
parameters and applications [16].
ETSs consist of sets of PIN diodes, mounted in microstrip circuits, having only
two states, i.e., on and off. Due to the physical distribution along microstrip lines,
ETSs generate irregular shapes of reflection factors with values up to 0.8 over a lim-
ited frequency range [16]. An EMT can generate a reflection factor (Γ ) of 0.75 in
standard form and up to 0.92 in a pre-matched or cascaded state from low frequen-
cies up to millimeter waves [17, 18].
ETSs are faster and can change states from impedance to impedance within mil-
liseconds, whereas EMTs take at least a few seconds to tune from one state to an-
other. However, it should be noted that ETS can save only around 10 % of the
measurement time compared to that of EMT during the complete load-pull mea-
surements, due to the time needed to read the instrumental setup via general pur-
pose interface bus (GPIB) [16]. ETSs are also miniature, compact and substantially
2.3 Load-Pull Measurement 35

smaller compared to EMTs, making the ETS more suitable for applications prone to
mechanical vibrations, such as on-wafer device characterization and measurement.
The insertion loss of an ETS at the operating reflection factor is very high, due
to lossy microstrip lines. This requires high-power driver amplifiers at the source
side of the setup, which in turn raises the question of linearity of the PIN diodes and
the temperature drift in the ETS. An ETS operating loss of 12 dB at Γ = 0.8 is a
common observable fact, this results in raising the cost of the setup considerably,
whereas an EMT has only a few tenths of a dB at this Γ level [16].
The ETS has unpredictable impedance behavior at low frequencies, whereas the
EMT is low pass and presents 50  to the DUT. The ETS behavior is a high risk
factor for uncontrollable spurious oscillations outside the test band. An EMT does
not create parasitic oscillations, in principle, at low frequencies.
The tuning repeatability accuracy of ETS is around −70 to −80 dB, and the
tuning accuracy of a modern EMT is around −60 dB. However, for accurate noise
and load-pull measurements, tuning accuracies of −40 dB are sufficient [16].
In terms of tuning resolution, EMTs possess high resolution, thereby allowing
fine-tuning to the optimal performance of the DUT; whereas an ETS exhibits ir-
regular impedance patterns with the points jumping unpredictably when the diodes
are switched on and off and, therefore, is not able to fine-tune to the optimal perfor-
mance of the DUT. In noise measurements, the unpredictable jumping of impedance
states for small frequency changes prohibits the use of ETSs in double sideband se-
tups. This necessitates either YIG (yttrium iron garnet) filters or other expensive
single sideband noise receivers. They do not allow direct measurement of the min-
imum noise figure, NF min , by tuning to the optimal reflection factor Γopt ; whereas
EMT allows fine-tuning and can, therefore, directly measure NF min .
ETS is favored for on-wafer applications; however, EMT has been used in on-
wafer applications below 0.8 GHz, after performing appropriate vibration tests [3].

2.3 Load-Pull Measurement

The load-pull measurement procedure primarily involves the following three steps:
• Assembly of the system components to establish a load-pull measurement setup.
• Calibration of the load-pull measurement setup for correction of errors arising
from imperfect system components, dispersions, and mismatches. Calibration for
setting the appropriate calibrated reference planes.
• Measurement of relevant data at the calibrated reference planes and then the de-
embedding of these data, in order to refer them to the device planes so that the
behavior of the DUT can be accurately predicted.
It is imperative to understand the typical setup of a load-pull system, as shown
in Fig. 2.5, before getting into the details. As depicted, a standard load-pull system
consists of various equipments, such as a directional coupler, impedance tuners,
power meters, and a computer for automating and controlling the overall measure-
ment process.
36 2 Passive Load-Pull Systems

Fig. 2.5 Block diagram of a typical setup that depicts standard components required for carrying
out automated load-pull measurements

In the block diagram of Fig. 2.5, the tuners are shown to be close to the DUT
ports, ensuring that maximum possible reflection coefficients synthesized by the
tuners are directly seen by the DUTs. However, this is a non-real time setup, as all
variables must be fixed a priori and then de-embedding procedures must then be
applied before measurements are made [19, 20]. Such systems are slow in terms of
speeds as compared to real-time load-pull systems [21–24].
In real-time systems, the incident and reflected waves are directly collected from
the DUT ports; thus, the measurements are instantaneous with every change in the
DUT operating conditions [22–24]. Nonetheless, a non-real-time passive load-pull
system is frequently used to utilize maximum synthesizable reflection coefficients
from passive tuners [25]. However, with the advancements in passive tuner tech-
nologies [9, 18], real-time passive load-pull systems are also possible and will ob-
viously provide better performance in terms of speed, accuracy, flexibility, ease and
deployment of the type of tuners.

2.3.1 Load-Pull Setup

Essentially, a complete load-pull measurement setup consists of at least two funda-

mental tuners (one each at the output and input of the DUT), one signal source, a test
fixture or probe station, two power meters or a dual channel power meter, a power
supply, a spectrum analyzer, an RF test set (including bias tees, cables, attenuators,
couplers, power combiners and dividers), and access to a vector network analyzer
(VNA). A generic block diagram of a real-time passive load-pull measurement setup
is given in Fig. 2.6 [26].
2.3 Load-Pull Measurement 37

Fig. 2.6 Generic block diagram of a real-time passive load-pull setup [26], © IEEE 1984

The DUT is fed from a high-power microwave source, and its source and load
impedances are controlled by adjustable impedance tuners at the input and output. If
the DUT is an oscillator, only the output part of the system is needed. Uncorrected
large-signal reflection coefficients at the input and output of the DUT are monitored
using dual directional couplers and RF network analyzers. Attenuators ensure that
the signal levels at the inputs of the harmonic frequency converters are within the
safe operating range. If only one network analyzer is available, coaxial switches can
be used to connect it back and forth between the input and output circuits.
The power meters are used for the monitoring of the uncorrected power incident
on the input port of the DUT and the uncorrected power generated at the output port
of the DUT. Only one power meter at each port is required, because the reflection
coefficients generated by the tuners are known [27], given that the tuners are pre-
characterized using tuner calibration techniques [28–32]. The pre-characterization
of tuners ensures that all the imperfections associated with the stub movement in
the impedance tuners, central conductor of the tuners and the connectors at the tuner
ports are compensated. This step also helps in speeding up the overall measurement,
as it enables synthesis of the required reflection coefficient using interpolation in-
stead of multiple actual movements of stubs in the horizontal and vertical directions
[13].
The power dividers create reference signals for the respective network analyzers.
In principle, in an alternate configuration, the power divider and power meter can
be connected to other coupled ports of the respective directional couplers. However,
such a configuration results into reduced accuracy in the measurements for the con-
dition when the reflection coefficients at the DUT ports are close to zero and giving
very small power meter readings [26, 27].
An interfacing computer (not shown in the figure) is used for controlling the
instrument and acquiring and processing the measured data, and for error correction.
38 2 Passive Load-Pull Systems

Fig. 2.7 Power waves and

the reflection coefficients of
interest at the DUT reference
planes [26], © IEEE 1984

2.3.2 System Calibration

The main objective behind the load-pull characterization of DUTs is the accurate
prediction of device performance under realistic conditions, but the errors due to
imperfect system components, such as inherent directivity, mismatch and cross-
coupling errors associated with the network analyzer, cause uncertainty in the reflec-
tion coefficients and RF power at the specified reference planes, thereby defeating
the primary objective.
It is, therefore, extremely important to calibrate the load-pull setup, in order to
accurately set and measure the reflection coefficients and, hence, the design param-
eters, such as PAE, drain efficiency, and gain. The calibration, in principle, serves
a dual purpose, by removing the systematic errors arising from imperfections and
system components, as well as by shifting the reference plane from the network an-
alyzer measuring planes to the DUT reference planes. In essence, after calibration,
the system performs three functionalities, which are critical for achieving accurate
load-pull measurement results and can be summarized as:
• Setting a particular impedance value (reflection coefficients).
• Measuring the load reflection coefficients (impedances).
• Measuring the required device performance at the particular impedance.
In practical load-pull measurements, the input and output powers, Pin and Pout ,
at the DUT plane, as shown in Fig. 2.7, are given by:
 
Pin = |a1 |2 − |b1 |2 = |a1 |2 1 − |Γin |2 (2.4)
 
Pout = |b2 | − |a2 | = |b2 | 1 − |ΓL |
2 2 2 2
(2.5)
where Γin and ΓL are the reflection coefficients at the input and port of the DUT
and the load seen by the DUT, respectively. The traveling waves a1 , b1 , a2 and b2
are the incident and reflected waves at the DUT ports as shown in Fig. 2.7.
During load-pull measurement, the impedance tuners set the desired reflection
coefficients, whereas the vector network analyzers (VNAs) and power meters in
Fig. 2.6 give direct (but uncorrected) measurements of Γin , ΓL , |a1 |2 and |b2 |2 .
The errors introduced by the VNA and the hardware configuration in the measured
reflection coefficients significantly affect the accuracy in the measurements of in-
put and output RF power. The accuracy in the measured reflection coefficients and,
hence, the RF power is limited, due to the inherent directivity, mismatch, and cross-
coupling errors associated with the network analyzer components. These errors im-
pact the performance of the load-pull setup and, in turn, put a question mark on the
reliability of the measured data.
2.3 Load-Pull Measurement 39

Fig. 2.8 Signal flow error

model for the input port
incorporating the
imperfections contributed by
directional coupler, power
meter and network analyzer
[26], © IEEE 1984

For example, unknown tuner losses add uncertainty in the measured data, if the
output power meter is farther from the DUT [1, 33], with tuner losses as large as
several decibels [34]. On the other hand, finite coupler directivity and connector
mismatches can result in significant errors in the measured RF power in systems
where the RF power levels are determined using directional couplers at the input
and/or output of the DUT [35, 36]. With 25 dB directivity, errors in the measured
power can be as large as 1 dB [26]; and, these errors significantly affect the mea-
surement data, if the VSWR in the directional coupler is high.
Error correction in the measurement can be carried out using the error flow model
of the complete measurement system given in Fig. 2.6. The respective error flow
models of the input and output ports (as given in Figs. 2.8 and 2.9) incorporate all
the errors, such as the finite coupler directivity, connector mismatches, and cross-
coupling between the reference and VNA’s measuring ports. The models have been
simplified using the flow graph reduction technique [5]. The major assumption con-
sidered in this simplification is that the reflection coefficients of the power meter
sensor heads and the reference and VNA inputs are constant and independent of RF
power level. These models form the basis of a unified approach to the vector error
correction of RF power and reflection coefficient measurements. These error mod-
els are analogous to the error models commonly used for error correction in VNA
systems [37], but include some additional terms.
In the error flow model of the input side shown in Fig. 2.8, a1T represents the in-
put signal from the high-power microwave source, Γ1T is the reflection coefficient
presented to the input directional coupler by the input tuner, e16 represents coupling
to the input power meter, and e11 represents directivity errors in the reference chan-
nel of the input network analyzer. The term e13 represents the source mismatch,
e12 e14 denotes the transmission tracking, and e15 indicates the error occurring in
the measuring port of the VNA. The incident and reflected traveling waves at the
input VNA ports are a0 and b0 , whereas the actual incident and reflected traveling
waves at the input ports of DUT are represented by a1 and b1 , respectively. The
VNA measures the uncorrected reflection coefficient, ΓinU , as given in Eq. (2.6):
b0
ΓinU = (2.6)
a0
40 2 Passive Load-Pull Systems

Fig. 2.9 Signal flow error

model for the output port
incorporating the
imperfections contributed by
directional coupler, power
meter and network analyzer
[26], © IEEE 1984

The quantity of interest is, however, the input reflection coefficient, Γin , expressed
in Eq. (2.7), which relates the error terms and the input uncorrected reflection coef-
ficient measured at the input VNA plane in Fig. 2.7.
b1 ΓinU − e11
Γin = = (2.7)
a1 e12 e14 − e11 e13 + e13 ΓinU
There are three unknowns in Eq. (2.6), namely e13 , e11 and e12 e14 . These unknowns
can be determined by carrying out measurements on calibration standards, such as
open-short-load (OSL), by connecting these standards to the input DUT plane in
Fig. 2.6. The measurements on the calibration standard provide a system of three
equations (2.8) that relate the error terms and the reflection coefficients, which can
be solved to determine the error terms of e11 , e13 and e12 e14 [38].
  ⎡ 1 (Γ ) U ⎤ ⎡
−(Γin )open −1 (ΓinU )open
⎤
e11 in open (Γin )open
e13 = ⎣ 1 (Γin )short (ΓinU )short −(Γin )short ⎦ ⎣ (ΓinU )short ⎦ (2.8)
e 1 (Γin )load (ΓinU )load −(Γin )load (ΓinU )load
where e = e12 e14 − e11 e13 . Terms Γin with subscripts open, load and short refer
to the actual reflection coefficients of the OSL standards at the DUT plane; whereas
terms ΓinU with subscripts open, load and short are the corresponding measurements
at the VNA plane.
The determination of error terms using Eq. (2.8) enables the calibration of the
input port for the measurement of an accurate reflection coefficient at the input DUT
plane, as given by Eq. (2.7).
For the error model of the output network in Fig. 2.9, e26 represents coupling
to the power meter, e25 denotes directivity errors in the test channel of the network
analyzer, and Γ2T and a2T indicate the tuner and node, respectively. This is a general
form that can describe any passive or active tuner. A passive tuner has a non-zero
Γ2T term and zero a2T , while active tuners have a non-zero a2T . One step in the
calibration procedure, which is described in the following paragraphs, requires that
a test signal is injected from the right of the output coupler. In this case, a2T is
non-zero.
The simplification of signal flow model in Fig. 2.9 gives the error corrected re-
flection coefficients at the output reference plane, in terms of the uncorrected read-
ing, ΓLU , given in Eq. (2.10).
a3
ΓLU = (2.9)
b3
2.3 Load-Pull Measurement 41

e22 e24 ΓLU

ΓL = e23 + (2.10)
1 − e21 ΓLU
The simplifications of Figs. 2.8 and 2.9 also provide error corrected input and output
power, as per Eqs. (2.11) and (2.12).


2

2
e12
(1 − |Γin | )
2
Pin = |P1 |

(2.11)
e16 |1 − Γin e13 |2
|P2 |2  
Pout = |1 − e21 ΓL |2 1 − |ΓL |2 (2.12)
|e24 e26 |2

where |P1 |2 and |P2 |2 are the input and output power meter readings, respectively.
It is important to note that the equations are independent of the tuner reflection
coefficients, Γ1T and Γ2T , and the directivity terms, e15 and e25 . Therefore, it is not
necessary to obtain explicit values for these four terms.
The measurement of input power requires the determination of the error terms of
e13 , e11 and e12 e14 , which can be obtained by standard one-port VNA calibration
techniques in which the DUT is replaced by a series of calibration standards such
as a short circuit, an offset short circuit, and an open circuit [39]. The error term
|e12 /e16 |2 in Eq. (2.11) is obtained by connecting a matched power meter in place
of the DUT. The ratio of the power meter reading to the input power meter reading
is |e12 /e16 |2 .
For the measurement of output power, once again one-port VNA calibration [39]
can be performed to determine the error terms e21 , e23 , and e22 e24 . The input signal
for this stage of the calibration, a2T , is injected from the right in Fig. 2.6, with the
signal generator and an amplifier connected in place of the load. To determine the
error term, |e24 e26 |2 , the magnitude of the insertion loss of the output coupler and
power divider, between the DUT output reference plane and the output power meter,
is measured. If the test signal injected into the coupler at the DUT reference plane
is supplied from a matched source and if the output tuner is replaced by a matched
load, the power insertion loss, IP , is given by:
|e24 e26 |2
IP = (2.13)
|1 − e21 e25 |2
Since e21 and e25 are usually small, the insertion loss measurement gives |e24 e26 |2
directly.
As a guideline, for maximum accuracy, the system should be calibrated at coaxial
(APC-7 or APC-3.5) reference planes close to the test fixture. Then, appropriate de-
embedding using coaxial-to-microstrip transition [40] needs to be carried out to
transfer the reference plane into the microstrip test fixture.
The available signal power, Pa , at the output of the driver amplifier, in Fig. 2.6
is measured using a matched power meter. The driver amplifier is then connected at
the input port of the output directional coupler, in place of the DUT. The corrected
output power is obtained for a range of different output tuner settings across the
load plane. These power measurements are compared with the calculated output
42 2 Passive Load-Pull Systems

Fig. 2.10 The performance

of a 1 W GaAs FET for
|ΓL (2f0 )| = 1 with a variable
phase of the second
harmonic, while maintaining
optimal fundamental and
third harmonic terminations
[41], © IOP Journal of
Measurement Science and
Technology 2010

power, Pa (1 − |ΓL |2 ), which is obtained using the measured Pa and the corrected
ΓL . A maximum tolerance of around 0.15 dB between the corrected measured and
calculated output power values provides confidence in the calibration process [26].

2.4 Harmonic Load-Pull System

The terminal loading conditions at harmonic frequencies significantly affect device
performance. In principle, during the characterization and optimization of transis-
tor devices, the harmonic power generated by the device needs to be fully reflected
back at a given phase, in order to extract the best possible efficiency. In an ideal
scenario, all the harmonic power can be reflected back; however, this is not pos-
sible in practical load-pull systems, due to lossy transition between the tuner and
the device. During the characterization of devices, it is a common practice to vary
the phase and magnitude of the harmonic reflection coefficients, while keeping the
fundamental reflection coefficient constant, in order to obtain optimal efficiency.
Figure 2.10 illustrates an example of the effect of phase variation of ΓL (2f0 )
on the performance of a 1 W GaAs (gallium arsenide) FET (field effect transistor)
device, while keeping the reflection coefficients at the fundamental frequency and
third harmonic fixed [41]. It is apparent that, in this case, the variation of phase of the
second harmonic reflection coefficient can swing the efficiency by more than 20 %.
Similar effects can be observed when the phase of the third harmonic reflection
coefficient or the magnitudes of the second or third harmonic coefficients are varied
[42]. Improvements to the tune of over 15 % in efficiency can be achieved in the final
designed PA, while terminating the harmonic impedances in optimal terminations,
compared to 50  [43]. In addition to the harmonically tuned PAs, the harmonic
terminations play a key role in the design of switching-mode PAs [44].
Device characterization for the optimal design of either harmonically tuned PAs
or switching-mode PAs is carried out using a harmonic load-pull system [45–47].
One of the first reported passive harmonic load-pull setups [33], which is shown in
Fig. 2.11, was based around a circulator. The architecture used standard slide screw
tuners, filters and power meters to monitor and control the magnitude and phase of
the reflected signal components at the fundamental and second harmonic frequen-
cies. This setup, although very basic when compared to current developments, did
2.4 Harmonic Load-Pull System 43

Fig. 2.11 This architecture

was one of the first passive
harmonic load-pull setups
[33], © IEEE 1979

achieve tuning of impedances for fundamental and second harmonic frequencies. In

this setup, the transmitted wave, b2 , enters the circulator via port 1; and, the funda-
mental and second harmonic components exit through the appropriate low-pass and
high-pass filters at ports 2 and 3, respectively. The respective tuners reflect the trav-
eling waves, a2 (f0 ) and a2 (2f0 ), and subsequently synthesize the respective funda-
mental and second harmonic reflection factors at the load plane given by Eqs. (2.14)
and (2.15).

a2 (f0 )
ΓL (f0 ) = (2.14)
b2 (f0 )
a2 (2f0 )
ΓL (2f0 ) = (2.15)
b2 (2f0 )

The reflected traveling waves, a2 (f0 ) = ΓL (f0 )b2 (f0 ) and a2 (2f0 ) = ΓL (2f0 ) ×
b2 (2f0 ), then combine at port 1 of the circulator to form the overall reflected travel-
ing wave, a2 , as given in Eq. (2.16):

a2 = a2 (f0 ) + a2 (2f0 ) = ΓL (f0 )b2 (f0 ) + ΓL (2f0 )b2 (2f0 ) (2.16)

In theory, an n-port wideband circulator with appropriate band-pass filters can

be employed to extend this setup to n − 1 harmonics. However, in practice, it is
not feasible, considering the bandwidth limitation and substantial losses associated
with circulators beyond the second harmonic. Usually, high harmonic reflection co-
efficients are required, and the losses render the circulator based harmonic load-pull
system inappropriate in practical applications. In addition, the poor isolation of cir-
culators also affects the system’s ability to synthesize harmonically independent
reflection coefficients and, therefore, limits the overall usefulness of this harmonic
load-pull setup.
Over the years, passive harmonic load-pull architectures and setups have evolved
to address the two major issues of high reflection factor synthesis at harmonic fre-
quencies and harmonically independent reflection factor emulation [17, 18]. These
developments in the harmonic load-pull architectures can be grouped in three dis-
tinct categories, namely triplexer based harmonic load-pull, harmonic rejection
tuner harmonic load-pull and multipurpose single tuner harmonic load-pull [17].
44 2 Passive Load-Pull Systems

Fig. 2.12 Block diagram of triplexer based three-harmonic load-pull system

2.4.1 Triplexer Based Harmonic Load-Pull Setup

A block diagram of a typical three-harmonic triplexer based load-pull architecture

is shown in Fig. 2.12. It is apparent that an appropriate triplexer and tuners corre-
sponding to carrier and harmonic frequencies are required to build a triplexer based
harmonic load-pull system. The input and output biasing networks provide for the
biasing at the gate and source, respectively and, the input and output coupling net-
works are needed for capturing the transmitted and reflected traveling waves at the
input and output ports, respectively. The optional source tuner is used for matching
at the input port. The triplexer, which acts like filters for the respective harmonic
components, separates the harmonic components from the output generated by the
DUT. The subsequent filters at the output of the triplexer further improve the rejec-
tion of the out-of-band frequency components.
The respective load-tuners for harmonic components f0 , 2f0 , and 3f0 tune the
reflection environment as per specific requirements and reflect the respective a2 (f0 ),
a2 (2f0 ), and a2 (3f0 ) and, in the process, synthesize the harmonically independent
reflection factors by interacting with the respective components, b2 (f0 ), b2 (2f0 )
and b2 (3f0 ), of the traveling wave, b2 . The reflected traveling wave components,
having different weighting factors, contributed by the tuners, then combine to form
the overall a2 given in Eq. (2.17).

a2 = a2 (f0 ) + a2 (2f0 ) + a2 (3f0 ) (2.17)

The primary benefit of such a technique is that the reflection factors at different
harmonic frequencies can be set independently, i.e., the harmonic reflection factors
are independent of each other. It can thus be inferred that this method gives full
magnitude and phase control of all three harmonic frequencies, f0 , 2f0 , and 3f0 .
The drawback of this technique is the limited tuning coverage at the harmonics,
as shown in Fig. 2.13, due to the losses inherent in the triplexer. This is a serious
limitation for applications requiring high harmonic terminations, such as class-F and
inverse class-F mode PAs [48].
2.4 Harmonic Load-Pull System 45

Fig. 2.13 Depiction of tuning coverage at f0 , 2f0 and 3f0 for the triplexer based three-harmonic
load-pull system [17]

Fig. 2.14 The harmonic rejection tuner based three-harmonic load-pull architecture

2.4.2 Harmonic Rejection Tuner Based Harmonic Load-Pull Setup

The harmonic rejection tuner (HRT) based harmonic load-pull architecture incor-
porates an HRT in addition to a fundamental tuner, as shown in Fig. 2.14. In a
standard HRT based harmonic load-pull setup, pre-matched type harmonic tuners,
as explained in Chap. 5, such as passive harmonic tuners (PHT) [17] are used.
The fundamental tuner, possessing full magnitude and phase control for the re-
flection coefficient at the fundamental frequency, f0 , is placed farther from the DUT
output port, as the reflection coefficient required at the fundamental frequency is
smaller than those of the harmonic frequencies. The HRT, which is capable of full
phase control at harmonic frequencies 2f0 and 3f0 , is placed nearer to the DUT
port, in order to minimize the impact of the loss associated with the cables and con-
nectors. This helps in the synthesis of higher reflection coefficients at the harmonic
frequencies.
Furthermore, this technique overcomes the loss associated with the triplexer;
therefore. the reflection coefficient can cover a larger region on the Smith chart,
as shown in Fig. 2.15. The harmonic rejection tuners are low loss, which also con-
tributes to the synthesis of reflection coefficient with increased magnitude. It is also
evident that the HRTs, although are capable of providing higher reflection coeffi-
cients, do preclude a large portion of Smith chart. This limitation, however, does not
46 2 Passive Load-Pull Systems

Fig. 2.15 Depiction of tuning coverage at f0 , 2f0 and 3f0 for the HRT based three-harmonic
load-pull system [17]

Fig. 2.16 The multipurpose tuner based harmonic load-pull architecture

affect the usefulness in practical applications considering that the required harmonic
reflection coefficients usually fall nearer the boundary of the Smith chart.
The primary limitation of this technique is the poor isolation between the syn-
thesized reflection coefficients at the fundamental and harmonic frequencies. This
is due to the fact that the tuners are cascaded; therefore, setting of the tuner state at
one frequency affects the tuner state at the other frequencies. As a consequence, this
technique becomes impractical for more than three harmonic frequencies.

2.4.3 Single Tuner Harmonic Load-Pull Setup

Figure 2.16 shows a typical block diagram of a single tuner harmonic load-pull
setup, which also includes a source tuner and peripheral biasing and measuring net-
work. In this architecture, a multipurpose load tuner that covers the fundamental
frequency and the range of all of the relevant harmonic frequencies is employed
for the synthesis of harmonically independent fundamental and harmonic reflection
coefficients [17].
For example, a three-harmonic multipurpose tuner uses three independent wide-
band probes for the control of the magnitude and phase of the reflection coefficients
2.4 Harmonic Load-Pull System 47

Fig. 2.17 Depiction of tuning coverage at f0 , 2f0 and 3f0 for the single multipurpose tuner based
three-harmonic load-pull system [17]

at the three harmonic frequencies of f0 , 2f0 , and 3f0 . The proper positioning and
movement of the three probes in the horizontal and vertical directions allow inde-
pendent tuning at the three harmonic frequencies [17]. A multipurpose tuner gives
full magnitude and phase control for all three harmonic frequencies, as shown in
Fig. 2.17.
Since this setup utilizes only one tuner, the overall system loss associated with bi-
asing and measuring networks, cables and connectors is less; hence, the achievable
magnitude of the reflection coefficient is higher than other harmonic load-pull ar-
chitectures. This setup is compact and smaller in size compared to the types of other
setups. A major limitation of this technique includes the requirement for very pow-
erful computing facility to monitor and control the movement of the tuner probes. In
addition, this technique also suffers from inherent limited isolation, due to the series
configuration of all the probes. The limited isolation prevents this setup from synthe-
sizing harmonically independent harmonic reflection coefficients at all frequencies
and in all parts of the Smith chart.

2.4.4 Harmonic Load-Pull Comparisons

The harmonic tuning methods can be compared primarily in two aspects, namely
the tuning isolation and the effective Smith chart coverage.
An important factor of any load-pull device characterization system is the accu-
racy at which the impedance seen by the DUT at the fundamental frequency can be
controlled. If the impedance changes without the user’s knowledge or if it cannot be
kept constant for varying harmonic impedances, the measurement results obtained
are invalid. Hence, when considering harmonic load-pull systems, it is essential to
know the isolation across frequencies.
The triplexer based harmonic load-pull architecture relies on the isolation in-
herent within the isolator to mitigate the effects of variation in the setting of one
tuner from setting of another. Triplexers typically possess 50–60 dB of isolation at
the fundamental and harmonic frequencies; therefore, this technique can practically
48 2 Passive Load-Pull Systems

Fig. 2.18 Definition of

VSWR at load and tuner
planes for a fundamental
load-pull tuning setup

tune harmonic impedances independent of each other. In some applications, poor

isolation outside the fundamental and harmonic frequency band can cause spurious
oscillations, especially at low frequencies, due to high out-of-band reflections.
In the HRT based harmonic load-pull setup, the frequency isolation between the
harmonic resonators is around 30 dB. The poor frequency isolation is the result of
the sum of vectors generated by the reflections at f0 , 2f0 , and 3f0 , as well as the
residual reflection of the slab-line [17]. This makes setting of harmonically inde-
pendent impedances extremely difficult.
In the case of the multipurpose tuner, the fundamental and harmonic probes repo-
sition themselves for each desired reflection factor for the combination of f0 , 2f0 ,
and 3f0 , which helps in achieving very good tuning isolation. However, the mea-
surement using multipurpose based harmonic load-pull is relatively slow, due to
physical movement of probes for each set of harmonic impedances.
For tuning range, it is the effective voltage standing wave ratio (VSWR) at the
load plane that matters and not the tuner’s ability to synthesize the VSWR at the
tuner plane. The VSWR at the load plane depends on the interface between the
tuner and the DUT and gets reduced accordingly. Figure 2.18 provides a simple
relationship of the VSWR at the load and tuner planes. It is important to note that
there are always a passive measurement network and transition between the DUT
and the tuner, which reduce the achievable VSWR at the load plane compared to the
tuner plane.
In a triplexer based harmonic load-pull setup, as shown in Fig. 2.12, there is loss
due to the insertion loss of the triplexer, which is responsible for the reduced Smith
charge coverage shown in Fig. 2.13. The coverage gets further reduced at harmonic
frequencies, as the triplexer loss at harmonic frequencies is higher.
In the case of HRT based harmonic load-pull architecture, there are only mea-
suring passive networks between the tuner and the DUT, as shown in Fig. 2.14;
therefore, the reduction in the Smith chart coverage is smaller. However, there is a
transition between the fundamental and harmonic tuners, causing reduction in the
VSWR and further reducing the Smith chart coverage at the fundamental frequency,
f0 , as shown in Fig. 2.15, compared to the harmonic frequencies.
In a single multipurpose tuner, the reduction in the VSWR at the harmonic and
fundamental frequencies is almost similar, as all the tuning probes are contained
in a single casing. The reductions in the VSWR at the fundamental and harmonic
frequencies are caused only by the passive measuring network between the tuner
and the DUT.
2.5 Tuning Range Enhancement 49

Table 2.2 Features of harmonic tuning methods [17]

Tuning method Advantages Disadvantages

Triplexer based – High tuning isolation – Unsuitable for on-wafer

applications
– Simple extension of – Insertion loss of the triplexer
existing setup at all frequencies causes a
reduced tuning range
– Very good amplitude and – Out-of-band reflections in
phase control at all the the triplexer can cause
harmonic frequencies spurious oscillations

Harmonic rejection tuner – High tuning range – Very poor tuning isolation
based – High power handling – Unsuitable for broadband
capability applications
– Low insertion loss at – Limited to only three
fundamental frequency harmonics

Multipurpose single – High tuning range – Slow measurement time and

tuner based throughput
– High tuning isolation – Requires powerful
computing resource
– Ideal for on-wafer
applications
– Appropriate for broadband
applications

Table 2.2 summarizes the major features of the described three different har-
monic load-pull techniques.

2.5 Tuning Range Enhancement

The tuning range in a passive load-pull system is limited, due to the inherent losses
in the biasing and measuring network, connectors, cables, fixture and the tuner itself
[30]. Therefore, depending on the load-pull system architecture, a substantial part
of the Smith chart is precluded; and, as a consequence, not all relevant impedances
can be synthesized
Most of the low-power device characterization and PA design requirements are
met by standard passive load-pull setups. However, limitations of such systems
are apparent during high-power device characterization and PA design or harmonic
load-pull measurements, as these situations require highly reflective impedance en-
vironments. To address this issue, several solutions have been proposed [12].
The enhanced loop [49] and cascaded tuner architectures [18] are recent devel-
opments in passive load-pull architectures that have the ability to synthesize high
reflection coefficients. In both of these systems, the idea is the enhancement of the
50 2 Passive Load-Pull Systems

Fig. 2.19 The enhanced loop

load-pull architecture for
synthesizing a high reflection
coefficient [49], © IEEE 2010

Fig. 2.20 The signal flow

model depicting the
generation of reflection factor
using enhanced loop
load-pull architecture [49],
© IEEE 2010

tuning range by reflecting back the transmitted signal that is transmitted through the
main tuner.

2.5.1 Enhanced Loop Architecture

In order to achieve high reflection factors, this technique combines a passive tuner
and a passive loop, as shown in Fig. 2.19. It employs a low-loss circulator, Cir-2,
with very high directivity, so that the loss incurred in the setup is smaller and that
there is good isolation between signals a4 and b3 . The employed coupler, C2, and the
loop cable, L2 , are low loss. The coupling port of the coupler is connected to a power
meter, in order to monitor and measure the DUT performances, in terms of output
power or linearity, during load-pull characterization. The passive loop at the output
of Tuner2 generates reflection coefficient ΓLOADLOOP at plane-4 and augments the
overall reflection coefficient ΓL at plane-3. The flow model in Fig. 2.20 depicts the
generation of reflection factor ΓL at plane-3.
The following expressions relate the traveling waves in the enhanced loop load-
pull architecture of Fig. 2.19:

a 4 = k4 a 3 (2.18)
a5 = k5 a 4 (2.19)
b3 = k6 a 5 (2.20)
2.5 Tuning Range Enhancement 51

Fig. 2.21 Two tuners

cascaded in series to obtain
higher reflection factor [18],
© Maury Microwave
Corporation

where k4 , k5 and k6 are the complex factors dependent on the S-parameters of the
circulator, Cir-2, and the directional coupler, C2.
The mismatch between Tuner2 and the passive loop plane, Plane-4, is related by
the following expression:
b3
ΓLOADLOOP = KL = = |KL |e−2jβL2 (2.21)
a3
where the parameter KL (= k4 k5 k6 ) is a complex factor that depends on the pas-
sive loop structure characteristics, i.e., the transmission factors of coupler C2 and
circulator Cir-2.
It is clear from Eq. (2.21) that the reflection coefficient generated by the passive
loop is also dependent on the phase velocity, β, of the traveling waves and the length
of cable in the passive loop, L2 . The simplification of the flow model in Fig. 2.20
results in the following expression for the overall reflection coefficient at plane-3.
b2 S12TUNER2 S21TUNER2 KL
ΓL = = S11TUNER2 + (2.22)
a2 1 − S22TUNER2 KL
Equation (2.16) demonstrates that the total load reflection coefficient, ΓL , at plane-3
gets enhanced by the contribution from the passive loop. If the loop does not con-
tribute anything to the overall reflection coefficient, then S11 of Tuner2 is the total
reflection coefficient.

2.5.2 Cascaded Tuner

Figure 2.21 depicts the cascading of two passive tuners in series. The cascaded
tuners are capable of generating reflection factor, Γ1 , at the DUT plane, which is
governed by Eq. (2.23) [18].
S12 S21 Γ2
Γ1 = S11 + (2.23)
1 − S22 Γ2
where S11 , S21 , S12 , and S22 are the S-parameters of Tuner1 (closest to the DUT).
Γ2 is S11 looking into Tuner2 at a particular termination.
The overall reflection looking into the cascaded tuner combination is equal to the
complex reflection of the first tuner added to the complex reflection of the second
tuner and multiplied by some insertion/reflection factor. It is important to notice that
the effect of the second tuner is highly influenced by the S21 S12 product of the first
tuner. There are two extreme cases that best describe this effect [18]:
52 2 Passive Load-Pull Systems

• If Tuner1 is at a physical short circuit, i.e., S21 S12 = 0, Tuner2 would be com-
pletely masked.
• If Tuner1 is at initialized at 50 , i.e., S11 ∼ 0 and S21 S12 ∼ 1, Tuner2 would be
all that is seen.

References
1. J.M. Cusack, S.M. Perlow, B.S. Perlman, Automatic load contour mapping for microwave
power transistors. IEEE Trans. Microw. Theory Tech. 22, 1146–1152 (1974)
2. F. Secchi, R. Paglione, B. Perlman, J. Brown, A computer controlled microwave tuner for
automated load pull. RCA Rev. 44(4), 566–583 (1983)
3. Focus Microwave, Mechanical vibrations of CCMT tuners used in on-wafer load-pull testing,
Application Note AN-46, Oct. 2001
4. J. Sevic, Introduction to tuner-based measurement and characterization, Technical Note,
Maury Microwave Corporation, 5C-054
5. D.M. Pozar, Microwave Engineering, 3rd edn. (Wiley, New York, 2005). ISBN 0-471-17096-
8
6. Microlab, Mechanical tuners, Application Note, Oct. 2000
7. R. Tuijtelaars, Overview of device noise parameter measurement system, in VDE/ITG-
23.10.01 (2001), pp. 1–5
8. Maury Microwave Corporation, Device characterization with harmonic load and source pull,
Application Note: 5C-044, Dec. 2000
9. Focus Microwave, Load pull measurements on transistors with harmonic impedance control,
Technical Note, Aug. 1999
10. B.W. Leake, A programmable load for power and noise characterization, in IEEE Microwave
Theory and Techniques Society’s International Microwave Symposium Digest, Dallas, USA
(June 1982), pp. 348–350
11. Maury Microwave Corporation, LP series electronic tuner system, Technical Data, 4T-081,
2002
12. M.S. Hashmi, F.M. Ghannouchi, P.J. Tasker, K. Rawat, Highly reflective load-pull. IEEE Mi-
crow. Mag. 11(4), 96–107 (2011)
13. Focus Microwave, Computer controlled microwave tuner—CCMT, Product Note 41, Jan.
1998
14. Maury Microwave Corporation, Slide screw tuners, Technical Data, 2G-035A, Feb. 1998
15. Focus Microwaves, Algorithms for automatic high precision residual tuning to 50  using
programmable tuners, Application Note 45, May 2001
16. Focus Microwave, Electronic tuners (ETS) and electromechanical tuners (EMT)—a critical
comparison, Technical Note, Aug. 1998
17. Focus Microwave, Comparing harmonic load-pull techniques with regards to power-added
efficiency (PAE), Application Note 58, May 2007
18. Maury Microwave Corporatio, Cascading tuners for high-VSWR and harmonic applications,
Application Note 5C-081, Jan. 2009
19. F. Deshours, E. Bergeault, F. Blache, J.-P. Villotte, B. Villeforceix, Experimental comparison
of load-pull measurement systems for nonlinear power transistor characterization. IEEE Trans.
Instrum. Meas. 57(11), 1251–1255 (1997)
20. A. Ferrero, V. Teppati, Experimental comparison of active and passive load-pull measure-
ment technologies, in 30th European Microwave Conference Proceedings, Paris, France (Oct.
2000), pp. 1–4
21. C. Arnaud, J.L. Carbonero, J.M. Nebus, J.P. Teyssier, Comparison of active and passive load-
pull test benches, in 57th ARFTG Conference, vol. 39 (May 2001), pp. 1–4
References 53

22. M.S. Hashmi, A.L. Clarke, S.P. Woodington, J. Lees, J. Benedikt, P.J. Tasker, An accurate
calibrate-able multiharmonic active load-pull system based on the envelope load-pull concept.
IEEE Trans. Microw. Theory Tech. 58(3), 656–664 (2010)
23. W.S. El-Deeb, N. Boulejfen, F.M. Ghannouchi, A multiport measurement system for complex
distortion measurements of nonlinear microwave systems. IEEE Trans. Instrum. Meas. 59(5),
1406–1413 (2010)
24. D. Barataud, F. Blache, A. Mallet, P. Bouysse, J.-M. Nebus, J. Villotte, J. Obregon, J. Ver-
specht, P. Auxemery, Measurement and control of current/voltage waveforms of microwave
transistors using a harmonic load-pull system for the optimum design of high efficiency power
amplifiers. IEEE Trans. Instrum. Meas. 48(4), 835–842 (1999)
25. J.E. Mueller, B. Gyselinckx, Comparison of active versus passive on-wafer load-pull char-
acterization of microwave MMwave power devices, in IEEE Microwave Theory and Tech-
niques Society’s International Microwave Symposium Digest, San Diego, USA (June 1994),
pp. 1077–1080
26. R.S. Tucker, P.D. Bradley, Computer-aided error correction of large-signal load-pull measure-
ments. IEEE Trans. Microw. Theory Tech. 32(3), 296–300 (1984)
27. P.D. Bradley, R.S. Tucker, Computer-corrected load-pull characterization of power MES-
FETs, in IEEE Microwave Theory and Techniques Society’s International Microwave Sym-
posium Digest, Boston, USA (June 1983), pp. 224–226
28. C. Tsironis, Adaptable pre-matched tuner system and method, US Patent No. 6674293
29. G.R. Simpson, Impedance tuner systems and probes, US Patent No. 7589601
30. J. Sirois, B. Noori, Tuning range analysis of load-pull measurement systems and impedance
transforming networks, in 69th ARFTG Conference, Honolulu, USA (June 2007), pp. 1–5
31. C. Roff, J. Graham, J. Sirois, B. Noori, A new technique for decreasing the characterization
time of passive load-pull tuners to maximize measurement throughput, in 72nd ARFTG Con-
ference, Portland, USA (Dec. 2008), pp. 92–96
32. P. Hart, J. Wood, B. Noori, P. Aaen, Improving loadpull measurement time by intelligent
measurement interpolation and surface modeling techniques, in 67th ARFTG Conference, San
Francisco, USA (June 2006), pp. 69–72
33. R.B. Stancliff, D.B. Poulin, Harmonic load-pull, in IEEE Microwave Theory and Techniques
Society’s International Microwave Symposium Digest, Florida, USA (Apr. 1979), pp. 185–
187
34. E.W. Strid, Measurement of losses in noise-matching networks. IEEE Trans. Microw. Theory
Tech. 29(3), 247–252 (1981)
35. G.P. Bava, U. Pisani, V. Pozzolo, Active load technique for load-pull characterization at mi-
crowave frequencies. IEE Electron. Lett. 18(4), 178–180 (1982)
36. Y. Takayama, A new load pull characterization method for microwave power transistors, in
IEEE Microwave Theory and Techniques Society’s International Microwave Symposium Di-
gest, New Jersey, USA (June 1976), pp. 218–220
37. R.A. Hackborn, An automatic network analyzer system. Microw. J. May, 45–52 (1968)
38. W.S. El-Deeb, M.S. Hashmi, S. Bensmida, N. Boulejfen, F.M. Ghannouchi, Thru-less cali-
bration algorithm and measurement system for on-wafer large-signal characterization of mi-
crowave devices, IET J. Microw. Antenna Propag. 4(11), 1773–1781 (2010)
39. E.F. DaSilva, M.K. McPhun, Calibration technique for one-port measurements. Microw. J.
June, 97–100 (1978)
40. J.R. Souza, E.C. Talboys, S-parameter characterization of coaxial to microstrip transition. IEE
Proc. 129(Part H), 37–40 (1982)
41. M.S. Hashmi, A.L. Clarke, J. Lees, M. Helaoui, P.J. Tasker, F.M. Ghannouchi, Agile harmonic
envelope load-pull system enabling reliable and rapid device characterization. IOP J. Meas.
Sci. Technol. 21(055109), 1–9 (2010)
42. M.S. Hashmi, A.L. Clarke, S.P. Woodington, J. Lees, J. Benedikt, P.J. Tasker, Electronic multi-
harmonic load-pull system for experimentally driven power amplifier design optimization,
in IEEE Microwave Theory and Techniques Society’s International Microwave Symposium
Digest, Boston, USA, vols. 1–3 (June 2009), pp. 1549–1552
54 2 Passive Load-Pull Systems

43. F. Blache, J.-M. Nebus, P. Bouysse, L. Jallet, A novel computerized multiharmonic active
load-pull system for the optimization of high-efficiency operating classes in power transistor,
in IEEE International Microwave Symposium Digest, Orlando, USA (June 1995), pp. 1037–
1040
44. A. Grebennikov, N.O. Sokal, Switch Mode RF Power Amplifiers (Elsevier, Oxford, 2007)
45. F.M. Ghannouchi, F. Beauregard, A.B. Kouki, Power added efficiency and gain improvement
in MESFETs amplifiers using an active harmonic loading technique. Microw. Opt. Technol.
Lett. 7(13), 625–627 (1994)
46. R. Hajji, F.M. Ghannouchi, R.G. Bosisio, Large-signal microwave transistor modeling using
multiharmonic load-pull measurements. Microw. Opt. Technol. Lett. 5(11), 580–585 (1992)
47. F.M. Ghannouchi, R. Larose, R.G. Bosisio, A new multiharmonic loading method for large-
signal microwave and millimeter-wave transistor characterization. IEEE Trans. Microw. The-
ory Tech. 39(6), 986–992 (1991)
48. Y.Y. Woo, Y. Yang, B. Kim, Analysis and experiments for high efficiency class-F and inverse
class-F power amplifiers. IEEE Trans. Microw. Theory Tech. 54(5), 1969–1974 (2006)
49. F.M. Ghannouchi, M.S. Hashmi, S. Bensmida, M. Helaoui, Enhanced loop passive source-
and load-pull architecture for high reflection factor synthesis. IEEE Trans. Microw. Theory
Tech. 58(11), 2952–2959 (2010)
Chapter 3
Active Load-Pull Systems

This chapter describes active load-pull techniques, their design issues, and important
characteristics. The description involves two aspects namely the theoretical postula-
tions of various active load-pull methods and then the respective designs, features,
and practical issues encountered in realizations. The peripheral equipments used for
measurement of traveling waves at the device under test (DUT) ports have not been
included, as their configuration and calibration is similar to the description provided
previously.

3.1 Introduction

Passive load-pull systems, although provide direct characteristic impedance trans-

formation through passive tuners, are unable to synthesize high reflection coeffi-
cients due to inherent losses in the measurement network between DUTs and tuner
[1]. This drawback has a serious repercussion in the characterization of either high
power transistor devices or devices under harmonic terminations, where the required
reflection coefficients lie mostly on or near the border of the Smith chart [2, 3].
Active load-pull systems, which are based on injection of signals at the DUT
plane, possess the ability to synthesize high reflection coefficients by overcoming
the inherent losses in the passive technique, and therefore can provide full Smith
chart coverage [4, 5].
Closed-loop, feed-forward, and open-loop are the three broad categories of active
load-pull technique [6–8]. For the emulation of desired reflection coefficients at the
DUT plane the closed-loop, also known as feedback, load-pull modifies the DUT
output appropriately before feeding it back as reflected traveling wave [4, 6]. Feed-
forward or split signal load-pull uses a part of the input signal to generate coherent
modified signal to be reflected at the DUT plane [5, 7], whereas the open-loop load-
pull generates coherent modified reflected wave from external signal source [8], for
synthesis of reflection coefficients at the DUT plane.

F.M. Ghannouchi, M.S. Hashmi, Load-Pull Techniques with Applications to Power 55

Amplifier Design, Springer Series in Advanced Microelectronics 32
DOI 10.1007/978-94-007-4461-5_3, © Springer Science+Business Media Dordrecht 2013
56 3 Active Load-Pull Systems

Fig. 3.1 Diagram depicting

the concept of reflection
coefficient synthesis in an
ideal closed-loop active
load-pull system

3.2 Closed-Loop Load-Pull System

Figure 3.1 depicts block diagram of an ideal closed-loop active load-pull system. It
can be observed that it requires a three-port component, e.g. circulator, for forward-
ing the incident traveling wave, b2 , and reflected traveling wave, a2 , in the respective
directions. A complex variable ρej θ modifies the phase and magnitude of the inci-
dent traveling wave, b2 , before injecting it back into the form of reflected traveling
wave, a2 , so as to synthesize reflection coefficient given below in Eq. (3.1).
a2 b2 ρej θ
ΓLoad =
= = ρej θ (3.1)
b2 b2
where, ρ is the magnitude and θ is the phase of the desired reflection coefficient,
ΓLoad , at the DUT plane.

3.2.1 System Realization

Realization of the closed-loop load-pull system requires a variable attenuator, a

phase shifter, and a loop amplifier as shown in Fig. 3.2. The attenuator/loop am-
plifier combined together allows adjustment in the magnitude of reflected traveling
wave a2 , and the phase shifter modifies its phase, in order to emulate ΓLoad at the
DUT plane. The circulator directs the traveling waves b2 and a2 in their respective
direction. A directional coupler in place of circulator can also be employed, and will
be described in a later section, for a load-pull system requiring superior stability.
In an ideal situation, when the circulator is assumed ideal and the losses in the
system are ignored, the modified wave generated by the feedback loop can be given
by Eq. (3.2) and the reflection coefficient at the DUT plane by Eq. (3.3).
a2 = Gb2 (3.2)
a2
ΓLoad = =G (3.3)
b2
where, G is complex gain of the feedback loop and represents the overall change in
magnitude and phase of the incident traveling wave, b2 .
It is, thus, safe to assume that the magnitude of ΓLoad is proportional to G, estab-
lished through the settings of attenuator and loop amplifier, while the phase of ΓLoad
3.2 Closed-Loop Load-Pull System 57

Fig. 3.2 Block diagram of a

realized closed-loop active
load-pull system

is completely dependent on the setting of the phase shifter. Therefore, in principle,

the reflection coefficient, ΓLoad , at the DUT plane, given by Eq. (3.3), can be set to
any value, even larger than unity, by choosing an appropriate loop gain G. However,
the simplicity of Eq. (3.3) is slightly misleading considering that a high value of G
can create oscillating closed-loop load-pull [1, 4]. This is an unwanted situation as it
has the potential to damage either the DUT being characterized or the measurement
system itself.
The oscillation in the load-pull system can be eliminated by keeping the loop
gain G below one (G < 1). In order to keep G < 1, the amplifier gain should not
exceed the isolation of the circulator [6]. In theory, an ideal circulator has infinite
isolation and therefore the amplifier gain can be set to any value. However, for all
practical purposes, standard circulators possess isolation of the order of 30 dB and
therefore the loop amplifier must be set below this level in order to avoid instability
in the loop. In practical situations a narrow band filter, preferably waveguide type
selected on purposes to prevent oscillation at all frequencies lower than its cut-off
frequency, is inserted in the loop. The filter also helps in the mitigation of oscillation
at higher frequencies outside the band of characterization.

3.2.2 Analysis of Closed-Loop System

In a normal circumstance the closed-loop load-pull system, Fig. 3.2, must synthesize
high reflection coefficient, at the DUT plane, without encountering any difficulty.
However, experimental investigations have shown the following problem of such a
system during the characterization and measurement applications [4, 6, 9].
• A strong inter-dependency of magnitude, |ΓLoad |, and phase, ϕLoad , of the reflec-
tion coefficient, ΓLoad , on both the phase shifter and attenuator settings.
• A limited tuning range while synthesizing small magnitude, |ΓLoad |, of reflection
coefficient.
• Occurrence of oscillation in case of a broadband loop amplifier.
To gain better assessment on how these undesired effects occur and affect the
performance of the closed-loop active load-pull system, the analysis of the system
using signal flow model shown in Fig. 3.3 is carried out. The S-parameters depend
58 3 Active Load-Pull Systems

Fig. 3.3 Signal flow model

and the associated
S-parameter description of
the closed-loop active
load-pull system [1], © IEEE
1994

solely on the characteristics of the components used in the feedback loop. The circu-
lator is described by three port S-parameters Scr . The feedback network is described
by two-port S-parameters Sf , which are obtained from the cascade connection of the
S-parameter sets for the attenuator, Sat , phase shifter, Sps , and loop amplifier, Sa .
The reflection coefficient at the DUT plane, for linear operation of loop amplifier,
deduced from Fig. 3.3 is given in Eq. (3.4).
S12 S21
ΓLoad = S11 + (3.4)
1 − S22
where,
Sf 11 Scr13 Scr31 −Sf 11 Scr33 Scr11 +Sf 12 Scr13 Scr21 −Sf 12 Scr23 Scr11 +Scr11
S11 = 1−Sf 11 Scr23 −Sf 12 Scr23 (3.5)
Sf 11 Scr13 Scr32 −Sf 11 Scr33 Scr12 +Sf 12 Scr13 Scr22 −Sf 12 Scr23 Scr12 +Scr12
S12 = 1−Sf 11 Scr23 −Sf 12 Scr23 (3.6)
Sf 21 Scr33 + Sf 22 Scr23
S21 = Sf 21 γ + Sf 22 α + S11 (3.7)
Scr13
Sf 21 Scr33 + Sf 22 Scr23
S22 = Sf 21 δ + Sf 22 β + S12 (3.8)
Scr13
The parameters α, β, γ , and δ in the above equations are variables defined in
Eqs. (3.9)–(3.12) [1].
Scr11 Scr23
α = Scr21 − (3.9)
Scr13
Scr12 Scr23
β = Scr22 − (3.10)
Scr13
Scr33 Scr11
γ = Scr31 − (3.11)
Scr13
Scr33 Scr12
δ = Scr32 − (3.12)
Scr13
3.2 Closed-Loop Load-Pull System 59

It can be ascertained from Eqs. (3.4)–(3.12) that the non-ideal components em-
ployed in the design of the active closed-loop load-pull system plays key role in the
reflection coefficient synthesis ability of the system. The degree of the undesired
effects, due to the non-ideal components, obviously depends on the characteristics
of the respective circulator, attenuator, phase shifter, and the loop amplifier. For ex-
ample, typical S-parameters given in Eqs. (3.13)–(3.16) have been inserted into the
expressions (3.4)–(3.12) to determine the impact of imperfection in the loop com-
ponents on reflection coefficient ΓLoad [1].
 
0.09 0.9 0.1
Scr = 0.9 0.09 0.01 (3.13)
0.1 0.01 0.09
 
0.25 0.01
Sat = (3.14)
0.01 0.25
 
0.11 0.91
Sps = (3.15)
0.91 0.11
 
0.3 25.1
Sa = (3.16)
0.25 0.3
Figure 3.4 illustrates the dependence of magnitude and phase of ΓLoad on the
variations in the |Sat21 |, and Arg(Sps21 ). For example, at an attenuator setting
|Sat21 | = −22 dB, a change of the phase shifter setting Arg(Sps21 ) from 0° to 180°
results in no phase change,
ϕLoad , at all but in a magnitude change,
ΓLoad , of 0.7
in the reflection coefficient, ΓLoad .
It can also be observed from Fig. 3.4 that the synthesizable reflection coefficients
using closed-loop active load-pull are restricted. For example, in this case the system
cannot synthesize reflection coefficient smaller than 0.5 for the phase shifter setting
at Arg(Sps21 ) = 0° irrespective of the variation in the attenuator setting |Sat21 |. Re-
flection coefficient smaller than 0.5 can be synthesized using the active load-pull
system if loop components have lesser imperfections than Eqs. (3.13)–(3.16).
It is easier to comprehend the two problems through Eqs. (3.4)–(3.12), i.e., the
limitation of the system to synthesize smaller reflection coefficient and the strong
dependence of synthesized reflection coefficient on the attenuator and phase shifter
settings, but these equations are too abstract for interpreting the oscillation condi-
tions in the system. In order for comprehending the equations easily, first the ideal
situation is considered where all components are assumed to have matched input
and output ports and the circulator is assumed to have infinite directivity. The am-
plifier is also assumed to possess infinite reverse isolation. With these assumptions
the flow model of Fig. 3.4 greatly simplifies, and gives the expression, Eq. (3.17),
for synthesized reflection coefficient at the DUT plane.
ΓLoad = ΓLoad
o
= Scr31 Sat21 Sps21 Sa21 Scr12 (3.17)
o
where ΓLoad is the synthesized reflection coefficient, when all the components in
the system are ideal, whose magnitude is directly determined by the setting of the
attenuator and the phase by the setting of the phase shifter.
60 3 Active Load-Pull Systems

Fig. 3.4 Magnitude (top) and

phase (bottom) contours of
the reflection coefficient as a
function of the attenuator and
phase shifter settings [1],
© IEEE 1994

Next Eqs. (3.4)–(3.12) are used as a basis for a perturbation analysis. Cumulative
Step-by-step approach where only one non-ideal effect of the components used in
the feedback network is considered (e.g. poor input match Sa11 of amplifier (effect
(1) in Table 3.1)), leading to a set of expressions for ΓLoad which can be analyzed
simply.
Table 3.1 summarizes the results of perturbation analysis for the closed-loop ac-
tive load-pull system. The non-ideal effects (2), (5), and (7) can be neglected if an
isolator is used at the output of the loop amplifier. The input mismatch (8) of the
o
circulator Scr11 directly adds to ΓLoad and hence must be kept small. The effects
3.2 Closed-Loop Load-Pull System 61

Table 3.1 Results of perturbation analysis for closed-loop active load-pull system [1], © IEEE
1994
Effects Importance ΓLoad − ΓLoad
o

(1) Sa11 Minor, damped by Scr31 Sat21 Sps21 Sa11 Sat12 Sps12 Scr13
coupling factor
(2) Sa22 Major, isolator behind Scr21 Sa22 Scr12
amplifier helpful
(3) Sa12 Minor Scr21 Sa12 Sat12 Sps12 Scr13
(4) Sat11 Minor, damped by Scr31 Sat11 Scr13
coupling factor
(5) Sat22 Major, isolator behind Scr21 Sa12 Sps12 Sat22 Sps21 Sa21 Scr12
amplifier helpful
(6) Sps11 Minor, damped by Scr31 Sat11 Scr13
coupling factor
(7) Sps22 Major, isolator behind Scr21 Sa12 Sps22 Sa21 Scr12
amplifier helpful
(8) Scr11 Major, matched input Scr11
needed
(9) Scr22 Minor, damped by Scr31 Sat21 Sps21 Sa21 Scr22 Sa12 Sps12 Sat12 Scr13
coupling factor
(10) Scr33 Minor, damped by Scr21 Sat21 Sps21 Sa21 Scr33 Sa12 Sps12 Sat12 Scr12
coupling factor
Scr21 Sat12 Sps12 Sa12 Scr13
(11) Scr23 Major, high directivity is 1−Sa12 Sps12 Sat12 Scr23
a must
Scr31 Sat21 Sps21 Sa21 Scr12
(12) Scr32 Major, high directivity is 1−Sa21 Sps21 Sat21 Scr32
a must. Avoid
oscillations:
|Sa21 ||Scr32 | < 1

(1), (4), (6), (9), and (10) mainly concern the active closed-loop load-pull system
utilizing directional coupler as a three-port component. A high coupling factor will
dampen all these effects. In case of circulator, high isolation between the ports can
eliminate all these effects.
Effects (11) and (12) create condition of oscillations, when the forward gain of
the feedback loop gets higher than 1, and can be suppressed through the use of high
directivity circulator or directional coupler. A sufficient condition to avoid oscil-
lations under all attenuator and phase shifter settings (|Sat21 | < 1, |Sps21 | < 1) is
given in Eq. (3.18), which conveys that the amplifier gain has to be smaller than the
circulator directivity at all frequencies.
|Sat21 ||Scr32 | < 1 (3.18)
Further, Fig. 3.5 gives another condition, Eq. (3.19), which, if not satisfied, can
create instability in the load-pull system.
|ΓLoad ||ΓDUT | < 1 (3.19)
62 3 Active Load-Pull Systems

Fig. 3.5 A simplified signal flow graph for the closed-loop load pull

In practical applications |ΓDUT | is normally smaller than 1, ΓDUT < 1, con-

sequently active load-pull with the ability to synthesize |ΓLoad | higher than 1,
ΓLoad > 1, has the potential to nullify Eq. (3.19) and create unwanted oscillations
according to the basic stability theory [10]. This may happen if a broadband loop
amplifier with large gain ripple is used in the feedback loop. For example, a high
reflection coefficient synthesized by reducing attenuation in the loop at a specific
frequency f1 might create oscillation at another frequency f2 if the amplifier gain
at f2 is higher. It is also important to note that the stability condition, Eq. (3.19),
also affects the DUT, as both the load-pull and the DUT are coupled with each other
through their reflection coefficients and therefore any oscillation in the system easily
transfers to the DUT [10].
Simplification of Eqs. (3.3) and (3.19) suggests that the active load-pull loop-
gain G must be smaller than the return loss of the output of the DUT in order to
maintain stability in the measurement setup as expressed in Eq. (3.20).
G < 20 log |S22 | dB (3.20)
Equation (3.20) also identifies that a perfect match at the DUT output allows an
infinite gain G without any oscillations in the system. In practice, though, the gain
G is limited by the finite directivity of the circulator and therefore in order to avoid
oscillations, the loop gain G must be smaller than the directivity of the circulator so
as to fulfill the stability criterion in Eqs. (3.18) and (3.20).

3.3 Closed-Loop Load-Pull Architectures

Analysis of closed-loop system provides a suggestion that in its simplest form, it
is capable, in principles, of synthesizing high reflection coefficient, despite the fact
that it suffers from instability issues. In order to overcome the oscillation problem,
an isolator is required at the output of loop amplifier as shown in Fig. 3.6. The
incorporation of isolator is extremely useful, in particular, during characterization
and measurement of high power DUTs where the signal powers are extremely high
and slight mismatches condition at the circulator ports have the potential to cause
loop oscillation and damage the measurement system and/or the DUT.
3.3 Closed-Loop Load-Pull Architectures 63

Fig. 3.6 Block diagram of closed-loop active load pull incorporating isolator and narrow band
filter in order to eliminate oscillation problem [11], © IEEE 2007

Oscillations in closed-loop load-pull system, due to large gain ripple in the char-
acteristics of the loop amplifier, can be eliminated by inserting a highly selective
bandpass filter, such as yttrium iron garnet (YIG) filter, in the feedback loop as
shown in Fig. 3.6. The YIG filters possess flat gain over a specified bandwidth;
therefore eliminate the effects of large gain ripple in the response of loop amplifier.
The downside of the YIG filter insertion in the loop, however, is the resulting limited
bandwidth of the load-pull system.
Applications such as characterization and measurement of transistor devices for
high efficiency PA design require harmonic terminations on the border of Smith
chart [12–15]. The active closed-loop system can be easily extended to multi-
harmonic configurations, as shown in Fig. 3.7 for three-harmonic frequencies sys-
tem. Three-harmonic load-pull system requires a triplexer for separating the har-
monic components of the incident traveling wave, b2 , given in Eq. (3.21). The re-
spective harmonic components are modified in the feedback loop and then combined
by the other triplexer to create reflected traveling wave, a2 , given in Eq. (3.22). The
ratio of Eq. (3.21) and (3.22) give the reflection coefficients, at the respective har-
monic frequencies, according to Eq. (3.23).

b2 = b2 (f0 ) + b2 (2f0 ) + b2 (3f0 ) (3.21)

a2 = a2 (f0 ) + a2 (2f0 ) + a2 (3f0 ) (3.22)
a2 (nf0 )
ΓLoad (nf0 ) = (3.23)
b2 (nf0 )
where, n = 1, 2, 3.
The harmonic load-pull shown in Fig. 3.7 is capable of synthesizing any desired
reflection coefficient at various harmonic components, by enhancing power level of
reflected traveling wave a2 of specific harmonic component. The constraint on loop
amplifier, however, is the major bottleneck considering that it might go into com-
pression in order to raise the power level of the reflected traveling wave. Successful
operation of system, however, requires that the loop amplifier must be linear which
necessitates an appropriate corrective steps described in next section.
64 3 Active Load-Pull Systems

Fig. 3.7 Generic architecture of three-harmonic load-pull system based on active closed-loop
technique

Fig. 3.8 Block diagram of load-pull system combining active closed-loop and passive tuner [6],
© IEEE 2008

3.4 Optimized Closed-Loop Load-Pull System

In particular, the losses in the measuring network and the active load-pull loop have
dual impact: enhanced risk of oscillations [16] and increased output power require-
ment from the loop amplifier for the synthesis of a given ΓLoad . A simple technique
to address these problems is to replace the circulator with a low-loss three-port com-
ponent, such as load-pull head [17], as shown in Fig. 3.8. Note that the presence of
isolator in the loop is necessary to protect the loop amplifier, and to attain a stable
feedback loop.
Inclusion of a passive tuner in the loop, further, helps in reducing the output
power demand from the loop amplifier for the synthesis of low impedances. The
tuner in the loop could be two-port slug tuner (with one or more slugs) [18] so that
3.4 Optimized Closed-Loop Load-Pull System 65

Fig. 3.9 Simplified schematic of load-pull system combining active closed-loop and passive tuner
[6], © IEEE 2008

when the slugs are completely raised up, the tuner does not provide any contribu-
tions to ΓLoad and therefore is completely set by only the active loop. Under this
condition the power required from the loop amplifier is maximal. On the other hand
when the slugs are fully inserted, the loop is cut off, and the loop amplifier does
not contribute to ΓLoad regardless of its output power. As a conclusion, between
these two extreme cases, there must be an optimal situation (i.e., an optimal tuner
slug setting), where the required output power from loop amplifier reaches a mini-
mum. This aspect can be analyzed using the simplified schematic in Fig. 3.9, which
depicts the unavoidable losses due to cables in the loop by Lc and losses in the
measuring network by L. The schematic also includes the respective incident and
reflected traveling waves at various planes, and the scattering matrix (S-matrix), S,
of the passive tuner.
Following quantities are defined and/or assumed for simplification of the analysis
of closed-loop load-pull system schematic given in Fig. 3.9 [6].
• G = the overall loop gain (including the coupling factor of load-pull head, am-
plifier gain, and overall loop losses).
• The isolator is ideal and therefore bA = 0.
• ΓLo = the desired reflection coefficient at the DUT plane.
• ΓLt = the maximum reflection coefficient synthesized by only passive tuner at the
DUT plane (i.e., when the active loop is completely shut down (i.e., aA = 0).
• Pout = the power delivered by the DUT when ΓLoad = ΓLo .
Now the output power, PA , from loop amplifier can be expressed by Eq. (3.24).
PA = |aA |2 (3.24)
where,
aA = Gac = GLb2 (3.25)
Equations (3.24) and (3.25), when combined, give the expression for loop ampli-
fier output power in terms of the incident traveling wave, b2 , at the DUT plane and
the losses in the measuring network, L, given in Eq. (3.26).
PA = |G|2 |L|2 |b2 |2 (3.26)
66 3 Active Load-Pull Systems

The incident traveling wave, b2 , when ΓLoad = ΓLo is related to the DUT output,
Pout , by the expression given in Eq. (3.27).
 2 Pout
b  = (3.27)
2
1 − |ΓLo |2
Simplification of Eqs. (3.26) and (3.27) gives the loop amplifier output power, PA ,
Eq. (3.28), in terms of the feedback loop gain, G, power delivered by the DUT, Pout ,
and loss in the measuring network, L.
Pout
PA = |G|2 |L|2 (3.28)
1 − |ΓLo |2
At the tuner plane, S-parameter of the loop tuner, represented as Sij (i, j = 1, 2),
is related to the reflected traveling wave at the amplifier plane, aA , by Eq. (3.29).
bt = S11 at + S12 aA (3.29)
For load-pull head, with low coupling factor (< 20 dB), following assumptions
relating the losses in the measuring network, losses in the feedback loop, and the
traveling waves at the tuner reference plane can be made [6, 19]:
a2
bt = (3.30)
LLc
at = LLc b2 (3.31)
Substitution of Eqs. (3.25), (3.30) and (3.31) into Eq. (3.29) and followed by a
simplification provides the expression for the loop gain, G, given in Eq. (3.32).
ΓLo − S11 (LLc )2
G= (3.32)
S12 L(LLc )
Moreover, for a reciprocal lossless tuner S11 and S12 are related by Eq. (3.33) [19].
|S11 |2 + |S21 |2 = |S11 |2 + |S12 |2 = 1 (3.33)
In case of reciprocal low loss passive tuner in the loop, therefore, following as-
sumptions can be made:
|S11 |2 + |S12 |2 = |γ |2 (3.34)
where, |γ | is the maximum value that |S11 | can reach and should not be confused
with the propagation constant.
Equations (3.28), (3.32) and (3.34) can be solved to deduce the expression for
the output power from loop amplifier given in Eq. (3.35) [6].
|ΓLo − S11 (LLc )2 |2 Pout
PA = (3.35)
(|γ | − |S11 | |(LLc )| ) 1 − |ΓLo |2
2 2 2

Equation (3.35) gives insights on the behavior of the amplifier output power with
respect to the desired reflection coefficient, ΓLo , the pre-tuning provided by the loop
tuner, S11 , and overall losses, LLc , in the measurement setup. This equation can be
further analyzed by plotting PA versus S11 for different values of LLc , at a fixed
ΓLo , with the purpose of identifying the impact of loop tuner and load-pull head on
3.4 Optimized Closed-Loop Load-Pull System 67

Fig. 3.10 Normalized loop

amplifier output power PA
versus tuner S11 for different
values of overall losses.
PA
is the difference between the
minimum PA and its value
for S11 = 0 [6], © IEEE 2008

PA . For example, Fig. 3.10 shows a plot of PA versus S11 when ΓLo = 0.96 and
γ = 1. In this case S11 has been assumed purely real and S11 continuously varies
between zero and γ .
The plot contains information about PA for several values of LLc , from −2.4 dB
(typical for measurement setups with standard couplers) to −0.4 dB (setup with
load-pull head). The PA in this plot has been normalized with respect to its value
corresponding to LLc = −2.4 dB and S11 = 0.
It can be observed that the use of load-pull head in the active load-pull setup
reduces the requirements on loop amplifier power output by reducing the attenuation
terms LLc . For S11 = 0, i.e. when the tuner is not present in the loop, the required
PA to obtain ΓLo = 0.96 is 2 dB lower for LLc = −0.4 dB, with respect to the value
of the required power when LLc = −2.4 dB.
Furthermore, Fig. 3.10 also shows the effect of loop tuner as a prematching de-
vice in active closed-loop load-pull system. It is evident that the minimum required
PA show strong dependence on the loop tuner setting S11 . If term
PA is the dif-
ference between the minimum value of the PA and its value for S11 = 0, then it can
be deduced that
PA is 2 dB for LLc = −2.4 dB and is 10 dB for LLc = −0.4 dB.
This clearly demonstrates the advantage of using loop tuner and load-pull head in
the closed-loop active load-pull system.
The determination of the best tuner setting for the minimum PA requires zeroing
of the first derivative of (3.35) on the real axis, which results into Eq. (3.36).

|γ |2 (LLc )2
S11min = (3.36)
ΓLo

where, S11min is one of the tuner setting for minimum PA .

68 3 Active Load-Pull Systems

If the active loop is shut down, i.e., aA = 0, and the tuner S11 is set to S11min , the
reflection coefficient synthesized at the DUT Plane, ΓLoad , is smallest, ΓLmin , and is
given by Eq. (3.37) [6].
|γ |2 (LLc )4
ΓLmin = (3.37)
ΓLo
However, when aA = 0, the maximum |ΓLoad | obtainable with the tuner, |ΓLt |, is
given in Eq. (3.38) according to the initial definition/assumption [6].
|ΓLt | = |γ |(LLc )2 (3.38)
This means that by setting the tuner to |S11min |, the magnitude of reflection coef-
ficient synthesized at the DUT plane will be given by Eq. (3.39).
|ΓLt |2
|ΓLoad | = (3.39)
|ΓLo |
With these considerations, a practical procedure to find the best tuner position,
which minimizes PA , and to obtain the desired ΓLo is given below [6].
• Measure the maximum |ΓLt | obtainable with the tuner when the loop is shut
down.
• If the desired |ΓLo | < |ΓLt |, there is no need for active closed-loop load-pull.
• Otherwise, set the tuner, when the loop is shut down, so that |ΓLoad | =
(|ΓLt |2 /|ΓLo |) and arg(ΓLoad ) = arg(ΓLo ); this corresponds to setting the S11
of the tuner equal to the S11min .
• Finally, synthesize the desired ΓLo by properly setting the active-loop attenuator
and phase shifter.

3.5 Feed-Forward Load-Pull System

This technique involves splitting of source signal into two parts, with one forwarded
to the DUT to be measured while the other gets modified by attenuator, phase shifter,
and loop amplifier before being fed at the output of the DUT output port for achiev-
ing the load-pull functionality, as shown in Fig. 3.11. Because of high isolation pro-
vided by the power splitter and the additional isolator at the loop amplifier output,
there is only minor risk of oscillations in this load-pull technique as evident from
the ensuing analysis.
If the components of the load-pull are assumed ideal, i.e. power splitter with no
insertion or return loss and an infinite isolation between both of its output ports and
the connectors exhibiting no loss, then the modified traveling wave generated by it
is given by Eq. (3.40).
a2 = Ga1 (3.40)
where factor G represents the overall complex gain introduced by the feed-forward
load-pull components. Additionally Eq. (3.40) assumes a perfectly matched loop
amplifier; a condition provided by the isolator at the loop amplifier output.
3.5 Feed-Forward Load-Pull System 69

Fig. 3.11 A generic schematic of feed-forward active load-pull architecture

The incident traveling wave at the DUT plane, b2 , is related to the excitation,
a1 , and the reflected traveling wave at the DUT plane, a2 , by the expression in
Eq. (3.41).
b2 = S21 a1 + S22 a2 (3.41)
where S21 and S22 are the large-signal S-parameters of the DUT at the fundamental
frequency [21, 22].
Equations (3.41) and (3.42), when simplified, provide the expression for the syn-
thesized load reflection coefficient, ΓLoad , Eq. (3.42), at the DUT plane.
a2 1
ΓLoad = = S (3.42)
b2 ( + S22 )
21
G
It can be inferred from Eq. (3.42) that the synthesized reflection coefficients,
ΓLoad , are dependent not only on the load-pull gain, G, but also on the DUT large
signal S-parameters S21 and S22 . The synthesized ΓLoad will be zero when G = 0,
i.e. when the load-pull is not operated, and will be maximum when G = 1/S22 , i.e.
when the load-pull is in operation and very high power is injected at the DUT output
port by the feed-forward load-pull. The DUT operating in stable condition always
possess S22 ≤ 1 and therefore the maximum ΓLoad obtained from the feed-forward
load-pull system, according to Eq. (3.42), can cover the entire Smith chart.
However, the synthesis of ΓLoad by the feed-forward load-pull is unpredictable
due to its dependence on the large signal S-parameters S21 and S22 of an unknown
DUT. For example, in order to synthesize a specific ΓLoad the load-pull gain, G,
can be calculated, from Eq. (3.42), using the values S21 and S22 from previous
set of measurements. The large-signal S-parameters depend largely on the port
impedances at the DUT, and the changing DUT parameters such as drive power
and bias. It is, therefore, likely that both terms S21 and S22 will change with any
70 3 Active Load-Pull Systems

Fig. 3.12 Measurement flow

graph of feed-forward
load-pull system [20],
© IEEE 1994

Fig. 3.13 Loci of the

synthesizable reflection
coefficients by the
feed-forward load-pull
system [20], © IEEE 1994

variation in the setting of the load-pull components, drive power or bias, thereby
necessitating the calculation of new value for G. Thus the dependence of the large
signal S-parameters on port impedances, drive power, and bias level makes the re-
flection coefficient synthesis an iterative process. In principle the feed-forward load-
pull system requires convergence technique, described in a later section, in order to
achieve precision in the synthesized reflection coefficient.
The feed-forward load-pull system, although slow in synthesizing desired reflec-
tion coefficient due to iterative nature, exhibits increased stability as compared to
closed-loop active load-pull system. It remains stable as long as the complex gain,
G, remains less than the combined reverse transmission coefficient, S12 , of the DUT
and the isolation of the power splitter. In practical situations, a standard power split-
ter provides an isolation of more than 20 dB [23] and the reverse transmission co-
efficient, S12 , readily adds another 20 dB, consequently provides plenty of room to
the feed-forward load-pull gain, G. Furthermore, the output and input of the active
feed-forward loop are completely isolated and, as a result, its stability is indepen-
dent of the reflection coefficient S22 of the DUT. Consequently the condition for the
stable operation of the active feed-forward load-pull is decoupled from the stability
criterion for the DUT, therefore discards the need of loop amplifier with constant
magnitude and phase frequency response.
It can, additionally, be inferred from Eq. (3.42) that, although, the feed-forward
load-pull remains stable for majority of practical applications but can oscillate when
complex loop gain approaches extremely high value, which might be the case dur-
ing high power DUT measurements and characterization. As a result unconditional
stability of the DUT can’t be guaranteed; thereby necessitating fulfillment of an
additional condition at the DUT input so as to achieve stability.
3.6 Optimized Feed-Forward Load-Pull System 71

Fig. 3.14 A graphical

illustration of problems
encountered by the
feed-forward load-pull
technique [20], © IEEE 1994

Fig. 3.15 Modified schematic of feed-forward active load-pull system (top) and the associated
signal flow model (bottom) [26], © IEEE 1993

3.6 Optimized Feed-Forward Load-Pull System

The expression for reflection coefficient synthesized by a typical feed-forward load-

pull, given in Fig. 3.11 and whose measurement flow graph is illustrated in Fig. 3.12,
in term of its loop components is given by Eq. (3.43) [24]. In the flow graph, the
stimulus is represented by power source a1 , while the load-pull is represented by a2 .
a2
ΓLoad = = L1 Gej ϕ (3.43)
b2
72 3 Active Load-Pull Systems

where,
L1 = losses in the feed-forward loop
Gej ϕ = complex gain of the feed-forward loop
A feed-forward load-pull system output is normally matched to 50  and as a
result ΓLoad at the initialization is zero. Consequently, reflection coefficients syn-
thesized by it makes locus around the 50  impedance as shown in Fig. 3.13 [25].
This pattern of synthesized reflection coefficient is useful for weakly mismatched
DUTs, but may encounter three following major problems for highly mismatched
DUTs [20].
• The distribution of reflection coefficients may yield poor concentration, therefore
may not allow fine scan, of reflection coefficients in Smith chart regions where
the DUTs achieve their best matching conditions, Fig. 3.14. It has been found
that more than half of the experimental data determined from this system is not
useful in the case of highly mismatched DUTs [25], leading to a load-pull system
providing extremely poor measurement throughput.
• The required high mismatches result in reflection coefficients of large magnitude
and in such conditions slight error in their phase adjustments may result into
damage of the DUTs, as shown in Fig. 3.14.
• Large power sources are required to drive the output port of the DUT in order to
generate high reflection coefficients. For example, 8-watt DUT requires at least
40-watt power from the feed-forward load-pull source in order to synthesize re-
quired reflection coefficients, for accurate optimization and characterization [20].
To improve and overcome the three major problems of feed-forward load-pull
system, the output can be deliberately mismatched and moved away from 50 
by using an assembly of sliding-short circuit and a directional coupler as shown in
Fig. 3.15 [26]. The load-pull component is then connected to the coupled port of
the directional coupler, with coupling factor C, to synthesize the required reflection
coefficient at the DUT plane. Signal flow graph of the modified feed-forward load-
pull, depicted in Fig. 3.15, which also includes term Γ0 , the prematch established by
sliding-short circuit, and δa2 , the signal injected by the feed-forward load-pull, can
be analyzed to derive expression for synthesized reflection coefficients in Eqs. (3.44)
and (3.45).
a2 = Γ0 b2 + Cδa2 (3.44)
a2 δa2
= ΓLoad = Γ0 + C = Γ0 + δΓ (3.45)
b2 b2
It can be deduced from Eq. (3.45) that the reflection coefficient will be equal
to Γ0 , the value set by the sliding-short circuit, if the power from the feed-forward
load-pull is turned off. It is therefore a standard practice, in such a system, to load-
pull any DUT using the sliding-short circuit to first identify the Smith chart region
where the probable optimal reflection coefficient could lie, and then synthesize it
using the sliding-short circuit itself. Later the active component of feed-forward
load-pull injects δa2 to trace the locus of synthesized reflection coefficients around
3.6 Optimized Feed-Forward Load-Pull System 73

Fig. 3.16 Loci of the

synthesizable reflection
coefficient by the modified
feed-forward load-pull
architecture [20], © IEEE
1994

it, for example shown in Fig. 3.16. In other words, when feed-forward load-pull
is turned on, the reflected power, δa2 , generate reflection coefficients which act as
perturbations δΓ on the initial setting Γ0 of the sliding-short circuit as depicted in
Fig. 3.16.
Only small δa2 are required to generate perturbations δΓ , in a particular chosen
area of Smith chart, and therefore extremely small possibility exists for the modified
system to generate reflection coefficients beyond the safe region of Smith chart. As
a result, the incorporation of a sliding-short circuit in the feed-forward load-pull
allows a very safe load-pull operation resulting in reduced risk of damage to DUTs.
Moreover, modification in the feed-forward load-pull system results into substan-
tially reduced output power from the loop amplifier as it only need to provide the
additional perturbations δΓ on top of Γ0 . For example, if vector perturbation δ Γ is
assumed in phase with the vector Γ0 then Eq. (3.45) can be simplified to obtain the
expression for the power reflected by the load-pull source, 12 δa22 , given in Eq. (3.46).

1 1
|δa2 |2 = |b2 |2 |ΓLoad − Γ0 |2 (3.46)
2 |C|2

The power reflected by the load-pull source in standard feed-forward load-pull,

Fig. 3.11, can be given by expression:

1 1
|a2 |2 = |b2 |2 |ΓLoad |2 (3.47)
2 |C|2

The ratio, R, of the power reflected by the active feed-forward load-pull system in
the modified, Fig. 3.15, and original, Fig. 3.11, setup is given by:

1 |ΓLoad − Γ0 |2
R= (3.48)
|C|2 |ΓLoad |2
74 3 Active Load-Pull Systems

Fig. 3.17 A generic architecture of three-harmonic load-pull system based on feed-forward

method

If a perfect lossless 6 dB coupler is assumed, i.e., Γ0 = 0.75 and |C|2 = 0.25, then
following can be derived from Eq. (3.48) [20]:
ΓLoad = 0.9 → R = 11 %

ΓLoad = 0.95 → R = 17 % (3.49)

ΓLoad = 0.99 → R = 23.5 %

It is evident from Eq. (3.49) that the incorporation of sliding-short circuit and
a perfect 6 dB coupler in the standard feed-forward load-pull system results in the
reduction of required reflected power to the tune of almost 75 % while synthesizing
reflection coefficients between 0.9 to 0.99.

3.7 Harmonic Feed-Forward Load-Pull System

Extension of feed-forward load-pull to harmonic setup is simple, for example the

three-harmonic setup depicted in Fig. 3.17. The feed-forward active loop requires
frequency doubler and tripler to generate respective harmonic frequencies consider-
ing that the input signal has little or no power at the harmonics. Attenuators, phase
shifters, and loop amplifiers appropriately modify the harmonic components of re-
flected traveling wave, a2 , in order to synthesize the desired reflection coefficients
at the respective harmonics according to Eq. (3.23).
The modified components a2 (f0 ), a2 (2f0 ), and a2 (3f0 ) are then combined to-
gether to formulate a2 , according to Eq. (3.22), by the multiplexer placed at the
3.7 Harmonic Feed-Forward Load-Pull System 75

Fig. 3.18 Block diagram of 4-harmonic feed-forward active load-pull technique [27], © IEEE
2000

output of the feed-forward loop. The major limitation of this technique, however, is
the escalating requirement of higher power loop amplifiers with the increase in num-
ber of harmonic components to be load-pulled. To minimize required output power
from the loop amplifiers, vector generator based electronic load-module (ELM) can
be employed for the realization of feed-forward load-pull system, for example as
shown in Fig. 3.18.
ELM comprises a chain of IQ modulator (made up of double-balanced mixer) for
the fundamental frequency, and frequency multipliers for the harmonic components.
In this setup configuration, the source signal directed to the load-pull loop, a1 , first
gets divided into quadrature signals by a 3-dB hybrid which then get multiplied
with I and Q modulation signals by two double-balanced mixers, see Fig. 3.18,
and summed up to formulate reflected traveling wave, a2 , according to Eq. (3.22).
Suppose I and Q signals are DC voltages VI and VQ with positive or negative
polarity, the resulting harmonic components from each ELM is the output signal,
Vout (t), given in Eq. (3.50).
 
Vout (t)n = a2 (nf0 ) = cstV̂in VIn cos ω0 t + VQn sin ω0 t (3.50)
where n refers to the harmonic index, V̂in is the input signal amplitude, and cst the
conversion gain of the multipliers. VI controls the real part and VQ the imaginary
part of the output phasor in the complex plane. As a result the IQ modulator enables
the adjustment of the RF output voltage, i.e. magnitude and phase of the reflected
traveling wave, with arbitrary amplitude and phase relation to the input signal. The
multipliers and the control signals also provide gain, dependent on the input voltage
range of ELM, to respective harmonic components of a2 and consequently reduce
the required output power from loop amplifiers.
76 3 Active Load-Pull Systems

Fig. 3.19 A generic block diagram of active open-loop load-pull technique

The harmonic feed-forward load-pull, in addition to its slow reflection coeffi-

cient synthesis ability, is also not capable of emulating harmonically independent
reflection coefficients. This is due to the fact that the search for convergence of any
specific harmonic reflection coefficient involves inherent interactions between the
traveling wave components of other harmonics, which consequently leads to the
interaction among the reflection coefficients at harmonic frequencies.

3.8 Open-Loop Load-Pull System

Active open-loop and feed-forward load-pull techniques are almost similar, with the
only difference between the two being the source of reflected traveling wave, a2 .
The feed-forward load-pull utilizes same source for excitation and reflected wave,
whereas the open-loop system employs two separate sources for the excitation and
reflected wave as shown in Fig. 3.19. The sources are locked to a common refer-
ence signal, such as 10 MHz locking signal, in order to maintain phase coherence
between the transmitted and reflected waves at the DUT plane. Isolator prevents and
discards any potential damage to the loop amplifier, whereas attenuator and phase
shifter modifies the phase and magnitude of reflected wave.
The reflection coefficient, at the DUT plane, synthesized by an open-loop load-
pull is expressed as:
a2 a2 1
ΓLoad = = = a1 (3.51)
b2 S21 a1 + S22 a2 (S21 a2 + S22 )
where S21 and S22 are the large-signal S-parameters of the DUT at the fundamental
frequency [21, 22], and G represents the overall complex gain generated by the
open-loop active load-pull.
It is evident, from Eq. (3.51), that open-loop load-pull can synthesize ΓLoad from
0 to ∞ just by changing the magnitude of reflected traveling wave, a2 . Reflection
coefficient, ΓLoad , will be zero when the load-pull source is switched off resulting
into no reflected wave, a2 , while it will increase with the increasing a2 and will
reach a maximum of 1/S22 , exactly same as the feed-forward load-pull, usually
3.8 Open-Loop Load-Pull System 77

Fig. 3.20 Block diagram of a 3-harmonic active open-loop load-pull setup [8], © IEEE 2000

Fig. 3.21 Active open loop harmonic load-pull architecture employing doubler, tripler and just
one load-pull source

required for practical applications. Furthermore, synthesis of reflection coefficient

using open-loop load-pull being dependent on large signal S-parameters of the DUT,
which change with excitation, bias, and impedance environment at the ports, is an
iterative and slow process which necessitates convergence algorithm, described in
next section, for its precise setting.
The advantage of the open-loop load-pull, though, is its absolute stability, due to
the fact that it makes no loop when attached to the DUT and as a result eliminates
any possibility of loop oscillations. The absolute stability makes it suitable for high
power or harmonic measurements of DUTs; the applications requiring high reflec-
tion environments. The higher reflections can be obtained simply by increasing the
output power from the loop amplifier. As a further consequence of the absolute sta-
bility there are no constraints on the magnitude and phase response of the amplifier
as it is the case for closed-loop load-pull system.
Active open-loop load-pull is easily extendible to multi-harmonic setup, for ex-
ample the 3-harmonic load-pull depicted in Fig. 3.20. The sources used for gener-
ating the respective harmonic components of a2 need to be locked to a common
reference in order to maintain coherence between them. Triplexer combines the
78 3 Active Load-Pull Systems

fundamental, a2 (f0 ), second and third harmonic, a2 (2f0 ) and a2 (3f0 ), components
generated by the respective load-pull sources. In this architecture, the appropriate
changes in magnitude and phase of the harmonic components of reflected wave,
a2 , is carried out through the in-built function in the respective load-pull sources.
If load-pull sources are devoid of in-built functions for the control of magnitude
and phase, separate attenuators and phase shifters can be utilized for achieving the
amplitude and phase tuning functionality.
The serious drawback of open-loop active harmonic load-pull setup lies in the
practical limitations for scaling it to higher harmonics due to the exorbitantly high
costs of the load-pull sources. In order to avoid this limitation one must, either em-
ploy single load-pull source capable of generating multi-harmonic components from
their distinct ports or, use doublers and triplers in conjunction with a power divider
as shown in Fig. 3.21. However, in such a configuration, the required output power
from the loop amplifiers increases due to the power division by the divider.
Additionally, the harmonic open-loop active load-pull can’t synthesize harmon-
ically independent reflection coefficients due to interactions of different harmonic
components of a2 . The interactions between the harmonic components disturb the
settings of all the other reflection coefficients while searching for the convergence
of any specific harmonic reflection coefficient. The deflections in the established
harmonic reflection coefficients can be minimized through the use of high speed
algorithm which reduces the number of required iterations for convergence [28],
consequently resulting into reduced interactions between the harmonic frequency
components.

3.9 Convergence Algorithm for Open-Loop and Feed-Forward

Load-Pull Techniques

According to the active feed-forward and open-loop load-pull concept, reflection

coefficients synthesis in these systems is achieved by solving (3.52) for a2h , with h
indicating the harmonic index [28].
a2,h − Γh b2,h (a2,1 , a2,2 , . . . , a2,h ) = 0 (3.52)
Therefore, with no prior knowledge of the function b2,h (.), the solution to
Eq. (3.52) is often found using a numerical technique such as Newton-Raphson
[29]. This is an iterative process and often requires up to 5–10 iterations to converge
to a solution for the fundamental harmonic tone [30]. In the case of harmonic load-
pull, the number of iterations further increases due to change in the fundamental
tone causing distortions in higher harmonics. The problem is further compounded
by the fact that the numerical techniques present their own disadvantages such as
multiple roots and numerical oscillations [29].
Alternatively, Poly Harmonic Distortion Modeling (PHD) based framework [31,
32] is utilized in the determination of solutions to Eq. (3.52) [28]. The PHD models
contain magnitude and phase of spectral components of the input signal and also
3.9 Convergence Algorithm 79

introduce harmonic cross product information, which give the relationships between
harmonic frequencies for a given drive level and frequency. These information and
relationships are then used for achieving solution of Eq. (3.52) in lesser number of
iterations as compared to the numerical techniques.
The PHD modeling technique explained in [31] can be used to describe the output
b2,h of a non-linear DUT as a function of the respective components of reflected
traveling wave, a2,h , at the DUT plane, expressed in Eq. (3.53). Equations (3.52) and
(3.53), when combined, allow for the formulation of a new convergence algorithm,
which is much improved in terms of speed and convergence as compare to numerical
method, for the emulation of reflection coefficient in the feed-forward and open-loop
load-pull.
      ∗ 
b2,h = S21 |a1,1 | |a1,1 | + S22 |a1,1 | a2,h + T22,h |a1,1 | a2,h (3.53)
h h
where, S and T coefficients are functions only of the magnitudes of the stimulus
a-wave, a1 and a2 , fundamental components [33].
The formulation in Eq. (3.53) can be modified, as illustrated in [33], by describ-
ing P and Q as input and output a-wave harmonic phase operators, as shown in
Eqs. (3.54) and (3.55). This expression, a third order model, when generalized pro-
vides the formulation given in Eqs. (3.56) and (3.57) [28].

0
1
Q Q
b2,h = S21 |a1,1 | P+ S22 |a2,h | P
P P
h

−1
Q
+ T22,h |a2,h | P (3.54)
P
h
a1,1 a2,1
P= Q= (3.55)
|a1,1 | |a2,1 |
n=1 

n
Q
b2,h = P R2,h,n (3.56)
P
n=−1
 
R2,h,n = Gh,n |a1,1 |, |a2,1 |, . . . (3.57)
Adapting this generalized formulation and assuming that the magnitude of the input
signal, |a1,1 |, during this process is held constant; one can simplify it to Eq. (3.58)
by considering only the linear third order mixing terms. This is analogous to X-
parameter formulation described in [32].



Q P
b2,h = G2,0,h + G2,1,h |a2,h | + G2,−1,h |a2,h | (3.58)
P Q
In a measurement scenario, G2,0,h can be deduced from the output response of |a1,1 |
at the harmonic being load-pulled. Parameters G2,1,h and G2,−1,h are extracted by
applying a perturbation signal to the incident a2,1 wave, first of all with a zero degree
phase and then followed by the same signal with a 90 degree shift while keeping
|a1,1 | constant in both cases [31]. By utilizing the measured values of a2,h and b2,h
80 3 Active Load-Pull Systems

Fig. 3.22 Flow graph

showing the losses in the
active open-loop load-pull
system and the amplifier gain
[28], © IEEE 2010

at the center and the offset points, indicated by subscripts 0, 1 and 2, one can then
calculate the G parameters using the equation set (3.59)–(3.62).

1 = a2,h,1 − a2,h,0
2 = a2,h,2 − a2,h,0 (3.59)
(
∗2 )(b2,h,1 − b2,h,0 ) + (
∗1 )(b2,h,2 − b2,h,0 )
G2,1,h = (3.60)
(
1 )(
∗2 ) − (
∗1 )(
2 )
b2,h,1 − b2,h,0 − G2,1,h (
1 )
G2,−1,h = (3.61)
(
∗1 )
 ∗ 
G2,0,h = b2,h,0 − G2,1,h (a2,h,0 ) − G2,−1,h a2,h,0 (3.62)
The estimation of b2,h , from Eqs. (3.58)–(3.62), and the knowledge of target re-
flection coefficient, Γh , enables the computation of reflected signal, a2,h , from
Eq. (3.52). If the resulting value of a2,h doesn’t bring sufficient accuracy in the
target reflection coefficient, the process represented by Eqs. (3.53)–(3.62) can be
repeated.
In order to implement this algorithm for the feed-forward or open-loop load-pull
system, one has to include the following practical considerations.
Firstly, the reflected wave predicted using (3.47) does not take into account the
non-ideal behaviors of the load-pull realization. This is best described by the error
model illustrated in Fig. 3.22; where Ts,h accounts for the insertion gain/loss of
the load-pull amplifiers and couplers etc., while ΓL,h accounts for the impedance
mismatch of the measurement system, both of which can be dependent on a2set,h . It
is therefore imperative to re-adjust the reflected wave from calculated value of a2,h
to the compensated value of a2set.h in order to account for the physical state of the
system. To correctly achieve the target reflection coefficient, ΓT ,h , the characterized
system reflection coefficient at the harmonic being load-pulled, ΓL,h , and amplifier
gain, Ts,h , must be incorporated into (3.52) according to Eq. (3.63).
b2,h (ΓT ,h − ΓLoad,h )
a2set,h = (3.63)
Ts,h
Two distinct set of measurements are required, to compute the local model, so as to
solve Eq. (3.52) for prediction of reflected wave, a2,h . As a result, an optimization is
necessary to maximize the use of an existing set of G parameters, provided the input
drive, |a1,1 |, or biasing conditions remain unchanged. If the computed reflection
coefficient is found to be within the acceptable tolerance of the target reflection
coefficient, the existing model would have converged without requiring an update;
this is beneficial during load-pull of a reflection coefficient grid. The efficiency of
this algorithm can be calculated by comparing the number of useful to redundant
measurements.
3.9 Convergence Algorithm 81

Fig. 3.23 Flow chart

illustrating the
implementation of
convergence algorithm, based
on PHD modeling approach,
for the determination of
reflection coefficient in active
feed-forward and open-loop
load-pull systems [28],
© IEEE 2010

Finally, according to Eq. (3.52), any adjustments in the fundamental reflected

wave, a2,1 , distorts higher harmonic components, thereby requiring additional itera-
tions, if they were of interest. This effect can be compensated by utilizing harmonic
cross-product information from the local model generated by the fundamental tone.
For example, during load-pull of the fundamental, the second harmonic may be re-
quired to stay at constant impedance. In this case, the measured output response
of the second harmonic, b2,2 , can be used to calculate the adjustments required in
the reflected wave, a2,2 , to achieve constant reflection coefficient using the process
described above and the equation set (3.59)–(3.62).
The flow chart in Fig. 3.23 summarizes the implementation of the convergence
algorithm, including the optimization and adjustment steps, based on PHD modeling
approach, for the determination of reflection coefficient in active feed-forward and
open-loop load-pull systems. In order to illustrate how this algorithm works, a single
target was considered for emulation using active open-loop load-pull system [28].
This was followed by carrying out measurements to attain multiple targets on a
82 3 Active Load-Pull Systems

Fig. 3.24 Path taken by the

convergence algorithm, based
on PHD modeling approach,
for the emulation of target
impedance in an active
open-loop load-pull systems
[28], © IEEE 2010

Fig. 3.25 Convergence

results on a 5 × 5 square grid
of target impedances
emulated using active
open-loop load-pull system
[28], © IEEE 2010

square grid. The measurements were carried out on a 10 × 75 µm GaAs HEMT,

operating at 3 GHz in Class-B biasing conditions.
Figure 3.24 illustrates the path taken by the algorithm to achieve the target
impedance, while Fig. 3.25 shows the convergence results on a 5 × 5 square grid
of target impedances, and also highlights the points at which the algorithm required
the model to be updated. At the starting point, −0.2 + i0, two perturbations are
made to the a2,1 wave; thereby creating the two offset points, which allow the cal-
culation of a local model. The local model is then used to compute a value of a2set,h
which moves the load to a new position, −0.08 + i0.65, which in this case is not
within the tolerance range, set at 5 %, of the algorithm. The algorithm therefore
requires the model to update itself at this stage, hence the additional set of offset
points. The position of the load now obtained, −0.003 + i0.61, is within 1 % of the
target, 0 + 0.6i; implying the algorithm has now converged to a solution.
3.10 Comparison of Active Load-Pull Techniques 83

Table 3.2 Comparison of the three types of active load-pull systems

Characteristics Closed-loop Feed-forward Open-loop

Dynamic range limitation Low Medium High

(ΓLoad )min Limited No limitation No limitation
(ΓLoad )max 1 1 1
Adjustment of ΓLoad Fast Slow Slow
Oscillations possibility High Low None
Costs Low Medium High
Harmonic tuning Independent Dependent Dependent

The performance of this convergence algorithm can be quantified in terms of its

efficient use of the load-pull system. An ideal system, 100 % efficient, would re-
quire only one measurement per impedance point. The efficiency of the algorithm,
based on PHD modeling, to converge to a solution is 44 % for the emulated 5 × 5
square impedance grid considered in this case, averaging 2.26 measurements per
impedance point. It is significantly better as compared to Newton-Raphson conver-
gence algorithm whose typical efficiency is 5 % to 10 % [30].

3.10 Comparison of Active Load-Pull Techniques

The characteristics of the three types of active load-pull systems are summarized in
Table 3.2.
The dynamic range of the closed-loop system is limited by the requirement of flat
gain from loop amplifier in order to avoid loop oscillations. The feed-forward and
open-loop systems are not constrained by flat gain loop amplifier and therefore pos-
sess higher dynamic range. The feed-forward load-pull might get into oscillations
under some conditions and therefore open-loop has advantage while comparing the
dynamic range of these two.
In the closed-loop system, synthesis of smaller magnitude of reflection coeffi-
cients strongly depends on the resolution of phase shifter and attenuator settings.
The feed-forward and open-loop systems are not limited by this drawback and can,
therefore, synthesize any smallest required reflection coefficient. Ideally the max-
imum synthesizable reflection coefficient can be higher than one, but in practical
applications (ΓLoad )max equal to one is sufficient.
Reflection coefficient synthesis in closed-loop system is fast, whereas in the other
systems is slow due to iterative nature of synthesis process. Oscillations in closed-
loop system are inherent and therefore special measures are needed to prevent or
mitigate them. The feed-forward system has slight possibility of occurrence of oscil-
lations during the characterization of high power DUTs, while the open-loop system
is completely free from loop oscillations.
Harmonic reflection coefficients are dependent on each other in the feed-forward
and open-loop systems, whereas they are independent in closed-loop system.
84 3 Active Load-Pull Systems

References

1. J.-E. Muller, B. Gyselinckx, Comparison of active versus passive on-wafer load-pull character-
ization of microwave and MM-wave power devices, in IEEE MTT-S International Microwave
Symposium Digest (June 1994), pp. 1077–1080
2. Maury Microwave Corporation, Pulsed-bias pulsed-RF harmonic load-pull for gallium nitride
(GaN) and wide band gap (WBG), Application Note: 5A-043, Nov. 2009
3. C. Roff, J. Benedikt, P.J. Tasker, Design approach for realization of very high efficiency power
amplifiers, in IEEE MTT-S International Microwave Symposium Digest (June 2007), pp. 143–
146
4. G.P. Bava, U. Pisani, V. Pozzolo, Active load technique for load-pull characterization at mi-
crowave frequencies. IEE Electron. Lett. 18(4), 178–180 (1982)
5. Y. Takayama, A new load pull characterization method for microwave power transistors, in
IEEE/MTT-S International Microwave Symposium, New Jersey, USA (June 1976), pp. 218–
220
6. V. Teppati, A. Ferrero, U. Pisani, Recent advances in real-time load-pull systems. IEEE Trans.
Instrum. Meas. 57, 11 (2008)
7. M. Spirito, L.C.N. de Vreede, M. de Kok, M. Pelk, D. Hartskeerl, H.F.F. Jos, J.E. Mueller, J.
Burghartz, A novel active harmonic load-pull setup for on-wafer device characterization, in
IEEE/MTT-S International Microwave Symposium (June 1994), pp. 1217–1220
8. J. Benedikt, R. Gaddi, P.J. Tasker, M. Goss, M. Zadeh, High power time domain measurement
system with active harmonic load-pull for high efficiency base station amplifier design, in
IEEE/MTT-S International Microwave Symposium. Boston, USA (2000), pp. 1459–1462
9. D.D. Poulin, J.R. Mahon, J.-P. Lanterri, A high power on-wafer pulsed active load-pull system.
IEEE Trans. Microw. Theory Tech. 40(12), 2412–2417 (1992)
10. G.D. Vendelin, A.M. Pavio, U.L. Rohde, Microwave Circuit Design Using Linear and Non-
linear Techniques (Wiley, New York, 1990), p. 400
11. V. Camarchia, V. Teppati, S. Corbellini, M. Pirola, Microwave measurements part II—non-
linear measurements. IEEE Instrum. Meas. Mag. 10, 34–39 (2007)
12. Y.Y. Woo, Y. Yang, B. Kim, Analysis and experiments for high efficiency class-F and inverse
class-F power amplifiers. IEEE Trans. Microw. Theory Tech. 54(5), 2006 (1969–1974)
13. P. Colantonio, F. Giannini, R. Giofre, E. Limiti, A. Serino, M. Peroni, P. Romanini, C. Proietti,
A C-band high efficiency second harmonic tuned power amplifier in GaN technology. IEEE
Trans. Microw. Theory Tech. 54(6), 2713–2722 (2006)
14. M. Helaoui, F.M. Ghannouchi, Optimizing losses in distributed multi-harmonic matching net-
works applied to the design of an RF GaN power amplifier with higher than 80 % power-added
efficiency. IEEE Trans. Microw. Theory Tech. 57(2), 314–322 (2009)
15. F.M. Ghannouchi, M.S. Hashmi, Experimental investigation of the uncontrolled higher har-
monic impedances effect on the performance of high-power microwave devices. Microw. Opt.
Technol. Lett. 52(11), 2480–2482 (2010)
16. A. Ferrero, Active load or source impedance synthesis apparatus for measurement test set of
microwave components and systems, U.S. Patent 6 509 743, Jan. 21, 2003
17. V. Teppati, A. Ferrero, A new class of non-uniform, broadband, nonsymmetrical rectangular
coaxial-to-microstrip directional couplers for high power applications. IEEE Microw. Wirel.
Compon. Lett. 13(4), 152–154 (2003)
18. S. Dudkiewics, R. Meierer, Cascading tuners for high-VSWR and harmonic load-pull, Maury
Microwave Corporation, Application Note: 5C-081, Jan 2009
19. D.M. Pozar, Microwave Engineering, 3rd edn. (Wiley, New York, 2005). ISBN 0-471-17096-
8
20. P. Bouysse, J.M. Nebus, J.M. Coupat, J.P. Villotte, A novel, accurate load-pull setup allowing
the characterisation of highly mismatched power transistors. IEEE Trans. Microw. Theory
Tech. 42(2), 327–332 (1994)
References 85

21. J. Verspecht, F. Verbeyst, M.V. Bossche, Network analysis beyond S-parameters: characteriz-
ing and modeling component behavior under modulated large-signal operating conditions, in
56th ARFTG Conference Digest (Nov. 2000), pp. 1–4
22. J. Verspecht, Large-signal network analysis. IEEE Microw. Mag. 6(4), 82–92 (2005)
23. K. Rawat, F.M. Ghannouchi, A design methodology for miniaturized power dividers using
periodically loaded slow wave structure with dual-band applications. IEEE Trans. Microw.
Theory Tech. 57(12), 3380–3388 (2009)
24. F. Blache, J.M. Nebus, P. Bouysse, J.-P. Villotte, A novel computerized multiharmonic active
load-pull system for the optimization of high efficiency operating classes in power transistors,
in IEEE/MTT-S International Microwave Symposium, USA (1995), pp. 1037–1040
25. J.-M. Nebus, P. Bouysse, J.-P. Villotte, J. Obregon, Improvement of active load-pull technique
for the optimization of high power communication SSPAs. Int. J. Microw. Millimeter-wave
Computer-Aided Eng. 5(3), 149–160 (1995)
26. J.-M. Nebus, P. Bouysse, J.M. Coupat, J.-P. Villotte, An active load-pull setup for the large
signal characterization of highly mismatched microwave power transistors, in IEEE Instru-
mentation and Measurement Technology Conference, Irvine, USA (1993), pp. 2–5
27. F.V. Raay, G. Kompa, Waveform measurements—the load-pull concept, in 55th ARFTG Con-
ference, Boston, USA (2000), pp. 1–8
28. R.S. Saini, S. Woodington, J. Lees, J. Benedikt, P.J. Tasker, An intelligence driven active load-
pull system, in 75th ARFTG Microwave Measurement Conference, Anaheim, USA (2010), pp.
1–4
29. S.C. Chapra, Numerical Methods for Engineers, 6th edn. (McGraw Hill, New York, 2009)
30. D.J. Williams, P.J. Tasker, An automated active source and load pull measurement system, in
Proceedings of 6th IEEE High Frequency Postgraduate Colloquium, Cardiff, UK (2001), pp.
7–12
31. J. Verspecht, D.E. Root, Poly-harmonic distortion modeling. IEEE Microw. Mag. 7(3), 44–57
(2006)
32. J. Horn, D. Gunyan, L. Betts, J. Verspecht, D.E. Root, Measurement-based large-signal sim-
ulation of active components from automated nonlinear vector network analyzer data via X-
parameters, in IEEE International Conference on Microwaves, Communications, Antenna and
Electronic Systems (May 2008), pp. 1–6
33. S. Woodington, T. Williams, H. Qi, D. Williams, L. Pattison, A. Patterson, J. Lees, J. Benedikt,
P.J. Tasker, A novel measurement based method enabling rapid extraction of a RF waveform
look-up table based behavioral model, in IEEE/MTT-S International Microwave Symposium,
Boston, USA (June 2008), pp. 1453–1456
Chapter 4
Six-Port Based Load-Pull System

The previous couple of chapters explain the theory, concept and measurement tech-
niques of active and passive load-pull systems. This chapter provides detailed treat-
ment of six-port based setups for passive and active load/source-pull measurements,
as well as a description of six-port based harmonic load/source-pull measurements.
The load-pull measurements for on-wafer devices require special attention; there-
fore, an alternative approach for impedance and power flow calibration using a re-
flection based technique is described.

4.1 Introduction

As mentioned in previous chapters, the load/source-pull systems experimentally de-

termine the performances of a device in large-signal operations and identify the
optimal loading conditions that yield the desired performance, in terms of power,
power efficiency or linearity or a trade-off among them. Numerous active and pas-
sive load/source-pull setups have been developed and used in the past for such mea-
surements. Six-port (SP) junctions, extensively used for load-/source-pull configu-
rations [1–9], have been very flexible in their usage and applications.
In theory, any load-pull system should be able to measure the input impedance
of a device under test (DUT), the load impedance seen by the DUT at its output
(ZL ), the source impedance presented to the DUT at its input (ZS ), the input power
absorbed by the DUT (Pin ), and the power delivered to load (PL ). In addition, the
load-pull system also monitors the voltages and currents across both the input and
output of the DUT for any pair of (ZS , ZL ), in order to plot the power-added effi-
ciency (PAE) in the Smith chart.
In some circumstances, measurement and monitoring of the linearity of the de-
vice is also required. Several metrics of linearity are used depending on the appli-
cation, such as amplitude dependent amplitude distortion (AM/AM), amplitude de-
pendent phase distortion (AM/PM), carrier to third-order intermodulation products
(C/IMD3), the adjacent channel power ratio (ACPR) and the error vector magnitude

F.M. Ghannouchi, M.S. Hashmi, Load-Pull Techniques with Applications to Power 87

Amplifier Design, Springer Series in Advanced Microelectronics 32
DOI 10.1007/978-94-007-4461-5_4, © Springer Science+Business Media Dordrecht 2013
88 4 Six-Port Based Load-Pull System

Fig. 4.1 A typical SP based reflectometer with variable test port impedance

(EVM). An SP based load-pull setup enables all these measurements and, therefore,
possesses an appropriate and attractive position in the load/source-pull domain.

4.2 Impedance and Power Flow Measurement

Measurements of scattering parameters (S-parameters) using an SP based reflec-

tometer, such as shown in Fig. 4.1, were reported almost three decades ago [10, 11].
The measurements of complex S-parameters of an N-port network require 4 N
power readings [11, 12]. The phases of the S-parameters are directly obtained from
radio frequency (RF) signals. In addition, it has been shown that the calibration and
measurement calculations are not functions of the source impedance presented to the
input port of the SP reflectometer; therefore, one could change the source impedance
by using a passive tuner inserted before the input port of the SP reflectometer [12].
The power flow exciting the DUT connected to the measuring port of the SP
junction, as shown in Fig. 4.1, can be calculated using the following expressions
[12]:
w(f ) − e(f )
ΓDUT (f ) = (4.1)
−c(f )w(f ) + d(f )
k(f )PRef (f )
PDUT (f ) = (4.2)
|1 + c(f )ΓDUT (f )|2
where c(f ), d(f ) and e(f ) are the error box parameters, w(f ) is the embedded
reflection coefficient, k(f ) is the power calibration parameter that has to be deter-
mined at each measurement’s frequency, PRef (f ) is the power reading measured
by the reference port of the SP junction, and ΓDUT (f ) is the reflection coefficient
measured by the SP reflectometer.
In Fig. 4.1, ΓS (f ) is the source impedance seen by the DUT. The complex value
of ΓS (f ) is related to the tuner impedance, Γt (f ), and the S-parameters, sij , of the
two-port network delimited by the input and output ports of the SP junction when
4.3 SP in Reverse Configuration 89

the remaining ports of the SP junction are terminated by the 50  power sensors
[12].
s12 (f )s21 (f )Γt α(f )Γt + β(f )
ΓS (f ) = s22 (f ) + = (4.3)
1 − s11 (f )Γt μ(f )Γt + 1
where α(f ), β(f ) and μ(f ) are directly related to the two-port network delimited
by the input and output reference planes of the SP junction.
If the SP junction is designed to be “transparent” in such a way that the input and
measuring port of the SP junction are both situated in the direct line of the 50 
(|s12 | = |s21 | ∼
= 1; |s11 | ∼
= |s22 | ∼
= 0), a good approximation of the ΓS (f ) is given
by [12]:
   
ΓS (f ) ∼
= Γt (f ) (4.4)
It is thus evident from Eqs. (4.2), (4.3), (4.4) that the SP configuration can facili-
tate impedance and power flow measurements on the fly for arbitrarily changing the
impedance presented to the DUT. This assumes that the tuner is pre-calibrated and
that the impedance corresponding to Γt is known for any position of the stubs of the
used tuner.

4.3 SP in Reverse Configuration

An SP junction is usually built with physically assigned input, output and power
reference ports and three other power detection ports. The normal use of an SP
reflectometer requires the connection of the signal generator at the input port and
the DUT to the output port of the SP junction, as shown in Fig. 4.2(a) [13]. In such a
configuration, microwave signals are injected from the input to the output; and, the
SP reflectometer measures the reflection coefficient of the DUT and the power flow
at the output port according to Eqs. (4.1) and (4.2). In this measurement setup, also
known as a direct configuration, the signal generator impedance can be arbitrary and
has no effect on the SP calibration and measurement calculations [11].
Alternatively, if the signal generator is connected to the output and the DUT
to the input of the SP reflectometer, the SP junction is considered to be in a reverse
configuration, as shown in Fig. 4.2(b) [13]. In a reverse configuration, the microwave
signal is injected from the output port to the input port. In this configuration, the SP
measures the reflection coefficient of the DUT, seen at the reference plane located
at the output port (port 2) of the SP junction.

4.3.1 SP Calibration in Reverse Configuration

The calibration procedure of the SP reflectometer used in a reverse configuration is

divided into the following two steps [13]:
90 4 Six-Port Based Load-Pull System

Fig. 4.2 SP reflectometers: (a) direct, (b) reverse configuration [13]; © IEEE 1994

• The common SP to four-port reduction technique using thirteen unknown and

well-distributed loads
• A new error box procedure using three well-known standards
The determination of the error box parameters are based on those obtained in the
direct configuration and on some analytical manipulations [12]. Figure 4.3 depicts
the different steps required for the calibration, de-embedding and measurement pro-
cedures of the reverse SP reflectometer. The calibration and de-embedding proce-
dures characterize the SP junction and fix the reference plane for the measurements
shown in Figs. 4.3(a) and 4.3(b). A self-calibration approach [14, 15], contrary to
an explicit calibration method [11], can be used in this situation.
It is important to note that a −10 dB directional coupler is added to the output in
the de-embedding procedure. This coupler allows the use of the SP junction in the
reverse configuration, as shown in Fig. 4.3(c). This reverse configuration consists
of a generator connected to the output port of the SP reflectometer via the −10 dB
directional coupler. The SP junction measures the reflection coefficient of the tuner,
which is shifted from the input port (port 1) to the reference plane at the output port
(port 2).
4.3 SP in Reverse Configuration 91

Fig. 4.3 Reverse SP reflectometer measurements: (a) calibration in the direct configuration,
(b) de-embedding in direct configuration, (c) measuring in the reverse configuration [13]; © IEEE
1994

The microwave signal injected through the coupler excites the DUT. A part of
this signal is reflected toward the SP junction, enabling it to measure the reflection
coefficient of the tuner connected to the input port. It is assumed that the SP junction
is transparent (|s12 | = |s21 | ∼
= 1; |s11 | ∼
= |s22 | ∼
= 0), in order to minimize the insertion
loss and allow the tuner to cover a wider area on the Smith chart. The measured tuner
reflection coefficient is shifted to the reference plane defined by the de-embedding
procedure at the right-hand side of the SP junction and directional coupler. As the
use of an SP reflectometer in a reverse configuration requires calibration of SP re-
flectometer in a direct configuration, it is essential to establish a relation between
the calibration and the de-embedding of the direct and reverse SP junctions.
92 4 Six-Port Based Load-Pull System

The power readings at the SP junctions are given by following expressions:

PRef = P3 = β3 |1 + ξ3 ΓDUT |2 |b2 |2 (4.5)

Pi = βi |1 + ξi ΓDUT |2 |b2 |2 , i = 4, 5, 6 (4.6)
where βi are real constants, ξi are complex constants that characterize the SP junc-
tion, b2 is the outgoing wave at the output port (port 2, Fig. 4.3), and ΓDUT is the
reflection coefficient of the DUT.
The ratio of Pi /P3 can be written as [15]:
P4
= α4 |ΓDUT − Q4 |2 (4.7)
Pref
P5
= α5 |ΓDUT − Q5 |2 (4.8)
Pref
P6
= α6 |ΓDUT − Q6 |2 (4.9)
Pref
where αi and Qi are complex constants.
The reflection coefficient ΓDUT of the load is determined by the intersection of
three circles using the six-to-four port reduction technique [12]:
⎡ P4 ⎤
    − α4 |Q4 |2
|ΓDUT |2 α4 −α4 Q∗4 −α5 Q4 −1 Pref
∗ ⎢ P5 − α |Q |2 ⎥
ΓDUT = α5 −α5 Q∗5 −α5 Q5 ⎣ Pref 5 5 ⎦ (4.10)
ΓDUT α6 −α6 Q∗6 −α6 Q6 P6
Pref − α6 |Q6 |
2

In Eq. (4.10), the product of rows two and three is equal to row one. Therefore,
after simplification of Eq. (4.10), one can obtain the expression given in (4.11):
 2       
P4 2 P5 2 2 P6 2 P4 P5
p + q A2 + r B2 + (r − p − q)A2
Pref Pref Pref Pref Pref
     
2 P4 P6 2 2 P5 P6
+ (q − p − r)B + (p − q − r)A B
Pref Pref Pref Pref
   
P4 P 5
+ p(p − q − r) + q(q − p − r)A2
Pref Pref
 
2 P6
+ r(r − p − q)B + pqr = 0 (4.11)
Pref

where the real quantities of p, q, r, A2 and B 2 are the calibration constants related
to αi and Qi and associated with the physical structure of the SP junction [14].
These parameters are determined by the six-port to four-port reduction formulation.
For each specified frequency, at least five measurements for different distributed
loads, leading to a set of five nonlinear equations, are needed to solve this set of
equations [14, 15]. The explicit knowledge of the impedance values of these dis-
tributed loads is not required in the calculations of p, q, r, A2 and B 2 .
4.3 SP in Reverse Configuration 93

As mentioned earlier, in practice, thirteen loads are employed to improve the

accuracy of the calibration over the entire Smith chart [12]. Furthermore, the elim-
ination of |b2 |2 from the expression makes the quantities of p, q, r, A2 and B 2
completely independent of the choice of port 1 or port 2 for the test port of the SP
junction. Hence, if the power reference and the remaining detection ports are fixed,
the calibration of the SP junction, which is aimed at the determination of p, q, r, A2
and B 2 , normally gives the same value of these parameters for an SP reflectometer
used in either a direct or reverse configuration.

4.3.2 Error Box Calculation

Once the calibration constants are determined, the SP reflectometer can measure an
embedded reflection coefficient, w = g(Pi /Pref , p, q, r, A2 , B 2 ), given in Eq. (4.1)
at an unknown reference plane [14]. The error box calculation (or de-embedding
procedure) determines the three complex constants (c, d and e), which relate w to
the reflection coefficient, Γ , at a chosen reference plane with the following relation:
dΓ + e
w= (4.12)
cΓ + 1
With three well-known standards and their measured embedded reflection coeffi-
cients, wi , given by the SP reflectometer, one can easily find c, d and e by solving
the following system of linear equations:
w1 Γ1std c − Γ1std d − e = −w1

w2 Γ2std c − Γ2std d − e = −w2 (4.13)

w3 Γ3std c − Γ3std d − e = −w3

Referring to Fig. 4.3(c), for a reverse SP configuration, the microwave signal
must be injected from the output port (port 2) to the input port (port 1); and, there is
a need for three de-embedding standards at the left side of the reference plane, which
is physically impossible. To overcome this difficulty, a direct SP error box, as shown
in Fig. 4.3(b), with the three standards connected successively to the measuring
port at the right side of the reference plane is used. For the reverse SP error box
calculation, Eq. (4.13) changes to the following set of equations:
c d
w1 − − e = −w1
Γ1std Γ1std
c d
w2 − − e = −w2 (4.14)
Γ2std Γ2std
c d
w3 − − e = −w3
Γ3std Γ3std
94 4 Six-Port Based Load-Pull System

where the value Γistd of the three standards in (4.13) are replaced by 1/Γistd in
(4.14), since it is equivalent to using Γistd on the left side of the reference plane,
which is valid when Γ is referred to the same characteristic impedance (50 ).
In practice, it is usually convenient to use a short circuit, an open circuit and a
matched load of 50  (Γ = 0) as the three de-embedding standards. Unfortunately,
for a reverse SP error box procedure, the matched load needs to be replaced by
another standard to avoid the l/Γ singularity. To circumvent this difficulty, one can
determine the reverse SP error box by first finding coefficients c, d and e of the
conventional error box procedure and then deducing coefficients c , d  and e of
the reverse SP error box using the following relations [12], which are derived by
comparing (4.13) and (4.14):
c = 1/c; d  = 1/d; e = 1/e (4.15)
Measurement results obtained by a reverse SP calibrated and de-embedded accord-
ing to the above procedures confirm the validity of this technique [13]. The reverse
SP reflectometer is suitable to perform source-pull measurements at the input of the
DUT. The measurement technique consists of a signal injected into the DUT via the
directional coupler. The reflected signal from the input port of the DUT allows the
SP reflectometer to perform measurements in the opposite direction. It should be
mentioned that, when the DUT is well matched, the SP reflectometer is not powered
and will not operate. However, source-pull measurements are only useful when the
DUT input is mismatched.

4.3.3 Discussion

Considering Eqs. (4.12) and (4.15), it can be seen that Γ  = 1/Γ , where Γ  and
Γ are the reflection coefficients measured by an SP reflectometer in the reverse
and direct configurations, respectively. It may seem that this relation can be equally
obtained by just taking Γ = b/a and Γ  = a/b (where a and b are the incident and
reflected waves, respectively), which is not true. For example, in the case of a direct
SP measurement, 1/Γmeas does not lead to the impedance of the generator and has
no physical significance. In a direct SP configuration, the DUT is connected to port 2
of the SP reflectometer, which measures the ratio of a2 /b2 , i.e., the incoming and
outgoing waves at the output of the junction (port 2, Fig. 4.2). This ratio represents
the reflection coefficient of the DUT seen from the left side of the reference plane.
In a reverse SP configuration, the reference plane is always located at port 2, but the
output port of the reflectometer becomes port l.
The reverse SP junction, therefore, measures the ratio of a1 /b1 rather than that
of a2 /b2 . Ratio a1 /b1 , which is shifted to port 2 of the SP junction by the de-
embedding procedure, then represents the reflection coefficient of the test port of
the SP reflectometer seen from the load side of the reference plane. Therefore, the
inverse value of a1 /b1 cannot yield the reflection coefficient of the DUT connected
to port 2.
4.4 SP Based Source-Pull Configuration 95

Fig. 4.4 A passive source-pull measurement system based on an SP reflectometer

In summary, reverse SP reflectometer measurements require a conventional cal-

ibration procedure to characterize the SP junction and a special error box to fix the
reference plane. This error box can be obtained either by using the relations given in
Eq. (4.15) or by inverting the actual reflection coefficient, Γ , of the standards used
to calculate parameters c, d and e of the direct error box [12].

4.4 SP Based Source-Pull Configuration

The source-pull measurement is aimed at evaluating the linear and nonlinear device
performances as functions of source impedance [8]. In source-pull measurement,
the input impedance is varied, in order to determine the DUT behavior as a function
of source impedance. The source-pull measurement can be carried out using an SP
technique, in both passive and active systems.
Figure 4.4 presents a typical passive source-pull system based on an SP reflec-
tometer. In this setup, the input signal is injected to the DUT via a directional cou-
pler, and the source impedance is varied with the help of a stub tuner. The SP reflec-
tometer in such an arrangement is operated in reverse configuration and measures
the source impedance seen at the reference plane.
The setup in Fig. 4.5 depicts a typical setup of an active source-pull measure-
ment system based on an SP reflectometer. This setup utilizes a variable attenuator
and a variable phase shifter for varying the magnitude and phase of the reflection
coefficient being synthesized. The settings of the attenuator and phase shifter can be
altered during measurement applications. The SP reflectometer in this setup is used
in reverse configuration, allowing for the measurement of the source impedance
seen at the reference plane.
It is important to note that the active source-pull configuration has more flexi-
bility and greater accuracy in generating synthesized reflection coefficients, as it is
dependent on the settings of the phase shifter and attenuator. The development of
96 4 Six-Port Based Load-Pull System

Fig. 4.5 A typical setup of active source-pull measurement system based on an SP reflectometer

more accurate and automated passive tuners, however, has made passive load-pull
equally flexible and accurate.

4.5 SP Based Load-Pull Configuration

Both passive and active load-pull measurement setups can be realized using an SP
technique. The concept of active and passive load-pull measurements is similar to
the more conventional methods described in Chaps. 2 and 3. The only difference
lies in the measurement of the reflection coefficient. The more conventional load-
pull systems capture the traveling waves directly, whereas the SP based load-pull
depends on the ability of the SP reflectometer to measure the de-embedded reflection
coefficients.

4.5.1 Passive Load-Pull System

The SP based passive load-pull configuration, as shown in Fig. 4.6, includes two SP
junctions, SP1 and SP2, and two tuners, T1 and T2. This configuration allows for
the determination of the large-signal input impedance, Γin (f ), of the DUT and the
power delivered to it, Pin (f ), by means of SP1, according to Eqs. (4.16) and (4.17).
The source impedance seen by the DUT at its input can be changed by varying the
stub positions of T1.
Γin (f ) = b1 /a1 = Γ1 (f ) (4.16)

1  k1 f )PRef 1 (f )(1 − |Γin (f )|2 )
Pin (f ) = |a1 | − |b1 | =
2 2
(4.17)
2 |1 + c1 (f )Γin (f )|2
4.5 SP Based Load-Pull Configuration 97

Fig. 4.6 SP based passive load-pull measurement configuration

The determination of source impedance, Γs (f ), is not directly possible using con-

ventional SP calibration and measurement approaches in either a direct or reverse
configuration. Therefore a straightforward technique to solve this problem is the
pre-calibration of T1 using a vector network analyzer (VNA), which provides the
impedance of the tuner’s port connected to the DUT as a function of the tuner’s stub
positions. This technique requires a repeatable and high-quality computer-driven
tuner [12].
The reflection coefficient associated with the load impedance and the power ab-
sorbed by the load are measured with SP2 operated in direct configuration, where
port 1 is connected to the DUT and port 2 to the tuner T2, and can be calculated as
follows:

ΓDUT (f ) = a2 (f )/b2 (f ) (4.18)

1  k2 (f )PRef 2 (f )(1 − |ΓDUT (f )|2 )
PDUT (f ) = |b2 |2 − |a2 |2 = (4.19)
2 |1 + c2 (f )ΓDUT (f )|2
SP2 operated in a direct configuration enables the determination of ΓDUT according
to the following expression:
αΓ2 + β
ΓDUT = (4.20)
δΓ2 + 1
where α, β and δ are complex coefficients modeling the two-port network delimited
by the reference planes where ΓDUT and Γ2 are defined and measured.
Coefficients α, β and δ can be determined beforehand using the short, open and
load de-embedding technique [12]. Section 4.6 of this chapter provides the neces-
sary equations to de-embed both the reflection coefficient and the power flow mea-
surements between different reference planes and through different error networks.

4.5.2 Active Branch Load-Pull System

The active branch load-pull measurement system is basically an SP network an-

alyzer that also includes two SP junctions, including phase and amplitude signal
98 4 Six-Port Based Load-Pull System

Fig. 4.7 SP based active load-pull (a) measurement configuration and (b) measurement system

controllers inserted in one branch as shown in Fig. 4.7(a). The two SP junctions are
calibrated independently for impedance and power flow measurements in a reverse
configuration. The active load presented to the DUT, ZDUT (f ), can be varied by
changing the phase and amplitude of the injected signal, a2 (f ), at the output of the
DUT.
Equations (4.21) and (4.22) are used to determine the complex value of ZDUT (f )
from the measured reflection coefficient, Γ2 (f ), obtained from the second SP reflec-
tometer, SP2 [12].
1 + ΓDUT (f )
ZDUT (f ) = Z0 (4.21)
1 − ΓDUT (f )
1
ΓDUT (f ) = (4.22)
Γ2 (f )
The power absorbed by the active load-pull setup and presented to the DUT can be
calculated using the following expression:
 2
PDUT (f ) = PF 2 (f )Γ2 (f ) (4.23)
4.6 On-Wafer Load-Pull Measurements 99

Fig. 4.8 SP active loop load-pull measurement system

4.5.3 Active Loop Load-Pull System

An alternative technique for realizing an SP based active load-pull setup is the active
loop method, as shown in Fig. 4.8.
In this technique, a directional coupler, called a loop coupler, is placed at the
output of the DUT and takes part of the output signal and sends it to a variable
attenuator, a phase shifter, a tunable filter and an amplifier. The loop also uses a
circulator to re-inject the amplified signal back to the DUT output. In this config-
uration, the reflection coefficient provided to the DUT does not change with the
device output power if the loop amplifier is in the linear region and has a constant
gain [16].

4.6 On-Wafer Load-Pull Measurements

An absolute de-embedding technique for simultaneous impedance and power mea-
surements suitable for on-wafer load-pull measurements of microwave transistors is
required [6–9, 17]. This allows extraction of the impedance and power calibration of
a network analyzer directly from the embedded reflection coefficient measurements.
The de-embedding technique requires two sets of open-short-load (OSL) calibration
standards, one coaxial and one non-coaxial, such as a coplanar waveguide (CPW).
An absolute power calibration performed at the coaxial reference planes using a
standard power meter are also required for accurate power flow measurement pur-
poses at the coplanar reference planes of the microwave probes needed for on-wafer
measurements.
A general formulation of the on-wafer de-embedding problem is illustrated in
Fig. 4.9, where (A, B) two error box networks are delimited by the network ana-
lyzer reference planes and the coaxial reference planes, and (A , B  ) are two error
box networks delimited by the coaxial reference planes and the coplanar references
planes at the tips of the microwave probes. These error boxes are represented by
transmission matrices ([A], [B]) and ([A ], [B ]).
100 4 Six-Port Based Load-Pull System

Fig. 4.9 On-wafer impedance and absolute power calibration procedure

In Fig. 4.9, the reflection coefficients at both the coaxial and coplanar reference
planes are related to the reflection coefficients at the network analyzer reference
planes and coaxial reference planes, respectively, as follows [12]:
α1 ΓA + α2
ΓA = (4.24)
α3 ΓA + 1
α  Γ  + α
ΓA = 1 A 2
(4.25)
α3 ΓA + 1
β1 ΓB + β2
ΓB = (4.26)
β3 ΓB + 1
β  Γ  + β
ΓB = 1 B 2
(4.27)
β3 ΓB + 1
where αi , αi , βi and βi are the error box parameters that can be calculated using
four OSL calibrations, two at the coaxial references planes and two at the coplanar
references planes.
The two power flows at the coaxial reference planes can be calculated as follows
[6]:
 P
kA
PA =
SA
(4.28)
|1 + α3 ΓA |2


kB PSB
PB = (4.29)
|1 + β3 ΓB |2
where PSA and PSB are two power readings sampled by directional couplers at coax-
ial reference planes. These directional couplers can be placed within the network
analyzer or can be inserted between the coaxial reference planes and the coplanar
 and k  can be calculated by connect-
reference planes. Power calibration factors kA B
ing an absolute power meter successively at both coaxial reference planes as follows
[12]:

 PAPM |1 + α3 ΓAPM

 |2
kA =  (4.30)
PSAPM (1 − |ΓAPM |2 )

PBPM |1 + β  ΓBPM |2
kB =  (4.31)
PSBPM (1 − |ΓBPM |2 )
4.7 Applications of Source-Pull Setup 101

where PAPM and PBPM are the power readings of a standard power meter when it
is connected successively to the coaxial references planes, PSAPM and PSBPM are
the two sampled power readings, and ΓAPM 
and ΓBPM are the measured reflection
coefficients when the standard power meter is connected successively to the coaxial
reference planes.
The power flows at the coplanar reference planes can be calculated from [12]:
kA PSA
PA = (4.32)
|1 + α  ΓA |2
kB PSB
PB = (4.33)
|1 + β  ΓB |2
where the value of kA , kB , α  and β  can be calculated using the following ex-
pressions and without any need to perform extra power calibration at the coplanar
reference planes [12]:
|α1 − α2 α3 |kA
kA = (4.34)
|1 + α3 ΓA |2
α3 + α3 α1
α  = (4.35)
α3 α2 + 1
|β1 − β2 β3 |kB
kB = (4.36)
|1 + β3 ΓB |2
β3 + β3 β1
β  = (4.37)
β3 β2 + 1
It is evident that this technique requires two OSL calibration standards, one coax-
ial and one coplanar, for on-wafer de-embedding. A coaxial power detector is also
needed for power flow calibration purposes. The main advantage of this technique
is that a second power calibration at the tips of the coplanar probes is not necessary.
Such power calibration is almost impossible to perform, because there is no accurate
coplanar power sensor that can be connected to the tip of the probes to calculate kA
and kB .

4.7 Applications of Source-Pull Setup

In certain situations, such as in the characterization of mixers, oscillators or noise
measurements, it is more desirable to vary the generator impedance and measure its
reflection coefficient than to measure the reflection coefficient of the DUT. These
applications cannot be carried out using either an automatic VNA or the conven-
tional SP configuration.
This section contains measurement applications that specifically require an SP
reflectometer in reverse configuration. In such a configuration, the microwave source
and DUT are placed at the output of the SP junction; and, the measurement reference
plane also lies at the output of the SP junction. A tuner is connected at the input of
102 4 Six-Port Based Load-Pull System

Fig. 4.10 A typical noise characterization setup based on reverse SP source-pull configuration
[17], © IEEE 1995

the SP junction. In this arrangement, the SP reflectometer can simultaneously drive

the input of the DUT and perform source-pull measurements.

4.7.1 Low Noise Amplifier Characterization

An important part of low noise amplifier design is the measurement of the transis-
tor’s noise figure and the determination of noise parameters. Figure 4.10 depicts a
typical setup for noise characterization that is based on a reverse SP source-pull
configuration [17].
The SP junction is calibrated at all test frequencies, so that it is capable of accu-
rately measuring the source impedances seen by the DUT. During noise measure-
ments, the microwave generator must be turned off when the noise source is on. The
noise figure can be determined using the noise source, a noise meter and a standard
measurement procedure [18, 19] and calculated using Eq. (4.38). It is important to
note that the stability of the DUT is extremely important during characterization.
 
GDUT Nin + Nadded
NF DUT = 10 log (4.38)
GDUT Nin
where Nadded is the difference between the measured output noise level and the
noise level at the input of the device, GDUT is the gain of the DUT, and Nin is the
available noise power at the input of the DUT.
The advantage of SP source-pull based noise measurement and characterization
is that the SP junction in reverse configuration gives the source impedance for any
tuner position without any need to pre-calibrate the tuner using a network analyzer.
In addition, this setup does not require a highly repeatable tuner.
4.7 Applications of Source-Pull Setup 103

Fig. 4.11 Experimental setup for the source-pull characterization of a MESFET (metal semicon-
ductor field effect transistor) gate mixer [13], © IEEE 1994

4.7.2 Mixer Characterization

The nonlinearity of the active device mainly contributes to the generation of in-
termodulation products in the mixers. In essence, the behavior of mixers depends
strongly on the load and the local oscillator (LO) and RF source impedance termi-
nations. Computer-aided design approaches using harmonic balance methods allow
for the optimization of the input and output matching impedances, but the accuracy
of the analysis relies on the nonlinear models of the device.
In practice, it is possible to experimentally optimize the input and output
impedances simultaneously, by utilizing a reverse SP reflectometer to perform
source-pull measurements at the input and a direct SP reflectometer to perform load-
pull measurements at the output of the DUT [20] using the setup shown in Fig. 4.11
[13].
The mixer performance can be optimized by increasing its conversion gain, im-
proving the return losses at its ports, and improving its linearity. The advantage of
SP source-pull based mixer characterization is that the SP reflectometer in reverse
configuration gives the source impedance for any tuner position without any need to
pre-calibrate the tuner using a network analyzer. In addition, the tuner repeatability
is not important in this setup.
104 4 Six-Port Based Load-Pull System

Fig. 4.12 A generic SP based setup for power amplifier characterization

4.7.3 Power Amplifier Characterization

Figure 4.12 shows a generic SP based setup for power amplifier characterization.
The use of a reverse SP reflectometer allows for the measurement of the source
and load impedances presented to the transistor by the input and output tuning.
The measurement setup provides any designer with the ability to experimentally
verify the transistor sensitivity in power gain versus termination impedances and, as
a consequence, enables the designer to easily optimize the output power, amplifier
gain, power-added efficiency (PAE) or the intercept point of the transistor [21].
The expressions given in Eqs. (4.39)–(4.42) are relevant during power amplifier
characterization and optimization.
 2 
Pout (f ) = |b2 |2 1 − ΓL (f ) (4.39)
  
Pin (f ) = |a1 |2 1 − Γin (f )
2
(4.40)
Pout
Power_gain = (4.41)
Pin
Pout − Pin
PAE = (4.42)
Pdc
where it is assumed that the dissipated DC power is known to calculate the PAE.

4.8 Oscillator Measurements

The characterization of active microwave devices is generally done by the measure-

ment of S-parameters using automatic vector network analyzers (VNAs). The test
port impedance of the VNA seen by the active DUT is the characteristic impedance
of the measurement system (50 ). It must be noted that this test port impedance
cannot normally be varied during measurements. This constraint is not convenient
4.8 Oscillator Measurements 105

Fig. 4.13 Block diagram of the experimental set-up for source-/load-pull oscillator measurements
[22], © IEEE 1992

when measuring active devices having negative resistance (e.g., diodes and transis-
tors), because oscillations may occur during characterization whenever the negative
resistance of the DUT exceeds 50 . In such a situation, an SP reflectometer with
variable test port impedance could be very effective. There has been a report of
a specially designed SP junction that allows variation of the test port impedance
over the entire Smith chart, which would find immediate use in large-signal active
microwave device characterization, such as microwave source/load-pull oscillator
measurements [22].
By using the transparent SP assumption from Eq. (4.4), it can be seen that, in
order to obtain a variation of the test port impedance, it is sufficient to vary the in-
ternal impedance of the generator [22]. The variation of the test port impedance can
be performed with a three-stub tuner inserted between the signal generator and the
input port of the SP junction, as shown in Fig. 4.13. The generator and the three-stub
tuner can be considered equivalent to a generator with variable test port impedance.
Based on the above consideration and by using an appropriate calibration method
that is insensitive to the source power level variation and internal source impedance
[15], the calibration parameters of the SP reflectometer obtained for given tuner stub
positions are valid for any position of the stubs.
Source/load-pull oscillator measurements monitor both the oscillator power and
frequency as a function of the load impedance seen by the oscillator under test. The
impedance seen by the oscillator can be changed and, at the same time, known over
the quasi-entire Smith chart, by simultaneously changing and measuring the test
port impedance of the SP junction of the experimental setup shown in Fig. 4.13.
The reflection coefficient associated with the load impedance, ZDUT , seen by the
oscillator is directly deduced from the measured reflection coefficient, ΓDUT :

ΓDUT + 1
ZDUT = Z0 (4.43)
ΓDUT − 1
106 4 Six-Port Based Load-Pull System

The available power, Pa , from the oscillator can be calculated using the SP calibra-
tion parameters and the power calibration parameter as follows [23]:
k(f )P3 (f )|ΓDUT (f )|2
Pa (f ) = (4.44)
|1 + c(f )ΓDUT (f )|2
where c(f ) is the de-embedding calibration parameter obtained by the SP error box
calibration at oscillator frequency f , k(f ) is a scalar parameter obtained by the
power calibration at f , P3 (f ) is the power reading at port 3 of the SP junction at
f , and ΓDUT (f ) is the reflection coefficient measured by the SP reflectometer at
oscillator frequency f .
The impedance seen by the oscillator can be simultaneously varied and measured
over the quasi-entire Smith chart by changing the tuner stub positions and measur-
ing the test port impedances of the SP junction in Fig. 4.13. The available power
from the oscillator under test, Pa , is easily obtained at any tuner stub position and
at any operating oscillator frequency using Eq. (4.44). The oscillator frequency is
measured using a directional coupler and a frequency meter, as shown in Fig. 4.13.
The SP reflectometer is calibrated at discrete frequency points centered around the
nominal oscillator frequency. Measurement results of the oscillator characterization
using SP based setup are presented in [22].

4.9 AM/AM and AM/PM Measurements

For an accurate and quick design of linear solid-state power amplifiers (SSPAs)
and power limiters in hybrid microwave integrated circuit (HMIC) and monolithic
microwave integrated circuit (MMIC) technologies, transistor characterization, in
terms of phase distortion measurements, is highly recommended [24, 25]. The
AM/PM distortion behavior can be described by the relative phase shift versus input
power (φ vs Pin ) and the AM/PM conversion coefficient, k, in degrees per dB.
Traditionally, φ vs Pin characteristics have been obtained using two-carrier stim-
uli [24, 25], which is a tedious and complex method. Therefore, another alternate
method based on an SP active load-pull measurement system, which is less tedious
and requires only a single-carrier stimulus to obtain the φ vs Pin characteristic, is
more appropriate for microwave/millimeter wave amplifiers [5]. If needed, k can
then be found by derivation of φ with respect to Pin (dφ/dPin ). Actually, the mea-
sured φ vs Pin curve as a phase transfer function is more general for the character-
ization of the AM/PM distortion performance of microwave transistors and various
power amplifiers. In addition, φ vs Pin characterization is especially useful for mi-
crowave limiters, where the variation range of the phase shift over given power
ranges beyond saturation is the main concern [26].
This section utilizes a measurement setup that employs the principles of a dual
SP network analyzer [5] and the active loading technique [3]. In comparison to
heterodyne network analyzers based setups, the advantages of this measurement
setup are [12]:
4.9 AM/AM and AM/PM Measurements 107

Fig. 4.14 Block diagram of load-pull stage of distortion measurement system [5], © IEEE 1995

• The impedance and power measurements are performed at the actual power level
of the device: no extra attenuators are needed for power device characterization.
• The AM/PM characterizations are carried out at the input and output reference
planes of the DUT, loaded by arbitrary impedances.
• The cost of this system is much lower than the cost of setups using two automatic
network analyzers for AM/PM distortion load-pull measurements.

4.9.1 Principles of Operation

Figure 4.14 shows a block diagram of a distortion measurement system based on

an SP reflectometer. As shown, two SP junctions associated with the amplitude and
phase controllers are used for active load-pull techniques. Meanwhile, the measure-
ment system can be viewed as a three-port network with reference planes 1, 2 and 3.
In order to measure the input-output phase shift of the DUT, three two-port passive
standards, Z1, Z2 and Z3, are introduced to obtain the calibration coefficients.
Due to the fact that the three-port network has fixed network S-parameters during
the four measurements of the DUT, Z1, Z2 and Z3 for a fixed setting of the amplitude
and phase controllers, we can obtain the following equations expressed in matrix
format [14].
108 4 Six-Port Based Load-Pull System
    
S12 (1) (1) − Γ1 (1)S22 (1) S12 (1)Γ1 (1) x1 Γ1 (1) − S11 (1)
S12 (2) (2) − Γ1 (2)S22 (2) S12 (2)Γ1 (2) x2 = Γ1 (2) − S11 (2)
S12 (3) (3) − Γ1 (3)S22 (3) S12 (3)Γ1 (3) x3 Γ1 (3) − S11 (3)
(4.45)
where
• xi is related to the S-parameters of the three-port network for a given amplitude
and phase setting;
• (p) = S11 (p)S22 (p) − S12 (p)S21 (p) and Sij (p), where p = 1, 2 and 3, are the
known S-parameters of the three standards, Z1, Z2, and Z3;
• Γ1 (p), where p = 1, 2 and 3, are measured by SP1, when the test path is con-
nected to Z1, Z2, and Z3, respectively.
The three two-port standards, Z1, Z2, and Z3, are three coaxial transmission lines
with different lengths. These standards have been chosen in such a way that the
phase of S12 (1), S12 (2), and S12 (3) deviate from each other by about 120°, in order
to ensure a good condition for Eq. (4.45).
The following equation can also be derived [27] to calculate coefficient g, i.e.,
the ratio of the incident waves at reference planes 2 and 1, as shown in Fig. 4.15:
a2 x1 + x3 Γ1 (T )
g=  = (4.46)
a1 1 + x2 Γ2 (T )
where xi are complex constants [12], and Γ1 (T ) and Γ2 (T ) are measured by SP1
and SP2, respectively, when the test path is connected to the DUT. Coefficient g has
to be de-embedded to the input and output ports of the DUT (reference planes T1
and T2), in order to transfer this parameter to the respective planes, T1 and T2.
Network M and N in Fig. 4.15 consist of single-pole four-throw switches, SP4Tl
or SP4T2, bias tees and half parts (input or output) of the test fixture. For these
networks, following expressions can be obtained [12]:
b1 = S11
m 
a1 + S12
m
b1 (4.47)
m 
a1 = S21 a1 + S22
m
b1 (4.48)
n 
a2 = S11 b2
n
+ S12 a2 (4.49)
b2 = S21
n
b2 n 
+ S22 a2 (4.50)
where Sijm and Sijn are the S-parameters of networks M and N , and Γ1 (T ) = b1 /a1
and Γ2 (T ) = b2 /a2 .
From the above equations, a2 /a1 can be deduced and related to coefficient g as
given in Eq. (4.51).
n − n
S m Γ2 (T )S11
a2
= g. 12 . (4.51)
a1 S21 Γ1 (T )S22 − m
n m

where m = S11m S m − S m S m and n = S n S n − S n S n . Since switches SP4T1

22 12 21 11 22 12 21
and SP4T2 are operated under linear conditions, the S-parameters of networks M
and N are constants when Pin is swept. These S-parameters can be obtained using
a de-embedding method, such as the TRL (thru-reflect-line) technique [15].
4.9 AM/AM and AM/PM Measurements 109

Fig. 4.15 Error box model for de-embedding procedures and computations

According to the definition of normalized waves a and b, the ratio of the voltages
at the output and input ports of the DUT for any arbitrary load at a given Pin is
determined as follows:

V2 a2 + b2 a2 1 + 1/ΓDUT
= = . (4.52)
V1 a1 + b1 a1 1 + Γin

where a2 /a1 is given by Eq. (4.51), Γin is the input reflection coefficient of the DUT,
and ΓDUT is the load reflection coefficient presented to the DUT.
In brief, by means of switches, Γ1 (1), Γ1 (2), Γ1 (3) and Γ1 (T ), Γ2 (T ), can be
measured when Z1, Z2, Z3 and the DUT are tested in turn for a fixed setting of
the amplitude and phase controllers. Meanwhile, it is easy to obtain Γin , Pin , ΓDUT
and PDUT with de-embedding techniques, when Γ1 (T ) and Γ2 (T ) are measured
by SP1 and SP2 [5]. Therefore, the phase shift of the DUT can be determined by
Eqs. (4.45)–(4.52). The filter in Fig. 4.14 is used to filter out the harmonics, which
ensures that the signals detected by the power meter are only the fundamental com-
ponents.
The phase distortion, φ, is defined as an input-output phase shift, relative to the
ref
reference phase shift value at a given lower input power, Pin (small-signal oper-
ation mode), for a given load impedance. Thus, the pertinent relative phase shift
corresponds to the change of the angle of V2 /V1 , given in Eq. (4.52), when Pin in-
creases. The measurement of φ vs Pin of the AM/PM distortion behavior followed
by the calculation of the derivative (dφ/dPin ) provides the AM/PM conversion co-
efficient (k).
Similarly, the AM/AM distortion is described by the deviation of the gain from
its value at a small-signal input power level. Due to the fact that PDUT and Pin are
already measured, the operating power gain (G vs Pin ) can be deduced; and, the
AM/AM conversion coefficient can be determined by the derivative of dGp /dPin
[5, 24].
110 4 Six-Port Based Load-Pull System

4.9.2 Measurement Procedure

Traditionally, the load-pull measurement has to be carried out for each given Pin
with variable ΓDUT . This results in a large number of measurements and adjust-
ments. However, the SP based setup given in Fig. 4.14 enables active load-pull mea-
surements by fixing the positions of the amplitude and phase controllers and sweep-
ing the input power. In such a condition, ΓDUT does not change as Pin is swept. The
amplitude and phase controllers are then adjusted at new positions, and the swept
drive measurements are repeated. Finally, the load-pull contours for a given Pin can
be extracted from the measured data [5].
In addition, characteristics PDUT vs Pin and φ vs Pin for a given ΓDUT can also be
obtained from the measured result. The PAE and gain (GDUT ) can also be deduced.
In this method, there are mainly four error sources. First, the measurements are
made by increasing the input power and not by amplitude modulation as in the actual
operation mode. Therefore, the dynamic effect of the amplitude modulation, which
becomes more significant by increasing the modulation bandwidth, is ignored. Sec-
ond, the self-heating effect can introduce a drift in the electrical operational con-
ditions. To minimize this effect, the temperature of the fixture of the transistor in-
vestigated needs to be maintained as almost constant by circulating a flux of air
during measurements. Third, the measurement accuracy can also be diminished if
the power level exceeds the operational range where the switches are linear. In or-
der to circumvent this problem, it is preferable to use mechanical switches instead of
solid-state switches when characterizing high-power devices. Fourth, the extraction
of the pertinent data from the raw database and the post-measurement calculations
using interpolation routines may introduce error. This effect can be reduced by in-
creasing the amount of experimental data.

References
1. F.M. Ghannouchi, R. Larose, R.G. Bosisio, Y. Demers, A six-port network analyzer load-pull
system for active load tuning. IEEE Trans. Instrum. Meas. 39(4), 628–631 (1990)
2. R. Hajji, F.M. Ghannouchi, R.G. Bosisio, Large-signal microwave transistor modeling using
multiharmonic load-pull measurements. Microw. Opt. Technol. Lett. 5(11), 580–585 (1992)
3. F.M. Ghannouchi, R.G. Bosisio, An automated millimeter-wave active load-pull measurement
system based on six-port techniques. IEEE Trans. Instrum. Meas. 41(6), 957–962 (1992)
4. F.M. Ghannouchi, F. Beauregard, A.B. Kouki, Large-signal stability and spectrum character-
ization of a medium power HBT using active load-pull techniques. IEEE Microw. Guided
Wave Lett. 4(6), 191–193 (1994)
5. F.M. Ghannouchi, G. Zhao, F. Beauregard, Simultaneous load-pull of intermodulation and out-
put power under two tone excitation for accurate SSPA design. IEEE Trans. Microw. Theory
Tech. 42(6), 943–950 (1994)
6. F.M. Ghannouchi, F. Beauregard, R. Hajji, A. Brodeur, A de-embedding technique for
reflection-based s-parameters measurements of HMICs and MMICs. Microw. Opt. Technol.
Lett. 10(4), 218–222 (1995)
7. R. Hajji, F. Beanregard, F.M. Ghannouchi, Multitone power and intermodulation load-pull
characterization of microwave transistors suitable for linear Sspa’s design. IEEE Trans. Mi-
crow. Theory Tech. 45(7), 1093–1099 (1997)
References 111

8. D. Le, F.M. Ghannouchi, Multitone characterization and design of FET resistive mixers
based on combined active source-pull/load-pull techniques. IEEE Trans. Instrum. Meas. 46(9),
1201–1208 (1998)
9. G. Berghoff, E. Bergeault, B. Huyart, L. Jallet, Automated characterization of HF power
transistors by source-pull and multiharmonic load-pull measurements based on six-port tech-
niques. IEEE Trans. Microw. Theory Tech. 46(12), 2068–2073 (1998)
10. S. Jia, New application of a single six-port reflectometer. Electron. Lett. 20, 920–922 (1984)
11. J.D. Hunter, P.I. Slomo, S-parameter measurements with a single six-port. Electron. Lett. 21,
157–158 (1985)
12. F.M. Ghannouchi, A. Mohammadi, The Six-Port Technique (Artech House, Norwood, 2009)
13. D. Le, F.M. Ghannouchi, Source-pull measurements using reverse six-port reflectometers with
application to MESFET mixer design. IEEE Trans. Microw. Theory Tech. 42, 1589–1595
(1994)
14. T.E. Hodgetts, G.J. Griffin, A unified treatment of the six-port reflectometer calibration using
a minimum of standards, Royal Signals and Radar Establishment Rep. No. B3003, 1983
15. G.F. Engen, Calibrating the six-port reflectometer by means of sliding terminations. IEEE
Trans. Microw. Theory Tech. 26, 951–957 (1978)
16. V. Teppati, A. Ferrero, U. Pisani, Recent advances in real-time load-pull systems. IEEE Trans.
Instrum. Meas. 57, 2640–2646 (2008)
17. D. Le, F.M. Ghannouchi, Noise measurements of microwave transistors using an uncalibrated
mechanical stub tuner and a built-in reverse six-port reflectometer. IEEE Trans. Instrum. Meas.
44(4), 847–852 (1995)
18. R.Q. Lane, Determination of device noise parameters. Proc. IEEE 57, 1461–1462 (1969)
19. S.V. Bosch, L. Martens, Experimental verification of pattern selection for noise characteriza-
tion. IEEE Trans. Microw. Theory Tech. 48(1), 156–158 (2000)
20. F.M. Ghannouchi, R. Larose, R.G. Bosisio, A new multiharmonic loading method for large-
signal microwave transistor characterization. IEEE Trans. Microw. Theory Tech. 39, 986–992
(1991)
21. G. Gonzalez, Microwave Transistor Amplifiers Analysis and Design, 2nd edn. (Prentice-Hall,
Englewood Cliffs, 1996)
22. F.M. Ghannouchi, R.G. Bosisio, Source-pull/load-pull oscillator measurements at mi-
crowave/mm wave frequencies. IEEE Trans. Instrum. Meas. 41, 32–35 (1992)
23. T. Yakabe, M. Kinoshita, H. Yabe, Complete calibration of a six-port reflectometer with one
sliding load and one short. IEEE Trans. Microw. Theory Tech. 42, 2035–2039 (1994)
24. K. Koyama, T. Kawasaki, J.E. Hanely, Measurement of AM-PM conversion coefficients.
Telecommun. 12(6), 25–28 (1978)
25. J.F. Moss, AM-AM and AM-PM measurements using the PM null technique. IEEE Trans.
Microw. Theory Tech. MTT-35(8), 780–782 (1987)
26. T. Parra, X-band low phase distortion MMIC power limiter. IEEE Trans. Microw. Theory
Tech. 41(5), 876–879 (1993)
27. R.A. Soares, P. Gouzien, P. Legaud, G. Follot, A unified mathematical approach to two-port
calibration techniques and some applications. IEEE Trans. Microw. Theory Tech. MTT-37,
1660–1674 (1989)
Chapter 5
High-Power Load-Pull Systems

This chapter presents the theory and critical issues associated with high reflec-
tion load-pull systems required for the characterization and measurement of high
power transistor devices. Theory and analysis of the impedance transforming net-
work between the device under test (DUT) and the load-pull system, which is needed
to overcome problems encountered in high-power load-pull measurements, is dis-
cussed in detail. Finally, multi-tier calibration and de-embedding techniques, which
are required for accurate data extraction during high-power DUT characterization,
are also presented.

5.1 Introduction

Most high-power radio frequency (RF) transistor devices have extremely low output
impedance, in the order of 1  and, in some cases, even less than 1  [1–3]. The
measurement and characterization of such devices for optimal performance extrac-
tion, therefore, require load-pull systems capable of creating highly reflective load-
ing environments, where the magnitude of the reflection coefficient, |ΓL |, which is
associated with the load presented to the transistor, approaches unity. The require-
ment to establish highly reflective environments puts severe constraints on tradi-
tional passive and active load-pull systems [4, 5]; and, as a result, these systems are
limited in their use for the characterization and measurement of high-power devices.

5.2 Limitations of Existing Load-Pull Systems

The major drawback of the standard passive load-pull system is the limited maxi-
mum achievable reflection coefficient |ΓL | (maximum usually around 0.75–0.85),
due to the inherent insertion losses in the load tuner, measurement network and test
fixture [6]. Several developments, namely the quarter wave transformer technique,

F.M. Ghannouchi, M.S. Hashmi, Load-Pull Techniques with Applications to Power 113
Amplifier Design, Springer Series in Advanced Microelectronics 32
DOI 10.1007/978-94-007-4461-5_5, © Springer Science+Business Media Dordrecht 2013
114 5 High-Power Load-Pull Systems

Fig. 5.1 Generic model of a

passive load-pull
characterization setup

pre-matched tuning technique and active load-pull technique [7–14], have been in-
corporated to enhance the tuning range of the load-pull test benches to meet the
demands of high-power device characterization. The probing couplers [15], which
are used to sample and capture the incident and reflected traveling waves at the
DUT plane, are primarily used in the characterization of on-wafer devices. Direct
measurement of traveling waves at the DUT plane by probing couplers enables the
enhancement in the tuning range of the load reflection coefficient.
Tuning range enhancement is just one aspect in a high-power load-pull measure-
ment technique. The large mismatch between the output impedance of the high-
power DUT and the measurement system environment also creates a high volt-
age standing wave ratio (VSWR) in the measurement system. A high VSWR can
generate very large voltage and current peaks in the measurement system that can
damage the DUT and the measurement system. Furthermore, the load-pull power,
PLP , requirement in active load-pull systems increases manifold, due to the large
impedance mismatch between the output impedance of DUT and the load-pull test
set. The increase in PLP limits the usefulness of active load-pull systems in practical
applications due to increased cost.
In addition, the high impedance mismatch between the load-pull systems and
the DUT also leads to increased measurement uncertainty [16], which is caused
by the limitations of automatic vector network analyzers (VNAs) in characterizing
“high-reflection low-loss” two-ports, such as tuners. According to the well known
rule of thumb in metrology, the ratio between reference system impedance (usually
50 ) and DUT output should always remain less or equal to 10 (VSWR ≤ 10:1 or
|ΓL | ≤ 0.8) in order to obtain reliable load-pull measurement data [16]. In highly
reflective load-pull measurement systems, however, the required |ΓL | approaches
unity, thereby leading to measurement uncertainty.

5.2.1 Problems Due to High Standing Waves

The impact of a high VSWR on the load-pull system performance, in passive as well
as active systems, can be analyzed by considering a generic model of a passive load-
pull system, as shown in Fig. 5.1, and the associated signal flow model, as depicted
in Fig. 5.2. It can be observed that the DUT, which is represented by an equivalent
5.2 Limitations of Existing Load-Pull Systems 115

Fig. 5.2 Signal flow model

representing a passive
load-pull model

voltage source, Vd , and a series impedance, Zd , is connected to the passive tuner

via a 50  lossless transmission path of length l.
In a standard load-pull characterization setup, a passive network that consists of
measuring couplers is used for capturing the incident and reflected traveling waves
at the DUT ports; therefore, it always has a transmission path between the DUT
and the tuner. The terms Γd and Γlp represent the load reflection coefficients at the
DUT and load-pull planes, respectively. The DUT output reflection coefficient is
represented by ΓDUT ; bd is the maximum amplitude of the power wave generated
by the DUT; β is the propagation constant along the transmission path; and, b(z)
and a(z) are the incident and reflected waves, respectively.
Through the application of the standard transmission line theory, the relationship
for the average available power from the DUT, Pav , transferred at the DUT plane
can be expressed by Eq. (5.1), as follows:

|bd |2  
Pav = 1 − |ΓDUT Γd | (5.1)
2
The DUT delivers the maximum available power, Pd , at the DUT plane when it is
conjugately matched to the tuner transmission path, ΓDUT = Γd∗ .

|bd |2  
Pd = Pav
max
= 1 − |Γd |2 (5.2)
2
The voltage propagating on the transmission path, with the characteristic impedance,
Z0 , of 50 , is related to the respective incident and reflected traveling waves as
[17]:
  
V (z) = Z0 b(z) + a(z) (5.3)
where the incident and reflected traveling waves, b(z) and a(z), are given by
Eqs. (5.4) and (5.5).

b(z) = b(0)e−jβz (5.4)

a(z) = a(0)e jβz
(5.5)
where b(0) and a(0) are the incident and reflected traveling waves at the DUT plane.
The signal flow model in Fig. 5.2 can be used to derive the relationship between
the power wave, bd , and the incident wave, b(0), at the DUT plane, as given in
Eq. (5.8).
116 5 High-Power Load-Pull Systems

Fig. 5.3 Propagating voltage

on the transmission path for
three DUTs with power
ratings of 200 W, 100 W and
50 W

a(0) = b(0)Γd (5.6)

b(0) = bd + a(0)ΓDUT (5.7)
bd
b(0) = (5.8)
1 − |Γd |2
Equations (5.2) to (5.8), when solved and simplified, provide the voltage expres-
sion on the transmission path given in Eq. (5.9).

2Z0 Pd  
V (z) = 1 + |Γd |e2jβz e−jβz (5.9)
1 − |Γd |2

The propagating voltage, V (z), plotted using Eq. (5.9), along the transmission
path for 200 W, 100 W and 50 W devices with the respective assumed low output
impedances of 0.5 , 1  and 2  and a characteristic system impedance, Z0 , of
50  is given in Fig. 5.3. It can be observed that massive voltage peaks, which are
dependent on the power rating of the DUTs, are generated that can cause break-
downs between the transmission path and the ground, due to the “corona effect”
[18]. This has the potential to damage the tuner, couplers or bias tees used in the
measurement setup and/or the DUT itself. As a result, the load-pull setup in its
standard configuration is not suitable for characterization and measurement of high-
power DUTs with inherently low output resistance values.
The generated voltage peaks on the transmission path can be quantified by con-
sidering the magnitude of the voltage, Vmax , in Eq. (5.10), which is derived from the
expression of the propagating voltage in Eq. (5.9).

Vmax = 2Pd Z0 (VSWR) (5.10)
where
1 + |Γd |
VSWR = (5.11)
1 − |Γd |
5.2 Limitations of Existing Load-Pull Systems 117

Table 5.1 VSWR and

maximum generated voltage Device output Assumed output VSWR Vmax (volt)
for devices with low output power (watts) impedance ()
impedances in the
measurement system 50 2 24.974 353.37
transmission path 100 1 49.891 706.34
200 0.5 99.503 1410.69
250 0.4 124.786 1766.25
400 0.25 199 2821.35
500 0.2 249 3528.46

Fig. 5.4 Propagating voltage

in a passive load-pull
measurement system when
characterizing a 100 W (1 )
device at 2.1 GHz. Similar
curves can be obtained for an
active load-pull system

The data in Table 5.1 are the estimated peak voltages, Vmax , along the measure-
ment system transmission path obtained from Eq. (5.10) for transistor devices with
arbitrarily assumed low impedances. For a simpler analysis, only real impedances
have been assumed. It is apparent that a high VSWR in the load-pull setup used
for characterization of high-power devices can generate very high voltages in the
transmission path, which can damage the DUT and/or the measurement system.
It can be concluded from the above analysis that the tuner should be placed at
the DUT port, in order to avoid high voltage peaks; however, this is not feasible in
practical applications. For example, the propagating voltage, V (z), along the trans-
mission path for a 100 W device at a frequency of 2.1 GHz (as shown in Fig. 5.4)
can be as high as 707 V. To avoid this peak, the length of the transmission path be-
tween the DUT and the tuner should correspond to position 1 shown in Fig. 5.4, so
that the peak voltage never goes above 100 V, which is assumed to be the maximum
tolerable voltage for a DUT of 100 W with a 1  output impedance.
The length of the transmission path corresponding to position 1 is, however, too
small at 2.1 GHz; therefore, it is not feasible to place wave sampling apparatuses,
such as directional couplers, between the DUT and the load tuner, if the length of the
transmission path corresponding to position 1 is chosen. As a result, the closest the
118 5 High-Power Load-Pull Systems

Fig. 5.5 Generic model of an

open-loop active load-pull
measurement setup

tuner can be placed to the DUT corresponds to either to position 2 or 3 in Fig. 5.4,
thus introducing at least one voltage peak. The increasing operating frequency in-
creases the number of voltage peaks in the transmission path, due to decreasing
wavelength. Therefore, the standard active or passive load-pull approaches that em-
ploy 50  transmission lines between the DUT and the load tuner are not suitable
for high-power device characterization at high frequency.

5.2.2 Problem of Large Load-Pull Power

In an active load-pull technique, the reflected signal can be either from an external
signal generator [19, 20] or a modified form of the DUT generated signal [21–23].
In both cases, the synthesis of the desired reflection coefficient at the DUT plane
is dependent on the ability of the load-pull components to provide the necessary
load-pull power, PLP . In theory, active load-pull systems can synthesize reflection
coefficients at and beyond the boundary of the Smith chart, but this requires the load-
pull components to reflect a PLP of high magnitude, which creates a bottleneck in
the operation of the system.
The requirement of PLP for desired reflection coefficient synthesis at the DUT
plane can be analyzed through a generic model of an active load-pull setup, as shown
in Fig. 5.5. The DUT is represented by an equivalent voltage source, Vd , and a series
impedance, Zd . The terms VLP and ZLP represent the voltage and the impedance
of the active load-pull tuner, respectively. In this model, ZLP is set to the system
characteristic impedance, Z0 , of 50 , while the circulator isolates the DUT and the
active load-pull tuner. The circulator should be as close as possible to the DUT in
order to eliminate the occurrence of voltage peaks in the measurement system. In
reality, the presence of a circulator requires an additional length to the transmission
path, consequently leading to an increased possibility of voltage peak occurrence in
the measurement system.
From the load-pull model in Fig. 5.5, one can derive the expression for the reflec-
tion coefficient at the load-pull plane, Γlp , which is given by the ratio of the reflected
wave, a(z), to the incident wave, b(z), as:

a(z)|z=l PLP
Γlp = = (5.12)
b(z)|z=l Pgen
5.3 High-Power Load-Pull 119

Fig. 5.6 Load-pull power,

PLP , requirements as a
function of the load reflection
coefficient, Γlp , at the
load-pull reference plane for
a 100 W (1 ) device

where
Pgen = PLP + Pd (5.13)
The reflection coefficient at the DUT plane, Γd , is related to the reflection coef-
ficient at the load-pull plane by the relationship given in Eq. (5.14).
Γlp = Γd e−2jβz (5.14)
Simplification of Eqs. (5.12), (5.13), (5.14) provides the relationship between
the desired reflection coefficient at the DUT plane, Γd , and the required load-pull
power, PLP , to synthesize it.
|Γd |2
PLP = Pd (5.15)
1 − |Γd |2
Fig. 5.6 is a plot of Eq. (5.15) and relates the load-pull power, PLP , with the
reflection coefficient at the DUT plane, |Γd |. It can be observed that the PLP required
for load-pulling of the assumed 100 W device having 1  output impedance (i.e.
load reflection coefficient of 0.96) is around 1175 W. The required PLP of 1175 W
is, however, impractical for load-pulling of a 100 W DUT, due to the cost involved
in building such an active load-pull setup. It is these incurred costs that severely
limit the use of active load-pull systems in their standard configuration [19–23] for
the characterization of high-power DUTs in the order of 100 W.

5.3 High-Power Load-Pull

An appropriate load-pull for high-power device characterizations should be capa-
ble of synthesizing a high reflection factor and, at the same time, should possess
some mechanism to minimize the VSWR. In the case of the active load-pull setup,
120 5 High-Power Load-Pull Systems

Fig. 5.7 Mechanical drawing

of a pre-matched load-pull
tuning setup [25] (diagram
courtesy of Focus
Microwaves)

Fig. 5.8 Photograph of a

commercial pre-matching
load-pull tuner (photo
courtesy of Focus
Microwaves)

there is an additional requirement to minimize the load-pull power, PLP , for the
synthesis of the desired reflection coefficient. There is no generic approach to over-
come these problems concurrently; however, techniques have been developed that
are able to successfully fulfill the requirements of high-power device characteriza-
tion and measurement problems. The pre-matched tuning technique [10, 11], the
enhanced loop tuner [24], the quarter wave transformation technique [7] and the
broadband impedance transformer technique [19] are the most common approaches
in addressing the high-power device measurement and characterization problem.

5.3.1 Pre-matching Technique

Figure 5.7 shows a diagram of a pre-matched load-pull tuning setup. The pre-
matched load-pull tuners consist of two independent RF probes, called pre-matching
and tuning probes, respectively, placed side-by-side on a central conductor. These
two probes are capable of individually generating smaller reflection coefficients,
which combine together to enhance the maximum achievable reflection coefficient.
Figure 5.8 depicts a photograph of a commercial pre-matching tuner.
In this technique, the reflection coefficients generated by the pre-matching probe,
ΓPre-match , and the tuning probe, ΓProbe , combine together to create the eventual load
reflection coefficient, ΓTotal , given by Eq. (5.16).
5.3 High-Power Load-Pull 121

Fig. 5.9 Smith chart

demonstrating the concept of
the total reflection coefficient,
ΓTotal , generation by a
pre-matching load-pull setup
(diagram courtesy of Focus
Microwaves)

S12 S21 ΓProbe

ΓTotal = ΓPre-match + (5.16)
1 − S22 ΓProbe
where S12 , S21 , and S22 are the S-parameters of the pre-matching tuner.
A practical pre-matched load-pull setup, however, utilizes a low-loss central con-
ductor; therefore, it is reasonable to assume that S21 S12 approaches unity. Further-
more, the term S22 can be assumed to be approaching zero, considering that there
is no discontinuity in the central conductor. In such a situation, the reflection coeffi-
cient, ΓTotal , generated by the pre-matched load-pull setup is given as:

ΓTotal = ΓPre-match + ΓProbe (5.17)

The concept of reflection coefficient synthesis by pre-matched load-pull setup
using Eq. (5.17) is illustrated in Fig. 5.9. It can be observed that initially the pre-
matching probe moves the matched condition from 50  to some other desired
region of the Smith chart. This position of the pre-matching probe defines the re-
flection coefficient of the pre-matching probe, ΓPre-match , as shown in the figure.
The movement of the tuning probe in horizontal and vertical positions then gen-
erates the reflection coefficient of the tuning probe ΓProbe , whose vector addition
with ΓPre-match results into the overall synthesized reflection coefficient, ΓTotal . For
certain horizontal and vertical positions of the tuning probe, the overall synthesized
reflection coefficient achieves the highest value, denoted by Γmax .
It is evident from Fig. 5.9 that the overall synthesized reflection coefficient by
the pre-matching load-pull setup can be significantly higher for some horizontal
and vertical settings of the tuning probe. For example, pre-matched passive load-
pull systems [10, 11] can synthesize reflection coefficients in the order of 0.90–0.92
magnitudes, which are higher than the maximum achievable reflection coefficient
from a standard passive tuner, as can be seen in Fig. 5.10.
The pre-matching technique possesses an advantage over the quarter wave trans-
former and broadband impedance transformer techniques (described in Sects. 5.3.3
and 5.3.4, respectively), due to its capability of synthesizing any reflection coef-
ficient within the maximum achievable value. As a result, this technique is better
suited for high throughput load-pull measurement applications, in which a number
of optimal reflection coefficients need to be synthesized without changing the mea-
surement setup, such as the characterization and measurement of different DUTs
122 5 High-Power Load-Pull Systems

Fig. 5.10 Comparison of the

achieved reflection
coefficients using a standard
passive tuner and a
commercial pre-matched
tuner at 2.425 GHz [37],
© IEEE 2010

at varying frequencies. Additionally, the high bandwidth of the pre-matched load-

pull setup, which is about one decade, makes it suitable for high-power harmonic
load-pull applications.
The major drawback of this technique, however, is the strong dependence of the
accuracy of the synthesized reflection coefficients on interpolation algorithms when
the desired reflection coefficients lie farther from the calibration grid [26]. Prior
characterization of the pre-matched tuner with a dense calibration grid improves
the accuracy in the synthesized reflection coefficient. In some cases, separate prior
characterizations of tuners for distinct DUTs or different operating frequencies also
improve the accuracy in the synthesized reflection coefficients. However, in both
these scenarios, the overall measurement throughput gets reduced. Additionally, the
maximum achievable reflection coefficients get reduced, due to the adapters and
the insertion loss of the fixture. For example, a 0.1 dB insertion loss reduces the
reflection coefficient from 0.92 to 0.89, which is an increase in the synthesizable
impedance from 2.08  to 2.91  [27].

5.3.2 Enhanced Loop Architecture

An enhanced loop load-pull setup consists of an impedance tuner and a passive

loop cascaded together, as shown in Fig. 5.11 [24]. The impedance tuner, Tuner2,
is a standard low-loss passive tuner, whereas the passive loop is built using a high
directivity circulator and a coupler. In this technique, the passive loop first moves
the matched point farther from 50  by an amount ΓLOOP , which is regulated by
the length of the cable, L2 , and the position of probes in the impedance tuner. The
impedance tuner then adds its contribution to the reflection coefficient generated by
the passive loop, in order to synthesize the high reflection coefficient at the DUT
5.3 High-Power Load-Pull 123

Fig. 5.11 Block diagram of

an enhanced loop load-pull
tuner [24], © IEEE 2010

Fig. 5.12 Signal flow graph

of an enhanced loop load-pull
tuner [24], © IEEE 2010

plane. The signal flow graph in Fig. 5.12 illustrates the mechanism of reflection
coefficient synthesis by this technique.
The signal flow graph clearly identifies the dependence of the reflection coeffi-
cient synthesized at the DUT plane, ΓTotal , on the contributions from the impedance
tuner and the passive loop. The expression for ΓTotal derived from the above signal
flow graph is given as:
a2 S12TUNER S21TUNER ΓLOOP
ΓTotal = = S11TUNER + (5.18)
b2 1 − S22TUNER2 ΓLOOP
where
b3
ΓLOOP = = |ΓLOOP |e−2jβL2 (5.19)
a3
where the term ΓLOOP is the reflection coefficient generated by passive loop and is a
complex term dependent on the loop component characteristics, i.e. the transmission
factors of the coupler and the circulator, the phase velocity, β, of the traveling waves,
and the length of the loop, L2 .
The length of the loop and, hence, the reflection coefficient can be changed by
employing cable of appropriate lengths. It has been found that cables of only three
different lengths, L2 , can cover the entire Smith chart [24]. The need for only three
distinct cables results in reduced calibration and measurement time of the enhanced
loop load-pull system, compared to the pre-matched load-pull techniques, where
124 5 High-Power Load-Pull Systems

Fig. 5.13 Comparison of

achievable load reflection
coefficient ΓL using a
commercial pre-matched
tuner and the loop enhanced
tuner [24] at 2.425 GHz,
© IEEE 2010

separate pre-characterization of tuner is required for covering the desired Smith

chart region for the specified DUT. Additionally, the maximum synthesizable reflec-
tion coefficients of 0.97 using the enhanced loop load-pull system is higher than the
corresponding maximum values using the pre-matched load-pull system, as shown
in Fig. 5.13.
In this technique, the insertion loss of the fixture also reduces the maximum
synthesizable reflection coefficient, as in the pre-matched load-pull system [27].
Nonetheless, this technique is well suited for triplexer-based harmonic load-pull
systems, when the insertion loss of the triplexer and fixture are small [28]. In such
a situation, the all-passive harmonic load-pull setup can synthesize mutually inde-
pendent high reflection coefficients at fundamental and harmonic frequencies.

5.3.3 Quarter Wave Transformer Technique

The pre-matching and enhanced loop techniques are limited in their application, due
to two main reasons:
• The tuning ranges of these architectures are reduced by the adapters and the in-
sertion loss of the associated fixture.
• The high loss between the DUT and the tuner makes the associated measurement
error unpredictably high.
In many fundamental characterizations of high-power DUTs, the region of ap-
proximate optimal impedance lies is known. Therefore, in such applications, high
reflection coefficient tuning architectures are an unnecessary expense. Instead, the
inclusion of an impedance transforming network between the DUT and the tuner has
a significant advantage in such applications, as it brings down the impedance mis-
match to a level where the vector network analyzer (VNA) measurement uncertainty
is low; and, consequently, the measurement accuracy is high [29].
5.3 High-Power Load-Pull 125

Fig. 5.14 Pictorial representation of the quarter wave transformation technique

In contrast to the pre-matched and enhanced loop load-pull techniques,

impedance transforming network based load-pull setups do not experience dras-
tic degradation in the tuning range and the measurement accuracy caused by the
adapters and the insertion loss of the fixture. This is achieved due to the reduced
Smith chart region in which the reflection coefficients need to be synthesized,
thereby causing no or minimal error.
In principle, the quarter wave (λ/4) transformer technique is a special type of
pre-matched load-pull system, where the enhancement in the reflection coefficient
tuning range is obtained through the incorporation of a λ/4 impedance transformer
between the DUT and the tuner. The quarter wave moves the matched impedance
environment from 50  to some other smaller value, as shown in Fig. 5.14.
It can be observed that the maximum achievable reflection coefficient before
impedance transformation is smaller with the matched condition at point ‘a’ in the
Smith chart; whereas, the incorporation of a λ/4 transformer transforms the matched
condition to point ‘c’ in the Smith chart, which results in an enhanced tuning range,
albeit with reduced Smith chart coverage. It should be noted that the 50  line pro-
vides the match between the quarter wave transformer and the load-pull tuner.
The transformation of the measurement system’s impedance to some smaller
value results in a reduced impedance mismatch between the load-pull system and the
DUT, which reduces the VSWR and, hence, any potential damage to the DUT, tuner
and/or measurement system. The λ/4 transformer technique thus serves the dual
purpose of VSWR reduction and tuning range enhancement [7, 8]. Additionally, the
reduced impedance mismatch between the tuner and the DUT also significantly re-
duces the PLP requirement (described in Sect. 5.5) in the case of active load-pull
setups.
It must, however, be noted that the improvement in the tuning range and the re-
duction in the VSWR, which come at the cost of reduced Smith chart coverage,
necessitates the replacement of the λ/4 transformer for the cases when the charac-
terization of DUTs has to be carried out at multiple frequencies or when separate
DUTs need to be characterized successively, so that required Smith chart regions are
126 5 High-Power Load-Pull Systems

Fig. 5.15 Reduction in

Smith chart coverage with
increase in the transformation
ratio of the impedance
transformer [18]

not precluded. This results into additional hardware and calibration, which eventu-
ally increases size, cost and measurement time. Further, the narrow band nature of
a λ/4 transformer limits the usefulness of such systems in fundamental load-pull
characterizations and measurements of high-power DUTs.

5.3.4 Broadband Impedance Transformer Technique

Limitations of the quarter wave impedance transforming network, whose bandwidth

of only about 5 to 10 percent of the carrier frequency prevents harmonic load-pull
applications, are overcome through the incorporation of a broadband impedance
transformer, such as the “Klopfenstein” taper with bandwidth covering from a
few 100 MHz to 12 GHz [30], between the DUT and the tuner [19]. The broad-
band impedance transformer does reduce Smith chart coverage with increasing
impedance transformation ratios, as can be seen in Fig. 5.15, but this also results into
increased accuracy in the load-pull measurements [29] due to reduced impedance
mismatch between the DUT and the tuner. Furthermore, the broadband impedance
transformer incorporation in the load-pull setup not only enhances the maximum
achievable reflection coefficient, but also reduces the VSWR in both the active as
well as passive load-pull systems and the required load-pull power, PLP , in active
load-pull systems [17, 18].
The reduced Smith chart coverage sometimes requires very careful selection of
impedance transformers, because if the chosen impedance transformer precludes the
Smith chart region where the optimal reflection coefficient for a specified DUT lies,
then the very purpose of this whole exercise of impedance transformer incorporation
gets negated. This is especially important for high impedance transformation ratios,
because the smaller the impedance, the higher is the measurement sensitivity. It is,
therefore, a standard practice to load-pull an unknown DUT without an impedance
transformer, in order to determine the probable optimal impedance. It aids in the
identification of the Smith chart region and, hence, in the selection of the impedance
transformation ratio.
5.4 Impact of a Transformation Network on PLP and VSWR 127

Fig. 5.16 Modified representation for an open-loop active load-pull system with incorporation of
an impedance transformer [19], © IEEE 2005

5.4 Impact of a Transformation Network on PLP and VSWR

Impedance transformation networks between the DUT and the tuner enhance the
reflection coefficient tuning range, as can be seen in Figs. 5.14 and 5.15, and reduce
the generated VSWR and required PLP in the load-pull systems [7, 8, 18, 19]. This
section describes and quantifies the impact of Klopfenstein taper incorporation in an
open-loop active load-pull setup, using a generic model that is depicted in Fig. 5.16.
It is important to note that, although the analysis is based on an open-loop active
load-pull system, it is equally valid for other load-pull approaches.
In the load-pull model, the DUT is represented by an equivalent voltage source,
Vd , and a series impedance, Zd , that is connected to the larger geometry and, hence,
the smaller impedance of the impedance transformer. The circulator is connected
to the smaller geometry having 50  impedance. The line stretcher between the
impedance transformer and the DUT adjusts the phase of the synthesized reflection
coefficients. In several applications, the reflection coefficients phase adjustment is
essential, in order to achieve the appropriate matching condition to minimize the re-
quired load-pull power, PLP [18, 19, 22]. The parameter Γd represents the reflection
coefficient at the DUT plane, Γ the transformation ratio of the impedance trans-
former, Γlp the synthesized reflection coefficient at the load-pull plane, and ΓSyst
the system reflection coefficient arising due to mismatch between the circulator and
the smaller geometry of the impedance transformer.
Analysis of this load-pull model can be carried out using a simplified signal flow
graph, as given in Fig. 5.17, which assumes that the impedance transformer is loss-
less and its high impedance end is perfectly matched to the circulator, thereby lead-
ing to a further assumption of ΓSyst equaling zero.
In the signal flow graph, the term as ej ϕs stands for the signal injected at the
load-pull plane (i.e. at the high impedance end of the impedance transformer) by the
active load-pull source. The terms bd ej ϕb and ad ej ϕa represent the incident and re-
flected traveling waves at the DUT plane, respectively. Term  α represents the phase
associated with the line stretcher relating Γ to Γd , whereas 1 − |Γ |2 and γ rep-
resent the magnitude and phase of the transmission coefficient of the impedance
128 5 High-Power Load-Pull Systems

Fig. 5.17 Simplified signal

flow graph representing an
active load-pull setup with an
impedance transformer
between the DUT and the
load-pull signal generator
[38], © IEEE 2005

transformer. The synthesized reflection coefficient at the DUT plane, which is the
ratio of the reflected and incident traveling wave, can be derived from the signal
flow graph and expressed as:
ad ej ϕa
Γd = = |Γd |ej (ϕa −ϕb ) (5.20)
bd ej ϕb
It is evident from Eq. (5.20) that the load synthesized Γd at the DUT plane depends
on the relative phase between the incident and reflected traveling waves. Therefore,
the phase term ϕb can be taken as a reference and set to zero, in order to simplify
Eq. (5.18), which is then modified to:
ad ej ϕa
Γd = = |Γd |ej ϕa (5.21)
bd ej ϕb
The relationship given in Eq. (5.22) between the traveling waves present in the load-
pull system can be deduced from the signal flow graph and can be simplified to
determine the expression for the incident traveling wave, bd , at the DUT plane given
in Eq. (5.23).

ad ej ϕa = |Γ |ej α bd ej 0 + 1 − |Γ |2 ej γ as ej ϕs (5.22)

1 − |Γ |2 as ej (γ +ϕs )
bd = j α (5.23)
e (|Γd |ej (ϕa −α) − |Γ |)
The net power, Pd , at the DUT plane is related to the reflected traveling wave,
bd , and the incident traveling wave, ad , by Eq. (5.24).
 
Pd = |bd |2 − |ad |2 = |bd |2 1 − |Γd |2 (5.24)
Substitution of Eq. (5.23) into Eq. (5.24) and simplification gives the expression for
load-pull power, PLP , injected at the load-pull plane by the active load-pull source.
Pd (|Γd |2 + |Γ |2 − 2|Γd ||Γ | cos(ϕa − α))
|as |2 = |PLP | = (5.25)
(1 − |Γd |2 )(1 − |Γ |2 )
Equation (5.25) relates PLP to the transformation ratio of impedance transformer,
|Γ |ej α , and the desired reflection coefficient, |Γd |ej ϕa . Table 5.2 summarizes the
5.4 Impact of a Transformation Network on PLP and VSWR 129

Table 5.2 Impact of transformation ratio on the load-pull power for optimal synthesis of the load
reflection coefficient, for a 100 W device with a 1  output impedance
Transformation ratio Required PLP (watt) Transformation ratio Required PLP (watt)

50:5 78 50:20 442

50:7 126 50:30 686
50:9 174 50:40 931
50:10 198 50:50 1175

Fig. 5.18 Load-pull power

requirements as a function of
the relative phase difference
between the phases of Γd and
Γ for a 100 W (1 ) device
by incorporating a 50:7
impedance transformer

estimated PLP for load-pulling of a 100 W device and clearly identifies the sub-
stantial reduction in PLP for the synthesis of desired load reflection coefficients, Γd ,
with the increasing transformation ratio, Γ . In this case, it has been assumed that
the phase terms ϕa and α are equal; thus, the cosine term in Eq. (5.25) is unity.
In practical situations, it is not always possible to perfectly align the phase terms
ϕa and α. One can observe from Eq. (5.25) that the PLP get smaller as the phase
difference, ϕa − α, approaches zero and reaches minimum when α equal ϕa . In
order to visualize the impact of the term ϕa − α on PLP , Eq. (5.25) can be plotted
for the swept value of ϕa − α for a 100 W device and impedance transformer of
50:7, as shown in Fig. 5.18. It is apparent that a difference of 180° between ϕa
and α results into a very high requirement of PLP for the synthesis of a reflection
coefficient to match the 1  impedance of the 100 W device.
From Fig. 5.18, it is also evident that an active load-pull setup with 200 W of
available PLP requires the term ϕa −α to be below ±12°. In a practical measurement
setup, the phase difference, ϕa − α, is minimized by adding a low impedance line
or line stretcher between the DUT and the low impedance end of the transformer, as
shown in Fig. 5.16. The inclusion of a line stretcher, however, requires an additional
calibration step to remove any systematic error introduced by it [18, 31]. It is also
important to note that the phase term α is a function of the transformation ratio of the
130 5 High-Power Load-Pull Systems

Fig. 5.19 Voltage standing

wave for different
transformation ratios in active
and passive load-pull
measurement systems when
characterizing a 100 W (1 )
device at 2.1 GHz

impedance transformer; therefore, the selection of impedance transformer requires

careful attention, so as to minimize the PLP from the load-pull source.
The increasing transformation ratio also suppresses voltage peaks along the
transmission path of the load-pull measurement system. It is evident from the volt-
age (obtained using Eq. (5.9)) plot in Fig. 5.19, which depicts the voltage variation
along the transmission path for a 100 W (1 ) device for different transformation
ratios of the impedance transformer.
Fig. 5.19 and the data in Table 5.2 reveal that the higher the transformation ratio,
the lower are the required PLP and the VSWR within the measurement system. In
practice, however, the highest transformation ratio is not the optimal choice, as it
may preclude the desired Smith chart region. For an unknown device, the choice of
the optimal transformation ratio is dictated by its output impedance [19]. Therefore,
it is necessary to roughly estimate the output impedance of the DUT by measur-
ing its large signal, S22 , in order to determine the approximate Smith chart region
where the optimal reflection coefficient may lie. With this knowledge, an appropri-
ate impedance transformer is then selected to simultaneously minimize PLP demand
and the VSWR.

5.5 Hybrid Load-Pull System

In the preceding section, it was assumed that the circulator and smaller geometry
of impedance transformer in Fig. 5.16 were perfectly matched, due to both having
a characteristic impedance of exact 50 , thereby leading to ΓSyst equaling zero. In
practice, they do not possess perfect 50  impedance, resulting in a non-zero ΓSyst
and modification in the signal flow graph that is shown in Fig. 5.20.
In this signal flow graph (Fig. 5.20), the additional terms |ΓSyst |ej ϕSyst and |Γ |ejβ
represent the contributions of the impedance mismatch between the circulator and
5.5 Hybrid Load-Pull System 131

Fig. 5.20 Signal flow graph

for a load-pull model with an
impedance transformer
between the DUT and the
tuner and with an unmatched
circulator and impedance
transformer [19], © IEEE
2005.

impedance transformer and the impedance transformer reflection coefficient at the

high-impedance end of the impedance transformer, respectively. If it is assumed
that the impedance transformer is lossless and reciprocal, according to lossless re-
ciprocal network laws [30], its phase, β, can be related to its phase terms, α and γ ,
through the relationship expressed in Eq. (5.26)
β = 2γ − α ± nπ (5.26)
where n is an odd number.
The expression for PLP in Eq. (5.30) can be obtained after the simplification of
Eqs. (5.27) to (5.29), deduced from the signal flow graph in Fig. 5.20 and Eq. (5.24).
Once again the reference phase, ϕb , has been set to zero for arriving at the expression
for the PLP .
ad ej ϕa = Γin bd ej ϕb + ΓT as ej ϕs (5.27)
where
j (ϕSyst +2γ )
(1 − |Γ |2 )|ΓSyst |e
Γin = |Γ |ej α + j (ϕSyst +β)
(5.28)
(1 − |Γ |)|ΓSyst |e

1 − |Γ |2 ej γ
ΓT = j (β+ϕSyst )
(5.29)
(1 − |ΓSyst |)|Γ |e
⎛ ⎞
|Γd |2 + |Γ |2 + |ΓSyst |2 + |Γd |2 |Γ |2 |ΓSyst |2
⎜ − 2|ΓSyst | × |Γ | × (1 + |Γd |2 ) × cos(β + ϕSyst ) ⎟
⎜ ⎟
⎜
Pd ⎜ − 2|Γd | × |Γ | × (1 + |ΓSyst | ) × cos(ϕa − α)
2 ⎟
⎟
⎝ + 2|Γd | × |ΓSyst | × |Γ |2 × cos(β + ϕSyst + ϕa − α) ⎠
− 2|Γd | × |ΓSyst | × cos(ϕa − 2γ − ϕSyst )
PLP = |as |2 =
(1 − |Γd |2 )(1 − |Γ |2 )
(5.30)
The impact of ΓSyst on PLP can be observed in Fig. 5.21, which depicts the plot
of Eqs. (5.30) and (5.26) for two cases when |ΓSyst | equals 0 and 0.05, for a 100 W
(1 ) DUT and an impedance transformer with a 50:7 transformation ratio. The
phase terms ϕa , α, and γ have been set equal to π , as this provides the minimum
PLP ; whereas, the phase, ϕSyst , of the system reflection coefficient is swept from
132 5 High-Power Load-Pull Systems

Fig. 5.21 PLP as a function

of the phase of the system
reflection coefficient with a
50:7 impedance transformer
(for a 100 W (1 ) device)

Fig. 5.22 Typical block

diagram of a hybrid load-pull
measurement setup that
incorporates an impedance
transformer, open-loop active
load-pull components and a
passive tuner

−π to π for plotting Eq. (5.30). It is evident from the plot in Fig. 5.21 that a small
deviation in |ΓSyst | changes the PLP required from the load-pull source drastically
for the synthesis of a reflection coefficient to match the 100 W DUT.
The above plot also reveals a strong dependence of ϕSyst on PLP . In this case
of |ΓSyst | = 0.05, the minimization of PLP requires ϕSyst to be either smaller than
−100° or larger than 100°. The dependence of PLP on the mismatch between the
circulator and impedance transformer is utilized as an advantage in hybrid load-pull
systems [18, 22]. For example, a block diagram of a hybrid load-pull setup is illus-
trated in Fig. 5.22, which incorporates a passive tuner, in order to intentionally create
a mismatch between the impedance transformer and the circulator to minimize PLP
from the load-pull source. Its optimal operation in any specified application requires
the availability of an appropriate impedance transformer, a line stretcher between the
DUT and the impedance transformer, and a tuning resolution of the passive tuner.
Further, the available power from the active load-pull source plays a key role in the
selection of the transformation ratio of the impedance transformer.
5.6 Calibration and Data Extraction 133

Fig. 5.23 Required PLP as a

function of the system
reflection coefficient phase,
ϕSyst , for varying system
reflection coefficients when
using a 50:7 impedance
transformer (for a 100 W
(1 ) device)

In measurement applications, the load-pulling of a device is initially carried out

to identify the Smith chart region and then to select an appropriate impedance trans-
former for the DUT, as well as a line stretcher between the device and impedance
transformer. Upon selection of the impedance transformer and the line stretcher, the
system has two degrees of freedom, namely the passive tuner and the active load-
pull source. In such a case, the passive tuner is first tuned appropriately to create
a mismatch between the circulator and the impedance transformer; and, load-pull
measurements are then performed using the load-pull source. The mismatch created
by the passive tuner reduces the power demand from the load-pull source and, there-
fore, aids in the optimal operation of the hybrid load-pull system. For example, plot
in Fig. 5.23 displays the dependence of PLP on the phase and magnitude of ΓSyst
for load-pulling a 100 W (1 ) device using the hybrid load-pull setup with a 50:7
impedance transformer. The phase and magnitude of ΓSyst are tuned by changing
the setting of the passive tuner.
It can be observed in Fig. 5.23 that even a fine adjustment in the magnitude and
phase of ΓSyst significantly reduces PLP . Thus, passive tuners with fine resolution
in magnitude and phase variation are extremely important for reducing PLP from
the load-pull source. It can, therefore, be concluded that a hybrid load-pull system
with an appropriate impedance transformer and a high-resolution passive tuner is
very suitable for high-power DUT load-pull measurements, as it minimizes the PLP
demand from the load-pull source and also reduces the VSWR and, consequently,
the voltage peaks in the measurement system.

5.6 Calibration and Data Extraction

Incorporation of an impedance transformer between the DUT and the tuner neces-
sitates a two-tier calibration, in order for the high-power load-pull setup to function
properly. For example, the load-pull setup in Fig. 5.22 requires two error adapters
134 5 High-Power Load-Pull Systems

Fig. 5.24 Reference plane definition and error adapters at ports 1 and 2 of the DUT for removal
of the imperfections in the components, connectors, board launchers and impedance transformers

at both ports 1 and 2 of the DUT, in order to account for the imperfections in the
system components and the impedance transformation networks. As illustrated in
Fig. 5.24, error adapters X account for the systematic errors introduced by the sys-
tem components, as well as the connectors, and reset the reference plane at the coax-
ial connectors used for connecting the measurement setup and the impedance trans-
former. Error adapters Y remove the systematic errors contributed by the impedance
transformer, launchers and test fixture and reset the reference plane at the DUT pack-
age.
The resetting of the error corrections and reference planes can be achieved by
using either the measurement-based VNA error correction procedures, full wave
analysis using numerical techniques or analytical methods based on approximate
closed-form expressions [32–34]. The analytical method is based on approximations
that have built-in uncertainty [34], as are the numerical techniques [33] in the form
of material parameter uncertainty.
The VNA error correction techniques rely entirely on measurements and are
regulated by well-defined mathematical expressions for the extraction of the error
5.6 Calibration and Data Extraction 135

Fig. 5.25 Flow graph of the

combined error adapters at
both the ports 1 and 2 of the
DUT

terms for each half of the fixture [32]. The SOLT (Short-Open-Load-Thru) and
TRL (Thru-Reflect-Line) techniques are the most popular calibration procedures
for VNA error correction. SOLT standards, however, are difficult to implement for
VNA measurements of microstrip and similar circuits [17]. The TRL technique,
which relies only on the characteristic impedance of a short transmission, is sim-
ple and provides the highest calibration accuracy over a wide bandwidth [35] and,
therefore, is preferred over the traditional SOLT technique.
The calibration of a high-power load-pull setup involves a first-tier calibration
using coaxial TRL standards to determine the parameters of error adapters X, in
order to reset the reference plane at the coaxial connectors shown in Fig. 5.24. The
second-tier calibration for the determination of parameters of error adapters Y in-
volves TRL calibration using in-fixture standards attached to the impedance trans-
forming network, which then transfers the reference plane to the DUT ports.
The extraction of the measurement data at the DUT plane is a two-stage process.
The first stage involves transformation of the parameters of error adapters X and
Y to a common reference impedance [35]. It should be noted that, in practice, er-
ror adapters X are referenced to 50 , which is the characteristic impedance of the
calibration standards; whereas, error adapters Y are referenced to the characteristic
impedance of the impedance transformer, which is usually smaller than 50 , neces-
sitating transformation of error adapters Y to 50 . After transformation, the error
adapters are combined to form combined error adapters at both ports of the DUT,
as illustrated by the flow graph in Fig. 5.25. Equations (5.31) to (5.38), which are
derived from Fig. 5.25, provide the error parameters of the combined error adapters.
50
ex01 ex10 ey00
e00 = ex00 + (5.31)
1 − ex11 ey00
50

50 e50 e
ey01 y10 x11
e11 = ey11
50
+ (5.32)
1 − ex11 ey00
50

50 e
ey10 x10
e10 = (5.33)
1 − ex11 ey00
50

50 e
ey01 x01
e01 = (5.34)
1 − ex11 ey00
50
136 5 High-Power Load-Pull Systems

Fig. 5.26 Measurement

system error adapters in terms
of ABCD parameters

50 e50 e
ey23 y32 x22
e22 = ey22
50
+ (5.35)
1 − ey33
50 e
x22
50
ex23 ex32 ey33
e33 = ex33 + (5.36)
1 − ey33
50 e
x22
50 e
ey32 x32
e32 = (5.37)
1 − ey33
50 e
x22
50 e
ey23 x23
e23 = (5.38)
1 − ey33
50 e
x22

where the terms with the superscript of 50 refer to the parameters of error adapters
Y transformed to 50  impedance.
In the second stage, all the S-parameter data, the parameters of error adapters
and the corrected measured data required at the DUT plane are converted to ABCD
parameters [36], as they are easy to manipulate in the cascaded structure shown in
Fig. 5.26. For example, the actual measured data Am , Bm , Cm and Dm at the VNA
port is given by the cascade of the three ABCD matrices, according to Eq. (5.39).
   
Am Bm A1 B1 A B A2 B2
= (5.39)
Cm D m C1 D 1 C D C2 D 2
Simplification of Eq. (5.39) then provides the desired corrected ABCD parameter of
the DUT, as given in Eq. (5.40).
 −1  −1
A B A1 B1 Am Bm A2 B2
= (5.40)
C D C1 D 1 Cm D m C2 D 2
The ABCD parameters obtained from Eq. (5.40) are then converted to S-parameters,
in order to determine the error corrected measured data of the DUT at the DUT
plane [36]. This overall process of error corrected S-parameter extraction from the
measured data at the VNA ports is also called the de-embedding technique [32].

References
1. D. Miller, M. Drinkwine, High voltage microwave devices, in International Conference on
Compound Semiconductor Manufacturing Digest (2003), pp. 1–4
References 137

2. M.H. Wong, S. Rajan, R.M. Chu, T. Palacios, C.S. Suh, L.S. McCarthy, S. Keller, J.S. Speck,
U.K. Mishra, N-face high electron mobility transistors with GaN-spacer. Wiley InterScience
204(6), 2049–2053 (2007)
3. Maury Microwave Corporation, Pulsed-bias pulsed-rf harmonic load-pull for gallium nitride
(GaN) and wide band gap (WBG), Application Note: 5A-043, Nov. 2009
4. B. Bonte, C. Gacquiere, E. Bourcier, G. Lemeur, Y. Crosnier, An automated system for mea-
suring power devices in Ka-band. IEEE Trans. Microw. Theory Tech. 46(1), 70–75 (1998)
5. P. Bouysse, J. Nebus, J. Coupat, J. Villotte, A novel accurate load pull setup allowing the
characterization of highly mismatched power transistors. IEEE Trans. Microw. Theory Tech.
42(2), 327–332 (1994)
6. F. De Groote, J.-P. Teyssier, J. Verspecht, J. Faraj, Introduction to measurements for power
transistor characterization. IEEE Microw. Mag. 9(3), 70–85 (2008)
7. J. Sevic, A sub 1  load-pull quarter wave prematching network based on a two-tier TRL
calibration, in 52nd Automatic RF Techniques Group (ARFTG) Conference Digest, Rohnert
Park, USA (1998), pp. 73–81
8. S. Basu, M. Fennelly, J.E. Pence, E. Strid, Impedance matching probes for wireless applica-
tions, Application Note, AR126, Cascade Microtech, 1998
9. J. Hoversten, M. Roberg, Z. Popovic, Harmonic load pull for high power microwave de-
vices using fundamental only load pull tuners, in 75th Automatic Radio Frequency Techniques
Group (ARFTG) Conference Digest, Anaheim, USA (2010), pp. 73–81
10. Maury Microwave Corporation, Device characterization with harmonic load and source pull,
Application Note: 5C-044, Dec. 2000
11. Focus Microwaves, Load pull measurements on transistors with harmonic impedance control,
Technical Note, August 1999
12. P. Colantonio, A. Ferrero, F. Giannini, E. Limiti, V. Teppati, Harmonic load/source pull strate-
gies for high efficiency PAs design, in IEEE MTT-S International Microwave Symposium Di-
gest (2003), pp. 1807–1810
13. I. Yattoun, A. Peden, A new configuration to improve the active loop technique for transis-
tor large signal characterization in the Ka-band. Microw. Opt. Technol. Lett. 49(3), 589–593
(2007)
14. M.S. Hashmi, A.L. Clarke, S.P. Woodington, J. Lees, J. Benedikt, P.J. Tasker, An accurate
calibrate-able multi-harmonic active load-pull system based on the envelope load-pull concept.
IEEE Trans. Microw. Theory Tech. 58(3), 656–664 (2010)
15. F. De Groote, J.-P. Teyssier, J. Verspecht, J. Faraj, High power on-wafer capabilities of a time
domain load-pull setup, in IEEE MTT-S International Microwave Symposium, Atlanta, USA
(June 2008), pp. 100–102
16. Focus Microwaves, Accuracy and verification of load pull measurements, Application Note
18, Sept. 1994
17. R. Ludwig, P. Bretchko, RF Circuit Design—Theory and Applications, 1st edn. (Prentice Hall,
New York, 2000). ISBN 0-13-095323-7
18. Z. Aboush, Design, characterization and realization of thin film packaging for both broadband
and high power applications, PhD Thesis, Cardiff University, 2007
19. Z. Aboush, J. Lees, J. Benedikt, P.J. Tasker, Active harmonic load pull system for charac-
terizing highly mismatched high power transistors, in IEEE MTT-S International Microwave
Symposium Digest, vol. 3 (June 2005), pp. 1311–1314
20. F.V. Raay, G. Kompa, Waveform measurements—the load-pull concept, in 55th Automatic
Radio Frequency Techniques Group (ARFTG) Conference Digest, Boston, USA, vol. 37 (June
2000), pp. 1–8
21. F. De Groote, O. Jardel, J. Verspecht, D. Barataud, J.-P. Teyssier, R. Quere, Time domain
harmonic load-pull of an AlGaN/GaN HEMT, in 66th Automatic Radio Frequency Techniques
Group (ARFTG) Conference Proceedings, Washington, USA (December 2005)
22. V. Teppati, A. Ferrero, U. Pisani, Recent advances in real-time load-pull systems. IEEE Trans.
Instrum. Meas. 57(11), (2008)
138 5 High-Power Load-Pull Systems

23. M.S. Hashmi, A.L. Clarke, S.P. Woodington, J. Lees, J. Benedikt, P.J. Tasker, Electronic
multi-harmonic load-pull system for experimentally driven power amplifier design optimiza-
tion, in IEEE MTT-S International Microwave Symposium Digest, Boston, USA (June 2009),
pp. 1549–1552
24. F.M. Ghannouchi, M.S. Hashmi, S. Bensmida, M. Helaoui, Loop enhanced passive source-
and load-pull technique for high reflection factor synthesis. IEEE Trans. Microw. Theory Tech.
58(11), 2952–2959 (2010)
25. Focus Microwaves, Load-pull for power transistor characterization, Application Note, 2003
26. C. Roff, J. Graham, J. Sirois, B. Noori, A new technique for decreasing the characterization
time of passive load-pull tuners to maximize measurement throughput, in 72nd Automatic
Radio Frequency Techniques Group (ARFTG) Conference Digest, Portland, USA (Dec. 2008),
pp. 92–96
27. Focus Microwaves, Using transformers in harmonic load pull, Technical Note TN-3-2008,
Dec. 2008
28. Focus Microwaves, Comparing harmonic load-pull techniques with regards to power-added
efficiency (PAE), Application Note 58, May 2007
29. M. Golio, J. Golio, RF and Microwave Circuits, Measurements, and Modelling (CRC Press,
Boca Raton, 2008). ISBN 978-0-8493-7218-6
30. D.M. Pozar, Microwave Engineering, 3rd edn. (Wiley, New York, 2005). ISBN 0-471-17096-
8
31. G.F. Engen, C.A. Hoer, Thru-reflect-line: an improved technique for calibrating the dual six-
port automatic network analyzer. IEEE Trans. Microw. Theory Tech. 27, 987–993 (1979)
32. D. Rytting, Appendix to an analysis of vector measurement accuracy enhancement techniques,
in Hewlett-Packard RF Microwave Symposium Digest (March 1982), pp. 1–42
33. EM User’s Manual, v. 10, Sonnet Software Inc., Liverpool, NY, 2008
34. B. Wadell, Transmission Line Handbook (Artech House, Boston, 1991)
35. D. Williams, C.M. Wang, U. Arz, An optimal multiline TRL calibration algorithm, in IEEE
MTT-S International Microwave Symposium Digest, vol. 3 (June 2003), pp. 1819–1822
36. D.A. Fricky, Conversions between S, Z, Y, h, ABCD, and T parameters which are valid for
complex source and load impedances. IEEE Trans. Microw. Theory Tech. 42(2), 205–211
(1994)
37. M.S. Hashmi, F.M. Ghannouchi, P.J. Tasker, K. Rawat, Highly reflective load pull. IEEE Mi-
crow. Mag. 12(4), 96–107 (2011)
38. Z. Aboush, C. Jones, G. Knight, A. Sheikh, H. Lee, J. Lees, J. Benedikt, P.J. Tasker, High
power active harmonic load-pull system for characterization of high power 100 Watt transis-
tors, in IEEE 35th European Microwave Conference, vol. 1(4) (Oct. 2005), 4 pp.
Chapter 6
Envelope Load-Pull System

This chapter presents the fundamental concept of an active envelope load-pull (ELP)
system. Subsequently, the realization and design, along with its characteristics, are
presented and discussed. The calibration procedure of the ELP system, which is
rapid and increases load-pull measurement throughput, is also described in detail.
Finally, the harmonic ELP and its characteristics, which lead to valuable measure-
ments and investigations, are presented.

6.1 Introduction

Both traditional passive and active load-pull techniques are commercially available
[1–4]. In practice, however, most of these systems do not meet all the load-pull
measurement requirements appropriately, such as in the design and optimization of
harmonically tuned power amplifiers or high throughput radio frequency (RF) and
microwave load-pull measurement applications. For example, the inability of the
passive load-pull technique to synthesize harmonically independent reflection coef-
ficients limits the use of such systems in the design of high-efficiency power ampli-
fiers, such as class-F [5, 6]. Additionally, high load-pull measurement throughput
requires pre-calibrated tuners at all tuner positions and settings, which can turn out
to be lengthy and cumbersome procedures. Pre-calibration of passive tuners, albeit
slow, are possible before their deployment in load-pull systems; however, the pre-
calibration of classic active load-tuners is not possible.
An envelope load-pull (ELP) system can be calibrated rapidly and efficiently and
is also capable of synthesizing harmonically independent reflection coefficients [7–
9]. These two features make it most suitable for applications, such as semiconductor
device characterization for broadband power amplifiers design and their optimiza-
tion [10–12]. It is also suited for the evaluation of transistor devices for load- and
supply-modulation applications [13], as well as the rapid experimental investigation
of transistor devices for achieving high-efficiency and high-measurement through-
put [7, 14]. Additionally, the ELP system is also capable of synthesizing drive- and

F.M. Ghannouchi, M.S. Hashmi, Load-Pull Techniques with Applications to Power 139
Amplifier Design, Springer Series in Advanced Microelectronics 32
DOI 10.1007/978-94-007-4461-5_6, © Springer Science+Business Media Dordrecht 2013
140 6 Envelope Load-Pull System

Fig. 6.1 Generic block diagram illustrating the concept of the ELP technique

bias-level independent reflection coefficients and, thus, can work better in several
load-pull applications than other active load-pull methods. These applications may
be nonlinear device modeling [15, 16], determination of solid-state device perfor-
mance under varying drive and bias conditions [17–21], or waveform engineering
[22].

6.2 Envelope Load-Pull Concept

The operating methodology of the ELP system is similar to the active closed-loop
active load-pull technique. The major difference between the two techniques is the
manner in which the traveling wave is modified before being fed back for the synthe-
sis of the desired reflection coefficient at the DUT reference plane. The ELP utilizes
external control variables, X and Y , for modifying the phase and amplitude of the
reflected traveling wave, a2 , as shown in Fig. 6.1. Furthermore, in ELP system, the
modification in the forward traveling wave and then its feedback to the device takes
place at the baseband or at an intermediate frequency, which shields the system from
any oscillation at the carrier RF frequency.

6.2.1 Mathematical Formulation

Figure 6.1 depicts a generic block diagram of the ELP system. It consists of a
quadrature demodulator that down-converts the incident traveling wave into base-
band signals, a control unit that modifies the baseband signal components gener-
ated by the quadrature demodulator through external control variables X and Y ,
a quadrature modulator that up-converts the modified baseband signal components,
and a local oscillator (LO) that provides the desired signal to the quadrature de-
modulator and modulator. In principle, the circulator works as a perfect three-port
component with the ability to transfer signals only one way between two respective
6.2 Envelope Load-Pull Concept 141

Fig. 6.2 Block diagram illustrating a realized ELP setup [8], © IEEE 2010

ports. As shown in Fig. 6.1, this helps in completely directing the incident traveling
wave, b2 , to the quadrature demodulator and the up-converted reflected traveling
wave, a2 , to the DUT plane for the synthesis of the reflection coefficient given by
Eq. (6.1).
a2
ΓLoad = (6.1)
b2
The synthesized reflection coefficient at the DUT plane is regulated by the control
function, F (X, Y ), given in Eq. (6.2). The control function is also called the reflec-
tion coefficient, Γset , formulated by the control unit.
F (X, Y ) = Γset = X + j Y (6.2)
According to the ELP concept [23, 24], in an ideal situation, the synthesized reflec-
tion coefficient at the DUT plane, ΓLoad , should be equal to the control function,
F (X, Y ); therefore, Eq. (6.3) can be written as:
a2
X + jY = (6.3)
b2
Equations (6.4) and (6.5) give the incident and reflected waves, b2 and a2 , in
terms of their respective quadrature baseband components, Ib , Qb , Ia and Qa .

b2 (t) = Ib (t) + j Qb (t) (6.4)

a2 (t) = Ia (t) + j Qa (t) (6.5)
Equations (6.3), (6.4), (6.5) can be simplified to obtain the ELP equations given
in Eqs. (6.6) and (6.7).

Ia = Ib X − Qb Y (6.6)
Qa = Qb X + Ib Y (6.7)
The ELP equations relate control variables X and Y to the quadrature baseband
components of traveling waves b2 and a2 .
142 6 Envelope Load-Pull System

6.3 Practical Realization

Figure 6.2 shows a block diagram of a realized ELP system. It can be observed that
the realization of the system requires one quadrature demodulator, one quadrature
modulator, a control unit, one loop amplifier and some passive components, such as
cables, connectors, directional coupler or circulator, and attenuators.
Either an active or a passive demodulator can be used for the realization of the
system. An active direct conversion demodulator provides in principles a wider
bandwidth and higher operating dynamic range, compared to the passive demod-
ulator, but also suffers from a bias-dependent inherent DC offset. The presence of
varying DC offsets degrades the overall performance of the system [7, 8]. Ideally,
the complete removal of DC offsets is essential for the calibration of the setup. In
order to remove the DC offsets, the ELP system needs to be operated in a heterodyne
mode [9].
The control unit is an electronic circuit with the functionality to perform multi-
plication and addition/subtraction. It takes computer-generated control variables X
and Y , which act on the baseband components, Ib and Qb , in order to synthesize the
active ΓLoad using the ELP equations (Eqs. (6.6) and (6.7)). A directional coupler or
circulator can be employed as a three-port component for directing the incident and
reflected waves in the desired direction. For example, the realized setup employs a
directional coupler, as it provides better isolation between the incident and reflected
waves compared to a circulator.
Better isolation is key in achieving accurate calibration, as described in Sect. 6.4.
A circulator, however, can be used in place of the directional coupler in measure-
ment applications where the desired accuracy in the synthesized reflection coeffi-
cients is not too stringent.
The quadrature modulator used in the system can be either active or passive and
should be chosen according to the required measurement bandwidth, as an active
quadrature modulator provides higher bandwidth than that of a passive modulator.
A loop amplifier, AMP, is used to boost the amplitude of the reflected traveling wave,
in order to enhance the tuning range of the synthesized reflection coefficient at the
DUT plane; whereas, the attenuator, Atten, controls the level of the incident travel-
ing wave in the feedback loop of the system. The ELP system stability is heavily
dependent on the settings of AMP and Atten [8]; therefore, it is essential to choose
appropriate settings, in order to achieve and guarantee a stable operation over the
whole dynamic range and bandwidth of the system.

6.3.1 Design of Control Unit

The control unit processes the ELP equations (Eqs. (6.6) and (6.7)) consisting of
four multiplication, one addition and one subtraction terms. In an analogue tech-
nique, four multipliers, one adder and one subtractor are required, as shown in
Fig. 6.3.
6.3 Practical Realization 143

Fig. 6.3 Generic configuration of a control unit for the processing of the ELP equations

The control unit requires DC blocks at the input to suppress the DC offset if
an active demodulator is used for the design of the ELP setup. The DC blocks at
the output of the control unit discard any DC offsets generated by the multiplier or
adder/subtractor units, while passing only the useful information contained in the
baseband intermediate frequency (IF) signal to subsequent next stages for further
processing.
In the heterodyne mode, as expressed below, the quadrature demodulator down-
converts the incident traveling wave, given in Eq. (6.8), by mixing it with an LO
signal, given in Eq. (6.9), and generates the baseband signal components, given in
Eqs. (6.10) and (6.11). The higher mixing components are filtered out by low-pass
filters, LP
b2 (t) = R cos(ωs t − α) (6.8)
LO = S cos(ωs t + δωs t) (6.9)
  
RS
Ib (t) = LP 2 cos(ωs t − α) cos(ωs t + δωs t)
2
= A cos(δωs t + α) (6.10)
  
RS
Qb (t) = LP 2 cos(ωs t − α) sin(ωs t + δωs t)
2
= A sin(δωs t + α) (6.11)
where A = (RS/2) and α represent the magnitude and phase of the baseband signals,
ωs is the stimulus frequency, and δωs is the frequency offset between the stimulus
and the LO.
In practical realizations, however, there can be imbalances in the magnitude and
phase of the baseband components, Ib and Qb . For example, Eqs. (6.12) and (6.13)
are the quadrature baseband components, with the terms L and K representing the
amplitudes of Ib and Qb , respectively, when there is a magnitude imbalance be-
tween the two.
Ib (t) = L cos(δωs t + α) (6.12)
Qb (t) = K sin(δωs t + α) (6.13)
144 6 Envelope Load-Pull System

Fig. 6.4 Analogue domain implementation of a modified control unit to suppress the image cre-
ation in the ELP feedback loop [9], © IOP Measurement Science and Technology 2010

The processing of these baseband components in the control unit, through the use
of control variables X and Y , and then up-conversion by the quadrature modulator
using the LO signal in Eq. (6.9) produces the reflected traveling wave, a2 , given in
Eq. (6.14).
  
L+K
a2 (t) = cos(ωs t − α)
2
  
L−K
+ cos(ωs t − α + 2δωs t) |Γset |ej θ (6.14)
2
√
where |Γset | = X 2 + Y 2 is the magnitude and θ = tan−1 (Y/X) is the phase of the
control function F (X, Y ).
The reflected traveling wave in Eq. (6.14) contains two components—one is use-
ful, whereas the other is its image. The image generation is the result of an imbalance
in the magnitudes of baseband signals Ib and Qb . An imbalance in the phase of Ib
and Qb also creates an image signal in the system [9]. The image due to a phase
imbalance, however, falls far away from the carrier signal and, therefore, does not
affect the performance of the ELP system. However, the image due to an amplitude
imbalance lies very close to the carrier signal and, as a result, greatly affects the
performance of the system [9]. As a consequence, the imbalances between Ib and
Qb must be corrected, in order to suppress the image and attain good load-pull per-
formance from the system [7–9]. This can be achieved by balancing the magnitude
imbalance of Ib and Qb generated by the demodulator through the incorporation of
a balance bridge, as shown in Fig. 6.4.
The balancing in the amplitude of the baseband components suppresses the image
in the up-converted signal given in Eq. (6.15) and, therefore, creates the required
reflected traveling wave for the synthesis of a desired reflection coefficient at the
DUT plane.
 
a2 (t) = A cos(ωs t − α) |Γset |ej θ (6.15)
6.4 ELP Calibration 145

Fig. 6.5 FPGA-based

implementation of the control
unit

The control unit can also be designed using a digital platform, such as a field-
programmable gate array (FPGA), as shown in Fig. 6.5 [25]. Analogue-to-digital
converters (ADCs) transform the baseband Ib and Qb into digital bit streams,
which are then modified by on-board generated variables X and Y . The modified
bit streams are then transformed into the analogue domain by digital-to-analogue
converters (DACs). DC offsets from Ib and Qb are removed using digital filtering,
while the balancing is done on-board using different scaling for the I and Q paths.
The on-board global clock synchronizes the whole baseband processing required for
the implementation of the ELP equations (Eqs. (6.6) and (6.7)).

6.4 ELP Calibration

The purpose of ELP system calibration is the correction of the imperfections in-
troduced by the components, demodulator, control unit, modulator, directional cou-
pler, cables and connectors, in order for it to function properly. After calibration,
the reflection coefficients at the DUT plane can be precisely synthesized through
the variation of control variables X and Y .

6.4.1 Error Flow Model Formulation

The first step in ELP calibration is the formulation of the error flow model for the
system, which requires identification of the sources of systematic errors, as illus-
trated in Fig. 6.6. These errors in the system are introduced, due to the non-ideal
nature of demodulator and modulator, imperfect feedback effect caused by the di-
rectional coupler, and delays and losses in the cables. The term TD accounts for the
demodulator conversion gain and the losses and delays associated with the physical
setup on the demodulator side; and, the term TM refers to the modulator conver-
sion gain and the losses and delays associated with the cables and connectors on
the modulator side. The term Γ0 is the passive impedance of the network, while the
error term, ΓF , accounts for the imperfections in the feedback and system isolation.
The impact of systematic errors in the ELP system can be analyzed using the
error flow model illustrated in Fig. 6.7. The error model assumes that an active
demodulator has been used and, therefore, includes error terms D and M caused by
the DC offset present in the baseband signal components. Error terms D and M are
146 6 Envelope Load-Pull System

Fig. 6.6 Block diagram of an ELP system depicting the sources of systematic errors [8], © IEEE
2010

Fig. 6.7 Error flow model

accounting for the errors
introduced by the components
used in the design of the ELP
system [8], © IEEE 2010

absent from the error flow model in case a passive demodulator is employed in the
design of the system. The term Γset = X + j Y refers to the required load reflection
coefficient established by control variables X and Y .

6.4.2 Simplification of the Error Flow Model

The error flow model can be simplified in five steps, using the flow graph reduction
technique [26] shown in Fig. 6.8.
Equation (6.16), which relates traveling waves a2 and b2 , can be deduced from
the last stage of the simplification process.
   
1 M Γset TD TM
a2 = b2 + D+ + Γ0 (6.16)
TD Γset 1 − ΓF (Γset TD TM )
The presence of DC offset terms D and M does not allow for the determination of
the direct ratio of the traveling waves a2 and b2 and, hence, the reflection coeffi-
cient at the DUT plane, which prevents the straightforward calibration of the ELP
system. In order to discard the DC offset terms and to achieve a calibrated ELP sys-
tem, a passive demodulator should be used or an active demodulator and DC blocks
should be employed. The removal of terms D and M from Eq. (6.16) enables a
direct relationship between a2 and b2 (Eq. (6.17)) and, therefore, provides the ELP
6.4 ELP Calibration 147

Fig. 6.8 Simplification of the error flow model of the ELP system

system with the ability to directly relate the synthesized reflection coefficient, ΓLoad ,
to the parameters of the error model.
 
a2 Γset TD TM
= ΓLoad = + Γ0 (6.17)
b2 1 − ΓF (Γset TD TM )
Equation (6.17), which is a first-order control equation describing the behavior of
the ELP setup, can be simplified to obtain a linear equation, as given in Eq. (6.20).
 
Γset G
ΓLoad = + Γ0 (6.18)
1 − ΓF (Γset G)

ΓLoad − ΓLoad Γset (ΓF G) = Γ0 + Γset G(1 − Γ0 ΓF ) (6.19)
ΓLoad = A + BΓLoad Γset + CΓset (6.20)
where G represents the factor TD TM and is termed as the gain of the ELP loop, and
A = Γ0 , B = ΓF G and C = G(1 − Γ0 ΓF ).
It can be observed that the simplified equation (Eq. (6.20)), which is analogous to
the 1-port error model calibration equation of any standard vector network analyzer
(VNA) [27], relates the measured reflection coefficient at the DUT plane to the
148 6 Envelope Load-Pull System

Fig. 6.9 Block diagram of an

ELP setup integrated in a
waveform measurement
system for the purpose of
calibration [9], © IOP
Measurement Science and
Technology 2010

reflection coefficient set by the ELP system and the error terms introduced by the
imperfect system components.

6.4.3 Calibration Technique

The three unknown terms, A, B and C, in Eq. (6.20) need to be determined, in

order to calibrate the ELP system. In theory, calibration measurements on three
distinct values of reflection coefficients, which are called ELP calibration standards
and are defined through control variables X and Y , can enable the determination
of the unknowns, A, B and C. In practice, however, it is preferable to perform
calibration measurements on a higher number of calibration standards, in order to
minimize random errors, and then to use the least squares method [28] to determine
the unknowns, A, B and C. The ELP calibration standards can be increased just by
varying variables X and Y .
Determination of the unknowns, A, B and C, enable the extraction of the error
terms, Eqs. (6.21) to (6.23), of the error flow model.

Γ0 = A (6.21)
G = TD TM = C + BA (6.22)
B
ΓF = (6.23)
G
Finally, substitutions of the above equations into Eq. (6.17) allow for the pre-
diction of the precise values of X and Y needed for the synthesis of any reflection
coefficient, ΓLoad , at the DUT plane. Calibration of the ELP setup requires a vector
corrected network analyzer, such as a waveform measurement system [29–35], an
example of which is the two-channel, pre-calibrated, time-domain waveform mea-
surement system shown in Fig. 6.9, and a step-by-step calibration procedure, as
outlined in Fig. 6.10.
6.4 ELP Calibration 149

Fig. 6.10 Flow chart

outlining the steps for
systematically calibrating the
ELP system [8], © IEEE 2010

As a first step, distinct calibration standards of thirty points, which are obtained
by varying variables X and Y , are defined and measured. The calibration stan-
dards and the measured reflection coefficients shown in Fig. 6.11 do not match up
and, therefore, experimentally verify the need for ELP calibration. Error terms are
150 6 Envelope Load-Pull System

Fig. 6.11 Uncalibrated

performance of an ELP
system over a 30-point data
set [8], © IEEE 2010

calculated using Eqs. (6.18) to (6.23), and the system checks if the gain, G, is realiz-
able. G is dependent on the attenuator and loop amplifier settings (Fig. 6.6), which
need to be changed appropriately until the check returns positive confirmation.
With a positive result for the gain check, the system is checked for stability by
considering the denominator of Eq. (6.18). In order to pass the stability test, the
error terms must satisfy Eq. (6.24).

ΓF (Γset G)
< 1 (6.24)
In case the stability test returns a negative result, the components of the system
(Fig. 6.6) need to be readjusted suitably and re-measured until the test returns pos-
itive confirmation, by satisfying the stability condition in Eq. (6.24). The readjust-
ment may be in the settings of the attenuator and/or loop amplifier, in the power
level of the LOs feeding the modulator and demodulator, or in the chosen calibra-
tion standards. In the worst case, a directional coupler with better isolation between
its coupled and through ports may be required to achieve stability.
With a positive stability check, the calibration standards need to be re-measured,
but with a modified set given by Eq. (6.25), in order to verify the calibration proce-
dure.
 
1 (ΓLoad )meas − Γ0
Γset = (6.25)
G ΓF ({(ΓLoad )meas − Γ0 } + 1)
The results, as illustrated in Fig. 6.12, show a perfect match between the cali-
bration standards and the measured reflection coefficients and, therefore, verify the
calibration procedure.

6.4.4 Evaluation of the Calibration Technique

The accuracy of the calibration can be quantified in terms of percentage error, e,

using Eq. (6.26), where N is the number of calibration standards. It is important
6.4 ELP Calibration 151

Fig. 6.12 Measured

reflection coefficients after
incorporation of error terms
over a 30-point data set [8],
© IEEE 2010

Table 6.1 Average

difference between the No. of e (%) No. of e (%)
calibration standards and the calibration calibration
measured reflection standards, N standards, N
coefficients after calibration
and verification [8], © IEEE 3 0.0456 12 0.0258
2010 8 0.0337 20 0.0239
10 0.0271 30 0.0224

to note that a minimum of three calibration standards are needed to determine the
terms of the ELP error model; therefore, Eq. (6.26) starts from N = 3.
 
 N
|(ΓLoad )meas − Γset |
e= 1 N × 100 (6.26)
|Γset |
3

The percentage errors for different numbers of calibration standards are listed in
Table 6.1, which shows a substantial improvement in accuracy with the increasing
number of calibration standards. As a compromise between the measurement speed
and the calibration accuracy, it is reasonable to assume that a calibration procedure
utilizing 12–20 calibration standards provides suitably accurate results [7].
Table 6.2, which provides information about the ELP calibration at varying fre-
quencies, shows that the calibration accuracy is independent of the operating fre-
quency.
Figure 6.13 illustrates the comparison between the calibrated measured reflection
coefficients and the desired reflection coefficients at the extreme edge of the Smith
chart. This is a stringent measure to test the calibration accuracy; and, in this case,
the two sets of reflection coefficients overlay each other, thereby demonstrating good
calibration accuracy.
It can, therefore, be concluded that the ELP calibration, which is independent of
the operating frequency, the number of calibration standards, or the location of the
152 6 Envelope Load-Pull System

Table 6.2 Average

difference between the 12 Carrier e (%) Carrier e (%)
calibration standards and the frequency frequency
respective measured (MHz) (MHz)
reflection coefficients [8],
© IEEE 2010 850 0.0252 1800 0.0258
900 0.0255 2100 0.0259

Fig. 6.13 Evaluation of the

ELP calibration process over
a 36-point data set [16],
© IEEE 2010

desired reflection coefficients, is accurate and versatile and, thus, helps in designing
a reliable active load-pull measurement system as compared to open-loop active
load-pull systems [13, 14].
The ELP calibration speed is another key parameter in improving the load-pull
measurement throughput and speeding up of power amplifier design/optimization
[7]. For example, an ELP system that integrated into a two-channel waveform mea-
surement setup [33] takes 15 minutes when subjected to 30 point calibration stan-
dards; whereas, the calibration evaluation over 36 points takes a further 18 minutes
[7, 8]. Therefore, on average, an ELP system can be calibrated, evaluated and then
employed in load-pull measurement applications in just over 30 minutes. As a com-
parison, passive load-pull systems typically need pre-characterization of tuners over
hundreds of points, covering all frequency ranges, thus requiring significantly more
time [36].
The key features of ELP system calibration can be summarized in the following
points:
• The equation representing the error flow model of an ELP setup is similar to
1-port error flow model equation of any standard vector network analyzer.
• The calibration can be carried out with as low as three calibration standards.
6.5 Stability Analysis 153

Fig. 6.14 Frequency response of the control unit of a prototype ELP setup

• Calibration accuracy can be improved by increasing the number of calibration

standards. The standards are not physical, but can be created through variations
of control variables, X and Y .
• The accuracy of calibration is not dependent on the operating frequency.
• The calibration is rapid and, therefore, improves the load-pull measurement
throughput.

6.5 Stability Analysis

The ELP system is based on a closed-loop feedback concept and, thus, has the po-
tential of creating instability in the measurement setup [37]. Equation (6.18) can be
analyzed to address the stability issues in the ELP system. The system is free from
oscillations if the denominator in Eq. (6.18) satisfies the following relationship:

ΓF (Γset G)
< 1 ⇒ |Γset | < 1 (6.27)
|ΓF G|
From Eq. (6.27), it can be deduced that the ELP system never goes into oscilla-
tion as long as the magnitude of the desired reflection coefficient, Γset , synthesized
through the variation of X and Y does not go beyond the product of the error terms
ΓF and G. In order to achieve this condition, the directional coupler must have a
high-isolation capability and the gain/attenuation in the loop must be appropriately
regulated through the adjustment of Atten and AMP in Fig. 6.6, so as to obtain a
higher product of ΓF and G.
Alternatively, the frequency response of the baseband section of the system can
be monitored to identify and mitigate any oscillation in the loop. For example,
Fig. 6.14 illustrates the baseband frequency response of a typical prototype ELP
system [39]. It can be observed that the response rolls off beyond −160 kHz and
+160 kHz, without any peaking characteristics beyond this range. There is also a
stop band in the response, due to the DC blocks used in the ELP prototype. The setup
must be calibrated and operated in the maximum gain region, in this case between
−160 kHz and +160 kHz, in order to ensure that the oscillation never occurs.
154 6 Envelope Load-Pull System

Fig. 6.15 Measured

magnitude and phase of the
reflection coefficient on a
THRU standard at a carrier
frequency of 1.8 GHz for a
constant value of the control
variables X and Y , as drive
power is swept from
−38 dBm to −18 dBm [23],
© IEEE 2005

On a side note, the baseband section in the ELP setup functions as a default band-
pass filter, removing the need for any additional highly selective filters to achieve
stability, which are normally required in the traditional closed loop active load-pull
system [38].

6.6 Features of the Envelope Load-Pull System

Load-pull measurements and characterizations often involve input power, Pin ,
sweeps at several load reflection coefficients, ΓLoad , in order to either extract op-
timal performance from transistor devices or validate nonlinear transistor models at
varying drive levels. Any load-pull system exhibiting drive-level independent reflec-
tion coefficients synthesis capability is, therefore, appropriate for such applications,
as they can perform measurements in a rapid manner. An ELP system also fulfills
this criterion, as illustrated in Fig. 6.15, and, thus, satisfies the requirements of swept
power load-pull measurement applications. This result has been achieved using the
ELP configuration of Fig. 6.10; and, in this particular case, the setup synthesizes
non-varying reflection coefficients over a drive level from −38 to −18 dBm and,
thus, exhibits a dynamic range of 20 dB [23]. The dynamic range of the ELP system,
is dependent on the dynamic range of the components used to assemble the system
and can be enhanced by employing components that possess higher dynamic ranges
and better performance.
The ELP system can be easily reconfigured to meet the requirements of high-
power device measurements and characterizations. The required baseband process-
ing for reflection coefficient synthesis at the DUT plane for any device is the same,
except that with distinct devices it will require appropriate readjustment of the at-
tenuator and loop amplifier at the ELP module plane, as illustrated in Fig. 6.16.
However, it is important to note that any change in the settings of the attenuator
and/or loop amplifier invalidates the current calibration, necessitating recalibration
of the ELP setup.
Some applications, such as class-F mode power amplifier design [6], require a
load-pull system capable of synthesizing bias-level independent reflection coeffi-
cients. The ELP system, like the traditional passive load-pull systems, has the ability
to synthesize bias-level independent reflection coefficients [39].
6.7 Harmonic Envelope Load-Pull System 155

Fig. 6.16 Block diagram

depicting the ELP module
plane, which needs to be
reconfigured appropriately for
meeting the requirements of
high-power devices

Fig. 6.17 Configuration of

the 3-harmonic ELP system
within the two-channel
time-domain waveform
measurement system [8],
© IEEE 2010

6.7 Harmonic Envelope Load-Pull System

The ELP system can be easily extended for multi-harmonic applications. The ELP
approach is an active technique; therefore, a triplexer-based harmonic architecture,
as shown in Fig. 6.17, is preferred, since it provides better isolation between the
synthesized reflection coefficients at the fundamental and harmonic frequencies.
The three-harmonic ELP setup is similar to the fundamental setup, with additional
triplexers before and after the bank of three ELP modules to separate and then re-
combine the frequency components of the incident and reflected traveling waves, b2
and a2 respectively.
The utilization of a harmonic ELP system in measurement applications requires
calibration of each of the ELP modules based on the calibration technique described
in Sect. 6.4. Table 6.3 provides the calibration data for the three ELP modules. It
can be observed that the accuracy of the calibration is independent of the harmonic
power and frequency. This result, thus, conveys that the ELP calibration technique
is independent of the frequency and the power level of the traveling waves a2 and
b2 .
156 6 Envelope Load-Pull System

Table 6.3 Average

difference between the No. of Fundamental Second Third
calibration standards and the calibration (F0 ) harmonic harmonic
measured reflection standards, N (F2 ) (F3 )
coefficients at each harmonic
frequency, after calibration 12 0.0252 0.0258 0.0276
and verification at a 20 0.0237 0.0243 0.0259
fundamental frequency of
10 0.0271 30 0.0224
850 MHz [8], © IEEE 2010

Fig. 6.18 Measurement data

depicting drive and
harmonically independent
reflection coefficient
emulation capability of the
harmonic ELP system [7],
© IEEE 2009

In principle, harmonic load-pull applications require the capability of synthesiz-

ing harmonically independent reflection coefficients. This feature is essential for
accurate performance estimation of transistor devices. In practice, however, with
the exception of the traditional closed-loop active load-pull architecture, none of
the load-pull systems possess this capability. On the other hand, the ELP can read-
ily synthesize harmonically independent reflection coefficients [7, 8], as illustrated
in Fig. 6.18. This figure depicts measured harmonic load reflection coefficients on
a 1 W gallium arsenide (GaAs) field-effect transistor (FET) for the 20 dB drive
power sweep at the fundamental frequency of 850 MHz and shows that the har-
monic impedance does not move during the power sweep. This characteristic of the
harmonic ELP technique is unique and extremely useful in accurate, rapid and reli-
able harmonic load-pull measurement applications, such as the harmonic load-pull
optimization of the transistor device illustrated in Fig. 6.19.
The results in Fig. 6.19 show ELP harmonic load-pull measurement data for a
1 W GaAs FET device when the fundamental impedance has been swept over a 4×4
impedance grid; whereas, for each fundamental, the phase of the second-harmonic
impedance has been varied in steps of 45 degrees around the extreme edge of the
Smith chart, and the third-harmonic impedance has been held at 50 . It can be
observed that the harmonic impedances are uncoupled and unaffected by each other.
In this investigation, the whole sweep involved capturing only 128 measurements
using the harmonic ELP technique, which with an open-loop active harmonic load-
6.8 Unique Measurement Applications 157

Fig. 6.19 Measurement data

illustrating the independent
harmonic reflection
coefficient synthesis
capability of the harmonic
ELP system [8], © IEEE 2010

Fig. 6.20 Measurement of

PA design parameters by
sweeping the phase of the
second-harmonic reflection
coefficient, Γ (F2 ), while
maintaining constant Γ (F0 )
and Γ (F3 ) [9], © IOP
Measurement Science and
Technology 2010

pull system would require at least 15 times more (1920) measurements, due to the
system needing to iterate and converge on the harmonic impedances [22]. This fea-
ture of the ELP system results, in most cases, reduced characterization time and,
hence, increased measurement throughput.

6.8 Unique Measurement Applications

The independent harmonic reflection coefficient synthesis capability of the ELP sys-
tem enables investigation into the impact of harmonic terminations and their pre-
cise effects on the power amplifier (PA) design parameters, in order to optimize the
transistor devices for PA design applications. For example, the impact of a second-
harmonic termination on a 1 W GaAs FET device that has been subjected to its opti-
mal fundamental reflection coefficient, Γ (F0 ), and matched third-harmonic reflec-
tion coefficient, Γ (F3 ), into 50  is illustrated in Fig. 6.20. In this investigation, the
phase of the second-harmonic reflection coefficient, Γ (F2 ), has been swept every 10
degrees for |Γ (F2 )| = 1; and, data has been collected on a two-channel waveform
158 6 Envelope Load-Pull System

Fig. 6.21 Output fundamental power contours from sweeping third-harmonic reflection coeffi-
cients, Γ (F3 ), over 12 × 12 and 6 × 4 grids, holding Zopt (F0 ) = 36.3 + j 9.9  [7], © IEEE 2009

Fig. 6.22 Drain efficiency contours from sweeping third-harmonic reflection coefficients, Γ (F3 ),
over 12 × 12 and 6 × 4 grids, holding Zopt (F0 ) = 36.3 + j 9.9  [7], © IEEE 2009

measurement setup. It can be observed that the PA design parameters are highly
sensitive to variations of the phase of the second-harmonic reflection coefficient.
In practice, this investigation can be carried out using traditional active and pas-
sive load-pull systems; however, the accuracy and reliability in the measurement
data remain questionable, considering the inability of these systems to synthesize
harmonically independent reflection coefficients. Furthermore, this type of inves-
tigation usually requires significantly more time (tenfold) on a similar open-loop
harmonic active load-pull system compared to the harmonic ELP system [39].
6.8 Unique Measurement Applications 159

The rapid and harmonically independent reflection coefficient synthesis capabil-

ity of the ELP system can be employed in measurements where the effect of a large
number of impedances on a device needs to be investigated for their optimization
in PA design applications. Figures 6.21 and 6.22, for example, illustrate the effect
of third-harmonic reflection coefficient variation on the output power and drain ef-
ficiency, respectively, of a 1 W GaAs FET device at a fundamental frequency of
850 MHz.
In this investigation, fundamental and second-harmonic reflection coefficients
remain fixed to their optimal values and 50 , respectively. The measurements in-
volve a 12 × 12 third-harmonic reflection coefficient grid on a broader area of Smith
chart, in order to check the sensitivity of its variation on the output power and drain
efficiency and to determine the Smith chart region where the optimal position of
third-harmonic reflection coefficient may be. The measurement required a further
6 × 4 reflection coefficient grid sweep to find the precise location of Γopt (F3 ) for
the optimal output power and drain efficiency.
The determination of optimal third-harmonic reflection coefficients and, there-
fore, the respective third-harmonic impedances of 1.3 + j 100  and 0.6 + j 60.4 
for optimal output power and drain efficiency, which requires 168 measurements,
take just over 75 minutes for the measurements on an ELP system integrated into
a two-channel waveform measurement setup [39]. This type of measurement and
investigation is not usually possible through traditional passive and active load-pull
systems. Furthermore, an open-loop harmonic active load-pull system, irrespective
of its inability to synthesize independent harmonic reflection coefficients, would
require 10 times more measurement time on a similar setup.
The multi-harmonic ELP system can also be used to conduct load-pull measure-
ments to discover the continuous high efficiency and output power from a device,
through the control of fundamental and harmonic reflection coefficients in a pre-
cisely controlled manner [10, 12]. These measurements are essential in the design
of PAs, such as class-J and class-J* [40].
One such investigation, illustrated in Fig. 6.23, requires measurements of 15 spe-
cific fundamental and second-harmonic reflection coefficients at the package plane
of a 2 W gallium nitride (GaN) device. This measurement needs the fundamental
and second-harmonic reflection coefficients to be directly related, with the reactive
components of fundamental and second-harmonic impedances being in a ratio of
1:2 for fixed real components of the fundamental and harmonic impedances, in or-
der to obtain constant output power and drain efficiency from the device, as shown
in Fig. 6.24.
The results in Fig. 6.24 show the achieved contours of constant efficiency and
output power, with less than 1 % deviation in a drain efficiency around 70 % and
only 0.1 dB deviation in the 34.5 dBm output power over the selected range of
fundamental and harmonic reflection coefficient variation. This type of investigation
to achieve a highly efficient linear mode PA, using experimental device probing and
design, can only be carried out by employing an ELP system as other load-pull
systems are either incapable or too slow for this purpose.
160 6 Envelope Load-Pull System

Fig. 6.23 Depiction of

robust harmonic load
reflection coefficient control
by employing the developed
harmonic envelope load-pull
system [8], © IEEE 2010

Fig. 6.24 Demonstration of

constant power output and
efficiency with respect to
changing reactance through
the use of an envelope
load-pull system [8], © IEEE
2010

References
1. Focus Microwaves, Product catalogue, http://www.focus-microwaves.com/template.php?
unique=4. Accessed Oct. 2010
2. Maury Microwave Corporation, Product catalogue, http://www.maurymw.com/mmc_catalog/
mmccatalog06.htm. Accessed Oct. 2010
3. Mesuro, Product catalogue, http://www.mesuro.com/products.htm. Accessed Oct. 2010
4. Real Time Passive Source/Load Pull Systems, PAF product, http://www.se-group.com/doc/
paf.pdf. Accessed Oct. 2010
5. R. Negra, F.M. Ghannouchi, W. Bachtold, Study and design optimization of multiharmonic
transmission-line load networks for class-E and class-F K-band mmic power amplifiers. IEEE
Trans. Microw. Theory Tech. 55(6), 1390–1397 (2007)
References 161

6. C. Roff, J. Benedikt, P.J. Tasker, Design approach for realization of very high efficiency power
amplifiers, in IEEE MTT-S International Microwave Symposium Digest (June 2007), pp. 143–
146
7. M.S. Hashmi, A.L. Clarke, S.P. Woodington, J. Lees, J. Benedikt, P.J. Tasker, Electronic multi-
harmonic load-pull system for experimentally driven power amplifier design optimization,
in IEEE MTT-S International Microwave Symposium, Boston, USA, vols. 1–3 (June 2009),
pp. 1549–1552
8. M.S. Hashmi, A.L. Clarke, S.P. Woodington, J. Lees, J. Benedikt, P.J. Tasker, An accurate
calibrate-able multi-harmonic active load-pull system based on the envelope load-pull concept.
IEEE Trans. Microw. Theory Tech. 58(3), 656–664 (2010)
9. M.S. Hashmi, A.L. Clarke, J. Lees, M. Helaoui, P.J. Tasker, F.M. Ghannouchi, Agile harmonic
envelope load-pull system enabling reliable and rapid device characterization. IOP Meas. Sci.
Technol. 21(055109), 1–9 (2010)
10. S.C. Cripps, P.J. Tasker, A.L. Clarke, J. Lees, J. Benedikt, On the continuity of high efficiency
modes in linear RF power amplifiers. IEEE Microw. Wirel. Compon. Lett. 19(10), 665–667
(2009)
11. P. Wright, J. Lees, J. Benedikt, P.J. Tasker, S.C. Cripps, A methodology for realizing high
efficiency class-J in a linear and broadband PA. IEEE Trans. Microw. Theory Tech. 57(12),
3196–3204 (2009)
12. F.M. Ghannouchi, M.S. Hashmi, Experimental investigation of the uncontrolled higher har-
monic impedances effect on the performance of high-power microwave devices. Microw. Opt.
Technol. Lett. 52(11), 2480–2482 (2010)
13. H.M. Nemati, A.L. Clarke, S.C. Cripps, J. Benedikt, P.J. Tasker, C. Fager, J. Grahn, H. Zirath,
Evaluation of a GaN HEMT transistor for load- and supply-modulation applications using in-
trinsic waveform measurements, in IEEE MTT-S International Microwave Symposium Digest
(May 2010), pp. 509–512
14. A.L. Clarke, M. Akmal, J. Lees, J. Benedikt, P.J. Tasker, Investigation and analysis into device
optimization for attaining efficiencies in-excess of 90 % when accounting for higher harmon-
ics, in IEEE MTT-S International Microwave Symposium Digest (May 2010), pp. 1114–1117
15. R. Hajji, F.M. Ghannouchi, R.G. Bosisio, Large-signal microwave transistor modeling using
multiharmonic load-pull measurements. Microw. Opt. Technol. Lett. 5(11), 580–585 (1992)
16. H. Qi, J. Benedikt, P.J. Tasker, Nonlinear data utilization: from direct data lookup to behavioral
modeling. IEEE Trans. Microw. Theory Tech. 57(6), (2009)
17. P. Berini, F.M. Ghannouchi, R.G. Bosisio, Active load characterization of a microwave tran-
sistor for oscillator design, in IEEE Instrumentation & Measurements Technology Conference,
Ottawa, Canada (May 1997), pp. 668–673
18. F.M. Ghannouchi, F. Beauregard, A.B. Kouki, Power added efficiency and gain improvement
in MESFETs amplifiers using an active harmonic loading technique. Microw. Opt. Technol.
Lett. 7(13), 625–627 (1994)
19. G. Zhao, S. El-Rabaie, F.M. Ghannouchi, The effects of biasing and harmonic loading on
MESFET tripler performance. Microw. Opt. Technol. Lett. 9(4), 189–194 (1995)
20. E. Bergeault, O. Gibrat, S. Bensmida, B. Huyart, Multiharmonic source-pull/load-pull active
setup based on six-port reflectometers: influence of the second harmonic source impedance on
RF performances of power transistors. IEEE Trans. Microw. Theory Tech. 52(4), 1118–1124
(2004)
21. Y.Y. Woo, Y. Yang, B. Kim, Analysis and experiments for high efficiency class-F and inverse
class-F power amplifiers. IEEE Trans. Microw. Theory Tech. 54(5), 1969–1974 (2006)
22. J. Benedikt, R. Gaddi, P.J. Tasker, M. Goss, M. Zadeh, High power time domain measurement
system with active harmonic load-pull for high efficiency base station amplifier design, in
IEEE/MTT-S International Microwave Symposium, Boston, USA (June 2000), pp. 1459–1462
23. T. Williams, J. Benedikt, P. Tasker, Experimental evaluation of an active envelope load-pull
architecture for high speed device characterization, in IEEE/MTT-S International Microwave
Symposium, vol. 3, Long Beach, USA (June 2005), pp. 1509–1512
162 6 Envelope Load-Pull System

24. T. Williams, J. Benedikt, P.J. Tasker, Application of a novel active envelope load-pull archi-
tecture in large signal device characterization, in IEEE 35th European Microwave Conference
Digest, Paris, France (Oct. 2005), 4 pp.
25. S.J. Hashim, M.S. Hashmi, T. Williams, S. Woodington, J. Benedikt, P.J. Tasker, Active en-
velope load-pull for wideband multi-tone stimulus incorporating delay compensation, in 38th
IEEE European Microwave Conference, Amsterdam, The Netherlands (Oct. 2008), pp. 317–
320
26. D.M. Pozar, Microwave Engineering, 3rd edn. (Wiley, New York, 2005). ISBN 0-471-17096-
8
27. R. Ludwig, P. Bretchko, RF Circuit Design: Theory and Applications, 2nd edn. (Pearson,
Upper Saddle River, 2008). ISBN 13-978-013-13555-07
28. E. Kreyszig, Advanced Engineering Mathematics, 8th edn. (Wiley, New York, 2000). ISBN
13-9780471488859
29. M. Sipila, K. Lehtinen, V. Porra, High-frequency periodic time-domain waveform measure-
ment system. IEEE Trans. Microw. Theory Tech. 36(10), 1397–1405 (1988)
30. U. Lott, Measurement of magnitude and phase of harmonics generated in nonlinear microwave
two-ports. IEEE Trans. Microw. Theory Tech. 37(10), 1506–1511 (1989)
31. F. van Raay, G. Kompa, A new on-wafer large-signal waveform measurement system with 40
GHz harmonic bandwidth, in IEEE MTT-S International Microwave Symposium Digest, vol. 3
(June 1992), pp. 1435–1438
32. D. Barataud, C. Arnaud, B. Thibaud, M. Campovecchio, J.-M. Nebus, J.P. Villotte, Measure-
ments of time-domain voltage/current waveforms at RF and microwave frequencies based on
the use of a vector network analyzer for the characterization of nonlinear devices-application
to high-efficiency power amplifiers and frequency-multipliers optimization. IEEE Trans. In-
strum. Meas. 47(5), 1259–1264 (1998)
33. D.J. Williams, P.J. Tasker, An automated active source and load-pull measurement system, in
6th IEEE High Frequency Student Colloquium Digest (Sep. 2001), pp. 7–12
34. S. Bensmida, P. Poire, R. Negra, F.M. Ghannouchi, G. Brassard, New time-domain voltage and
current waveform measurement setup for power amplifier characterization and optimization.
IEEE Trans. Microw. Theory Tech. 56(1), 224–231 (2008)
35. W.S. El-Deeb, M.S. Hashmi, S. Bensmida, N. Boulejfen, F.M. Ghannouchi, Thru-less cali-
bration algorithm and measurement system for on-wafer large-signal characterization of mi-
crowave devices, in IET Microwaves, Antenna and Propagation (2010)
36. C. Roff, J. Graham, J. Sirois, B. Noori, A new technique for decreasing the characterization
time of passive load-pull tuners to maximize measurement throughput, in 72nd Automatic
Radio Frequency Techniques Group (ARFTG) Conference Digest, Portland, USA (Dec. 2008),
pp. 92–96
37. K. Ogata, Modern Control Engineering, 5th edn. (Prentice Hall, New York, 2009). ISBN 13-
978-013-6156734
38. A. Ferrero, V. Teppati, G.L. Madonna, U. Pisani, Overview of modern load-pull and other
non-linear measurement systems, in Automatic RF Techniques Group Conference (ARFTG)
Digest (June 2001), pp. 1–5
39. M.S. Hashmi, Analysis, realization and evaluation of envelope load-pull system both for CW
and multi-tone applications, PhD Thesis, Cardiff University, 2009
40. J.D. Rhodes, Output universality in maximum efficiency linear power amplifiers. Int. J. Circuit
Theory Appl. 31, 385–405 (2003)
Chapter 7
Waveform Measurement and Engineering

In this chapter, techniques to develop high-frequency nonlinear time-domain wave-

form measurement systems are discussed along with critical issues related to these
systems. The description entails two aspects, namely the development of a test set
and the subsequent calibration mechanism for determining the absolute values of
magnitudes and phases of the individual harmonic components when the transis-
tor devices are operated in the nonlinear region. Finally, the waveform engineering
technique along with some of its applications in the context of power amplifiers is
discussed.

7.1 Introduction

The study of nonlinear semiconductor devices for obtaining the best insight into the
device behavior often requires detailed consideration of the time-varying current
and voltage waveforms on the terminals of such devices. For example, the operation
mode of radio frequency power amplifiers (RFPAs) can be effectively identified
by looking at the time-domain current and voltage (I-V) waveforms at the output
terminal of transistor devices. Figure 7.1 depicts such a scenario for a current-mode
class-D amplifier.
When active microwave devices are subjected to periodic nonlinear operation,
there is high harmonic content in the I-V waveforms. If these waveforms are mea-
sured in such a manner that they contain all the relevant associated harmonics, valu-
able information would be obtained for the understanding and modeling of these
devices. For example, if the terminal voltages and currents waveforms of an RFPA
are measured, it can be optimized for optimal design parameters, such as output
power, direct current (DC) to RF (radio frequency) power conversion efficiency,
gain [1–4]. In addition, these terminal waveforms also enable appropriate nonlinear
device models for RFPA applications [5].
As a result, it is suggested that, for an RFPA, the terminal I-V waveforms are the
unifying theoretical link between transistor technology, circuit design and system

F.M. Ghannouchi, M.S. Hashmi, Load-Pull Techniques with Applications to Power 163
Amplifier Design, Springer Series in Advanced Microelectronics 32
DOI 10.1007/978-94-007-4461-5_7, © Springer Science+Business Media Dordrecht 2013
164 7 Waveform Measurement and Engineering

Fig. 7.1 Current mode

class-D amplifier schematic
and the ideal respective
current and voltage
waveforms across each
transistor

Fig. 7.2 Depiction of

traveling waves at a nonlinear
two-port DUT

performance [6, 7]. The investigation of such a hypothesis requires an appropriate

measurement system that can measure I-V waveforms at high frequencies. It should
be understood that the time-domain waveforms at low frequencies can be easily
measured using any standard oscilloscope. At RF and microwave frequencies, how-
ever, the scenario is completely different.
The difficulty in measuring high-frequency I-V waveforms stems not just from
the inability of the standard scopes to measure all the harmonic components due
to the limitation of the bandwidth of the scopes: the problem mainly occurs from
the generic technological limitations that are very common at high frequencies. The
standard probes at microwave frequencies generate reflections, due to the compara-
ble size of probe and the wavelength of the waves being measured, which have the
potential to damage the device under test (DUT) and the measurement system due
to high VSWR that can occur in the transmission line of the system resulting from
the reflections.
High-speed sampling oscilloscopes can fulfill the bandwidth requirements; how-
ever, in the microwave region, there are systematic errors that are introduced by
the frequency response of the measurement system, mismatches, multiple reflec-
tions and other factors. As a consequence, appropriate calibration algorithms are
employed to correct for these errors [8–10] at fundamental frequency and the rele-
vant harmonics present in the I-V waveforms.

7.2 Theoretical Formulation

The periodic current and voltage waveforms at the terminals of the DUT consist of
(n) (n)
fundamental and harmonic waves ak and bk , as depicted in Fig. 7.2, with k and
n representing the number of ports and harmonics, respectively.
7.3 Historical Perspectives 165

The normalized traveling waves are defined by Eqs. (7.1) and (7.2):
(n) (n)
(n) Vk +I
ak = √ k (7.1)
2 2Z0
(n) Vk(n) − Ik(n)
bk = √ (7.2)
2 2Z0
(n)
where Z0 is the reference impedance (usually 50  for all harmonics); and, Vk and
(n)
Ik are the nth Fourier coefficients that need to be measured, which are given in
Eqs. (7.3) and (7.4), for the construction of the voltage and current waveforms at the
DUT’s terminals.
(n)
  (n) (n) 
Vk = 2Z0 ak + bk (7.3)
(n)
  (n) (n) 
Ik = 2/Z0 ak − bk (7.4)
Once these Fourier coefficients are known, the time-domain voltage and current
waveforms at the DUT’s terminals can be obtained by employing Eqs. (7.5) and
(7.6), respectively.

(0)

Nh
(n)  (n) 
vk (t) = Vk + Vk cos 2πnf t − ϕk (7.5)
n=1
(0) (n)
where Vk is the direct current component at kth port, Vk is the magnitude of
the nth harmonic at kth port, Nh is the number of harmonics, f is the fundamental
(n)
frequency, and ϕκ is the phase of the nth harmonic.

(0)

Nh
(n)  (n) 
ik (t) = Ik + Ik cos 2πnf t − θk (7.6)
n=1
(0) (n)
where Ik the direct current component at kth port is, Ik is the magnitude of the
nth harmonic at kth port, Nh is the number of harmonics, f is the fundamental
(n)
frequency, and θκ is the phase of the nth harmonic.
It is evident from Eqs. (7.5) and (7.6) that the construction of time-domain cur-
rent and voltage waveforms at the DUT’s terminals requires absolute values of the
magnitude and phase of each harmonic component. Therefore, the most appropriate
time-domain waveform measurement system is the one that can accurately provide
the phase and magnitude information at the harmonic component of interest.

7.3 Historical Perspectives

At low frequencies, the terminal voltage can be directly measured using a high-
impedance probe, and the current through a node can be measured using a low-
impedance probe in series. At high frequencies, the high and low impedances of the
166 7 Waveform Measurement and Engineering

probes can only be maintained across a narrow bandwidth, and conditionally sta-
ble circuits can become unstable during the measurement. In order to address this
problem, non-intrusive measurement techniques are employed in which the incident
and reflected energy of traveling voltage waveforms is measured [11–13]. This al-
lows a constant measurement impedance environment (Z0 = 50 ) across a broad
frequency bandwidth. These traveling waves are measured in terms of scattering pa-
rameters (S-parameters) [14] using a vector network analyzer (VNA) [15], which
can then be converted to impedance and admittance using conversion matrices. The
impedance and admittance matrices relate the sum of the voltage and current to the
measurement ports.
Despite the valuable information gained from S-parameter data, the measurement
technique can only be applied when the superposition principle holds true [16]. This
principle prevents energy being transferred from the stimulus frequency to other
harmonic frequencies. Therefore, the measurements are performed by exciting the
DUT with a number of sine waves, one frequency at a time, in order to detect the
response of the DUT at each stimulus frequency. This produces limitations when
dealing with nonlinear networks containing active devices operating at high power
levels, where the superposition principle cannot be applied, as nonlinear behavior
produces additional harmonic frequency and intermodulation components.
These nonlinear effects have traditionally been measured in the frequency do-
main using spectrum analyzers (SAs) to estimate and assess the device performance
at the fundamental and harmonic frequencies. The SA is a scalar instrument and can
measure a broad spectrum in real time, while achieving very high dynamic range. In
cases where modulated signals are used, an SA can measure the magnitudes of the
fundamental and harmonic frequencies and the intermodulation components. How-
ever, the main limitation of such instruments is the inability to measure the phase in-
formation, thus limiting their usefulness in nonlinear device modeling applications.
Nonetheless, the real-time broadband behavior of the SA provides the capability for
detecting and measuring spurious signals that result from device instabilities and
oscillations. All these measurements can be carried out without calibration or prior
knowledge of the circuit behavior, making the SA an ideal instrument for evaluating
the performance of circuits and devices operating in a nonlinear mode of operation.
The nonlinear behavior of a device or circuit can be obtained from the direct
measurement of the voltage and current waveforms in the time domain using a con-
ventional oscilloscope. The Fourier transformation of the measured time-domain
waveforms gives both the magnitude and phase information for the fundamental
frequency and each harmonic component. The absolute phase information gives ad-
ditional information for estimating the DUT’s behavior.
At high frequencies, limitations are introduced by the restricted sampling rates of
the analog-to-digital converter (ADC) employed in the oscilloscope. This results in
a reduced bit resolution at high frequencies, which reduces the dynamic range of the
oscilloscope. Additionally, the measurement system must be fully error corrected,
in terms of magnitude and phase, in order to achieve the error coefficients at each
harmonic of interest before reconstruction of the I-V waveforms can be achieved at
the DUT’s terminals.
7.3 Historical Perspectives 167

Fig. 7.3 Waveform measurement system developed using a two-channel sampling scope [17],
© IEEE 1988

The first nonlinear I-V waveform measurement system that was capable of pro-
viding the absolute values of the magnitude and phase of each harmonic content of
the waveform was reported in 1988 [17]. The measurement system, as depicted in
Fig. 7.3, uses a two-channel high-frequency sampling oscilloscope and collects two
of the four traveling waves: the reflected input wave, b1 , and the transmitted output
wave, b2 . Using knowledge of the S-parameters of the linear input and output cou-
pling networks at the fundamental and harmonic frequencies, the terminal current
and voltage (I-V) waveforms are determined.
In this system, the calibration is performed by measuring the S-parameters of the
whole system, in order to remove the error introduced by losses, mismatches and
imperfect directivities in the system. The measured traveling waves are converted
into the frequency domain for vector error correction, and the inverse transform
is then performed to retrieve the corrected traveling waves. This approach is seen
as important in the development of large-signal measurement systems, as it allows
for the adoption of calibration procedures akin to those developed for the more
established VNA technology.
The disadvantages of this system include the requirement to measure the S-
parameters of the input and output coupling networks, leading to an increased like-
lihood of error in measurements. The system also suffers from noise introduced
by trigger jitter, reducing the dynamic range compared to the mixer based approach
adopted in VNAs. These errors become quite large at frequencies above 5 GHz, lim-
iting the application of this system to frequencies of less than 2 GHz when accurate
information up to the third harmonic component is required.
An alternate system, as depicted in Fig. 7.4, that utilizes a modified VNA archi-
tecture for the capture of time-domain I-V waveforms was reported in 1989 [18].
The most important component in the system is the calibrated VNA that measures
the magnitude and phase of a constant wave signal at each frequency of interest.
This system utilizes a high-frequency Schottky diode as a reference for the pur-
pose of determining the absolute phase of nth harmonic component with respect to
the fundamental signal. The Schottky diode produces a calibration reference for the
168 7 Waveform Measurement and Engineering

Fig. 7.4 VNA based time-domain waveform measurement system presented in [18], © IEEE 1989

absolute magnitude and phase required for reconstructing the output waveforms in
the time domain.
This measurement technique offers greater dynamic range compared to that
achieved using an oscilloscope, due to the use of a VNA. This system, however,
is limited in its application, as it allows construction of time-domain waveforms
only at the output of the DUT.
A significantly improved system based on this concept was reported in 1998
[19]. It enables the measurement of time-domain I-V waveforms at both the input
and output ports of the DUT. The system employs a step recovery diode (SRD) for
producing a fixed reference signal. In this system, the amplitude calibration can be
achieved by attaching known calibration standards to the measurement ports; and,
the SRD facilitates phase calibration, as it produces a signal with very well behaved
and known phase relationship at the fundamental and harmonic frequencies.
This system, however, is inherently slow, due to presence of a VNA, considering
that a frequency sweep has to be carried out for each harmonic to be measured.
In addition, accurate phase coherence is difficult to achieve for fundamental and
harmonic frequencies. However, it is worth mentioning that the concept employed in
this system has led to the development of more advanced VNAs and is commercially
available in the form of PNA-X [20].
The microwave transition analyzer (MTA) operates as a wideband, time-domain
sampling oscilloscope in its main mode of operation and has the capability to con-
duct narrowband swept frequency measurements and acts as a VNA to measure S-
parameters with a very high dynamic range. It is advantageous over other sampling
oscilloscopes for two main reasons. Firstly, the sample-and-hold trigger employed
in conventional sampling oscilloscopes is replaced with a mixing process in the
MTA, which uses a local oscillator (LO) signal that is phase locked to one of the
input signals.
The MTA is, therefore, immune to the errors introduced by trigger jitter, and
time-base error is defined only by the quality of the LO signal. Thus, measurement
using an MTA is significantly better than that of the sampling oscilloscopes. Several
7.4 Practical Waveform Measurement System 169

Fig. 7.5 A typical time-domain waveform measurement system built around a microwave transi-
tion analyzer (MTA)

I-V waveform measurement systems with an MTA as the main instrument have been
reported [21–25] in the past two decades. One such system, as depicted in Fig. 7.5,
utilizes the MTA as a multi-harmonic receiver (MHR).
As can be seen, a multiplexing network converts the two-channel MTA into a
four-channel receiver, allowing measurement of all the four traveling waves in one
shot. Recent advances employ more advanced MTA type receivers, such as a non-
linear vector network analyzer (NLVNA) [26] that consists of four channels and
reduces the complexity of the measurement setup.
An alternative approach for determining time-domain I-V waveforms at the
DUT’s ports is the six-port based waveform correlator [27]. In such a system,
the concept of frequency-domain reflectometry is employed; therefore, such setups
when appropriately configured can also be useful for linear measurements [28].

7.4 Practical Waveform Measurement System

A typical waveform measurement setup is depicted in Fig. 7.6. The system consists
of an MTA used as a complex MHR from DC to 40 GHz [29], an RF generator,
four RF two-way (single pole, double throw, i.e., SPDT) switches, a control circuit
for the switches, a power supply, one power divider, and two directional couplers.
The entire system is controlled by a personal computer (PC) via general purpose
interface bus (GPIB).
170 7 Waveform Measurement and Engineering

Fig. 7.6 Block diagram of a typical MTA based measurement system

A harmonic receiver operates in a similar way to a conventional super heterodyne

receiver; the main difference is in the mixing process. In a harmonic receiver, the
LO signal is first passed through a comb generator prior to being applied to the
mixer. This effectively generates multiple harmonics of the LO signal, meaning that
multiple sections of the RF spectrum can be captured simultaneously. The MTA
works on a similar principle. In this setup, the power divider splits the signal from
the generator into two parts. The first part excites the DUT’s ports, depending on the
position of switch 1 (SW1); and, the second part is fed to channel 2 (CH2) of the
MTA for a reference signal. Channel 1 (CH1) of the MTA serves as the measuring
port in this setup, although there have been measurement setups based around an
MTA in which the channels have been used differently [21–23]. This particular setup
provides measurements as ratios between CH1, as the receiver of the incident and
reflected waves to be measured, and CH2, as a reference channel.
To fully exploit the capability of any high-frequency I-V waveform measurement
system, full vector error correction needs to be carried out, in order to remove the
systematic errors introduced by mismatches, losses, and imperfect components.

7.5 System Calibration

For the accurate large-signal waveforms measurements, the error terms of the error
flow model between the MHR measuring plane and the DUT’s terminals, described
for the nth harmonic frequency and depicted in Fig. 7.7, need to be fully determined.
This is a simplified version of the error model: more extensive details can be found
in [23].
As a general rule, if there is no error introduced by the system imperfections
and mismatches, the traveling waves at the DUT’s ports are exactly the same as
7.5 System Calibration 171

Fig. 7.7 Error flow graph depicting the error terms between the DUT and MHR ports

measured at the MHR port. In such an ideal scenario, the I-V waveforms can be di-
rectly extracted by employing Eqs. (7.3) and (7.4). However, in practical situations,
the error terms between the DUT and MHR’s ports change this direct relationship.
For example, the traveling waves at the first port of DUT and MHR ports are related
by Eqs. (7.7) and (7.8).

(n)  (n) (n) (n) (n)  (n) (n) (n)  (n) 
a1dut = e01 e10 − e00 e11 a1m + e11 b1m e01 (7.7)
(n)  
(n) (n) (n)  (n) 
b1dut = −e00 a1m + b1m e01 (7.8)

It is apparent from the above equations that the extraction of terminal I-V wave-
forms requires determination of individual error terms separately, thereby necessi-
tating a different calibration process from the standard VNA calibration. In essence,
the calibration approach for the waveform extraction is a three-step process where
one of the steps is conventional VNA calibration. The three steps are discussed in
the following three subsections.

7.5.1 First Step: Power Flow Calibration

The purpose of this step is to relate the power at the DUT port to the power sampled
at the MHR port through a power scaling factor, α. This factor takes into account
the losses incurred in the cabling and all other peripheral components employed
in the measurement setup. The correct scaling factor changes the Fourier coeffi-
cient Eqs. (7.3) and (7.4) to a new set of equations, i.e., Eqs. (7.9) and (7.10), and
consequently helps in determining the exact magnitude of the voltage and current
waveforms.

(n)
 (n) (n)  (n)
V1 = 2Z0 a1dut + b1dut = αV1 (7.9)
  (n) (n) 
I1(n) = 2/Z0 a1dut − b1dut = αI1(n) (7.10)
172 7 Waveform Measurement and Engineering

Fig. 7.8 Error flow graph

depicting the error terms
between the DUT and MHR
ports

Figure 7.8 depicts the power flow at the MHR plane, Pm , and the DUT plane,
Pdut . They are related to the error parameters, the reflection coefficient at coaxial
plane, Γdut , and the power scaling factor, α, as follows [30]:
1  (n) 2 αPm
Pkdut = akdut  = (n) (n)
(7.11)
2 |1 + (−e11 )Γkdut |2
The power flow calibration factor, α, is obtained by connecting a power meter to the
DUT port [31]. The power measured by the power meter, PPWM , is related by:

1  (n) 2   (n) 2  (n)

αPm (1 − |Γkdut |2 )
PPWM = akdut  1 − Γkdut  = (n) (n)
(7.12)
2 |1 + (−e11 )Γkdut |2
Simplification of Eqs. (7.11) and (7.12) gives the power calibration factor (α).
(n) (n)
PPWM |1 + (−e11 )Γkdut |2
α= (n)
(7.13)
Pm (1 − |Γkdut |2 )

7.5.2 Second Step: S-Parameter Calibration

The second step is the conventional VNA type calibration for the determination
(n) (n) (n) (n) (n) (n) (n) (n)
of error terms e00 , e11 and e01 e10 for first port and e22 , e33 and e23 e32 for the
second port, as shown in Fig. 7.7.
For the first port, Fig. 7.7 can be simplified to derive an expression for the mea-
(n)
sured reflection coefficient at the MHR port, Γ1m , in terms of error parameters and
(n)
the reflection coefficient at the first port of DUT, Γ1dut .
(n) (n) (n) (n) (n) (n)
(n) (n) e01 e10 Γ1dut − e1 Γ1dut + e00
Γ1m = e00 + (n) (n)
= (n) (n)
(7.14)
1 − e11 Γ1dut −e11 Γ1dut + 1
where
(n)  (n) (n) (n) (n) 
e1 = e00 e11 − e01 e10 . (7.15)
(n) (n) (n) (n)
For determining the three unknowns e00 , e11 and e01 e10 in Eq. (7.14), the sim-
plest method is to measure the open-short-load (OSL) standards [32] at the first port
7.5 System Calibration 173

Fig. 7.9 Error flow graph depicting the error terms between the DUT and MHR ports

of the DUT. These measurements provide three sets of linear equations, which are
summarized in matrix form in Eq. (7.16).
⎡ (n) ⎤  
e00 1 (Γ1dut )O (Γm1 )O −(Γ1dut )O −1 (Γm1 )O
⎣ e (n) ⎦ = 1 (Γ1dut )S (Γm1 )S −(Γ1dut )S × (Γm1 )S (7.16)
11
e
(n) 1 (Γ ) (Γ
1dut L m1 L) −(Γ )
1dut L (Γ )
m1 L
1
where the terms with subscript O, S and L are the quantities for the calibration
standards open, short and load, respectively.
(n) (n) (n) (n)
Equation (7.16) once solved gives the error terms e00 , e11 and e01 e10 between
(n) (n)
port 1 of the DUT and the MHR, while the determination of error terms e22 , e33
(n) (n)
and e23 e32 between port 2 and the MHR requires a thru standard between port 1
and port 2, as shown in Fig. 7.9. The simplification of error flow model in Fig. 7.9
(n) (n) (n) (n)
gives expressions for error terms e22 , e33 and e23 e32 as a function of the measured
S-parameters of the thru standard given in Eqs. (7.17)–(7.20).
(n) (n)
(n) S11T − e00
e33 = (n) (n) (n) (n)
(7.17)
t11 + e11 (S11T − e00 )
(n) (n)
(n) (n) t22 e11
e22 = S22T − (n) (n)
(7.18)
1 − e11 e33
(n) (n) (n) (n) (n) (n)
t11 = e01 e10 = e00 e11 − e1 (7.19)
(n) (n)
(n) (n) (n) (n) (n) (1 − e11 e33 )2
t22 = e32 e23 = S12T S21T (n)
(7.20)
t11
(n) (n) (n) (n)
where S11T , S12T , S21T and S22T are the measured S-parameters of the thru stan-
(n) (n)
dard connected between ports 1 and 2, and t11 and t22 are the reflection tracking
parameters representing ports 1 and 2, respectively.
(n) (n) (n) (n)
Equations (7.16)–(7.20) allows the calculation of error terms e00 , e11 , e01 e10 ,
(n) (n) (n) (n)
e22 , e33 and e23 e32 of the error model between the MHR and DUT ports and
is, therefore, referred as the S-parameter calibration stage. However, this step does
(n) (n) (n) (n)
not allow for the determination of individual terms e01 , e10 , e23 , and e32 and,
174 7 Waveform Measurement and Engineering

Fig. 7.10 Error flow graph of

the setup with coaxial cable
between the MHR and port 1
of the DUT

consequently, does not provide enough information for accurate determination of

the traveling waves at the DUT’s ports, e.g., the traveling waves at the first port of
DUT given by Eqs. (7.7) and (7.8), and hence for the extraction of terminal I-V
waveforms.

7.5.3 Third Step: Enhanced Calibration

This calibration step is necessary to separate the individual terms from the tracking
(n) (n) (n) (n)
error parameters, e01 e10 and e23 e32 . It is an enhancement of the second step and,
as a result, is called enhanced calibration.
In this step, with the thru standard connected between ports 1 and 2, as shown
(n) (n)
in Fig. 7.9, the error flow model when solved gives two possible values of e10 e32
through following expressions [33]:
(n)  (n) (n) 2 (n)  (n) (n)  (n) (n) 
S21T e10 e32 − S12T e01 e10 e32 e23 = 0 (7.21)

 (n) (n) (n) (n) (n)
 S (e e )(e e )
e10 e32 = ± 12T 01 10
(n) (n) 32 23
(n)
(7.22)
S21T
The appropriate solution for the transmission tracking between the first and sec-
(n) (n)
ond DUT ports, e10 e32 , is dependent on the length of the thru standard and it is
given by [33]:
 −γ l (n) 
(e )
Re (n)
>0 (7.23)
S21T
(n)
where γ is the propagation constant, l is the length of the thru standard and S21T is
the measured S-parameter of the thru standard connected between first and second
ports of the DUT.
Next, the thru standard is replaced by a coaxial cable and its S-parameters,
(n) (n) (n) (n)
S11coax , S12coax , S21coax and S22coax , are measured. Subsequently, the end of coax-
ial cable connecting the second port of DUT is disconnected and connected to first
channel of the MHR, the error flow model of which is shown in Fig. 7.10, in order
(n)
to measure the traveling wave, bcoax .
Simplification of the error flow model in Fig. 7.10 provides an expression for the
(n)
error term e10 given in Eq. (7.24) [25]
7.5 System Calibration 175

 (n)  (n) (n) (n) (n) 

(n) bcoax (1 − e11 S11coax )(1 − ΓMHR S22coax )
e10 = (n) (n)
a1m S21coax
 (n)  (n) (n) (n) (n) 
bcoax (e11 ΓMHR S21coax S12coax )
− (n) (n)
(7.24)
a1m S21coax
(n)
The determination of the absolute value of individual term e10 enables the calcu-
(n) (n) (n)
lation of error parameters e01 , e23 and e32 from Eqs. (7.25), (7.26), (7.27), which
are directly deduced from Eqs. (7.19), (7.20) and (7.22).
(n) (n) (n)
(n) e01 e10 t11
e01 = (n)
= (n)
(7.25)
e10 e10
(n) (n)
(n) e10 e32
e32 = (n)
(7.26)
e10
(n) (n) (n)
(n) e23 e32 t22
e23 = (n)
= (n)
(7.27)
e32 e32
The determination of the absolute values of error parameters enables accurate
measurements of incident and reflected traveling waves, given by Eqs. (7.28)–
(7.31), at the DUT’s ports.
 (n) (n) (n) (n)   (n) 
(n) e01 e10 − e00 e11 (n) e (n)
a1dut = (n)
a1m + 11 (n)
b1m (7.28)
e01 e01
 (n)   
(n) −e00 (n) 1 (n)
b1dut = (n)
a 1m + (n)
b1m (7.29)
e01 e01
 (n) (n) (n) (n)   (n) 
(n) e e −e e (n) e (n)
a2dut = 32 23 (n) 33 22 a2m + 33 (n)
b2m (7.30)
e23 e23
 (n)   
(n) −e33 (n) 1 (n)
b2dut = (n)
a2m + (n) b2m (7.31)
e23 e23
The measured incident and traveling waves along with the power calibration fac-
tor, α, obtained from power calibration in the first step gives complete information
for the extraction of time-domain current and voltage waveforms at the DUT’s ports
given by Eqs. (7.9) and (7.10).

7.5.4 Calibration Evaluation

The fully calibrated system, when used to measure terminal voltage waveform at the
output port of a Mini-Circuits ZHL-42 W power amplifier biased at 17 V and excited
176 7 Waveform Measurement and Engineering

Fig. 7.11 Output voltage

waveforms of the ZHL-42 W
power amplifier operating in
linear mode

Fig. 7.12 Output voltage

waveforms of the ZHL-42 W
power amplifier operating in
nonlinear mode

by a 0.5 GHz signal, gives the results shown in Fig. 7.11 during linear operation,
while considering up to five harmonics.
It is also evident that the voltage waveforms measured using the system and a
standard commercial scope, i.e., 4 Gbps digital oscilloscope (Tektronix TDS 794D)
[34], show good agreement. However, the output voltage waveform is distinct even
when the amplifier is pushed into slightly nonlinear region, as shown in Fig. 7.12.
The difference is due to the fact that the system captured all the spectral components
of the voltage waveform, whereas the commercial scope could capture spectral lines
only up to four harmonics for the chosen excitation of 500 MHz, due to limited
bandwidth of the commercial scope.
7.6 Six-Port Based Waveform Measurement System 177

7.6 Six-Port Based Waveform Measurement System

By accurately determining the magnitude of the DC component (V (0) , I (0) ), as well

as the magnitude and phase of the fundamental frequency (V (1) , I (1) ) and harmonic
components (V (n) , I (n) ), the time-domain voltage and current waveforms can be
calculated at the port of the DUT according to Eqs. (7.32) and (7.33).
 m 

v(t) = Re V (n) e(j 2πnf0 t) (7.32)
n=0
 

m
i(t) = Re I (n) (j 2πnf0 t)
e (7.33)
n=0

where n denotes the harmonic order, f0 is the fundamental frequency, and m is the
total number of harmonics included in the measurement.
The DC components, V (0) and I (0) , are routinely measured directly with a bi-
asing instrument; however, the measurement of the complex Fourier coefficients,
V (n) and I (n) , for n ≥ 1 is more difficult. A traditional setup developed using a
multi-harmonic receiver (MHR) has evolved in the form of a large-signal network
analyzer (LSNA) [35, 36], which shows promise.
Alternatively, a deliberately modified six-port (SP) reflectometer, which works as
homodyne VNA, can also determine the complex Fourier coefficients of the waves
in the frequency domain [27]. To be able to determine waveforms v(t) and i(t),
complex voltage traveling waves, also called pseudowaves [37], at least at the fun-
damental frequency (f0 ) and second (2f0 ), and third (3f0 ) harmonic frequencies
should be measured. At a given reference plane, V (n) and I (n) in Eqs. (7.32) and
(7.33) are related to the incident traveling wave, a (n) , and reflected traveling wave,
b(n) , by [37, 38]:

V (n) = a (n) + b(n) (7.34)

a (n) − b(n)
I (n) = (7.35)
Zc
where Zc is the characteristic impedance of the system (typically 50 ).
By measuring the pseudo reflection coefficients, Γ (n) = a (n) /b(n) , at a given
plane using an SP reflectometer, the magnitude and phase of traveling wave a (n)
can be easily determined if traveling wave b(n) is from a reference generator, i.e.,
whose magnitude and phase is known. We propose to measure pseudo reflection
coefficient Γ (n) at a given plane by means of an SP reflectometer. In the proposed
source/load-pull setup, the reflection coefficient measured by the SP reflectometer
is equal to Γ (n) , since the reference impedance used to define the traveling waves
is equal to the characteristic impedance of the system [37]. Subsequently, the volt-
age and current frequency components, V (n) and I (n) , can be deduced according to
Eqs. (7.34) and (7.35).
178 7 Waveform Measurement and Engineering

Fig. 7.13 Schematic of the

used SRD based
multi-harmonic reference
generator [27], © IEEE 2008

Fig. 7.14 Measured

multi-harmonic reference
generator voltage waveforms
for different input levels [27],
© IEEE 2008

7.6.1 Multi-harmonic Reference Generator

The multi-harmonic generator shown in Fig. 7.13 generates the harmonic-rich volt-
age wave, b(n) , with known magnitude and phase components and has been em-
ployed as a reference signal generator for time-domain waveform measurements.
It consists of an SRD excited by a continuous wave RF signal at the fundamental
frequency, f0 , of 4 GHz. A high-gain amplifier is used to enhance the signal level
to the extent that the output of the SRD is a harmonic-rich RF signal with spectral
components at f0 and its higher harmonics. It is evident from the waveforms are
spectrally rich at the output of the SRD, as shown in Fig. 7.14, measured using an
LSNA for input drive levels from −10 dBm to 12 dBm.

7.6.2 SP Reflectometer Principle

The SP reflectometer is an interferometric circuit with six ports. Measurements are

performed by connecting a generator to Port 1, also called the generator port, and
an unknown load to Port 2, denoted as the measurement port. In this configuration,
the RF powers detected at Ports 3–6 are related to the reflection coefficient, Γ ,
presented by the unknown load.
An SP reflectometer can be used to determine both the source and the reflection
coefficients, as illustrated in Fig. 7.15. Referring to Fig. 7.15(a), the SP reflectometer
operates in the forward configuration, i.e., it is driven by an RF source at Port 1 and
7.6 Six-Port Based Waveform Measurement System 179

Fig. 7.15 Typical setup for

reflection coefficient
measurement using an SP
reflectometer in (a) forward
configuration and (b) reverse
configuration [27], © IEEE
2008

measures the source reflection coefficient, ΓS , as given in Eq. (7.36), of the DUT
connected to Port 2
b
ΓS = (7.36)
a
In the reverse configuration, as shown in Fig. 7.15(b), the SP reflectometer mea-
sures the load reflection coefficient, ΓL , as given in Eq. (7.37), at the DUT plane,
presented by the impedance seen through the SP reflectometer
a
ΓL = (7.37)
b
Moreover, the SP reflectometer is able to measure the incident power, POUT , de-
livered to a load connected to the measurement port and is related to the detected
power at Port 3 [39, 40].

7.6.3 Multi-harmonic SP Reflectometer Architecture

Typical broadband SP architectures based on branch-line couplers and a power di-

vider are available for several octaves of bandwidth. However, the detection block
of a conventional SP reflectometer is not able to discriminate between the different
spectral components, because it only detects the total power at the detection ports.
Time-domain waveform measurements, however, require the measurement of a few
complex Fourier coefficients of the signal. To be able to perform this measurement,
the power detection block of the SP reflectometer is modified, as shown in Fig. 7.16,
to become frequency selective and, hence, allow for multi-harmonic measurements.
The power detection block consists of a single pole, four through (SP4T) switch,
a tunable yttrium-iron-garnet (YIG) filter, and a power meter. For harmonic mea-
surements, the YIG filter is adjusted to pass only one harmonic frequency compo-
nent at a time from the output of the SP4T switch to the power meter. Using the
SP4T switch to switch successively from Port 3 to Port 6 and detect the powers, the
180 7 Waveform Measurement and Engineering

Fig. 7.16 Multi-harmonic SP

reflectometer topology [27],
© IEEE 2008

magnitude and phase of reflection coefficient Γ = b/a can be measured through the
procedure described in the following paragraph.
The YIG filter is centered at f0 . Port 3 is selected by means of the S4PT switch,
and the power is measured. Before tuning the YIG filter to 2f0 , the powers at
Port 4, 5 and 6 are measured. In the next measurement cycle, the filter is adjusted
to 2f0 , and the powers at Ports 3, 4, 5 and 6 are measured. Finally, the filter is set
to 3f0 , and the power detection steps are repeated. Before every measurement, the
repeatability of the YIG filter adjustment and the switch contact quality has to be
verified. The multi-harmonic SP reflectometer is calibrated at f0 , 2f0 and 3f0 .
(n) (n)
The relationships between the measured powers at Port 4, 5 and 6, P3 , P4 ,
(n) (n) (n) (n)
P5 , and P6 , and the output power, POUT , and the reflection coefficient, ΓL , are
given by:
 
Pin  1 + Ain ΓS(n) 2

= kin   (7.38)
(n) 
P3(n) 1 + A(n)
3 ΓS
(n) 1
ΓS = (n)
(7.39)
ΓL
P3
(n)
  (n) 2 
1 − ΓS 
(n) (n)
POUT = kP (n) (n)
(7.40)
1 + A3 ΓS
7.6 Six-Port Based Waveform Measurement System 181

Fig. 7.17 Definition of the

calibration principle of the SP
reflectometer for time-domain
voltage and current waveform
measurements [27], © IEEE
2008

Fig. 7.18 Flow graph of the

calibration principle of the SP
reflectometer [27], © IEEE
2008

where n is the harmonic order, constants kin and system related constants Ain (for
(n)
i = 4, 5, 6) are determined by the SP calibration procedure, and constants kP are
determined by an absolute power calibration method using a reference power meter.

7.6.4 Multi-harmonic SP Reflectometer Calibration

Figure 7.17 shows a block diagram of the SP reflectometer calibration principle

for time-domain voltage and current waveform measurements in a source/load-pull
environment. The Fourier coefficients of the harmonic-rich reference voltage wave,
A(n) , which is injected by reference Gref , is known in both magnitude and phase.
Unlike A(n) , the spectral content of B (n) , generated by an unknown generator Gunk ,
does not have to be known. Therefore, Gunk can be any arbitrary multi-harmonic
generator.
In Fig. 7.17, Gunk consists of three active branches that form the load impedance
tuning system [41]. a (n) and b(n) are total traveling waves that travel toward the
SP and Gref , respectively, in the measurement reference plane. a (n) is the sum of
the reference signal, A(n) , for each harmonic generated by Gref and any reflected
(n)
waves. The same relation is valid for B (n) and b(n) . ΓG represents the reflection
(n)
coefficient presented by the source impedance of Gref when A(n) = 0, and Γ6P
represents the reflection coefficient presented by the source impedance of the SP
(n)
reflectometer when Gunk is off, i.e., B (n) = 0. Finally, ΓL is the reflection coeffi-
cient presented by the load impedance at the respective fundamental and harmonic
frequencies, synthesized by Gunk at the measurement plane and measured by the SP
reflectometer.
Since Gref delivers wave A(n) with a known magnitude and phase, the SP reflec-
tometer should be calibrated to determine the magnitude and phase of B (n) for ex-
traction of the time-domain waveform. The error flow model, as shown in Fig. 7.18,
182 7 Waveform Measurement and Engineering

of the SP reflectometer calibration setup can be used to deduce the expression for
B (n) through the following steps.
If the source delivering B (n) is turned off, then:
(n)
a (n) = A(n) + b(n) ΓG (7.41)
(n)
b(n) = a (n) Γ6P (7.42)
Combining Eqs. (7.41) and (7.42) leads to:
A(n)
a (n) = (n) (n)
(7.43)
1 − Γ6P ΓG
In the same way, if the source delivering A(n) is off, then a (n) can be written as:
(n)
ΓG B (n)
a (n)
= (n) (n)
(7.44)
1 − Γ6P ΓG
The superposition of Eqs. (7.43) and (7.44) for the coexistence of A(n) and B (n)
results in:
(n)
A(n) + ΓG B (n)
a (n) = (n) (n)
(7.45)
1 − Γ6P ΓG
Similarly, the expression for b(n) given in (7.46) can also be derived from:
(n) (n)
B (n) + Γ6P A
b(n) = (n) (n)
(7.46)
1 − Γ6P ΓG
(n)
The measured reflection coefficient, ΓL , provided by the synthesized load of Gunk
is obtained from the ratio of Eqs. (7.46) and (7.45):
(n)
(n) b(n) B (n) + Γ6P A(n)
ΓL = = (7.47)
a (n) A(n) + Γ (n) B (n)
G
Rearrangement of Eq. (7.47) gives the expression for unknown quantity B (n) :
(n) (n)
(Γ6P − ΓL )
B (n) = A(n) (n) (n)
(7.48)
(ΓG ΓL − 1)
(n) (n)
In Eq. (7.48), since A(n) is the known wave generated by Gref , only Γ6P , ΓG
and ΓL(n) need to be measured. Γ6P(n)
, and ΓG(n) can be measured by an SP reflectome-
(n)
ter or a VNA, whereas ΓL can be measured as a pseudo reflection coefficient by
means of an SP reflectometer in the reverse configuration, as shown in Fig. 7.15(b).

7.6.5 Calibration Verification

The multi-harmonic calibration of the SP reflectometer can be verified by measuring

reference signals A(n) at a different power level from the one used to determine
7.7 Waveform Engineering 183

Fig. 7.19 Comparison

between SP reflectometer and
LSNA measurements of the
time-domain waveform at the
output of the harmonic
generator [27], © IEEE 2008

(n)
B (n) during the calibration procedure. Since the SP reflectometer measures ΓL ,
traveling waves A(n) can be extracted using Eq. (7.48).
The voltage waveforms deduced from the SP reflectometer measurements are
then compared with the waveforms previously measured with an LSNA, as depicted
in Fig. 7.14. The measured voltage waveforms at the output of the harmonic genera-
tor obtained by the calibrated SP reflectometer and an LSNA are shown in Fig. 7.19.
Although the SP takes into account only the first three harmonic components of the
signal (f0 , 2f0 and 3f0 ), the relative error compared to the LSNA measurement
based on a total of five harmonics is less than 2.5 %. This result shows the accuracy
and efficiency of the SP reflectometer calibration procedure.

7.7 Waveform Engineering

The coupling of the waveform measurement setup with impedance controlling ele-
ments results into a typical waveform engineering system, such as the one depicted
in Fig. 7.20. These systems are capable of measuring high-frequency waveforms in
variable impedance environments at the terminals of the DUT.
In principle, waveform engineering is the technique in which the current and
voltage waveforms at the DUT’s terminals can be modified in a quantified manner
to optimize the performance of the DUT [42]. Through theoretical and experimental
investigations, it has been established that the terminal current and voltage wave-
forms are the unifying link between transistor technology, circuit design and RFPA
performance [3, 4]. Therefore, it is imperative to consider engineering of the wave-
form as fundamental force driving both the measurement and mathematical methods
in the overall RFPA design process, i.e., transistor optimization, circuit design, and
system integration.
The waveform engineering technique can be applicable in the design investiga-
tion/evaluation loop of either power transistor technology or power amplifiers [42].
This eliminates the black box design process and allows for new intelligent design
processes that are completely founded on theoretical waveform analysis. Alterna-
tively, it can also indirectly support the amplifier investigation/evaluation loop by
184 7 Waveform Measurement and Engineering

Fig. 7.20 Block diagram of a typical waveform engineering system

supporting the development or improving the accuracy of nonlinear transistor mod-

els or by providing computer-aided design (CAD) accessible behavioral or behav-
ioral model parameter datasets [42].

7.8 Applications of Waveform Engineering

7.8.1 Transistor Characterization

RF waveform engineering systems find extensive use in the determination of non-

linear dynamic responses of transistor devices. For example, such a system can be
used in the investigation of RF power performance of a gallium nitride (GaN) het-
erojunction field effect transistor (HFET). The poor performance of GaN HFETs
is often associated with knee walkout or current collapse problems. Knee walkout
is defined as the condition when the knee voltage (minimum drain voltage at max-
imum drain current) is increased under RF excitation over that determined under
DC excitation. On the other hand, the current collapse is the condition in which
the maximum drain current under an RF excitation is reduced compared to that
observed under DC excitation. Traditional microwave measurements, linear or non-
linear, are unable to provide visual observations; rather, they can only measure the
consequences, e.g., transconductance and output conductance dispersion, RF power
loss, and/or decreased efficiency [42].
However, RF waveform measurement and engineering systems [43] help in over-
coming such problems and readily allow observation of the knee walkout and cur-
rent collapse problem, as shown in Fig. 7.21. The plot in this figure shows the en-
gineered waveforms obtained by varying the RF fundamental load impedance from
7.8 Applications of Waveform Engineering 185

Fig. 7.21 One GHz fan

diagrams, showing measured
observation of both knee
walkout and poor pinch-off
[43], © IEEE 2006

Fig. 7.22 One GHz fan

diagrams, showing measured
knee walkout and optimized
pinch-off [43], © IEEE 2006

a low (short) to a high (open) impedance for four different drain bias voltages, in-
creasing the values of the DC quiescent (10, 20, 30 and 40 V). The RF knee walkout
and its extension with increased drain voltage are evident in these measured results.
The results in Fig. 7.21 also provide an indication that this particular technol-
ogy also suffers from poor pinch-off (minimum drain current was not zero at high
drain voltages). On the contrary, Fig. 7.22 shows similar sets of measurements with
identical knee walkout problems, but no pinch-off issues. Through these plots, one
can easily identify the role of iron doping the buffer on the dynamic transistor re-
sponse. Thus, in this case the RF waveform and engineering aided in technology
optimization, i.e., iron doped buffer design [43].
Other applications of RF current and voltage waveform measurement and engi-
neering systems in the transistor characterization domain include technology selec-
tion, reliability investigations, and engineered RF stress testing.

7.8.2 CAD Incorporation

RF waveform measurement systems are also considered a tool for the optimiza-
tion of nonlinear transistor models: therefore, they can also support nonlinear CAD.
186 7 Waveform Measurement and Engineering

Waveform engineering systems further advance the nonlinear model development

and nonlinear CAD. These systems enable the full device I-V plane to be dynam-
ically investigated without the previous requirement of DC bias point variations.
Therefore, these systems can generate more robust data for either the extraction of a
transistor’s nonlinear state functions [44] or the validation of conventional analytical
models [45].
Furthermore, engineered current and voltage waveforms measured as functions
of load impedance can be used within the CAD tool that is relevant for PA design
or behavior modeling of transistor devices [46]. The incorporation of measured data
directly into a CAD environment provides an accurate alternative to the traditional
solution requiring a nonlinear device model. This concept is analogous to the direct
use of measured S-parameter datasets rather than, for example, small-signal equiv-
alent transistor circuit models in linear CAD.

7.8.3 Power Amplifier Design

Waveform measurement and engineering systems are very useful in determining the
appropriate matching impedance required at device ports to achieve the required
performance. This performance is directly linked to the mode of operation (volt-
age and current waveforms) realized, which can be directly obtained. Through the
independent control of bias and drive, one can engineer the shape of current wave-
forms; whereas, the voltage waveforms can be engineered by varying the impedance
through a load-pull system. Therefore, these systems can be utilized to achieve the
target current and voltage waveforms obtained from theoretical analysis, in order
to reach the theoretically predicted performance. Once the desired performance is
achieved, linear CAD tools can then be employed to design a suitable matching
network.
As an example, let us consider the first-pass design of a class-J high-efficiency
broadband PA based on GaN HFET technology. Class J is defined as a mode in
which the voltage has harmonic components that make it behave asymptotically
towards a half-wave rectified sine wave. This, in practice, can be usefully approx-
imated by a suitably phased second harmonic component. Therefore, a class-J am-
plifier can be realized by engineering the waveforms using appropriate fundamental
and second harmonic terminations. In this way, a higher fundamental component can
significantly outweigh the loss in power implied by the reactive load. Consequently,
class J displays approximate half-wave rectified sinusoidal output current and volt-
age waveforms with a phase overlap between the two, as shown in Fig. 7.23 [47].
The engineered waveforms achieved after systematic waveform optimization
clearly demonstrate that the transistor can support the class-J mode of operation
and, hence, deliver the expected RF output power along with very high power-added
efficiency (PAE) across a broad bandwidth [47]. In addition, once the targeted wave-
forms are achieved, the system provides the designer with the desired input and out-
put matching circuit impedances necessary at the fundamental frequency and second
and third harmonic frequencies.
References 187

Fig. 7.23 Measured typical

class-J waveforms on a 10 W
GaN HEMT [47], © IEEE
2009

Fig. 7.24 Realized class-J

amplifier [47], © IEEE 2009

The designer now has all the information necessary to design an appropriate mi-
crowave matching circuit and assemble the amplifier shown in Fig. 7.24. Once as-
sembled and measured, the amplifier gives a measured performance that is identical
to that predicted, i.e., a first-pass design success [47].

References
1. A. Ramadan, A. Martin, T. Reveyrand, J.-M. Nebus, P. Bouysse, L. Lapierre, J.-F. Villemazet,
S. Forestier, Efficiency enhancement of GaN power HEMTs by controlling gate-source volt-
age waveform shape, in 39th European Microwave Conference Digest, Rome, Italy (Sept.
2009), pp. 1840–1843
2. Y.Y. Woo, Y. Yang, B. Kim, Analysis and experiments for high efficiency class-F and inverse
class-F power amplifiers. IEEE Trans. Microw. Theory Tech. 54(5), 1969–1974 (2006)
3. H.M. Nemati, A.L. Clarke, S.C. Cripps, J. Benedikt, P.J. Tasker, C. Fager, J. Grahn, H. Zirath,
Evaluation of a GaN HEMT transistor for load- and supply-modulation applications using
intrinsic waveform measurements, in IEEE Microwave Theory and Techniques Society’s In-
ternational Microwave Symposium Digest, Anaheim, USA (May 2010), pp. 509–512
4. C.J. Wei, P. DiCarlo, Y.A. Tkachenko, R. McMorrow, D. Bartle, Analysis and experimental
waveform study on inverse class-F mode of microwave power FETs, in IEEE Microwave
Theory and Techniques Society’s International Microwave Symposium Digest, Boston, USA
(June 2000), pp. 525–528
188 7 Waveform Measurement and Engineering

5. W. Grabinski, B. Nauwelaers, D. Schreurs (eds.), Transistor Level Modeling for Analog/RF

IC Design (Springer, Dordrecht, 2006)
6. D.M. Snider, A theoretical analysis and experimental confirmation of the optimally loaded
and overdriven RF power amplifier. IEEE Trans. Electron Devices, 851–857 (1967)
7. J.D. Rhodes, Output universality in maximum efficiency linear power amplifiers. Int. J. Circuit
Theory Appl. 31, 385–405 (2003)
8. K.A. Remley, D.F. Williams, Sampling oscilloscope models and calibrations, in IEEE
Microwave Theory and Techniques Society’s International Microwave Symposium Digest,
Philadelphia, USA (June 2003), pp. 1507–1510
9. T.S. Clements, P.D. Hale, D.F. Williams, C.M. Wang, A. Dienstfrey, D.A. Keenan, Calibration
of sampling oscilloscopes with high-speed photodiodes. IEEE Trans. Microw. Theory Tech.
54(8), 3173–3181 (2006)
10. J. Verspecht, P. Debie, A. Barel, L. Martens, Accurate on-wafer measurement of phase and
amplitude of the spectral components of incident and scattered voltage waves at the signal
ports of a nonlinear microwave device, in IEEE Microwave Theory and Techniques Society’s
International Microwave Symposium Digest, Orlando, USA (May 1995), pp. 1029–1032
11. J. Fitzpatrick, Error models for systems measurement. Microw. J. May, 63–66 (1978)
12. R.S. Tucker, P.D. Bradley, Computer-aided error correction of large-signal load-pull measure-
ments. IEEE Trans. Microw. Theory Tech. 32(3), 296–300 (1984)
13. W.R. Scott Jr., G.S. Smith, Error corrections for an automated time-domain network analyzer.
IEEE Trans. Instrum. Meas. 35, 300–303 (1986)
14. D.M. Pozar, Microwave Engineering, 3rd edn. (Wiley, New York). ISBN 0-471-44878-8
15. Understanding the fundamental principles of vector network analysis, Application Note, Agi-
lent AN 1287-1
16. S.A. Mass, Nonlinear Microwave Circuits, 2nd edn. (Artech House, Norwood, 2003)
17. M. Sipila, K. Lehtinen, V. Porra, High-frequency periodic time-domain waveform measure-
ment system. IEEE Trans. Microw. Theory Tech. 36(10), 1397–1405 (1988)
18. U. Lott, Measurement of magnitude and phase of harmonics generated in nonlinear microwave
two-ports. IEEE Trans. Microw. Theory Tech. 37(10), 1506–1511 (1989)
19. D. Barataud, C. Arnaud, B. Thibaud, M. Campovecchio, J.-M. Nebus, J.P. Villotte, Measure-
ments of time-domain voltage/current waveforms at RF and microwave frequencies based on
the use of a vector network analyzer for the characterization of nonlinear devices—application
to high-efficiency power amplifiers and frequency multipliers. IEEE Trans. Instrum. Meas.
47(5), 1259–1264 (1998)
20. Agilent Technical Data Sheet, PNA-X series network analyzer, http://cp.literature.agilent.
com/litweb/pdf/N5242-90007.pdf (Online)
21. F. van Raay, G. Kompa, A new on-wafer large signal waveform measurement system with
40 GHz harmonic bandwidth, in IEEE Microwave Theory and Techniques Society’s Interna-
tional Microwave Symposium Digest, Albuquerque, USA (June 1992), pp. 1435–1438
22. M. Demmler, P.J. Tasker, M. Schlechtweg, A vector corrected high power on-wafer measure-
ment system with a frequency range for the higher harmonics up to 40 GHz, in 24th European
Microwave Conference, Cannes, France (Sept. 1994), pp. 1367–1372
23. J. Benedikt, R. Gaddi, P.J. Tasker, M. Goss, High-power time-domain measurement system
with active harmonic load-pull for high-efficiency base-station amplifier design. IEEE Trans.
Microw. Theory Tech. 48(12), 2617–2624 (2000)
24. W.S. El-Deeb, M.S. Hashmi, S. Bensmida, N. Boulejfen, F.M. Ghannouchi, Thru-less cali-
bration algorithm and measurement system for on-wafer large-signal characterization of mi-
crowave devices. IET Microw. Antenna Propag. 4(11), 1773–1781 (2010)
25. W.S. El-Deeb, M.S. Hashmi, N. Boulejfen, F.M. Ghannouchi, Small-signal, complex dis-
tortion and waveform measurement system for multiport microwave devices. IEEE Instrum.
Meas. Mag. 14(3), 2833 (2011)
26. Agilent Technical Data Sheet, Agilent nonlinear vector network analyzer, http://cp.literature.
agilent.com/litweb/pdf/5989-8575EN.pdf (Online)
References 189

27. S. Bensmida, P. Poire, R. Negra, F.M. Ghannouchi, New time-domain voltage and current
waveform measurement setup for power amplifier characterization and optimization. IEEE
Trans. Microw. Theory Tech. 56(1), 224–231 (2008)
28. F.M. Ghannouchi, A. Mohammadi, The Six-Port Technique with Microwave and Wireless Ap-
plications (Artech House, Norwood, 2009)
29. The Microwave Transition Analyser, Measure 25 ps transitions in switched and pulsed mi-
crowave components, Hewlett Packard Product Note 70820-2, 1991
30. W.S. El-Deeb, S. Bensmida, N. Boulejfen, F.M. Ghannouchi, An impedance and power flow
measurement system suitable for on-wafer microwave device large-signal characterization.
Int. J. RF Microw. Computer-Aided Eng. 4(11), 306–312 (2010)
31. W.S. El-Deeb, S. Bensmida, F.M. Ghannouchi, A de-embedding technique for on-wafer simul-
taneous impedance and power flow measurements, in IEEE Instrumentation and Measurement
Technology Conference, Victoria, Canada (May 2008), pp. 58–61
32. J.R. Juroshek, D.X. LeGolvan, Correcting for systematic errors in one-port calibration stan-
dards, in 62nd ARFTG Microwave Measurement Conference, Boulder, USA (Dec. 2003),
pp. 119–126
33. P.S. Blockley, J.G. Rathmell, Towards generic calibration, in 65th ARFTG Measurement Con-
ference, Long Beach, USA (June 2005), pp. 1–4
34. Tektronix, TDS794D user’s manual, http://www2.tek.com/cmswpt/psdetails.lotr?ct=PS&cs
=psu&ci=13446&lc=EN
35. J.G. Lecky, A.D. Patterson, J.A.C. Stewart, A vector corrected waveform and load line mea-
surement system for large signal transistor characterization, in IEEE Microwave Theory and
Techniques Society’s International Microwave Symposium Digest, Orlando, USA (May 1995),
pp. 1243–1246
36. D.J. Williams, P.J. Tasker, An automated active source- and load-pull measurement system, in
6th IEEE High Frequency Student Colloquium, Cardiff, UK (Sept. 2001), pp. 7–12
37. R.B. Marks, D.F. Williams, A general waveguide circuit theory. J. Res. Natl. Inst. Stand.
Technol. 97(5), 533–562 (1992)
38. J. Verspecht, Large signal network analysis. IEEE Microw. Mag. 6(4), 82–92 (2005)
39. S. Bensmida, P. Poire, F.M. Ghannouchi, Source-pull/load-pull measurement system based on
RF and baseband coherent active branches using broadband six-port reflectometers, in 37th
European Microwave Conference, Munich, Germany (Oct. 2007), pp. 953–956
40. E. Bergeault, B. Huyart, G. Geneves, L. Jallet, Accuracy analysis for six-port automated net-
work analyzers. IEEE Trans. Microw. Theory Tech. 38(3), 492–496 (1990)
41. F.M. Ghannouchi, R. Larose, R.G. Bosisio, A new multiharmonic loading method for large-
signal microwave and millimeter-wave transistor characterization. IEEE Trans. Microw. The-
ory Tech. 39(6), 986–992 (1991)
42. P.J. Tasker, Practical waveform engineering. IEEE Microw. Mag. 10(7), 65–76 (2009)
43. C. Roff, P. McGovern, J. Benedikt, P.J. Tasker, R.S. Balmer, D.J. Wallis, K.P. Hilton, J.O.
Maclean, D.G. Hayes, M.J. Uren, T. Martin, Detailed analysis of dc-RF dispersion in Al-
GaN/GaN HFETs using waveform measurements, in 1st European Microwave Integrated Cir-
cuits Conference, Manchester, UK (Sept. 2006), pp. 43–45
44. R. Gaddi, J.A. Pla, J. Benedikt, P.J. Tasker, LDMOS electro-thermal model validation from
large-signal time-domain measurements, in IEEE Microwave Theory and Techniques Society’s
International Microwave Symposium Digest, Phoenix, USA (May 2001), pp. 399–402
45. D.G. Morgan, G.D. Edwards, A. Phillips, P.J. Tasker, Full extraction of PHEMT state func-
tions using time domain measurements, in IEEE Microwave Theory and Techniques Society’s
International Microwave Symposium Digest, Phoenix, USA (May 2001), pp. 823–826
46. G. Simpson, J. Horn, D. Gunyan, D.E. Root, Load-pull + NVNA = enhanced X-parameters
for PA designs with high mismatch and technology-independent large-signal device models,
in 72nd ARFTG Microwave Measurement Conference, Portland, USA (Dec. 2008), pp. 88–91
47. P. Wright, J. Lees, J. Benedikt, P.J. Tasker, S.C. Cripps, A methodology for realizing high
efficiency class J in a linear and broadband PA. IEEE Trans. Microw. Theory Tech. 57(12),
3196–3204 (2009)
Chapter 8
Advanced Configurations and Applications

This chapter presents some advanced applications of load-pull and source-pull sys-
tems. The major emphasis of this chapter is on multi-tone and modulated load-pull
systems and their applications. Subsequently, a noise characterization system is de-
scribed. Finally, a description of a mixer characterization technique is presented.

8.1 Introduction
Radio frequency power amplifiers (RFPAs) are generally characterized using load-
pull measurement techniques to determine the parametric values of efficiency,
power-added efficiency (PAE), gain, power output, etc., of the power amplifier (PA).
In these measurements, various complex load conditions are presented to the device,
in order to determine the optimal loading condition for the desired parameters. In
the majority of applications, the signals used are single continuous-wave (CW) sig-
nals. However, in practical situations, such as load-pull measurements for optimal
intermodulation (IMD) components, broadband matching impedance synthesis and
device characterization under digitally modulated stimulus, are extremely difficult
to carry out using standard load-pull setups.
The major problem in traditional load-pull setups when employing a multi-tone
and modulated stimulus stems from the different delays caused by individual tones
in the spectrum of the stimulus in the measurement setup, while transmitting and
reflecting the modulated traveling waves [1–3]. In order to overcome such prob-
lems, requirement/application-driven load-pull setups have been developed [4–12].
These setups have proven to be highly effective in characterizing devices for the
determination of optimal parametric values for the desired application.

8.2 Multi-tone Load-Pull Technique

Intermodulation distortion (IMD) analysis is a good way to assess the nonlinear
behaviors of RF and microwave amplifiers [5, 13]. IMD analysis is relatively well

F.M. Ghannouchi, M.S. Hashmi, Load-Pull Techniques with Applications to Power 191
Amplifier Design, Springer Series in Advanced Microelectronics 32
DOI 10.1007/978-94-007-4461-5_8, © Springer Science+Business Media Dordrecht 2013
192 8 Advanced Configurations and Applications

Fig. 8.1 Diagram of the multi-tone load-pull measurement setup developed by Ghannouchi et al.
[4], © IEEE 1997

studied, theoretically and experimentally, under two-tone excitations [14, 15]. How-
ever, IMD characterization and analysis under multi-tone excitations have not yet
been well investigated in open literature [16].
Theoretical postulations about IMD are available [15, 16], but those Ghannouchi
et al. [4, 17] were among the first to develop multi-tone signal test setups to ex-
perimentally evaluate those theoretical postulations. The setup in [17] focuses on
multi-tone device characterization when the tones have an optimal phase relation-
ship [18]; whereas the setup reported in [4] and shown in Fig. 8.1 is capable of IMD
characterization with different multi-tone phase distributions.
For multi-tone load-pull measurements, it is important to first generate a very
clean spectrum of the multi-tone excitation to be applied to the transistor. In Fig. 8.1,
the arbitrary waveform generator (AWG) generates the desired number of tones with
the desired frequency spacing, power level and phase distribution. The phase distri-
bution of the tones can be random, uniform, or fixed by the user. Using an AWG
reduces the complexity of the system as the number of tones increases, and the syn-
chronization problem between the tones is avoided. The other indispensable instru-
ment used in the setup is the microwave transition analyzer (MTA, e.g., HP70820),
which acts as a network analyzer with its home-designed test set.
The mixer up-converts the baseband signal generated by the AWG to around the
microwave operating signal available from the microwave synthesizer connected to
the other input of the mixer. The highly selective narrowband filter placed at the out-
put of the mixer is used to pass only the principal N -tone spectrum, where N is the
number of tones. The multi-tone signal at the output of the filter is obtained with an
intermodulation rejection (IMR) of 55 dBc. The filtered spectrum gets amplified by
a linear amplifier to reach the desired power level. The variable attenuator performs
a power sweep, and the mechanical tuner adjusts the input reflection coefficient for
maximum power transfer to the transistor.
The incident and reflected waves at the input and output of the transistor are
sampled by two dual couplers with small coupling factors (20 dB). These sampled
waves are received and measured by the MTA via a switching stage. The output
8.2 Multi-tone Load-Pull Technique 193

Fig. 8.2 Various parameter

variation options in the setup
developed by Ghannouchi et
al. [4], © IEEE 1997

mechanical tuner passively tunes the load. The insertion losses in the tuner limit
the maximum achievable reflection coefficient magnitude to 0.9. However, use of
recently developed load-pull configurations in this setup can improve the maximum
achievable reflection coefficient [19]. A Hewlett-Packard interface bus (HP-IB) con-
trolled through a personal computer (PC) aids in the data acquisition from the MTA,
system calibration, and the control of single pole, double throw (SPDT) switches.
De-embedding of the measured raw data taken by the MTA to the reference mea-
surement planes of the transistor can be achieved by carrying out short-open-load-
thru (SOLT) [20] or thru-reflect-line (TRL) [21] techniques. For power corrections,
the measuring ports can be calibrated with an independent power meter.
Overall, the setup in Fig. 8.1 is appropriate for both packaged and on-wafer de-
vices. In addition, it is essential to note that the measurement systems [5, 13] are
only useful for two-tone excitations where two separate signal generators are em-
ployed. These systems have severe limitations, due to the increasing complexity if
the number of tones increases, as one signal generator is required for each tone.
However, the Ghannouchi technique overcomes this problem by using an AWG and
a simple mixer to generate the spectrum of any desired tones; therefore, it elimi-
nates the need for a large number of signal generators and the associated problem
of signal synchronization in those setups.
This configuration enables several kinds of multi-tone measurements, as shown
in Fig. 8.2, through the variation of independent parameters of the system. It is
evident that this setup allows for the characterization of DUTs in terms of most
of the common parameters, such as input power, number of tones, carrier phase
distribution, biasing conditions, load impedance and IMR measurement mode (in-
band/out-band). By varying various parameters, these measurements can generate a
complete database that could be useful in the development of different charts for a
given DUT in terms of nonlinear characteristics.
To demonstrate the usefulness of such a setup, a metal semiconductor field effect
transistor (MESFET, i.e., SCK0151P) exhibiting a 1 dB compression point (P1dB )
194 8 Advanced Configurations and Applications

Fig. 8.3 IMR degradation

with equal power at each tone
[4], © IEEE 1997

Fig. 8.4 Required power

backoff to maintain constant
IMR [4], © IEEE 1997

of 23 dBm was characterized for assessing the effect of various parameters on the
overall IMR and validation of the theoretical intermodulation distortion analysis
[16]. The device was biased at Vds = 10 V and Ids = 100 mA for class-A operation.
The first experiment constituted the evaluation of IMR degradation as the num-
ber of tones increases, as well as the required backoff level in the input power to
maintain a certain constant IMR. All the measurements were carried out for input
signals having 2, 4, 8, 16 and 32 tones, with spacing of 100 kHz between tones.
The load was set at its optimal value of ΓL = 0.58∠172°, which was obtained from
load-pull measurement data.
Using the MTA, the input reflection coefficient, absorbed input power per tone,
output reflection coefficient, output power per tone, and third-order IMR of the tran-
sistor were measured. For each number of tones, the IMR parameter was measured
for an input power sweep up to the 1 dB compression region of the transistor. Such
8.2 Multi-tone Load-Pull Technique 195

measurement was repeated for ten random phase distributions of the carriers. The
experimental results, as shown in Figs. 8.3 and 8.4 [4], correspond to the two ex-
treme cases of the phased distribution among the ten cases considered: one is the
best case, and the other is the worst case.
Figure 8.3 illustrates a comparison between the measured and theoretical [16]
IMR degradation as a function of the number of tones for the same total input power
level (Pin/total = 7 dBm). The total input power in dB is calculated from Eq. (8.1).
Pin/total = Pin/tone + 10 log(N ) (8.1)
where N is the number of tones in the input signal.
It is apparent that the theoretical IMR degradation falls in the range defined be-
tween the best and worst cases of the carriers’ phase distribution. This result proves
that the phase distribution of the carriers has a strong influence on the IMR; and,
in this particular example, the difference is up to 20 dBc between the IMR degra-
dation values of the two extreme cases. Thus, it supports the hypothesis that the
power characteristics of amplifiers in a multichannel operation strongly depend on
the behavior of the carrier signal of each channel.
Figure 8.4 shows the measured and theoretical [16] total input power backoff
required to maintain a constant IMR. Again, the theoretical results fall in the range
bounded by the best and worst cases. It is also important to note that, for a certain
phase distribution of the carriers, a backoff in the total input power is not needed to
maintain a constant IMR, as it can be seen for the best case scenario where the power
backoff required is very small (around 0.2 to 0.4 dB). This is very important in the
design of highly efficient linear amplifiers, where the transistor has to be operated
near its compression region to achieve the desired high power-added efficiency.
In the second experiment, the effect of amplifier class on the IMR was mea-
sured. In this measurement, the carriers’ phase distribution was kept uniform for
all performed operating classes (i.e., identical absolute phase for each tone). The
transistor was loaded with the optimal impedance for maximum output power
(ΓL = 0.58∠172°). The IMR measurements were performed at seven bias points,
from 30 % Idss to 70 % Idss with a 10 % step, where Idss is the saturation current.
The measurement results were obtained by carrying out an input power sweep at
each bias point for N = 2 and 8 tones.
It can be observed from the experimental result that the IMR slope, as a func-
tion of drain current, varies at a certain bias point depending on the applied input
power. This experimental result is not readily explainable. To further investigate this
experimental observation, multi-tone harmonic-balance simulations [22] were per-
formed using Agilent’s Advanced Design System (ADS) software on a MESFET
model extracted from the measured DC and scattering parameter (S-parameter) data
by applying using the cubic model [23].
The comparison of the measured and simulated results, as shown in Figs. 8.5 and
8.6, for values of N of 2 and 8, respectively, reveals that there is an acceptable agree-
ment in terms of the IMR levels, mainly with respect to the IMR variation slope.
Discrepancies between the measurement and simulation results increase when the
transistor moves toward its pinch-off region. This may be attributed to limitations
of the model extracted from the measured data.
196 8 Advanced Configurations and Applications

Fig. 8.5 Effects of changing

the saturation current (Idss )
on IMR for N = 2 [4],
© IEEE 1997

Fig. 8.6 Effects of changing

the saturation current (Idss )
on IMR for N = 8 [4],
© IEEE 1997

It can be observed that the IMR is optimal in class-A operation (50 % Idss ) in
small-signal operating mode (low-input power level), whereas the IMR worsens to-
ward the clipping and pinch-off regions. In the large-signal operating mode (near or
in the compression region), the IMR degradation increases with the increase in drain
current and tends to become constant at high Idss . The multi-tone setup developed
by Ghannouchi et al. [4], therefore, enables the optimization of power and linearity
performances of amplifiers through biasing conditions.
8.3 Real-Time Multi-harmonic Load-Pull Technique 197

Fig. 8.7 Real-time multi-harmonic load-pull setup developed by Cui et al. [24], © IEEE 2006

8.3 Real-Time Multi-harmonic Load-Pull Technique

Cui et al. [24] developed an efficient real-time open-loop multi-harmonic load-pull
system, as shown in Fig. 8.7, based on the large signal network analyzer (LSNA).
The system is capable of rapidly synthesizing a wide range of fundamental and har-
monic impedances without the usual stability problem associated with traditional
active closed-loop load-pull systems [25] and the slowness inherent in the conven-
tional open-loop load-pull system [26]. In essence, this setup is an advanced version
of the systems reported in [27]. Roblin et al. [28] developed a further advancement
on this system.
In Fig. 8.7, Ports 1 and 2 are used for the gate and drain of the DUT, respectively.
The bias tees and DC power supplies set the quiescent operating point of the DUT.
The voltage and current sensors detect the time variations of the device bias voltage
and current, which gets displayed on the oscilloscope. A group-delay calibration
procedure is needed to synchronize the RF envelope measured by the LSNA with
the time dependence of the acquired bias current signal (with a period of ω). The
ESG generates a fundamental signal (ω0 ) used to excite the DUT input. The two
harmonic sources (PSG and Anritsu) inject 2ω0 + ω and 3ω0 + ω in the real-
time tuning case (i.e., swept phase measurement case) or generate 2ω0 and 3ω0
in the constant phase measurement scenario. The 10 MHz reference signals are tied
together for phase synchronization. The diplexer provides 50  termination at Port 2
for the fundamental frequency, while offering a path for the injected second and third
harmonics at the DUT output. The LSNA acquires incident and reflected waves in
the frequency domain at the de-embedded calibration planes. The LSNA is operated
in the modulation mode to deal with the modulation effects brought about by ω.
This system employs a real-time approach in which the tuning at the nth har-
monic utilizes a CW tone a2 (nω0 + ω) with a frequency offset ω (see Fig. 8.7)
from the targeted nth harmonic frequency injected at the device output, so that the
phase of the harmonic load reflection coefficient is continuously swept in a single
198 8 Advanced Configurations and Applications

Fig. 8.8 Output power contour (left) and PAE contour (right) in ΓL (ω0 ) plane, obtained from the
ω0 + ω real-time swept phase measurements [24], © IEEE 2006

LSNA measurement [24]. The power of the CW tone |a2 (nω0 + ω)|2 can be swept
using a computer via a general purpose interface bus (GPIB). It aids in varying the
radius of the reflection coefficient loci, which in turn can produce a large set of
harmonic reflective load coefficients with amplitudes smaller or larger than unity
mapping of the extended Smith chart.
As an example to demonstrate the capability of this setup, a GaN HEMT (gal-
lium nitride high electron mobility transistor) device was considered with a quies-
cent DC bias of VGS = −2.77 V, VDS = 4.25 V and IDS = 1.7 mA for the interactive
design of a high-efficiency PA operating in class F. The LSNA was used to acquire
frequency domain data up to the fourth harmonic (N = 4). In principle, before car-
rying out real-time second and third harmonic tuning, it is imperative to run the
real-time active load-pull measurement at the fundamental frequency, so that an op-
timal load at ω0 can be obtained. This optimal load at ω0 eventually forms the basis
for subsequent harmonic tuning.
In the fundamental load-pull case, the ESG provides a 2 GHz fundamental signal
(ω0 ), and the PSG source injects a ω0 + ω [ω = 200 KHz] signal at the DUT’s
output. The obtained fundamental output power and PAE contours given in Fig. 8.8
show a strong dependence on the loading condition. It is also apparent that the opti-
mal loading for this device for both the PAE and output power lie in the same region
of Smith chart.
Case Z marked in Fig. 8.8 represents the impedance point for achieving a high
PAE while keeping a large output power. Although both the PAE and output power
show progressive trends in the first quadrant of Smith chart, one can identify that
there are some termination impedances, which are potentially unstable because the
corresponding magnitude of |Γin (ω0 )| is greater than one, as shown in Fig. 8.9.
This situation indicates negative input impedances, which potentially could lead to
unstable operation.
8.3 Real-Time Multi-harmonic Load-Pull Technique 199

Fig. 8.9 |Γin (ω0 )| contour

obtained from the ω0 + ω
real-time swept phase
measurements [24], © IEEE
2006

Table 8.1 Comparison of three typical cases in Fig. 8.8: cases X, Y and Z [24], © IEEE 2006
ΓL (ω0 ) PAE (%) Output power (mW)
Swept phase Constant phase Swept phase Constant Phase

Case X: 0 47.9 49.1 69.9 72.8

Case Y: 0.2∠0° 55.4 57.6 67.1 72.4
Case Z: 0.53∠19.3° 80.0 70.6 53.0 56.2

In this situation, study terminations with only positive input impedances are in-
vestigated; therefore, three typical cases (X, Y and Z) are selected from the PAE
contour plot in Fig. 8.9 for efficiency comparison as shown in Table 8.1. For these
cases, the PAE and output power data extracted from the real-time swept phase
measurement correlates to the results obtained from the constant phase measure-
ment. Note that in the constant phase measurement the ESG source injects 2 GHz
fundamental signal at the drain output without any frequency offset.
It can be observed in Table 8.1 that, in case Z, a higher PAE is extracted from the
swept phase measurement than is obtained from the constant phase measurement,
due to the large swept power at (ω0 + ω). For cases X or Y, the PAE difference
is less between the swept and constant phase measurements, because smaller power
at (ω0 + ω) is applied. Due to the high PAE (70.6 %) and stability consideration,
the load reflection coefficient of case Z was chosen as the optimal fundamental
impedance for further analysis at the second and third harmonics.
For load-pulling the harmonic, the PSG source injects (2ω0 + ω) at the DUT’s
output, whereas the ESG source provides the 2 GHz (ω0 ) excitation at the DUT’s
input. In this case, the fundamental impedance is kept to the previously identified
optimal impedance corresponding to point Z, whereas the third harmonic is left
arbitrarily terminated. The sensitivity of the second harmonic termination on the
obtained PAE and output power is given in Fig. 8.10.
200 8 Advanced Configurations and Applications

Fig. 8.10 PAE contour (left) and output power contour (right) in the ΓL (2ω0 ) plane, obtained
from the 2ω0 + ω real-time swept phase measurements [24], © IEEE 2006

Table 8.2 Comparison of three typical cases in Fig. 8.10: cases A, B and C [24], © IEEE 2006
ΓL (ω0 ) PAE (%) Output power (mW)
Swept phase Constant phase Swept phase Constant phase

Case A: 0.985∠ −156° 74.95 72.03 61.15 57.7

Case B: 0.976∠ −180° 74.18 71.82 59.79 57.6
Case C: 0 67.13 66.47 54.6 53.4

It is evident from Fig. 8.10 that the second harmonic termination with the highest
PAE (case A) is close to the short circuit condition and is in agreement with the
class-F PA design theory. This high PAE can be further verified from the constant
phase measurement, which turns out to be 72.03 %. The corresponding output power
using the developed real-time load-pull system and using standard constant phase
measurements are 61.15 mW and 57.7 mW, respectively, as given in Table 8.2.
In Table 8.2, the PAEs for three typical cases (A, B, C) are picked from the PAE
contours in Fig. 8.10 for the sake of comparison. For the constant phase measure-
ment, the PSG source simply injects a 2ω0 signal to the DUT’s output. In each of
these three cases, both the PAE and output power results calculated from the real-
time swept phase measurement are consistent with those from the constant phase
measurement. The degradations of PAE and output power from the swept phase to
the constant phase are much less than those in Table 8.1. This fact indicates that the
memory effect brought about by the 200 kHz frequency is prominent only at the
fundamental frequency, rather than in the second harmonic [24].
The third harmonic real-time load-pull measurement utilizes the optimized load
conditions acquired from the tests at the fundamental frequency and harmonic
(points Z and A in Figs. 8.8 and 8.10, respectively). In this case, the ESG gives
8.4 Modulated Signal Load-Pull Technique 201

Fig. 8.11 PAE contour (left) and output power contour (right) in the ΓL (3ω0 ) plane, obtained
from the 3ω0 + ω real-time swept phase measurements [24], © IEEE 2006

the 2 GHz (ω0 ) excitation at the DUT’s input, while the second harmonic (2ω0 )
and the third harmonic with frequency offset (3ω0 + ω) are injected by the two
harmonic sources to the DUT’s output. The load tuner is properly adjusted to the op-
timal fundamental termination. The PAE and output power obtained from the swept
phase measurements in the ΓL (3ω0 ) plane are given in Fig. 8.11.
In Fig. 8.11, the high PAE (around 80 %) contour lines are mainly distributed
around the rightmost part (nearer to the open circuit condition) of the Smith chart,
which is in close agreement with class-F PA theory. In this particular case, the opti-
mum for class F turns out to be case W with ΓL (3ω0 ) of 0.97∠17.3°, which is close
to the edge of Smith chart.
The setup developed by Cui et al. [24] thus enables device characterization for
high-efficiency PA design. It offers great potential in expediting the device charac-
terization and, therefore, minimizing the power amplifier design time.

8.4 Modulated Signal Load-Pull Technique

Load-pull measurements have been generally limited to single- and two-tone signals
for optimizing the output power and third-order intermodulation products (IMD3)
performances of transistors. However, with the advancement in modulation tech-
niques, traditional figures of merits, such as the one dB compression point (P1dB )
and third-order inception point (IP3 ), obtained from one and two-tone stimuli can no
longer be used to fully predict the behavior and performance of the transceiver. This
is due to the fact that signals represented by discrete spectra do not reflect the com-
plex digitally modulated signals. Furthermore, it has been thoroughly studied and
found that the performance of a transistor is strongly dependent on the type of ex-
citation used during characterization and measurement [29, 30]. In order to address
202 8 Advanced Configurations and Applications

Fig. 8.12 Illustration of a load-pull setup developed by Ghanipour et al. [6], © IEEE 2007

these concerns and investigate the device performance under digitally modulated
stimuli, Ghanipour et al. [6] developed a load-pull setup using standard equipment,
as shown in Fig. 8.12.
The load-pull system illustrated in Fig. 8.12 utilizes an automatic tuner system
(ATS) from Maury Microwave [29]. A vector signal generator (VGS) (Rohde and
Schwarz SMIQ03B) uploads and generates the desired digitally modulated stim-
uli. A traveling wave tube amplifier (1277H) amplifies the stimuli, whereas power
meters (HP437B) measure the output and reflected powers. The vector signal ana-
lyzer (VSA) (Rohde and Schwarz FSQ8) measures the adjacent channel power ratio
(ACPR) and IMD3 of the signals. The ATS software controls all the instruments in
the measurement setup.
To show the usefulness of this setup, a high-power laterally diffused metal ox-
ide semiconductor (LDMOS) device (MRF21030LR3) with a typical P1dB output
power of 44.6 dBm was selected. The experiment was carried out at 2.14 GHz using
WCDMA (wideband code division multiple access) modulated signals (peak-to-
average power ratio (PAPR) values of 9.4, 8.6, 7.6, 7.6, 6.5 and 6.1 dB), OFDM (or-
thogonal frequency division multiplexing) signals (PAPR values of 9.4 and 6.1 dB),
and a two-tone signal. In order to resemble the bandwidth of a single-carrier
WCDMA signal, the OFDM signal had a 5 MHz bandwidth, and the two-tone signal
generated had tone spacing of 5 MHz. For each type of source signal, the average
output power was adjusted, so that the peak output power for all eight signals was
40.4 dBm at 4 dB backoff.
Power gain and lower channel ACPR (ACLR)/IMD3 contours, as shown in
Fig. 8.13, for the WCDMA signal with a PAPR of 9.1 dB and a two-tone excita-
tion reveal that the size and shape of the contours are very much dependent on the
type of stimuli used during the load-pull measurement.
It is apparent that the optimal loading conditions for two-tone and WCDMA
stimuli are quite different. To quantify the difference, the optimal impedance
8.4 Modulated Signal Load-Pull Technique 203

Fig. 8.13 Gain and IMD3 contours for an MRF21030 device with a two-tone source (left) and gain
and ACLR contours for an MRF21030 device with a WCDMA (PAR = 9.4 dB) source (right) [6],
© IEEE 2007

Fig. 8.14 Optimal gain

(shown in the upper half of
the Smith chart) and optimal
ACLR/IMD3 (shown in the
lower half of the Smith chart)
impedances for two-tone and
WCDMA (with PAPR values
of 9.4, 8.6, 7.6, 6.5 and
6.1 dB) signals [6], © IEEE
2007

obtained from all the stimuli under consideration, as shown in Fig. 8.14, can be
analyzed. In this figure, the optimal gain impedances are located in the upper part
of the Smith chart, while the optimal ACLR and IMD3 impedances are located in
the lower section. It can be seen that the reactance of the optimal gain impedance
decreases as the average power of the excitation signal is increased (to compensate
for the decreasing PAPR). Similarly, the resistance of the optimal ACLR impedance
for WCDMA signals tends to increase as the average power is increased. The ACLR
and gain contours and the location of the respective optimal impedances for OFDM
signals (not shown here) also matched those for WCDMA signals with similar PAPR
values.
204 8 Advanced Configurations and Applications

Fig. 8.15 Optimal gain

(shown in the upper half of
the Smith chart) and optimal
ACLR/IMD3 (shown in the
lower half of the Smith chart)
impedances for two-tone and
WCDMA (PAPR = 6.1 dB)
signals at various output
power levels [6], © IEEE
2007

It can, therefore, be concluded that device performance is predominately depen-

dent on stimuli’s power, PAPR and bandwidth and are largely independent of the
type of excitation, i.e., OFDM or WCDMA signal.
To enhance the understanding of whether the shift in the optimal gain and ACLR
impedances was due to an increase in the average power or due to a decrease in
PAPR, another experiment was performed, in which load-pull measurements of the
WCDMA signal with a PAPR of 6.1 dB were performed with the average output
power at 31, 34.3 and 37.4 dBm and compared to two-tone load-pull measurements
with the same average power.
The results in Fig. 8.15 show that, even though the optimal impedances for a two-
tone signal are significantly different from the optimal impedances for a modulated
signal, both are similarly affected by an increase in output power. As the required
output power is increased, the reactance of the optimal gain impedance tends to
decrease and is in agreement with theoretical results [32]. This experiment also
revealed a new insight: there is increased resistance of the optimal ACLR impedance
as the output power is increased, which can be very useful in the optimization of
devices for PA design applications.

8.5 Multi-tone Envelope Load-Pull Technique

As mobile communications have become more complex with the transmission of
huge volumes of data, the modulation schemes used to transmit the information
have also become more complex often requiring transmission over very wide band-
widths. In order to fully characterize and design the required PA properly, it is vital
that large-signal measurement systems adapt to accommodate these complex mod-
ulations.
The ideal way to predict the performance of a designed PA for modern com-
munication system is a priori characterization of devices under modulated excita-
tions. A load-pull setup capable of synthesizing constant impedance over a mod-
ulation bandwidth and beyond is a logical requirement for characterizing devices
8.5 Multi-tone Envelope Load-Pull Technique 205

Fig. 8.16 Generic block diagram illustrating the concept of the ELP technique

for such applications. Hashmi et al. [33] proposed a concept to emulate broadband
impedance, and the technique has been further enhanced [11, 34] and shown to pro-
vide significant characterization data [35].
The envelope load-pull (ELP) system (a block diagram of which is illustrated in
Fig. 8.16) explained in Chap. 6 emulates reflection coefficient given by Eq. (8.2).
However, for a modulated signal of modulation frequency ωm with n tones, this
relationship changes to Eq. (8.3) [11].
a2 (ω)
Γload (ω) = = X + jY (8.2)
b2 (ω)
a2 (ω)
Γload (ωc − nωm ) = = (X + j Y )ej nωm τ (8.3)
b2 (ω)
Parameter τ in Eq. (8.3) is the group delay around the loop of the load-pull system. It
is apparent from Eq. (8.3) that the load reflection coefficient becomes dependent on
the modulation frequency through a phase operator, ej nωm τ , and that it leads to phase
spread between the loads at ωc − nωm and ωc + nωm frequency pairs, respectively.
This spread can be alternatively attributed to the delayed envelope of a2 (reflected
traveling wave) at the reference plane when it traverses through the feedback loop.
Equation (8.4) gives the envelope domain expression for a CW signal, whereas the
envelope domain expression changes to Eq. (8.5) for a modulated signal.
 
Ia (t) + j Qa (t) = (X + j Y ) Ib (t) + j Qb (t) (8.4)
 
Ia (t) + j Qa (t) = (X + j Y ) Ib (t − τ ) + j Qb (t − τ ) (8.5)
Experimentally, the effect of group delay can be observed in Fig. 8.17, which
was measured using a time-domain modulated waveform system [36]. These results
are from an experiment carried out on a thru line in place of a DUT with three-tone
excitation (on a 1.8 GHz carrier) having a tone spacing of 100 kHz for emulation
of the reflection coefficients of 0.9∠90° and 0.9∠270°. It is evident that the loss
of synchronization between the envelopes of transmitted traveling wave, b2 , and
reflected traveling wave, a2 , cause spread in the synthesized reflection coefficient
and thus agree perfectly with Eqs. (8.3) and (8.5).
206 8 Advanced Configurations and Applications

Fig. 8.17 Measured

three-tone envelope (top) and
two emulated reflection
coefficients in the Smith chart
(bottom) [11], © IEEE, 2008

Fig. 8.18 Block diagram of the envelope load-pull system with a delay compensator developed by
Hashmi et al. [33] for synchronizing the second cycle of the transmitted traveling wave, b2 , with
the first cycle of the reflected traveling wave, a2

In general, time-domain waveform measurement systems utilized in nonlinear

device characterization employ repetitive signals. Therefore, Hashmi et al. [33]
identified that a delay compensator in the feedback loop of ELP system, as shown in
Fig. 8.18, can be incorporated in such a way that the envelope domain expressions
for transmitted traveling wave, b2 , and reflected traveling wave, a2 , are related by
Eq. (8.6) [11, 33].
 
Ia (t) + j Qa (t) = (X + j Y ) Ib (t − τ − ) + j Qb (t − τ − ) (8.6)
8.5 Multi-tone Envelope Load-Pull Technique 207

Fig. 8.19 Reflected traveling wave (a2 ) and transmitted traveling wave (b2 ) and a clear depiction
of their synchronized envelopes [34], © IEEE 2008

Fig. 8.20 Measured six-tone

reflection coefficient over a
6 MHz bandwidth after delay
compensation in the
envelopes (i.e.,
synchronization of the
transmitted and reflected
traveling waves) [34],
© IEEE 2008

where  = T − τ is the additional delay provided by the delay compensator to

synchronize the envelope of the transmitted traveling wave, b2 , and the reflected
traveling wave, a2 ; and, T is the time period of the excitation signal.
This compensation technique works equally well with any excitation, as can be
seen in Fig. 8.19 [34]. It is evident that the transmitted and reflected traveling waves
are synchronized as a result of delay compensation in their envelopes. The stimulus
in this case was a six-tone signal with a tone separation of 1.2 MHz, and the target
reflection coefficient was 0.25∠225°. This enabled the synthesis of the constant
reflection coefficient (i.e., constant load impedance), as shown in Fig. 8.20, around
the 6 MHz bandwidth of the six-tone stimulus.
As an example, this setup finds immediate use in the investigation of the effect
of RF impedance variation around the fundamentals on PA distortion characteristics
[35]. Such an experiment allows for the determination of device behavior in terms
of linearity when subjected to variable impedances or constant impedance around
a specific bandwidth and also aids in the associated trade-off between linearity and
efficiency.
208 8 Advanced Configurations and Applications

8.6 Wideband Load-Pull Technique

The use of signals with high peak-to-average power ratios (PAPRs) for increased
spectral efficiency in modern wireless systems results in the requirement that PAs
have both good linearity and efficiency. Traditionally, class-AB (where all higher
harmonics are shorted) or inverse class-AB operation (where all higher harmonics
are open [49]) are used and operated in backoff, until an acceptable linearity perfor-
mance is obtained. However, this approach is suboptimal, as it results in poor design
for signals with high PAPRs.
Alternatively, linearization techniques that utilize out-of-band terminations at the
baseband and second harmonic frequencies are used to improve the in-band linearity
of transistor devices employed in the design of PAs [51, 52]. To evaluate device
technologies for their linearity, a dedicated large-signal characterization setup is
required that must provide:

• Calibrated testing of highly linear devices at all frequencies of interest (e.g., f0 ,

2f0 , 3f0 , and third- and fifth-order intermodulation frequencies).
• Flexible and independent control of the baseband, fundamental and second har-
monic terminations at both the DUT’s input and output over the full Smith chart.
• Constant impedance for all frequency components.

Unfortunately, commercially available passive load-pull systems [52, 53] cannot

control the baseband (BB) impedance and do not cover the complete Smith chart.
On the other hand, most of the active load-pull systems can cover the entire Smith
chart, but are not appropriate for linearity characterization with wideband signals,
due to the inherent electrical delay associated with these systems that causes spread
in the phase of synthesized reflection coefficients. To overcome these problems,
Spirito et al. [7] developed an active load-pull setup, which was further improved
by Marchetti et al. [9].
The setup possesses a high dynamic range and also provides a solution that over-
comes the losses, electrical delay, power handling and linearity limitations present
in traditional passive and active load-pull systems. The setup is based on wideband
data acquisition and wideband signal injection of the incident and device generated
power waves at the frequencies of interest.
By monitoring and controlling the spectral content of these waves in the fun-
damental and harmonic frequency bands, it is possible to synthesize user defined
reflection coefficients at the DUT’s reference planes. This allows emulation of any
matching network based on its frequency scattering parameters (S-parameters). Fur-
thermore, the monitoring and control of the spectral content of the waves at the
DUT’s reference planes enables the automatic compensation of the signal distor-
tion resulting from the loop amplifier nonlinearities [9]. In theory, the elimination
of amplifier nonlinearities and the electrical delay allows this setup to characterize
devices under wideband stimuli.
8.6 Wideband Load-Pull Technique 209

Fig. 8.21 Principle of the wideband open-loop active load-pull approach. [9], © IEEE 2008

8.6.1 Wideband Load-Pull Approach

As the reflection coefficients represent the ratio of waves, there is a requirement to

control the linear ratios of the incident and device generated waves of the DUT in
the BB, fundamental and harmonic frequency bands described in Fig. 8.21.
When the nonlinear DUT is excited with a user defined modulated signal, as ,
it generates signals in the baseband, fundamental and higher harmonic frequency
bands. By measuring the device’s generated waves (b1,n and b2,n ) as well as the in-
cident waves, the waves to be injected are estimated at every iteration. When the re-
quired reflection coefficient versus frequency (at every controlled band) is achieved,
the iteration has converged, and the large signal parameters (e.g., power-added effi-
ciency, output power, intermodulation distortion) are measured.
The only wave known prior to the iteration procedure is the source signal (as ),
whereas the device generated waves (b waves) are the result of the interaction of
as with the unknown nonlinear properties of the device. It is important to note that
the spectral content of the generated waves include not only the fundamental fre-
quency band, but also BB, higher order harmonics, and intermodulation distortion.
However, the accurate measurement of the spectral content of the device generates
waves along with the knowledge of user defined reflection coefficient and allows for
the estimate of the spectral content of all the injected waves (a waves) at the DUT’s
port. The expression in Eq. (8.7) relates the a and b waves and the user defined
reflection coefficient [9].
ax,n (fn ) = bx,n (fn )Γx,n (fn ) (8.7)
where x is an index to indicate source (s) or load (l); n indicates the frequency
band, e.g., BB (0), fundamental (1) and harmonic (2 and up); and, Γx,n (fn ) is the
210 8 Advanced Configurations and Applications

Fig. 8.22 Simplified block diagram of the wideband active harmonic load-pull system developed
by Marchetti et al. [9], © IEEE 2008

user defined reflection coefficient of the source or load versus frequency at the BB,
fundamental or harmonic frequency.
It is imperative to understand that, although this load-pull technique seems sim-
ple, in practice, it requires attention to a few issues. First of all, it requires very fast
data acquisition with high linearity and dynamic range, in order to handle the spec-
tral content of the complex modulated signals with their related distortion products.
Secondly, the a waves need to be generated with a high dynamic range and must be
optimized for their spectral content, in order to satisfy Eq. (8.7). Both these require-
ments place high demands on the capabilities of the hardware configuration, as well
as on the related software [9].

8.6.2 Setup Description

A simplified block diagram of the measurement setup is shown in Fig. 8.22. The
S-parameter test set, which is based on a five-coupler configuration [54], allows for
simultaneous measurement of the source, input and load reflection coefficients at
the DUT’s reference planes.
Wideband analog-to-digital (A/D) converters (or ADCs) with a sampling fre-
quency of 100 MHz are used to acquire the down-converted waveforms, facilitating
the measurement of the device reflection coefficients over a wide bandwidth in a
8.6 Wideband Load-Pull Technique 211

Fig. 8.23 Simplified block

diagram of the wideband
active loads with phase
coherent frequency
up-conversion [9], © IEEE
2008

single data acquisition. Custom bias tees with low inductance are placed directly
at the wafer probe, in order to minimize the electrical delay of the BB impedance.
They are implemented as a passive impedance switch bank [55]. The BB circuit also
includes the low-frequency test set for the calibrated BB impedance measurement.
The source signal and all injection signals needed to create the user defined
reflection coefficients at the DUT’s reference planes, originate from fully syn-
chronized (200 MS/s) AWGs having same the time base. Based on the phase
coherency requirement between fundamental and harmonic injection signals, the
in-phase/quadrature (IQ) up-conversion is preferred over digital intermediate fre-
quency (IF) techniques [56]. This allows for the use of a single local oscillator (LO)
to generate the high-frequency signals at fundamental and harmonic frequencies by
means of multipliers (e.g., 32 for the second harmonic) in the LO path, as depicted
in Fig. 8.23. This guarantees that the active loads and driving signal are phase co-
herent, since this LO does not need to be swept. Consequently, the source and all
injection signals are up-converted to the fundamental and harmonic frequencies and
fed to the DUT to establish the driving signal and reflection coefficients [9].
Another advantage of the IQ approach, compared to other known signal genera-
tion techniques, is the relative limited length of the data records needed to fulfill the
standard model requirements of complex modulated signals (e.g., WCDMA [57]),
yielding a significant speed advantage in practical measurement situations.
In Fig. 8.23, computer controlled attenuators and high power amplifiers are
placed in the signal path, in order to control the level the power of the injection
signals. This allows for the utilization of the maximum dynamic range of the AWGs
at all times, something that proves to be essential for meeting the spectral require-
ments of modern communication signals.
In this method, the accurate representation of the spectral content of the RF waves
at fundamental and harmonic frequencies can be obtained if the measured power
waves are down-converted to a lower frequency prior to data acquisition. This is
also a common technique used in vector network analyzers (VNAs) and allows the
highest possible dynamic range achievable.
For modulated signals, it is very important that the detection path is free from
nonlinear errors that cannot be corrected by linear calibration techniques. The de-
tection dynamic range of the harmonic frequency components are optimized through
the use of power splitters and high-pass filters at the detection ports of the couplers
in the input and output sections, as shown in Fig. 8.22 [7, 9].
212 8 Advanced Configurations and Applications

By high-pass filtering the higher harmonic components, the mixer used for the
down-conversion in the signal path is protected from the high power of the fun-
damental signal. This drastically relaxes the mixer linearity requirements for the
second and higher harmonics, improving the quality of the acquired signals. By
stepping the LO frequency, the frequency band centered on the harmonic of interest
is down-converted to its low IF representation for data acquisition.
The detection dynamic range of the system in the fundamental frequency band
is maximized by using variable attenuators in the RF path and high-power high-
linearity mixers. Note that DUT nonlinearities also give rise to BB signals, which
do not need frequency conversion prior to the data acquisition. The resulting low IF
signals and BB signals from the DUT are fed to high-speed multiplexing switches
to reduce the required number of high-performance wideband ADCs [9].
Finally, the system calibration can be carried out using the combination of the
techniques described in [54, 55, 58]. In the first step, the combination of standards
at the source and DUT input reference planes allows for simultaneous measurement
of the source and DUT input reflection coefficients [54]. At the same time, use of
an open-short-load (OSL) standard at the DUT’s input and output reference planes
enables the calibrated measurement of the BB impedance [55].
In the second step, use of a OSL standard at the load reference plane, when
a thru line is in place of a DUT, allows for the measurement of the DUT’s load
reflection coefficient. Finally, connection of power meter at the load reference plane
aids in absolute power calibration [58]. IQ modulators can be calibrated for leakage
minimization through the use of DC offsets, while balance and quadrature errors
can be corrected through digital pre-compensation [59].

8.7 Noise Characterization

Noise parameter determination methods have evolved in order to cater to the require-
ments of emerging low-noise transistor technology. In principle, noise characteriza-
tion always requires a highly repeatable electronic or automated mechanical tuner
along with a VNA. Ghannouchi et al. [37] developed a noise characterization setup
that removes the need of a repeatable impedance tuner and discards the requirement
for a VNA. It employs a reverse six-port (SP) reflectometer and a standard spectrum
analyzer and.

8.7.1 Noise Parameter Measurement

The measurement of the noise parameter consists of measuring the noise factor of
the DUT for several different noise source impedances. The noise factor of a net-
work, as defined by Friis, is the ratio of the input signal-to-noise ratio (SNR) to the
output SNR. The output signal is equal to the input signal multiplied by the gain of
8.7 Noise Characterization 213

Fig. 8.24 Block diagram of

source-pull setup in a reverse
six-port configuration for the
measurement of noise source
impedance [37], © IEEE
1995

the device; therefore, the noise factor (F) can be expressed as per Eq. (8.8) and the
noise figure (NF) can be expressed as per Eq. (8.9).
Nout
F= (8.8)
G.Nin
N F = 10 log(F ) (8.9)
with Nin = kT0 B and Nout = GkT0 B + Nadd , where k is Boltzman’s constant, and
B is the measurement bandwidth in hertz.
In a 1 Hz bandwidth for a reference temperature, T0 , defined as 290 K, the refer-
ence noise level, kT0 B, is equal to −174 dBm. Nadd represents the noise contribu-
tion of the DUT and is independent of the noise power injected to the input of the
DUT. When measuring the noise factor with an available input noise, Nin , different
from −174 dBm, the obtained noise factor given by Eq. (8.8) is wrong. However,
for a known input noise power level, appropriate corrections help in reaching the
true noise factor of the device [37].
The measurement of the noise figure using a non-repeatable tuner experiences
two critical issues. The first relates to the determination of an accurate impedance
of a noise source for each position of the uncalibrated tuner, while the other corre-
sponds to the accurate evaluation of the available noise power at the input of DUT
associated with the impedance of a noise source. A SP reflectometer in reverse con-
figuration, as shown in Fig. 8.24, can measure its own test port measurement (Γ2 )
as described in Chap. 4.
Experimental verification shows that the reflection coefficient (Γ2 ) measured by
a reverse SP reflectometer is identical to that measured by a VNA connected at
plane 2 in Fig. 8.24. The advantage of the reverse SP configuration is that it can
measure the source impedance’s reflection coefficient (Γ2 ) at plane 2, under any
experimental condition. As a consequence, the repeatability and calibration of the
tuner are not required. For a transparent SP reflectometer (i.e., S11 ≈ S22 ≈ 0 and
S12 ≈ S21 ≈ 1), the tuner and SP junction can cover a wide area of the Smith chart
[38]. Noise can be injected to the DUT by switching alternatively between the signal
generator and a commercial 50  noise source.
214 8 Advanced Configurations and Applications

The available signal or noise power at plane 2 can be determined once the reflec-
tion coefficient (Γ2 ) and the scattering matrix (S[3 × 3]) of the directional coupler
are known. Scattering matrix S[3 × 3] can be reduced to matrix S  [2 × 2] in such
a way that its entries are functions of the reflection coefficient (Γ3 ) at plane 3 in
Fig. 8.24 [37]. This reflection coefficient (Γ3 ) can be obtained from the measured
reflection coefficient (Γ2 ) and the scattering matrix S[3 × 3] of the coupler. Matrix
S  [2 × 2] represents the tuner, SP and coupler as an equivalent two-port network de-
fined between planes 1 and 2. The available power (P1 ) of the generator connected
at plane 1 is related to the available power at plane 2 (P2 ) by the expression given
in Eq. (8.10) [37]. The term G21 in this expression represents the available power
gain in the two-port network between planes 1 and 2.
1 − |Γ1 |2    1
P2 = G21 P1 =  S21 P1 (8.10)
|1 − S11 Γ1 |2 1 − |Γ2 |2
where Γ1 is the reflection coefficient of the generator connected at plane 1, Γ2 is
measured by the reverse SP junction at plane 2, and:
 S13 S31
S11 = S11 +
(1/Γ3 ) − S33
 S23 S32
S21 = S21 +
(1/Γ3 ) − S33
Γ2 − S22
Γ3 =
S23 S32 + (Γ2 − S22 )S33
The available noise power at plane 2 can be expressed as a function of the compo-
nent noise temperature given in Eq. (8.11). It is assumed that the noise contribution
from the directional coupler, SP junction and tuner are maintained at the same con-
stant physical temperature (Tc ).
 
N2 = TS α + (1 − α)Tc kB (8.11)
where TS is the noise temperature of the noise source with kTS B = N1 ; and, α is
the coupling loss between planes 1 and 2 and is equal to available gain G21 given
by Eq. (8.10).
As the power available at the input of the DUT is not at −174 dBm, the noise
factor cannot be determined from Eq. (8.8). Instead, a comparison of the measured
output noise (Nout ) and the calculated input noise (N2 GDUT ) yields the noise con-
tribution, Nadd , of the DUT. GDUT represents the gain of the DUT.
Nadd = (Nout )measured − N2 GDUT (8.12)
Equations (8.8), (8.9) and (8.12) can be simplified at reference temperature T0 to
obtain the expression for the noise figure (NF) of the DUT as:
 
Nout + Nadd
NFDUT = 10 log
GDUT kT0 B
 
(Nout )measured + GDUT (kT0 B − N2 )
= 10 log (8.13)
GDUT kT0 B
where Nout = GDUT Nin = GDUT kT0 B.
8.7 Noise Characterization 215

Fig. 8.25 Noise parameter

test setup using an
uncalibrated tuner [37],
© IEEE 1995

Equation (8.13) requires measurements at only one noise temperature; therefore,

a noise receiver with high measurement accuracy is needed. In the absence of such a
quality noise receiver, an alternative approach is to use Y factor for the noise figure
calculation, which given in Eq. (8.14).
( TTH0 − 1) − Y ( TTC0 − 1)
N FDUT = 10 log (8.14)
Y −1
TH and TC are deduced from N2 , and Y is deduced from Nout as follows:
N2 (H ) N2 (C) Nout (H )
TH = ; TC = ; Y= (8.15)
kB kB Nout (C)

8.7.2 Noise Parameter Test Setup

Dedicated noise receivers are the most suitable instrument for measuring noise, but
they often require costly converters for high-frequency measurements. Furthermore,
noise receivers use a fixed measurement bandwidth of a few MHz; and, this may
become a problem when measuring narrowband devices. In such circumstances, a
multipurpose spectrum analyzer offers variable measurement bandwidth and per-
forms noise measurements with acceptable accuracy. Figure 8.25 shows a block
diagram of an experimental setup that utilizes a multipurpose spectrum analyzer in
the noise characterization of microwave transistors.
In this setup, a spectrum analyzer, with a 2 MHz resolution bandwidth centered
on the test frequency, measures the available noise power. The noise level is rel-
atively low at the DUT’s output; therefore, a low-noise pre-amplifier is necessary
to increase the sensitivity of the spectrum analyzer. The isolator, with |S12 | smaller
than −30 dB, helps in maintaining a constant noise figure of the noise receiver (com-
prising of the isolator, pre-amplifier, and spectrum analyzer shown in Fig. 8.26).
A 2–8 GHz transparent SP reflectometer in reverse configuration measures the
source reflection coefficient, Γmeas , presented by the non-repeatable mechanical stub
216 8 Advanced Configurations and Applications

Fig. 8.26 Model of spectrum

analyzer based noise receiver
[37], © IEEE 1995

tuner seen at the reference plane. Power at the four detection ports of the SP junction
is measured using a four-channel power meter. This task can also be performed by
the spectrum analyzer with computer-controlled switches. The losses in the SP junc-
tion, 10 dB directional coupler and test fixture limit the source impedance coverage
up to a radius circle of 0.75 on the Smith chart, but this is sufficient for noise pa-
rameter measurement. Microwave signals and white noise are injected alternatively
into the SP reflectometer and into the DUT using a PC-controlled switch.
To determine the gain and noise figure of the noise receiver, a conventional cal-
ibration technique consists of measuring two noise standards, which correspond to
the hot and cold states of the noise source. An alternative technique for spectrum
analyzer based noise receivers, as shown in Fig. 8.26, discards the use of noise stan-
dards [39]. However, in this technique, the gain and noise figure of the pre-amplifier
must be known; and, if the input mismatch factor is taken into account the noise
power available at the reference plane is given by Eq. (8.16) [37].
 
(PSA )measured |1 − ΓR Γout |2
NReceiver = − Ni (8.16)
Gamplifier 1 − |Γout |2
where ΓR and Γout are the reflection coefficients at the input of the noise receiver
and output of the DUT, respectively; and, Ni is the noise contribution of the receiver
given by:
 
Ni = 100.1N FR (dB) − 1 kT0 B (8.17)
The power measured by the spectrum analyzer, (PSA )measured , and Eq. (8.15) can
then be utilized to evaluate the noise at plane 2 (N2 ) given in Eq. (8.11) as detailed
in the following paragraph.
The available noise power of the noise source connected at plane 1 is pre-
measured (see Fig. 8.24). The available noise power at plane 2 is then calculated
as a function of the tuner position during measurement according to Eqs. (8.10) and
(8.11) and measured using the receiver according to Eq. (8.16). As an example, Ta-
ble 8.3 lists the measured and calculated values for eight different tuner settings for
an experimented carried out on a GaAs MESFET device operated in low noise am-
plifier mode. Both the values agree well with only a slight discrepancy that can be
8.8 Mixer Characterization 217

Table 8.3 Comparison

between the calculated and ΓS Calculated Measured
measured noise power [37], Amplitude/phase dBm dBm
© IEEE 1995
0.6710/ −102.5° −73.82 −73.42
0.6739/ +152.2° −73.12 −72.91
0.4308/ +161.0° −75.05 −74.45
0.7433/ −131.9° −72.88 −72.62
0.1293/ +152.9° −75.92 −75.72
0.1722/ −155.8° −75.76 −75.69
0.2236/ +088.8° −75.64 −75.54
0.0966/ −012.2° −75.97 −75.82

attributed to the measurement uncertainty of the SP reflectometer and noise receiver.

This offset in the measured results can be reduced by taking measurements at two
different noise temperatures.

8.8 Mixer Characterization

In-band intermodulation (IBIM) is an essential characteristic of microwave/milli-
meter-wave front-end mixers. In modern wireless systems, IBIM products gener-
ated by the mixer can significantly deteriorate the main IF signal. The field effect
transistor (FET) resistive mixers [40] substantially overcome this problem, as they
exhibit good performance, in terms of conversion loss, higher operating frequency,
and a low level of intermodulation products [41–44]. These performances are usu-
ally evaluated using setups that are complex and difficult to operate [5, 13, 45].
Ghannouchi et al. [8], therefore, developed a mixer characterization setup that is
simple to operate and enables measurement and design of wide dynamic range (lin-
ear) FET resistive mixers.

8.8.1 Measurement Setup

The mixer characterization setup, given in Fig. 8.27, makes use of an active source-
pull measurement (ASPM) system [46], as shown in Fig. 8.28. In the ASPM system,
a part of the input signal that is well controlled in phase and magnitude is injected
as a reflected wave to synthesize the reflection coefficient, ΓS , at the DUT port.
The advantage of ASPM over the traditional passive tuner lies in the fact that a truly
unitary ΓS , which is often needed for mixer characterization, can be generated using
ASPM, whereas a passive tuner fails to do so.
Figure 8.27, which depicts a multi-frequency test setup, is a dual SP network
analyzer appropriately modified to simultaneously carry out ASPM and load-pull
218 8 Advanced Configurations and Applications

Fig. 8.27 Schematic diagram of multi-frequency test setup for characterization of mixers [8],
© IEEE 1998

Fig. 8.28 Schematic diagram

of an active source-pull
measurement (ASPM) system
based on SP reverse
configuration [8], © IEEE
1998

measurements on a FET resistive mixer. The test setup is built such that the LO and
RF signals are received by the gate and drain, respectively. The IF signal gets ex-
tracted from the drain. It is imperative to note that the drain is not biased; therefore,
the transistor works as a time-varying resistance controlled by the LO signal.
The power at four detection side arms of SP#1 is measured by a four-channel
power meter. A computer-controlled SP4T (single pole, four through) switch allows
successive measurements of the powers at the four detection ports of SP#2. Since
many frequencies are simultaneously present at the output, there are two possible
ways to take power readings at the ports of SP#2: either by a power probe along
with a computer-controlled yttrium iron garnet (YIG) filter or by a pre-calibrated
spectrum analyzer controlled via an IEEE-488 bus. The measurement by the first
technique has advantages in terms of accuracy and speed, but suffers from three
following problems:
8.8 Mixer Characterization 219

• The stability regulation of the computer-controlled YIG filter versus the varia-
tions in the temperature and polarized current, although possible, is tedious and
complex.
• In a two-tone test, the third- and fifth-order IBIM products remain inside the
30 MHz pass band of the YIG filter and, therefore, cause problems in measure-
ments.
• The characterization of low-power nonlinear devices, such as FET resistive mix-
ers, often requires detection of signals below −50 dBm, which can be over-
whelmed by the noise of the YIG filter and power sensor. The dynamic range
of detection (i.e., the variation of power at each detection port of SP#2) with a
power meter and YIG filter may be as low as 15 dB, which is insufficient for
practical needs.
The pre-calibrated spectrum analyzer (SA#1 in Fig. 8.27) can circumvent these
problems. A very weak signal becomes measurable when a high-resolution band-
width is used. The dynamic range of SP#2 can be increased to 60 dB, which enables
measurements over the frequencies of the intermodulation products of interest. In
addition, the presence of a spectrum analyzer in SP#2 allows for observation of any
higher order products of the LO and RF signals (spurious responses). Furthermore,
it has been demonstrated that an appropriate setting of the spectrum analyzer could
give both reasonable accuracy and speed of measurement [39].
SP#1 performs LO active source-pulling at the transistor gate, while SP#2 simul-
taneously holds ASPM at the input RF frequency and active load-pull measurements
at the output LO, 2LO and IF frequencies, both at the drain of the transistor. The
power available from the source and the power absorbed by the load can be evaluated
by SP#1 and SP#2, respectively, according to the expressions given in Eqs. (8.18)
and (8.19) [37, 48].
1 − |ΓS |2 1
(P )SP#1 = S21 PSource (8.18)
|1 − S11 ΓS |2 |1 − ΓS |2
k(PRef -port )SP#2
(PAbsored )SP#1 = (8.19)
|1 − cΓL |2
where Sij are the S-parameters of the two-port network defined between the gener-
ator and the measurement plane of SP#1. ΓS and ΓL are the reflection coefficients
measured by SP#1 and SP#2, respectively, k is a real power calibration constant,
and c is the complex error box constant.
A triplexer separates the LO, RF and IF frequencies into three branches, while
the 2LO active load-pull branch is realized by injecting the 2LO signal via a wide-
band 16 dB directional coupler. The IF active load-pulling is performed by a closed
loop, in order to ensure the frequency coherence of the IF re-injected signal. The in-
sertion of a computer-controlled YIG filter in this closed loop eliminates any risk of
oscillation. In practice, a quasi-unitary reflection coefficient can be emulated at the
DUT plane by increasing the synthesized signals to compensate for the combined
insertion loss of the triplexer, SP junction, 16 dB directional couplers, bias tee, and
test fixture.
220 8 Advanced Configurations and Applications

The whole system is PC controlled via a HPIB controller. The YIG filter and
SP4T switch are controlled directly by a digital-to-analog 16-channel converter
card. This setup can be configured to provide fully automatic measurements by re-
placing the manual phase shifters and variable attenuators with electronic vector
modulators.

8.8.2 Experimental Procedure

To carry out experiments, SP junctions SP#1 and SP#2 need to be calibrated us-
ing any appropriate technique, such as the conventional six-to-four-port reduction
technique [47]. For a transistor mounted on a microstrip line test fixture, a thru-
reflection-line (TRL) method can be employed to de-embed the measurements at
the input and output (I/O) planes.
Since only the LO signal is injected into the gate and several signals are present
at the drain of the transistor, it is necessary to calibrate the SP junctions for a set
of frequencies. For example, during the design of a C-band mixer, the required
frequencies are SP#1 at 2.225 GHz (LO frequency), SP#2 at 2.225 GHz, SP#2 at
3.600–3.900–4.200 GHz (IF frequencies), SP#2 at 4.450 GHz (2LO frequency), snf
SP#2 at 5.825–6.125–6.425 GHz (RF frequencies) [8].
The setup contains two 2–18 GHz signal generators, which provide +15 dBm
and −4 dBm powers for the respective LO and RF signals, which need to be mea-
sured when the source impedance is set to 50 , in order to be matched to the 50 
power probe. A quick measurement can then be performed to verify that the mea-
sured reflection coefficient of a fixed impedance remains quasi-constant when the
source powers (LO and RF) are swept from −35 dBm to their maximum available
values, in order to identify the dynamic range of the setup.
In the design and characterization of mixers, usually the target is the optimization
of the linearity as a function of termination impedances, while maintaining a typical
conversion loss. The overall measurement involves a series of measurements that
involves sweeping of different parameters with the most important parameters first
and the less significant ones last. The optimization starts with a biasing point and
input power levels (LO and RF), followed by LO and RF source impedances, and
finally by the load impedances. These load impedances, in turn, are in order of
importance: the load at the IF, the load at the LO, and finally, at the second harmonic
2LO frequencies of the output signal.
Once the effects of the termination impedances at a single-tone RF excitation are
known, a two-tone RF excitation can then be applied to determine the IBIM perfor-
mance of the transistor. However, it is imperative to understand that there is a need
of some hardware modifications to adapt the test set to two-tone measurements, as
shown in Fig. 8.29. First, the second harmonic 2LO load-pull branch is replaced by
a second spectrum analyzer, SA#2. It is connected to the DUT output via a 16 dB di-
rectional coupler and detects a higher magnitude signal than that detected by SA#1.
In two-tone measurements, the terminating impedances are fixed as (ZS )RF1 ∼ =
(ZS )RF2 and (ZS )IF1 ∼ = (ZS )IF2 obtained in single-tone characterization, since the
References 221

Fig. 8.29 Modified setup for two-tone measurements [8], © IEEE 1998

variation in frequency is very small. In theory, the active loop can still be used
for the synthesis of IF impedances, but this setup (Fig. 8.29) utilizing mechanical
tuners is simpler and sufficient for measurement requirements, considering that the
range of optimal IF impedances are now known from the single-tone measurements.
A detailed description of a mixer characterization and design example is available
in [8].

References

1. M.J. Pelk, L.C.N. de Vreede, M. Spirito, J.H. Jos, Base-band impedance control and calibra-
tion for on-wafer linearity measurements, in 63rd ARFTG Conference (June 1994), pp. 35–40
2. H. Arthaber, M.L. Mayer, G. Mageri, An active load-pull setup for broadband signals us-
ing digital baseband processing for the active loop. Int. J. RF Microw. Computer-Aided Eng.
18(6), 574–581 (2008)
3. M.S. Hashmi, J. Benedikt, P.J. Tasker, Impact of group delay on the load emulation in multi-
tone active envelope load-pull system. Aust. J. Electr. Electron. Eng. 7(1), 83–88 (2010)
4. R. Hajji, F. Beauregard, F.M. Ghannouchi, Multi-tone power intermodulation load-pull char-
acterization of microwave transistors suitable for linear SSPA’s design. IEEE Trans. Microw.
Theory Tech. 45(7), 1093–1099 (1997)
5. F.M. Ghannouchi, G. Zhao, F. Beauregard, Simultaneous load-pull of intermodulation and out-
put power under two-tone excitation for accurate SSPAs design. IEEE Trans. Microw. Theory
Tech. 42(6), 929–934 (1994)
6. P. Ghanipour, S. Stapleton, J. Kim, Load-pull characterization using different digitally modu-
lated stimuli. IEEE Microw. Wirel. Compon. Lett. 17(5), 400–402 (2007)
7. M. Spirito, M.J. Pelk, F. van Rijs, S.J.C.H. Theeuwen, D. Hartskeeri, L.C.N. de Vreede, Ac-
tive harmonic load-pull for on-wafer out-of-band device linearity optimization. IEEE Trans.
Microw. Theory Tech. 54(12), 4225–4235 (2006)
222 8 Advanced Configurations and Applications

8. D. Le, F.M. Ghannouchi, Multitone characterization and design of FET resistive mixers based
on combined active source-pull/load-pull techniques. IEEE Trans. Microw. Theory Tech.
46(9), 1201–1208 (1998)
9. M. Marchetti, M.J. Pelk, K. Buisman, W. Neo, M. Spirito, L. de Vreede, Active harmonic
load-pull with realistic wideband communications signals. IEEE Trans. Microw. Theory Tech.
56(12), 2979–2988 (2008)
10. B. Noori, P. Hart, J. Wood, P.H. Aaen, M. Guyonnet, M. Lefevre, J.A. Pla, J. Jones, Load-
pull measurements using modulated signals, in 36th European Microwave Conference, Paris,
France (2005), pp. 1594–1597
11. M.S. Hashmi, S.J. Hashim, S. Woodington, T. Williams, J. Benedikt, P.J. Tasker, Active en-
velope load pull system suitable for high modulation rate multi-tone applications, in 3rd In-
ternational Microwave Monolithic Integrated Circuit Workshop, Malaga, Spain (Nov. 2008),
pp. 93–96
12. Maury Microwave Corporation, Active harmonic load-pull with realistic wideband communi-
cations signals, Application Note 5A-044, June 2010
13. C. Tsironis, Two-tone intermodulation measurements using computer controlled microwave
tuner. Microw. J. 32, 161 (1989)
14. J. Jacobi, IMD still unclear after 20 years, in Microwaves & RF (Nov. 1986), pp. 119–126
15. J.G. Freed, Equations provide accurate third-order IMD analysis, in Microwaves & RF (Aug.
1992), pp. 75–84
16. M. Leffel, Intermodulation distortion in a multi-signal environment, in RF Design (June 1995),
pp. 78–83
17. R. Hajji, F. Beauregard, F.M. Ghannouchi, Multi-tone characterization for intermodulation
and distortion analysis, in IEEE Microwave Theory and Techniques Society’s International
Microwave Symposium Digest, San Francisco, CA (June 1996), pp. 1691–1694
18. L. Cellai, A. Greco, Optimal phase relationship aids IM testing, in Microwaves & RF (July
1996), pp. 56–61
19. M.S. Hashmi, F.M. Ghannouchi, P.J. Tasker, K. Rawat, Highly reflective load-pull. IEEE Mi-
crow. Mag. 12(4), 96–107 (2011)
20. G.P. Locatelli, De-embedding techniques for device characterization. Alta Freq. 57(5), 267–
272 (1988)
21. M.A. Maury Jr., S.L. Mar, G.R. Simpson, TRL calibration of vector automatic network ana-
lyzers. Microw. J. 30(5), 382–387 (1987)
22. K.S. Kundert, A.S. Vincentelli, Simulation of nonlinear circuits in the frequency domain. IEEE
Trans. Computer-Aided Des. 5, 521–535 (1986)
23. W.R. Curtice, GaAs MESFET modeling and nonlinear CAD. IEEE Trans. Microw. Theory
Tech. 36, 220–230 (1988)
24. X. Cui, S.J. Doo, P. Roblin, G.H. Jessen, J. Strahler, Real-time active load-pull of the 2nd and
3rd harmonics for interactive design of non-linear power amplifiers, in 68th ARFTG Confer-
ence, Colorado, USA (Nov. 2006), pp. 29–42
25. G.P. Bava, U. Pisani, V. Pozzolo, Active load technique for load-pull characterization at mi-
crowave frequencies. IET Electron. Lett. 18(4), 178–180 (1982)
26. Y. Takayama, A new load-pull characterisation method for microwave power transistors, in
IEEE Microwave Theory and Techniques Society’s International Microwave Symposium Di-
gest (1976), pp. 218–220
27. F. Verbeyst, M.V. Bossche, Real-time and optimal PA characterization speeds up PA design,
in 34th European Microwave Conference, Amsterdam, Netherlands (2004), pp. 431–434
28. P. Roblin, Seok Joo Doo, Xian Cui, G.H. Jessen, D. Chaillot, J. Strahler, New ultra-fast real-
time active load-pull measurements for high speed RF power amplifier design, in IEEE Mi-
crowave Theory and Techniques Society’s International Microwave Symposium Digest, Hon-
olulu, USA (June 2007), pp. 1493–1496
29. J. Sevic, M.B. Steer, A.M. Pavio, Large-signal automated load-pull of adjacent-channel power
ratio for digital wireless communication systems, in IEEE Microwave Theory and Techniques
Society’s International Microwave Symposium Digest (1996), pp. 763–766
References 223

30. C. Clark, Time-domain envelope measurement technique with application to wideband power
amplifier modelling. IEEE Trans. Microw. Theory Tech. 46(12), 2351–2540 (1998)
31. O. Väänänen, J. Vankka, K. Halonen, Effect of Clipping in Wideband CDMA System and
Simple Algorithm for Peak Windowing (Helsinki University of Technology, Helsinki, 2002)
32. D.E. Stoneking, R.J. Trew, J.B. Yan, Load pull characteristics of GaAs MESFETs calculated
using an analytic, physics based large signal device model, in IEEE Microwave Theory and
Techniques Society’s International Microwave Symposium Digest (1998), pp. 1057–1060
33. M.S. Hashmi et al., Active envelope load pull solution addressing the RF device multi-tone
characterization problem, in IEEE MTT-11 Design Competition, Atlanta, USA (June 2008).
http://www.mtt.org/mtt11/ims_award.htm
34. S.J. Hashim, M.S. Hashmi, T. Williams, S. Woodington, J. Benedikt, P.J. Tasker, Active enve-
lope load-pull for wide-band multi-tone stimulus incorporating delay compensation, in 34th
European Microwave Conference, Amsterdam, Netherlands (2008), pp. 317–320
35. S.J. Hashim, M.S. Hashmi, J. Benedikt, P.J. Tasker, Effect of impedance variation around the
fundamental on pa distortions characteristics under wideband stimuli, in IEEE Asia Pacific
Conference on Circuits and Systems, Kuala Lumpur, Malaysia (Dec. 2010), pp. 1115–1118
36. T. Williams, O. Mojon, S. Woodington, J. Lees, M.F. Barciela, J. Benedikt, P.J. Tasker, A ro-
bust approach for comparison and validation of large signal measurement systems, in IEEE
Microwave Theory and Techniques Society’s International Microwave Symposium Digest, At-
lanta, USA (June 2008), pp. 257–260
37. D. Le, F.M. Ghannouchi, Noise measurements of microwave transistors using an uncalibrated
mechanical stub tuner and a built-in reverse six-port reflectometer. IEEE Trans. Instrum. Meas.
44(4), 847–852 (1995)
38. F.M. Ghannouchi, R.G. Bosisio, Source-pull/load-pull oscillator measurements at mi-
crowave/MMwave frequencies. IEEE Trans. Instrum. Meas. 41(1), 32–35 (1992)
39. B. Peterson, Spectrum analysis basics, Hewlett-Packard Application Note 150, p. 2634, and
pp. 38–40
40. S.A. Maas, A GaAs MESFET mixer with very low intermodulation. IEEE Trans. Microw.
Theory Tech. 35, 429–435 (1987)
41. H.H.G. Zirath, C.-Y. Chi, N. Rorsman, G.M. Rebeiz, A 40 GHz integrated quasi-optical slot
HFET mixer. IEEE Trans. Microw. Theory Tech. 42, 2492–2497 (1994)
42. K. Yhland, N. Rorsman, H.H.G. Zirath, Novel single device balanced resistive HEMT mixers.
IEEE Trans. Microw. Theory Tech. 43, 2863–2867 (1995)
43. R.S. Virk, S.A. Maas, Modeling MESFET for intermodulation analysis of resistive FET mix-
ers. in IEEE Microwave Theory and Techniques Society’s International Microwave Sympo-
sium Digest, Orlando, USA, vol. 3 (May 1995), pp. 1247–1250
44. E.W. Lin, W.H. Ku, Device considerations and modeling for the design of an INP-based MOD-
FET millimeter-wave resistive mixer with superior conversion efficiency. IEEE Trans. Microw.
Theory Tech. 43(Aug.), 1951–1959 (1995)
45. L. Ricco, G.P. Locatelli, F. Calzavara, Constant intermodulation Loci measure for power de-
vices using HP-8510 network analyzer, in IEEE Microwave Theory and Techniques Society’s
International Microwave Symposium Digest, New York, USA (May 1988), pp. 221–224
46. D. Le, P. Poire, F.M. Ghannouchi, Six-port based active source-pull measurement technique.
IOP Meas. Sci. Technol. 9, 1336–1342 (1998)
47. J.D. Hunter, P.I. Somlo, Explicit six-port calibration method using five standards. IEEE Trans.
Microw. Theory Tech. MTT-39, 69–72 (1985)
48. F.M. Ghannouchi, R. Larose, R.G. Bosisio, A new multiharmonic loading method for large
signal microwave transistor characterization. IEEE Trans. Microw. Theory Tech. 39(June),
986–992 (1991)
49. F. van Rijs, R. Dekker, H.A. Visser, H.G.A. Huizing, D. Hartskeer, P.H.C. Magnee, R. Don-
dero, Influence of output impedance on power added efficiency of si-bipolar power transistors,
in IEEE Microwave Theory and Techniques Society’s International Microwave Symposium
Digest, Boston, USA (June 2000), pp. 1945–1948
224 8 Advanced Configurations and Applications

50. S. Liwei, L.E. Larson, An Si-SiGe BiCMOS direct-conversion mixer with second-order and
third-order nonlinearity cancellation for WCDMA applications. IEEE Trans. Microw. Theory
Tech. 51(11), 2211–2220 (2003)
51. V. Aparin, C. Persico, Effect of out-of-band terminations on intermodulation distortion in
common-emitter circuits, in IEEE Microwave Theory and Techniques Society’s International
Microwave Symposium Digest, Anaheim, USA (June 1999), pp. 977–980
52. Maury Microwave Corporation, http://www.maurymw.com/
53. Focus Microwaves, http://www.focus-microwaves.com/
54. G.L. Madonna, M. Pirola, A. Ferrero, U. Pisani, Testing microwave devices under different
source impedance values—a novel technique for on-line measurement of source and device
reflection coefficients. IEEE Trans. Instrum. Meas. 49(2), 285–289 (2000)
55. M.J. Pelk, L.C.N. de Vreede, M. Spirito, J.H. Jos, Base-band impedance control and cali-
bration for on-wafer linearity measurements, in 63rd ARFTG Conference, Forth Worth, USA
(June 2004), pp. 35–40
56. W.C.E. Neo, J. Qureshi, M.J. Pelk, J.R. Gajadharsing, L.C.N. de Vreede, A mixed-signal
approach towards linear and efficient N-way Doherty amplifiers. IEEE Trans. Microw. Theory
Tech. 55(5), 866–879 (2007)
57. 3G TS 25.141 base station conformance testing (FDD), Tech. specification group radio access
networks, 3rd generation partnership project, Valbonne, France, Tech. Spec., Rev. V3.1.0,
2000
58. A. Ferrero, U. Pisani, An improved calibration technique for on-wafer large-signal transistor
characterization. IEEE Trans. Instrum. Meas. 42(2), 360–364 (1993)
59. R. Marchesani, Digital precompensation of imperfections in quadrature modulators. IEEE
Trans. Commun. 48(4), 552–556 (2000)
Authors

Fadhel M. Ghannouchi is professor and AITF/CRC Chair in the Department of

Electrical and Computer Engineering, Schulich School of Engineering, University
of Calgary, Canada, and Director of the Intelligent RF Radio Laboratory. He has
held numerous invited positions at several academic and research institutions in
Europe, North America and Japan. He has provided consulting services to a number
of microwave and wireless communications companies. His research interests are
in the areas of microwave instrumentation and measurements, nonlinear modeling
of microwave devices and communications systems, design of power and spectrum
efficient microwave amplification systems and design of intelligent RF transceivers
and software-defined radio systems for wireless and satellite communications. His
research activities have led to over 500 publications and 14 US patents (6 pending)
and two books. He is Fellow of IEEE and he has been a distinguished microwave
lecturer of IEEE MTT-S since 2009.
Mohammad S. Hashmi received MS degree from Darmstadt University of
Technology, Germany and PhD degree from Cardiff University, UK. He is now an
adjunct researcher at the iRadio Lab, University of Calgary, Canada and Assistant
Professor at IIIT Delhi, India. He was previously associated with Philips Semicon-
ductors and Thales Electronics in Germany during which time he was involved in the
field of RF circuits and systems. His current research interests are related to nonlin-
ear microwave instrumentation, microwave device characterization, and lineariza-
tion of power amplifiers for mobile and satellite applications. He was the recipient
of 2008 Automatic Radio Frequency Techniques Group (ARFTG) Microwave Mea-
surement Fellowship, and 3rd place winner in the novel and creative instrument
design competition organized by IEEE MTT-11 for the year 2008. His research has
led to over 40 publications and 3 US patents (pending).

F.M. Ghannouchi, M.S. Hashmi, Load-Pull Techniques with Applications to Power 225
Amplifier Design, Springer Series in Advanced Microelectronics 32
DOI 10.1007/978-94-007-4461-5, © Springer Science+Business Media Dordrecht 2013
About the Book

This first book on load-pull systems is intended for readers with a broad knowledge
of high frequency transistor device characterization, nonlinear and linear microwave
measurements, RF power amplifiers and transmitters. It fulfills the demands of
users, designers, and researchers both from industry and academia who have felt
the need of a book on this topic. It presents a comprehensive reference spanning
different load-pull measurement systems, waveform measurement and engineering
systems, and associated calibration procedures for accurate large signal character-
ization. Besides, this book also provides in-depth practical considerations required
in the realization and usage of load-pull and waveform engineering systems. In ad-
dition, it also provides procedure to design application specific load-pull setup and
includes several case studies where the user can customize architecture of load-pull
setups to meet any specific measurement requirements. Furthermore, the materials
covered in this book can be part of a full semester graduate course on microwave
device characterization and power amplifier design.

F.M. Ghannouchi, M.S. Hashmi, Load-Pull Techniques with Applications to Power 227
Amplifier Design, Springer Series in Advanced Microelectronics 32
DOI 10.1007/978-94-007-4461-5, © Springer Science+Business Media Dordrecht 2013
Index

0-9 C
1-port error flow model, 147, 152 CAD-based PA Design, 14, 15
CAD incorporation, 185
A Calibration, vii, viii, 21, 29, 35, 37, 38, 40–42,
ABCD Parameter, 136 55, 87–93, 95, 97, 99–101, 105–107,
Accurate calibration, 142 113, 122, 123, 126, 129, 133, 135, 139,
Active Impedance Tuner, 17 142, 145–156, 163, 164, 166–168,
Active load-pull, viii, 18, 20, 22, 55–62, 64, 170–175, 181–183, 193, 197, 211–213,
67, 69–71, 75, 76, 78, 83, 96, 98, 99, 216, 219, 227
106, 107, 110, 113, 114, 117–119, Cascaded Tuner, 49, 51
125–129, 132, 133, 139, 140, 152, 154, CDMA, 1
156, 158, 159, 198, 208, 209, 219 Characteristic impedance, 17, 32, 55, 94, 104,
Active Source-Pull, 95 115, 118, 130, 135, 177
Active Tuner, 40 Characterization, vii, viii, 14, 17–22, 35, 37,
Adjacent Channel Power Ratio (ACPR), 5, 8, 38, 42, 47, 49, 50, 55, 57, 62, 63, 70,
9, 87, 202 72, 83, 101–107, 113–122, 124, 125,
AM/AM, viii, 87, 106, 109 139, 152, 157, 184, 185, 191–193, 201,
AM/AM Measurement, viii, 87, 106 204–206, 208, 212, 215, 217–221, 225,
AM/PM, viii, 87, 106, 107, 109 227
AM/PM Measurement, viii, 87, 106, 107 C/IMD3, 87
Arbitrary waveform generator (AWG), 192, Class-A, 10–12, 194, 196
193 Class-AB, 208
Attenuator, 56–61, 68, 76, 83, 95, 99, 142, Class-B, 82
150, 154, 192 Class-C, 10, 12
Automatic tuner system (ATS), 202 Class-D, 12, 13, 163, 164
Class-E, 12, 13
B Class-F, 13
Baseband, 140–145, 153, 154, 192, 208, 209 Class-J, 13, 159, 186, 187
signal, 140, 143, 145, 192 Closed-loop Load Pull, 62
Bias Current, 3, 10, 12, 13, 197 Coaxial environment, 32
Bias Voltage, 3, 13, 197 Common Reference, 76, 77, 135
Blocking Capacitor, 3 Communication, vii, 1, 2, 7, 9, 204, 211
Breakdown Voltage, 4 Complex Load, 4, 191
Broadband Impedance, 120, 121, 126, 205 Computer-Aided Design (CAD), 14, 15, 18,
Broadband matching, 191 103, 184–186

F.M. Ghannouchi, M.S. Hashmi, Load-Pull Techniques with Applications to Power 229
Amplifier Design, Springer Series in Advanced Microelectronics 32
DOI 10.1007/978-94-007-4461-5, © Springer Science+Business Media Dordrecht 2013
230 Index

Conduction Angle, 12, 13 Envelope Load Pull, viii

Continuous wave (CW), 178, 191, 197, 198, Error
205 parameter, 172, 174
Control term, 40, 41, 145, 148–151, 153, 170–174
unit, 140–145, 153 Error adapter, 133
variable, 193 Error flow model formulation, 145
Convergence, viii, 70, 76–79, 81–83 Error Model, 39, 40, 80, 145, 147, 151, 170,
Convergence Algorithm, 77–79, 81–83 173
Coplanar Reference Plane, 101 Error Vector Magnitude (EVM), 5, 9
Coplanar waveguide (CPW), 99 ESG, 197–200
Current waveform, 10, 181 Extraction, 14, 17, 99, 110, 113, 133–136, 148,
171, 174, 175, 181, 186
D
DC Power, 1, 104, 197 F
De-embed, 97, 220 Fan-diagram, 185
De-embedding, 35, 36, 41, 90, 91, 93, 94, 97, Feed-Forward, 68, 69, 71, 74, 78
99, 101, 106, 108, 109, 113, 136, 193 Feed forward Active Load Pull, viii, 55, 69, 70
Device Characterization, 18, 19, 21, 22, 35, 42, Feedback, 55, 56, 58, 60–64, 66, 140, 142,
47, 49, 105, 107, 114, 118, 120, 139, 144, 145, 153, 205, 206
191, 192, 201, 206, 225, 227 Feedback Loop, 56, 58, 61–64, 66, 142, 144,
Device Under Test, 16, 17, 29, 55, 87, 113, 164 205, 206
Digital, 1, 145, 166, 176, 210–212, 220 Feedback Network, 58, 60
Direct SP Configuration, 94 First-order, 147
Directivity, 38–41, 50, 59, 61, 62, 122 Flow Graph, 39, 62, 70–72, 80, 123, 127, 128,
Doherty, 2 130, 131, 135, 146, 171–174, 181
Drain Efficiency, 4–6, 11–14, 38, 158, 159 Fourier coefficient, 171
DUT, 17, 20, 22, 29, 30, 32, 34–41, 44, 45, 47, Fundamental, 5–7, 10, 12, 36, 42, 43, 45-49,
48, 50, 51, 55–59, 61, 62, 65, 66, 69, 75, 76, 78, 79, 81, 109, 124, 126,
68–70, 72, 76, 77, 79, 87–89, 91, 92, 139, 155–159, 164–168, 177, 178, 181,
94-99, 101–110, 113–119, 122–136, 183, 184, 186, 197–201, 208–212
140–142, 144–148, 154, 164–166, Fundamental frequency, 5, 6, 10, 12, 42,
168–175, 177, 179, 183, 193, 197–201, 45–49, 69, 75, 76, 156, 159, 164–166,
205, 208–217, 219, 220 177, 178, 186, 197, 198, 200, 209, 212
plane, 38, 40, 51, 55-059, 65, 68, 69, 72,
76, 79, 114, 115, 118, 119, 123, 127, G
128, 135, 136, 141, 142, 144–148, 154, GaAs, 42, 82, 156, 159, 216
172, 179, 219 GPIB, 34, 169, 198
Dynamic range, 7, 83, 142, 154, 166–168, 208,
210–212, 217, 219, 220 H
Harmonic
E Active Open-Loop Load-Pull, 77
Efficiency, 1, 2, 4–6, 10–16, 30, 38, 42, 63, 80, Closed-Loop Load-Pull, 63, 64
83, 87, 104, 139, 158–160, 163, 183, Harmonic Distortion, 4, 6–8, 78
184, 186, 191, 195, 198, 199, 201, Harmonic ELP, 139, 155–159
207–209 Harmonic Feed-Forward Load-Pull, 74
Electromehanical tuner (EMT), vii, 29–32, 34, Harmonic frequency, 37, 48, 78, 156, 166, 170,
35 179, 197, 208–211
Electronic circuit, 33, 142 Harmonic Index, 75, 78
Electronic Load Module (ELM), 75 Harmonic Load-Pull, 42, 44–47, 197
Electronic Tuner (ETS), vii, 29, 33–35 Harmonic Tuning, 13, 47, 49, 83, 198
ELP, viii, 139–159, 205, 206 HEMT, 82, 187, 198
equations, 141–143, 145 Heterodyne receiver, 170
Enhanced calibration, 174 High Frequency, 22, 118, 227
Enhanced Loop, 49, 50, 120, 122–125 High-power DUT, 113, 114, 133
Index 231

High Power Load Pull, 18, 19, 113, 114, 122, Linear Network, 2, 186
133, 135 Linearity, 1, 2, 6, 7, 12, 18, 35, 50, 87, 103,
Hybrid Load-Pull, 130 196, 207, 208, 210, 212, 220
Load Impedance, 4, 11, 13, 22, 87, 97, 105,
I 109, 181, 184, 186, 193, 207
Impedance grid, 83, 156 Load Line, 3, 4, 11
Impedance sweep, 156 Load-Pull, 1, 17, 18, 22, 29, 30, 35, 36, 42,
Impedance Synthesis, 191 44–47, 55, 56, 62, 64, 68, 71, 74, 76,
Impedance Transformer, 120, 121, 125–135 78, 83, 87, 96, 97, 99, 113, 118, 119,
IMR, 192–196 130, 140, 154, 155, 191, 197, 201, 204,
In-band intermodulation, 217, 219, 220 208, 209, 225
Incident, 17, 29, 30, 36–39, 56, 63, 65, 66, 69, Load-Pull Gain, 15, 69, 70
79, 94, 108, 114, 115, 118, 127, 128, Load-Pull Head, 66, 67
140–143, 155, 166, 170, 175, 177, 179, Load-pull power, 114, 118–120, 126–129
192, 197, 208, 209 Load Reflection Coefficient, 17, 51, 69, 109,
Incident Traveling Wave, 56, 63, 65, 66, 69, 114, 119, 120, 124, 129, 146, 160, 179,
128, 140–143, 177 197, 199, 205, 212
Injected, 29, 40, 41, 69, 72, 89, 91, 93–95, 98, Load Resistance, 10
127, 128, 181, 197, 201, 209, 213, 216, Local Model, 80–82
217, 219, 220 Local Oscillator (LO), 103, 140, 168, 211
Injected Traveling Wave, 127 Loop, viii, 49–51, 55–68, 70–84, 99, 118, 120,
Input Impedance, 87, 95, 96 122–125, 127, 132, 140, 142, 144, 147,
Input noise, 213, 214 150, 152-154, 156, 158, 159, 183, 197,
Insertion Loss, 20, 34, 35, 41, 48, 49, 91, 122, 205, 206, 208, 209, 219, 221
124, 125, 219 Loop Amplifier, 56–60, 62–68, 70, 73, 76, 77,
Intermodulation Distortion (IMD), 4, 7, 8, 87, 83, 99, 142, 150, 154, 208
191, 192, 194, 201–204, 209 Low Impedance Line, 129
IQ demodulator, 140–143 Low Loss Passive Tuner, 66
IQ Modulator, 75 Low-pass filter, 143
Isolation, 43, 46–50, 57, 59, 61, 68, 70, 142,
145, 150, 153, 155 M
Iterative Process, 14, 30, 70, 78 Magnitude, 5, 9, 30, 31, 41, 42, 44–47, 56, 57,
I-V 59, 60, 68, 70, 72, 75–79, 83, 87, 95,
plane, 186 113, 116, 118, 127, 133, 143, 144, 153,
waveform, 167, 169, 170 154, 165–168, 171, 177, 178, 180, 181,
193, 198, 217, 220
K Matching Circuit, 4, 19, 186, 187
Kahn, 2 Matching Impedance, 17, 186, 191
Klopfenstein, 126, 127 Matching Network, 5, 19, 186, 208
Knee Voltage, 4, 12, 184 Measurement, vii, viii, 1, 9, 14–16, 18, 20–22,
29, 34–42, 47–49, 55, 57, 62, 63, 66,
L 67, 70–72, 79, 80, 83, 87–90, 94–99,
Large Periphery, viii 101, 102, 104, 106, 107, 109, 110, 113,
Large-Signal, vii, 2, 3, 11, 14–16, 37, 69, 76, 114, 116–118, 120–126, 129, 130,
87, 96, 105, 167, 170, 177, 196, 204, 132–136, 139, 142, 144, 148, 151–159,
208 163–172, 174, 176–186, 191–195,
Large-Signal Model, 14, 170, 196 197–202, 204, 206, 209–213, 215–221,
Large Signal Network Analyzer (LSNA), 177, 225, 227
178, 183, 197, 198 Measurement-based PA Design, 16
Large-signal waveform, 16 Measurement plane, 181, 219
Least Squares Method, 148 Measuring port, 39, 88, 89, 93, 170
LINC, 2 Memory effect, 200
Line Stretcher, 127, 129, 132, 133 MESFET, 103, 193, 195, 216
Linear mode PA, 159 Microstrip, 34, 35, 41, 135, 220
232 Index

Microwave, vii, viii, 14, 18, 30, 37, 39, 51, 89, Oscillator Measurements, 104, 105
91, 93, 99, 101, 102, 104–106, 139, Oscilloscope, 164, 166–168, 176, 197
163, 164, 168, 169, 184, 187, 191, 192, Output Impedance, 113, 114, 117, 119, 129,
202, 215–217, 225, 227 130
Microwave Transition Analyzer (MTA), Output noise, 102, 214
168–170, 192–194 Output Power, 2, 3, 5, 6, 11–14, 18, 20, 29, 39,
Mismatch, 11, 30, 31, 38, 39, 51, 60, 80, 114, 41, 42, 50, 64–67, 73, 75, 77, 78, 99,
124–127, 130, 132, 133, 216 104, 117, 159, 163, 180, 186, 194, 195,
Mixer, vii, viii, 75, 103, 167, 170, 191–193, 198–202, 204, 209
212, 217, 218, 220, 221
Modulated, viii, 8, 9, 18, 21, 166, 191, 201, P
202, 204, 205, 209–211 Passive demodulator, 142, 146
Modulated Load Pull, viii, 191, 201, 202 Passive Impedance Tuner, 22, 88, 122, 132,
Multi-harmonic Load-pull, 197 133
Multi-harmonic Receiver (MHR), 169–174, Passive Load-Pull, 29, 30, 96
177 Passive Source-Pull, 95, 96
Multi-harmonic SP Reflectometer, 179–181 Passive Tuner, 20, 29, 36, 40, 50, 64–66, 88,
Multi-tone, viii, 191–193, 195, 196, 204 115, 121, 122, 132, 133, 217
Multi-tone Load-Pull, 191 Peak Voltage, 13, 117
Multipurpose Tuner (MPT), 46–48 Phase, 30, 31, 42, 44–47, 49, 51, 56-61, 68, 70,
72–79, 83, 87, 95, 97–99, 106–110,
N 123, 127–129, 131–133, 140, 143, 144,
Network analyzer, 19, 29, 36–40, 97, 99, 100, 154, 156–158, 165–168, 177, 178, 180,
102, 103, 106, 124, 147, 148, 152, 166, 181, 186, 192, 193, 195, 197–201, 205,
169, 177, 192, 197, 217 208, 211, 217, 220
Newton-Raphson, 78, 83 Phase Shifter, 56–61, 68, 76, 83, 95, 99
Noise PHD, 78, 79, 81–83
meter, 102 model, 78, 79, 81–83
parameter, 212, 215, 216 PIN diode, 30, 33–35
source, 102, 212–214, 216 PNA-X, 168
Noise Characterization, viii, 102, 191, 212, Poly Harmonic Distortion Modeling, 78
215 Pout , 5–7, 12, 29, 30, 38, 41, 65, 66, 104, 179
Noise Measurement, viii, 34, 102 Power Added Efficiency (PAE), 4–6, 18, 20,
Nonlinear device modeling, 140, 166 30, 38, 87, 104, 110, 186, 191, 198–201
Nonlinear I-V waveform, 167 Power Amplifier (PA), vii, 1–7, 10–16, 18, 29,
Nonlinear network, 2 42, 49, 55, 63, 65–68, 101, 104, 113,
Nonlinear Vector Network Analyzer 152, 154, 157–159, 163, 175, 176, 186,
(NLVNA), 169 191, 198, 200, 201, 204, 207, 227
N-port network, 88 Power Amplifier Characterization, 104
Numerical method, 79 Power meter, 29, 36, 37, 39–41, 50, 99–101,
109, 172, 179, 181, 193, 212, 216, 218,
O 219
On-wafer, 18, 21, 32, 33, 35, 49, 87, 99-101, Pre-match, 120, 121
114, 193 Pre-matching, 120, 121, 124
Open-Loop Active Load-Pull, viii, 55, 76, 81, Pre-matching Probe, 120, 121
82 Probe, 21, 22, 30–32, 36, 120, 121, 164, 165,
Open-loop harmonic, 158, 159 211, 218, 220
Open-short-load (OSL), 40, 99–101, 172, 212 Pseudowave, 177
Optimal second-harmonic, 157, 159
Optimal third-harmonic, 159 Q
Optimization, vii, 2, 42, 72, 80, 81, 103, 104, Quadrature Demodulator, 140–143
139, 152, 156, 159, 183, 185, 186, 196, Quadrature Modulator, 140, 142, 144
204, 220 Quarter wave Transformer, 113, 121, 124, 125
Oscillation, 57, 59, 62, 63, 140, 153, 219 Quiescent Point, 4
Index 233

R Source Impedance, 87, 88, 95–97, 102, 103,

Radio Frequency (RF), vii, 1–4, 6, 12, 36–39, 105, 181, 213, 216, 220
75, 88, 103, 113, 120, 139, 140, 163, Source-Pull, 95, 101
164, 169, 170, 178, 184–186, 191, 197, SP, 87–99, 101–110, 177–183, 212–220
207, 211, 212, 218–220, 225, 227 SP Calibration, 89, 97, 106, 181
Reciprocal Network, 131 SP Junction, 88–95, 101, 102, 105, 106, 213,
Reference 214, 216, 219
port, 88 SP Network Analyzer, 97, 106, 217
signal, 76, 168, 170, 178, 181 SP Reflectometer, 88–91, 93–96, 98, 101–107,
Reference Plane, 30–32, 38, 40, 41, 66, 89–91, 177–183, 213, 215–217
93–95, 101, 119, 134, 135, 140, 177, SP1, 96, 108, 109
181, 205, 212, 216 SP2, 96–98, 108, 109
Reflected, 17, 29, 30, 36, 38, 39, 42–44, 55, S-Parameter, 172
56, 63, 65, 66, 69, 73–76, 78–81, 91, S-parameter calibration, 172, 173
94, 114, 115, 118, 127, 128, 140–142, Spectrum analyzer, 36, 212, 215, 216, 218–220
144, 155, 166, 167, 170, 175, 177, 181, Spurious Free Dynamic Range, 7
192, 197, 202, 205–207, 217 Stability, 56, 62, 70, 77, 102, 142, 150, 153,
Reflected Traveling Wave, 43, 44, 55, 56, 63, 154, 197, 199, 219
66, 69, 74-76, 79, 128, 140–142, 144, analysis, 153
177, 205–207 Step Recovery Diode (SRD), 168, 178
Reflection Coefficient, 17, 19, 22, 29, 31–33, Switched-Mode Power Amplifier, 13
37, 39, 40, 42, 45, 47, 50, 51, 56–60, Switching Amplifier, 13, 14
62, 63, 65, 66, 68–73, 76–81, 83, Synthesized reflection coefficient, 59, 70, 72,
88–99, 101, 105, 106, 109, 113–115, 121, 122, 127, 128, 141, 142, 147, 205
118–133, 140–142, 144, 146–148, 153,
154, 156–160, 172, 177–182, 192–194, T
197–199, 205, 207, 209, 210, 212–215, Test Frequency, 215
217, 219, 220 Third-harmonic, 42, 78, 156–159, 167, 186,
Reflection Factor, 19, 30–32, 34, 35, 43, 48, 198–201
50, 51, 119 Third order
Reflectometer, viii, 88–91, 93–96, 98, mixing, 79
101–107, 177–183, 212, 213, 215–217 model, 79
Reverse Configuration, 89–91, 93, 95, 97, 98, Third-order Intercept (IP3), 7, 201
101–103, 179, 182, 213, 215 Third-order Intermodulation (IMD3), 7, 87,
RF Power, 1, 2, 4, 38, 39, 184, 227 201–204
RFPA, 1, 163, 183 Three-port, 56, 61, 64, 107, 108, 140, 142
component, 56, 61, 64, 140, 142
S Thru-Reflect-Line (TRL), 135
Sampling Time-domain waveform, viii, 148, 155, 163,
oscilloscope, 167, 168 165, 168, 169, 178, 179, 181, 183, 206
rate, 166 Total Efficiency, 6
Second-harmonic, 42, 43, 81, 156-159, 186, Total Harmonic Distortion (THD), 7
199–201, 208, 211, 220 Transconductance, 10, 12, 13, 184
Short-Open-Load-Thru (SOLT), 135, 193 Transistor characterization, 20, 106, 184, 185
Signal Flow Graph, 62, 72, 123, 127, 128, 130, Transmission line, vii, 30, 32, 33, 115, 164
131 Traveling Wave, 30, 43, 44, 55, 56, 63, 65, 66,
Signal Flow Model, 40, 50, 57, 58, 71, 114, 68, 69, 74–76, 79, 128, 140–144, 174,
115 177, 202, 205–207
Single-stub, 30 Triplexer, 43–45, 47–49, 63, 77, 124, 155, 219
Six-Port, 87, 177 TRL Standard, 135
Six-Port Load-Pull, 87 Tuner Bandwidth, 21
Sliding-Short, 72–74 Tuner Resolution, 20
Small-Signal, 11, 109, 186, 196 Tuner Size, 21, 34
Small-signal Model, 186, 196 Tuning Accuracy, 21, 35
234 Index

Tuning Probe, 120, 121 Voltage waveform, 12, 13, 175, 176, 185
Tuning Range, viii, 19, 20, 32, 48–50, 57, 114,
125, 127, 142
Tuning Speed, 20, 33 W
Two-port network, 88, 89, 97, 214, 219 Waveform, vii, viii, 9, 10, 12, 13, 16, 140, 148,
Two-tone, 7, 192, 193, 201–204, 219–221 152, 155, 157, 159, 163–172, 174–186,
192, 205, 206, 227
U engineering, vii, viii, 140, 163, 183, 184,
Uncertainty, 38, 39, 114, 124, 134, 217 186, 227
Unique measurement, viii, 157 measurement, 16, 148, 152, 155, 159, 163,
165, 167–170, 177, 183–186, 206, 227
V
Vector Network Analyzer (VNA), 19, 29, 36, WCDMA, 1, 202-204, 211
38–41, 97, 101, 104, 124, 134–136, Wide Bandwidth, 135, 210
147, 152, 166–169, 171, 172, 177, 182, Wideband, 1, 43, 46, 168, 202, 208–212, 219
212, 213 Wideband Load-Pull, 208, 209
VNA Wireless, vii, 1, 7, 208, 217, 225
error correction, 134, 135 Wireless Communication, 1, 7
plane, 40
Voltage Standing Wave Ratio (VSWR), 11, 32,
39, 48, 114, 116, 117, 119, 125–127, Y
129, 130, 133, 164 YIG Filter, 63, 179, 180, 219, 220