Teuvo Suntio

Dynamic Prole of
Switched-Mode Converter
Related Titles

Batarseh, I.

Power Electronic Circuits

2006
ISBN: 978-0-471-12662-1

Mohan, N., Robbins, W. P., Undeland, T. M.

Power Electronics
Converters, Applications and Design, Media Enhanced. International Edition

2003
ISBN: 978-0-471-42908-1
Teuvo Suntio

Dynamic Prole of Switched-Mode Converter

Modeling, Analysis and Control

The Author
All books published by Wiley-VCH are
carefully produced. Nevertheless, authors,
Prof. Teuvo Suntio editors, and publisher do not warrant the
Tampere University of Technology information contained in these books,
Dept. of Electrical Energy Engineering including this book, to be free of errors.
Tampere, Finland Readers are advised to keep in mind that
teuvo.suntio@tut. statements, data, illustrations, procedural
details or other items may inadvertently be
inaccurate.

Cover-Picture Library of Congress Card No.: applied for

Spiesz Design, Neu-Ulm
British Library Cataloguing-in-Publication Data
A catalogue record for this book is available
from the British Library.

Bibliographic information published by

the Deutsche Nationalbibliothek
The Deutsche Nationalbibliothek lists this
publication in the Deutsche National-
bibliograe; detailed bibliographic data are
available in the Internet at http://dnb.d-nb.de.

2009 WILEY-VCH Verlag GmbH & Co.

KGaA, Weinheim

All rights reserved (including those of

translation into other languages). No part of
this book may be reproduced in any form by
photoprinting, microlm, or any other
means nor transmitted or translated into a
machine language without written permission
from the publishers. Registered names,
trademarks, etc. used in this book, even when
not specically marked as such, are not to be
considered unprotected by law.

Typesetting Laserwords, Chennai, India

Printing Strauss GmbH, Morlenbach

Binding Litges & Dopf Buchbinderei GmbH,

Heppenheim

Printed in the Federal Republic of Germany

Printed on acid-free paper

ISBN: 978-3-527-40708-8
 V

Contents

Preface XI

1 Introduction 1
1.1 Introduction 1
1.2 Dynamic Modeling of Switched-Mode Converters 4
1.3 Dynamic Analysis of Interconnected Systems 6
1.4 Canonical Equivalent Circuit 8
1.5 Load-Response-Based Dynamic Analysis 9
1.6 Content Review 12

2 Basis for Dynamic Analysis and Control Dynamics 17

2.1 Introduction 17
2.2 Dynamic Representations at Open Loop 17
2.2.1 State Space 19
2.2.2 Two-Port Models 21
2.2.3 Control-Block Diagrams 23
2.3 Dynamic Representations at a Closed Loop 23
2.3.1 Voltage-Output Converter 26
2.3.2 Current-Output Converter 27
2.4 Load and Source Effects 28
2.4.1 Voltage-Output Converter 29
2.4.2 Current-Output Converter 31
2.5 An Example LC Circuit 33
2.5.1 Voltage-Output Circuit 33
2.5.2 Current-Output Circuit 35
2.6 Review of Basic Mathematical Tools 37
2.6.1 Linearization 37
2.6.2 Transfer Functions 38
2.6.2.1 Single Zero 39
2.6.2.2 Single Pole 40
2.6.2.3 Second-Order Transfer Function 40
2.6.2.4 Example 43
2.6.3 Stability and Performance 45

Dynamic Prole of Switched-Mode Converter. Teuvo Suntio

2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
ISBN: 978-3-527-40708-8
VI Contents

2.6.3.1 Stability 46
2.6.3.2 Loop-Gain-Related Dynamic Indices 48
2.6.3.3 Right-Half-Plane Zero and Pole 50
2.6.4 Matrix Algebra 50
2.6.4.1 Addition of Matrices 53
2.6.4.2 Multiplication by Scalar 53
2.6.4.3 Matrix Multiplication 54
2.6.4.4 Matrix Determinant 54
2.6.4.5 Matrix Inversion 55
2.7 Operational and Control Modes 55

3 Average and Small-Signal Modeling of Direct-On-Time Controlled

Converters 59
3.1 Introduction 59
3.2 Direct-on-Time Control 60
3.3 Generalized Modeling Technique 62
3.3.1 Buck Converter 64
3.3.2 Boost Converter 66
3.3.3 BuckBoost Converter 68
3.4 Fixed-Frequency Operation in CCM 70
3.4.1 Synchronous Buck Converter 71
3.4.2 Dynamic Descriptions of Buck, Boost, and BuckBoost
Converters 76
3.4.2.1 Diode-Switched Buck (Figure 3.6a) 76
3.4.2.2 Diode-Switched Boost (Figure 3.8a) 77
3.4.2.3 Synchronous Boost (Figure 3.8b) 79
3.4.2.4 Diode-Switched BuckBoost (Figure 3.10a) 80
3.4.2.5 Synchronous BuckBoost (Figure 3.10b) 81
3.4.3 Steady-State and Small-Signal Equivalent Circuits 82
3.5 Fixed-Frequency Operation in DCM 85
3.5.1 Buck Converter 87
3.5.2 Dynamic Models for Boost and BuckBoost Converters 92
3.5.2.1 Boost Converter (Figure 3.8a) 92
3.5.2.2 BuckBoost Converter (Figure3.10a) 94
3.6 Dynamic Review 95
3.6.1 Buck Converter 96
3.6.1.1 Control-to-Output Transfer Function 96
3.6.1.2 Output Impedance 98
3.6.1.3 Input-to-Output Transfer Function 100
3.6.1.4 Input Admittance 103
3.6.1.5 Ideal Input Admittance 104
3.6.1.6 Short-Circuit Input Admittance 106
3.6.2 Boost Converter 106
3.6.2.1 Control-to-Output Transfer Function 108
3.6.2.2 Output Impedance 109
 Contents VII

3.6.2.3 Input-to-Output Transfer Functions 111

3.6.2.4 Input Admittance 112
3.6.2.5 Ideal Input Admittance 114
3.6.2.6 Short-Circuit Input Admittance 116

4 Average and Small-Signal Modeling of Peak-Current-Mode

Control 121
4.1 Introduction 121
4.2 PCM-Control Principle 122
4.3 Modeling in CCM 124
4.3.1 Duty-Ratio Constraints for Buck, Boost, and BuckBoost
Converters 126
4.3.1.1 Buck Converter 126
4.3.1.2 Boost Converter 126
4.3.1.3 BuckBoost 128
4.3.1.4 General CCM Transfer Functions 129
4.3.2 Specic Transfer Functions for the Basic Converters 131
4.3.2.1 Buck Converter 131
4.3.2.2 Boost Converter 133
4.3.2.3 BuckBoost Converter 134
4.3.3 Origin and Consequences of Mode Limit in CCM 136
4.4 Modeling in DCM 139
4.4.1 Duty-Ratio Constraints for Basic Converters 140
4.4.1.1 Buck Converter 140
4.4.1.2 Boost Converter 142
4.4.1.3 Buckboost Converter 142
4.4.2 Small-Signal PCMC State Spaces for the Basic Converters 143
4.4.3 Origin and Consequences of Mode Limit in DCM 144
4.5 Dynamic Review 146
4.5.1 Buck Converter 147
4.5.1.1 Control-to-Output Transfer Function 148
4.5.1.2 Output Impedance 151
4.5.1.3 Input-to-Output Transfer Function 152
4.5.1.4 Input Admittance 153
4.5.1.5 Ideal Input Admittance 155
4.5.1.6 Short-Circuit Input Admittance 155
4.5.2 Boost Converter 157
4.5.2.1 Control-to-Output Transfer Function 159
4.5.2.2 Output Impedance 161
4.5.2.3 Input-to-Output Transfer Function 163
4.5.2.4 Input Admittance 164
4.5.2.5 Ideal Input Admittance 165
4.5.2.6 Short-Circuit Input Admittance 166
VIII Contents

5 Average and Small-Signal Modeling of Average-Current-Mode

Control 169
5.1 Introduction 169
5.2 ACM-Control Principle 169
5.3 Modeling with Full-Ripple-Current Feedback 171
5.4 Dynamic Review 175
5.4.1 Control-to-Output Transfer Function 177
5.4.2 Output Impedance 177
5.4.3 Input-to-Output Transfer Function 179
5.4.4 Input Admittance 183
5.5 Effect of Current-Loop High-Frequency Pole 183

6 Average and Small-Signal Modeling of Self-Oscillation Control 189

6.1 Introduction 189
6.2 Self-Oscillation Modeling 189
6.2.1 Averaged Direct-on-Time Model 190
6.2.2 Small-Signal Direct-on-Time Model 193
6.2.3 Small-Signal PCM Models 194
6.3 Dynamic Review 198
6.3.1 Buck Converter 198
6.3.1.1 Control-to-Output Transfer Function 199
6.3.1.2 Output Impedance 199
6.3.1.3 Input-to-Output Transfer Function 201
6.3.1.4 Input Admittances 201
6.3.2 Flyback Converter 202
6.3.2.1 Control-to-Output Transfer Function 203
6.3.2.2 Output Impedance 206
6.3.2.3 Input-to-Output Transfer Function 206
6.3.2.4 Input Admittance 207
6.3.2.5 Ideal and Short-Circuit Admittances 208

7 Dynamic Modeling and Analysis of Current-Output

Converters 211
7.1 Introduction 211
7.2 Dynamic Models for Current-Output Converter 212
7.2.1 Modied-State-Space-Averaging Technique 213
7.2.2 General Dynamic Models 215
7.3 Load and Supply Interactions 216
7.4 Cascaded Voltage-Current Loops 218
7.5 Dynamic Review 219

8 Interconnected Systems 225

8.1 Introduction 225
8.2 Theoretical Interaction Formulation 226
8.2.1 Load and Supply Interactions 227
 Contents IX

8.2.2 Internal and InputOutput Stabilities 230

8.2.3 Output Voltage Remote Sensing 234
8.2.4 Input EMI Filter 236
8.3 Review of Methods to Reduce the Interactions 238
8.3.1 Input-Voltage Feedforward 238
8.3.2 Output-Current Feedforward 240
8.4 Experimental Dynamic Review 241
8.4.1 Load and Supply Interactions 243
8.4.2 Remote Sensing 251
8.4.3 System Stability 255

9 Control Design Issues 261

9.1 Introduction 261
9.2 Feedback-Loop-Design Constraints 265
9.2.1 Phase and Gain Margins 266
9.2.2 RHP Zeros and Poles 268
9.2.3 Minimum and Maximum Loop Crossover Frequencies 268
9.2.4 Internal Gain of an Operational Amplier 270
9.3 Controller Implementations 271
9.4 Optocoupler Isolation 272
9.5 Shunt-Regulator-Based Control Systems 274
9.5.1 Dynamic Model 274
9.5.2 Two-Loop Control System 281
9.6 Simple Control-Design Method 284
9.6.1 Control Design Example: VMC Buck Converter 285
9.6.2 Control Design Example: PCMC Buck Converter 290
9.6.3 Control Design Example: VMC Boost Converter 295
9.6.4 Control Design Example: PCMC Boost Converter 298
9.7 Conclusions 302

10 The Fourth-Order Converter Superbuck 307

10.1 Introduction 307
10.2 Basic Dynamics 309
10.2.1 Averaged Models 311
10.2.1.1 Averaged State Space 311
10.2.1.2 Steady-State Operating Point 311
10.2.1.3 Boundary Conduction Mode 312
10.2.2 Small-Signal Models 312
10.2.2.1 Small-Signal State Space 312
10.2.2.2 Transfer Functions 313
10.2.3 RHP Poles 315
10.2.4 Design Considerations 317
10.3 Coupled-Inductor Superbuck 318
10.3.1 Small-Signal Models 319
X Contents

10.3.2 RHP Poles 322

10.3.3 Input-Current-Ripple Reduction 322
10.3.4 Design Considerations 325
10.4 PCM-Controlled Superbuck 325
10.4.1 Small-Signal Models 326
10.4.2 Design Considerations 330
10.4.2.1 Inductor-Current-Feedback Compensation 330
10.4.2.2 Avoiding RHP Poles 330
10.5 Coupled-Inductor PCM-Controlled Superbuck 331
10.5.1 Small-Signal Models 331
10.5.2 Design Considerations 336
10.6 Dynamic Review 337
10.6.1 Superbuck I: 1520 V/10 V/2.5 A 337
10.6.2 Superbuck II: 69 V/3.4 V/12 A 345
10.7 Summary 348

Index 351
 XI

Preface

Twenty-ve years in industry engaged in the practical design of

switched-mode converters and the systems based on them and especially
the multitude of problems we faced convinced me that there has to be some-
thing we do not really understand. After joining the academy, I decided to nd
out what are the reasons for the problems we were confronted with almost
daily. After ten years in academy, I understand many of the problems but there
still seems to be a lot to solve. The biggest challenge is to convince the other
designers and academics that the electrical circuits have an internal dynamic
sole which dictates the way the circuits behave and contribute to the behavior of
the systems composed of them. The plain sole has to be known for being able
to understand and predict the behavior and especially to avoid the undesired
consequences such as instability and deteriorated transient performance. The
book is intended to introduce the dynamic features the different converter
topologies may incorporate and the changes the different control methods
and operation modes may create in them in addition to the introduction of
the methods to model analytically the dynamic behavior and to design the
controllers.
Many individuals have helped me to create the book and especially to
understand the extent of the problems and to nd the solutions for them:
Professor Dr. D. R. Vij as the consultant editor and the staff of Wiley-VCH
have provided me the opportunity to publish the book and patient guidance
during the process. Dr. Kai Zenger has guided me into the secrets of control
engineering but a lot of work is still left. My former doctorate students Dr. Idris
Gadoura, Dr. Ander Tenno, Dr. Mikko Hankaniemi, Dr. Ali Altowati, and Dr.
Matti Karppanen have contributed substantially to the solutions presented in
the book and especially to their experimental validation. I am very grateful to
my wife Sirpa for her love and patience during the process.

Teuvo Suntio

Dynamic Prole of Switched-Mode Converter. Teuvo Suntio

2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
ISBN: 978-3-527-40708-8
 1

1
Introduction

1.1
Introduction

The switched-mode converters can be divided into two main classes

such as voltage- (Figure 1.1) and current-sourced (Figure 1.2) converters [1],
where either the output voltage (Figures 1.1a and 1.2b) or output current
(Figures 1.1b and 1.2a) is kept constant [2]. As a consequence, there are
four different main types of converters namely voltage-to-voltage, voltage-
to-current, current-to-current, and current-to-voltage converters having dif-
ferent dynamic features. The most usual converter is the voltage-to-voltage
converter (Figure 1.1a) because most of the energy sources are voltage sources
and the loads current sinks [3]. Sometimes storage batteries are connected
at the output of the voltage-sourced converter, which requires to limiting
the maximum output current for preventing the converters from damage
due to the extremely low internal impedance of a storage battery [48]. The
operation at current-limiting mode changes the voltage-to-voltage converter to
voltage-to-current converter (Figure 1.1b). Current-sourced converters can be
used to interface solar arrays and magnetic energy storage systems due to the
current-output nature of those energy sources [9, 10]. Such a basic converter
is naturally the current-to-current converter (Figure 1.2a). If the maximum-
output voltage limiting is used, the current-to-current converter changes to a
current-to-voltage converter (Figure 1.2b).
Every switched-mode converter has a unique dynamic prole or internal
dynamics, which would determine the obtainable transient dynamics and
robustness of stability as well as the converters sensitivity to the external
source and load impedances [1113]. The dynamic prole can be changed
by means of certain internal feedback or feedforward arrangements but not
much in practice by means of the feedback-loop control design. The internal
dynamics can be characterized by means of a certain set of open-loop transfer
functions constituting the circuit theoretical two-port parameters known as G
(Figure 1.1a), Y (Figure 1.1b), H (Figure 1.2a), or Z (Figure 1.2b) depending on
the input source and the type of the converter output [1115]. The different sets
do characterize only one main type of a converter and are not interchangeable

Dynamic Prole of Switched-Mode Converter. Teuvo Suntio

2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
ISBN: 978-3-527-40708-8
2 1 Introduction

Figure 1.1 Voltage-sourced

converter (a) at voltage-output
mode and (b) at current-output
mode.

but the parameters within the main converter class (i.e., G and Y, H and Z)
can be computed from each other. In addition with the open-loop transfer
functions, certain admittance or impedance parameters have to be dened for
obtaining the full picture of the internal dynamic prole [11].
The term internal means that the transfer functions constituting the sets are
to be such that all the effects of the source and load impedances are removed
from them. The analytical models can be easily derived to be such, when
knowing the correct load yielding the internal models (Figures 1.1. and 1.2).
The dynamic parameter sets for the voltage-to-voltage and current-to-current
converters can also be usually measured by means of frequency response
analyzers but certain internal control modes may change the open-loop
converter such that it cannot operate at the dened load or the required
ideal load is not available. In such cases, a resistive load has to be used
and the internal models have to be solved computationally [11, 16, 17]. It is,
however, extremely important to obtain those internal models because they
only characterize the converter not the source- or load-affected models.
A large number of power electronics text books are available such as [1826],
which tend to give a comprehensive picture of all the issues related to the design
of switched-mode converters both in AC and DC applications. Therefore, it is
understandable that the dynamic issues are typically not treated adequately.
The exceptions are [27] and [28], which mainly concentrates on the dynamic
 1.1 Introduction 3

Figure 1.2 Current-sourced

converter (a) at current-output
mode and (b) at voltage-output
mode.

issues. The main deciency of the dynamic analyses in the aforementioned

text books is the inclusion of the load usually as a resistor in the presented
dynamic models, which may effectively hide the true dynamics and thereby
made the output of the system-level interaction analyses useless. A describing
example of the misunderstanding such a treatment can cause is the prevailing
understanding that the damping of the resonant behavior in a converter
would decrease, when the resistive load is decreased [29]. The phenomenon
is naturally true from the external point of view but the internal dynamics
does not, however, change if the operating point is maintained. Therefore,
it may be a big surprise when the converter behaves nicely in the laboratory
but dynamic problems arise when connected into a real application. Such
an experience might be very common among the industrial switched-mode-
converter designers leading easily to frustration and blaming the customer of
abusing the converter.
The main goal of the book is to provide the reader with the tools by means of
which the challenging dynamics of the systems comprising of switched-mode
converters can be made more understandable and the design of them more
deterministic. It is natural that the key element is the building block of such
a system the switched-mode converter. The most fundamental issue behind
the ideas provided in the book is the observation that each electrical device
4 1 Introduction

or circuit has its unique internal dynamic prole similar to the psychological
prole of a human being [11]: the prole determines how the device or circuit
would behave as a part of the system under different external interactions and
how it would affect the other subsystems within the overall system. The internal
prole cannot be basically changed by applying external feedback control but
only by providing internal feedback or feedforward from the input, output
and/or state variables constituting the dynamic constellation of the device.
An illustrative example is the application of inductor current to produce the
duty ratio in a peak-current-mode-controlled (PCMC) converter [30], which
changes profoundly the converter dynamics compared to the corresponding
direct-duty-ratio or voltage-mode-controlled (VMC) converter, where the duty
ratio is produced using a constant ramp voltage: The resonant nature of the
VMC converter disappears, the input-noise attenuation may be substantially
increased, the internal open-loop output impedance is increased but the
nonminimum nature if existing in the VMC converter would not be removed.
A multitude of similar examples can be given, which actually proves the
existence of such a prole.
During the time of writing the book, the analog control is still dominating
but digital control with all the opportunities involved in it is evidently coming
and may dominate the future converter applications. The fact is, however,
that the power stage does not change and, therefore, the basic dynamic prole
related to the power stage does not change. The digital control with the physical
resolution and time limitations may cause more dynamic problems or equally
also improvements, which can be revealed and analyzed using the methods and
information based on the corresponding continuous-time processes treated in
this book.
The issues related to the dynamic proles are briey discussed and claried
in the subsequent subsections in order to make the reader familiar with
the issues treated in the subsequent chapters. Even if we discussed on the
current-sourced converters in the beginning of the chapter, we will limit our
discussions on the voltage-sourced converters within the rest of the book.

1.2
Dynamic Modeling of Switched-Mode Converters

The dynamic analysis of the voltage-output switched-mode converters dates

back to the early 1970s [31], when the foundation for the state-space-averaging
(SSA) method [32] was laid down. It was observed that the dynamics associated
with the direct-duty-ratio or VMC converter in continuous conduction mode
(CCM) could be quite accurately captured up to half the switching frequency
by averaging the converter variables within a switching cycle and computing
the small-signal models from the corresponding averaged state space by
means of linearization. The dynamic behavior of a converter was represented
by means of the canonical equivalent circuit shown in Figure 1.3 for the
 1.2 Dynamic Modeling of Switched-Mode Converters 5

Figure 1.3 Small-signal

canonical equivalent circuit for a
two-memory-element converter.

two-memory-element converters, where the different circuit elements are

dened according to a specic converter. It may be obvious that the equivalent
circuit in Figure 1.3 provides real physical insight into the dynamic processes
inside the converter and has, therefore, promoted the acceptance of the
theoretical method providing the model. Similar equivalent circuit to Figure 1.3
can also be naturally constructed for the higher order converters.
The rst attempt to model the dynamics associated with a VMC converter
operating in discontinuous mode (DCM) is presented in [33] but it failed to
capture the true full-order dynamics due to the lack of proper understanding of
the dynamical processes inside a converter. The accurate small-signal models
for the DCM operation were developed in the late 1990s [34]. A unied method
based on the SSA method was nally developed in the early 2000s providing
consistent modeling tools for xed and variable-frequency operation both
in DCM, CCM, and even in the combination of them [35]. The pulsewidth
modulation (PWM) process would not produce linear responses but only
at rather low frequencies (i.e., 1/10 of switching frequency) for sinus
excitations [3638]. Therefore, the responses measured through the PWM
input (i.e., control-to-input and control-to-output) may have more phase lag
than the models derived using the SSA method would predict. Further studies
on the topic are needed in order to nd the correct dynamic behavior of the
converter also at the frequencies approaching half the switching frequency.
This is important because the desired loop crossover frequencies tend to
approach ever higher frequencies beyond those typically used in the past.
The small-signal models of the VMC operation are important because the
other control modes would usually only change the dynamics associated with
the duty-ratio generation and, therefore, the corresponding dynamic models
can be derived from the VMC state-space representation by substituting the
perturbed duty ratio with the developed relation between the new control
variable and the duty ratio known as duty-ratio constraints [22].
In reality, the controlled variable is usually the length of the on-time of the
main switch [35]. Under xed-frequency operation, the dynamical information
incorporated into the on-time is equal to that of the duty ratio because of
constant cycle time. Under variable-frequency operation, the duty ratio is
nonlinear and, therefore, the on-time has to be used as the control variable
under the VMC mode of operation. A comprehensive survey of the modeling
issues can be found from [39].
6 1 Introduction

1.3
Dynamic Analysis of Interconnected Systems

The rst system-level analysis was actually applied to a system comprising

of an EMI (electromagnetic interference) lter and a regulated converter in
the mid-1970s [40]. The analysis yielded the design rules for the EMI-lter
design. The equivalent circuit of Figure 1.3 was effectively utilized. It was
noticed that the EMI lter would inuence the converter dynamics through
its output impedance if certain impedance overlaps take place. The developed
design rules were straightforward: avoid impedance overlap with a substantial
margin. It was also stated that the stability of the converter can be deduced
based on the ratio of the lter output impedance and the closed-loop input
impedance of the converter by applying the Nyquist stability criterion. The
impedance ratio is commonly known as minor-loop gain according to [40].
The analyses of interconnected regulated systems [41] have been based on
the minor-loop-gain concept. It was concluded that the design rules given
in [40] are too conservative and they may lead to unnecessary costs if applied
as such. Typically, the design rules are given as a certain forbidden region in a
complex plane out of which the minor-loop gain should stay to avoid instability
and performance degradation. The shaded area in Figure 1.4 is the minimum
area out of which the minor-loop gain should stay for stability to exist. This
criterion is known as ESAC criterion [41]. The design rules of [40] would dene
the forbidden region as the area outside the circle having a radius of 1/GM,
where GM stands for gain margin related to the minor-loop gain.

Figure 1.4 Forbidden regions.

 1.3 Dynamic Analysis of Interconnected Systems 7

Figure 1.5 Different system interfaces.

It has turned out that avoiding the stated forbidden regions does not
actually ensure that the transient performance of the converter would stay
intact but more detailed considerations should be carried out [42, 43]: The
practical power systems consist of several interfaces at which the minor-loop
gain can be dened as depicted in Figure 1.5 (i.e., A1 AN ) containing also
different information on the dynamics of the overall system. The interfaces
that exist at the direct input or output of the power stage of the converter
would contain the most useful information as actually has been demonstrated
in [40]. The existence of stability can be concluded equally based on any of
the dened minor-loop gains within the system [44]. The existence of stability
even with a good margin (i.e., no impedance overlap) does not necessarily
ensure that the transient performance of the associated converters is acceptable
[42, 43].
Typically the converters are equipped with EMI lters or capacitors at
the input side further complicating the performance analysis based on the
measurable information at system level (Figure 1.6) due to hiding effects of
those components [43]. The converter modules may be also provided with
output-voltage remote sensing [45]. The application of the remote sensing
may profoundly change the dynamics of the associated converter depending
on what kind of external passive circuit elements are connected inside the
converter (Figure 1.6) as demonstrated in [46].

Figure 1.6 Internal system interfaces inside a converter.

8 1 Introduction

As a summary we can state that the stability analysis of an interconnected-

regulated system would be deterministic but to conclude whether the transient
performance of the system is satisfactory or not is a more complicated issue and
perfect information on it is difcult to obtain. Therefore, the methods by means
of which the interactions can be reduced or totally eliminated are of great im-
portance and worth to be considered as explained and demonstrated in [4750].

1.4
Canonical Equivalent Circuit

The equivalent circuit introduced in Figure 1.3 as a canonical circuit is not a

true canonical equivalent circuit, because it can only represent the dynamics of
a two-memory-element VMC converter operating in CCM. A two-port model
(Figure 1.7), where the input port is a Norton equivalent circuit and the
output port a Thevenin equivalent circuit, would provide a real canonical
representation of the dynamics associated with a voltage-inputvoltage-output
converter [11, 51]: The input and output-port parameters constitute of a set
known as G-parameters, which can be proved to exist always and thus they
can be dened for any voltage-input-voltage-output electrical system [14]. In
practice, the set composes of the well-known transfer functions typically
used to characterize switched-mode converters. It is essential that the transfer
functions are dened in such a way that the source and load effects are removed
in order to represent the real internal dynamics. Such transfer functions are
commonly known as unterminated transfer functions [51].
In addition, the dynamic representation of the current-output converter
can be derived from the two-port model of the voltage-output converter by
transforming its Thevenin output port to an equivalent Norton representation
shown in Figure 1.8 [52, 53]. The parameters of the current-output model (i.e.,
admittance or Y-parameters) can be derived as a function of the well-known
voltage-output transfer functions (i.e., the G-parameters).
It is known that the source and load impedances, Figure 1.9, may affect the
dynamics of a converter [11]. Mathematical formulation describing analytically
the effects can be solved by using either circuit theory [15] or by applying the
extra element theorem (EET) described in [16]. The EET method would provide
useful formulation but may be difcult to apply.

Figure 1.7 Canonical equivalent

circuit of voltage-output converter.
 1.5 Load-Response-Based Dynamic Analysis 9

Figure 1.8 Canonical equivalent

circuit for current-output converter.

Figure 1.9 Nonideal source and

load: (a) voltage-output converter
and (b) current-output converter.

1.5
Load-Response-Based Dynamic Analysis

The frequency response of the voltage or current loop can be measured

injecting an excitation signal into the corresponding loop, which has to be
disconnected in such a way that only the DC signal can pass through it.
This means that the internal circuitry of the converter has to be manipulated.
Therefore, the methods not requiring the tear-down approach would be
desirable such as the load-transient analysis [5456]. The transient-based-
analysis technique is commonly used in the control engineering textbooks
such as [54]. It should, however, be noticed that the control engineering
textbooks usually discuss on the transients resulting from applying excitations
into the reference input, which is usually not available in the power
electronic converters. The transients resulting from applying excitations into
the disturbance inputs (i.e., input voltage or load current) also contain the effect
of the corresponding open-loop transfer function in addition with the loop
behavior. Those effects usually dominate and they would hide the information
from the loop behavior. An illustrative example of such a phenomenon is
shown in Figure 1.10 [16], where a certain converter under three different
control principles is subjected to a constant-current-type load change. Some
similarities and differences are clearly observable:
10 1 Introduction

Figure 1.10 Output-voltage responses of a buck converter under VMC, PCM, and PCM
with output-current-feedforward (OCF) control to an output- current change.

The setup time of the PCM converter seems to be very long compared
to the VMC converter interpreted easily as a substantial difference in the
voltage-loop crossover frequencies if following the information given for
example in [56]. The output transient of the PCMC-OCF converter is extremely
small and recovers quickly. This could be interpreted as a sign of very
high control bandwidth. The output-voltage loop gains of the converters
are, however, designed in a comparable manner as shown in Figure 1.11.
Therefore, Figure 1.10 clearly demonstrates that the time-domain transients

Figure 1.11 Measured voltage-loop gains (PM = phase margin).

 1.5 Load-Response-Based Dynamic Analysis 11

Figure 1.12 Measured open-loop output impedances.

do not provide the desired information on the voltage-loop properties because

of the dominating effect of the disturbance input.
The measured open-loop output impedances of the converters are shown
in Figure 1.12 and the corresponding closed-loop output impedances in
Figure 1.13, where the voltage-loop-gain effect is observable. The closed-
loop impedances provides the explanations for the observed load transients
as explained in detail in [5759], but basically the origin of the observed
differences is the internal open-loop output impedances (Figure 1.12).

Figure 1.13 Measured closed-loop output impedances.

12 1 Introduction

The internal open- and closed-loop output impedances are important sources
of information, because they can be used to predict the dynamic behavior of
the converter in different load environment and eventually to choose a best
type of converter for the specic application [12].

1.6
Content Review

The content of the subsequent chapters is briey reviewed in order to clarify

the message each chapter contains:
The conceptual and theoretical basis of the book is provided in Chapter 2 in a
simple and practical manner without using difcult mathematical treatments.
The same theoretical formulas are repeated in the associate chapters if deemed
to be necessary for understanding the message. The denition of different
stability concepts, the inuence of zeros and poles in the transfer functions,
and the denition of the open-loop condition in a converter are especially
important to be fully understood in order to understand the messages the book
will provide.
The unied dynamic modeling of the direct-on-time control is provided in
Chapter 3. The method is applied more in detail to the basic converters (i.e.,
buck, boost, and buckboost) in the xed-frequency mode of operation both in
continuous (CCM) and discontinuous conduction modes (DCM). The variable-
frequency operation is treated separately in Chapter 6. Extensive dynamic
review is provided at the end of the chapter based both on experimental and
theoretical evidence. The emphasis is in introducing the dynamical changes
the operation in CCM and DCM would provide. The modeling of the direct-on-
time control is important, because the dynamical models of the other control
modes would be mainly derived based on it. The presented methods are also
easily applicable to modeling of higher order converters.
The dynamic modeling of peak-current-mode control (PCMC) is provided
in Chapter 4 and applied to the same converters as in Chapter 3. The origin
of the peculiar phenomena observed in the operation of the PCM-controlled
converters is fully explained. Extensive dynamic review is provided at the
end of the chapter based both on experimental and theoretical evidence.
The dynamic differences of the VM- and PCM-controlled buck and boost
converters in CCM are compared. The PCM control is widely applied in
controlling the switched-mode converters due to providing advantageous
features but its modeling is the subject of intensive discussions. Unanimously
accepted modeling method does not exist. The method presented in the book
is based on the natural processes taking place in the converter without any
kind of curve tting or similar approaches. Therefore, the resulting dynamic
models naturally represent the dynamics of the converter well and also
provide the scientically sound explanations for the phenomena observed in a
PCM-controlled converter.
 1.6 Content Review 13

The dynamic modeling of average-current-mode control (ACMC) is provided

in Chapter 5 and applied to a buck converter in CCM. The ACM control has
naturally similarities to the PCM control, because both of the control methods
use the inductor current for producing the duty ratio. Therefore, the presented
modeling method applies the PCMC modeling presented in Chapter 4. The
full ripple-feedback case is treated in detail. The effect of the high-frequency
pole in the inductor-current-loop amplier is discussed more in detail. The
comparisons between VM, PCM, and ACM control in a buck converter are
provided in order to highlight the dynamic changes and features the ACM
control would provide.
The dynamic modeling of self-oscillation control (i.e., variable-frequency
operation) is provided in Chapter 6 and applied to the basic converters. The
dynamic review is provided for buck and buckboost or yback converters.
The self-oscillation control is usually a form of PCM control requiring similar
modeling steps as the xed-frequency PCM control introduced in Chapter 4.
The dynamics associated with the current-output converters is treated
in Chapter 7. The dynamic review is provided for a buck converter under
VM and PCM controls based on the experimental and theoretical evidence.
The observed peculiar dynamical behavior is fully explained. The chapter
concentrates on the single-feedback-loop case but the method to solve the
dynamic modeling of cascaded cases is also provided.
The dynamic issues related to the interconnected systems are treated in
Chapter 8. The theoretical interaction formalism is briey reviewed. The con-
cepts of intermediate and inputoutput stabilities are consistently dened by
applying system theory and shown to be related to the impedance ratio known
as minor-loop gain. The analysis of the output-voltage remote-sensing and EMI-
lter effects are briey discussed and the theoretical formulation to treat them
is provided. Practical evidence is provided to support the theoretical ndings.
The control-related issues such as different controller implementations,
factors affecting the transient, responses and limiting the maximum loop
crossover frequency are discussed in Chapter 9. In addition, the dynamic
constraints related to the simple control systems based on the adjustable shunt
regulator TL431 are treated. Finally, a consistent method to shape the loop
gain in order to achieve the desired loop dynamics is proposed and veried
experimentally.
Dynamic modeling and analysis of a fourth-order converter known as two-
inductor buck, current-sourced buck, and superbuck converter is provided in
Chapter 10 as an example of the higher order converters and the possible
dynamic anomalies involved in them. The superbuck converter has similar
features as the conventional buck converter but its input and output currents
are continuous. The input-current ripple can be further reduced by coupling
the inductors. Consistent and easy-to-apply analysis methods for the coupled-
inductor technique are given and applied to the VM- and PCM-controlled
converter. Practical evidence on the dynamics of the PCM-controlled superbuck
converter is provided.
14 1 Introduction

References

1. S. Cuk, General topological properties 11. M. Hankaniemi, Dynamical prole of

of switching structures, in Proc. IEEE switched-mode converter fact of
Power Electronics Specialists Conf., ction, PhD Thesis, Tampere
1979, pp. 109130. University of Technology, 2007,
2. Y. Huang and C.K. Tse, Circuit Publication No. 687.
theoretical classication of parallel 12. T. Roinila, M. Hankaniemi, T. Suntio,
connected DCDC converters, IEEE M. Sippola, and M. Vilkko, Dynamical
Trans. Circuits Syst. I: Regular Papers, prole of a switched-mode converter
vol. 54, no. 5, 2007, pp. 10991108. Reality or imagination, in Proc. IEEE
3. R. Tymerski and V. Vorperian, International Telecommunications
Generation and classication of PWM Energy Conf., 2007, pp. 420427.
DC-to-DC converters, IEEE Trans. 13. M. Hankaniemi, T. Suntio, and
Aerosp. Electron. Syst., vol. 24, no. 6, M. Sippola, Characterization of
1988, pp. 743754. regulated converters to ensure stability
4. V.J. Thottuvelil, Modeling and and performance in distributed power
analysis of power converter systems supply system, in Proc. IEEE
with batteries, in Proc. IEEE International Telecommunications
International Telecommunications Energy Conf., 2005, pp. 533538.
Energy Conf., 1997, pp. 517522. 14. P.G. Maranesi, V. Tavazzi, and
5. T. Suntio, A. Glad, and P. Waltari, V. Varanoli, Two-port modeling of
Constant-current vs. constant-power PWM voltage regulators at low
protected rectier as a DC UPS frequencies, IEEE Trans. Indust.
systems building block, in Proc IEEE Electron., vol. 35, no. 3, 1988,
International Telecommunications pp. 444450.
15. C.K. Tse, Linear Circuit Analysis,
Energy Conf., 2006, pp. 227233.
Addison-Wesley Longman, Harlow,
6. T. Suntio, I. Gadoura, J. Lempinen,
UK, 1998.
and K. Zenger, Practical design issues
16. R.D. Middlebrook, Null double
of multiloop controller for a telecom
injection and the extra element
rectier, in Proc. IEEE Telecommuni-
theorem, IEEE Trans. Edu., vol. 32,
cations Energy Special Conf., 2000,
no. 3, pp. 167180, 1989.
pp. 197201.
17. T. Suntio, M. Hankaniemi, and
7. J. Lempinen and T. Suntio, Small-
M. Karppanen, Analysing the
signal modeling for design of robust dynamics of regulated converters, IEE
variable-frequency yback battery Proc. Electr. Power Appl., vol. 153,
chargers for portable device appli- no. 6, 2006, pp. 905910.
cations, in Proc. IEEE Applied Power 18. R. Severn and G. Bloom, Modern
Electronics Conf., 2001, pp. 548554. DC-to-DC Switch Mode Power Converter
8. A. Tenno, R. Tenno, and T. Suntio, Circuits, Van Nostrand Reinhold, New
Method for battery impedance York, USA, 1985.
analysis, J. Electrochem. Soc., vol. 151, 19. N. Mohan, T.M. Undeland, and W.P.
no. 6, 2004, pp. A806A824. Robbins, Power Electronics
9. S. Lyi and R.A. Dougal, Dynamics and Converters, Applications, and Design,
multiphysics model of solar array, John Wiley & Sons, New York, USA,
IEEE Trans. Energy Conv., vol. 17, 1989, 2nd Edition.
no. 2, 2002, pp. 285294. 20. J.G. Kassakian, M.F. Schlecht, and
10. D. Shmilovitz and S. Singer, A G.C. Verghese, Principles of Power
switched-mode converter suitable for Electronics, Addison-Wesley, Reading,
superconductive magnetic energy MA, USA, 1991.
storage (SMES) systems, in Proc. 21. P.T. Krein, Elements of Power
IEEE Applied Power Electronics Conf., Electronics, Oxford University Press,
2002, pp. 630634. New York, USA, 1998.
 References 15

22. R.W. Erickson and D. Maksimovic, modeling of PWM converters in

Fundamentals of Power Electronics, discontinuous modes, IEEE Trans.
Kluwer, Norwell, MA, USA, 2001, 2nd Power Electron., vol. 16, no. 4, 2001,
Edition. pp. 482492.
23. M.H. Rashid, Power Electronics 35. T. Suntio, Unied average and
Circuits, Devices and Applications, small-signal modeling of direct-
Pearson Education, Upper Saddle on-time control, IEEE Trans. Indust.
River, NJ, USA, 2004, 3rd Edition. Electron., vol. 53, no. 1, 2006,
24. S. Maniktala, Switching Power Supply pp. 287295.
Design & Optimization, McGraw-Hill, 36. D.M. Mitchell, Pulsewidth modulator
New York, USA, 2004. phase shift, IEEE Trans. Aerosp.
25. S. Ang and A. Oliva, Power-Switching Electron. Syst., vol. AES-16, no. 3, 1980,
Converters, Taylor & Francis, Boca pp. 272278.
Raton, FL, USA, 2005, 2nd Edition. 37. J. Sun, Small-signal modeling of
26. I. Batarseh, Power Electronic Circuits, variable-frequency pulsewidth
John Wiley & Sons, New York, USA, modulators, IEEE Trans. Aerosp.
2004. Electron. Syst., vol. 38, no. 3, 2002,
27. A.S. Kislovski, R. Redl, and N.O. pp. 11041108.
Sokal, Dynamic Analysis of Switching- 38. Y. Qiu, M. Xu, K. Yao, J. Sun, and F.C.
Mode DC/DC Converters, Van Lee, Multifrequency small-signal
Nostrand Reinhold, New York, USA, model for buck and multiphase buck
1991. converters, IEEE Trans. Power
28. D.M. Mitchell, DCDC Switching Electron., vol. 21, no. 5, 2006,
Regulator Analysis, DMMitchell pp. 11851192.
Consultants, Cedar Rapids, IA, USA, 39. D. Maksimovic, A.M. Stankovic, V.J.
1992. Tottuvelil, and G.C. Verghese,
29. J.Y. Zhu, Interpreting small signal Modeling and simulation of power
behavior of the synchronous buck electronic converters, Proc. IEEE,
converter at light load, IEEE Power vol. 89, no. 6, 2001, pp. 898912.
Electron. Lett., vol. 3, no. 4, 2005, 40. R.D. Middlebrook, Input lter
pp. 144147. considerations in design and
30. C.W. Deisch, Simple switching application of switching regulators, in
control method changes power Proc. IEEE Industry Application Society
converter into a current source, in Annual Conf., 1976, pp. 366382.
Proc. IEEE Power Electronics Special 41. S.D. Sudhoff, S.F. Glover, P.T. Lamm,
Conf., 1978, pp. 300306. D.H. Schmucker, and D.E. Delisle,
31. G.W. Wester and R.D. Middlebrook, Admittance space stability analysis
Low-frequency characterization of and power electronic systems, IEEE
switched-mode dcdc converters, Trans. Aerosp. Electron. Syst., vol. 36,
IEEE Trans. Aerosp. Electron. Syst., no. 3, 2000, pp. 965973.
vol. AES-9, no. 3, 1973, pp. 376385. 42. M. Hankaniemi, M. Karppanen, and
32. R.D. Middlebrook and S. Cuk, A T. Suntio, Load imposed instability
general unied approach to modeling and performance degradation in a
switching-converter power stages, Int. regulated converter, IEE Proc. Electr.
J. Electron., vol. 42, no. 6, 1977, Power Appl., vol. 153, no. 6, 2006,
pp. 521550. pp. 781786.
33. S. Cuk and R.D. Middlebrook, A 43. M. Karppanen, M. Sippola, and
general unied approach in modeling T. Suntio, Source-imposed instability
switching DC-to-DC converters in and performance degradation in a
discontinuous conduction mode, in regulated converter, in Proc. IEEE
Proc. IEEE Power Electronics Special Power Electronics Specialists Conf.,
Conf., 1977, pp. 3657. 2007, pp. 194200.
34. J. Sun, D.M. Mitchell, M.F. Greuel, 44. K. Zenger, A. Altowati, and T. Suntio,
P.T. Krein, and R.M. Bass, Average Dynamic properties of interconnected
16 1 Introduction

power systems A system theoretic Thesis, Virginia Polytechnic Institute

approach, in Proc. IEEE Industrial and State University, USA, 1985,
Electronics and Applications Conf., 181 pp.
2006, pp. 835840. 52. M. Hankaniemi and T. Suntio,
45. C. Gezgin, W.C. Bowman, and V.J. Small-signal models for constant-
Thottuvelil, A stability analysis tool current regulated converters, in Proc.
for DCDC converters, in Proc. IEEE IEEE Industrial Electronics Society
Applied Power Electronics Conf., 2003, Annual Conf., 2006, 20372042.
pp. 10141020. 53. M. Hankaniemi, M. Sippola, and
46. M. Karppanen, T. Suntio, and T. Suntio, Analysis of the load
M. Sippola, Impact of output-voltage interactions in constant-current-
remote sensing on converter controlled buck converter, in Proc.
dynamics, Int. Rev. Electr. Eng., vol. 2, IEEE International Telecommunications
no. 2, 2007, pp. 196202. Energy Conf., 2006, pp. 343348.
47. T. Suntio, K. Kostov, T. Tepsa, and 54. K. Ogata, Modern Control Engineering,
J. Kyyra, Using input invariance as a Prentice-Hall, Upper Saddle River, NJ,
method to facilitate system design in USA, 1997, 3rd Edition.
DPS applications, J. Circuits, Syst., 55. J.C. Basio and S.R. Matos, Design of
Comput., vol. 13, no. 4, 2004, PI and PID controllers with transient
pp. 707723. performance specications, IEEE
48. M. Karppanen, M. Hankaniemi, and Trans. Educ., vol. 45, no. 4, 2002,
T. Suntio, Load and supply invariance pp. 364370.
in a regulated converter, in Proc. IEEE 56. W.H. Tuttle, Relating converter
Power Electronics Specialists Conf, 2006, transient response characteristics to
pp. 26632668. feedback loop control design, in Proc.
49. M. Karppanen, M. Hankaniemi, Powercon 11, 1984, pp. 10.110.12.
T. Suntio, and M. Sippola, Dynamical 57. B. Choi, Step load response of a
characterization of peak-current- current-mode-controlled DC-to-DC
mode-controlled buck converter with converter, IEEE Trans. Aerosp.
output-current feedforward, IEEE Electron. Syst., vol. 33, no. 4, 1997,
Trans. Power Electron., vol. 22, no. 2, pp. 11151121.
2007, pp. 444451. 58. C. Gezgin, Predicting load transient
50. M. Karppanen, T. Suntio, and response of output voltage in DCDC
M. Sippola, Dynamical converters, in Proc. IEEE Applied
characterization of input-voltage- Power Electronics Conf., 2004,
feedforward-controlled buck pp. 13391343.
converter, IEEE Trans. Indust. 59. J. Betten and R. Kollman, Easy
Electron., vol. 54, no. 2, 2007, calculation yields load transient
pp. 10051013. response, Power Electronics Technology,
51. B.H. Cho, Modeling and analysis of February 2005, pp. 4048.
spacecraft power systems, PhD
 17

2
Basis for Dynamic Analysis and Control Dynamics

2.1
Introduction

The chapter introduces the basis behind the dynamic analysis and control
dynamics, denes the concepts of open- and closed-loop systems, and presents
the sets of transfer functions for voltage- and current-output converters used
to characterize their dynamics (i.e., G- and Y-parameters) in matrix form as a
two-port network and control-engineering block diagrams. The basic dynamic
representation is given to correspond to the true internal dynamics of the
associated converter. The methods to address the effect of a nonideal source and
load are shortly introduced. The control-engineering block diagrams are used to
derive the closed-loop representations for the output and input dynamics. The
stability and performance indices such as control bandwidth, loop crossover
frequency, phase and gain margins, sensitivity and complementary sensitivity
functions as well as instability, unconditional and conditional stabilities are
reviewed. The meaning and consequences of zeros and poles in the transfer
functions are explained. Especially, the right-half-plane zeros and poles are
discussed in the light of the control bandwidth constraints they provide.
A short review of matrix algebra is also provided.

2.2
Dynamic Representations at Open Loop

The open-loop converter is the basis for the models utilized in the control
design. The open- loop condition is dened to be such a condition where the
external feedback from the controlled output variable (i.e., usually the output
voltage or current) does not affect the dynamic behavior of the converter
[1]. The open-loop representations of the voltage-output and current-output
converters are shown in Figure 2.1. The voltage-output converter is such a
converter (Figure 2.1a) where the control system regulates the output voltage
(i.e., Figure 2.2, the external feedback is from the output voltage uo ). The
current-output converter is such a converter (Figure 2.1b) where the control

Dynamic Prole of Switched-Mode Converter. Teuvo Suntio

2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
ISBN: 978-3-527-40708-8
18 2 Basis for Dynamic Analysis and Control Dynamics

Figure 2.1 Converters at open

loop: (a) voltage-output converter
and (b) current-output converter.

system regulates the output current [3, 4]. Sometimes the converter may
change the operation mode from the voltage output to the current output
[4] to protect itself from damage due to excess load current generated for
example by a storage battery [5] when charged. The outer active feedback loop
determines the mode of operation. The converter may have internal feedback
or feedforward loops connected in order to change its dynamic behavior, but
despite these connections the converter is dened to operate at open loop,
whenever the outer external feedback loop is open as depicted in Figure 2.1.
Typical representatives of such a condition are the peak-current-mode
control (PCMC) [6] where the pulsewidth modulation (PWM) is accomplished
using the up-slope of the inductor current, the average-current-mode (ACM)
control [7], where the inductor current affects the modulation process through

Figure 2.2 A second-order

converter known as the buck or
step-down converter.
 2.2 Dynamic Representations at Open Loop 19

an average lter, and the input-voltage-feedforward (IVFF) control [8], where

the PWM ramp is modulated by means of the input voltage. All these control
modes are known to change the open-loop dynamics of the associated converter
compared to the corresponding direct-on-time or voltage-mode (VM) controlled
converter [911], where the PWM is accomplished using a constant-slope PWM
signal.
The dynamic description of the converter is usually given in such a way that
the load impedance (i.e., ZL in Figure 2.1, a resistor) is included in the derived
transfer functions [1214]. Such models are load-affected models, which do
not characterize the true internal dynamics of the converter and consequently
do not sufce for analyzing and predicting the converter behavior in its
practical environment, that is, as a subsystem in an interconnected system.
The models corresponding to the internal dynamics are called unterminated
models [1517]. They can be usually measured and/or derived by using an
ideal voltage source (i.e., Zs = 0 in Figure 2.1) at the input of the converter,
and an ideal current sink (i.e., ZL = in Figure 2.1a) or an ideal voltage
source (i.e., ZL = 0 in Figure 2.1b) as a load. The open-loop transfer functions
of the current-output converter cannot be usually measured due to the short-
circuit nature of the required ideal voltage source at output but only in
case of the converters having current-output nature at open loop such as
peak-current-mode or average-current-mode-controlled converters [6, 7]. The
practical electronic loads providing the voltage-source capability have usually
nonzero internal impedance (i.e., ZL > 0, Figure 2.1b), which would produce
load-affected transfer functions [2, 3].

2.2.1
State Space

The variables associated with the converter dynamics are called state variables,
which are usually the currents (iLi ) and voltages (uCk ) of the inductors and
capacitors incorporated in the power stage (Figure 2.1), input variables, which
are the input voltage (uin ), output current (io ), and the control variable (c) in the
case of the voltage-output converter (Figure 2.1a), and the input voltage (uin ),
the output voltage (uo ), and the control variable (c) in the case of the current-
output converter (Figure 2.1b) as well as the output variables, which are the
input current (iin ) and the output voltage (uo ) in the case of the voltage-output
converter (Figure 2.1a) and the input current (iin ) and the output current (io )
in the case of the current-output converter (Figure 2.1b). The ports, to which
the input variables are applied, are also known as disturbance inputs [1].
The switched-mode converters are variable-structure systems where the
topological circuit structure varies depending on the state of the semiconductor
switches in the power stage producing nonlinearity (Figure 2.2). The dynamics
associated with them can be effectively captured, however, by averaging the
converter behavior within one switching cycle and linearizing the averaged
20 2 Basis for Dynamic Analysis and Control Dynamics

behavior at the desired operating point [911]. The modeling issues are
discussed more in detail in the subsequent chapters.
The small-signal state-space representation is constructed by expressing the
derivatives of the state variables and the output variables as a function of the state
and input variables as illustrated in (2.1), where the linearized state space is
developed corresponding to a second-order voltage-output converter (i.e., the
power stage consists of one inductor and one capacitor),

diL
     uin
dt
= a11 a12 iL b
+ 11
b12 b13
io
a21 a22 uC b21 b22 b23
duC c
dt (2.1)
       uin
iin c11 c12 iL d11 d12 d13
uo
=
c21 c22 uC
+
d21 d22 d23 io
c
and in (2.2), where the linearized state space represents a similar current-
output converter, respectively. The hat over the variables denotes small-signal
behavior, that is, a small perturbation is applied to the variables at the dened
operating point.

 uin

diL
   
dt a a i b b b
= 11 12 L
+ 11 12 13
uo
a21 a22 uC b21 b22 b23
duC c
dt (2.2)

     uin

iin c11 c12 iL d11 d12 d13

= + uo
io c21 c22 uC d21 d22 d23
c
The state spaces in (2.1) and (2.2) are typically given in matrix form [18] by
x(t) = Ax(t) + Bu(t)
(2.3)
y(t) = Cx(t) + Du(t)
where x(t) is the state-variable vector, the dot over x(t) denotes derivative, u(t)
is the input vector, and y(t) is the output vector. The entries in the coefcient
matrices A, B, C, and D are scalars for a time-invariant linear system [18]. The
time-domain state space (2.3) can be solved in frequency domain by applying
Laplace transform with zero initial conditions yielding the input-to-output
description as
 
y(s) = C [sI A]1 B + D u(s) (2.4)

where I is an identity matrix and superscript 1 denotes the matrix inverse.

The matrix G(s) = C[sI A]1 B + D is called the transfer matrix constituting
the G- or Y-parameter representation of the converter inputoutput dynamics.
The use of G-parameters is well justied, because their existence is always
quarantined [19].
 2.2 Dynamic Representations at Open Loop 21

To solve the dynamics associated with the peak and average-current-mode

control may require to solve also the input-to-state description by applying
Laplace transformation to the state equation in (2.3) yielding

x(s) = [sI A]1 Bu(s) (2.5)

The matrix
(s) = [sI A]1 is called state-transition matrix, and its determi-
nant
(s) = |sI A| denes the denominator of the transfer functions in (2.4)
and (2.5). The roots of
(s) are known as the poles of the transfer function
dening the basic nature of the internal dynamics of the open-loop system,
and also its order (i.e., the order of the system is usually the number of state
variables).
The open-loop transfer matrices of voltage-output (2.6) and current-
output (2.7) converters are usually of the form
 
Yino Toio Gci
G(s) = (2.6)
Gioo Zoo Gco
 
Yino Toio Gci
G(s) = (2.7)
Gioo Yoo Gco

where
Yino = input admittance (i.e., iin
uin
)
Toio = reverse or output-to-input transfer function (i.e., iin
or iin
uo
)
io
Gci = control-to-input transfer function (i.e., iinc )
Gioo = forward, input-to-output, line-to-output transfer function or audio-
susceptibility (i.e., uuino or uioo )
Zoo = output impedance (i.e., uo )
io

Yoo = output admittance (i.e., io

uo
)
Gco = control-to-output transfer function (i.e., uo
c
or ico ).
The open-loop transfer matrices in (2.6) and (2.7) describe fully the dynamics
associated with the open-loop converter. Physically the G- and Y- parameters
of the voltage- and current-output modes differ from each other but the output
1
impedance and output admittance are related as Zoo = Yoo . The differences
are described more in detail in the next subsection.

2.2.2
Two-Port Models

The internal dynamics of the voltage-output converter can also be represented

equally to (2.6) by means of a linear two-port network, Figure 2.3, where the
input port is a Norton equivalent circuit corresponding to

iin = Yino uin + Toio io + Gci c (2.8)

22 2 Basis for Dynamic Analysis and Control Dynamics

Figure 2.3 Two-port model of a

voltage-output converter.

Figure 2.4 Two-port model for a

current-output converter.

and the output port a Thevenin equivalent circuit corresponding to

uo = Gioo uin Zoo io + Gco c (2.9)

Similarly, the dynamics of the current-output converter can be represented

by using a two-port network, Figure 2.4, where the input port is a Norton
equivalent circuit corresponding to

iin = Yino
i
uin + Toio
i
uo + Gici c (2.10)

and the output port a similar Norton equivalent circuit corresponding to

io = Giioo uin Yoo uo + Gico c (2.11)

The relation between the voltage (2.6) and current-output (2.7) parameters
can be found by applying the Thevenin-to-Norton transformation technique
dened in [19] to the output port of the voltage-output-converter model in
Figure 2.3 yielding [2]
i i

Yino Toio Gici

G (s) =
i
Giioo Yoo Gico
Gioo Toio Toio Gco Toio
Yino + Gci +
Zoo Zoo Zoo
=

(2.12)
Gioo 1 Gco

Zoo Zoo Zoo
 2.3 Dynamic Representations at a Closed Loop 23

Figure 2.5 Control-block diagrams at open loop for a voltage-output converter:

(a) output dynamics and (b) input dynamics.

where the superscript i denotes the transfer function of the current-output

converter. Equation (2.12) shows that the internal dynamics of the voltage-
output converter would change radically, when it enters into the current-output
mode of operation [3].

2.2.3
Control-Block Diagrams

The input and output dynamics of the converter can also be represented
by using the control-engineering block diagrams [1] shown for a voltage-
output converter in Figure 2.5 and for a current-output converter in
Figure 2.6, respectively. The key for constructing the representations is
Eqs. (2.8)(2.11). The block diagrams are useful in deriving the closed-loop
dynamic representations.

2.3
Dynamic Representations at a Closed Loop

The closed-loop converter is naturally a converter, where the external feedback

loop or loops are connected as depicted in Figure 2.7. In practice, the converter

Figure 2.6 Control-block diagrams for a current-output converter: (a) output dynamics
and (b) input dynamics.
24 2 Basis for Dynamic Analysis and Control Dynamics

Figure 2.7 Closed-loop converter.

may have a multiloop control system, where the voltage loop is the main
or inner loop, and the current loop is the outer loop activated in the case
of excess load current by reducing the voltage-loop reference in order to
make the output power or current constant as described in detail in [4].
The overload protection is typically needed in the applications where the
converter has to recharge a storage battery [5]. The multiloop control system
is naturally implemented in a current-output converter in such a way that the
output voltage and current are interchanged compared to the voltage-output
converter.
The loop gain of a converter is denoted by Li (s), where the subscript i
denes whether the loop is related to the output voltage (i.e., v) or the
output current (i.e., c), respectively. The internal loop gain consists of the
dynamic elements along the closed loop from the feedback variable back
to the feedback variable (Figure 2.7) such as the sensor gain, error amplier,
the PWM process, and the power stage depending on the application: In
the isolated converters, the PWM process is usually located in the primary
side of the converter, and the control system in the secondary side requiring
also to isolate the control loop [2022] for safety reasons. A typical control-
loop isolation medium is an optocoupler, which would affect the dynamical
behavior and has to be carefully considered [22]. Sometimes the simple control
 2.3 Dynamic Representations at a Closed Loop 25

Figure 2.8 The closed-loop control block diagrams of the voltage-output converter for
(a) the output dynamics and (b) the input dynamics.

Figure 2.9 The closed-loop control-block diagrams of the current-output converter for
(a) output dynamics and (b) input dynamics.

systems based for example on the popular shunt regulator TL431 are actually
extremely complicated systems requiring special attention as discussed in
[2326].
The closed-loop dynamic representation of the converter can be solved most
conveniently applying the corresponding open-loop block diagrams dened
in Figures 2.5 and 2.6 as shown in Figures 2.8 and 2.9. The loop gain can be
given generally by

Li (s) = Gse Gcc Ga Gco (2.13)

where the transfer functions correspond to the control-system blocks dened

in Figure 2.7 starting from the sensor gain Gse , respectively.
26 2 Basis for Dynamic Analysis and Control Dynamics

2.3.1
Voltage-Output Converter

The closed-loop output voltage (i.e., uo , Figure 2.8a) can be computed to be

Gioo Zoo
uo = uin io
1 + Gse Gcc Ga Gco 1 + Gse Gcc Ga Gco
Gcc Ga Gco
+ ur (2.14)
1 + Gse Gcc Ga Gco
The application of the loop-gain denition (2.13) yields
Gioo Zoo 1 Lv (s)
uo = uin io + ur (2.15)
1 + Lv (s) 1 + Lv (s) Gse 1 + Lv (s)

The closed-loop input current (i.e., iin , Figure 2.8b) can be computed to be

iin = Yino uin + Toio io Gse Gcc Ga Gci uo + Gcc Ga Gci ur (2.16)

The output voltage in (2.16) has to be substituted with (2.14) yielding with the
application of the loop-gain denition (2.13)

Gioo Gci Lv (s)
iin = Yino uin
Gco 1 + Lv (s)
 (2.17)
Zoo Gci Lv (s) Gci Lv (s)
+ Toio + io + ur
Gco 1 + Lv (s) Gse Gco 1 + Lv (s)
If the voltage reference (ur ) is constant as is usually the case in single-loop
converters, (2.15) and (2.17) reduce to
Gioo Zoo
uo = uin io
1 + Lv (s) 1 + Lv (s)

Gioo Gci Lv (s)
iin = Yino uin (2.18)
Gco 1 + Lv (s)

Zoo Gci Lv (s)
+ Toio + io
Gco 1 + Lv (s)
which denes the usual closed-loop transfer matrix of the voltage-output
converter to be
 
Yinc Toic
G(s) =
Gioc Zoc
Gioo Gci Lv (s) Zoo Gci Lv (s)
Yino Toio +
Gco 1 + Lv (s) Gco 1 + Lv (s)
=
Gioo Zoo

1 + Lv (s) 1 + Lv (s)
(2.19)
 2.3 Dynamic Representations at a Closed Loop 27

In the case of the multiloop operation, the full-order representations in (2.15)

and (2.17) have to be applied. The two-port network dened in Figure 2.3 would
equally represent also the closed-loop voltage-output converter, when the input
port is dened using (2.17) and the output port using (2.15), corresponding to
the following transfer matrix:
 
Yinc Toic Gcic
G(s) =
Gioc Zoc Gcoc
Gioo Gci Lv (s) Zoo Gci Lv (s) Gci Lv (s)
Yino Toio +
Gco 1 + Lv (s) Gco 1 + Lv (s) Gse Gco 1 + Lv (s)
=


Gioo Zoo 1 Lv (s)

1 + Lv (s) 1 + Lv (s) Gse 1 + Lv (s)
(2.20)
where Gcoc and Gcic are the reference-to-output and reference-to-input
transfer functions, respectively. In the case of the cascaded control system, the
current-output model can be derived from (2.20) utilizing (2.12).

2.3.2
Current-Output Converter

The closed-loop output current (i.e., io , Figure 2.9a) can be computed to be

Giioo i
Yoo
io = uin uo
1 + Gise Gicc Gia Gico 1 + Gise Gicc Gia Gico
Gicc Gia Gico
+ ui (2.21)
1 + Gise Gicc Gia Gico r
The application of the loop-gain denition (2.13) yields
Giioo i
Yoo 1 Lc (s)
io = uin uo + i ui (2.22)
1 + Lc (s) 1 + Lc (s) Gse 1 + Lc (s) r

The closed-loop input current (i.e., iin , Figure 2.9b) can be computed to be

iin = Yino
i
uin + Toio
i
uo Gise Gicc Gia Gici io + Gicc Gia Gici uri (2.23)

The output current in (2.23) has to be substituted with (2.22) yielding with the
application of the loop-gain denition (2.13)
 
G i
ioo G i
ci L c (s)
iin = Yino
i
uin
Gico 1 + Lc (s)
 
i
Yoo Gici Lc (s) Gi Lc (s)
+ Toio +
i
uo + i ci i ui
Gcoi 1 + Lc (s) Gse Gco 1 + Lc (s) r
(2.24)
 28 2 Basis for Dynamic Analysis and Control Dynamics

If the current reference (uri ) is constant as is usually the case in single-loop

converters, (2.22) and (2.24) reduce to
Giioo i
Yoo
io = uin uo
1 + Lc (s) 1 + Lc (s)
 
Giioo Gici Lc (s)
iin = Yino
i
uin
Gico 1 + Lc (s)
 
i
Yoo Gici Lc (s)
+ Toio +
i
uo (2.25)
Gico 1 + Lc (s)

which denes the usual closed-loop transfer matrix of the current-output

converter to be
i i

Yinc Toic
G (s) =
i
Giioc Yoc
i

Gi Gi Lc (s) Y i Gi Lc (s)
i
Yino iooi ci i
Toio + ooi ci
Gco 1 + Lc (s) Gco 1 + Lc (s)
=


i i
Gioo Yoo

1 + Lc (s) 1 + Lc (s)
(2.26)
In the case of multiloop operation, the full-order representations in (2.22)
and (2.24) have to be applied. The two-port network dened in Figure 2.4
would equally represent also the closed-loop current-output converter when
the input port is dened using (2.24) and the output port using (2.22) yielding
the following input-to-output transfer matrix:
i i

Yinc Toic Gicic

G(s) =
Giioc Yoc
i
Gicoc

Giioo Gici Lc (s) i
Yoo Gici Lc (s) Gici Lc (s)
Yino +
i i
Toio
Gico 1 + Lc (s) Gico 1 + Lc (s) Gise Gico 1 + Lc (s)
=
Giioo i
Yoo 1 Lc (s)

1 + Lc (s) 1 + Lc (s) Gise 1 + Lc (s)
(2.27)

2.4
Load and Source Effects

The previous sections treated the pure internal dynamics from which all the
effects of nonideal load and source were removed. In practice, the open-
loop transfer functions describing the pure internal dynamics can be usually
 2.4 Load and Source Effects 29

Figure 2.10 Two-port models

with nonideal source (i.e.,Zs , uins )
and nonideal load (i.e.,ZL , jo ):
(a) voltage-output converter and
(b) current-output converter.

measured by supplying the converter by means of an ideal voltage source,

and loading the converter with an ideal current sink or voltage source [26].
The peak-current-mode controlled converter [6] is a current source at open
loop, and therefore, the voltage-output transfer functions cannot be measured
as described above but using a resistor as load [27]. The internal dynamics
determines the external interactions, and therefore, knowing it is of prime
importance. Actually computational methods have to be applied in such cases
as instructed in [27].
It is well known that the nonideal source and load may affect the dynamics of
the converter [28, 29]. Consequently, its dynamical performance in the practical
applications may differ signicantly from that measured in the laboratory. The
two-port models introduced earlier can be used to solve analytically the load
and source effects. Circuit theory provides the media to solve the required
equations easily. Basically, the origins of the load and source effects are the
impedances associated with the load and source as depicted in Figure 2.10.
We provide here only the analytical equations describing the interactions, and
the more detailed analysis would be provided later in Chapter 8.

2.4.1
Voltage-Output Converter

The two-port model of the voltage-output converter (Figure 2.10a) is character-

ized by means of uo and iin as follows:

uo = Gioo uin Zoo io + Gco c

(2.28)
iin = Yino uin + Toio io + Gci c
30 2 Basis for Dynamic Analysis and Control Dynamics

The load effect on the converter dynamics can be solved by computing io from
Figure 2.10a, which yields

Gioo u in + Gco c ZL jo
io = (2.29)
Zoo + ZL
Substituting it in (2.28) with (2.29) yields the desired load-affected dynamic
model of the voltage-output converter as follows:
Gioo Toio ZL Toio Gco Toio
Yino + Gci +
  Zoo + ZL Zoo + ZL Zoo + ZL uin
iin
= Gioo Zoo Gco jo
uo
Zoo Zoo Zoo c
1+ 1+ 1+
ZL ZL ZL
(2.30)
The source effect on the converter dynamics can be solved by computing uin
from Figure 2.10a, which yields
1 Zs
uin = uins (Toio io + Gci c) (2.31)
1 + Zs Yino 1 + Zs Yino
Substituting it in (2.28) yields the source-affected dynamic model for the
voltage-output converter as follows:
Yino Toio Gci
 
iin 1 + Zs Yino 1 + Zs Yino 1 + Zs Yino
= 
uo Gioo Zs Gioo Toio Zs Gioo Gci
Zoo + Gco
1 + Zs Yino 1 + Zs Yino 1 + Zs Yino

uins
i o (2.32)
c
The form of (2.32) is not most convenient but can be transformed to correspond
the formulation given in [30] yielding (2.33), where the special admittances
Yin (ideal input admittance) and Yinsc (short-circuit input admittance) are
dened in (2.34), respectively.
Yino Toio Gci
  uins
iin 1 + Zs Yino 1 + Zs Yino 1 + Zs Yino
= io
uo Gioo 1 + Zs Yinsc 1 + Zs Yin
Zoo Gco c
1 + Zs Yino 1 + Zs Yino 1 + Zs Yino
(2.33)
Gioo Gci
Yin = Yino
Gco
(2.34)
Gioo Toio
Yinsc = Yino +
Zoo
 2.4 Load and Source Effects 31

2.4.2
Current-Output Converter

The two-port model of the current-output converter (Figure 2.10b) is

characterized by means of io and iin as follows, where the superscript i
denotes the current-output transfer function:
io = Giioo uin Yoo
i
uo + Gico c
(2.35)
iin = Yino uin + Toio uo + Gici c
i i

The load effect on the converter dynamics can be found by computing uo from
Figure 2.10b, which yields
ZL   1
uo = Giioo uin + Gico c + eo (2.36)
1 + ZL Yoo
i
1 + ZL Yoo
i

Substituting it in (2.35) yields the desired load-affected dynamic models of the

current-output converter as follows:
ZL Giioo Toio
i i
Toio i
ZL Gico Toio

ino Y i
+ G i
+
iin 1 + Yoo
i
ZL 1 + Yoo
i
ZL ci
1 + Yoo
i
ZL

= i
Gioo 1 i
Gco

io
ZL ZL + (Yoo
i
)1 ZL
1 + i 1 1 + i 1
(Yoo ) (Yoo )
u

in
eo (2.37)
c
The load-affected current-output model (2.37) can also be given by using the
transfer functions of the corresponding voltage-output converter, which would
give much more information due to the usually well-known transfer functions
applying the information given in Section 2.2.2 (Eq. (2.12)) [2]:
Gioo Toio Toio Gco Toio
Yoo + Gci +

ZL + Zoo ZL + Zoo ZL + Zoo

iin Gioo Gco
=

io Zoo 1 Zoo

ZL ZL + Zoo ZL
1+ 1+
Zoo Zoo
u

in
eo (2.38)
c
The source effect on the converter dynamics can be found by computing uin
from Figure 2.10b, which yields
1 Zs
uin = uins (Toio
i
uo + Gici c) (2.39)
1 + Zs Yino
i
1 + Zs Yino
i
32 2 Basis for Dynamic Analysis and Control Dynamics

Substituting it in (2.35) yields the source-affected dynamical models as

follows:
i i
Yino Toio Gici

iin 1 + Z Yi 1 + Zs Yino
i
1 + Zs Yino
i
s ino
 
= i
io Giioo Z s G i
ioo T i
oio Z s G i
ioo G ci
Yooi
+ Gico
1 + Zs Yino
i
1 + Zs Yinoi
1 + Zs Yino
i

ins
uo (2.40)
c

The models in (2.40) can be transformed to the formulation described in [30]

yielding
i
Yino i
Toio Gici

iin 1 + Zs Yino
i
1 + Zs Yino
i
1 + Zs Yino
i
=



io Giioo 1 + Zs Yinoc
i
1 + Zs Yin
i
Yoo
i
Gico
1 + Zs Yino
i
1 + Zs Yino
i
1 + Zs Yino
i
u

ins
uo (2.41)
c
i i
where the special admittances Yin (ideal input admittance) and Yinoc
(open-circuit input admittance) are dened as follows.

Giioo Gici
i
Yin = Yino
i

Gico
(2.42)
Giioo Toio
i
i
Yinoc = i
Yino + i
Yoo

The source-affected current-output model (2.41) can also be given by using

the transfer functions of the corresponding voltage-output converter, which
may be more informative compared to (2.41) due to the usually well-known
transfer functions, applying the information given in Section 2.2.2 (Eq. 2.12):
Toio Gco Toio
i Gci +

Yino Zoo Zoo

iin 1 + Zs Yino
i
1 + Zs Yino
i
1 + Zs Yino
i
=
Gioo


io
Zoo 1 + Zs Yino 1 1 + Zs Yin Gco

1 + Zs Yino
i
1 + Zs Yino Zoo
i
1 + Zs Yino
i Zoo
u

ins
uo (2.43)
c
 2.5 An Example LC Circuit 33

where the ideal input admittance Yin is dened in (2.34). The ideal
i
current-output input admittance Yin equals the corresponding ideal input
admittance Yin of the voltage-output converter. The open-circuit input
admittance (Yinoc
i
) equals the internal open-loop input admittance (Yino ) of
the corresponding voltage-output converter, respectively.

2.5
An Example LC Circuit

We derive the state space for the system comprising of a LC circuit both at
voltage-output (Figure 2.11a) and current-output (Figure 2.11b) modes, and
we solve its input-to-output (G(s)) and input-to-state (
(s)) descriptions as
an example to illustrate the use of the theoretical formulations dened in the
previous sections.
According to Section 2.2, the state variables of the system are usually the
inductor current (iL ) and the capacitor voltage (uC ), and their derivatives have
to be evaluated as a function of the state variables and the input variables. The
input variables of the voltage-output circuit, Figure 2.11a, are the input voltage
(uin ) and the output current (io ). The input variables of the current-output
circuit, Figure 2.11b, are the input voltage (uin ) and the output voltage (uo ). In
addition to the derivatives of the state variables, the output variables have to
be dened as a function of the state and input variables. The output voltage
(uo ) and the input current (iin ) are the output variables of the voltage-output
circuit, and the output current (io ) and the input current (iin ) are the output
variables of the current-output circuit, respectively. According to the circuit
theory [19], the voltage across the inductor can be given by uL = L didtL and the
current through the capacitor by iC = C dudtC , which explicitly dene the desired
derivatives of the state variables.

2.5.1
Voltage-Output Circuit

In order to construct the state-space representation for the system of

Figure 2.11a, we have to solve uL and uo applying Kirchhoff s voltage law

Figure 2.11 An example LC circuit: (a) voltage-output mode and (b) current-output mode.
34 2 Basis for Dynamic Analysis and Control Dynamics

yielding

uL = uin rL iL uo
(2.44)
uo = uC + rC iC

and iC and iin applying Kirchhoff s current law yielding

iC = iL io
(2.45)
iin = iL

According to (2.44) and (2.45), the state space can be solved to be

diL rL + rC 1 1 rC
= iL uC + uin + io
dt L L L L
duC 1 1
= iL io (2.46)
dt C C
iin = iL
uo = rC iL + uC rC io

The LC circuit is a linear system. Therefore, the state space (2.46) equally
represents also the small-signal state space without any additional actions, and
the desired transfer matrices can be solved directly applying Laplace transforms
on it. We transform the state space (2.46) to the usual matrix form according
to (2.3) yielding

diL rL + rC 1 1 rC

 
dt L L iL L L uin
= +
1 uC 1
duC io
0 0
dt C C
(2.47)
       u

iin 1 0 iL 0 0 in
= +
uo rC 1 uC 0 rC io

In order to solve the state-transition and transfer matrices, the Laplace

transform is applied to the small-signal state space (2.47) yielding
rL + rC 1 1 rC
   

iL (s) L L iL (s) L L uin (s)

s = +
uC (s) 1 uC (s) 1 io (s)
0 0
C C
(2.48)
       u (s)

iin (s) 1 0 iL (s) 0 0 in

= +
uo (s) rC 1 uC (s) 0 rC io (s)
 2.5 An Example LC Circuit 35

Applying matrix manipulation techniques to (2.48) yields the state-transition

matrix
(s)
1
s
L
1 rL + rC
s+
C L

(s) = (2.49)
rL + rC 1
s2 + s +
L LC
the transfer matrix G(s)
s 1 + srC C
L LC

1 + srC C (rL + sL)(1 + srC C)
 
Yin Toi LC LC
G(s) = = (2.50)
Gio Zo rL + rC 1
s2 + s +
L LC
and the characteristic equation (s)
rL + rC 1
(s) = s2 + s + (2.51)
L LC
The equivalent series resistances (ESR) rL and rC have to be small for an
effective lter, and therefore, the roots of the characteristic polynomial are
complex. As a consequence, the LC-circuit system, Figure 2.11a, would exhibit
resonant-type behavior.
The transfer functions in (2.49) and (2.50) can be easily derived directly
from the circuit without the state-space methods by applying superposition
principle.

2.5.2
Current-Output Circuit

In order to construct the state-space representation for the system of

Figure 2.11b, we have to solve uL and iC by applying Kirchhoff s voltage
law yielding

uL = uin rL iL uo
(2.52)
uo uC
iC =
rC
and io and iin by applying Kirchhoff s current law yielding

io = iL iC
(2.53)
iin = iL
36 2 Basis for Dynamic Analysis and Control Dynamics

According to (2.52) and (2.53), the state space can be solved to be

diL rL 1 1
= iL + uin uo
dt L L L
duC 1 1
= uC + uo
dt rC C rC C
(2.54)
iin = iL
1 1
io = iL + uC uo
rC rC
We transform the state space (2.54) to the usual matrix form according to (2.3)
yielding
r 1
diL L 1
0    
dt L iL L L uin
= +
1 uC 1 uo
duC 0 0
dt rC C rC C
(2.55)

1 0   0 0

 
iin iL uin
= 1 u + 1 u
io 1 C 0 o
rC rC
In order to solve the state-transition and transfer matrices, the Laplace
transform is applied to the small-signal state space (2.55) yielding
rL 1 1
  0    
i (s) iL (s) L L uin (s)
s L = L 1 +
uC (s) 0 uC (s) 1 uo (s)
rC C 0
rC C
(2.56)
  1 0   0 0  
iin (s) iL (s) uin (s)
= 1 u (s) + 1 u (s)
uo (s) 1 C 0 o
rC rC
Applying matrix manipulation techniques to (2.56) yields the state-transition
matrix
(s),
 
1 + srC C (1 + srC C)
0 rL + sL

(s) = (2.57)
(rL + sL)(1 + srC C)
the transfer matrix G(s)

1 + srC C (1 + srC C)

rL + rC 1
i

1 + srC C (s2 + s + )
Yin Toii L LC
G(s) = = (2.58)
Giio Yoi (rL + sL)(1 + srC C)
 2.6 Review of Basic Mathematical Tools 37

and the characteristic equation (s)

(s) = (rL + sL)(1 + srC C) (2.59)

The roots of the characteristic equation (2.59) are real, and all the other transfer
functions except the output admittance (Yoo i
) are clearly rst-order transfer
functions. The output admittance would, however, exhibit resonant behavior
similar to that of the output impedance (Zoo ) of the voltage-output circuit,
because Yoi = 1/Zoo .
According to Section 2.2.2, the current-output transfer functions can be
derived from the corresponding transfer functions of the voltage-output circuit
by applying (2.60)
Gioo Toio Toio
i
Yino +
i
Yino Toio Zoo Zoo
Gi (s) = =
(2.60)
Giioo Yooi Gioo 1

Zoo Zoo
which yields

1 + srC C (1 + srC C)

rL + rC 1
1 + srC C s2 + s +
L LC
Gi (s) = (2.61)
(rL + sL) (1 + srC C)
The transfer functions in (2.61) are obviously equal to (2.58). It may be
obvious that the change of mode from voltage output to current output would
signicantly change the dynamics of the associated system, and would be
induced by the change of load type.

2.6
Review of Basic Mathematical Tools

In this subsection, we will review some basic mathematical and control-

engineering concepts, which are vital to produce dynamical models and to
understand the dynamics associated with a converter.

2.6.1
Linearization

The models representing the converter dynamics are constructed by linearizing

the averaged behavior around the dened operating point. In power electronics,
the recommended method is to substitute the averaged values by means of
the sum of a DC value and small perturbation [14]. This method works well
38 2 Basis for Dynamic Analysis and Control Dynamics

when the averaged model is only slightly nonlinear as in the case of continuous
mode of operation, but usually fails when the average model is highly nonlinear
as in the case of discontinuous mode of operation. In control engineering [31],
the linearization is typically carried out computing the Jacobian matrix of
the function y = f (t, x), where x contains the state and input variables. The
Jacobian matrix represents the partial derivatives of the function with respect
to all the variables. In the one-dimensional case, the linearized function y can
be given as follows:
y y
y = x1 + + xn
x1 xn
x
1
 f f 

y = (t, X) (t, X) (2.62)
x1 xn

xn
where X in the Jacobian matrix contains the steady-state values of the
corresponding variables at the operating point. As an example, we consider
u2 i
the function y = uinC L , which is highly nonlinear. We consider that the steady-
state values of the variables uin , iL , and uC are Uin , IL , and UC , respectively.
According to these assumptions, the linearized function y can be given by
2
Uin U 2 IL 2Uin IL
y = iL in2 uC + uin
UC UC UC

In practice, this means that we treat each variable at a time and consider
the other variables to be constant when developing the required derivative
according to the basic mathematics.

2.6.2
Transfer Functions

A transfer function is usually given as a ratio of two polynomials in s (2.63),

where s is the Laplace variable. The roots of the numerator polynomial zi are
called zeros, and the roots of the denominator polynomial pi poles. The zeros
and poles may be real or complex numbers and they are given with respect
to angular frequency (rad/s). The frequency (f ) in Hz and the angular
frequency () in rad/s are related by f = /2:

an sn + an1 sn1 + + ao
G(s) =
bm sm + bm1 sm1 + + bo
(2.63)
(s z1 )(s z2 ) (s zn )
G(s) = K
(s p1 )(s p2 ) (s pm )
 2.6 Review of Basic Mathematical Tools 39

The magnitude of the transfer function (|G(s)|) is commonly expressed in

dB, that is, |G(s)|dB = 20 log10 (|G(s)|), and the phase ( G(s)) in degrees. The
logarithmic magnitude means that the combined effect of zeros and poles
can be found adding together the dB-values of the zeros and subtracting the
dB-values of the poles, respectively. The phase of the transfer function can
be found similarly adding together the phase contributions of the zeros and
subtracting the phase contributions of the poles. The zeros and poles locating
closest to the origin (i.e., zero frequency) are called dominant zeros and poles
having the strongest effect on the time-domain behaviour of the corresponding
system.

2.6.2.1 Single Zero

Let us consider the single-zero
 transfer function G(s) = 1 + s/z . Its
2
magnitude (i.e., |G(j)| = 1 + 2 ), and phase (i.e.,  G(j) = arctan
 z

z
), may be given at certain interesting angular-frequency points as
follows:
z
=
10

|G(j)| = 1.01 = 20 log10 ( 1.01) = 0.04 dB

 G(j) = arctan(0.1) = 5.7
= z

|G(j)| = 2 = 20 log10 ( 2) = 3 dB

 G(j) = arctan(1) = 45
= 10 z

|G(j)| = 10 = 20 log10 (10) = 20 dB

 G(j) = arctan(10) = 84.3

As a consequence, the behavior of the single-zero transfer functions is typically

considered to be such that its magnitude is unity (i.e., 0 dB) up to = z ,
and starts increasing at a slope of +20 dB/decade (i.e., +20 dB for a 10 times
increase in frequency) from = z . Its phase is zero up to = z /10,
and starts increasing at a slope of +45 /decade up to 10z , and stays
constant at +90 after that. The phase is exactly equal to +45 at = z .
If G(s) = 1 s/z , the magnitude behavior is same but the phase has the
negative sign compared to G(s) = 1 + s/z . s = z is called left-half-plane
(LHP) zero and s = z right-half-plane (RHP) zero, based on their locations in
the complex plane. The RHP zero would have a profound impact on the control
loop design, and would exist also in the certain-type of the switched-mode
converters.
40 2 Basis for Dynamic Analysis and Control Dynamics

2.6.2.2 Single Pole

Lets consider the single-pole
 transfer function G(s) = (1 + s/p )1 . Its magni-
 
tude (i.e., |G(j)| = 1/ 1 + 2 ), and phase (i.e.,  G(j) = arctan z ), may
2

z
be given at certain interesting angular-frequency points as follows:
z
=
10

|G(j)| = 1/ 1.01= 20 log10 ( 1.01) = 0.04 dB

 G(j) = arctan(0.1) = 5.7
= z

|G(j)| = 1/ 2= 20 log10 ( 2) = 3 dB

 G(j) = arctan(1) = 45
= 10 z

|G(j)| = 1/10= 20 log10 (10) = 20 dB

 G(j) = arctan(10) = 84.3

As a consequence, the behavior of the single-pole transfer functions is typically

considered to be such that its magnitude is unity (i.e., 0 dB) up to = z , and
starts decreasing at a slope of 20 dB/decade from = z . Its phase is zero up
to = z /10, and starts decreasing at a slope of 45 /decade up to 10z , and
stays constant at 90 after that. The phase is exactly equal to 45 at = z .
If G(s) = (1 s/p )1 , the magnitude behavior is same but the phase has
positive sign compared to G(s) = (1 + s/p )1 .s = p is called left-half-plane
(LHP) pole and s = p right-half-plane (RHP) pole based on their location in
the complex plane. An RHP pole means that the corresponding system is
unstable.
Most important is to recognize that the zeros and poles in the transfer
functions are given in the angular frequency and 6.28 f ; the phase of the
transfer function starts changing already 10 times earlier than the location of
the corresponding zero or pole, and keep on changing up to 10 times higher
frequencies than the location of the zero or pole. These facts are forgotten very
often.

2.6.2.3 Second-Order Transfer Function

The second-order polynomial is typically expressed as s2 + s 2 n + n2 or as
s2 + s Qn + n2 comprising either zeros or poles in the corresponding transfer
function, where is called damping factor, n undamped natural frequency,
and Q quality factor. The second-order polynomials are common in power
electronics, and will be recognized because of the special features they will
bring to the control design, the load and source interactions, and so on.
 2.6 Review of Basic Mathematical Tools 41

The roots of the second-order polynomial can be expressed as s1,2 =

n n 2 1. The system characterized by it (i.e., the second-order
polynomial is the denominator) may be classied according to the value of the
damping factor as (1) underdamped, when 0 < < 1, (2) critically damped,
when = 1, (3) overdamped, when > 1, and (4) oscillatory, when = 0. The
examples in Figure 2.12 correspond to the case where the transfer function is
of the form n2 /(s2 + s 2 n + n2 ).

Figure 2.12 Effect of damping factor on (a) step response,

(b) magnitude of the frequency response, and (c) phase of the
frequency response of a second-order transfer function of the form
n2 /(s2 + s 2 n + n2 ).
42 2 Basis for Dynamic Analysis and Control Dynamics

Figure 2.12 (Continued).

1. Underdamped Case: 0 < < 1

The roots of the second-order function are complex conjugates
 of each other,
which can be expressed by s1,2 = n jd , where d = n 1 2 is called
damped natural frequency. Step response of such a system applied to the
reference input includes decaying oscillatory behavior, where the oscillation
frequency is d (Figure 2.12, = 0.1).
2. Critically Damped Case: = 1
The second-order function has a real double root, which can be expressed
by s1.2 = n . The step response resembles an exponential response but is
faster (Figure 2.12, = 1).
3. Overdamped Case: > 1
The second-orderfunction has two real roots, which can be expressed by
s1,2 = n n 2 1. The greater the damping factor, the more the roots
are separated from each other. The step response of such a system contains
only exponential behavior (Figure 2.12, = 2).
4. Oscillatory Case: = 0
The second-order function has two pure imaginary roots, which can be
expressed by s1,2 = jn . The system will oscillate at the underdamped natural
frequency n (Figure 2.12, = 0).

The damping factor affects the system step response (Figure 2.12a)
and the behavior of the corresponding transfer function as shown in
Figures 2.12b and c, where the second-order system is assumed to be of
the form G(s) = n2 /(s2 + s 2 n + n2 ). The damping factor = 0 would
produce an innite value in the magnitude of the transfer function, and
therefore, = 0.001 is used in Figures 2.12b and c to demonstrate the effect
 2.6 Review of Basic Mathematical Tools 43

of the decrease in the damping factor instead of zero. Characteristic to the

magnitude of the second-order transfer function is that its high-frequency-
magnitude slope is -40 dB/decade and the overall change in the phase is
180 depending on whether the poles are LHP or RHP, respectively. If
the second-order polynomial is a numerator polynomial in the corresponding
transfer function, then its effect is 40 dB/decade and 180 , respectively. If
the zeros or poles are complex, then there exists either dipping or peaking in
the magnitude response (Figure 2.12b) and the phase changes rapidly 180 .
Such a behavior is called resonant behavior. The resonant frequency (o ) (i.e.,
the undamped natural frequency n ) can be found at a frequency where the
phase has changed exactly 90 (Figure 2.12c). The maximum peak/dip value
of the magnitude of the resonant-type transfer function corresponds to the
 = 1/2 (i.e., the difference between 0 dB and the peak/dip
quality factor Q
value) at = 1 2 2 n . This means that the maximum peak/dip value
does not locate
 at the resonant frequency but at a slightly lower frequency (i.e.,
n versus 1 2 2 n ).

2.6.2.4 Example
The practical transfer functions associated with the power electronic converters
are typically a combination of rst-order and second-order polynomials. The
internal output impedance (Zoo ) of the converter, shown in Figure 2.2, can
be given symbolically by

(rL + Drds1 + D rds2 ) L
1+s (1 + srC C)
LC rL + Drds1 + D rds2
Zoo =
rL + rC + Drds1 + D rds2 1
s2 + s +
L LC
(2.64)
A practical converter, Figure 2.2, operating at the switching frequency of
100 kHz, the output power of 25 W, and the duty ratio (D) of 0.5, may
have the circuit element values such as follows: L = 100 H, C = 330 F, rL =
20 m, rC = 33 m, and rds1,2 = 200 m, where rL , rC , and rds1,2 are called
equivalent series resistances (ESR) of the associated circuit element, and rds1,2
includes also the switching losses of the associated semiconductor switches.
Substituting the symbolical element values in (2.64) with the dened physical
values yields

6.667 106 (1 + s 4.546 104 )(1 + s 1.089 105 )

Zoo = (2.65)
s2 + s 2.53 103 + 3.03 107
According to (2.64), we can conclude that the low-frequency value of |Zoo |
equals rL + Drds1 + D rds2 by setting s = 0. The corresponding high-frequency
value can be found by letting s to approach innity (i.e., s = ) equaling rC ,
respectively. The damping factor ( ) and the resonant frequency (o ) or the
44 2 Basis for Dynamic Analysis and Control Dynamics

Figure 2.13 The frequency responses of the zeros and poles of Zoo (a) magnitude and
(b) phase.

undamped natural frequency (n ) of the physical circuit can be computed

to be 0.23 and 5.5045 krad/s. This means that the system poles are complex
(i.e., < 1), and they can be given according to the denominator of (2.65)
as sp1,2 = (1.265 103 j5.3572 103 ) rad/s. The magnitude of sp1,2 (i.e.,
|sp1,2 |) equals n = 5.5045 krad/s (i.e., 876 Hz). The system comprises also two
zeros sz1 = 2.1997 103 rad/s (i.e., 350 Hz) and sz2 = 9.1827 104 rad/s
(i.e., 14.615 kHz). The frequency responses of the dened transfer functions
and their combination or Zoo are shown in Figure 2.13. The location of
the poles and zeros can be found from Figure 2.13b according to the phase
behavior, that is, the pole location, where the phase equals 90 , and the zero
location, where the phase equals 45 , respectively.
 2.6 Review of Basic Mathematical Tools 45

Figure 2.14 Typical parameters specifying the transient response in the control
engineering textbooks.

2.6.3
Stability and Performance

The control engineering textbooks [1, 32] usually characterize the transient
responses assuming a second-order system behavior (Figure 2.14) by means
of parameters (2.66) such as rise time (tr ), peak time (tp ), settling time (ts ),
maximum overshoot (Mp ), and maximum peak value (Mr ) in order to relate
the time-domain behavior of the system to the frequency-domain loop-gain
behavior. Similar attempts have also been pursued in [33].

1 d
tr = arctan
d n

tp =
d
4 3
ts = (2.5%); (5%) (2.66)
d d
 
/ 1 2
Mp = e 100%
1
Mr = 
2 1 2
Close correlation between the loop characteristics and the time-domain
transient behavior can be obtained, however, only for the responses excited
46 2 Basis for Dynamic Analysis and Control Dynamics

through the reference input, which is the most usual case in control
engineering. In the switched-mode converters, the voltage or current reference
is usually not available for transient testing, and consequently, the transients
have to be injected through the input voltage or load current. Therefore,
the transients recorded at the output voltage also contain the effect of the
corresponding internal transfer functions associated with the input voltage
and output current as well as the loop gain as dened in Section 2.3. The
internal transfer functions tend to dominate in the responses, and therefore,
the time-domain transients do not provide accurate information on the loop
behavior [3437].

2.6.3.1 Stability
The closed-loop output dynamics of the voltage-output converter are
characterized (Section 2.2.1) by
Gioo Zoo 1 Lv (s)
uo = uin io + ur (2.67)
1 + Lv (s) 1 + Lv (s) Gse 1 + Lv (s)
where Lv (s) is the voltage-loop gain. The denominator term 1 + Lv (s) is
called the closed-loop systems characteristic polynomial. The inverse of the
characteristic polynomial is called sensitivity function S(s), and Lv (s) S(s)
the complementary sensitivity function T(s) [1, 31, 32]. It is obvious that
S(s) + T(s) = 1. Consequently, the output dynamics can be dened by
1
uo = Gioo S(s) uin Zoo S(s) io + T(s) ur (2.68)
Gse
For a stable system, the roots of the characteristic polynomial (1 + Lv (s)) have
to be located in the open left-half plane of the complex plane. A system having
roots in the imaginary axis is considered to be marginally stable in control
engineering, but in power electronics, a system with pure imaginary roots is
deemed to be unstable.
If an accurate analytical expression for the loop gain L(s) is available, then
the study of the location of the roots of the characteristic polynomial can be
made. In practice, the frequency response of the loop gain may be available
only, from which the poles and zeros cannot be reliably extracted. Therefore,
other methods based directly on the loop frequency response have to be used.
The usual visualization methods of the loop frequency behavior are polar and
Bode plots. The polar plot (Figure 2.15a) is constructed by plotting the locus of
the magnitude of the loop gain in the complex plane with the x axis containing
the real part of the loop gain and the y axis containing the imaginary part.
Usually the locus tends to zero when the frequency approaches innity. The
frequency is not explicitly shown in the polar plot but only the direction of the
increasing frequency. In order to study the stability, the polar plot is constructed
both for the positive (Figure 2.15a, solid line) and negative (Figure 2.15a,
 2.6 Review of Basic Mathematical Tools 47

Figure 2.15 Visualization methods: (a) polar plot and (b) Bode plot.

dashed line) frequencies. In practice, this means that the imaginary part is
as it is in the original loop gain for the positive frequencies, and its sign is
changed negative for the negative frequencies producing a mirror effect with
respect to the x axis. The Bode plot (Figure 2.15b) is constructed by plotting
the magnitude in dB and the phase in degrees usually in the separate subplots
in respect to frequency, where the x axis is the frequency in the logarithmic
scale and the y axis the magnitude and phase in the linear scale. The polar and
Bode plots in Figure 2.15 are drawn for the same loop gain, but the polar plot
shows only the frequencies higher than 10 kHz when the Bode plot shows a
much higher frequency range.
48 2 Basis for Dynamic Analysis and Control Dynamics

The stability of the closed-loop system can be studied by applying the

Nyquist stability criterion [1, 32] to the loop gain L(s) by means of a polar
plot constructed both for the positive and negative frequencies as shown in
Figure 2.15a. Such a polar plot is called Nyquist plot. According to the Nyquist
stability criterion, the system is stable if the locus of the loop gain does not
pass through the point (1, 0) or encircle it in the clockwise direction when
both of the loci are considered. The situation may be much more complicated
for concluding the state of stability as explained in detail in [32] (pp. 521542).
Consider the following loop gain:
K
L(s) = (2.69)
s(1 + s 103 )(1 + s 102 )
where K is a constant. According to Figure 2.16a, the closed-loop system
having the loop gain according to (2.69) would become unstable for high
enough values of K because they either pass through the point (1, 0) or
encircle it in the clockwise direction.
The closed-loop system may also be conditionally stable as shown in
Figure 2.17, where the encirclement of the point (1, 0) takes place
counterclockwise indicating stability, but the system may become unstable
when the gain is slightly increased or decreased. Such a condition may take
place, for instance, when the converter starts up due to a reduction of a gain
in the associated control circuitry.

2.6.3.2 Loop-Gain-Related Dynamic Indices

The robustness of the stability is typically related to gain (GM) and phase (PM)
margins and the dynamical transient performance to the control bandwidth.
The closed-loop systems usually employ negative feedback, which is also taken
into account when constructing the characteristic polynomial (i.e., 1 + L(s)):
if the loop gain is physically measured, the overall phase change for unstable
operation is 360 , which is also the typical reading (i.e., 0 ) in the frequency
response analyzers. For complying with the control engineering domain, 180
is subtracted from the reading or data produced by the analyzers. According
to this, the PM (Figure 2.18a) is dened by

PM =  L(s) + 180 (2.70)

at the frequency where |L(s)| = 1 (i.e., the loop-gain crossover frequency, gco
or fgco ). Similarly, the GM (Figure 2.18a) is dened by

1
GM = (2.71)
|L(s)|
at the frequency where  L(s) = 180 (i.e., the phase-crossover frequency,
phco or fphco ). Figure 2.18b shows the same denitions using the Bode plot.
 2.6 Review of Basic Mathematical Tools 49

Figure 2.16 Stable (K = 50) and unstable (K = 200) systems: (a) Nyquist plot and
(b) Bode plot.

In power electronics, the loop crossover frequency (fgco ), Figure 2.19, is

typically called control bandwidth, but the theoretical control bandwidth is
dened to be the frequency range from zero to the frequency at which the
sensitivity function (S(s)) equals 3 dB (i.e., 1/ 2). The loop crossover
frequency is naturally slightly higher than the control bandwidth (i.e.,
Figure 2.19; fgco = 2.36 kHz and fs3 dB = 1.79 kHz). Another denition can
also be found based on the corner frequency of the complementary sensitivity
function (T(s)).
50 2 Basis for Dynamic Analysis and Control Dynamics

Figure 2.17 Conditionally stable

system.

2.6.3.3 Right-Half-Plane Zero and Pole

Certain switched-mode converters contain an RHP zero in their control-to-
output transfer function (Gco ) [3840]. Such a converter is also known as
a nonminimum-phase system. Characteristic to the RHP zero is that it will
increase the loop gain and degrease the loop phase, which would make
the corresponding converters more susceptible to instability without proper
control design actions. The effect of RHP zero cannot be removed [4143]
even if such claims have been presented [3840]. The RHP zero would limit
the achievable maximum control bandwidth at the frequency of the RHP zero,
and even lower if proper phase and gain margins are to be obtained [4143].
Sometimes, the internal dynamics of a switched-mode converter may contain
an RHP pole or it may be imposed by the load impedance [44]. The RHP pole
would determine the minimum control bandwidth for stability to exist [41]. As
a consequence, the control bandwidth should be designed as high as possible
to minimize the possible load-imposed stability problems.
Figure 2.20 shows a control-to-output transfer function containing an RHP
zero: the at portion of the magnitude at higher frequencies and a phase less
than 180 implies that the compensator gain and, consequently, the control
bandwidth would be limited. It was claimed in [40] that the application of
peak-current-mode control would remove the effect of RHP zero, but in reality
it would not do that.

2.6.4
Matrix Algebra

Manipulation of the simultaneous linear equations constituting the small-

signal state space is carried out using a matrix technique. Such a group
 2.6 Review of Basic Mathematical Tools 51

Figure 2.18 Gain and phase margins: (a) Nyquist plot and (b) Bode plot.

of linear equations can be viewed as the transformation of one vector

into another. Consider, for example, the n simultaneous linear equations:

y1 = a11 x1 + a12 x2 + + a1n xn

y2 = a21 x1 + a22 x2 + + a2n xn
.. (2.72)
.
yn = an1 x1 + an2 x2 + + ann xn
52 2 Basis for Dynamic Analysis and Control Dynamics

Figure 2.19 Control bandwidth (fs3 dB ) versus loop crossover frequency (fgco ).

Figure 2.20 Typical effect of RHP zero on the control-to-output transfer function.

If we dene two column vectors x and y by

x y
1 1
x y
x = 2 ,y = 2 (2.73)

xn yn
 2.6 Review of Basic Mathematical Tools 53

then the linear transformation in (2.72) can be given by

y a a a x
1 11 12 1n 1
y2 a21 a22 a2n x2
= (2.74)

yn an1 an2 ann xn
or y = A x, where the operator stands for multiplication and will be dened
later.
An identity matrix I is dened such that its diagonal elements or entries
(i.e., aii ) equals 1, and all the other elements are zeros as shown in (2.75).
1 0 0 0
0 1 0 0
I= (2.75)
0 0 1 0
0 0 0 1

Transpose of a matrix B = (bij )mn denoted by BT = (bji )mn is such that the
diagonal elements of the original matrix and its transpose are the same but
the upper and lower triangular elements are interchanged as illustrated in
(2.76).
b b b
b b b

11 12 13 11 21 31
B= b21 b22 b23 BT = b12 b22 b32 (2.76)
b31 b32 b33 b13 b23 b33

2.6.4.1 Addition of Matrices

If we have two matrices A and B, both of the same order (i.e., m n)
a a a
b b b

11 12 13 11 12 13
A= a21 a22 a23 B= b21 b22 b23 (2.77)
a31 a32 a33 b31 b32 b33
then the sum A B is dened by
a b a b a13 b13

11 11 12 12
AB= a21 b21 a22 b22 a23 b23 (2.78)
a31 b31 a32 b32 a33 b33

2.6.4.2 Multiplication by Scalar

The multiplication of a matrix by a scalar is dened by
ca ca ca

11 12 13
cA = ca21 ca22 ca23 (2.79)
ca31 ca32 ca33
54 2 Basis for Dynamic Analysis and Control Dynamics

2.6.4.3 Matrix Multiplication

If the product AB is dened by C = AB, where cij , the element of C in the ith
row and jth column, is found by adding the products of the elements of A in
the ith row with the corresponding elements of B in the jth column. Thus

cij = ai1 b1j + ai2 b2j + + ain bnj (2.80)

It will be noted that the number of rows in A has to be the same as the number
of columns in B. The order of the multiplication is also important because
usually AB  = BA.
For example, if
   
a a12 b b12
A = 11 B = 11 (2.81)
a21 a22 b21 b22

then the product

 
a b + a12 b21 a11 b12 + a12 b22
AB = 11 11 (2.82)
a21 b11 + a22 b21 a21 b12 + a22 b22

2.6.4.4 Matrix Determinant

The determinant of the matrix A denoted by det A or |A| can be found as
follows:

n
det A = aij (1)i+j cij (2.83)
i=1

where cij is the cofactor of the element aij in the matrix A. The cofactor cij
can be found by eliminating the ith row and jth column associated with the
element aij . The summation is carried out for example for the chosen ith row.
Matrix determinant exists only for a square matrix, that is, the matrix order
has to be n n.
For example, if
a a a

11 12 13
A= a21 a22 a23 (2.84)
a31 a32 a33
then the determinant can be given by
   
 a22 a23  a a23 

det A = (1) a11 
2  + (1) a21  21
3
a32 a33  a31 a33 
 
a a22 
+ (1)4 a31  21 (2.85)
a31 a32 
det A = a11 (a22 a33 a23 a33 ) a21 (a21 a33 a23 a31 ) + a32 (a21 a32 a22 a31 )
 2.7 Operational and Control Modes 55

2.6.4.5 Matrix Inversion

The inverse of the matrix A is denoted by A1 . The inverse exists only when
det A  = 0. The inverse can be found as follows:
adj A
A1 = (2.86)
det A
where adj A is the adjugate of A. adj A is a matrix where all the elements of A
are rst replaced by their corresponding cofactors cij and the resulting matrix
is transposed.
For example, if
a a12 a13

11
A= a21 a22 a23 (2.87)
a31 a32 a33

then det A is as dened in (2.85), and

     

2  a22 a23 
 
3  a21 a23  
4  a21 a22  T
(1)   (1)  (1) 
a
 32 a 33   a31 a33   a31 a32 
  a  a
a a13  a23  a12 
adj A = (1)3  12  (1)4  11  (1)5  11
 a32 a33   a31 a33   a31 a32 
  
4  a12 a13  5  a11 a13  
6  a11 a12 
(1)   (1)   (1) 
a22 a23 a21 a23 a21 a22 
(2.88)
as well as the inverse of A
a a a a a12 a33 + a13 a32 a12 a33 a13 a22

22 33 23 32
a21 a33 + a23 a31 a11 a33 a23 a31 a11 a23 + a13 a21
a21 a32 a22 a31 a11 a32 + a12 a31 a11 a22 a12 a22
A1 =
a11 (a22 a33 a23 a32 ) a12 (a21 a33 a23 a31 ) + a13 (a21 a32 a22 a31 )
(2.89)

2.7
Operational and Control Modes

The operation of a converter is classied as continuous, boundary, or

discontinuous based on the behavior of the inductor current during a switching
cycle (Ts ), where the cycle time is an inverse of the switching frequency (fs ).
The operation mode is typically dened to be the continuous conduction
mode (CCM) [17] when the inductor during the switching cycle is always
greater than zero (Figure 2.21, CCM-1). This denition has become obsolete
because the synchronous rectication (Figure 2.2) has replaced the diodes
usually used earlier to implement part of the PWM switch, and consequently,
the inductor current can have values less than zero (Figure 2.2, CCM-2). As
56 2 Basis for Dynamic Analysis and Control Dynamics

Figure 2.21 Converter

operational modes.

a consequence, the operation is continuous if the inductor current has ex-

actly two different slopes within a switching cycle. Even this denition is not
covering all the cases but denes well most of them.
The operation mode is dened to be the boundary conduction mode (BCM,
Figure 2.2) when the inductor current momentarily touches the zero level but
does not cross it. This operation mode is also called the critical or transition
mode.
The operation mode is dened to be discontinuous when the inductor
current is equal to zero during a part of the switching cycle. Replacing the
low-side FET switch in Figure 2.2 with a diode would limit the inductor current
to zero by effectively disconnecting the current carrying path [14].
The way of producing the control of the pulsewidth (i.e., the length of the
on-time (Ton ), Figure 2.21) during which the main switch (i.e., the high-side
switch in Figure 2.2) is turned on would dene the control mode. The basic
control mode is the direct-on-time or voltage-mode control [11], which is also called
the direct-duty-ratio control under xed-switching-frequency operation [9]. The
direct-on-time control is implemented in such a way that the length of the on-
time is controlled based on the constant-slope ramp signal. The current-mode
control is such that the length of the on-time is directly or partly generated
using the instantaneous inductor current [6, 7]. The hysteretic control is such
that the length of the on-time and the off-time are controlled based on either
inductor current or output voltage by providing upper and lower limits for
the associated variable. The self-oscillation or boundary-mode control is a class
of hysteretic control, where the upper limit is controlled and the lower limit
is zero, that is, the converter is forced to operate in BCM. The voltage-mode
control is usually misinterpreted to mean the constant-voltage feedback control
but is related to the way of generating the pulsewidth as described above.
 References 57

References

1. R.C. Dorf and R.H. Bishop, Modern 11. T. Suntio, Unied average and
Control Systems, Addison-Wesley, small-signal modeling of direct-on-
Menlo Park, CA, USA, 1998, 8th time control, IEEE Trans. Indust.
Edition. Electron., vol. 53, no. 1, 2006,
2. M. Hankaniemi and T. Suntio, pp. 287295.
Small-signal models for constant- 12. A.S. Kislovski, R. Redl, and N.O.
current regulated converters, in Proc. Sokal, Dynamic Analysis of Switching-
IEEE Industrial Electronics Society Mode DC/DC Converters, Van Nostr-
Annual Conf., 2006, pp. 20372042. and Reinhold, New York, USA, 1991.
3. M. Hankaniemi, M. Sippola, and 13. D.M. Mitchell, DCDC Switching
T. Suntio, Analysis of load Regulator Analysis, DMMitchell
interactions in constant-current- Consultants, Cedar Rapids, IA, USA,
controlled buck converter, in Proc. 1992.
IEEE International Telecommunications 14. R.W. Erickson and D. Maksimovic,
Energy Conf., 2006, pp. 343348. Fundamentals of Power Electronics,
4. T. Suntio, I. Gadoura, J. Lempinen, Kluwer, Norwell, MA, USA, 2001,
and K. Zenger, Practical design issues 2nd Edition.
15. T. Suntio and I. Gadoura, Dynamic
of multiloop controller for a telecom
analysis of switched-mode converters
rectier, in Proc. IEEE Telecommuni-
using two-port modeling technique,
cations Energy Special Conf., 2000,
in Proc. Power Conversion and Intelli-
pp. 197201.
gent Motion Conf., 2002, pp. 387392.
5. A. Tenno, R. Tenno, and T. Suntio,
16. B.H. Cho, Modeling and analysis of
Battery impedance and its relation to
spacecraft power systems, PhD
battery characteristics, in Proc. IEEE
Thesis, Virginia Polytechnic Institute
International Telecommunications
and State University, 1985, 181 pp.
Energy Conf., 2002, pp. 176183. 17. M. Shoyama, Y. Hamafuku,
6. C.W. Deisch, Simple switching N. Matsuzaki, and T. Ninomiya,
control method changes power Simplication of transfer function in
converter into a current source, in switching converter with general load
Proc. IEEE Power Electronics Specialists impedance, in Proc. IEEE Power
Conf., 1978, pp. 300306. Electronics and Drives Conf., 1995,
7. L. Dixon, Average current mode pp. 155161.
control, in Proc. Unitrode Power Supply 18. C.T. Chen, Linear System Theory and
Design Seminar, 1991, pp. C1-1C1-14. Design, Oxford University Press, New
8. M. Karppanen, T. Suntio, and York, USA, 1999, 3rd Edition.
M. Sippola, Dynamical characteri- 19. C.K. Tse, Linear Circuit Analysis,
zation of input-voltage-feedforward- Addison-Wesley Longman, Harlow,
controlled buck converter, IEEE UK, 1998.
Trans. Indust. Electron., vol. 54, no. 2, 20. R. Mammano, Isolating the control
2007, pp. 10051013. loop, in Proc. Unitrode Power Supply
9. R.D. Middlebrook and S. Cuk, A Seminar, SEM-1000, 1994,
general unied approach to modeling pp. C21C215.
switching-converter power stages, Int. 21. M.P. Sayani, R.V. White, D.N. Nason,
J. Electron., vol. 42, no. 6, 1977, and W.A. Taylor, Isolated feedback for
pp. 521550. off-line switching power supplies with
10. J. Sun, D.M. Mitchell, M.F. primary-side control, in Proc. IEEE
Greuel, P.T. Krein, and R.M. Bass, Applied Power Electronics Conf., 1988,
Average modeling of PWM converters pp. 203211.
in discontinuous modes, IEEE Trans. 22. Y. Panov and M. Jovanovic, Small-
Power Electron., vol. 16, no. 4, 2001, signal analysis and control design of
pp. 482492. isolated power supplies with
58 2 Basis for Dynamic Analysis and Control Dynamics

optocoupler feedback, IEEE Trans. dynamics of regulated converters, IEE

Power Electron., vol. 20, no. 4, 2005, Proc. Electric Power Appl., vol. 153, no.
pp. 823832. 6, 2006, pp. 905910.
23. T. Tepsa and T. Suntio, Adjustable 35. B. Choi, Step load response of a
shunt regulator based control current-mode-controlled DC-to-DC
systems, IEEE Power Electron. Lett., converter, IEEE Trans. Aerosp.
vol. 1, no. 4, 2003, pp. 9396. Electron. Syst., vol. 33, no. 4, 1997,
24. R. Kollman and J. Betten, Closing the pp. 11151121.
loop with a popular shunt regulator, 36. C. Gezgin, Predicting load transient
Power Electronics Technology, response of output voltage in DCDC
September 2003, pp. 3036. converters, in Proc. IEEE Applied
25. C. Basso, Biasing the TL431 for Power Electronics Conf., 2004,
improved output impedance, Power pp. 13391343.
Electronics Technology, January 2005, 37. J. Betten and R. Kollman, Easy
pp. 5657. calculation yields load transient
26. Y. Panov and M. Jovanovic, Small- response, Power Electronics Technology,
signal measurement techniques in 2005, pp. 4048.
switching power supplies, in Proc. 38. D.M. Sable, B.H. Cho, and R.B. Ridley,
IEEE Applied Power Electronics Conf., Use of leading-edge modulation to
2004, pp. 770776. transform boost and yback
27. T. Suntio and M. Hankaniemi, converters into minimum-phase-zero
Unied small-signal model for PCM systems, IEEE Trans. Power Electron.,
control in CCM Unterminated vol. 6, no. 4, 1991, pp. 704711.
modeling approach, HIT J. Sci. Eng. 39. W.C. Wu, R.M. Bass, and J.R.
B, vol. 2, nos. 34, 2005, pp. 452475. Yeargan, Eliminating the effects of
28. R.D. Middlebrook, Input lter the right-hand plane zero in xed
considerations in design and frequency boost converter, in Proc.
application of switching regulators, in IEEE Power Electron. Specialists Conf.,
Proc. IEEE Industry Application Society 1988, pp. 362366.
Annual Conf., 1976, pp. 366382. 40. D.M. Mitchell, Tricks of the trade:
29. M. Hankaniemi, M. Karppanen, and Understanding the right-hand-plane
T. Suntio, Load imposed instability zero in the small-signal DC/DC con-
and performance degradation in a verter models, IEEE Power Electron.
regulated converter, IEE Proc. Electric Soc. Newsletter, January 2001, pp 56.
Power Appl., vol. 153, no. 6, 2006, 41. J.M. Freudenberg and D.P. Looze,
pp. 905910. Right half plane poles and zeros and
30. R.D. Middlebrook, Null double design tradeoffs in feedback systems,
injection and the extra element IEEE Trans. Autom. Control, vol.
theorem, IEEE Trans. Educ., vol. 32, AC-30, no. 6, 1985, pp. 555565.
no. 3, 1989, pp. 167180. 42. J.M. Maciejowski, Multivariable
31. P.J. Antsaklis and A.N. Michel, Linear Feedback Design, Addison-Wesley,
Systems, McGraw-Hill, New York, Reading, MA, USA, 1989.
USA, 1997. 43. S. Skogestad and I. Postlethwaite,
32. K. Ogata, Modern Control Engineering, Multivariable Feedback Control,
Prentice-Hall, Upper Saddle River, NJ, Analysis and Design, Wiley, New York,
USA, 1997. USA, 1996.
33. W.H. Tutle, Relating converter 44. V. Grigore, J. Hatonen, J. Kyyra, and
transient response characteristics to T. Suntio, Dynamics of a buck
feedback loop control design, in Proc. converter with a constant power load,
Powercon 11, 1984, pp. 10.110.12. in Proc. IEEE Power Electronics
34. T. Suntio, M. Hankaniemi, and Specialists Conf., 1998, pp. 7278.
M. Karppanen, Analysing the
 59

3
Average and Small-Signal Modeling of Direct-On-Time
Controlled Converters

3.1
Introduction

The chapter provides the unied basis for the average and small-signal
modeling of the voltage-output switching-mode converters under direct
-on-time control in continuous (CCM) and discontinuous (DCM) conduction
modes. Direct on-time control is commonly known as voltage-mode control
(VMC) and also direct-duty-ratio control in the case of constant switching
frequency [1]. It will be shown that the classical state-space-averaging (SSA)
technique [2, 3] yields the same xed-frequency dynamic models in CCM as
the more general method introduced in [10, 11]. The classical SSA method [5]
failed to produce accurate small-signal models when applied directly to the
converters operating in DCM. The method of producing the more accurate
DCM models was developed in the late 1990s [79] after 20 years of intensive
efforts [6, 12]. The dynamic models associated with the direct-on-time control
are important because they would provide the basis for modeling the other
control modes such as current-mode control, self-oscillation control, hysteretic
control, and so on.
The SSA method was developed without considering the dynamic processes
inside the converter and, therefore, the good results in CCM were actually
obtained by accident, which the failing in DCM proves: The time-averaged
state variables where the averaging is done over one switching cycle would
mainly contribute to the dynamics observed at the output or input of the
converter [11]. In practice, this means that the real state variables are the
time-averaged values of those variables which are also continuous signals of
time regardless of the operation mode. If feedback or feedforward signals are
applied either to implementing the usual feedback control or to producing
different control modes such as current-mode control, the ripple components
superimposed on the time-averaged signals may produce ripple effects on the
converter dynamics [1316], which may even lead to instability under certain
conditions [17].
It has been observed that the pulsewidth modulation (PWM) process would
corrupt the sinus-form excitation signals [20]. The results of the corruption

Dynamic Prole of Switched-Mode Converter. Teuvo Suntio

2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
ISBN: 978-3-527-40708-8
60 3 Average and Small-Signal Modeling of Direct-On-Time Controlled Converters

start appearing clearly when the excitation signal approaches half the switching
frequency. Typically, the phenomenon would cause an excess phase lag and
slight increase in the magnitude compared to the predictions where the
phenomenon is not considered [18, 19, 21, 22].
The modeling and the models are typically presented including the load
impedance (i.e., usually a resistor) in them [18]. As a consequence, those
models do not represent the internal dynamics of the associated converter, and
they may even totally hide its real dynamic behavior. Such an approach has led
to widespread misunderstanding. A good example is [23], where it is explicitly
claimed that the resonant nature in a VMC buck converter would become
more severe (i.e., damping decreases) when the load decreases. Actually the
internal dynamics does not change if the other operational conditions are kept
constant.
The concept of unterminated or internal load-and-source-independent
models has been well known for a long time [24, 25] but their meaning
has not been understood until [26]. The average- and small-signal models
presented in this chapter are the real internal models. We derive and present
the state spaces in general form applicable both to variable and xed-frequency
operation, but the transfer functions we derive are only applicable to xed-
frequency operation. The variable-frequency operation will be treated more in
detail in Chapter 6. The converters we treat more in detail are buck, boost, and
buckboost converters. The introduced methods are also readily applicable to
the more complicated converters.

3.2
Direct-on-Time Control

The direct on-time control is the basic control method of a switched-mode

converter (Figure 3.1) where the length of the on-time (Figure 3.1b, ton ) is
controlled to keep the output voltage or current constant at the predened
level [1]. During the on-time, the main switch (Figure. 3.1a, the high-side
switch) is conducting and the other switch (Figure 3.1a, the low-side switch)
is turned off. During the off-time (Figure 3.1b, toff ), the main switch is
turned off, and the other switch is conducting. This process is repeated in
a sequential manner, that is, the operation is periodical. If the switching
frequency (Figure 3.1a, fs ) is constant, then the on-time control equals direct-
duty-ratio control, where the duty ratio (d) can be expressed by tTons , and
its complement (d
= 1 d) by tToffs . The inverse of switching frequency
(fs ) denoted by Ts (Figure 3.1b, Ts = ton + toff ) is called cycle time. The
xed-frequency operation (i.e., constant cycle time) is the usual choice of
controlling the converter, although other controlling methods do exist yielding
variable switching frequency, for example, self-oscillation, constant-on-time,
and hysteretic control [1]. In reality, the variable-frequency direct-on-time-
control mode is ctive serving only the modeling purposes.
 3.2 Direct-on-Time Control 61

Figure 3.1 Direct-duty-ratio-controlled synchronous buck converter: (a) schematics and

(b) PWM generation.

The on-time or PWM under the xed-frequency VMC is implemented

(Figure 3.1b) by comparing the voltage provided by the controller (i.e., uco )
to a constant-slope PWM ramp signal (i.e., MPWM = VM /Ts ) provided by the
modulator. The pulse is switched on in the beginning of the cycle by means of
the pace signal applied to the set input (S) of the RS-ip-op. The pace signal
also determines the switching frequency (fs ). The pulse will be terminated when
the PWM ramp exceeds the control signal (uco ) by turning the comparator
output high and resetting the RS-ip-op. The ip-op controls the gate drivers,
which in turn provide the gate-control signals of the associated switches.
When considering the situation from the dynamic point of view, the slope
of the PWM ramp signal would affect the gain (i.e., Ga , Figure 3.2) of the
dynamic actions through the modulator. This gain is usually called modulator
gain. It can be found by constructing the comparator equations (Figure 3.1a),
which dene the length of the on-time, as follows:

VM
uco = ton (3.1)
Ts
62 3 Average and Small-Signal Modeling of Direct-On-Time Controlled Converters

Figure 3.2 Direct-on-time-controlled converter at open loop with a PWM modulator.

and developing the partial derivatives of (3.1), which yields in general form
Ts
ton = uco (3.2)
VM
Under xed-frequency operation, Eq. (3.2) becomes
1
d = uco (3.3)
VM
1
and consequently, the modulator gain (Ga ) equals VM [1]. Due to the PWM
process, the modulator gain may also contain some frequency-dependent
components [1822] in addition to the constant gain dened above.
When dening the dynamic models for the direct-on-time-controlled power
stages in the subsequent sections, we will assume that the modulator gain
equals 1 as illustrated in Figure 3.2. The derived models are also the internal
models, that is, the source impedance (i.e., Zs , Figure 3.2) is assumed to be
zero and the load impedance (i.e., ZL , Figure 3.2) is assumed to be innite
when deriving the corresponding state spaces.

3.3
Generalized Modeling Technique

The switched-mode converters are nonlinear variable-structure systems where

different structures are periodically switched on and off. The nonlinearity can
be removed by averaging the converter operation within one switching cycle.
Linearizing the obtained averaged model by developing the partial derivatives
with respect to the state and input variables at a certain operating point would
yield the required dynamic representation of the converter. It is obvious that
the models are valid only up to half the switching frequency due to the nature
 3.3 Generalized Modeling Technique 63

Figure 3.3 Waveforms of the inductor current

at an arbitrary switching cycle.

of averaging. In order to construct the averaged state space, we have to consider

the variables as the time-varying averages of the corresponding instantaneous
values (x), which we denote by x.
The behavior of the time-averaged inductor current iL  would determine
most of the dynamics in a converter. The portion of the inductor current having
positive up-slope (m1i ) and negative down-slope (m2i ) is presented in Figure 3.3
for an arbitrary cycle. The corresponding portions of the cycle time (ts ) are
denoted by ton and toff1 . It should be noted that the cycle time ts = ton + toff1 in
CCM but ts > ton + toff1 in DCM (see Figure 2.21, Chapter 2). The slopes m1i ,
m2i are local averages within the on-time or off-time, respectively.
The time-averaged inductor current iLi  is a continuous signal of time
regardless of the conduction mode but its charge potential within those
subcycles is distributed as
ton
iLi on = iLi 
ton + toff1
(3.4)
toff1
iLi off = iLi 
ton + toff1
For developing the averaged state space, the derivatives of the time-averaged
state variables have to be dened. The derivative of the time-averaged
inductor current can be approximated according to the average slope of
the instantaneous inductor current (Figure 3.3) by
diLi  ton toff1
= m1i m2i (3.5)
dt ts ts

where m1i = uLion

Li
and m2i = uLioff1
Li
as well as uLion , the voltage across the
inductor during the on-time (ton ), and uLioff1 , the voltage across the inductor
during the part (toff1 ) of the off-time (toff ) when the inductor current is
nonzero. This relation may be obvious because uL = L didtL . The bar over the
voltage denotes that the variables constituting the voltage are the corresponding
local averages during the corresponding subcycle.
The derivative of the time-averaged capacitor voltage can be approximated
(Figure 3.4) by
duCi  i+  i 
= (3.6)
dt Ci Ci
64 3 Average and Small-Signal Modeling of Direct-On-Time Controlled Converters

Figure 3.4 Capacitor charge balance.

Figure 3.5 Converter output stage.

where i+  is the time-averaged current charging the capacitor, and i  is the
time-averaged current discharging the capacitor within a cycle. This relation
is obvious because ic = C dudtC and iC = i+ i according to Kirchhoff s current
law (Figure 3.4).
The output of the converter is most often provided with an output capacitor as
illustrated in Figure 3.5. Therefore, the average output voltage can be given by

uo  = uCi  + rCi i+  rCi io  (3.7)

or by
duCi 
uo  = uCi  + rCi Ci (3.8)
dt
which is a useful form in the nal state space for the output voltage.
The time-varying averaged input current (iin ) is usually the current of a
certain circuit inductor or part of it, and therefore, its value can be found
according to (3.4). The output voltage (uo ) may affect the slope of the inductor
current. The slopes (mij ) are the local averages taken over the corresponding
subcycle, and therefore, the effect of the output voltage will be considered
correspondingly: the inductor current shall be taken always as such if involved
in the process not as an average over the whole cycle. This statement will be
claried when we will present the derived state spaces for the basic converters
in the subsequent subsections.

3.3.1
Buck Converter

A buck or step-down converter (Figure 3.6) is a converter where the input

voltage has to be higher than the output voltage for a proper operation. We
 3.3 Generalized Modeling Technique 65

Figure 3.6 Buck converter

using (a) diode switching and
(b) synchronous switching.

consider the modeling of two type of buck converters differing from each
other with respect to the implementation of the freewheeling or low-side
switch either with a diode (Figure 3.6a) or with a MOSFET (Figure 3.6b). The
use of the diode means that the converter may operate in both CCM and
DCM. The use of MOSFET means that the converter operates only in CCM
without any special control arrangements. The internal parasitic resistances
(i.e., rd , rds1 , rds2 ) in the switching elements also include the switching losses.
The high-side switch is the main switch, which is turned on during the on-time.
According to the generalized method, we divide the switching cycle into two
subcycles: during the on-time both of the buck converters (Figure 3.6) have the
same structure as shown in Figure 3.7a. During the off-time1 (i.e., during the

Figure 3.7 The subcircuit

structures of the buck converter
during the (a) on-time and
off-time1 with (b) diode
switching and (c) synchronous
switching.
66 3 Average and Small-Signal Modeling of Direct-On-Time Controlled Converters

part of the off-time when the inductor current slope is negative (Figure 3.3)),
the structures are basically the same except the loss components: Figure 3.7b,
with the diode switching, and Figure 3.7c, with the synchronous switching.
According to Figure 3.7, we may conclude that the current i+  charging the
output capacitor equals the time-averaged inductor current iL  because the
inductor current charges the output during both of the subcycles. For the same
reason, the local average of the output voltage during the on- and off-times can
be given by

uoon/off = rC iL  + uC  rC io  (3.9)

The time-averaged input current (iin ) equals the on-time inductor current,
and therefore, the rst equation in (3.4) applies.
According to these conclusions and applying Eqs. (3.4)(3.8), the general
averaged state space can be given for the buck converter by
diL  ton toff1
= m1 m2
dt ts ts
duC  iL  io 
=
dt C C
(3.10)
ton
iin  = iL 
ton + toff1
duC 
uo  = uC  + rC C
dt
where the up-slope
uin  (rL + rds1 + rC )iL  uC  + rC io 
m1 = (3.11)
L
the down-slope of the diode-switching converter
(rL + rd + rC )iL  + uC  rC io  + UD
m2 = (3.12)
L
and the down-slope of the synchronous-switching converter
(rL + rds2 + rC )iL  + uC  rC io 
m2 = (3.13)
L

3.3.2
Boost Converter

The boost or step-up converter is a converter where the input voltage has to be
lower than the output voltage for a proper operation. We consider the modeling
of both the diode-switching (Figure 3.8a) and the synchronous-switching
 3.3 Generalized Modeling Technique 67

Figure 3.8 Boost converter

using (a) diode switching and
(b) synchronous switching.

(Figure 3.8b) boost converter. The parasitic loss resistances (rd , rds1 , rds2 ) also
include the switching losses. The main switch is the low-side MOSFET. The
converter may operate in both CCM and DCM when the high-side switch is a
diode but only in CCM when the high-side switch is a MOSFET.
During the on-time both of the boost converters (Figure 3.8) have the
same structure as shown in Figure 3.9a. During the off-time1, the structures
are basically the same but the loss components are different as shown in
Figures 3.9b (diode switching) and 3.9c (synchronous switching).
From Figure 3.9, we may conclude that the current i+  charging the output
capacitor equals the time-averaged off-time inductor current iL off dened in

Figure 3.9 The subcircuit

structures of the boost converter
during the (a) on-time and
off-time1 using (b) diode switching
and (c) synchronous switching.
68 3 Average and Small-Signal Modeling of Direct-On-Time Controlled Converters

(3.4) because the inductor current charges the output only during the off-time.
For the same reason, the local average of the output voltage during the on- and
off-times can be given by

uoon = uC  rC io 

(3.14)
uooff = rC iL  + uC  rC io 

The time-averaged input current (iin ) equals the time-averaged inductor

current.
According to these conclusions and applying Eqs. (3.4)(3.8), the general
averaged state space can be given for the boost converter by
diL  ton toff1
= m1 m2
dt ts ts
duC  toff1 iL  io 
= + toff1
dt ton C C
(3.15)
iin  = iL 
duC 
uo  = uC  + rC C
dt
where the up-slope
uin  (rL + rds1 )iL 
m1 = (3.16)
L
the down-slope of the diode-switching converter
(rL + rd + rC )iL  + uC  rC io  + UD uin 
m2 = (3.17)
L
and the down-slope of the synchronous-switching converter
(rL + rds2 + rC )iL  + uC  rC io  uin 
m2 = (3.18)
L

3.3.3
BuckBoost Converter

The buckboost, step-up/down or yback converter is a converter where the

input voltage can be lower or higher than the output voltage and where
the polarity of the output voltage is negative compared to the polarity of
the input voltage. We consider the modeling of both the diode-switched
(Figure 3.10a) and synchronous-switched (Figure 3.10b) versions. We have
arranged the polarities of the output voltage and current so that the associated
equations would have positive sign as can be concluded from Figure 3.10, when
comparing to Figures. 3.6 and 3.8. The parasitic loss resistances (rd , rds1 , rds2 )
 3.3 Generalized Modeling Technique 69

Figure 3.10 A buckboost

converter.

of the switching components also include the switching losses. The main
switch is the high-side MOSFET. The converter can operate in both CCM
and DCM when the diode switching is in use, but only in CCM when the
synchronous switching is in use.
During the on-time, the converter has the structure shown in Figure 3.11a
and during the off-time1 the structures are the same in principle but the loss
components are different as shown in Figures 3.11b (diode switching) and
Figure 3.11c (synchronous switching).

Figure 3.11 The subcircuit

structures of the buckboost
converter during the
(a) on-time and off-time1
using (b) diode switching and
(c) synchronous switching.
70 3 Average and Small-Signal Modeling of Direct-On-Time Controlled Converters

From Figure 3.10, we may conclude that the current i+  charging the output
capacitor equals the time-averaged off-time inductor current iL off dened in
(3.4), because the inductor current charges the output only during the off-time.
For the same reason, the local average of the output voltage during the on- and
off-times can be given by

uoon = uC  rC io 

(3.19)
uooff = rC iL  + uC  rC io 

The time-averaged input current (iin ) equals the time-averaged on-time

inductor current dened in (3.4). According to these conclusions and applying
Eqs. (3.4)(3.8), the general averaged state space can be given for the
buckboost converter by
diL  ton toff1
= m1 m2
dt ts ts
duC  toff1 iL  io 
=
dt ton + toff1 C C
(3.20)
ton
iin  = iL 
ton + toff1
duC 
uo  = uC  + rC C
dt
where the up-slope
uin  (rL + rds1 )iL 
m1 = (3.21)
L
the down-slope of the diode-switching converter
(rL + rd + rC )iL  + uC  rC io  + UD
m2 = (3.22)
L
and the down-slope of the synchronous-switching converter
(rL + rds2 + rC )iL  + uC  rC io 
m2 = (3.23)
L
According to Sections 3.1 and 3.2, the buckboost converter has inherited
features from the buck and the boost converters as its name implies.

3.4
Fixed-Frequency Operation in CCM

The only practical voltage-mode-control principle is the xed-frequency

operation mode described in Section 3.2. The variable-frequency-operation
 3.4 Fixed-Frequency Operation in CCM 71

modes are ctive in that respect. The generalized modeling method would
give exactly the same models as the classical state-space averaging [4] in CCM
because the time-varying averaged inductor current lies exactly in the middle
of the inductor-current ripple [11], and consequently, its charge distribution
would be directly related to the length of the on-time and off-time according
to (3.4). In practice, this means that we can apply the circuit theory without
necessity to considering the nature of the inductor current or the other variables
more in detail. Consequently, the averaging can be done by multiplying the
on-time equations with the duty ratio d and the off-time equations with
its complement d
as instructed in [3, 4]. Naturally, the direct use of the
circuit theory (i.e., applying Kirchhoff s voltage and current laws) is the most
convenient method to solve the required averaged state spaces, and also a
recommended method. The steady-state values of the circuit variables can be
solved from the averaged state spaces by setting the derivatives to zero. This
method would give more accurate steady-state models compared for example
to those given in [1], because the averaging methods presented in [1] cannot
correctly model the steady-state behavior of the converter. Consequently, the
effect of some important loss mechanisms such as the ESR of the circuit
capacitors does not appear in the results. The synchronous buck converter
will be treated more in detail in Section 3.4.1 as an example. The dynamical
models for the other converters will be given in Section 3.4.2 including also the
canonical small-signal and steady-state equivalent circuits introduced in [3, 4].

3.4.1
Synchronous Buck Converter

We apply the state-space averaging technique to the synchronous buck

converter shown in Figure 3.12. The converter can operate only in CCM.
Therefore, two subcircuit structures have to be studied. In order to construct
the state space, the derivatives of the inductor current and the capacitor voltage
have to be dened as well as the equations for the output variables (iin , uo )
utilizing basic circuit theory and Kirchhoff s voltage and current laws.
The on-time subcircuit is shown in Figure 3.13a and the off-time subcircuit
in Figure. 3.13b. We have to solve the equations for uL , iC , iin , and uo during
both of the subcycles as follows:

Figure 3.12 Synchronous

buck converter.
72 3 Average and Small-Signal Modeling of Direct-On-Time Controlled Converters

Figure 3.13 Subcircuit

structures for (a) on-time and
(b) off-time.

On-time:

uL = uin (rL + rds1 )iL uo

iC = iL io
iin = iL
uo = uC + rC iC

Off-time:

uL = (rL + rds2 )iL uo

iC = iL io
iin = iL
uo = uC + rC iC

When arranging the on-time and off-time equations as required by the state-
space representation and applying the identities didtL = uLL and dudtC = iCC , we get
On-time:
diL (rL + rds1 + rC ) 1 1 rC
= iL uC + uin + io
dt L L L L
duC iL io
=
dt C C
iin = iL
uo = rC iL + uC rC io

Off-time:
diL (rL + rds2 + rC ) 1 rC
= iL uC + io
dt L L L
 3.4 Fixed-Frequency Operation in CCM 73

duC iL io
=
dt C C
iin = iL
uo = rC iL + uC rC io

The averaging is done by multiplying the on-time equations with d and the
off-time equations with d
, adding them together, and applying the identity
d + d
= 1 if applicable. This procedure yields

diL (rL + drds1 + d
rds2 + rC ) 1 d rC

= iL uC + uin + io
dt L L L L
duC iL io
=
dt C C
(3.24)
iin = diL
uo = rC iL + uC rC io

The averaged state space (3.24) is nonlinear due to the product of two variables
(i.e., duin and diL ), and needs to be linearized by developing the partial
derivatives as instructed in Chapter 2 (Section 2.6.1): this procedure yields

diL (rL + Drds1 + D
rds2 + rC ) 1 D rC

= iL uC + uin + io
dt L L L L
Uin + (rds2 rds1 )IL
+ d
L
duC iL io
= (3.25)
dt C C
iin = DiL + IL d
uo = rC iL + uC rC io

As discussed earlier, the output-voltage

equation in (3.25) would be most
convenient to be substituted with uC 1 + rC C dtd when solving the state space
in frequency domain (i.e., applying Laplace transformation). This yields as the
nal time-domain state space

diL (rL + Drds1 + D
rds2 + rC ) 1 D rC

= iL uC + uin + io
dt L L L L
Uin + (rds2 rds1 )IL
+ d
L
duC iL io
= (3.26)
dt C C
74 3 Average and Small-Signal Modeling of Direct-On-Time Controlled Converters

iin = DiL + IL d
 
d
uo = 1 + rC C uC
dt
The set of equations in (3.26) can be presented in matrix form as

diL rL + Drds1 + D
rds2 + rC 1  

dt L L iL
=
duC 1 uC
0
dt C
D rC Uin + (rds2 rds1 )IL u
in
+ L L L io (3.27)
1
0 0 d
C

  D 0     uin
iin iL 0 0 IL
= d + io
uo 0 1 + rC C uC 0 0 0
dt d
Applying Laplace transform to (3.26) yields

  rL + Drds1 + D
rds2 + rC 1  
iL (s)
L L iL (s)
s =
uC (s) 1 uC (s)
0
C
D rC Uin + (rds2 rds1 )IL u (s)
in
+ L L L io (s) (3.28)
1
0 0 d(s)
C

       uin (s)
iin (s) D 0 iL (s) 0 0 IL
= + io (s)
uo (s) 0 1 + srC C uC (s) 0 0 0
d(s)
and

  rL + Drds1 + D
rds2 + rC 1 1  
iL (s) s+ iL (s)
L L
=
uC (s) 1 uC (s)
s
C
D rC Uin + (rds2 rds1 )IL u (s)
in
+ L L L io (s) (3.29)
1
0 0 d(s)
C

       uin (s)
iin (s) D 0 iL (s) 0 0 IL
= + io (s)
uo (s) 0 1 + srC C uC (s) 0 0 0
d(s)
 3.4 Fixed-Frequency Operation in CCM 75

The determinant of the inverse matrix (i.e., [ ]1 ) determines the characteristic

polynomial of the transfer functions to be
rL + Drds1 + D
rds2 + rC 1

= s2 + s + (3.30)
L LC
The transfer functions from the input variables to the state variables (i.e.,
[sI A]1 B, Chapter 2, Section 2.2) can be computed to be
Ds 1 + srC C
L LC
D rL + Drds1 + D
rds2 + sL
 
GiLo GoLo
= LC LC
GiCo GoCo rL + Drds1 + D
rds2 + rC 1
s +s
2 +
L LC
(3.31)
(Uin + (rds2 rds1 )IL )s
L
  Uin + (rds2 rds1 )IL
GcL LC
=
GcC rL + Drds1 + D
rds2 + rC 1
s +s
2 +
L LC
where
GiLo = transfer function from input voltage to inductor current
GoLo = transfer function from output current to inductor current
GiCo = transfer function from input voltage to output capacitor voltage
GoCo = transfer function from output current to output capacitor voltage
GcL = transfer function from control to inductor current
GcC = transfer function from control to inductor current.
The control-to-inductor-current transfer function (GcL ) of the set of the input-
to-state transfer functions (3.31) may be used to construct the dynamical
models for the current-mode controls introduced in Chapters 4 and 5.
The input-to-output description (i.e., C[sI A]1 B + D, Chapter 2, Section
2.2) can be found to be
D2 s D(1 + sr C)
C
L LC

(rL + Drds1 + D
rds2 + sL)
D(1 + sr C)
  C (1 + srC C)
Yino Toio
= LC LC
Gioo Zoo rL + Drds1 + D
rds2 + rC 1
s +s
2 +
L LC
D(U + (r r )I )s
in ds2 ds1 L
L
(U + (r r )I )(1 + sr C)
  in ds2 ds1 L C  
Gci LC IL
=
+ (3.32)
Gco rL + Dr ds1 + D rds2 + rC 1 0
s2 + s +
L LC
76 3 Average and Small-Signal Modeling of Direct-On-Time Controlled Converters

The steady-state values of the variables can be found setting the derivatives
diL
dt
and dudtC in (3.24) to zero. This procedure yields

IL = Io
Iin = DIo
Uo = UC (3.33)


Uo = DUin (rL + Drds1 + D rds2 )Io
Uo + (rL + rds2 )Io
D=
Uin + (rds2 rds1 )Io
The transfer functions in (3.32) would represent the open-loop internal
dynamics of a synchronous buck converter at a certain operating point dened
in (3.33). The operating-point equations (3.33) show that the ESR of the output
capacitor does not affect the operating point. It may be obvious that the internal
dynamics does not change if the operating point is kept constant. The parasitic
resistances (rds1 and rds2 ) are dependent on the level of output current and
input voltage as well as the gate-driver implementation and the dead time
applied to the switching control, which would affect the damping factor of the
resonant behavior as discussed in [23].
The ideal input admittance (Yin = Yino Gioo Gci /Gco ) has a signicant
role in the source interactions, and therefore, it has to be known: it can be
given by
DIL
Yin = (3.34)
Uin + UD + (rds2 rds1 )IL
Similarly, the short-circuit input admittance (Yinsc = Yino + Gioo Toio /
Zoo ) has a signicant role in the source interactions, and therefore, it has to
be known also: It can be given by
D2
Yinsc = (3.35)
rL + Drds1 + D
rds2 + sL

3.4.2
Dynamic Descriptions of Buck, Boost, and BuckBoost Converters

3.4.2.1 Diode-Switched Buck (Figure 3.6a)

diL (rL + Drds1 + D
rd + rC ) 1 D rC

= iL uC + uin + io
dt L L L L
Uin + UD + (rd rds1 )IL
+ d
L
duC iL io
=
dt C C
 3.4 Fixed-Frequency Operation in CCM 77

iin = DiL + IL d
 
d
uo = 1 + rC C uC (3.36)
dt

IL = Io
Iin = DIo
Uo = UC (3.37)



Uo = DUin D UD (rL + Drds1 + D rd )Io
Uo + UD + (rL + rd )Io
D=
Uin + UD + (rd rds1 )Io

D2 s D(1 + srC C)
L LC

(rL + Drds1 + D
rd + sL)
Y Toio D(1 + srC C)
ino (1 + srC C)

LC LC
=
rL + Drds1 + D
rd + rC 1
s +s
2 +
Gioo Zoo L LC
D(Uin + UD + (rd rds1 )IL )s
L
  (Uin + UD + (rd rds1 )IL )(1 + srC C)  
Gci LC IL
=
+
Gco rL + Drds1 + D rd + rC 1 0
s2 + s +
L LC

(Uin + UD + (rd rds1 )IL )s

GcL = L (3.38)
rL + Drds1 + D
rd + rC 1
s +s
2 +
L LC
DIL
Yin =
Uin + UD + (rd rds1 )IL

D2 (3.39)
Yinsc =
rL + Drds1 + D
rd + sL

3.4.2.2 Diode-Switched Boost (Figure 3.8a)

diL (rL + Drds1 + D
rd + D
rC ) D
1 D
rC

= iL uC + uin + io
dt L L L L
Uo + UD + (rd + DrC rds1 )IL
+ d
L
78 3 Average and Small-Signal Modeling of Direct-On-Time Controlled Converters

duC D
1 IL
= iL io d
dt C C C
iin = iL (3.40)
 
d
uo = 1 + rC C uC
dt

Io
IL =
D

Io
Iin =

D
Uo = UC
(3.41)
Uin (rL + Drds1 + D
rd + DD
rC )
Uo =

UD Io
D D
2
(Uo +UD rC Io )D
2 (Uin (rd rds1 + rC )Io )D

+ (rL + rds1 )Io = 0

s D
(1 + srC C)
L LC

(r + Dr + D

r

L ds1 d
D (1 + srC C) + DD
rC + sL)(1 + srC C)
 
Yino Toio LC LC
=
Gioo Zoo rL + Drds1 + D (rd + rC ) D
2


s2 + s +
L LC

   
D
IL Uo + UD rd + DrC rds1
1+s + C
LC D
IL D



2
  (D (U o + U D ) (rL + r ds1 + D rC L sLIL )(1 + srC C)
)I
Gci LC
=
Gco rL + Drds1 + D
(rd + rC ) D
2
s +s
2 +
L LC
   
D
IL Uo + UD rd + DrC rds1
1+s + C
LC D
IL D

GcL = (3.42)
rL + Drds1 + D
(rd + rC ) D
2
s2 + s +
L LC

IL
Yin =
D
(Uo + UD ) (rL + rds1 + D
2 rC )IL sLIL
(3.43)
1
Yinsc =
rL + Drds1 + D
rd + DD
rC + sL
 3.4 Fixed-Frequency Operation in CCM 79

3.4.2.3 Synchronous Boost (Figure 3.8b)

diL (rL + Drds1 + D
rds2 + D
rC ) D

= iL uC
dt L L
1 D
rC Uo + (rds2 + DrC rds1 )IL
+ uin + io + d
L L L
duC D
1 IL
= iL io d (3.44)
dt C C C
iin = iL
 
d
uo = 1 + rC C uC
dt

Io
IL =
D

Io
Iin =

D
Uo = UC
(3.45)
Uin (rL + Drds1 + D
rds2 + DD
rC )
Uo =

Io
D D
2
(Uo rC Io )D
2 (Uin (rds2 rds1
+ rC )Io )D
+ (rL + rds1 )Io = 0

s D
(1 + srC C)
L LC
(rL + Drds1 + D
rds2


  D (1 + srC C) +DD

r + sL)(1 + sr C)
Yino Toio C C

= LC LC
Gioo Zoo rL + Dr ds1 + D

(r ds2 + rC ) D
2
s2 + s +
L LC
D
I    
L Uo rds2 + DrC rds1
1+s + C
LC D
IL D


(D
Uo (rL + rds1 + D
2 rC )IL sLIL )(1 + srC C)
 
Gci LC
=
Gco rL + Dr ds1 + D
(rds2 + rC ) D
2
s2 + s +
L LC
   
D
IL Uo rds2 + DrC rds1
1+s + C
LC D
IL D

GcL = (3.46)
rL + Drds1 + D
(rds2 + rC ) D
2
s2 + s +
L LC
 80 3 Average and Small-Signal Modeling of Direct-On-Time Controlled Converters

IL
Yin =
D
Uo (rL + rds1 + D
2 rC )IL sLIL

1 (3.47)
Yinsc =
rL + Drds1 + D
rds2 + DD
rC + sL

3.4.2.4 Diode-Switched BuckBoost (Figure 3.10a)

diL (rL + Drds1 + D
rd + D
rC ) D
D

= iL uC + uin
dt L L L
D
rC Uin + Uo + UD + (rd + DrC rds1 )IL
+ io + d
L L
duC D
1 IL
= iL io d (3.48)
dt C C C
iin = D
 iL + IL d
d
uo = 1 + rC C uC
dt

Io
IL =
D

D
Iin =
Io
D
Uo = UC
(3.49)
DUin (rL + Drds1 + D
rd + DD
rC )
Uo = UD Io
D
D
2
(Uin + Uo + UD rC Io )D
2 (Uin (rd rds1 + rC )Io )D

+ (rL + rds1 )Io = 0

D2 s DD
(1 + srC C)
L LC

(rL + Drds1 + D
rd + DD
rC + sL)

  DD
(1 + srC C) (1 + srC C)
Yino Toio
= LC LC
Gioo Zoo rL + Dr ds1 + D

(rd + rC ) D
2
s2 + s +
L LC
DD
I  
Uin + Uo + UD rd + DrC rds1
 
L
1+s + C
LC D
IL D


(D
(Uin + Uo + UD ) (rL + rds1 + D
2 rC )IL sLIL )(1 + srC C)
   
Gci LC I
=

2
+ L
Gco rL + Drds1 + D (rd + rC ) D 0
s2 + s +
L LC
 3.4 Fixed-Frequency Operation in CCM 81
  
D
IL Uin + Uo + UD rd + DrC rds1
1+s +
LC D
IL D

GcL =
rL + Drds1 + D (rd + rC ) D
2


s2 + s +
L LC
(3.50)

DIL
Yin =
D
(Uin + Uo + UD ) (rL + rds1 + D
2 rC )IL sLIL
(3.51)
D2
Yinsc =
rL + Drds1 + D
rd + DD
rC + sL

3.4.2.5 Synchronous BuckBoost (Figure 3.10b)

diL (rL + Drds1 + D
rds2 + D
rC ) D
D

= iL uC + uin
dt L L L
D
rC Uin + Uo + (rds2 + DrC rds1 )IL
+ io + d
L L
duC D
1 IL
= iL io d (3.52)
dt C C C
iin = D iL + IL d
 
d
uo = 1 + rC C uC
dt

Io
IL =
D

D
Iin =
Io
D
Uo = UC (3.53)



DUin (rL + Drds1 + D rds2 + DD rC )
Uo =

Io
D D
2
(Uin + Uo rC Io )D
2 (Uin (rds2 rds1 + rC )Io )D

+ (rL + rds1 )Io = 0

D2 s DD
(1 + srC C)
L LC

(rL + Drds1 + D
rds2
DD
(1 + sr C)
  C +DD
rC + sL)(1 + srC C)
Yino Toio
= LC LC
Gioo Zoo rL + Drds1 + D

(rds2 + rC ) D
2
s2 + s +
L LC
82 3 Average and Small-Signal Modeling of Direct-On-Time Controlled Converters

   
DD
IL Uin + Uo rds2 + DrC rds1
1+s + C
LC D
IL D



2
  (D (Uin + Uo ) (rL + rds1 + D rC )I L sLI L )(1 + srC C)  
Gci LC IL
=

2
+
Gco rL + Dr ds1 + D (rds2 + r C ) D 0
s2 + s +
L LC
  
D
IL Uin + Uo rds2 + DrC rds1
1+s +
LC D
IL D

GcL = (3.54)
rL + Drds1 + D (rds2 + rC ) D
2


s2 + s +
L LC

DIL
Yin =
D
(Uin + Uo ) (rL + rds1 + D
2 rC )IL sLIL

D2 (3.55)
Yinsc =
rL + Drds1 + D
rds2 + DD
rC + sL
The operating point equations (3.41), (3.45), (3.49), and (3.53) show that the
ESR of the output capacitor affects the steady-state operation of the boost and
buckboost converters: The effect is reected as a series resistance (DrC /D
) at
the output of the converter, which increases the required duty ratio for a given
output voltage and has to be considered in the design.

3.4.3
Steady-State and Small-Signal Equivalent Circuits

The steady-sate operation point of the converter can be solved using an

equivalent circuit shown in Figure 3.14 in CCM, when the source and/or load
conditions are changed. The parameter values are given in Table 3.1 for the
buck, boost, and buckboost converters treated in Sections 3.4.1 and 3.4.2
based on (3.33), (3.37), (3.41), (3.45), (3.49) and (3.53).

Figure 3.14 Steady-state equivalent circuit.

 3.4 Fixed-Frequency Operation in CCM 83

Table 3.1 Steady-state equivalent-circuit parameters for buck,

boost and buckboost converters.

Converter M(D) UE rE

Buck D
Diode D
UD rL + Drds1 + D
rd
Synch 0 rL + Drds1 + D
rds2
1
Boost
D

rL Drds1 rd DrC
Diode UD + +
+

D
2 D
2 D D
rL Drds1 rds2 DrC
Synch 0 + +
+

D
2 D
2 D D
D
Buckboost

D
rL Drds1 rd DrC
Diode UD + +
+

D
2 D
2 D D
rL Drds1 rds2 DrC
Synch 0 + +
+

D
2 D
2 D D

The small-signal dynamical behavior of the basic second-order converters

can be given as an equivalent circuit shown in Figure 3.15 derived originally
in [3, 4]. The equivalent circuit presented in [3, 4] is not correctly constructed,
because the effect of the load resistor is added in the parameters dening
the internal model. The equivalent-circuit parameters for the basic converters
treated in Sections 3.4.1 and 3.4.2 are given in Tables 3.2 and 3.3.
The equivalent circuit in Figure 3.15 can be utilized to develop a general
set of transfer functions for the corresponding converters applying the circuit
theory. This procedure yields

M(D)2 s M(D)(1 + srC C)
Le Le C

M(D)(1 + srC C) (re + sLe )(1 + srC C)
 
Yino Toio Le C Le C
=
Gioo Zoo re + rC 1
s2 + s +
Le Le C

(3.56)
M(D)2 e(s)s
Le

M(D)e(s)(1 + srC C)
   
Gci Le C j(s)
= +
Gco re + rC 1 0
s2 + s +
Le Le C
The negative sign of s in the voltage-source-type parameter (e(s), Table 3.3)
reects the existence of a RHP zero in the control-to-output transfer function
84 3 Average and Small-Signal Modeling of Direct-On-Time Controlled Converters

Figure 3.15 Small-signal equivalent circuit.

Table 3.2 Small-signal-equivalent-circuit output-side parameters

for buck, boost and buckboost converters.

Converter M(D) Le re

Buck D L
Diode rL + Drds1 + D
rd
Synch rL + Drds1 + D
rds2
1 1
Boost
D
D
2
rL Drds1 rd DrC
Diode + +
+

D
2 D
2 D D
rL Drds1 rds2 DrC
Synch + +
+

D
2 D
2 D D
D L
Buckboost
D
D
2
rL Drds1 rd DrC
Diode + +
+

D
2 D
2 D D
rL Drds1 rds2 DrC
Synch + +
+

D
2 D
2 D D

(Gco ) and consequently, a nonminimum-phase system with the associated

control-bandwidth limitation as discussed in Chapter 2 (Section 2.6.3). The
peak values of the open-loop output impedance and the input admittance can
be given in general form for the basic converters by
 
r 2
r2
R2oe 1 + e2 1 + C2
Roe Roe
|Zoo |max =
re + rC

(3.57)
M(D)2
|Yino |max =
re + sLe
 3.5 Fixed-Frequency Operation in DCM 85

Table 3.3 Small-signal-equivalent-circuit output-side parameters

for buck, boost and buckboost converters .

Converter e(s) j(s)

Buck IL
Um + UD + (rd rds1 )IL
Diode
D
Uin + (rds2 rds1 )IL
Synch
D
IL
Boost
D

IL LIL
Diode Uo + UD (rL + rds1 + D
2 rC )
s

D D
IL LIL
Synch Uo (rL + rds1 + D
2 rC )
s

D D
IL
Buckboost
D

IL LIL
Um + Uo + UD (rL + rds1 + D
2 rC ) s

Diode D
D
D
IL LIL
Um + Uo (rL + rds1 + D
2 rC )
s

Synch D D
D


where Roe = LCe is the equivalent characteristic impedance of the averaging
lter. When the source effects on the converter dynamics in Chapter 2 (Section
2.4) were introduced, the ideal input admittance (Yin ) and the short-circuit
input admittance (Yinsc ) were dened as
Gioo Gci
Yin = Yino
Gco
(3.58)
Gioo Toio
Yinsc = Yino +
Zoo
As a consequence, the admittances can be given for the basic converters by

j(s)
Yin =
e(s)
(3.59)
M(D)2
Yinsc =
re + sLe

3.5
Fixed-Frequency Operation in DCM

The diode-switched converters (i.e., Figures. 3.6a, 3.8a, and 3.10a) will enter
into DCM when the inductor current tends to become negative: the diode in the
86 3 Average and Small-Signal Modeling of Direct-On-Time Controlled Converters

Figure 3.16 Inductor-current

waveforms in DCM.

current-carrying path will become reverse biased, which effectively disconnects

the path. As a consequence, the converters will have three different subcircuit
structures during the switching cycle: one for the on-time, one for the off-time1,
and one for the off-time2 (Figure 3.16). The inductor-current waveforms in
DCM are shown in Figure 3.16. The time-averaged inductor current (iL ) is
naturally continuous within the switching cycle and less than half the peak
inductor current.
When considering the general state-space equations we dened in Section
3.3, the only unknown variable is the length of the off-time1 (toff1 ). The
dynamics associated with toff1 can be recovered by computing its relation to
iL [8] according to the waveforms of Figure 3.16, which gives

ts
1 1
iL  = iL (t)dt = m1 ton (ton + toff1 ) (3.60)
ts 2ts
0

and which under the xed-frequency operation equals

1
iL  = m1 d(d + d1 )Ts (3.61)
2

According to (3.61), we nd that

2iL 
d1 = d (3.62)
dm1 Ts

In order to nd the xed-frequency averaged models in DCM, toff1 or d1 has

to be replaced with (3.62) in the general averaged state-space equations given
in Section 3.3. We treat more in detail the diode-switched buck converter in
Section 3.5.1 and give the corresponding averaged state spaces and transfer
functions for the boost and buckboost converters without going into the
details in Section 3.5.2.
 3.5 Fixed-Frequency Operation in DCM 87

3.5.1
Buck Converter

The set of equations dening the averaged state space for a buck converter in
DCM can be given as follows (Section 3.3.1):

diL 
= dm1 d1 m2
dt
duC  iL  io 
=
dt C C
d
iin  = iL 
d + d1
duC 
uo  = uC  + rC C (3.63)
dt
uin  (rL + rds1 + rC )iL  uC  + rC io 
m1 =
L
(rL + rd + rC )iL  + uC  rC io  + UD
m2 =
L
2iL 
d1 = d
dTs m1

Substituting m1 and m2 in (3.63) with their dened formulas in (3.11) and

(3.12) yields

diL  d(uin  + (rd rds1 )iL  + UD )

=
dt L
2iL  (rL + rd + rC )iL  + uC  rC io  + UD

dTs uin  (rL + rds1 + rC )iL  uC  + rC io 
duC  iL  io 
= (3.64)
dt C C
d2 Ts
iin  = (uin  (rL + rds1 + rC )iL  uC  + rC io )
2L
duC 
uo  = uC  + rC C
dt

which is extremely nonlinear but can be naturally linearized using the methods
introduced in Chapter 2 (Section 2.6.1). The parasitic losses do not, however,
signicantly affect the dynamical behavior except the zero from the output-
capacitor ESR (rC ), and therefore, the parasitics may be neglected [11]: the
DCM operation would create in small-signal sense a lossless resistor in series
with the other resistive losses at the output, which is much greater than
the other parasitic resistive losses together. As a consequence, the simplied
88 3 Average and Small-Signal Modeling of Direct-On-Time Controlled Converters

averaged state space can be given by

diL  duin  2iL uC 
=
dt L dTs (uin  uC )
duC  iL  io 
=
dt C C

(3.65)
d2 Ts (uin  uC )
iin  =
2L
duC 
uo  = uC  + rC C
dt
The parasitic elements may, however, affect the steady-state operating point,
which can be solved from (3.64) by setting the derivatives to zero. The
inputoutput relation (i.e., M(D, K) in [1]) cannot be dened directly with the
parasitic losses included. Usually the input and output voltages as well as the
output current are known in the specications, and we have to solve only the
duty ratio D. As a consequence, the operation point can be given as follows:

IL = Io
((rL + rd )Io + Uo + UD )
Iin = Io
(Uin + (rd rds1 )Io + UD )
Uo = UC

2LIo (rL + rd )Io + Uo + UD
D= .
Ts (Uin (rL + rds1 )Io Uo )(Uin + (rd rds1 )Io + UD )
(3.66)
The commonly available equations [1] describing the steady-state and small-
signal behavior includes two notations M and K in order to simplify the nal
equations: the steady-state inputoutput relation (Uo /Uin ) equals M, and the
dimensionless value K = Ts2L Req
, where Req = Uo /Io . (Note: commonly Req is
denoted using R because of assuming resistive load. The same relation is valid,
however, with arbitrary loads, and therefore, we use the general notation Req ,
which would also suite well for describing the internal dynamics.)
Applying the above-dened notations and (3.66), the operating point can be
given by

IL = Io
 
(rL + rd ) UD
1+ +
Req Uo
Iin =   MIo
(rd rds1 ) UD
1+ M+
Req Uin
Uo = UC
 3.5 Fixed-Frequency Operation in DCM 89


 (rL + rd ) UD
 1+ +
 Req Uo
D = M
 K      
 (rL + rds1 ) (rd rds1 ) UD
1 1+ M 1+ M+
Req Req Uin

KM
D1 =   D.
rL + rds1
D 1M
Req
(3.67)
According to (3.67), it may be obvious that the losses do not signicantly affect
the operating point. Therefore, it may be justied to neglect them, which yields
the commonly given formulation [1] as follows:

IL = Io
Iin = MIo
Uo = UC
(3.68)
2
M= 
4K
1+ 1+
D2

K
D=M
1M

D1 = K(1 M)

We derive the small-signal state space without the losses based on the simplied
averaged state space given in (3.65). The linearization yields

diL 2UC 2Uin IL

= iL uC
dt DTs (Uin UC ) DTs (Uin UC )2
   
D 2UC IL Uin 2UC IL
+ + uin + + d
L DTs (Uin UC )2 L D2 Ts (Uin UC )
duC iL io
= (3.69)
dt C C
D2 Ts D2 Ts DTs (Uin UC )
iin = uC + uin + d
2L 2L 2L
duC
uo = uC + rC C
dt
90 3 Average and Small-Signal Modeling of Direct-On-Time Controlled Converters

Applying the operation-point denitions (3.68) and the notations M and K

yields
 
diL Req K 1 K
= iL uC
dt L 1M L(1 M) 1 M

(2 M)M K 2Uin
+ uin + d
L(1 M) 1 M L
duC iL io
= (3.70)
dt C C

M2 M2 2Uo 1M
iin = uC + uin + d
Req (1 M) Req (1 M) Req K
duC
uo = uC + rC C
dt

The set of equations in (3.70) can be given in matrix form as follows:

Req  K 1

K

diL  
dt L 1M L(1 M) 1 M iL
=
uC
duC 1
0
dt C 

(2 M)M K 2Uin uin
0
+ L(1 M) 1 M L io
1
0 0 d
C
(3.71)
  M2  
0
iin Req (1 M) iL
=
uo d uC
0 1 + rC C
dt 
u
M2 2Uo 1 M in
0
+ Req (1 M) Req K io
0 0 0 d
Applying Laplace transform to (3.71) yields
  1
Req K 1 K
  s +
iL (s) L 1 M L(1 M) 1 M
=



uC (s) 1
s
C
(2 M)M  K 2Uin uin (s)
0
L i (s)
L(1 M) 1 M o
1
0 0 d(s)
C
 3.5 Fixed-Frequency Operation in DCM 91

  M2  
iin (s) 0 i (s)
= Req (1 M) L
uo (s) uC (s)
0 1 + srC C

M2 2Uo 1 M uin (s)
0
+ Req (1 M) Req K io (s) (3.72)
0 0 0 d(s)

The determinant of the matrix inverse (i.e., [ ]1 ) determines the

characteristic polynomial as
 
Req K 1 K 1

=s +s 2
+ (3.73)
L 1 M LC 1M 1M

The transfer functions from the input variables to the state variables (i.e.,
[sI A]1 B, Chapter 2, Section 2.2) can be computed to be
M(2 M)  K 1

K
s
L(1 M) 1 M LC(1  M) 1 M 

M(2 M) K 1 K
  Req + sL
GiLo GoLo LC(1 M) 1 M LC 1 M
=  
GiCo GoCo Req K 1 K 1
s +s
2 +
L 1 M LC 1M 1M
2Uin s
L
  2Uin
GcL
=  LC 
GcC Req K 1 K 1
s +s
2 +
L 1 M LC 1M 1M
(3.74)

The input-to-output description (i.e., C[sI A]1 B + D, Chapter 2, Section 2.2)

can be found to be
 
Yino Toio
Gioo Zoo
   

M3 (2 M) K M2 Req K
s +
LCReq (1 M)2 1 M CReq (1 M) L 1M
  

K
 sL + R (1 + sr C)
M(2 M) K
eq
1M
C
(1 + srC C)
LC(1 M) 1 M LC
=  
R eq K 1 K 1
s2 + s +
L 1 M LC 1M 1M
92 3 Average and Small-Signal Modeling of Direct-On-Time Controlled Converters

M2
0
+ Req (1 M)
0 0

2Uin M2
LCReq (1 M)

2U (1 + sr C)
  in C
Gci LC
=  
Gco R eq K 1 K 1
s +s
2 +
L 1 M LC 1M 1M

2Uo 1 M
+ Req K (3.75)
0
The transfer functions in (3.75) would represent the open-loop internal
dynamics of the buck converter at a certain operating point dened in (3.68).
According to the operating point denition, the ERS (rC ) of the output capacitor
does not
 affect it. The lossless resistor providing the damping can be found to
K
be Req 1M according to the output impedance (Zoo ) shown in (3.75). The
lossless resistor is so large that the roots of the characteristic polynomial (3.73)
are typically real and well separated from each other .

3.5.2
Dynamic Models for Boost and BuckBoost Converters

The averaged sate space with parasitic elements, the operation point as well
as the transfer functions dening the internal dynamics for the boost and
buckboost converters will be provided in this subsection.

3.5.2.1 Boost Converter (Figure 3.8a)

Full-order averaged state space:
diL  d((rd + rC rds1 )iL  + uC  rC io  + UD )
=
dt L
2iL  (rL + rd + rC )iL  + uC  rC io  + UD uin 

dTs uin  (rL + rds1 )iL 
duC  iL  d2 Ts io 
= (uin  (rL + rds1 )iL ) (3.76)
dt C 2LC C
iin  = iL 
duC 
uo  = uC  + rC C
dt
 3.5 Fixed-Frequency Operation in DCM 93

Simplied state space:

diL  duC  2iL (uC  uin )
=
dt L dTs uin 
duC  iL  d2 Ts uin  io 
=
dt C 2LC C
(3.77)
iin  = iL 
duC 
uo  = uC  + rC C
dt

Simplied operating point:

IL = MIo
Iin = MIo
Uo = UC
 (3.78)
4D2
1+ 1+
M= K
2

D= KM(M 1)
1
D1 =
M1
Dynamic description:
 
Yino Toio
Gioo Zoo
   
M2 KM M1 1 KM
s+
L M1 MReq C LC M 1
 
K(M 1)
  sL + Req (1 + srC C)

2M 1 KM M(M 1) M
s (1 + srC C)
LCM M1 Req C LC
=  
R eq K(M 1) 1 KM
s2 + s +
L M LC M1
 
2Uo 1
s+
L Req C
  

L M(M 1)
2Uin 1 s (1 + srC C)
R K
 eq
Gci
=  LC 
Gco R eq K(M 1) 1 KM
s +s
2 +
L M LC M1
94 3 Average and Small-Signal Modeling of Direct-On-Time Controlled Converters
 
2Uo 1
s+
L Req C
GcL =   (3.79)
Req K(M 1) 1 KM
s +s
2 +
L M LC M1

The transfer functions in (3.79) would represent the open-loop internal

dynamics of the boost converter at a certain operating point dened in
 (3.78).
The lossless resistor providing the damping can be found to be Req K(M1) M
according to the output impedance (Zoo ) shown in (3.79). The lossless resistor
is so large that the roots of the characteristic polynomial (i.e., the denominator
in (3.79)) are typically real and well separated from each other.

3.5.2.2 BuckBoost Converter (Figure3.10a)

Full-order averaged state space:
diL  d((rd + rC rds1 )iL  + uC  rC io  + UD + uin )
=
dt L
2iL  (rL + rd + rC )iL  + uC  rC io  + UD

dTs uin  (rL + rds1 )iL 
duC  iL  d2 Ts io 
= (uin  (rL + rds1 )iL ) (3.80)
dt C 2LC C
d2 Ts
iin  = (uin  (rL + rds1 )iL )
2L
duC 
uo  = uC  + rC C
dt
Simplied averaged state space:
diL  d(uC  + uin ) 2iL uC 
=
dt L dTs uin 
duC  iL  d2 Ts uin  io 
=
dt C 2LC C
(3.81)
d2 Ts uin 
iin  =
2L
duC 
uo  = uC  + rC C
dt

Simplied operating point:

IL = (1 + M)Io
Iin = MIo
Uo = UC
 3.6 Dynamic Review 95

D
M=
K

D=M K

D1 = K (3.82)
Dynamic description:
 
Yino Toio
Gioo Zoo

0 0
 
2M K LM (sL + Req K)(1 + srC C)
1s (1 + srC C)
LC 2Req C K LC
=
R eq K K
s2 + s +
L LC
M2 
0
+ Req
0 0

 0 
2Uin
  1s
LM
(1 + srC C) 2Uo
Gci LC Req K
= + Req K
Gco Req K K
s2 + s + 0
L LC
 
2Uo 1 C
+s
LC Req M
GcL = (3.83)
Req K K
s +s
2 +
L LC
The transfer functions in (3.83) would represent the open-loop internal
dynamics of the buck converter at a certain operating point dened in
(3.82).
The lossless resistor providing the damping can be found to equal
Req K according to the output impedance (Zoo ) shown in (3.83). The lossless
resistor is so large that the roots of the characteristic polynomial (i.e., the
denominator in (3.83)) are typically real and well separated from each other.

3.6
Dynamic Review

The buck and boost converters are dynamically reviewed. We compare the
dynamic changes when the converters are designed to operate in CCM or
DCM. Experimental frequency-response data are provided for a buck converter
in both CCM and DCM to support the theoretical ndings and to prove the
accuracy of the modeling approach.
96 3 Average and Small-Signal Modeling of Direct-On-Time Controlled Converters

3.6.1
Buck Converter

The dynamic prole of the buck converter shown in Figure 3.17 [27] is analyzed
in CCM and DCM at the output current of 2.5 A, and the input voltage of 20 V
and 50 V, respectively. The power stage is same in both of the operation modes
except the size of the inductor (i.e., 105 H versus 5 H). The experimental
frequency responses have been measured using Venable Industries frequency-
response analyzer Model 3120 with an impedance measurement kit. The
measurement data have been imported into Matlab for efcient gure
handling.
The duty ratio (D) giving the desired output voltage (i.e., 10 V) can be
computed according to (3.37) and (3.66) as follows: CCM, 0.53 (20 V), 0.21
(50); and DCM, 0.38 (20 V) and 0.116 (50 V). The CCM values are very accurate,
and the DCM values within 2% of the exact values according to the simulation
by means of the switching models. The value of K in DCM equals 0.25.

3.6.1.1 Control-to-Output Transfer Function

The control-to-output transfer functions are shown symbolically in (3.84).
The magnitude of the DCM transfer function has near two times higher
dependence on the input voltage than does the magnitude of the CCM transfer
function has. It may be obvious that the controller should be designed for the
maximum loop crossover frequency at the high input voltage in order to avoid
the high-frequency ripple at the output voltage to cause instability [17]:
(Uin + UD + (rd rds1 )IL )(+srC C)
GCCM = L
co
r L + Drds1 + D
rd + rC 1
s2 + s +
L LC
(3.84)
2Uin (1 + srC C)
GDCM =  LC 
co
R eq K 1 K 1
s2 + s +
L 1 M LC 1M 1M

Figure 3.17 Experimental buck converter.

 3.6 Dynamic Review 97

The numerical values of the transfer functions at 20 V are given by

s 6.2 103 + 5.93 108

GCCM =
co
s2 + s 2.08 103 + 3 107
(3.85)
s 2.64 105 + 2.53 1010
GDCM = 2
co
s + s 5.66 105 + 8.95 108
and at 50 V by

s 1.6 104 + 1.5 109

GCCM =
co
+ s 1.33 103 + 3 107
s2
(3.86)
s 6.6 105 + 6.33 1010
GDCM = 2
co
s + s 4.47 105 + 4.42 108
The damping factor ( ) (Chapter 2, Section 2.6.2) in the CCM converter varies
from 0.19 to 0.12 and in the DCM converter from 9.5 to 10.6, where the
rst value corresponds to 20 V and the last to 50 V. This means that the CCM
converter exhibits resonant behavior and, the DCM converter has real and well-
separated roots of the characteristic polynomial (i.e., the denominator in (3.85)
and (3.86)). The resonant frequency of the CCM converter is close to 872 Hz.
The low-frequency pole of the DCM converter can be approximated to be close
R
to flow 2Req C(1M)
1
, and the high-frequency pole close to fhigh 2L
eq K
1M
.
The high-frequency pole locates typically at the frequencies higher than the
switching frequency yielding effectively rst-order transfer functions.
The frequency responses of the transfer functions are shown in Figures 3.18a
(CCM) and 3.18b (DCM) conrming the observations. The phase behavior
of the CCM converter (Figure 3.18a) requires using a proportional-integral-
derivative or PID-type compensator in order to ensure stability due to the
resonant behavior. The phase behavior of the DCM converter depicts that
a proportional-integral or PI-type compensator can be used. The at high-
frequency magnitude behavior (Figure 3.18b) may indicate difculties to
design proper control system for large inputoutput voltage ratios and high
control bandwidth.
The predicted and measured CCM and DCM frequency responses are
shown in Figure 3.19. The transfer functions also include the modulator
gain (Ga = 1/3), and therefore, their magnitudes are about 10 dB lower
than the corresponding magnitudes in Figure 3.18. The predictions and the
measurements comply well with each other. In CCM, the losses are higher than
those expected at the high input voltage yielding inaccuracy in the prediction.
The high-frequency phase is clearly lower than the analytical predictions yield.
The predictions could be corrected placing a zero at the switching frequency.
The cause for the extra phase lag could be the modulation process [1822]
or simply a stray capacitor in the PWM modulator, which was a peak-
current-mode modulator containing a Zener diode limiting the maximum
instantaneous current signal approximately to 1 V.
98 3 Average and Small-Signal Modeling of Direct-On-Time Controlled Converters

Figure 3.18 Predicted frequency responses of the control-to-output transfer functions:

(a) CCM converter and (b) DCM converter.

3.6.1.2 Output Impedance

The open-loop output impedances are shown symbolically in (3.87) in CCM
and DCM. The low-frequency value of the CCM output impedance  equals
rL + Drds1 + D
rd , and that of the DCM output impedance to Req 1M K
. The
high-frequency output impedance corresponds to rC for both of the converters:

(rL + Drds1 + D
rd + sL)(1 + srC C)

CCM
Zoo = LC
rL + Drds1 + D
rd + rC 1
s +s
2 +
L LC
 3.6 Dynamic Review 99

Figure 3.19 Predicted (solid lines) and measured (dashed lines)

frequency responses of the control-to-output transfer functions
including the modulator gain (i.e., 1/3) at the input voltages of 20
and 50 V: (a) CCM, and (b) DCM.

  
K
sL + Req (1 + srC C)
1M
DCM
Zoo =  LC  (3.87)
Req K 1 K 1
s2 + s +
L 1 M LC 1M 1M
100 3 Average and Small-Signal Modeling of Direct-On-Time Controlled Converters

The CCM output impedance has the highest value close to the resonant
frequency (fo = 2 1 LC ), which can be approximated by
 
r2 rC2
R2o 1 + e2 1+
Ro R2o R2o
|Zoo
CCM
|max = (3.88)
re + rC re + rC

where re = rL + Drds1 + D
rd and Ro = CL is the characteristic impedance of
an LC circuit.
The numerical values of the output impedances are given at 20 V by

s2 3.3 102 + s 3.22 103 + 5.6 106

CCM
Zoo =
s2 + s 2.08 103 + 3 107
(3.89)
s2 3.3 102 + s 2.18 104 + 1.79 109
DCM
Zoo =
s2 + s 5.66 105 + 8.95 108
and at 50 V by

s2 3.3 102 + s 3.2 103 + 3.23 106

CCM
Zoo =
s2 + s 1.33 103 + 3 107
(3.90)
s 3.3 102 + s 1.8 104 + 1.42 109
2
DCM
Zoo =
s2 + s 4.47 105 + 4.42 108
The predicted and measured frequency responses of the CCM output
impedances are shown in Figure 3.20a. The measured output impedance
does not change when the input voltage is changed, due to the increase in the
switching losses in the high-side switch. The measured impedances are shown
in the middle of the corresponding predicted values. The predicted (solid
line) and measured (dashed line) frequency responses of the DCM output
impedances are shown in Figure 3.20b. The rst-order nature is obvious. The
analytical predictions and the measurements have good compliance.

3.6.1.3 Input-to-Output Transfer Function

The open-loop input-to-output transfer functions are shown symbolically in
(3.91),
D(1 + srC C)
GCCM = LC
ioo
rL + Drds1 + D
rd + rC 1
s +s
2 +
L LC
 (3.91)
M(2 M) K
(1 + srC C)
LC(1 M) 1 M
GDCM
ioo =  
Req K 1 K 1
s +s
2 +
L 1 M LC 1M 1M
 3.6 Dynamic Review 101

Figure 3.20 Predicted (solid line) and measured (dashed line) frequency responses of
the output impedances at the input voltages of 20 and 50 V: (a) CCM and (b) DCM.

and the numerical values are given at 20 V by

s 167 + 1.6 107

ioo =
GCCM
s2 + s 2.08 103 + 3 107
(3.92)
s 7 103 + 6.7 108
GDCM = 2
ioo
s + s 5.66 105 + 8.95 108

and at 50 V by

s 66 + 6.3 106
ioo =
GCCM
s2 + s 1.33 103 + 3 107
102 3 Average and Small-Signal Modeling of Direct-On-Time Controlled Converters

s 2 103 + 1.9 108

ioo =
GDCM (3.93)
s2 + s 4.47 105 + 4.42 108
The predicted (solid line) and measured (dashed line) frequency responses
of the CCM input-to-output transfer functions are shown in Figure 3.21a.
The value of the duty ratio determines the low-frequency attenuation, which
is poorest at the resonant frequency due to resonant peaking. The predicted
(solid line) and measured (dashed line) frequency responses of the DCM input-
to-output transfer functions are shown in Figure 3.21b. The low-frequency

Figure 3.21 Measured and predicted frequency responses of the input-to-output transfer
functions at the input voltages of 20 and 50 V: (a) CCM, and (b) DCM.
 3.6 Dynamic Review 103

attenuation is lightly less than the corresponding attenuation in CCM. The

reason for the high modeling inaccuracy at the high frequencies is the
measurement problems and parasitic resonance.

3.6.1.4 Input Admittance

The open-loop input admittances are given symbolically by

D2 s
CCM
Yino = L
rL + Drds1 + D
rd + rC 1
s +s
2 +
L LC
   
M 2 R eq K 1 M K
s2 + s +
Req (1 M) L 1M LC 1M
DCM
Yino =   .
Req K 1 K 1
s2 + s +
L 1 M LC 1M 1M
(3.94)

The corresponding numerical values are given at 20 V by

s 2.68 103
CCM
Yino =
s2 + s 2.08 103 + 3 107
(3.95)
s2 + s 5.66 105 + 2.24 108
DCM
Yino = 0.125 2
s + s 5.66 105 + 8.95 108
and at 50 V by

s 4.2 102
CCM
Yino =
s2 + s 1.33 103 + 3 107
(3.96)
s2 + s 4.47 105 + 2.83 108
DCM
Zoo = 0.0125 2
s + s 4.47 105 + 4.42 108
The predicted (solid line) and measured (dashed line) frequency responses of
the CCM input admittances are shown in Figure 3.22a. The modeling accuracy
is obvious. The reason for the high-frequency inaccuracy is the measurement
problems and the parasitic resonance. The corresponding input impedances
can be found by changing the sign of the magnitude and phase. The peak
value of the admittance corresponds to

D2
|Yino
CCM
|max = (3.97)
re + rC
where re = rL + Drds1 + D
rd . The resonant dip value of the corresponding
input impedance is naturally the inverse of (3.97). The predicted (solid line)
and measured (dashed line) frequency responses of the DCM input admittance
104 3 Average and Small-Signal Modeling of Direct-On-Time Controlled Converters

Figure 3.22 Predicted (solid line) and measured (dashed line) frequency responses of the
open-loop input admittance at the input voltages of 20 and 50 V: (a) CCM and (b) DCM.

are shown in Figure 3.22b. The reason for the modeling inaccuracy is the
measurement problems at low signal values.

3.6.1.5 Ideal Input Admittance

The ideal input admittances are symbolically given by
DIL Iin
CCM
Yin =
Uin + UD + (rd rds )IL Uin
 3.6 Dynamic Review 105

M2 Iin
DCM
Yin = (3.98)
Req Uin

The corresponding numerical values are 6.25 102 1 at 20 V and

102 1 at 50 V. The predicted (solid line) and the measured (i.e., computed
according to the measured data based on Yin = Yino Gioo
Gco
Gci
) (dashed line)
frequency responses of the CCM ideal admittances are shown in Figure 3.23a
and those of the DCM admittances in Figure 3.23b. The corresponding

Figure 3.23 Predicted (solid line) and computed (dashed line) frequency responses of
the ideal admittances at the input voltages of 20 and 50 V: (a) CCM and (b) DCM.
106 3 Average and Small-Signal Modeling of Direct-On-Time Controlled Converters

impedances values can be obtained by changing the sign of the magnitude

and phase. Figure 3.23 proves that the ideal admittance is a topology-based
constant independent of operation mode but naturally dependent on the level
of output power.

3.6.1.6 Short-Circuit Input Admittance

The short-circuit input admittances are given symbolically in (3.99) and the
corresponding numerical values are given at 20 V in (3.100) and at 50 V in
(3.101). The corresponding impedances are naturally inverses of the given
admittances.
D2
CCM
Yinsc =
sL + rL + Drds1 + D
rd
(3.99)
M2
DCM
Yinsc =
Req (1 M)

0.28
CCM
Yinsc =
s 1.05 104 + 0.186
(3.100)
DCM
Yinsc = 0.125

0.044
CCM
Yinsc =
s 1.05 104 + 0.1075
(3.101)
DCM
Yinsc = 0.0125

The corresponding frequency responses in CCM are shown in Figure 3.24a,

and those in DCM in Figure 3.24b. The experimental-data-based responses
(i.e., those computed according to Yinsc = Yino + Gioo Toio
Zoo
) are shown with
the dashed lines. The predictions and measurements comply well with each
other.

3.6.2
Boost Converter

The dynamical prole of the boost converter shown in Figure 3.25 is analyzed
in both CCM and DCM at the output current of 1.5 A, and the input voltage
of 20 and 50 V. The power stage is the same in both of the operation modes
except the size of the inductor (i.e., 350 H versus 9 H).
The duty ratio (D) giving the desired output voltage (i.e., 75 V) can be
computed according to (3.41) in CCM to be 0.748 at 20 V and 0.339 at 50 V.
The value of K in DCM equals 0.036, and Req = 50 . The corresponding
duty-ratio values (3.78) are 0.66 and 0.17.
 3.6 Dynamic Review 107

Figure 3.24 Predicted (solid line) and computed (dashed line) frequency responses of
the short-circuit admittances at the input voltages of 20 and 50 V: (a) CCM and (b) DCM.

Figure 3.25 A boost converter.

108 3 Average and Small-Signal Modeling of Direct-On-Time Controlled Converters

3.6.2.1 Control-to-Output Transfer Function

The control-to-output transfer functions are shown symbolically in (3.102).
The magnitude of the DCM transfer function has close to two times higher
dependence on the input voltage than does the magnitude of the CCM transfer
function:
(D
(Uo + UD ) (rL + rds1 + D
2 rC )IL sLIL )(1 + srC C)
GCCM = LC
co
rL + Drds1 + D
(rd + rC ) D
2
s2 + s +
L LC
   (3.102)
L M(M 1)
2Uin 1 s (1 + srC C)
Req K

GDCM =  LC 
co
Req K(M 1) 1 KM
s2 + s +
L M LC M1

According to (3.102), the RHP zero in CCM is approximately Uin /LIL , which
is closest to the origin when the input voltage is at minimum and the output
current at maximum (i.e., IL = Io /D
and D
Uin /Uo ). The RHP zero of the
DCM converter can also be given by DTs /2, which is closest to the origin when
the input voltage is at minimum and the output current at maximum. This
implies that the controllers should be designed under the same condition to
give satisfactory performance due to the control-bandwidth limitation imposed
by the RHP zero (Chapter 2, Section 2.6.3).
It is also obvious that the resonant
frequency in CCM (i.e., fo = D
/2 LC) would change according to the
changes in the duty ratio, which would also affect the controller design. The
low-frequency pole of the DCM converter can be approximated to beclose to
R
flow 2R1eq C M1
M
, and the high-frequency pole close to fhigh 2L
eq
K(M1)
M
.
The high-frequency pole locates typically at the frequencies higher than the
switching frequency yielding effectively rst-order transfer functions. This
means that the RHP zero may have more severe effect than would be expected.
The corresponding numerical values are given at 20 V by

s2 0.2 s 1.72 104 + 1.61 108

GCCM =
co
s2 + s 494 + 5.75 105
(3.103)
s 0.447 + s 1.039 105 + 1.41 1010
2
GDCM =
co
s2 + s 9.027 105 + 7.79 107

and at 50 V by

s2 7.49 102 s 2.54 103 + 4.45 108

GCCM =
co
s2 + s 363 + 3.95 106
(3.104)
s 0.3 + s 3.38 105 + 3.52 1010
2
GDCM =
co
s2 + s 6.086 105 + 1.156 108
 3.6 Dynamic Review 109

According to the numerical values, the damping factor ( ) in CCM varies from
0.0326 to 0.0913, and in DCM from 51.1 to 28.3. This means that the roots of
the characteristic polynomial in CCM are complex and in DCM real and well
separated. The resonant frequency of the CCM converter is 120.7 Hz at 20 V
and 316 Hz at 50 V. The RHP zero of the CCM converter is at 1.36 kHz at
20 V and at 9.86 kHz at 50 V. The low-frequency pole of the DCM converter is
at 14.3 Hz at 20 V and at 30.2 Hz at 50 V. The corresponding high-frequency
poles are at 143.7 and 96.8 kHz. The given approximated values of the DCM
poles would predict quite close the same locations. The RHP zero of the DCM
converter is at 52.2 kHz at 20 V and at 193.7 kHz at 50 V. As a consequence,
the maximum control bandwidths would be limited approximately to the
frequency corresponding to half the minimum RHP-zero location.
The frequency responses of the transfer functions are shown in Figures 3.26a
(CCM) and 3.26b (DCM). The dots and squares represent the responses
extracted from the switching model by means of simulation, which indicates
that the analytical models are quite accurate. The at high-frequency gain
implies limitation on the achievable maximum crossover frequency for
maintaining proper gain margin (i.e., typically 6 dB at least). Figure 3.26b
implies that the controller of the DCM converter should be designed under
the high-input voltage condition due to the highest high-frequency gain to
avoid instability caused by the output-voltage ripple effects. This means that
the lowest location of the RHP zero does not necessarily determine the worst
case for the controller design, but the authentic frequency responses have to
be studied carefully for the correct decision. In the case of the CCM converter
(Figure 3.26a), the RHP zero clearly determines the worst case.

3.6.2.2 Output Impedance

The open-loop output impedances are symbolically shown in (3.105) in
CCM and DCM. The low-frequency value of the CCM output impedance
equals DrL
2 + DrDds1

2 + D
+ D
, and the DCM output impedance to
rd DrC M1
M
Req ,
respectively. The high-frequency value corresponds to rC for both of the
converters.
(rL + Drds1 + D
rd + DD
rC + sL)(1 + srC C)
CCM
Zoo = LC
rL + Drds1 + D
(rd + rC ) D
2
s2 + s +
L LC
   (3.105)
K(M 1)
sL + Req (1 + srC C)
M
DCM
Zoo =  LC 
Req K(M 1) 1 KM
s2 + s +
L M LC M1
110 3 Average and Small-Signal Modeling of Direct-On-Time Controlled Converters

Figure 3.26 Predicted frequency responses of the control-to-

output transfer functions: (a) CCM converter and (b) DCM
converter. The dots and squares represent the switching-model-
based frequency responses obtained by means of simulation.

The CCM
output impedance

 has the highest value close to the resonant


frequency fo = DLC , which can be approximated by
 
r 2
r2
R2oe 1 + e2 1 + C2
Roe Roe R2oe
|Zoo
CCM
|max = (3.106)
re + rC re + rC

where re = DrL
2 + DrDds1

2 + D
+ D
,
rd DrC
and Roe = D1
CL is the equivalent
characteristic impedance of the LC circuit of the boost converter.
 3.6 Dynamic Review 111

The numerical values of the output impedance at 20 V are

s2 3.3 102 + s 3.18 103 + 1.54 106
CCM
Zoo =
s2 + s 494 + 5.75 105
(3.107)
s2 3.3 102 + s 3.3 104 + 2.86 109
DCM
Zoo =
s2 + s 9.027 105 + 7.79 107
and at 50 V
s2 3.3 102 + s 3.18 103 + 1.02 106
CCM
Zoo =
s2 + s 363 + 3.95 106
(3.108)
s 3.3 102 + s 2.32 104 + 1.93 109
2
DCM
Zoo =
s2 + s 6.086 105 + 1.156 108
The predicted frequency responses of the CCM output impedances are given
in Figure 3.27a, where the effect of the varying duty cycle is observable only
at low frequencies. The corresponding DCM output impedances are shown in
Figure 3.27b, where the effect of the varying duty cycle is also observable at
low frequencies but is much smaller than that in CCM.

3.6.2.3 Input-to-Output Transfer Functions

The input-to-output transfer functions are symbolically given in (3.109). The
corresponding numerical values at 20 V are given in (3.110) and at 50 V
in (3.111). The predicted frequency responses are given in Figure 3.28.
Characteristic to the input-to-output transfer functions is that they do not
provided input-noise attenuation but would amplify the noise:
D
(1 + srC C)
GCCM = LC
ioo
rL + Drds1 + D
(rd + rC ) D
2
s2 + s +
L LC
   (3.109)
2M 1 KM M(M 1)
s (1 + srC C)
LCM M1 Req C
GDCM
ioo =  
Req K(M 1) 1 KM
s +s
2 +
L M LC M1

s 23.76 + 2.28 106

ioo =
GCCM
s2 + s 494 + 5.75 105
(3.110)
s2 6.81 103 + s 7.55 102 + 1.35 108
GDCM =
ioo
s2 + s 9.027 105 + 7.79 107
s 62.32 + 5.98 106
ioo =
GCCM
s2 + s 363 + 3.95 106
(3.111)
s2 4.95 104 + s 4.77 103 + 4.62 108
GDCM =
ioo
s2 + s 6.086 105 + 1.156 108
112 3 Average and Small-Signal Modeling of Direct-On-Time Controlled Converters

Figure 3.27 The predicted open-loop output impedances at 20 and 50 V: (a) CCM
converter and (b) DCM converter.

3.6.2.4 Input Admittance

The open-loop input admittances are shown symbolically in (3.112). The
resonant peak value in CCM can be given as shown in (3.113). The inverse of
it corresponds to the resonant dip value of the open-loop input impedance:
s
CCM
Yino = L
rL + Drds1 + D
(rd + rC ) D
2
s +s
2 +
L LC
 3.6 Dynamic Review 113

Figure 3.28 The predicted open-loop input-to-output transfer functions at 20 and 50 V:

(a) CCM converter and (b) DCM converter.

  
M2 KM M1
s+
L M1 MReq C
DCM
Yino =   (3.112)
Req K(M 1) 1 KM
s +s
2 +
L M LC M1
1

2 1
|Yino
CCM
|max = D = (3.113)
re + rC rL + Drds1 + D
(rd + rC )
114 3 Average and Small-Signal Modeling of Direct-On-Time Controlled Converters

The numerical values of the input admittance at 20 V are given by

s 2.86 103
CCM
Yino =
s2 + s 494 + 5.75 105

(3.114)
s 3.46 105 + 1.61 107
DCM
Yino = 2
s + s 9.027 105 + 7.79 107
and at 50 V by

s 2.86 103
CCM
Yino =
s2 + s 363 + 3.95 106
(3.115)
s 8.22 10 + 1.73 10
4 6
DCM
Yino =
s2 + s 6.086 105 + 1.156 108
The corresponding frequency responses are shown in Figure 3.29. The
resonant nature of the input admittance in CCM (Figure 3.29a) is obvious.
The DCM input admittance implies close to resistive nature.

3.6.2.5 Ideal Input Admittance

The ideal input admittances are shown symbolically in (3.116). The
corresponding numerical values are given at 20 V in (3.117) and at 50 V
in (3.118). Theoretically, the numerators in (3.117) should be equal but
the observed deviation is the result of omitting the effect of the parasitics
in DCM (i.e., the numerator of (3.117) for an ideal converter would
be 0.283). The corresponding predicted frequency responses are shown
in Figure 3.30. The phase behavior of the ideal input admittance of
the CCM converter (Figure 3.30a) implies increased sensitivity to source
interactions:
IL
 
IL
D
Uo + UD (rL + rds1 + D
2 rC )

D
CCM
Yin =
LIL
1s  
IL
D
Uo + UD (rL + rds1 + D
2 rC )

D
(3.116)
M2
Req
DCM
Yin = 
L M(M 1)
1s
Req K
 3.6 Dynamic Review 115

Figure 3.29 The predicted open-loop input admittances at 20 and 50 V: (a) CCM
converter and (b) DCM converter.

0.3347
CCM
Yin = 1
1 s 1.172 104
(3.117)
0.2813
DCM
Yin = 1
1 s 3.047 106
0.046
CCM
Yin = 1
1 s 1.61 105
(3.118)
0.045
DCM
Yin = 1
1 s 8.216 107
116 3 Average and Small-Signal Modeling of Direct-On-Time Controlled Converters

Figure 3.30 The predicted ideal input admittances at 20 and 50 V: (a) CCM converter and
(b) DCM converter.

3.6.2.6 Short-Circuit Input Admittance

The short-circuit input admittances are shown symbolically in (3.119), and the
corresponding numerical values at 20 V in (3.120) and at 50 V in (3.121). The
predicted frequency responses are shown in Figure 3.31. The clear difference
in CCM and DCM is obvious:
1
rL + Drds1 + D
(rd + DrC )
CCM
Yinsc =
L
1+s
rL + Drds1 + D
(rd + DrC )
 3.6 Dynamic Review 117

Figure 3.31 The predicted short-circuit admittances at 20 and 50 V: (a) CCM converter
and (b) DCM converter.

M2
(M 1)Req
DCM
Yinsc =  (3.119)
L M
1+s
Req K(M 1)
5.86
CCM
Yinsc = 1
1 + s 2.05 103
(3.120)
0.1
DCM
Yinsc = 1
1 + s 1.1 106
118 3 Average and Small-Signal Modeling of Direct-On-Time Controlled Converters

8.9
CCM
Yinsc = 1
1 + s 3.11 103
(3.121)
0.09
DCM
Yinsc = 1
1 + s 1.64 106

References
9. D. Maksimovic, Computer-aided
1. R.W. Erickson and D. Maksimovic, small-signal analysis based on the
Fundamentals of Power Electronics, impulse response of DC/DC
Kluwer, Norwell, MA, USA, 2001, 2nd switching converters, IEEE Trans.
Edition. Power Electron., vol. 15,, no. 6, 2000,
2. G.W. Wester and R.D. Middlebrook, pp. 11831191.
Low-frequency characterization of 10. T. Suntio, Small-signal modeling of
switched-mode dcdc converters, switched-mode converters under
IEEE Trans. Aerosp. Electron. Syst. vol. direct-on-time control A unied
AES-9,, no. 3, 1973, pp. 376385. approach, in Proc. IEEE Indust.
3. R.D. Middlebrook and S. Cuk, A Electron. Society Annual Conf., 2002,
general unied approach to modeling pp. 479484.
switching-converter power stages, in 11. T. Suntio, Unied average and
Proc. IEEE Power Electronics Specialists small-signal modeling of
Conf., 1976, pp. 1834. direct-on-time control, IEEE Trans.
4. R.D. Middlebrook and S. Cuk, A Indust. Electron., vol. 53,, no. 1, 2006,
general unied approach to modeling pp. 287295.
switching-converter power stages, Int. 12. D. Maksimovic, A.M. Stankovic,
J. Electron., vol. 42,, no. 6, 1977, V.J. Thottuvelil, and G.C. Verghese,
pp. 521550. Modeling and simulation of power
5. S. Cuk and R.D. Middlebrook, A electronic converters, Proc. IEEE, vol.
general unied approach to modeling 89,, no. 6, 2001, pp. 898912.
switching DC-to-DC converters in 13. P.T. Krein, J. Bentsman, R.M. Bass,
discontinuous conduction mode, in and B.L. Lesieutre, On the use of ave-
Proc. IEEE Power Electron. Specialists rag ing for analysis of power electronic
Conf., 1977, pp. 3657. systems, IEEE Trans. Power Electron.,
6. D. Maksimovic and S. Cuk, A unied vol. 5,, no. 2, 1990, pp. 182190.
analysis of PWM converters in 14. P.T. Krein and R.M. Bass, A new
discontinuous modes, IEEE Trans. approach to fast simulation of
Power Electron., vol. 6,, no. 3, 1991, periodically switching power
pp. 476490. converters, in Proc. IEEE Industry
7. J. Sun, D.M. Mitchell, M.F. Greuel, Applications Society Annual Conf.,
P.T. Krein, and R.M. Bass, Modeling 1990, pp. 11851189.
of PWM converters in discontinuous 15. B. Lehman and R.M. Bass, Switching
conduction mode- reexamination, in frequency dependent averaged models
Proc. IEEE Power Electron. Specialists for PWM DCDC converters, IEEE
Conf., 1998, pp. 615622. Trans. Power Electron., vol. 11,, no 1,
8. J. Sun, D.M. Mitchell, M.F. Greuel, 1996, pp. 8998.
P.T. Krein, and R.M. Bass, Averaged 16. R.M. Bass and J. Sun, Large-signal
modeling of PWM converters averaging methods under large signal
operating in discontinuous conduction ripple conditions, in Proc. IEEE Power
mode, IEEE Trans. Power Electron., Electron. Specialists Conf., 1998,
vol. 16,, no. 4, 2001, pp. 482492. pp. 630632.
 121

4
Average and Small-Signal Modeling of Peak-Current-Mode
Control

4.1
Introduction

Peak-current-mode (PCM) control was invented in 1970s [1, 2] and has

been a popular method to control the switched-mode converter due to the
dynamic features it provides, such as increased input-noise attenuation,
effective rst-order transfer functions, pulse-by-pulse current limiting, and
easiness to parallel converters using a common reference current [3]. It has
also been claimed that the PCM control would remove the effect of RHP
zero [4], but this claim does not hold. Limited duty-ratio range at the basic
switching frequency, high open-loop output impedance, and sensitivity to
high-frequency noise in the current loop have been considered as its main
disadvantages [3, 5]. The operation of the converter in the subharmonic mode
has been considered earlier unstable [68] but proved to be stable by applying
chaos theories [9, 10].
The dynamics associated with the PCM control has fascinated the engineers
since its invention. The rst dynamic models appeared in the late 1970s [6]
but were rather inaccurate. The modeling method presented in [11, 12] has
been considered to be the most accurate one. The basic idea behind the
method is to address the existence of the mode limit to sampling effect,
which causes a resonant pole pair at half the switching frequency, and
consequently, high gain in the inductor-current loop. Applying the proposed
method [11] would, however, produce models, which do not correspond
to the measured frequency responses. The author of [11] has also clearly
observed the same discrepancies and manipulated certain transfer functions
from those resulting by applying the method in order to have better model
accuracy.
The observed phenomena in the behavior of a PCM-controlled converter are
caused by an innite frequency-independent small-signal duty-ratio gain in the
inductor-current loop both in continuous (CCM) [1317] and discontinuous
(DCM) [21] conduction modes. The correctness of the proposed models has
been widely disputed [18, 19] based on different arguments, but the consistency
and accuracy of the modeling have been, however, clearly proved in [17, 20].

Dynamic Prole of Switched-Mode Converter. Teuvo Suntio

2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
ISBN: 978-3-527-40708-8
122 4 Average and Small-Signal Modeling of Peak-Current-Mode Control

4.2
PCM-Control Principle

Under PCM control, the on-time (ton ) or duty ratio (d) is generated comparing
the inductor-current and the control current (ico ) (i.e., ico = uco /Rs in Figure 4.1,
where Rs is the equivalent inductor-current sensing resistor). In order to
extend the duty ratio (d) beyond the mode limit, the control current has to be
compensated using an articial ramp Mc . In practice, the compensation ramp
is added to the sensed inductor-current signal, but the analysis would be more
convenient considering the compensation ramp to be subtracted from the
control current as depicted in Figure 4.1. The duty ratio (d) is established when
the inductor current reaches the compensated control current (Figure 4.1b).
The PCM control is a method to change the internal dynamics of a converter
by applying feedback from the inductor current as depicted in Figure 4.1: as

Figure 4.1 PCM-control principles: (a) circuit schematics with a buck converter and (b)
duty-ratio generation.
 4.2 PCM-Control Principle 123

a result of the inductor-current feedback, the dynamics associated with the

on-time (ton ) or duty ratio (d) would be changed from the dynamics under VM
control. Therefore, it may be obvious that the analytical models can be derived
from the models or the state-space representation of the corresponding VMC


converter [22] by replacing the perturbed duty ratio d with the expression
describing its dependence on the new constellation. This dependence is
known as duty-ratio constraints [8]. The duty-ratio constraints are typically
given [8] by
 
d = Fm ico qc iL qo uo qi uin (4.1)

where Fm is the duty-ratio gain and qk is the feedforward or feedback gain

from the dened variable. If the duty-ratio constraints are applied

 to the
corresponding VMC state space, the perturbed output voltage uo has to be
substituted with its relation to the state and input variables [17] yielding the
second-order converters
 
sp sp
d = Fm
sp
ico qsp
c iL qo uc qi uin qio io
sp
(4.2)

where the superscript sp means state space. As a consequence, the

fundamental issue in the PCM modeling is to nd the proper denition
for the duty-ratio constraints.
The time-averaged inductor current
iL  contributes to the dynamics of the
converter as discussed in Chapter 3. Therefore, the desired feedback signal
would be also
iL , but the real duty-ratio generation takes place as illustrated
in Figure 4.2. From Figure 4.2, we can compute that at t = (k + d)Ts holds

ico mc dTs =
iL  +
iL (4.3)

where
iL is the difference between the peak inductor current and its averaged
value at t = (k + d)Ts . Equation (4.3) is termed as comparator equation due to
its physical realization as shown in Figure 4.1a. This means that the duty-ratio
constraints can be determined if
iL can be found. It may be obvious that
iL
would be affected by the operation mode (i.e., CCM or DCM).

Figure 4.2 Duty-ratio generation

based on the inductor-current up
slope.
124 4 Average and Small-Signal Modeling of Peak-Current-Mode Control

4.3
Modeling in CCM

In CCM, the averaged inductor current lies exactly in the middle of the ripple
band. It may be approximated by means of a rst-order function of time within
a switching cycle. Its derivative can be approximated by means of the average
slope of the instantaneous inductor current as derived in Chapter 3. Therefore,
we may express
iL  by

dd Ts

iL  = (dm1 d m2 )t + (m1 + m2 ) + iL (kTs ) (4.4)
2
where iL (kTs ) is the value of the averaged inductor current at the beginning
of the cycle (Figure 4.2). (More detailed derivation of (4.4) can be found
from [17]). The on-time instantaneous inductor current iLon (Figure 4.2) can
be approximated by

iLon = m1 t + iL (kTs ) (4.5)

As a consequence,
iL can be found by

iL = iLon ((k + d)Ts )
iL ((k + d)T)s  (4.6)

yielding

dd Ts

iL = (m1 + m2 ) (4.7)
2
and the comparator equation as

dd Ts
ico Mc dTs =
iL  + (m1 + m2 ) (4.8)
2
when the articial compensation is assumed to be constant. The duty-ratio
constraints can be developed from (4.8) by substituting the up and down slopes
with the topology-dependent values and linearizing it at the dened operating
point (i.e., developing the partial derivatives as instructed in Chapter 2). The
same
iL as (4.7) is implicitly found also in [1315] based on different methods.
If several inductor currents constitute the feedback signal, the overall
iL
can be dened to be

dd Ts 
n

iL = (m1i + m2i ) (4.9)
2
i=1

According to (4.9), the PCMC modeling of the higher order converters can
be easily made as demonstrated in Chapter 10. Essential is to remember that
the up and down slopes are the local averages within the dened part of the
switching cycle as instructed in Chapter 3.
 4.3 Modeling in CCM 125

Transformer isolation is often used for safety reasons and/or for scaling the
input voltage to obtain more optimal duty-ratio range [8]. The inductor-current
feedback is commonly taken from the transformer primary current containing
the reected inductor current (iL ) and the transformer magnetizing current
(iLM ) as depicted in Figure 4.3 in the case of an active-reset forward converter.
Similar conditions also apply to the other transformer-isolated converters, but
the shape of magnetizing current may vary and, consequently, its effect on
the duty-ratio constraints may be different [17]. It may be obvious that the
magnetizing current would have similar effect as the articial compensation
has. As a consequence, the comparator equation may be given by

kdTs uin
ico mc dTs  =
iL  +
iL (4.10)
LM

where uin and LM are the corresponding values given at the secondary side, and
k is the coefcient taking into account the shape of the magnetizing current
(i.e., normal forward converter: k = 1, active-reset forward, full and half-bridge,
push-pull converters: k = 1/2). This means that the duty-ratio gain (Fm ) and
the input-voltage feedforward gain (qi ) are to be changed compared to the
corresponding basic converter.

Figure 4.3 Active-reset forward converter: (a) schematics and (b) transformer on-time
(dTs ) current.
126 4 Average and Small-Signal Modeling of Peak-Current-Mode Control

4.3.1
Duty-Ratio Constraints for Buck, Boost, and BuckBoost Converters

We develop the duty-ratio constraints for the basic converters corresponding

to (4.1) and (4.2) utilizing the material given in Chapter 3 (Section 3.3). The
corresponding converters are dened in Figures 3.6a (for buck), 3.8a (for
boost), and 3.10a (for buckboost).

4.3.1.1 Buck Converter

uin +UD +(rd rds1 )
iL 
According to (3.11) and (3.12), m1 + m2 = L
. Therefore, the
comparator equation (4.8) becomes
dd Ts
ico Mc dTs =
iL  + (
uin  + UD + (rd rds1 )
iL ) (4.11)
2L
The developing of the partial derivatives yields the duty-ratio constraints as
1
d =   
(D D)(Uin + UD + (rd rds1 )IL
Ts Mc +
2L
    
DD Ts DD Ts
ico 1 + (rd rds1 ) iL uin (4.12)
2L 2L
from which we nd that
1
Fm =  
(D D)(Uin + UD + (rd rds1 )IL )
Ts Mc +
2L
DD Ts
qc = 1 + (rd rds1 )
2L
(4.13)
qo = 0
DD Ts
qi =
2L
The duty-ratio constraints coefcients in (4.13) would dene equally the
sp sp sp sp
coefcients in (4.1) and (4.2) (i.e., Fm = Fm , qc = qc , qo = qo , qi = qi , and
sp
qio = 0).

4.3.1.2 Boost Converter

uo +UD +(rd rds1 )
iL 
According to (3.16) and (3.17), m1 + m2 = L
or m1 + m2 =

uc +UD +(rd +rC rds1 )
iL rC
io 
L
,
when the output voltage is substituted with (3.14).
Therefore, the comparator equation can be given either as
dd Ts
ico Mc dTs =
iL  + (
uo  + UD + (rd rds1 )
iL ) (4.14)
2L
 4.3 Modeling in CCM 127

or
dd Ts
ico Mc dTs =
iL  + (
uc  + UD + (rd + rC rds1 )
iL  rC
io )
2L
(4.15)
Linearizing (4.14) yields
1
d =  
(D D)(Uo + UD + (rd rds1 )IL )
Ts Mc +
2L
    
DD Ts DD Ts
ico 1 + (rd rds1 ) iL uo (4.16)
2L 2L
from which the duty-ratio-constraints coefcients corresponding to (4.1) can
be found to be
1
Fm =   
(D D)(Uo + UD + (rd rds1 )IL )
Ts Mc +
2L
DD Ts
qc = 1 + (rd rds1 )
2L
(4.17)
DD Ts
qo =
2L
qi = 0
Linearizing (4.15) yields
1
d =  
(D D)(Uo + UD + (rd + rC rds1 )IL rC Io )
Ts Mc +
2L
   
DD Ts DD Ts DD Ts
ico 1 + (rd + rC rds1 ) iL uC + rC io
2L 2L 2L
(4.18)
from which the duty-ratio-constraints coefcients corresponding to (4.2) can
be found to be
1
sp
Fm =  
(D D)(Uo + UD + (rd + rC rds1 )IL rC Io )
Ts Mc +
2L
DD Ts
c =1+
qsp (rd + rC rds1 )
2L
DD Ts
qsp
o = (4.19)
2L
sp
qi = 0
sp DD Ts
qio = rC
2L
128 4 Average and Small-Signal Modeling of Peak-Current-Mode Control

4.3.1.3 BuckBoost

uin +
uo +UD +(rd rds1 )
iL 
According to (3.21) and (3.23), m1 + m2 = L
or m1 + m2 =

uin +
uc +UD +(rd +rC rds1 )
iL rC
io 
L
,
when the output voltage is substituted with
(3.19). Therefore, the comparator equation (4.8) becomes

dd Ts
ico Mc dTs =
iL  + (
uin  +
uo  + UD + (rd rds1 )
iL ) (4.20)
2L
or
dd Ts
ico Mc dTs =
iL  + (
uin  +
uc  + UD
2L
+ (rd + rC rds1 )
iL  rC
io ) (4.21)

Linearizing (4.20) yields

1
d =   
(D D)(Uin + Uo + UD + (rd rds1 )IL
Ts Mc +
2L
    
DD Ts DD Ts DD Ts
ico 1 + (rd rds1 ) iL uo uin
2L 2L 2L
(4.22)
from which the duty-ratio-constraints coefcients corresponding to (4.1) can
be found to be
1
Fm =  
(D D)(Uin + Uo + UD + (rd rds1 )IL )
Ts Mc +
2L
DD Ts
qc = 1 + (rd rds1 )
2L
(4.23)
DD Ts
qo =
2L
DD Ts
qi =
2L
Linearizing (4.21) yields

1
d =  
(D D)(Uin + Uo + UD + (rd + rC rds1 )IL rC Io )
Ts Mc +
2L
   
DD Ts
ico 1 + (rd + rC rds1 ) iL
2L
 
DD Ts DD Ts DD Ts
uC uin + rC io (4.24)
2L 2L 2L
 4.3 Modeling in CCM 129

from which the duty-ratio-constraints coefcients corresponding to (4.2) can

be found to be
1
sp
Fm =  
(D D)(Uin + Uo + UD + (rd + rC rds1 )IL rC Io )
Ts Mc +
2L
DD Ts
c =1+
qsp (rd + rC rds1 )
2L
DD Ts
o =
qsp (4.25)
2L
sp DD Ts
qi =
2L

sp DD Ts
qio = rC
2L
According to the derived duty-ratio constraints, it may be obvious that the duty-
ratio gain (Fm ) would become innite without compensation (i.e., Mc = 0) at
D = 0.5, which is also the observed mode limit [8]. More detailed discussions
on the subject will be given in Section 4.3.3.

4.3.1.4 General CCM Transfer Functions

The dynamic description of the basic converters can be given as a function of the
corresponding VMC-converter transfer functions, the duty-ratio-constraints
coefcients corresponding to (4.1), and the coefcients A and B based on
the block diagrams shown in Figure 4.4, where ZC is the impedance of the
output capacitor, GcL the control-to-inductor-current transfer function, and the
coefcients A and B as follows: Buck: A = 1, B = 0; Boost and Buckboost:
A = D , B = IL . The other transfer functions are irrelevant and used only to
detect the original VMC transfer functions.
From Figure 4.4, we may compute that the general transfer functions
dening the internal dynamics of the PCM-controlled buck, boost, and
buckboost converters in CCM can be given as follows:
 
BFm qc
1+ Gvioo Fm qi Gvco
A
Gioo =
1 + Lc (s) + Lv (s)
 
BFm qc Fm qc v
1+ v
Zoo + Gco
A A
Zoo =
1 + Lc (s) + Lv (s)
Fm Gvco
Gco =
1 + Lc (s) + Lv (s)
  
qc
Fm qo + Gvioo + qi Gvci
AZC
Yino = Yino
v

1 + Lc (s) + Lv (s)
130 4 Average and Small-Signal Modeling of Peak-Current-Mode Control

Figure 4.4 Block diagrams for

PCM control in CCM for the basic
converters: (a) output dynamics
and (b) input dynamics.

  
qc qc
Fm qo + v
Zoo Gvci
AZC A
Toio = Toio
v
+
1 + Lc (s) + Lv (s)
Fm Gvci
Gci = (4.26)
1 + Lc (s) + Lv (s)
where the superscript v denotes the VMC transfer functions dened in
Chapter 3. The internal inductor-current-loop gain Lc (s), and the internal
output-voltage-loop gain Lv (s) are dened by

Lc (s) = Fm qc GvcL
(4.27)
Lv (s) = Fm qo Gvco

where GvcL is the control-to-inductor-current transfer function.

The general PCMC transfer functions dened in (4.26) are actually quite
useful, because several conclusions can be easily made based on them:
 4.3 Modeling in CCM 131

1. If the VMC control-to-output transfer function (Gco ) contains, for

instance, RHP zero, the same zero also exists in the PCMC converter
with the corresponding control-bandwidth limitation.
2. If qi = 0, then the input-to-output transfer function (Gioo ) cannot be
made to be zero by means of the articial compensation.
Gv Gvci
3. If Gioo = 0, then Yino = Yino
v
iooGv
= Yin , and Yinc = Yin .
co

4.3.2
Specic Transfer Functions for the Basic Converters

The dynamic description of the PCMC converter can be derived from

the corresponding VMC state-space representation
 given in Chapter 3 by
substituting the perturbed duty ratio d with the corresponding state-space
duty-ratio constraints (4.2) given in Subsection 4.3.1.
From Figure 4.2, we may expect that the PCM control would provide extra
damping into the process similar to the operation in DCM and make the
roots of the characteristic equation real, because the control current (ico ) would
prevent the inductor current to move as freely as under the direct-duty-ratio
control, where such a limiting does not exist. In small-signal sense, the increase
in damping would be caused by a lossless resistor connected in series with the
sp sp
equivalent inductor, which can be generally given by Fm UE qc and is typically
much larger than the other restive losses. When the converter approaches the
mode limit, the duty-ratio gain Fm approaches innity. As a consequence, the
inductor would be effectively disconnected, that is, the time-averaged inductor
current would loose its status as a state variable. The PCMC transfer functions
are given below, and their dynamical implications are discussed and compared
to the corresponding VMC transfer functions in Section 4.5.

4.3.2.1 Buck Converter

The schematic of the diode-switched buck converter is given in Chapter 3
in Figure 3.6a. Substituting the perturbed duty ratio d with the derived
duty-ratio constraints
 
sp
d = Fm
sp
ico qsp
c iL qi uin

1
sp
Fm =  
(D D)UE
Ts Mc +
2L
(4.28)
DD Ts
qsp
c =1+ (rd rds1 )
2L
sp DD Ts
qi =
2L
132 4 Average and Small-Signal Modeling of Peak-Current-Mode Control

yields the PCMC small-signal state space as

sp sp 
diL rE + rC + Fm UE qc 1
= iL uC
dt L L
sp sp sp
D Fm UE qi rC Fm UE
+ uin + io + ico
L L L
duC iL io
=
dt C C

 (4.29)
iin = D Fm sp
IL qsp
c iL + Fm
sp
IL ico
duC
uo = uC + rC C
dt
UE = Uin + UD + (rd rds1 )IL
rE = rL + Drds1 + D rd

Applying Laplace transformation and matrix algebra to (4.29), the input

dynamics of the PCMC buck converter can be solved to be

sp sp 
sp sp 
D Fm UE qi D Fm IL qc s
Yino = L sp sp
Fm qi IL



sp sp 
D Fm IL qc (1 + s rC C)
Toio = LC (4.30)


sp
sp sp 
Fm UE D Fm IL qc s
Gci = L + Fmsp
IL


and the output dynamics as

sp sp 
D Fm UE qi (1 + s rC C)
Gioo = LC



sp sp 
rE + Fm UE qc + sL (1 + s rC C)
Zoo = LC (4.31)


sp
Fm UE (1 + s rC C)
Gco = LC


where the denominator (
) is dened by
sp sp
rE + Fm UE qc + rC 1

= s2 + s + (4.32)
L LC
 4.3 Modeling in CCM 133

4.3.2.2 Boost Converter

The schematic of the diode-switching boost converter   is given in Chapter 3 in
Figure 3.8a. Substituting the perturbed duty ratio d in the VMC small-signal
state space with the derived duty-ratio constraints
 
sp
d = Fm
sp
ico qsp
c iL qo uC qio io
sp

1
sp
Fm =  
(D D)UE
Ts Mc +
2L
DD Ts
c =1+
qsp (rd + rC rds1 ) (4.33)
2L
DD Ts
o =
qsp
2L
sp DD Ts
qio = rC
2L
yields the PCMC small-signal state space as
sp sp sp sp
diL rE + D2 rC + Fm UE qc D + Fm UE qo 1
= iL uC + uin
dt L L L
sp sp
D rC Fm UE qio
sp
Fm UE
+ io + ico
L L
sp sp sp sp sp sp sp
duC D + Fm IL qc Fm IL qo 1 Fm IL qio Fm IL
= iL + uC io ico
dt C C C C
iin = iL (4.34)
duC
uo = uC + rC C
dt
UE = Uo + UD + (rd + rC rds1 )IL rC Io
rE = rL + Drds1 + D (rd + DrC )
Applying Laplace transformation and matrix algebra to (4.34), the input
dynamics of the PCMC boost converter can be solved to be as
 sp sp 
1 Fm IL qo
s
L C
Yino =


sp sp

 sp sp 

 D + Fm UE qo L
D rC Fm UE qio
s+
 sp sp 
L D rC Fm UE qio C
Toio = (4.35)


sp  
Fm UE D IL
s+
L CUE
Gci =


 134 4 Average and Small-Signal Modeling of Peak-Current-Mode Control

and the output dynamics as

 sp sp 
D + Fm IL qc
(1 + s rC C)
Gioo = LC


sp sp sp
rE + Fm UE qc + Fm



 
1 Fm IL qio
sp sp sp  2 sp 
qio D UE IL (rE + D rC ) qc IL D rC
sp sp + s L (1 + s rC C)
LC 1 Fm IL qio

Zoo =


sp
 
Fm D UE IL (rE + D2 rC ) LIL
1s  (1 + s rC C)
LC D UE IL (rE + D2 rC )
Gco =


(4.36)
where the denominator (
) is dened by
 sp sp sp sp 
rE + D2 rC + Fm UE qc Fm IL qo

= s2 + s
L C
 sp

sp sp  sp sp
D + D Fm UE qo + IL qc Fm IL qo (rE + D2 rC )
2
+ (4.37)
LC

4.3.2.3 BuckBoost Converter

The schematic of the diode-switched buckboost converter is given  in
Chapter 3 in Figure 3.10a. Substituting the perturbed duty ratio d with
the dened duty-ratio constraints
 
sp
d = Fm sp ico qsp
c iL qsp
o uC qio io

1
sp
Fm =  
(D D)UE
Ts Mc +
2L
DD Ts
c =1+
qsp (rd + rC rds1 )
2L
(4.38)
DD Ts
qsp
o =
2L
sp DD Ts
qi =
2L
sp DD Ts
qio = rC
2L
 4.3 Modeling in CCM 135

yields the PCMC small-signal state space as

sp sp sp sp
diL rE + D2 rC + Fm UE qc D + Fm UE qo
= iL uC
dt L L
sp sp sp sp
D rC Fm UE qio
sp
D Fm UE qi Fm UE
+ uin + io + ico
L L L
sp sp sp sp sp sp
duC D + Fm IL qc Fm IL qo Fm IL qi
= iL + uC + uin
dt C C C
sp sp sp
1 Fm IL qio Fm IL
io ico (4.39)
C C
sp
iin = (D Fm
sp
c ) iL Fm IL qo uC Fm IL qi uin
IL qsp sp sp sp

sp
Fm
sp
IL qio io + Fm
sp
IL ico
duC
uo = uC + rC C
dt
UE = Uin + Uo + UD + (rd + rC rds1 )IL rC Io
rE = rL + Drds1 + D (rd + DrC )

Applying Laplace transformation and matrix algebra to (4.39), the input

dynamics of the PCMC buckboost converter can be solved, but the explicit
transfer functions are very complex and long and, therefore, not given here. The
transfer functions describing the output dynamics are, however, manageable
and dened as

sp sp 
D D + Fm IL qc


+ Fm qi (rE + D2 rC )IL D UE
sp sp sp sp
Fm IL qi
s + sp sp (1 + s rC C)
C LFm IL qi

Gioo =


sp sp sp
rE + Fm UE qc + Fm

sp 
sp 
1 Fm IL qio
sp  2 sp 
qio (D UE IL (rE + D rC )) qc IL D rC
sp sp + s L (1 + s rC C)
LC 1 Fm IL qio

Zoo =


 
sp LIL
Fm (D UE IL (rE + D2 rC )) 1 s (1 + s rC C)
D UE IL (rE + D2 rC )
Gco = LC


(4.40)
136 4 Average and Small-Signal Modeling of Peak-Current-Mode Control

where the denominator (
) is dened by

 sp sp sp sp 
rE + D2 rC + Fm UE qc Fm IL qo

= s2 + s
L C

sp sp  sp

D + D Fm UE qo + IL qc Fm IL qo rE + D2 rC
2 
+ (4.41)
LC
It may be obvious that the symbolic forms of Zoo , Gco , and the characteristic
equation (
) are same for the boost and buckboost converters.

4.3.3
Origin and Consequences of Mode Limit in CCM

We may compute according to the averaged comparator Eq. (4.8) at steady state
that the difference between the control current (Ico ) and the average inductor
current (IL ) in terms of the duty ratio (D) can be given by
 
(M1 + M2 )Ts M1 + M2
Ico IL = D + Mc +
2
Ts D (4.42)
2 2
which has a minima according to
 
Mc Ts Mc (M1 + M2 )Ts
|Ico IL |min = 1+ + (4.43)
2 M1 + M2 8
at the duty ratio (D)
1 Mc
D= + (4.44)
2 M1 + M2
The duty ratio (4.44), where the minima takes place, is the same value, where
the small-signal duty-ratio gain would become innite, because Fm can be
given generally as
1
Fm =   (4.45)
(D D)(M1 + M2 )
Ts Mc +
2
according to (4.8). This means also that the 100% of the duty-ratio range
requires the compensation to be designed as
M1 + M2
Mc = (4.46)
2
The parabola shape of (4.42) dictates that the difference would decrease along
the increase in the duty ratio until the duty ratio dened by (4.44) is reached.
After that, the difference should start increasing again, but it is physically
impossible within a single cycle, because the increasing duty ratio requires
 4.3 Modeling in CCM 137

Figure 4.5 Simulated

inductor-current waveforms
normal (solid line) and
subharmonic modes
(second-harmonic: dashed
line, fourth-harmonic: dash-dot
line) based on an
uncompensated buck
converter.

the difference to keep on decreasing as well. As a consequence, the converter

is forced to enter into the second harmonic mode of operation, where the
difference naturally increases and a stable operating point would be found as
shown in Figure 4.5. The possible stable subharmonic operation modes may
exist at fs /2n , where fs is the switching frequency and n = 1, 2, 3, . . . , until the
converter enters into a chaotic operation mode [17].
Equation (4.42) may be developed also in terms of D as
 
2Mc 2(Ico IL )
D 1+
2
D+ =0 (4.47)
M1 + M2 Ts (M1 + M2 )
having roots at
 2
1 Mc 1 Mc 2(Ico IL )
D1,2 = + + (4.48)
2 M1 + M2 2 M1 + M2 Ts (M1 + M2 )

If the difference (Ico IL ) in (4.48) is substituted with |Ico IL |min (4.43), the
quadratic equation in (4.47) would have a double root coinciding with (4.44),
which means that the real-valued solution exists only up to the duty ratio
equaling (4.44). Therefore, it may be obvious that the average comparator
Eq. (4.8) and the small-signal duty-ratio gain (Fm ) would predict correctly the
location and existence of the mode limit as also discussed in [20].
It is observed that the inductor current up (M1 ) and down (M2 ) slopes
maintain certain relation in the subharmonic mode, which is clearly visible in
Figure 4.6 as well. The formula dening mathematically the relation can be
derived as follows: the small-signal inductor-current loop (see Figure 4.4) has
innite duty-ratio gain (i.e., Fm = ) at the mode limit. As a consequence,
the perturbation in the inductor current would follow exactly the perturbation
in the control current, which is zero at open loop. This means that the
derivative of the time-averaged inductor current (
iL ) has to be zero, i.e.,
138 4 Average and Small-Signal Modeling of Peak-Current-Mode Control

Figure 4.6 Inductor-current waveforms in second-harmonic

mode: (a) measured inductor current with zero compensation and
(b) simulated inductor current with certain amount of
compensation.


dmax m1 dmax m2 = 0, where dmax equals (4.44). If we substitute dmax with
(4.44), we get

M2 = M1 + 2Mc (4.49)

which shows that the absolute values of the up and down slopes are same in
an uncompensated (i.e., Mc = 0) converter as also shown in Figure 4.6a.
 4.4 Modeling in DCM 139

From Figure 4.6b, we may compute that

(M1 + Mc )D1 = M2 D2 + Mc D2

(4.50)
(M1 + Mc )D2 = M2 D1 + Mc D1

which yields the average duty ratio (Dav ) as

D1 + D2 M2
Dav = = (4.51)
2 M1 + M2
If we denote M = Uo /Uin and consider the ideal converters, then we will get
by applying (4.51) that M = Dav for a buck converter, M = 1/Dav for a boost
converter, and M = Dav /Dav for a buckboost converter, which are similar to
the ideal modulos (i.e., M(D)) dened in [8]. The average duty ratio (4.51) can be
shown to be equal to the maximum duty ratio dened in (4.44) by substituting
M2 in (4.51) with (4.49) and applying some mathematical manipulation. This
means in practice that the output dynamics of the converter entered into a
subharmonic mode would be limited by Dav .

4.4
Modeling in DCM

The modeling of PCM control in DCM is the similar process as in CCM. The
same comparator equation (4.3) applies, and the main task is to nd a proper
denition for
iL , which can be solved from the inductor current waveforms
shown in Figure 4.7 according to its denition at t = (k + d)Ts . This procedure
yields
m1 d(d + d1 )Ts

iL = m1 dTs , (4.52)
2
where the rst term corresponds to the peak inductor current and the last term
to the average inductor current [16, 21].
The unknown duty ratio (d1 ) can be solved from the equality dm1 = d1 m2 ,
which yields

Figure 4.7 Inductor-current

waveforms in DCM.
140 4 Average and Small-Signal Modeling of Peak-Current-Mode Control

m1
d1 = d (4.53)
m2
Substituting d1 in (4.52) with (4.53) yields
m1 (m1 + m2 )d2 Ts

iL = m1 dTs (4.54)
2m2
and consequently the comparator equation (4.8) becomes as
m1 (m1 + m2 )d2 Ts
ico Mc dTs =
iL  + m1 dTs (4.55)
2m2
The coefcients of the duty-ratio constraints of (4.1) or (4.2) can be solved
from (4.55) by substituting the up and down slopes with their topology-based
values introduced in Chapter 3 (Section 3.5), and developing the proper partial
derivatives. It may be obvious that the small-signal duty-ratio gain (Fm ) can be
generally given by
1
Fm =   (4.56)
M1 (M2 (M1 + M2 )D
Ts Mc +
M2
It is obvious that the duty-ratio gain (Fm ) may become innite also in DCM
similarly to CCM, when the duty ratio is
M2 M2 Mc
D= + (4.57)
M1 + M2 M1 (M1 + M2 )
The value M2 /(M1 + M2 ) actually denes the mode limit between the DCM
and CCM operation and is symbolically same as the similar maximum value in
CCM (4.51). The simulated inductor-current waveforms shown in Figure 4.8
based on an uncompensated DCM buck converter having K = 0.45 (i.e.,
L = 9 H, Req = 4 , fs = 100 kHz) prove that the subharmonic operation
would also take place in DCM [21]. The possible subharmonic frequencies are
all the even and odd harmonics of the switching frequencies.

4.4.1
Duty-Ratio Constraints for Basic Converters

We develop the duty-ratio constraints for the basic converters corresponding to

(4.1) by utilizing the material provided in Chapter 3 (Section 3.5) and omitting
the effect of circuit parasitic elements.

4.4.1.1 Buck Converter

The inductor-current up slope is m1 = (
uin 
uo )/L and down slope
m2 =
uo /L. As a consequence, the averaged comparator equation can be
 4.4 Modeling in DCM 141

Figure 4.8 Simulated inductor-current waveforms in

subharmonic mode in a DCM buck converter (basic switching
frequency: solid line, second-harmonic: dashed line,
third-harmonic: dash-dot line).

given by

(
uin 
uo )dTs (
uin 
uo )
uin d2 Ts

ico Mc dTs =
iL  +
L 2L
uo 
(4.58)
and the corresponding small-signal duty-ratio constraints by
  
1 DTs D
d =   ico iL 1 uo
Uin (1 M)(M D) L 2M2
Ts Mc +
LM
  
DTs (2 M)D
1 uin (4.59)
L 2M
where M = Uo /Uin .
According to (4.59), the duty-ratio-constraints coefcients are:
1
Fm =  
Uin (1 M)(M D)
Ts Mc +
LM
qc = 1
  (4.60)
DTs D
qo = 1
L 2M2
 
DTs (2 M)D
qi = 1
L 2M
142 4 Average and Small-Signal Modeling of Peak-Current-Mode Control

4.4.1.2 Boost Converter

The inductor-current up slope is m1 =
uin /L and down slope m2 =
(
uo 
uin )/L. As a consequence, the averaged comparator equation can
be given by

uin dTs
uin 
uo d2 Ts

ico Mc dTs =
iL  + (4.61)
L 2L(
uo 
uin )
and the corresponding small-signal duty-ratio constraints by
1
d =  
Uin (D M 1)
Ts Mc +
L(M 1)
   
D2 Ts DTs DM2
ico iL uo 1 uin
2L(M 1)2 L 2(M 1)
(4.62)
where M = Uo /Uin .
According to (4.62), the duty-ratio-constraints coefcients are:
1
Fm =  
Uin (D M 1)
Ts Mc +
L(M 1)
qc = 1
  (4.63)
D2 Ts
qo =
2L(M 1)2
 
DTs DM2
qi = 1
L 2(M 1)

4.4.1.3 Buckboost Converter

The inductor-current up slope is m1 =
uin /L and down slope m2 =
uo /L.
As a consequence, the averaged comparator equation can be given by

uin dTs
uin (
uin  +
uo )d2 Ts

ico Mc dTs =
iL  + (4.64)
L 2L
uo 
and the corresponding small-signal duty-ratio constraints by
1
d =
Uin (D M D)
Ts (Mc + )
LM
   
D2 Ts DTs D(2 + M)
ico iL uo 1 uin (4.65)
2LM2 L 2M
where M = Uo /Uin .
 4.4 Modeling in DCM 143

According to (4.65), the duty-ratio-constraints coefcients are:

1
Fm =  
Uin (D M D)
Ts Mc +
LM
qc = 1
(4.66)
D2 Ts
qo =
2LM2
 
DTs D(2 + M)
qi = 1
L 2M

4.4.2
Small-Signal PCMC State Spaces for the Basic Converters
 
The state spaces are derived by substituting the perturbed duty ratio d in the
corresponding VMC state spaces (Chapter 3, Section 3.5) with
 
d = Fm ico iL qo uC qi uin (4.67)

where the coefcients are derived in Section 4.4.1 for the basic converters. The
transfer functions describing the converter dynamics can be solved naturally
from the presented state spaces, but the solving is left for the reader. The
dynamics of the PCMC converter in DCM is highly damped due to the upper
limit provided by the control current and the bottom limit provided by the
zero level. Therefore, the transfer functions would have mainly a rst-order
nature [21].
Buck:
   
diL Req K 2Fm Uin 1 K 2Fm qo
= + iL + uC
dt L 1M L L(1 M) 1 M L
 
(2 M)M K 2Fm Uin qi 2Fm Uin
+ uin + ico
L(1 M) 1 M L L

duC iL io
=
dt C C
 
2Fm Uo 1 M M2 2Fm Uo qo 1 M
iin = iL ( + ) uC
Req K Req (1 M) Req K
  
M2 2Fm Uo 1 M 2Fm Uo 1 M
+ uin + ico
Req (1 M) Req K Req K
duC
uo = uC + rC C (4.68)
dt
144 4 Average and Small-Signal Modeling of Peak-Current-Mode Control

Boost:
   
diL Req K(M 1) 2Fm Uo 1 KM 2Fm Uo qo
= + iL + uC
dt L M L L M1 L
 
M2 KM 2Fm Uo qi 2Fm Uo
+ uin + ico
L M1 L L
  
duC 2Fm Uo M 1 iL 2Fm Uo qo M 1 1
= 1+ + uC
dt Req KM C Req C KM Req C
  
M1 io 2Fm Uo M 1
M(M 1) 2Fm Uo qo uin ico
KM C Req C KM

iin = iL
duC
uo = uC + rC C (4.69)
dt
BuckBoost:
 
diL Req K 2Fm (Uin + Uo ) K 2Fm (Uin + Uo )qo
= + iL + uC
dt L L L L

M(M + 2) K 2Fm (Uin + Uo )qi 2Fm (Uin + Uo )
+ uin + ico
L L L
  
duC 2Fm Uo iL 2Fm Uo qo 1 2Fm Uo qi
= 1+ + uC M2 uin
dt Req K C Req C K Req C K

io 2Fm Uo
ico
C Req C K

2Fm Uo 2Fm Uo M2 2Fm Uo qi 2Fm Uo
iin = iL uC + uin + ico
Req K Req K Req Req K Req K
duC
uo = uC + rC C (4.70)
dt

4.4.3
Origin and Consequences of Mode Limit in DCM

The averaged comparator equation (4.55) may be presented also at steady state,
when the compensation is set to zero, by
M1 (M1 + M2 )Ts
Ico IL = M1 Ts D D2 (4.71)
2M2
 4.4 Modeling in DCM 145

Equation (4.71) is a similar quadratic equation as in CCM (4.42), which dictates

that the difference (Ico IL ) shall decrease along the increase in the duty ratio
until the minimum value
M1 M2 Ts
|Ico IL |min = (4.72)
2(M1 + M2 )
is reached. The corresponding maximum duty ratio (Dmax ) is
M2
Dmax = (4.73)
M1 + M2
which is known to dene the mode limit between the DCM and CCM
operation [8]. The averaged comparator equation can also be developed in
terms of duty ratio as
2M2 2M2 (Ico IL )
D2 D+ =0 (4.74)
M1 + M2 Ts M1 (M1 + M2 )
It may be obvious that (4.74) has a real-valued solution only up to the duty
ratio corresponding to (4.73).
The steady-state comparator equation (4.55) can also be developed in 
terms
of the output-to-input relation (M), when taking into account that D = M 1MKM

for a buck converter, D = KM(M 1) for a boost converter, and D = M K
for a buckboost converter. This procedure yields at resistive load for a buck
converter
 
Ico R 2
M M +K
3 2
=0 (4.75)
Uin

for a boost converter

 2
K Ico R
M2 M =0 (4.76)
Uin 2
and for a buckboost converter

K Ico
M= (4.77)
2Uin
Equation (4.75) can be developed further yielding
    
2 2 3Ico R 2
M M =0 (4.78)
3 4Uin

which shows that there exists a double root at M = 23 , which means that there
are no real-valued solutions for M after M = 23 at open loop as also discussed
in [8]. It can be proved theoretically that the mode limit at M = 23 exists only at
open loop: according to the inductor-current waveforms (Figure 4.6, Mco = 0),
146 4 Average and Small-Signal Modeling of Peak-Current-Mode Control

we can compute substituting D with M 1M K
and D1 with K(1 M) that a
real-valued solution for the control current can always be found as


 1 Uo
2Uo  Uin
Ico = (4.79)
R K
when the feedback loop is connected. This proves that the corresponding mode
limit does not exist under feedback control. According to (4.76) and (4.77),
any special open-loop mode limit does not exist in boost and buckboost
converters, because the roots of the corresponding equations are always real.
This also means that the articial compensation is not necessary in DCM,
when the converters are designed to operate in such a way that their duty ratio
is always less than Dmax dened in (4.73).

4.5
Dynamic Review

The buck and boost converters operating in CCM will be analyzed in detail.
The dynamical features of the PCM control in a buck converter are compared
experimentally to the corresponding VMC buck converter as analyzed in
Chapter 3 (Section 3.6). The PCM-controlled DCM converters are not analyzed
dynamically at all due to their limited importance in the eld.
A PCM-controlled converter cannot operate at open loop at a constant-current
load due to their current-output nature. This means that the internal dynamics
are not directly measurable, but we have to measure rst the load-affected
transfer functions and then compute the internal transfer functions from
the measured data by applying the load-affected transfer functions dened
in Chapter 2 (Eq. (2.30)). The commercially available electronic loads do not
usually provide the ideal features they are supposed to provide. Therefore, it is
highly recommended to use a pure resistor as a load. The output impedance
can be, however, measured directly, which would also make the computations
more deterministic.
As an example, the measured and computed control-to-output transfer
functions of a PCMC buck converter at the duty ratio of 0.3 and 0.4 are shown
in Figure 4.9, where the articial compensation (Mc ) is set to zero. In practice,
the measured control-to-output transfer functions also contain the effect of
the PWM modulator and the equivalent inductor-current-sensing resistor,
which are removed computationally from the responses shown in Figure 4.9.
At low frequencies, the load resistor totally hides the real behavior of the
control-to-output transfer function but not anymore at the higher frequencies.
The effect of the increasing duty ratio on the dynamics of the PCMC buck
converter is clearly visible in the internal control-to-output transfer function:
the DC gain increases, the low-frequency pole moves toward origin, and the
 4.5 Dynamic Review 147

Figure 4.9 Measured control-to-output transfer function versus

the internal control-to-output transfer function at the duty ratio of
0.3 and 0.4, where the load resistor has been 4 .

high-frequency pole toward innity. If the conclusions are drawn based on the
load-affected transfer functions, then it is obvious that the low-frequency gain
corresponds to the magnitude of the load resistor (i.e., 12 dB) and the low-
frequency pole is totally controlled by the load resistor (i.e., flow = 1/2RC).
The moving of the high-frequency pole to the higher frequencies is visible
but could be interpreted as a consequence of some other phenomenon as for
example the sampling effect [11].

4.5.1
Buck Converter

The dynamic prole of the PCM-controlled buck converter as shown in

Figure 4.10 is analyzed in CCM and compared to the dynamics associated with
the VMC converter reviewed earlier in Chapter 3 (Section 3.6.1). The PCM
control is implemented using a commercial PCM modulator UC3842 where
the PWM modulator contributes an extra gain of 1/3. The on-time inductor
current is measured using a current transformer having the turns ratio of 1/100
as shown in Figure 4.10. The equivalent inductor-current sensing resistor (Rs )
equals 75 m. The used articial compensation (Mc ) equals 1.45 Uo /2L.
The analysis would be done at the input voltages of 20 and 50 V, but
the practical measurements have been carried out at the input voltage of
20 V. The value of the duty ratio (D) at the input voltage of 20 V is 0.53
and 0.21 at 50 V as dened in Chapter 3 (Section 3.6.1). As a consequence,
sp sp
the duty-ratio gain (Fm ), the inductor-current-feedback gain (qc ), and the
sp
input-voltage-feedforward gain (qi ) (4.28) are at 20 V as
148 4 Average and Small-Signal Modeling of Peak-Current-Mode Control

Figure 4.10 Experimental

PCM-controlled buck converter.

sp
Fm = 1.576728

c = 0.99709
qsp (4.80)
= 1.186 102
sp
qi

and at 50 V as
sp
Fm = 0.484776

c = 0.99806
qsp (4.81)
= 7.9 103
sp
qi

4.5.1.1 Control-to-Output Transfer Function

The control-to-output transfer function (Gco ) is symbolically presented in
sp sp
(4.82). The increased damping due to the lossless resistor (Fm UE qc ) means
that the roots of the characteristic equation (i.e., the denominator of (4.82)) or
the poles are real and well separated.
sp
Fm UE (1 + s rC C)
LC (4.82)
sp sp
rE + Fm UE qc + rC 1
s2 + s +
L LC
As a consequence, the low-frequency pole can be approximated to be as
1 1
fplow sp sp (4.83)
2 Fm UE qc C

and the high-frequency pole as

sp sp
1 Fm UE qc
fphigh (4.84)
2 L
sp
The low-frequency gain of the transfer function equals Fm UE , respectively.
 4.5 Dynamic Review 149

The corresponding resistive-load-affected transfer function can be given as

sp
R Fm UE (1 + s rC C)

R + rC LC
GRco = sp sp (4.85)
(R(rE + Fm UE qc + rC )
sp sp sp sp
+ rC (rE + Fm UE qc ))C + L R + rE + Fm UE qc
s2 + s +
LC(R + rC ) LC(R + rC )

from which the low- and high-frequency poles and the low-frequency gain can
sp
be approximated, when Fm UE  rE , to be as
 
1 1 1
fplow sp sp +
2 Fm UE qc C RC
sp sp
1 Fm UE qc
fphigh (4.86)
2 L
sp
RFm UE
|GRco |low sp sp
R + Fm UE qc
sp
Typically, Fm UE  R. Therefore, fplow approaches to 1/2RC, and the low-
frequency gain to R as demonstrated in Figure 4.9. As a consequence, the
internal dynamics is not fully observable.
The measured control-to-output transfer functions at D = 0.3 and 0.4
(Figure 4.8) are repeated in Figure 4.11, where the dots and squares are
the corresponding predictions using (4.82). The modeling accuracy is obvious.

Figure 4.11 The measured internal control-to-output transfer functions at D = 0.3 and
0.4, where the dots and squares denote the predicted responses.
150 4 Average and Small-Signal Modeling of Peak-Current-Mode Control

In order to compare the dynamic changes imposed by the PCM control to

the VMC converter, we give the control-to-output transfer functions of both of
the converters at 20 V as
s 6.2 103 + 5.93 108
GVMC =
co
s2 + s 2.08 103 + 3 107
(4.87)
s 9.756 103 + 9.356 108
GPCMC = 2
co
s + s 2.9687 105 + 3 107
and at 50 V as
s 1.6 104 + 1.5 109
GVMC =
co
s2 + s 1.33 103 + 3 107
(4.88)
s 7.57 103 + 7.26 108
GPCMC = 2
co
s + s 2.303 105 + 3 107
The damping factor ( ) (Chapter 2, Section 2.6.2) varies from 0.19 to 0.12, and
from 27.1 to 21.0, for the VMC and PCMC converters, respectively, where the
rst value corresponds to 20 V and the second value to 50 V. The main reason
sp sp
for the increased damping is the lossless series resistor (Fm UE qc ), which is
equal to 30.95  at 20 V and 24.04  at 50 V. As a consequence, the poles of
the PCMC converter are real and well separated. The corresponding frequency
responses are shown in Figure 4.12: typical to the PCMC transfer function
is the rst-order nature and the high independence from the changes in the
input voltage. This means that the crossover frequency of the voltage-loop
gain does not change when the input voltage changes as is the case with the
VMC transfer function. The high-frequency phase lag of the PCMC transfer

Figure 4.12 The predicted control-to-output transfer functions of a VMC and PCMC buck
converters at the input voltages of 20 and 50 V.
 4.5 Dynamic Review 151

function is highest at the high-line condition and, therefore, the controller

should be designed also at the high line to ensure robust stability.

4.5.1.2 Output Impedance

The open-loop output impedance (Zoo ) of the PCMC converter is presented
sp sp
symbolically in (4.89). Its low-frequency value equals rE + Fm UE qc , where rE
is the low-frequency output impedance of the corresponding VMC converter.
Its high-frequency value equals rC . It may be obvious that the high-frequency
pole (4.84) and the inductor-based zero would effectively cancel each other,
sp sp
when Fm UE qc  rE .
sp sp
(rE + Fm UE qc + sL)(1 + s rC C)
LC (4.89)
sp sp
rE + Fm UE qc + rC 1
s2 + s +
L LC
The measured internal output impedances corresponding to the control-to-
output transfer functions of Figure 4.9 are shown in Figure 4.13, where the
dots and the squares denote the predicted responses. It is obvious that any
resonant-like behavior does not exist at half the switching frequency, as the
method used in [11] would predict. The output impedance has clearly rst-
order nature as discussed above. The model-based prediction (i.e., the dots
(D = 0.3) and squares (D = 0.4)) would also produce quite accurate results.
Along the increase in the duty ratio, the low-frequency gain increases and the
low-frequency pole moves closer to the origin.

Figure 4.13 Directly measured internal output impedance at D = 0.3 and 0.4. The dots
and squares denote the predicted responses.
152 4 Average and Small-Signal Modeling of Peak-Current-Mode Control

Figure 4.14 The predicted internal output impedances of VMC and PCMC buck converter
at the input voltage of 50 V.

The VMC and PCMC output impedances are given numerically in (4.90)
and the corresponding frequency responses in Figure 4.14 only at the input
voltage of 50 V due to minimal changes in their behavior with respect to the
input voltage.
s2 3.3 102 + s 3.2 103 + 3.23 106
VMC
Zoo =
s2 + s 1.33 103 + 3 107
(4.90)
s2 3.3 102 + s 1.075 104 + 7.28 108
PCMC
Zoo =
s2 + s 2.303 105 + 3 107
The load impedance will affect the converter dynamics as if the open-loop
output impedance would behave as a weighting gain as discussed in Chapters
2 and 3 (i.e., GLco = Gco /(1 + Zoo /ZL )). Therefore, we can predict the load
sensitivity of the converter according to the shape of the open-loop output
impedance (i.e., both the magnitude and phase), [23]: the VMC converter is
sensitive to capacitive load especially at the frequencies close to the resonant
frequency. The PCMC converter cannot become unstable due to capacitive
load, although its voltage-loop crossover frequency may be reduced easily. We
will discuss more in detail the dynamical load effects in Chapter 8.

4.5.1.3 Input-to-Output Transfer Function

The open-loop input-to-output transfer function (Gioo ) can be given
symbolically by
sp sp
(D Fm UE qi )(1 + s rC C)
Gioo = LC (4.91)
sp sp
rE + Fm UE qc + rC 1
s2 + s +
L LC
 4.5 Dynamic Review 153

It may be obvious that the input noise attenuation may be theoretically very
sp sp
high if D Fm UE qi = 0. This means that the articial compensation (Mc )
had to be set to DUE /2L. According to (4.29), UE is dependent on the operating
point and, therefore, the ideal compensation cannot be maintained. This
compensation value is known as an optimal compensation and is typically
expressed as Uo /2L[8]. Physically, it means that the average inductor current
is maintained constant at open loop and, therefore, the output voltage does
not change when the input voltage changes. As we discussed earlier, we had to
use light overcompensation (i.e., Mc 1.45 Uo /2L) in order to ensure signal
integrity in the inductor-current loop and, therefore, the attenuation would
also be nite.
The numerical values of the input-to-output transfer functions of the VMC
and PCMC converters are given at 20 V by
s 167 + 1.6 107
ioo =
GVMC
s2 + s 2.08 103 + 3 107
(4.92)
s 50.84 + 4.876 106
GPCMC = 2
ioo
s + s 2.9687 105 + 3 107
and at 50 V by
s 66 + 6.3 106
ioo =
GVMC
s2 + s 1.33 103 + 3 107
(4.93)
s 6.19 + 5.94 105
GPCMC = 2
ioo
s + s 2.303 105 + 3 107
The corresponding frequency responses are shown in Figure 4.15a, where the
upper curve within the group (i.e., VMC or PCMC) corresponds to 20 V and
the lower curve to 50 V. It is obvious that the PCM control provides much
higher attenuation at all frequencies than the VM control. The frequency
responses measured (solid lines) and predicted (dashed line) at 20 V are
shown in Figure 4.15b. The reason for the deviation between the measured
and predicted responses of the PCMC converter is the assumed level of
articial compensation, which deviates from the real compensation.
The input-to-output transfer function determines the level of the source
interactions and the propagation of the load interactions through the
converter [2426]. Therefore, it may be obvious that the interactions taken
place at the frequencies close to the resonant frequency would affect easily
the VMC converter but not the PCMC converter. The interactions are treated
more in detail in Chapter 8.

4.5.1.4 Input Admittance

The open-loop input admittance (Yino ) is symbolically given in (4.94).
If the converter were optimally compensated (i.e., Mc = DUE /2L), then
sp sp
Yino = Fm qi IL = DIL /UE . In practice, such a condition is not possible as
154 4 Average and Small-Signal Modeling of Peak-Current-Mode Control

Figure 4.15 The input-to-output transfer curves within the group correspond to 20 V)
functions of the VMC and PCMC converters: and (b) the measured and predicted transfer
(a) the predicted transfer functions at the functions at 20 V.
input voltages of 20 and 50 V (the upper

discussed earlier.

sp sp 
sp sp 
D Fm UE qi D Fm IL qc s
L sp sp
Fm qi IL (4.94)
sp sp
r E + F U
m E c q + rC 1
s +s
2 +
L LC
 4.5 Dynamic Review 155

The corresponding numerical values of the input admittances of the VMC and
PCMC converters are given at 20 V by

s 2.86 103
VMC
Yino =
s2 + s 2.08 103 + 3 107
(4.95)
s2 4.676 102 + s 1.91 104 + 1.4 106
PCMC
Yino =
s2 + s 2.9687 105 + 3 107
and at 50 V by

s 4.2 102
VMC
Yino =
s2 + s 1.33 103 + 3 107
(4.96)
s2 9.58 103 + s 4.08 103 + 2.872 105
PCMC
Yino =
s2 + s 2.303 105 + 3 107
The corresponding frequency responses are shown in Figure 4.16a, where the
upper curves within the group (i.e., VMC or PCMC) correspond to 20 V and
the bottom curves to 50 V. The frequency responses measured (dashed lines)
and predicted (solid lines) are shown in Figure 4.16b. The high-frequency
deviation between the measured and predicted responses is due to the circuit
parasitic elements in the practical converter.
The open-loop input admittance or its inverse impedance is one of
the decisive factors describing the effect of the source interactions in a
converter [24, 26]. The lack of resonant peaking in the PCMC converter
implies improved insensitivity to the source interactions compared to the VMC
converter. More detailed treatment of the subject is presented in Chapter 8.

4.5.1.5 Ideal Input Admittance

Gioo Gci 
The ideal admittance (Yin ) i.e., Yino Gco
is same for the VMC and
PCMC converters and can be given by
DIL
(4.97)
UE
The corresponding measured-data-based frequency responses are shown in
Chapter 3 (Section 3.6.1, Figure 3.23). The open-loop input admittance of the
PCMC buck converter is quite close to the ideal input admittance, even if the
articial compensation is not ideal as shown in Figure 4.17 because of a rather
high attenuation provided by the open-loop input-to-output transfer function.

4.5.1.6 Short-Circuit Input Admittance
Gioo Toio 

The short-circuit input admittance (Yinsc ) i.e., Yino + Zoo
can be
symbolically given by
156 4 Average and Small-Signal Modeling of Peak-Current-Mode Control

Figure 4.16 The input admittances of the VMC and PCMC

converters: (a) the predicted responses at the input voltages of
20 V (the upper curves within the group) and 50 V (the bottom
curves within the group) and (b) the measured and predicted
responses at 20 V.

sp sp 
sp sp 
D Fm UE qi D Fm IL qc sp
sp sp Fm
sp
qi IL (4.98)
rE + Fm UE qc + sL

Yinsc equals Yin for an optimally compensated converter due to Gioo = 0.

The corresponding numerical values of the VMC and PCMC converters are
 4.5 Dynamic Review 157

Figure 4.17 The ideal (solid line) and open-loop input admittance (dashed line) of the
PCMC converter at the input voltage of 20 V.

given at 20 V by
1.505
VMC
Yinsc =
1 + s 5.65 104
(4.99)
s 1.58 107 + 6.44 102
PCMC
Yinsc =
1 + s 3.37 106
and at 50 V by
0.409
VMC
Yinsc =
1 + s 9.77 104
(4.100)
s 4.16 108 + 1.039 102
PCMC
Yinsc =
1 + s 4.34 106
The corresponding predicted frequency responses are given in Figure 4.18,
where the upper curves within the group correspond to 20 V and the bottom
curves to 50 V.

4.5.2
Boost Converter

The dynamic prole of the PCM-controlled boost converter shown in

Figure 4.19 is analyzed in CCM and compared to the dynamics associated with
the corresponding VMC converter reviewed earlier in Chapter 3 (Section 3.6.2).
The PCM control is assumed to be implemented using the same commercial
158 4 Average and Small-Signal Modeling of Peak-Current-Mode Control

Figure 4.18 The predicted frequency responses of the

short-circuit input admittance of the VMC and PCMC converters
at the input voltages of 20 V (the upper curves within the group)
and at 50 V (the bottom curves within the group).

Figure 4.19 PCM-controlled

boost converter.

PWM modulator as in the case of the buck converter. The on-time inductor
current is measured using the equivalent sensing resistor of 150 m as shown
in Figure 4.19. The articial compensation (Mc ) is set to Uo /2L, which would
provide a 100% duty-ratio range for all the input-voltage values. The required
compensation value can be computed by setting the denominator of the
duty-ratio gain (Fm ) to zero.
The analysis would be done at the input voltages of 20 and 50 V, where the
duty ratio (D) equals 0.748 and 0.339, and the inductor current (IL ) to 5.95 A and
2.27 A, respectively, as dened in Chapter 3 (Section 3.6.2). As a consequence,
sp sp
the duty-ratio gain (Fm ), the inductor-current-feedback gain (qc ), the output-
sp sp
voltage-feedback gain (qo ), and the output-current-feedforward gain (qio ) (4.33)
are at 20 V as
sp
Fm = 1.8411

c = 0.9997
qsp
 4.5 Dynamic Review 159

3
o = 2.694 10
qsp
qio = 8.889 105
sp
(4.101)

and at 50 V as
sp
Fm = 0.70644

c = 0.99964
qsp
(4.102)
3
o = 3.1992 10
qsp
qio = 1.056 104
sp

4.5.2.1 Control-to-Output Transfer Function

The control-to-output transfer function (Gco ) is symbolically presented in
sp
(4.103). The increased damping due to the lossless resistor (Fm UE qc ) means
that the roots of the characteristic equation (i.e., the denominator of (4.103))
or the poles are real and well separated. The RHP zero present in the transfer
function is the same as the RHP zero of the corresponding VMC converter.
sp  
Fm (D UE IL (rE + D2 rC )) LIL
1s  (1 + s rC C)
LC D UE IL (rE + D2 rC )


 2 sp sp sp sp 
rE + D rC + Fm UE qc Fm IL qo

=s +s
2

L C
sp sp sp sp sp
D2 + D Fm (UE qo + IL qc ) Fm IL qo (rE + D2 rC )
+ (4.103)
LC
Because of the well-separated roots, the low-frequency pole can be approxi-
mated by
sp sp sp
D2 + D Fm (UE qo + IL qc )
fplow sp (4.104)
Fm UE qc C
and the high-frequency pole by
sp sp
Fm UE qc
fphigh (4.105)
L
The low-frequency gain can be approximated by
UE
|Gco |low sp sp (4.106)
UE qo + IL qc
According to (4.106), we may conclude that the increase in the duty-ratio
gain (Fm ), when approaching the mode limit, is not clearly visible in the
low-frequency gain similarly to the gain of the PCMC buck converter (4.86). It
160 4 Average and Small-Signal Modeling of Peak-Current-Mode Control

is, however, obvious that the low-frequency pole (4.104) would move toward
the origin and the high-frequency pole (4.105) toward the innity, when the
duty-ratio gain increases.
In order to compare the dynamics imposed by the PCM control in the VMC
boost converter, we give numerically the control-to-output transfer functions
of both of the converters at 20 V as
s2 0.2 s 1.72 102 + 1.61 108
GVMC =
co
s2 + s 494 + 5.75 105
(4.107)
s2 0.36 s 3.158 104 + 2.96 108
GPCMC =
co
s2 + s 3.926 105 + 2.633 107
and at 50 V as
s2 7.49 102 s 2.54 103 + 4.45 108
GVMC =
co
s2 + s 363 + 3.95 106
(4.108)
s2 5.29 102 s 1.794 103 + 3.146 108
GPCMC =
co
s2 + s 1.517 105 + 1.453 107
The corresponding frequency responses of the PCMC control-to-output
transfer functions are shown in Figure 4.20, where the dots and squares
are the switching-model-based simulated responses. The match between the
predictions and simulated responses is obvious.
The comparison between the VMC and PCMC transfer functions are shown
in Figure 4.21: the high-frequency magnitude behavior of the PCMC transfer
function at 20 V implies that the voltage-loop crossover frequency would be

Figure 4.20 Predicted and switching-model-based simulated

(dots and squares) frequency responses of the control-to-output
transfer function of the PCMC boost converter at the input
voltages of 20 V (solid line) and 50 V (dashed line).
 4.5 Dynamic Review 161

Figure 4.21 The frequency responses of the control-to-output

transfer functions of the VMC and the PCMC boost converters at
the input voltages of 20 V (solid lines) and 50 V (dashed lines).

shorter in the PCMC converter than in the VMC converter in order to maintain
sufcient gain margin. If considering the situation from the viewpoint of the
phase, the conclusion might be different but the magnitude behavior would
be most decisive. The stability of the converter would also be quite sensitive
to the size of the ESR (rC ) of the output capacitor: an increase in the ESR is
directly reected in an increase in the high-frequency gain.

4.5.2.2 Output Impedance

The open-loop output impedance (Zoo ) of the PCMC boost converter is
presented symbolically in (4.109). Its low-frequency value can be approximated
sp sp sp sp sp
to equal Fm UE qc /(D2 + D Fm (UE qo + IL qc )), and the high-frequency value
to rC . It may be obvious that the high-frequency zero and pole would cancel
each other effectively and, therefore, the response would resemble a rst-order
response.
  sp
 2
1 q (D UE I L (rE + D rC ))
rE + Fmsp
c + Fm
UE qsp sp io
sp + s L (1 + s rC C)
LC qc IL D rC


 2 sp sp
sp sp 
rE + D rC + Fm IL qo
Fm UE qc

=s +s2

L C
sp
sp sp  sp sp

D2 + D Fm UE qo + IL qc Fm IL qo rE + D2 rC
+ (4.109)
LC
162 4 Average and Small-Signal Modeling of Peak-Current-Mode Control

In order to compare the dynamic changes imposed by the PCM control in the
VMC boost converter, we give numerically the output impedances of both of
the converters at 20 V as
s2 3.3 102 + s 3.18 103 + 1.54 106
VMC
Zoo =
s2 + s 494 + 5.75 105
(4.110)
s2 3.3 102 + s 1.612 104 + 1.24 109
PCMC
Zoo =
s2 + s 3.926 105 + 2.633 107
and at 50 V as
s2 3.3 102 + s 3.18 103 + 1.02 106
VMC
Zoo =
s2 + s 363 + 3.95 106
(4.111)
s2 3.3 102 + s 8.17 103 + 4.8 108
PCMC
Zoo =
s2 + s 1.517 105 + 1.453 107
The frequency responses of the output impedance at the input voltages of
20 V (solid line) and 50 V (dashed line) are shown in Figure 4.22, where
the switching-model-based responses are denoted by dots and squares. The
match with the analytical predictions and the switch-model-based responses
is obvious as well as the rst-order nature.
The comparison between the VMC and PCMC output impedances is
shown in Figure 4.23. The effect of PCM control is essentially similar to the
corresponding effect in the buck converter: the low-frequency gain increases
and the output impedance changes to have more capacitive nature compared

Figure 4.22 Predicted and switching-model-based (dots and

squares) frequency responses of the output impedance of the
PCMC boost converter at the input voltages of 20 V (solid line)
and 50 V (dashed line).
 4.5 Dynamic Review 163

Figure 4.23 The frequency responses of the output impedances of the VMC and the
PCMC boost converters at the input voltages of 20 V (solid lines) and 50 V (dashed lines).

to the corresponding VMC converter. The capacitive nature of the output

impedance means that the instability sensitivity to the capacitive load naturally
reduces.

4.5.2.3 Input-to-Output Transfer Function

The input-to-output transfer function (Gioo ) can be symbolically given by
sp sp
(D + Fm IL qc )
(1 + s rC C)
LC


 sp sp sp sp 
rE + D2 rC + Fm UE qc Fm IL qo

= s2 + s
L C
sp sp sp sp sp
D2 + D Fm (UE qo + IL qc ) Fm IL qo (rE + D2 rC )
+ (4.112)
LC
In order to compare the dynamic changes imposed by the PCM control into
the VMC converter, we give numerically the input-to-output transfer functions
of both the converters at 20 V as
s 17.79 + 1.7 106
ioo =
GVMC
s2 + s 494 + 5.75 105
(4.113)
s 1.056 103 + 1.013 108
GPCMC = 2
ioo
s + s 3.926 105 + 2.633 107
164 4 Average and Small-Signal Modeling of Peak-Current-Mode Control

and at 50 V as
s 21.14 + 2.03 106
ioo =
GVMC
s2 + s 363 + 3.95 106
(4.114)
s 2.134 102 + 2.047 107
GPCMC = 2
ioo
s + s 1.517 105 + 1.453 107
The corresponding frequency responses are given in Figure 4.24 at the input
voltages of 20 and 50 V. It is obvious that low-frequency input-noise attenuation
does not exist, but it does at the high input voltage in the VMC converter. This
means that the boost converter is sensitive to the source interactions and the
load interactions would also easily make the source interactions more severe.

4.5.2.4 Input Admittance

The open-loop input admittance (Yino ) can be symbolically given by
 sp sp 
1 Fm IL qo
s
L C


 sp sp sp sp 
rE + D2 rC + Fm UE qc Fm IL qo

= s2 + s (4.115)
L C
sp
sp sp  sp sp

D2 + D Fm UE qo + IL qc Fm IL qo rE + D2 rC
+
LC

Figure 4.24 The frequency responses of the input-to-output

transfer functions of the VMC and the PCMC boost converters at
the input voltages of 20 V (solid lines) and 50 V (dashed lines).
 4.5 Dynamic Review 165

The VMC and PCMC input admittances are given numerically for comparison
at 20 V as
s 2.86 103
VMC
Yino =
s2 + s 494 + 5.75 105
(4.116)
s 2.86 103 2.668 105
PCMC
Yino = 2
s + s 3.926 105 + 2.633 107
and at 50 V as
s 2.86 103
VMC
Yino =
s2 + s 363 + 3.95 106
(4.117)
s 2.86 103 4.64 104
PCMC
Yino = 2
s + s 1.517 105 + 1.453 107
The corresponding frequency responses are shown in Figure 4.25. The lack
of resonant peaking in the PCMC converter would make the converter less
sensitive to the source interactions than the corresponding VMC converter.
The effect is similar as in the buck converter.

4.5.2.5 Ideal Input Admittance

The ideal input admittance (Yin ) can be given by
IL 1

 (4.118)
D UE IL rE + D2 rC LIL


1s
D UE IL rE + D2 rC


Figure 4.25 The frequency responses of the input admittances of the VMC and the
PCMC boost converters at the input voltages of 20 V (solid lines) and 50 V (dashed lines).
166 4 Average and Small-Signal Modeling of Peak-Current-Mode Control

which is exactly the same as in the corresponding VMC converter. The

frequency responses are presented in Chapter 3 (Section 3.6.2, Figure 3.30).

4.5.2.6 Short-Circuit Input Admittance

The short-circuit admittance (Yinsc ) can be symbolically given by
sp sp
1 Fm IL qio

sp
  
sp 
rE + Fm UE qc + Fm qio D UE IL (rE + D2 rC ) qsp
sp sp sp 
c IL D rC + s L 1 Fm IL qio
sp

(4.119)

The VMC and PCMC short-circuit admittances are given for comparison at
20 V as
5.86
VMC
Yinsc =
1 + s 2.05 103
(4.120)
7.276 103
PCMC
Yinsc =
1 + s 2.547 106
and at 50 V as
8.9
VMC
Yinsc =
1 + s 3.11 103
(4.121)
1.887 102
PCMC
Yinsc =
1 + s 6.604 106

Figure 4.26 The frequency responses of the short-circuit input

admittances of the VMC and the PCMC boost converters at the
input voltages of 20 V (solid lines) and 50 V (dashed lines).
 References 167

as well as the corresponding frequency responses in Figure 4.26. The much

lower value of the short-circuit input admittance in the PCMC converter
compared to the VMC converter implies reduced sensitivity to the source
interactions.

References

1. C.W. Deisch, Simple switching 9. S. Banerjee and G.C. Verghese,

control method changes power Nonlinear Phenomena in Power
converter into a current source, in Electronics, IEEE Press, NJ, USA, 2001.
Proc. IEEE Power Electronics Specialists 10. C.K. Tse, Complex Behaviour of
Conf., 1978, pp. 300306. Switching Power Converters, CRC Press,
2. A. Capel, G. Ferrante, D. OSullivan, Boca Raton, FL, USA, 2004.
and A. Wienberg, Application of the 11. R.B. Ridley, A new continuous-time
injected current model for the model for current-mode control, IEEE
dynamic analysis of switching Trans. Power Electron., vol. 6, no. 2,
regulators with the new concept of 1991, pp. 271280.
LC3 modulator, in Proc. IEEE Power 12. R.B. Ridley, A new continuous-time
Electronics Specialists Conf., 1978, model for current-mode control with
pp. 135147. constant frequency, constant on-time,
3. R. Redl and N.O. Sokal, What a and constant off-time, in CCM and
design engineer should know about DCM, in Proc. IEEE Power Electronics
current mode control, in Proc. IEEE Specialists Conf., 1990, pp. 382389.
Power Electronics Design Conference, 13. C.P. Schultz, A unied model of
1985, pp. 1833. constant frequency switching
4. D.M. Mitchell, Tricks of the trade: regulators using multiloop feedback
understanding the right-half-plane control, in Proc. Power Conversion and
zero in small-signal DCDC converter Intelligent Motion Conf., 1993,
models, IEEE Power Electronics Society pp. 319329.
Newsletter, January 2001, pp. 56. 14. F.D. Tan and R.D. Middlebrook, A
5. G.K. Schoneman and D.M. Mitchell, unied model for current-
Output impedance considerations for programmed converters, IEEE Trans
switching regulators with current Power Electron., vol. 10, no. 4, July
injected control, IEEE Trans. Power 1995, pp. 397408.
Electron., vol. 4, no. 1, 1989, pp. 2535. 15. J. Sun and R.M. Bass, A new
6. S.-H. Hsu, A. Brown, L. Rensink, approach to average modeling of PWM
and R.D. Middlebrook, Modeling and converters with current-mode control,
analysis of switching DC-to-DC in Proc. IEEE Industrial Electronics
converters in constant-frequency Annual Conf., 1997, pp. 599604.
current programmed mode, in Proc. 16. T. Suntio, Unied derivation and
IEEE Power Electronics Specialists Conf., analysis of duty-ratio constraints for
1979, pp. 248301. peak-current-mode control in
7. A.S. Kislovski, Unied model for continuous and discontinuous modes,
open-loop instability, in Proc. IEEE in Proc. IEEE Industrial Electronics
Applied Power Electronics Conf., 1991, Annual Conf., 2002, pp. 13981403.
pp. 459465. 17. T. Suntio, M. Hankaniemi,
8. R.W. Erickson and D. Maksimovic, and T. Roinila Dynamical modelling
Fundamentals of Power Electronics, of peak-current-mode controlled
Kluwer, Norwell, MA, USA, 2001, 2nd converter in continuous conduction
Edition. mode, J. Simul. Modelling Pract.
168 4 Average and Small-Signal Modeling of Peak-Current-Mode Control

Theory, vol. 15, no. 10, 2007, direct-on-time control, IEEE Trans.
pp. 13201337. Ind. Electron., vol. 53, no. 1, 2006,
18. M.K. Kazimierczuk, Transfer pp. 287295.
function of current modulator in 23. M. Hankaniemi, M. Karppanen,
PWM converters with current-mode and T. Suntio, Load imposed
control, IEEE Trans. Circuits Syst. I, instability and performance
Fundam. Theory Appl., vol. 47, no. 9, degradation in a regulated converter,
2000, pp. 14071412. IEE Proc. Electr. Power Appl., vol. 153,
19. B. Johansson, A comparison and no. 6, 2006, pp. 781786.
improvement of two-continuous time 24. T. Suntio, M. Hankaniemi,
models for current-mode control, in and M. Karppanen, Analysing
Proc. IEEE International dynamics of regulated converters, IEE
Telecommunications Energy Conf., 2002, Proc. Elect. Power Appl., vol. 153, no. 6,
pp. 552559. 2006, pp. 905910.
20. J. Sun and B. Choi, Average modeling 25. T. Suntio, K. Kostov, T. Tepsa,
and switching instability prediction for and J. Kyyra, Using input invariance
peak-current control, in Proc. IEEE as a method to facilitate system design
Power Electronics Specialists Conf., in DPS applications, J. Circuits Syst.
2005, pp. 27642770. Comput., vol. 13, no. 4, 2004,
21. T. Suntio, Analysis and modelling of pp. 707723.
peak-current-mode controlled buck 26. M. Hankaniemi, M. Karppanen,
converter in DICM, IEEE Trans. Ind. T. Suntio, A. Altowati, and K. Zenger,
Electron., vol. 48, no. 1, 2001, Source-reected load interactions in a
pp. 127135. regulated converter, in Proc. IEEE
22. T. Suntio, Unied average and Industrial Electronics Annual Conf.,
small-signal modeling of 2006, pp. 28932898.
 169

5
Average and Small-Signal Modeling of Average-Current-Mode
Control

5.1
Introduction

Average-current-mode (ACM) control was introduced in the early 1990s [1] to

alleviate the disadvantages related to the peak-current-mode (PCM) control
such as the limited-duty-ratio range and high-frequency noise sensitivity [2, 3].
ACM control is typically used in power-factor-correction applications [4] and
in the converters interfacing the solar panels [5, 6]. The duty-ratio generation
is basically identical to the method utilized in the VMC control, but the control
signal contains both the output-voltage-loop control signal and the averaged
inductor current. The internal dynamics of the converter would naturally
change by application of ACM control but may resemble either the dynamics
of VM or PCM control depending on the level of inductor-current ripple left
in the duty-ratio generation. It may be obvious that the dynamical modeling
is quite similar to the PCM modeling introduced in Chapter 4. Therefore, it
is also natural that the proposed modeling methods in [7, 8, 11, 13, and, 14]
would utilize the modeling of PCM control [9, 12, 15]. We will introduce the
ACM modeling only in the continuous conduction mode. The DCM models
can be naturally derived by applying the proposed method and using the
results developed for the PCM control in DCM (Chapter 4, Section 4.4).

5.2
ACM-Control Principle

Under ACM control, the duty ratio (d) is generated comparing the output
signal (uca ) of the current-loop amplier and the constant ramp signal (Rs Mc )
provided by the PWM modulator as shown in Figure 5.1, where Rs is the
inductor-current equivalent sensing resistor, and Mc the slope of the PWM
ramp in current domain. The duty ratio is established when the output signal
(uca ) of the current-loop amplier reaches the PWM ramp signal. The output
signal of the current-loop amplier (uca ) can be given by

uca = uco + Gca (uco Rs iL ) (5.1)

Dynamic Prole of Switched-Mode Converter. Teuvo Suntio

2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
ISBN: 978-3-527-40708-8
170 5 Average and Small-Signal Modeling of Average-Current-Mode Control

Figure 5.1 ACM-control principles: (a) circuit schematics with a buck converter and
(b) duty-ratio generation.

where uco is the output signal of the voltage-loop amplier and Gca the
transfer function of the current-loop amplier. The instantaneous inductor
current (iL ) can be presented as a sum of the time-averaged inductor current

iL  and the triangular-shaped switching ripple (iL-ripple ) ripple. Therefore, the
current-amplier output signal can be given, according to [13], by

uca = uco + Gca (uco Rs
iL ) Gca iL-ripple (5.2)

The current-loop amplier is typically a PI-type controller with an extra high-

frequency pole as depicted in Figure 5.1a. It is obvious that the inductor-current
ripple may affect the converter dynamics, if the extra pole does not remove
the ripple-current term in (5.2): the basic dynamics of the converter is dened
assuming that the duty ratio is generated based on the averaged behavior
 5.3 Modeling with Full-Ripple-Current Feedback 171

of (5.2), i.e.,
uca  =
uco  + Gca (
uco 
iL ). If the ripple term is not zero,
the dynamics would be changed. The dynamical models presented in [10] are
based on the assumption of the zero-ripple conditions.
In practice, the proper dynamical models exist only either with the full- [13]
or with zero-ripple conditions [10]. The full-ripple effect would take place when
the current-loop amplier is implemented as a perfect PI controller without
the extra high-frequency pole.

5.3
Modeling with Full-Ripple-Current Feedback

ACM control is also a method to change the dynamics imposed by the direct-
duty-ratio or VM control similar to PCM control. Therefore, the fundamental
issue is to nd the proper duty-ratio constraints.
The inductor-current loop (Figure 5.1a) is provided with an average lter
or PI controller, which is shown in Figure 5.2a in current domain, i.e., the
voltage signals are replaced with the corresponding current signals to facilitate
the development of the duty-ratio constraints. Its transfer function (Gca ) can
be given by
1 + s R f Cf
Gca =   (5.3)

 Cf Cp
s Rin Cf + Cp 1 + s Rf
C f + Cp

and the corresponding frequency response is presented conceptually in

Figure 5.2b, where the effect of the location of the extra high-frequency
pole (fp ) is explicitly shown. When doing the modeling, we assume that fp is
much higher than the switching frequency (fs ). This means that the amplier
gain at the switching frequency is Kf Rf /Rin (see Figure 5.2b), and conse-
quently, the inductor-current switching-frequency ripple would be weighted
by Kf . Therefore, the amplier output signal (5.2) can be presented in current
domain by

ica = (1 + Gca )ico Gca
iL  Kf iL-ripple (5.4)

The corresponding duty-ratio generation process, when taking into account

(5.4), is shown in Figure 5.3, where
ica  = (1 + Gca )
ico  Gca
iL , m1 and
m2 are the corresponding inductor-current up and down slopes as well as
Kf is the current-loop-amplier gain at the switching frequency as dened
above.
According to the above-presented duty-ratio-generation process, it may be
obvious that
iL is the same value as dened in the conjunction with the PCM
modeling, that is,
iL = dd Ts /2(m1 + m2 ). This means that the duty-ratio
constraints can be given by
172 5 Average and Small-Signal Modeling of Average-Current-Mode Control

Figure 5.2 Current-loop PI controller: (a) physical amplier and (b) its frequency
response.

Figure 5.3 The duty-ratio generation

process in current domain.
 5.3 Modeling with Full-Ripple-Current Feedback 173

1
d = 
 
Kf D D UE
Ts Mc +
2L
 
(1 + Gca ) ico Gca iL Kf qo uo Kf qi uin (5.5)

by applying the duty-ratio constraints developed for the PCM control, where
UE , qo , and qi are dened in Chapter 4 (Section 4.3.1) for the buck, boost, and
buckboost converters.
According to (5.5), it is obvious that the current-loop-amplier gain at the
switching frequency would have a fundamental effect on the ripple effects. It
may also be obvious that the similar mode limit observed in the PCM control
does not easily exist, because the converter is automatically compensated by
means of Mc . Reference [11] provides excessive data on determining the value
of Fm for an ACM-controlled buck converter, and coming up to the conclusion
that Fm = 1/Rs Ts Mc = 1/VM , where VM is the peak-to-peak voltage of the
PWM ramp. The tests were carried out at the operating point close to D = 0.5,
and in addition by using very low value of Kf . According to (5.5), the conclusion
is evident, because Mc = VM /Rs Ts .
The ACM transfer functions cannot be easily obtained anymore by
substituting the perturbed duty ratio (d) in the corresponding VMC small-
signal state space by means of (5.5) because of the current-loop-amplier
transfer function but the block-diagram technique introduced in conjunction
with the PCM modeling has to be applied.
Figure 5.4 shows the required block diagrams, where the coefcients A and B
for the basic converters are as follows: buck A = 1, B = 0, boost and buckboost
A = D , B = IL , and the output-voltage-feedback gain (qo ) and the input-voltage-
feedforward gain (qi ) as dened above based on the corresponding PCM gains.
From Figure 5.4a, we may compute the transfer functions dening the
internal dynamics of the ACM-controlled converter as a function of the VMC
transfer functions and the duty-ratio-constraints coefcients to be as follows:
 
BFm Gca
1+ Gvioo Fm qi Gvco
A
Gioo =
1 + Lc (s) + Lv (s)
 
BFm Gca Fm Gca v
1+ v
Zoo + Gco
A A
Zoo =
1 + Lc (s) + Lv (s)
(1 + Gca ) Fm Gvco
Gco =
1 + Lc (s) + Lv (s)
  
Gca
Fm qo + Gvioo + qi Gvci
AZC
Yino = Yino
v

1 + Lc (s) + Lv (s)
174 5 Average and Small-Signal Modeling of Average-Current-Mode Control

Gca v Gca v
Fm ((qo + )Z )Gci
AZC oo A
Toio = Toio
v
+
1 + Lc (s) + Lv (s)
(1 + Gca )Fm Gvci
Gci = (5.6)
1 + Lc (s) + Lv (s)
where the inductor-current-loop gain (Lc (s), and the output-voltage-loop gain
(Lv (s)) are as follows:

Lc (s) = Gca Fm GvcL

(5.7)
Lv (s) = Fm qo Gvco

Figure 5.4 Block diagrams for

ACM control in CCM: (a) output
dynamics and (b) input
dynamics.
 5.4 Dynamic Review 175

According to (5.6), we may conclude the following:

1. The control-to-output transfer function (Gco ) incorporates the same

RHP zero as in the corresponding VMC converter if any.
2. The buck converter may have high input-noise attenuation similar to
the corresponding PCM converter depending on Kf .
3. If Kf is small, then the transfer functions will have features resembling
the VM control.
4. If Kf is high, then the transfer functions will have features resembling
the PCM control.

5.4
Dynamic Review

The dynamic effects imposed by the ACM control are analyzed by using the
buck converter as shown in Figure 5.5. The peak-to-peak voltage swing (VM )
of the PWM modulator is set to 3 V, and consequently, Mc = 40 A/10 s
(i.e., VM /Rs Ts ). The proposed dynamical models (5.6) are veried by means of
simulated frequency responses. The dynamics of the ACM-controlled converter
is compared to the corresponding VM- and PCM-controlled converters. The
internal transfer functions cannot be obtained directly by simulation because
of the current-output nature of the converter at open loop, but they have to be
computed similarly to the PCM transfer functions from the measured data.
The current-loop amplier is designed to have a zero (i.e., 1/Rf Cf ,
Rf = 10 k, Cf = 47 nF) at half the resonant frequency (i.e., 339 Hz). Its high-
frequency gain (Rf /Rin ) is varied from 1 to 10 by changing Rin accordingly as
shown in Figure 5.6. The high-frequency pole is placed to innity (i.e., Cp = 0).
The corresponding inductor-current-loop gains (i.e., Lc (s)(5.7)) are shown
in Figure 5.7 at the input voltages of 20 and 50 V. It may be obvious that Kf

Figure 5.5 ACM-controlled

buck converter.
176 5 Average and Small-Signal Modeling of Average-Current-Mode Control

Figure 5.6 Current-loop-amplier frequency responses.

Figure 5.7 The inductor-current-loop gains at the input voltages of 20 and 50 V with
varying current-amplier high-frequency gain (Kf ).

should be higher than unity for a proper low-frequency gain but not much
higher than 10 to ensure proper crossover frequency. It is dened in [1] that
the maximum gain should not exceed the value given by Kf = Mc /M2 , where
M2 is the down slope of the inductor current yielding 42. The loop phase
margin is always at least 90 .
 5.4 Dynamic Review 177

5.4.1
Control-to-Output Transfer Function

The control-to-output transfer function (Gco ) can be presented symbolically by

(1 + Gca )Fm Gvco
(5.8)
1 + Gca Fm GvcL

because the output-voltage feedback gain (qo ) is zero for a buck converter, where
Gvco and GvcL are the VMC control-to-output and control-to-inductor current
transfer functions, respectively. In this case, the control-to-output transfer
function will be given in voltage domain by Gco /Rs , where Rs = 75 m as
shown in Figure 5.5, because Mc is dependent of Rs . The predicted (solid and
dashed lines) and simulated (dots and squares) internal transfer functions are
shown in Figure 5.8 at the input voltages of 20 and 50 V. It is obvious that the
predictions and simulated responses have a good match.
Figure 5.9 shows the load-affected (dash-dot lines) and internal (solid and
dashed lines) transfer functions at the input voltage of 20 V. The differences
are observable at the frequencies lower than the resonant frequency. If making
conclusions based on the load-affected transfer functions (i.e., ZL = 4 ), the
low-frequency phase of 90 implying tendency to low-frequency instability
in the case of traditional PI-type controllers would not be noticed.
Figure 5.10 shows the control-to-output transfer function of the ACM, VM,
and PCM-controlled converters in voltage domain at the input voltage of 20 V.
It may be obvious that the high value of Kf induces PCM-like features and the
low value of Kf induces VM-like features.

5.4.2
Output Impedance

The output impedance (Zoo ) can be given symbolically by

v
Zoo + Fm Gca Gvco
(5.9)
1 + Fm Gca GvcL

The predicted and simulated (dots, squares, and triangles) frequency responses
at the input voltages of 20 V (solid lines) and 50 V (dashed lines) are shown
in Figure 5.11. The simulated responses match perfectly the corresponding
predicted responses. The ACM control clearly increases the low-frequency
impedance similarly to the PCM control.
The comparison between the corresponding VM, PCM, and ACM (solid
lines) output impedances at the output voltage of 20 V are shown in Figure 5.12
verifying the PCM features of the ACM output impedance.
The load-affected (ZL = 4 ) (dashed and dash-dot lines) and the internal
(solid lines) output impedances are shown in Figure 5.13. The differences are
178 5 Average and Small-Signal Modeling of Average-Current-Mode Control

Figure 5.8 The frequency responses of the internal control-to-output transfer functions at
the input voltages of 20 and 50 V: (a) Kf = 1 and (b) Kf = 10.

observable up to the frequencies lower than the resonant frequency of the

output lter. The differences between the load-affected and the internal output
impedances are signicant and could lead to totally erroneous analytical
conclusions on the load interactions if based on the load-affected transfer
functions.
 5.4 Dynamic Review 179

Figure 5.9 Load-affected (dash-dot lines) and internal (solid and dashed lines) transfer
functions at the input voltage of 20 V.

Figure 5.10 Comparison of VM, PCM, and ACM (solid lines) transfer functions at the
input voltage of 20 V.

5.4.3
Input-to-Output Transfer Function

The input-to-output transfer function (Gioo ) can be given symbolically by

Gvioo Fm qi Gvco
(5.10)
1 + Fm Gca GvcL
180 5 Average and Small-Signal Modeling of Average-Current-Mode Control

Figure 5.11 The frequency responses of the internal output impedances at the input
voltages of 20 V (solid lines) and 50 V (dashed lines), when Kf varies from 1 to 10.

Figure 5.12 The comparison of VM, PCM, and ACM (solid lines) output impedances at
the input voltage of 20 V, when Kf is varying from 1 to 10.

The predicted and simulated (dots, squares, and triangles) internal input-
to-output transfer functions at the input voltages of 20 V (solid lines) and
50 V (dashed lines) are shown in Figure 5.14. The predicted and simulated
responses have very good match validating the accuracy of the proposed
modeling method. It may be obvious that the increase in the ripple feedback
also increases the input-noise attenuation.
 5.4 Dynamic Review 181

Figure 5.13 Load-affected (dashed lines) and internal (solid lines) output impedances at
the output voltage of 20 V, when Kf varies from 1 to 10.

Figure 5.14 The predicted and simulated (dots and squares)

frequency responses of the internal input-to-output transfer
function at the input voltage of 20 V (solid lines) and 50 V (dashed
lines), when Kf varies from 1 to 10.

The comparison of the corresponding VM, PCM, and ACM (solid

lines) transfer functions at the input voltage of 20 V is shown in
Figure 5.15, which clearly indicates how the low value of Kf makes the
converter resemble VM converter and the higher value of Kf makes
the converter resemble PCM converter. It may be obvious that the
PCM control has the highest high-frequency attenuation, which would
182 5 Average and Small-Signal Modeling of Average-Current-Mode Control

Figure 5.15 The comparison of the VM, PCM, and ACM (solid lines) input-to-output
transfer functions at the input voltage of 20 V, when Kf varies from 1 to 10.

make it more insensitive to the source interactions than the other

control modes.
The load-affected (ZL = 4 ) (dashed and dash-dot lines) and the internal
(solid lines) input-to-output transfer functions are shown in Figure 5.16. The
differences are observable up to the resonant frequency of the output lter. The
low-frequency input-noise attenuation of the load-affected transfer functions
is much higher than the real internal transfer functions due to the effect of

Figure 5.16 Load-affected (dashed lines) and internal (solid lines) input-to-output
transfer functions at the output voltage of 20 V, when Kf varies from 1 to 10.
 5.5 Effect of Current-Loop High-Frequency Pole 183

RL /(RL + Zoo ), which may lead to erroneous conclusions on the possible

source and reected-load interactions.

5.4.4
Input Admittance

The input admittance (Yin ) can be symbolically given by

Gca
Fm ( Gvioo + qi )Gvci
ZC
Yino
v
(5.11)
1 + Fm Gca GvcL
The predicted internal input admittance (solid line Kf = 1, dashed line Kf =
10) and the corresponding VM and PCM input admittances (dash-dot lines) are
shown in Figure 5.17. Depending on Kf , the ACM input admittance may have
feature resembling VM converter or PCM converter. The source interactions
may, however, be worse than that in the corresponding VM or PCM converter.

5.5
Effect of Current-Loop High-Frequency Pole

The ripple component of the inductor current is a triangle-shaped signal,

which can be presented by means of its Fourier components by


iL-pp
iL-ripple = (2 sin2 nD cos no t
n=1
2 2 n2 DD
+ sin n2D sin no t) (5.12)

Figure 5.17 The comparison of the VM, PCM, and ACM (solid and dashed lines) input
admittances at the input voltage of 20 V, when Kf varies from 1 to 10.
184 5 Average and Small-Signal Modeling of Average-Current-Mode Control

Figure 5.18 The simulated duty-ratio generation process, where

the current-amplier high-frequency pole (fp ) is placed at innity,
switching frequency, and 1/4 of the switching frequency.
(a) Uin = 20 V and (b) Uin = 50 V.

where
iL-pp is the peak-to-peak value of the ripple current. The inverse
relation to the square of n indicates that the rst harmonic components
have a very strong effect and, therefore, the high-frequency pole (fp )
would shape easily the ripple component and, consequently, affect the
dynamics of the ACM-controlled converter more than would be expected.
We do not give analytical models capable to perfectly predict the converter
dynamics. The boundaries for the ripple effects are naturally the full-ripple
 5.5 Effect of Current-Loop High-Frequency Pole 185

Figure 5.19 Effect of current-amplier high-frequency pole on (a) amplier transfer

function and (b) inductor-current loop gain.

effects [13] introduced in this chapter and zero-ripple effects introduced in [10]
setting Kf = 0.
The simulated duty-ratio generation process (5.4) at the input voltages of
20 and 50 V are shown in Figure 5.18, where Cp = 0 corresponds to fp = ,
Cp = 160 pF to fp = fs , and Cp = 646 pF to fp = 0.25 fs . The dashed horizontal
lines correspond to the averaged current-amplier output signal (
uca ). The
control voltage (uco ) has been kept constant. The PWM ramp signal (Rs Mc )
186 5 Average and Small-Signal Modeling of Average-Current-Mode Control

Figure 5.20 The effect of high-frequency pole (fp ) in the

inductor-current loop on (a) control-to-output transfer function
and (b) input-to-output transfer function, where the dash-dot lines
correspond to fp = , dashed lines to fp = fs , and solid lines to
fp = 0.25 fs .

has a voltage swing of 3 V. The strong attenuating effect of the high-frequency

pole is obvious as discussed above.
The corresponding current-amplier transfer function (Gca ) and the
inductor-current loop gain (Lc (s)) are shown in Figure 5.19, when Kf = 10.
It may be obvious that the crossover frequency and phase margin of the
inductor-current loop would reduce along the decrease in the high-frequency
 References 187

pole (fp ), which are 15.8 kHz and 57.3 (Figure 5.19b) for fp = 0.25 fs ,
respectively.
The effect of the location of the high-frequency pole on the control-to-output
transfer function is shown in Figure 5.20a and on the input-to-output transfer
function in Figure 5.20b. The simulated frequency responses are marked
with dots (Cp = 0), squares (Cp = 160 pF), and triangles (Cp = 646 pF). The
predicted responses marked with dash-dot line (Cp = 0) are computed by using
Kf = 10. The other predicted responses (Cp = 160 pF and Cp = 646 pF) are
computed by using Kf = 0. According to the simulated responses, the high-
frequency pole affects mainly the high-frequency behavior of the corresponding
transfer functions. The observed peaking at 15.8 kHz is the consequence of
the phase margin (i.e., <60 ) in the inductor-current loop as discussed earlier.
It may be obvious that the use of Kf = 0, that is, omitting the ripple effect,
would produce quite accurate predictions.

References

1. L.H. Dixon, Average current mode current-mode control, IEEE Trans.

control of switching power supplies, Power Electron., vol. 8, no. 2, 1993, pp.
in Proc. Power Supply Design Seminar 112119.
SEM 700, Unitrode Corporation, 1990, 8. Y.-S. Jung and M.-J. Youn, Sampling
pp. 5.15.14. effect in continuous-time small-signal
2. C.W. Deisch, Simple switching modeling of average-current mode
control method changes power control, IEE Proc. Electr. Power Appl.,
converter into a current source, in vol. 149, no. 4, 2002, pp. 311316.
Proc. IEEE Power Electronics Specialists 9. R.B. Ridley, A new continuous-time
Conf., 1978, pp. 300306. model for current-mode control, IEEE
3. R. Redl and N.O. Sokal, What a Trans. Power Electron., vol. 6, no. 2,
design engineer should know about 1991, pp. 271280.
current mode control, in Proc. IEEE 10. J. Sun and R.M. Bass, Modeling and
Power Electronics Design Conference, practical design issues for average
1985, pp. 1833. current control, in Proc. IEEE Applied
4. R. Redl, Power-factor correction in Power Electronics Specialists Conf.,
single-phase switching-mode power 1999, pp. 980986.
supplies An overview, Int. J. 11. P. Cooke, Modeling average current
Electron., vol. 77, no. 5, 1994, pp. mode control, inProc. IEEE Applied
555582. Power Electronics Specialists Conf.,
5. F. Tonicello, The control problem of 2000, pp. 256262.
maximum point power tracking in 12. R.W. Erickson and D. Maksimovic,
power systems, in Proc. Seventh Fundamentals of Power Electronics,
European Space Power Conference, Kluwer, Norwell, MA, USA, 2001, 2nd
2005, 7 pp. Edition.
6. J.H. Lee, H.S. Bae, S.H. Park, 13. T. Suntio, J. Lempinen, I. Gadoura,
and B.H. Cho, Constant resistance and K. Zenger, Dynamic effects of
control of solar array regulator using inductor current ripple in average
average current mode control, in Proc. current mode control, in Proc. IEEE
IEEE Applied Power Electronics Conf., Power Electronics Specialists Conf.,
2006, pp. 15441548. 2001, pp. 12591264.
7. W. Tang, F.C. Lee, and R.B. Ridley, 14. T. Suntio, M. Rahkala, I. Gadoura,
Small-signal modeling of average and K. Zenger, Dynamic effects of
188 5 Average and Small-Signal Modeling of Average-Current-Mode Control

inductor current ripple in peak-current peak-current-mode control in

and average-current mode control, in continuous and discontinuous
Proc. IEEE Industrial Electronics Society modes, in Proc. IEEE Industrial
Annual Conf., 2001, pp. 10721077. Electronics Society Annual Conf., 2002,
15. T. Suntio, Unied derivation and pp. 13981403.
analysis of duty-ratio constraints for
 189

6
Average and Small-Signal Modeling of Self-Oscillation Control

6.1
Introduction

Self-oscillation control is typically used in the applications requiring low-EMI

or low-cost solutions such as power-factor correction [1] as well as mobile
phone charging and notebook battery charging [24]. Boost and yback
converters [1, 4] are the most common converter topologies applied under
the self-oscillation control. The self-oscillation control is a form of peak-
current-mode control, where the cycle is initiated, when the inductor current
reaches the zero level and terminated, when the inductor current reaches the
control current. A determined delay is usually applied before switching on
the next cycle to prevent the switching frequency to increasing indenitely.
The resulting switching frequency is naturally variable. The operation mode is
commonly known as boundary conduction mode (BCM), critical conduction or
transition mode. A simple self-oscillating yback converter used in the mobile
phone applications is shown in Figure 6.1 [3] illustrating the implementation
of the control principle. A detailed operational analysis of the self-oscillating
yback converter can be found from [4].
It has been usually assumed that the dynamics associated with the self-
oscillation control is similar to the xed-frequency VM control [5, 6]. The
PCM control and variable switching frequency would, however, change the
dynamics requiring special modeling methods to capture it [710]. The method
presented in this chapter is based on [811]. We present the dynamic review
only for buck and buckboost or yback converters.

6.2
Self-Oscillation Modeling

The self-oscillation control is a derivative of PCM control in which the resulting

inductor-current waveforms are as shown in Figure. 6.2, when a determined
switching delay (TD ) is applied. Similarly to PCM control, we have to nd the
proper duty-ratio constraints under variable switching frequency. However, the

Dynamic Prole of Switched-Mode Converter. Teuvo Suntio

2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
ISBN: 978-3-527-40708-8
190 6 Average and Small-Signal Modeling of Self-Oscillation Control

Figure 6.1 A simple self-

oscillating yback converter.

Figure 6.2 Inductor-current

waveforms under self-oscillation
control.

small-signal direct-on-time-controlled state space has to be solved before we can

proceed with the PCM modeling. The direct-on-time control is not a physical
control mode but a ctive mode serving only the purposes of modeling [11].

6.2.1
Averaged Direct-on-Time Model

According to Chapter 3, the derivative of the average inductor current
iL  can
be given by
d
iL  ton toff1
= m1 m2 (6.1)
dt ts ts
From the inductor-current waveforms in Figure 6.2, we may solve toff1 by
means of
iL  yielding
2
iL 
toff1 =
 (6.2)
TD
m2 1
ts
Substituting toff1 in (6.1) with (6.2) yields
d
iL  ton 2
= m1
iL  (6.3)
dt ts ts TD
The derivative for the average capacitor voltage (
uC ) can be generally given by
d
uC  q1
iL 
io 
= (6.4)
dt C C
where the coefcient q1 is as follows: buck q1 = 1, boost and buckboost
q1 = 1 ton /(ts TD ).
 6.2 Self-Oscillation Modeling 191

Figure 6.3 The steady-state

equivalent circuit for the
self-oscillating converters.

The average output voltage (
uo ) can be given by

d
uC 

uo  =
uC  + rC C (6.5)
dt
and the average input current (
iin ) by

iin  = q2
iL  (6.6)

where the coefcient q2 is as follows: buck and buckboost q2 = ton /(ts TD ),

boost q2 = 1.
As a summary, the general direct-on-time averaged state space can be given
by
d
iL  ton 2
= m1
iL 
dt ts ts TD
d
uC  q1
iL 
io 
= (6.7)
dt C C

iin  = q2
iL 
d
uC 

uo  =
uC  + rC C
dt
where q1 and q2 are as dened above.
The steady-sate operating point of the converter can be solved from (6.7)
setting the derivatives to zero and substituting the inductor current up slope
(m1 ) with its topology-based values given in Chapter 3 (Section 3.3), which can
be given in the form of an equivalent circuit shown in Figure 6.3.
The coefcients of the equivalent circuit and the other operating-point
parameters can be dened for the basic converters as follows:

Buck:


TD D
M D, =
Ts TD
1
Ts
TD
D
Ts
UE = UD or 0
TD
1
Ts
192 6 Average and Small-Signal Modeling of Self-Oscillation Control

TD
D
D Ts
rE = rL + r + (rd or rds2 )
TD ds1 TD
1 1
Ts Ts
IL = Io
D
Iin = I
TD o
1
Ts

Boost:

TD

 1
TD Ts
M D, =
Ts 
TD
D
Ts
UE = UD or 0


TD 2 TD
1 D(1 )
Ts Ts
rE =
2 rL +
2 rds1
TD TD
D D
Ts Ts
TD
1
Ts D
+ (rd or rds2 ) + r
T D TD C
D D
Ts Ts
TD
1
Ts
IL = I

TD o
D
Ts
TD
1
Ts
Iin = I
TD o
D
Ts

BuckBoost:


TD D
M D, =
Ts TD
D
Ts
UE = UD or 0
 6.2 Self-Oscillation Modeling 193


TD 2 TD TD
1 D(1 ) 1
Ts Ts Ts
rE =
2 rL +
2 rds1 +
TD TD TD
D D D
Ts Ts Ts
D
(rd or rds2 ) + rC
TD
D
Ts
TD
1
Ts
IL = I

TD o
D
Ts
D
Iin = I
TD o
D
Ts

The steady-state cycle time (Ts ) can be solved from the inductor current
waveforms in Figure 6.2 yielding


2IL (M1 + M2 )
Ts2 2TD + Ts + TD2 = 0 (6.8)
M1 M2
and substituting the topology-based up and down slopes with the values
dened in Chapter 3 (Section 3.3).

6.2.2
Small-Signal Direct-on-Time Model

The small-signal state space can be derived from the averaged state space (6.7)
by developing the proper partial derivatives as instructed in Chapter 2. Under
self-oscillation control or generally in the variable-frequency-operation mode,
the cycle time (ts ) is also variable. Therefore, we also have to develop proper
dynamic constraints (i.e., cycle-time constraints) for the varying cycle time in
order to introduce its effect on the dynamics of the converter.
The cycle-time constraints may be obtained from the denition of the cycle
time ts = ton + toff1 + TD by substituting toff1 with (6.2) yielding
2
iL 
ts = ton +
 + TD (6.9)
TD
m2 1
ts
and by developing the proper partial derivatives, and formulating the
results as

ts = Fm
c
(ton + qcc iL + qco uo + qci uin ) (6.10)
194 6 Average and Small-Signal Modeling of Self-Oscillation Control

The nal small-signal state space results when the perturbed cycle time (ts ) is
substituted with (6.10).
It has been observed that the circuit parasitic elements and the delay (TD ) do
not contribute much onto the converter dynamics [10]. Therefore, we may omit
their effect and develop the constraints coefcients in a most convenient form.
The parasitic elements and the delay (TD ) would, however, signicantly affect
the steady-state operating point, and should not be omitted when dening the
steady-state parameters.
The coefcients of the cycle-time constraints (6.10) can be given for the basic
converters as shown in Table 6.1.
We do not give explicitly the small-signal state spaces but would proceed
directly to the required PCM modeling.

6.2.3
Small-Signal PCM Models

In order to obtain the PCM small-signal state space, we have to develop the
on-time constraints relating the perturbed on-time (ton ) to the control current
(ico ), the other circuit variables and elements similarly to the xed-frequency
PCM modeling treated in Chapter 4. From Figure 6.2, the comparator equation
dening the length of the on-time at t = kts + ton can be given by
ico =
iL  +
iL (6.11)
The difference (
iL ) between the peak inductor current and its average value
can be computed to be


ton m1 TD

iL = 1+ (6.12)
2 ts
Therefore, the comparator equation (6.11) becomes


ton m1 TD
ico =
iL  + 1+ (6.13)
2 ts

Table 6.1 Cycle-time-constraints coefcients for buck, boost, and

buckboost converters.

Converter F0m q0c q00 q0i

2L DTs DTs
Buck 1
Um U0 2L 2L
2L DTs
Boost 1 0
Um 2L
2L DTs
Buckboost 1 0
Um 2L
 6.2 Self-Oscillation Modeling 195

Table 6.2 On-time-constraints coefcients for buck, boost, and

buckboost converters.

Converter Fm
c
qcc qco qci

2L D(Um U0 )Ts
Buck 1 0
U0 U02
2L DUm Ts DUm Ts
Boost 1
U0 Um (U0 Um )2 (U0 Um )2
2L DUm Ts
Buckboost 1 0
U0 U02

from which the on-time constraints can be solved by developing the proper
partial derivatives and applying the cycle-time constraints (6.10) yielding
 
ton = Fm
o
ico qoc iL qoo uo qoi uin (6.14)

The coefcients of the on-time constraints (6.14) can be given for the basic
converters as shown in Table 6.2.
The small-signal state spaces and the corresponding sets of the transfer
functions corresponding to the self-oscillation control can be given for the
basic converters as follows:
Buck:
diL 4 2
= iL + ico
dt Ts Ts
duC 1 1
= iL io
dt C C (6.15)
DD Ts D2 D Ts 
iin = D(2D 1) iL + uC uin + DD ico
2L 2L
duC
uo = uC + rC C
dt
D2 D T
DIL

DD Ts


DIL

 
s
= =
Yino Toio 2L Uin s 2LC s Uo C
=
Gioo Zoo 1 + s rC C
0
sC
(6.16)


2 1
DD s2 + s  +
D Ts LC


  4
Gci s s +
= Ts (6.17)
Gco
+
2(1


s rC C)


4
s Ts C s +
Ts
where we have applied the identity Ts = 2LIL Uin /(Uin Uo )Uo .
196 6 Average and Small-Signal Modeling of Self-Oscillation Control

In boundary mode, the average inductor current stays constant at half the
peak inductor current, which is dened by the control current. Therefore, it is
natural that the input-to-output transfer function (Gioo ) of a buck converter
is zero, because IL = Io , and the input admittance (Yino ) equals the ideal
input admittance (Yin DIL /Uin ). The transfer functions are mainly of
rst order except the control-related transfer functions, which are the expected
results due to the tightly bounded inductor current.
Boost:
diL 4 2
= iL + ico
dt Ts Ts
duC D (2D + 1) DD2 Ts DD Ts 1 DD
= iL uC + uin io ico
dt C 2LC 2LC C C
iin = iL
duC
uo = uC + rC C (6.18)
dt

0 0
 
Yino Toio DD Ts (1 + s rC C) 1 + s rC C
=


 (6.19)
Gioo Zoo DD2 Ts DD2 Ts
2LC s + C s+
2LC 2LC
2


Ts s +
4

  Ts
Gci

=
DT (6.20)
(1 + s rC C)
 s
Gco 2D 1 s 2



2
DD Ts 4
Ts C s + s+
2LC Ts

where we have applied the identity Ts = 2LIL Uo /(Uo Uin )Uin . As discussed
earlier, the average inductor current stays constant at Ico /2. Therefore, it is
natural that the input admittance (Yino ) and the output-to-input transfer
function (Toio ) are zero as well as the control-to-input transfer function (Gci )
is of rst order. The control-to-output transfer function (Gco ) contains an
RHP zero at 2/DTs , which is symbolically same as the RHP zero in the
DCM VM control-to-output transfer function. It may be obvious that the
RHP zero does not much affect the magnitude but does affect the phase
behavior (i.e., the phase starts affecting from = 0.2/DTs , which corresponds
to f = (0.064/D) fs ). It may be obvious that the maximum control bandwidth
would be limited, and the stability of the converter would be sensitive to the
variation in the output-capacitor ESR (rC ).
 6.2 Self-Oscillation Modeling 197

BuckBoost:

diL 4 2
= iL + ico
dt Ts Ts
duC D (2D + 1) DD2 Ts D2 D Ts 1 DD
= iL uC + uin io ico
dt C 2LC 2LC C C
DD2 Ts D2 D Ts
iin = D(2D 1) iL + uC uin + DD ico
2L 2L
duC
uo = uC + rC C (6.21)
dt

D2 D Ts s DD2 Ts s



DD2 Ts DD2 Ts
 
2L s + 2LC 2LC s +

Yino Toio 2LC
= (6.22)
Gioo Zoo D2 D Ts (1 + s rC C) 1 + s r C



C

2
DD Ts DD Ts2
2LC s + C s+
2LC 2LC


 2 D
DD s + s  +
2
D Ts LC



DD2 Ts 4

 s+ s+
2LC T
Gci s
=
 (6.23)
Gco DT
2D 1 s s
(1 + s rC C)

2

2 

DD Ts 4
Ts C s + s+
2LC Ts

where we have applied the identity Ts = 2LIL (Uin + Uo )/Uin Uo . The control-
to-output transfer function (Gco ) contains an RHP zero at 2/DTs , which is
symbolically same as the RHP zero in the DCM VM control-to-output transfer
function. It may be obvious that the RHP zero does not much affect the
magnitude but does affect the phase behavior (i.e., the phase starts affecting
from = 0.2/DTs , which corresponds to f = (0.064/D) fs ). The buckboost
converter does not have similar special features as the buck and boost
converters, because both the input and output currents are discontinuous.
It may be obvious that the maximum control bandwidth would be limited
and the stability of the converter would be sensitive to the variation in the
output-capacitor ESR (rC ) as discussed in [10]. The rst-order nature of (6.22)
is natural due to tightly bounded inductor current.
198 6 Average and Small-Signal Modeling of Self-Oscillation Control

6.3
Dynamic Review

The dynamic effects imposed by the self-oscillation control are reviewed by

using the same buck converter as in the other chapters and a yback or
buckboost converter introduced in [8, 10]. The converter can operate at
open loop only at resistive load similarly to the other current-mode-controlled
converters. Therefore, the internal transfer functions have to be computed
from the measured or simulated load-affected responses. The resistive load
would effectively hide the dynamic low-frequency behavior.

6.3.1
Buck Converter

The buck converter shown in Figure 6.4 is subjected to the dynamic analyses.
There are available commercial critical (CM) or transition (TM) mode PWM
modulators as introduced in [1], which can be used to implement the self-
oscillation control.
The required control input (Ico ) in current domain can be approximated by
means of
2Io M1 M2
2
Ico 2Io Ico TD = 0 (6.24)
M1 + M2
where M1 and M2 are the up and down slopes of the inductor current,
respectively. Similarly the cycle time (Ts ) can be solved by means of
M1 M2
Ts = TD + Ico (6.25)
M1 + M2
and the duty ratio (D) by means of
Ico
D= (6.26)
M1 Ts
Equations (6.24)(6.26) are solved from the inductor current waveforms shown
in Figure 6.2.

Figure 6.4 A buck converter

under self-oscillation control.
 6.3 Dynamic Review 199

The buck converter is analyzed at the input voltage of 20 and 50 V by varying

the delay time (TD ) from 0 to 2 s. The resulting switching frequencies and
duty ratios are as follows:

20 V : 98 kHz, 0.53(TD = 0), 72 kHz, 0.454(TD = 2 s)

50 V : 165 kHz, 0.21(TD = 0), 104 kHz, 0.17(TD = 2 s)

6.3.1.1 Control-to-Output Transfer Function

The internal control-to-output transfer function (Gco ) can be given symbolically
by
2 (1 + s rC C)

 (6.27)
4
s Ts C s +
Ts
having a pole at the origin, which implies possibility of low-frequency
instability, if conventional PI or PID controllers are used due to of phase
lag 180 at low frequencies. The frequency responses of the control-to-output
transfer function at the input voltage of 20 V (solid lines) and 50 V (dashed
lines) are shown in Figure 6.5, where the dots and squares denote the simulated
responses at the input voltage of 20 V with TD = 0 and TD = 2 s, respectively.
The dash-dot lines of the phase response denote the effect of the delay (i.e.,
TD = 2 s corresponding to 20% of the cycle time). The simple model predicts
quite well the control-to-output transfer function, but the phase has less phase
lag than the model predicts at high frequency due to the modulation-process
nonlinearities as discussed in [12].
The load-affected (dashed line) and internal (solid line) control-to-output
transfer functions at the input voltage of 20V are shown in Figure 6.6. It
may be obvious that the resistive load effectively hides the real behavior
of the transfer function at the low frequencies and, therefore, the critical
low-frequency phase behavior would not be visible as discussed earlier.
Figure 6.7 shows the comparison of the VM, PCM, and self-oscillation
control-to-output transfer functions at the input voltage of 20 V. It may be
obvious that the self-oscillation control imposes similar features than the PCM
control except the low-frequency phase.

6.3.1.2 Output Impedance

The output impedance (Zoo ) can be symbolically given by
1 + s rC C
(6.28)
sC
which is the impedance of the output capacitor. It may be obvious that the
result is as could be expected, because the average inductor current would stay
200 6 Average and Small-Signal Modeling of Self-Oscillation Control

Figure 6.5 The predicted and simulated (dots and squares)

frequency responses of the control-to-output transfer function at
the input voltage of 20 V (solid lines) and 50 V (dashed lines). The
dots correspond to TD = 0 and the squares to TD = 2 s,
respectively.

Figure 6.6 The resistive-load-affected and internal control-to-output transfer function at

the input voltage of 20 V and TD = 0.

constant at half the control current. This also means that the output impedance
does not change along the operation point as is usual with the other control
modes. The predicted (solid line) and simulated (dots) frequency responses of
the output impedance are shown in Figure 6.8a conrming the validity of the
 6.3 Dynamic Review 201

Figure 6.7 The comparison of the VM (dashed line), PCM (dash-dot line), and
self-oscillation (solid line) control-to-output transfer functions at the input voltage of 20 V.

prediction. The corresponding VM and PCM output impedances compared to

the self-oscillation output impedance at the input voltage of 20 V are shown in
Figure 6.8b revealing that the PCM output impedance also resembles greatly
the impedances of its output capacitor.

6.3.1.3 Input-to-Output Transfer Function

The input-to-output transfer function (Gioo ) is naturally zero, because the
average inductor current and, consequently, the output current would stay
constant regardless the changes in the input voltage.

6.3.1.4 Input Admittances

The open-loop input admittance (Yino ), the ideal input admittance (Yin ),
and the short-circuit input admittance (Yinsc ) are equal because Gioo = 0.
They can be symbolically given by

DIo

 (6.29)
TD
1 Uin
Ts

which equals the general ideal input admittance (Iin /Uin ) of the buck
converter.
202 6 Average and Small-Signal Modeling of Self-Oscillation Control

Figure 6.8 The internal output impedance of

(a) self-oscillation-controlled buck converter (solid lines), and
(b) compared to the output impedances of the corresponding VM
and PCM buck converters (dashed lines). The dots denote the
simulated responses.

6.3.2
Flyback Converter

The yback or buckboost converter shown in Figure 6.9 is subjected to the

dynamic analyses. The experimental frequency responses are shown only for
 6.3 Dynamic Review 203

Figure 6.9 A yback converter.

the output-voltage loop gain due to the difculties to measure the open-loop
responses reliably.
The required control current (Ico ) can be approximated by means of
2Io (M1 + M2 ) 2Io M2
2
Ico Ico TD = 0 (6.30)
M1 L
and the cycle time (Ts ) and the duty ratio (D) according to (6.25) and (6.26),
respectively.
The open-loop transfer functions are analyzed assuming that the input-side
parameters of the yback converter (Figure 6.9) are transformed into the
secondary side: Uin = 27.5 V and L = 64 H. The measured voltage loop gain
(L(s)) is shown in Figure 6.10a by varying the delay time from 0 to 2 s [10],
and the corresponding predictions in Figure 6.10b. The switching frequency
was approximately 64 kHz and the load of a 12  resistor. It may be obvious
that the delay does not much affect the dynamics, when its value is less than
12% of the cycle time. The extra low-frequency ripple in the measurements is
due to the rectied line voltage having frequency of 50 Hz.

6.3.2.1 Control-to-Output Transfer Function

The control-to-output transfer function (Gco ) can be symbolically given by


 DTs
2D 1 s (1 + s rC C)
2


 (6.31)
DD2 Ts 4
Ts C s + s+
2LC Ts
where the low-frequency value L/DD Ts corresponds to D Uo /2DIo or Uin /2Io .
This shows that the low-frequency gain would be highest at the high input
voltage and low output current.
The converter is analyzed at the input voltage of 275 V, which corresponds
to 27.5 V at the secondary side. The delay (TD ) is set to zero and 5 s
204 6 Average and Small-Signal Modeling of Self-Oscillation Control

Figure 6.10 The frequency response of the voltage loop gain: (a) experimental
measurement and (b) predictions.

corresponding to the switching frequencies of 63 and 40 kHz as well as the

duty ratios of 0.19 and 0.15, respectively. The corresponding predicted (solid
line) and simulated (dots: TD = 0, squares: TD = 5 s) frequency responses are
shown in Figure 6.11. The delay of 5 s corresponds to about 20% of the cycle
time. When the delay is small compared to the cycle time, the simple transfer
function predicts the dynamics well. The tendency of the increasing delay is
to lower the low-frequency gain and increase the phase lag as is visible in
Figure 6.11 (the squares). The modulator nonlinearities would cause the high-
frequency phase lag to be less than predicted as discussed in [12]. In the case
of rather long delay, it is necessary to use the proposed modeling technique
without simplifying the results in order to get more accurate predictions.
The at high-frequency gain indicates that the control bandwidth cannot be
designed very high in order to maintain sufcient gain margin. It may be also
obvious that the high-frequency gain has very strong dependence on the ESR
 6.3 Dynamic Review 205

Figure 6.11 The predicted (solid line) and simulated (dots: TD = 0, and squares:
TD = 5 s) frequency responses of the control-to-output transfer function.

of the output capacitor, and the increase in it could reduce the gain margin
to be unacceptable. As discussed above, the high input voltage and low output
current would also increase the gain, which may lead to unacceptable low gain
margin or even instability.
The load-affected transfer function compared to the internal transfer function
is shown in Figure 6.12. In this case, the internal transfer function does not
contain any surprising features as was the case with the buck converter.

Figure 6.12 Load-affected (dashed line) and internal control-to-output transfer function.
206 6 Average and Small-Signal Modeling of Self-Oscillation Control

6.3.2.2 Output Impedance

The output impedance (Zoo ) can be symbolically given by
1 + s rC C

 (6.32)
DD2 Ts
C s+
2LC
where the low-frequency value 2L/Ts DD corresponds to Uo /DIo [8]. This shows
that the output impedance would increase, when the input voltage increases.
The predicted (solid and dashed lines) and simulated (dots and squares)
frequency responses of the output impedance are shown in Figure 6.13, where
the solid lines and dots correspond to TD = 0 and the dashed lines and squares
to TD = 5 s. It may be obvious that the proposed model would predict
accurately the response at the low delay values but the longer delays would
require the application of the proposed method without simplications for
obtaining more accurate predictions.

6.3.2.3 Input-to-Output Transfer Function

The input-to-output transfer function (Gioo ) can be symbolically given by
D2 D Ts (1 + s rC C)

 (6.33)
DD2 Ts
2LC s +
2LC
The low-frequency value of the transfer function is D/D , which is the same
as the value of the corresponding xed-frequency converter. It means that

Figure 6.13 The predicted (solid line: TD = 0, dashed line: TD = 5 s) and simulated
(dots: TD = 0, and squares: TD = 5 s) frequency responses of the output impedance.
 6.3 Dynamic Review 207

the converter attenuates the input noise at the low frequencies, when the
duty ratio is less than 0.5. The maximum designed duty ratio of the yback
converter (Figure 6.9) is typically 0.5 in order to keep the component stresses
at a reasonable level.
The predicted (solid and dashed lines) and simulated (dots: TD = 0, squares:
TD = 5 s) frequency responses of the input-to-output transfer function are
shown in Figure 6.14. The high-frequency phase responses are missing due
to the problems to extract them, when the switching-frequency ripple is much
higher than the response signal. The proposed model accurately predicts the
response, when the delay is small. The tendency of the delay is to increase the
attenuation and decrease the phase at low frequencies.
The resistive-load-affected transfer function (dashed line) compared to the
internal transfer function (solid line) is shown in Figure 6.15. It may be obvious
that the load-affected transfer function would give misleading information on
the attenuation potential of the converter. The extra attenuation is from the
attenuating effect of the output impedance and the load resistor.

6.3.2.4 Input Admittance

The internal input admittance (Yino ) can be symbolically given by

D2 D Ts s

 (6.34)
DD2 Ts
2L s +
2LC

Figure 6.14 The predicted (solid line: TD = 0, dashed line:

TD = 5 s) and simulated (dots: TD = 0, and squares: TD = 5 s)
frequency responses of the input-to-output transfer function.
208 6 Average and Small-Signal Modeling of Self-Oscillation Control

Figure 6.15 Load-affected (dashed line) and internal input-to-output transfer function.

The corresponding frequency responses are shown in Figure 6.16, where

the solid and dashed lines correspond to the predicted responses (Yino )
(i.e., TD = 0, and TD = 5 s), the dash-dotted lines to the predicted resistive-
R
load-affected transfer functions (Yino ), and the dots (TD = 0) and squares
(TD = 5 s) the simulated frequency responses, respectively. The load affects
the responses only at low frequencies. The simulated responses are the load-
affected responses. It may be obvious that the effect of the delay is insignicant.
The phase-behavior implies that the yback converter can be unstable even
at open loop when the input lter is connected, if the lter is not properly
designed (i.e., no impedance overlap is allowed).

6.3.2.5 Ideal and Short-Circuit Admittances

The ideal input admittance (Yin ) can be symbolically given by
D2 Ts 1 DIL 1
(6.35)
2L DTs Uin DTs
1s 1s
2 2
and the short-circuit admittance (Yinsc ) by
D2 D Ts DD IL
(6.36)
2L Uin
The corresponding frequency responses are shown in Figure 6.17. The RHP
zero of the control-to-output transfer function (Gco ) forms the RHP pole in
the ideal input admittance (6.35). The high-frequency phase behavior of the
ideal input admittance implies slightly increased sensitivity to the source
interactions.
 6.3 Dynamic Review 209

Figure 6.16 Predicted load-affected (dash-dotted lines) and

internal (solid lies: TD = 0, dashed lines: TD = 5 s) input
admittances. The dots (TD = 0) and squares (TD = 5 s) denote
the simulated responses.

Figure 6.17 The predicted frequency responses of the ideal

(Yin ) and short-circuit (Yinsc ) input admittance, where the solid
lines correspond to TD = 0, and the dashed lines to TD = 5 s.
210 6 Average and Small-Signal Modeling of Self-Oscillation Control

References

1. R. Redl, Power-factor correction in modeling of boundary conduction

single-phase switching-mode power mode DC-to-DC converters, in Proc.
supplies An overview, Int. J. IEEE Industrial Electronics Society
Electron., vol. 77, no. 5, 1994, pp. Annual Conf., 2001, pp. 844849.
555582. 8. J. Lempinen and T. Suntio,
2. Y. Panov and M. Jovanovic, Small-signal modeling for design of
Performance evaluation of 70-W robust variable-frequency yback
two-stage adapters for notebook battery chargers for portable device
computers, in Proc. IEEE Applied applications, in Proc. IEEE Applied
Power Electronics Conf., 1999, pp. Power Electronics Conf., 2001, pp.
10591065. 548554.
3. Use of rectied image voltage for 9. T. Suntio, J. Lempinen, K. Hynynen,
controlling the switch on the primary and P. Silventoinen, Analysis and
side of a switched-mode power supply, small-signal modeling of
U.S. Patent 6608768030B2, August self-oscillating converters with applied
19, 2003. delay, in Proc. IEEE Applied Power
4. B. Irving and M. Jovanovic, Analysis Electronics Conf., 2002, pp. 395401.
and design of self-oscillating yback 10. T. Suntio, Average and small-signal
converter, in Proc. IEEE Applied Power modeling of self-oscillating yback
Electronics Conf., 2002, pp. 897903. converter with applied switching
5. G. Spiazzi, D. Tagliavia, and delay, IEEE Trans. Power Electron., vol.
S. Spampinato, DCDC yback 21, no. 2, 2006, pp. 479485.
converters in the critical conduction 11. T. Suntio, Unied average and
mode: A re-examination, in Proc. small-signal modeling of
IEEE Industry Applications Society direct-on-time control, IEEE Trans.
Annual Conf., 2000, pp. 24262432. Indust. Electron., vol. 53, no. 1, 2006,
6. B. Irving, Y. Panov, and M. Jovanovic, pp. 287295.
Small-signal model of variable- 12. J. Sun, Small-signal modeling of
frequency yback converter, in Proc. variable-frequency pulsewidth
IEEE Applied Power Electronics Conf., modulators, IEEE Trans. Aerosp.
2003, pp. 977982. Electron. Syst., vol. 38, no. 3, 2002, pp.
7. J. Chen, R. Erickson, and 11041108.
D. Maksimovic, Averaged switch
 211

7
Dynamic Modeling and Analysis of Current-Output
Converters

7.1
Introduction

Majority of the switched-mode converters operate at voltage-output mode

providing constant output voltage. In the applications, where the storage
batteries are used either to ensure uninterrupted operation of the end load as
in the telecom DC UPS (uninterruptible power supply, Figure 7.1) systems [1]
or as a standalone power source in mobile phones [2], the switched-mode
converters recharging the storage batteries have to be provided with over-
current or overload limiting for protecting them from damage. The over-current
limiting changes the operation of the converter into the current-output mode.
The overload limiting is typically implemented using either constant-current
or modied-constant-power limiting schemes as illustrated in Figure 7.2 [35].
The constant-current limiting can be accomplished either by using only the
output current signal (i.e., single loop) or both the output voltage and the output
current in cascade where the inner loop is the voltage loop and the outer loop the
current loop. The modied constant-power limiting is typically implemented
in such a way that the reference of the current loop is gradually increased
along the decrease in the output voltage until the maximum dened output
current is reached. After that the limiting follows the constant-current scheme
as described in detail in [4].
The change of mode from the voltage output to the current output would
also drastically change the dynamic characteristics of the converter from
those observed in the voltage-output mode. It has been observed [6] that the
crossover frequency of the current loop may increase even a decade or two
compared to the measured frequency response at a resistive load, when the
storage battery is connected as a load due to low internal impedance of the
battery [7]. Several attempts have been made earlier to model the dynamics of
the current-output converter [810], but the real dynamics were not found. The
consistent methods to model the dynamics of the current-output converters
are presented in [1113] based on the models describing the dynamics of the
voltage-output converter. The modeling technique applied in this chapter is
extracted from [11].

Dynamic Prole of Switched-Mode Converter. Teuvo Suntio

2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
ISBN: 978-3-527-40708-8
212 7 Dynamic Modeling and Analysis of Current-Output Converters

Figure 7.1 DC UPS system.

Figure 7.2 Typical output-

voltage/current characteristics
of a converter used in the
battery-charging applications.

7.2
Dynamic Models for Current-Output Converter

The general load of a voltage-output converter consists of a parallel connection

of an impedance (ZL ) and a constant-current sink (jo ) as depicted in Figure 7.3a.
The current-output converter cannot operate properly with such a load.
Therefore, its general load would consist of a series connection of an impedance
(ZL ) and a constant-voltage source (eo ) as shown in Figure 7.3b. The internal
dynamics of the current-output converter can be found assuming the load
to be a pure constant-voltage source, which would be denoted by uo . It may
be obvious that the ideal load would effectively short-circuit the converter
and, therefore, the converter cannot operate without damage at open loop.
As a consequence, the internal dynamics is not fully directly measurable:
near to ideal dynamic constant-voltage load can be implemented connecting
a proper resistor and large capacitor with a low ESR in parallel, which would
yield accurate enough frequency responses for the practical purposes. The
dynamical models of the current-output converter can also be derived directly
from the power stage changing the load as described above and by applying the
modeling methods described in Chapters 36. The most convenient method
would be, however, to derive the models based on the well-known transfer
functions of the corresponding voltage-output converter.
 7.2 Dynamic Models for Current-Output Converter 213

Figure 7.3 General load and supply

conditions of (a) voltage-output
converter and (b) current-output
converter.

Figure 7.4 Experimental

VM/PCM-controlled buck
converter at current-output
mode.

The application of the modied state-space-averaging method is demon-

strated in Section 7.2.1 by using a CCM buck converter (Figure 7.4). The
general dynamic models are derived in Section 7.2.2 based on the voltage-
output converter.

7.2.1
Modied-State-Space-Averaging Technique

We demonstrate the application of the modied-state-space-averaging tech-

nique [14, 15] on the VMC buck converter as shown in Figure 7.4. The
open-loop conditions we assume, when developing the models, are natu-
rally ctive due to the reasons discussed earlier. We also assume that the
output-current-sensing resistor (Rs1 ) is zero (note that the physical output-
current-sensing resistor actually acts as a load and, therefore, it would be
treated as a load resistor).
The variables of the current-output converter are basically same as the
variables of the voltage-output converter except the output voltage (uo ) and the
output current (io ), which would change their roles from the output variable to
the input variable (uo ) and from the input variable to the output variable (io ).
As a consequence, the small-signal state space can be given by (see Chapter 3,
Section 3.3)
214 7 Dynamic Modeling and Analysis of Current-Output Converters

diL rE D 1 UE
= iL + uin uin + d
dt L L L L
duc 1 1
= uc + uo
dt rc C rc C
(7.1)
iin = DiL + IL d
1 1
io = iL + uc uo
rc rc

where rE and UE are given by

rE = rL + Drds + D
rd
(7.2)
UE = Uin + UD + (rd rds )Io

The set of transfer functions dening the dynamics of the current-output

converter can be obtained from (7.1) by rst applying Laplace transform and
secondly matrix algebra yielding



i i  uin
iin Yino Toio Gici
= uo
io Giioo Yoo
i
Gico
d

i i 
Yino Toio Gici
Gioo Yoo Gico
i i

 2
(7.3)
D (1 + src C)  D(1 + src C) DUE (1 + src C)
rE + rc 1
D(1 + src C) LC s2 + s + UE (1 + src C)
L LC
=
(rE + sL)(1 + src C)


0 0 IL
+
0 0 0

According to (7.3), all the transfer functions except the output admittance
i
(Yoo ) are of the rst order, which is as could be expected due to the dynamic
short-circuit at the output of the converter. We also give for comparison
the set of transfer functions dening the dynamics of the corresponding
voltage-output converter as follows:


 v
v  uin
iin Yino Toio Gvci
= io
uo Giioo Zoov
Gvco
d

v v 
Yino Toio Gvci
Gvioo Zoo
v
Gvco
 7.2 Dynamic Models for Current-Output Converter 215
D2 s D(1 + src C) DUE s
L LC L
D(1 + src C) (rE + sL)(1 + src C) UE (1 + src C)

= LC LC LC (7.4)
rE + rc 1
s2 + s +

L LC

0 0 IL
+
0 0 0
i
According to (7.2) and (7.4), the output admittance (Yoo ) is exactly the inverse
of the output impedance (Zoo ) of the corresponding voltage-output converter.
This means that the output admittance is the only directly measurable transfer
function in a current-output converter.
The current-output mode has clearly and signicantly changed the internal
dynamics of the converter by removing its resonant nature from all the other
transfer functions except from the output admittance. It may also be obvious
that the gain of the internal current loop (Lc (s)) may be much higher than the
corresponding gain of the voltage loop (Lv (s)) due to the changes in the control-
to-output transfer function. We will later show in Section 7.3 that the load
would easily recover the voltage-output nature of the transfer functions, which
also explains the observed unexpected changes in the crossover frequency of
the current loop observed in [6] as discussed earlier in Section 7.1. In short,
the close-to-ideal voltage-type load such as storage batteries [4, 7] or a large
capacitor would recover the internal dynamics of the converter into effect with
the observed consequences.
The modied state-space-averaging technique can be equally applied to all
the converters, but we have only demonstrated the use of the technique with
the VMC buck converter. The more general and may be more convenient
treatment of the modeling would be given in the next section.

7.2.2
General Dynamic Models

The general dynamic models of the current-output converter can be developed

by applying the two-port model of the voltage-output converter (Figure 7.5a)
and duality, which yield the two-port model as shown in Figure 7.5b.
The duality means that the Thevenin output port of the voltage-output
converter would be substituted by means of the corresponding Norton
equivalent circuit, and io in the input port by
uT uo
io = (7.5)
ZT
When the described procedures are followed, the internal dynamic models of
the current-output convert can be given based on the transfer functions of the
corresponding voltage-output converter as follows:
216 7 Dynamic Modeling and Analysis of Current-Output Converters

Figure 7.5 Two-port linear models

of (a) voltage-output converter and
(b) current-output converter.

Gvioo Toio
v v v
Gvco Toio
Toio

 v
Yino + Gvci + uin
iin v
Zoo v
Zoo v
Zoo
=
v v
uo
(7.6)
io Gioo 1 Gco
v v v c
Zoo Zoo Zoo
According to (7.6), it is clear that the output admittance of the current-output
converter is the inverse of the output impedance of the corresponding voltage-
output converter as we observed already in the previous section. The order
of the internal control-to-output (Gico ), input-to-output (Giioo ), and output-to-
i
input (Toio ) transfer functions are surely reduced compared to the order of the
i
corresponding voltage-output converter, but the input admittance (Yino ) and
i
the control-to-input (Gci ) may or may not have the same order and the features
as the corresponding voltage-output converter. The internal input admittance
i
(Yino ) is the same as the short-circuit input admittance (Yinsc ) dened for
the voltage-output converter, but the input admittance is not anymore load
independent, which may be concluded according to Figure 7.3b.

7.3
Load and Supply Interactions

The effect of the nonideal load (ZL , eo ) can be found by computing uo at the
presence of the load from Figure 7.6, which yields

ZL (Gvioo uin + Gvco c) + Zoo

v
eo
uo = (7.7)
ZL + Zoov
 7.3 Load and Supply Interactions 217

and substituting it in (7.6) with (7.7). These procedures yield

Gvioo Toio
v v
Toio v
Gvco Toio
v
Yino + Gvci +
ZL + Zoov
ZL + Zoo
v
ZL + Zoov


 uin
iin v
Gioo Gvco
=
v v
eo

io Zoo

1 Zoo c

ZL ZL + Zoo
v ZL
1+ v 1+ v
Zoo Zoo
(7.8)
The effect of the nonideal source (Zs , uins ) can be found by computing uin at
the presence of the source from Figure 7.6, which yields
 v  v
Toio Gvco Toio
uins + Zs v eo Gci +
v
v c
Zoo Zoo
uin = (7.9)
1 + Zs Yino
i

and substituting it in (7.6) with (7.9). These procedures yield

v
Toio v
Gvco Toio
G v
+
i
Yino v ci v


Zoo Zoo uins
1 + Z Yi 1 + Zs Yino
i
1 + Zs Yino
i
iin
uo
s ino
=
io v
Gioo
c
v
Zoo 1 + Zs Yinoc
i
1 1 + Zs Yin
i
Gv
vco
1 + Zs Yino
i
1 + Zs Yino Zoo 1 + Zs Yino Zoo
i v i

(7.10)
i
where Yin (the ideal input admittance) is equal to the ideal input admittance
Gv Gv
i
of the voltage-output converter (i.e., Yin = Yin
v
= Yino
v
ioo
Gvco
ci
), and
i
Yinoc (the open-circuit input admittance) is equal to the internal open-loop
i
input admittance of the voltage-output converter (i.e.,Yinoc = Yino
v
). If the
source interactions are to be dened for the load-affected converter, then all
the parameters in (7.10) would be the corresponding load-affected transfer
v
functions dened in (7.8). The ideal input admittance (Yin ) and the open-
i
circuit input admittance (Yinoc ) will be the only parameters, which will stay
intact.

Figure 7.6 Two-port model of a

current-output converter with
nonideal source and load.
218 7 Dynamic Modeling and Analysis of Current-Output Converters

According to (7.6), the load-affected control-to-output transfer function (GiL

co )
can be given by
Gvco
v
Zoo
(7.11)
ZL
1+ v
Zoo

and the corresponding voltage-output transfer function (GvL

co ) by

Gvco
(7.12)
Zv
1 + oo
ZL
Equation (7.11) shows that the load does not affect the control-to-output
transfer function, when ZL  Zoo v
. Equation (7.12) shows that the load does
not affect the control-to-output transfer function, when ZL  Zoo v
. In both the
cases, the resulting effect is typically a reduction of crossover frequency of the
current or voltage loop.
According to (7.11), it may be obvious that the resistive load would recover
the dynamical features of the voltage-output converter into the current-output
Gvco
co ZL , when ZL > Zoo . Similarly, the low-impedance
converter, because GiL v

load would change the voltage-output converter to have the features resembling
Gvco
current-output converter, when ZL < Zoo v
, because GvL
co ZL Zoo .

7.4
Cascaded Voltage-Current Loops

The rectiers in the Telecom DC-UPS applications (Figure 7.1) [4] are normally
operating at voltage-output mode, and when recharging the parallel connected
battery, the current-output mode is automatically switched on. The required
control system may be implemented either by using independent controllers
for the voltage and current loops or using a cascaded conguration, where the
voltage loop is the inner loop and the current loop the outer loop. The rst
case means that the voltage and current loops do not interact and, therefore,
the basic current-output dynamical modeling presented in Section 7.2 applies.
The second case means that the voltage and current loops are interconnected
and, therefore, a different technique has to be applied to nd the required
dynamical models.
In the cascaded case, the closed-loop voltage-output converter has to be
converted into the current-output mode. This can be accomplished by using
similar methods as with the open-loop converter, but the open-loop transfer
functions in (7.6) have to be replaced with the corresponding closed-loop
transfer functions as follows:
 7.5 Dynamic Review 219
Gvioc Toic
v
Tv v
Gvcoc Toic
v
+ oic Gvcic +

 Yinc uin
iin v
Zoc v
Zoc v
Zoc
=

uo (7.13)

io Gvioc 1 Gvcoc
v v v
c
Zoc Zoc Zoc
where the subscript extension c denotes closed loop and the corresponding
transfer functions are given in (7.14) (Gse = voltage sensing gain and
Lv (s) = voltage loop gain)

v v 
Yinc Toic Gvcic
Gvioc Zoc v
Gvcoc
Gv Gv Lv (s) Z v Gv Lv (s) Gvci Lv (s)
v
Yino ioov ci v
Toio + oov ci
Gco 1 + Lv (s) Gs 1 + Lv (s) Gse Gco 1 + Lv (s)
v
=


v
Gioo v
Zoo 1 Lv (s)

1 + Lv (s) 1 + Lv (s) Gse 1 + Lv (s)
(7.14)

7.5
Dynamic Review

The behavior of the control-to-output transfer function (Gico ) and the current-
loop gain (Lc (c)) of the current-output VMC and PCMC buck converter in
CCM would be analyzed dynamically in this section. The experimental buck
converter is shown earlier in Figure 7.4.
The predicted frequency responses of the internal (Gico ) and resistive-load-
affected (GiR
co ) control-to-output transfer functions are shown in Figure 7.7 at
the input voltage of 50 V and at the load of 2.5 A based on the analytical models
derived earlier in Chapters 3 and 4.
The solid lines represent the internal dynamics and the dashed lines the
4-
-resistor-affected dynamics. The VMC control-to-output transfer function
is multiplied by the modulator gain 1/3, and the PCMC control-to-output
transfer function is divided by the inductor-current-sensing resistor of 75 m
.
According to Figure 7.7a, the current-loop crossover frequency has clear ten-
dency to reduce approximately one decade, when the resistive load is connected
compared to the internal crossover frequency. According to Figure 7.7b, the
change of crossover frequency under PCM control could be approximately two
decades or more. These are clear indications and explain the observed phe-
nomena in [6]. This also means that the real reason for the observed problems
has been the incorrect control design. It is discussed in [6] that the unexpected
phenomenon can be avoided if the instantaneous current limiting of the PCM
control is only applied. This is true because the internal instantaneous current
limiting forces the converter to enter into the open-loop mode of operation
and, therefore, no feedback loop from the output current does exist.
220 7 Dynamic Modeling and Analysis of Current-Output Converters

Figure 7.7 The frequency responses current-output buck converter at the input
of the internal (solid lines) and resistive- voltage of 50 V and at the load current of 2.5
load-affected (dashed lines) control- A under (a) VM control and (b) PCM
to-output transfer functions of the control.

The current-loop frequency responses were measured by injecting the

excitation signal between the PWM block and the current-loop amplier
(20) (Figure 7.4). The load was a parallel connection of a pure 4-
resistor
and a 5-mF capacitor mimicking the dynamic short-circuit. The corresponding
frequency responses are shown in Figure 7.8, which shows that the predictions
 7.5 Dynamic Review 221

Figure 7.8 Measured frequency responses load (R, dashed lines) of a current-output
of the current-loop gain at the input voltage buck converter under (a) VM control and
of 50 V and the output load current of 2.5 A (b) PCM control.
using RC load (solid lines) and pure resistor

made earlier in conjunction with Figure 7.7 are correct with respect to the
crossover-frequency changes as well as the effect of a resistive load on the
converter dynamics. The shown internal crossover frequencies are actually
slightly load-affected due to the output-current-sensing resistor of 50 m

(Figure 7.4), which is higher than the high-frequency output impedance of the
converters causing slight reduction of the loop crossover frequency.
222 7 Dynamic Modeling and Analysis of Current-Output Converters

Figure 7.9 The internal impedance of the electronic load (dash-dot lines) and the output
impedances of the experimental converters (VMC solid lines, PCMC dashed lines).

There are available electronic loads providing constant-voltage-type-load

features. There are naturally high temptations to use them during the
design phase due to the exibility they provide. Figure 7.9 shows the internal
impedance (dash-dot lines) of a certain electronic load compared to the output
impedances of the experimental converters (VMC solid lines, PCMC dashed
lines).

Figure 7.10 The measured current loop gains at the RC load (solid lines) and the
electronic voltage-type load (dashed lines).
 References 223

Figure 7.9 shows that the internal impedance of the electronic load is
quite high and higher than the internal output impedances of the converters
in the frequencies starting from 300 Hz. This means that the measured
loop gain is load affected and does not give the correct information on the
dynamics of the loop gain. The corresponding measured current-loop gains
are shown in Figure 7.10, which conrms the conclusions made according to
the impedances. The current loop indicates very low crossover frequency and
phase margin (PM), which would lead to redesign of the controller. It may be
obvious that the redesign could lead even to a worse design by extending the
crossover frequency beyond the switching frequency. As a consequence, it is
highly recommended to use an RC load with sufciently large capacitor for
the validation purposes in order to avoid the problems originating from the
nonideal characteristics of the electronic loads.

References

1. T. Suntio and A. Glad, The batteries International Telecommunications

as a principal component in DC UPS Energy Conf., 1997, pp. 517522.
systems, in Proc. IEEE International 7. A. Tenno, R. Tenno, and T. Suntio,
Telecommunications Energy Conf., 1990, Battery impedance and its relation to
pp. 400411. battery characteristics, in Proc. IEEE
2. J. Lempinen and T. Suntio, International Telecommunications
Small-signal modeling for design of Energy Conf., 2002, pp. 176183.
robust variable-frequency yback 8. T. Sato, T. Nakano, and K. Harada,
battery chargers for portable device The overload-protection
applications, in Proc. IEEE Applied characteristics of the current-mode
Power Electronics Conf., 2001, pp. DC-to-DC converter: Analysis and
548554. improvements, in Proc. IEEE Power
3. T. Suntio, A. Glad, and P. Waltari, Electronics Specialists Conf., 1988, pp.
Constant-current vs. constant-power 830835.
protected rectier as a DC UPS 9. H. Matsuo, F. Kurokawa,
building block, in Proc. IEEE and T. Takeda, Analysis of the
International Telecommunications dynamic characteristics in the
Energy Conf., 1996, pp. 227233. over-current limited mode of the
4. T. Suntio, I. Gadoura, J. Lempinen, DCDC converter, IEEE Trans. Power
and K. Zenger, Practical design issues Electron., vol. 4, no. 2, 1989, pp.
of multi-loop controller for a telecom 175180.
rectier, in Proc. IEEE 10. G.C. Hsieh, Y.H. Lin, and H.C. Tasi,
Telecommunications Energy Special Control strategy for constant
Conf., 2000, pp. 197201. voltage/constant current
5. I. Gadoura and T. Suntio, switching-mode instrumentation
Implementation of optimal output power supply, in Proc. IEEE Industrial
characteristics for a telecom power Electronics Annual Conf., 2000, pp.
supply: Fuzzy-logic approach, Int. J. 983988.
Cont. Intell. Syst., vol. 30, no. 3, 2002, 11. M. Hankaniemi and T. Suntio,
pp. 134139. Small-signal models for
6. V.J. Thottuvelil, Modeling and constant-current regulated converters,
analysis of power converter systems in Proc. IEEE Industrial Electronics
with batteries, in Proc. IEEE Annual Conf., 2006, pp. 20372042.
224 7 Dynamic Modeling and Analysis of Current-Output Converters

12. M. Hankaniemi, T. Suntio, 14. T. Suntio, Unied average and

and M. Sippola, Analysis of the load small-signal modeling of
interactions in constant-current- direct-on-time control, IEEE Trans.
controlled buck converter, in Proc. Indust. Electron., vol. 53, no. 1, 2006,
IEEE International Telecommunications pp. 287295.
Energy Conf., 2006, pp. 343348. 15. R.D. Middlebrook and S. Cuk, A
13. M. Hankaniemi and T. Suntio, general unied approach in modeling
Dynamical modeling and control of switching-converter power stages, Int.
current-output converters, Int. Rev. J. Electron., vol. 42, no. 6, 1977,
Electr. Eng., vol. 2, no. 5, 2007, pp. 521550.
pp. 671680.
 225

8
Interconnected Systems

8.1
Introduction

The switched-mode converter is usually a subsystem within a system either

in a single equipment or in a wider constellation known as a distributed
power system [13] forming a dynamically complicated interconnected overall
system, which is prone to instability and performance degradation. The use
of distributed power architectures is continuously growing and would become
gradually the most popular way of applying the converters [3]. The distributed
systems can be found, for example, on electronic circuit boards [4, 5], in
space [6], and automotive [7] power applications, and so on. The stability
and performance problems are actually general problems, which can also be
encountered equally in the AC power systems [8, 9] and are supposed to be
increasing when more and more power electronic equipments are deployed
into the AC grid systems.
The switched-mode converters are a natural source of electromagnetic
interference (EMI) due to its operating principles. Therefore, the converters
are typically equipped with an input EMI lter to suppress the EMI noise to
an acceptable level in order to comply with the stringent international EMI
standards. The converter and the EMI lter form a complicated interconnected
system, which may easily affect the stability and dynamic performance of
the converter if not carefully designed [1012]. Those effects are not very
well understood [1315]. Basically it is a question of the differences in the
open-loop input admittance (Yino ), the ideal (Yin ) and short-circuit (Yinsc )
input admittances, which dictate the way the converter would react on the
source or EMI-lter interactions [16].
The stability analysis of the interconnected systems [1720] is typically based
on the application of the methods developed for the EMI-lter analysis and
design [10, 11] as the impedance ratio consisting of the output impedance of the
supplying subsystem and the input impedance of the load subsystem, which is
commonly known as minor-loop gain according to [10]. It is shown in [21, 22]
that the question is actually on the internal stability of the interconnected
system, in which the stability can be related to the minor-loop gain with

Dynamic Prole of Switched-Mode Converter. Teuvo Suntio

2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
ISBN: 978-3-527-40708-8
226 8 Interconnected Systems

certain conditions. As a consequence, the internal stability can be concluded

by applying the Nyquist stability criterion to the minor-loop gain. The results of
the analysis are usually expressed as a certain forbidden region on the left-half
plane of the complex plane out of which the minor-loop gain should stay in
order to ensure stability to exist. It is natural that an internally unstable system
is also unstable in the inputoutput sense, but the inputoutput instability
may not be necessarily explicitly visible.
It is well known that the load [20, 2326] and the source [27, 31, 32] as well
as even the remote sensing [28, 29] may signicantly change the dynamics of
a converter, and lead to the degradation of the transient performance or even
to instability. The combined effects of the load and source are usually most
surprising [32].
It may be difcult to know accurately enough the state of the parameters,
which may cause the stability and transient performance problems. Therefore,
it is obvious that the methods by means of which the interactions can be
reduced or totally eliminated are of importance [3335]. It is known [36]
that a high input-output-noise attenuation (i.e., Gioo 0) would make the
converter highly invariant to the source interactions [37, 38] as well as would
prevent the load interactions to propagate through the converter [32]. The
load interactions would be greatly reduced if the open-loop output impedance
(Zoo ) can be made very small [36] as demonstrated in [38, 39], which naturally
also gives very small closed-loop output impedance and highly improved
load-transient dynamics. In theory, it is possible to develop methods based
on the electrical interconnection of the subsystems [40], which would reduce
the interactions, but in practice such methods would not work or are not
applicable.
It may be obvious that most of the problems encountered in the practical
applications are due to the system interactions, and cannot be easily reproduce
in the laboratory for nding the reasons for the observed phenomena.
Therefore, it is important that the possible causes for such phenomena
are understood and recognized in advance. The knowing of the internal
prole of the associated converters is the key for understanding and successful
analyses. The internal dynamic prole can be found analytically by applying
the modeling methods described in detail in Chapters 37 or by measuring the
frequency responses of the dened transfer functions. The analytical models
and the measured transfer functions can be combined by means of proper
software packages such as Matlab.

8.2
Theoretical Interaction Formulation

The internal dynamics of a voltage-output convert can be represented by means

of a linear two-port model as shown in Figure 8.1 both at open and closed loop
 8.2 Theoretical Interaction Formulation 227

Figure 8.1 General internal linear

two-port model of an electrical
voltage-output system.

[41, 43]. The two-port model of Figure 8.1 can also be represented equally in a
matrix form at open loop by


  uin 
iin Yino Toio Gci
= io (8.1)
uo Gioo Zoo Gco
c
and at closed loop by
Gioo Gci Lv (s) Zoo Gci Lv (s) Gci Lv (s)

 Y Toio +
iin ino G 1 + Lv (s) Gco 1 + Lv (s) Gse Gco 1 + Lv (s)
= co

uo Gioo Zoo 1 Lv (s)

1 + Lv (s) 1 + Lv (s) Gse 1 + Lv (s)
 u 
in
io (8.2)
c
where Lv (s) denotes the output-voltage loop gain and Gse the output voltage
sensing gain. If the electrical system is unregulated, the control variable (c)
and the voltage-loop gain Lv (s) equals zero, and thus the dynamic description
would consist of only four transfer functions instead of six. The control
variable (c) corresponds to the reference voltage (ur ) at closed loop and may
be usually zero for a constant-voltage converter. Therefore, the closed-loop
system is most usually represented by using only four transfer functions
(i.e., Yinc , Toic , Gioc , and Zoc ), which are the four left-most elements
in (8.2).
Any electrical system can be represented by means of the two-port model of
Figure 8.1. The models can be connected in series and/or parallel. We would
only treat the systems comprising of series-connected subsystems of which
the intermediate bus architecture (IBA, Figure 8.2) is a typical example and
widely used in powering different electronics loads [35].

8.2.1
Load and Supply Interactions

The dynamic constellation of series-connected subsystems can be analyzed by

considering the source or upstream subsystem represented by its output port
228 8 Interconnected Systems

Figure 8.2 Intermediate-bus-architecture (IBA) based distributed power system.

Figure 8.3 Two-port model

with a nonideal input port
(i.e.,uins , Zs ) and nonideal
output port (i.e., jo , ZL ).

of the corresponding two-port model, and the load or downstream subsystem

by its input port, as illustrated in Figure 8.3.
The load and source effect on the converter dynamics can be solved either
by applying the extra-element-theorem method described in [42] or simply by
computing io and uin from Figure 8.3 in the presence of the nonideal source
and load, and substituting them in (8.2) and (8.3) with the computed formulas.
When following the described procedures, the load-affected sets of transfer
functions become
Gioo Toio ZL Toio Gco Toio

 Y ino + G ci +
Zoo + ZL Zoo + ZL Zoo + ZL
iin
= Gioo Zoo Gco
uo
Zoo Zoo Zoo
1+ 1+ 1+
ZL ZL ZL
 u 
in
jo (8.3)
c
 8.2 Theoretical Interaction Formulation 229

and
Gioc Toic ZL Toic Gcoc Toic
Yinc + Gcic +

 Zoc + ZL Zoc + ZL Zoc + ZL
iin
= Gioc Zoc Gcoc
uo
Zoc Zoc Zoc
1+ 1+ 1+
ZL ZL ZL
 u 
in
jo (8.4)
c

where the closed-loop internal transfer functions are dened explicitly in (8.2).
Similarly, the source-affected sets of transfer functions can be given by
Yino Toio Gci


iin 1 + Zs Yino 1 + Zs Yino 1 + Zs Yino
=
uo Gioo 1 + Zs Yinsc 1 + Zs Yin
Zoo Gco
1 + Zs Yino 1 + Zs Yino 1 + Zs Yino
 u 
ins
io (8.5)
c

and by
Yinc Toic Gcic


iin 1 + Zs Yinc 1 + Zs Yinc 1 + Zs Yinc
=
uo Gioc 1 + Zs Yinsc 1 + Zs Yin
Zoc Gcoc
1 + Zs Yinc 1 + Zs Yinc 1 + Zs Yinc
 u 
ins
io (8.6)
c

where Yin is the ideal input admittance and Yinsc the short-circuit input
admittance dened in (8.7). Both these special input admittances are load
independent, and the ideal input admittance is operation- and control-mode
independent as well [36]. The special admittances are explicitly presented
for the basic converters in Chapter 3 in conjunction with the corresponding
dynamical models, and naturally always computable by means of (8.7) for
any converter. The short-circuit input admittance can be measured accurately
enough by using the parallel connected resistor and large capacitor as a load
as described in Chapter 7. The ideal input admittance cannot be measured
directly because it is the internal feature of the converter.
Gioo Gci Gioc Gcic
Yin = Yino = Yinc
Gco Gcoc
(8.7)
Gioo Toio Gioc Toic
Yinsc = Yino + = Yinc +
Zoo Zoc
230 8 Interconnected Systems

If the open-loop input-to-output transfer function (Gioo ) can be made very

small, then it may be obvious that the corresponding converter would be
insensitive to the source interactions, because Yino = Yinc = Yinsc = Yin
and Yin is totally independent on the load impedance. The fact is, however,
that the total independence cannot be achieved due to the possible changes
in the output-to-input (Toio ) and control-to-input (Gci ) transfer functions.
Sometimes the small value of Toio or Gci would yield similar results.
If the open-loop output impedance (Zoo ) can be made very small,
it is obvious that the load effects would be highly reduced. The load
independence would not, however, be perfect, because the load may change
the input admittance (Yino ) and thus also the control-to-output transfer
function (Gco ) in addition with the output-to-input (Toio ) and control-to-
input (Gci ) transfer functions [32]. The small internal output impedance
makes actually the short-circuit input admittance in (8.7) to become innite,
and consequently the source-interaction formulation given in (8.6) is not
anymore valid. This also means that the extra-element-theorem-based method
may not be either applicable. We have actually used the circuit theory in
developing (8.6) and, therefore, the source-affected output impedance can be
S
given by Zoo = Zs Gioo Toio /(1 + Zs Yino
L
), where the superscripts S and
L denote the source and load, respectively. This means that the zero-output
impedance can be ensured only if Gioo 0.
According to (8.3) and (8.4), the control-to-output transfer function (Gco )
and thus the voltage-loop gain (Lv (s)) would be affected, if the open-loop
output impedance (Zoo ) is close to or greater than the load impedance (ZL ).
The load-affected loop gain LLv (s) can be given approximately by ZL Lv (s)/Zoo ,
which can be used to approximate how the loop characteristics would change.
It may be obvious that the output impedance having positive phase would
increase the phase lag of the loop. If the phase of the output impedance is
negative, then the phase lag will reduce. This means that the load effect may
be quite easily predicted when the open-loop output impedance is known.

8.2.2
Internal and InputOutput Stabilities

If the input-to-output transfer function (Gioo ) is zero or very small, then

the control-to-input transfer function (Gco ) and the output impedance (Zoo )
would stay intact under different source and load conditions as discussed
in the previous section. This also means that the output-voltage loop gain
(Lv (s)) and consequently the output transient dynamics would stay as they are
designed to be. This does not, however, mean that the system using this kind
of converters would be always stable. The same applies to the unregulated or
open-loop converters, which are often considered to be always stable due to
the lack of feedback.
 8.2 Theoretical Interaction Formulation 231

Figure 8.4 A cascaded system.

The concepts of internal and inputoutput stabilities would provide tools

[21, 22] to analyze the stability of interconnected systems. Consider the
cascaded system shown in Figure 8.4, where the source subsystem is denoted
by S and the load subsystem by L. The subsystems may be either open-
or closed-loop subsystems. Their internal dynamics can be represented by
means of four transfer functions from the input variables (i.e., uini , ioi ) to
the corresponding output variables (i.e., uoi , iini ). We assume without loss
of generality that the subsystems are of voltage-output type. We denote the
internal variables by u (i.e., uo1 , uin2 ) and i (i.e., io1 , iin2 ), because they are same
by denition.
The dynamic descriptions of the subsystems are given as follows:




iin1 S11 S12 uin1
=
u S21 S22 i



 (8.8)
i L11 L12 u
=
uo2 L21 L22 io2
when the intermediate variables are taken into account. The internal and
inputoutput stabilities can be studied by developing the mappings from
the system input variables (uin1 , io2 ) to the intermediate variables (u, i)
(i.e., internal stability) and to the system output variables (uo2 , iin1 ) (i.e.,
inputoutput stability). Theses procedures yield
S21 L11 L12



i 1 S22 L11 1 S22 L11 uin1
= (8.9)
u S21 S22 L12 io2
1 S22 L11 1 S22 L11
and
S12 S21 L11 S12 L12

 S11 +

iin1 1 S22 L11 1 S22 L11 uin1
= (8.10)
uo2 S21 L21 S22 L12 L21 io2
L22 +
1 S22 L11 1 S22 L11
Some of the converters may have S21 = 0 or L21 = 0. This means that some of
the entries in the matrices of (8.9) or (8.10) may be zero or equal to the internal
transfer functions of the original subsystems (i.e., S11 or L22 ) and consequently
providing no additional information on the stability of the cascaded system. We
assume that the subsystems are internally stable. Consequently, the intercon-
nected system is stable if the characteristic polynomial 1 S22 L11 does not have
232 8 Interconnected Systems

roots in the closed right-half plane of the complex plane. The stability can also
be inferred applying the Nyquist stability criterion to S22 L11 . S22 L11 is usu-
ally known as the minor-loop gain according to [10] and constitutes, in practice,
of the output impedance of subsystem S and the input impedance of subsystem
L, respectively. The presented system-theoretic approach is scientically sound
and proofs that the minor-loop gain can be used to study the stability of cascaded
system, but it does not take any stand on the performance of the system (i.e., the
gain and phase margins of the minor-loop gain do not necessarily comply to the
same output-voltage-loop-gain margins of the supply or load converter) [30, 31].
In the case of supply converter (i.e.,S), the minor-loop gain can be naturally
given by Zoc /ZL , which equals Zoo /(1 + Lv (s))/ZL . If the voltage-loop gain
is high, then the minor-loop gain can be approximated by Zoo /Lv (s)/ZL ,
which has an inverse relation to the real load-affected loop gain as discussed
in the previous section. If the loop gain is small, then the minor-loop gain
would correspond to Zoo /ZL , which does not have any relation to the real
load-affected loop gain [30]. In practice, this means that the gain margin of
the minor-loop gain corresponds to the gain margin of the voltage loop of the
supply converter at low frequencies, but more gain margin is required at those
frequencies close to and higher than the crossover frequency of the internal
voltage loop in order to ensure that the dynamic characteristics related to the
voltage loop would stay intact.
Similar explicit theoretical formulation between the minor-loop gain of the
source impedance and the input impedance of the supply converter is difcult
to develop, but quite the same as discussed above would apply [31].
The minor-loop gain is typically used to construct the forbidden region in
the complex plane out of which the minor-loop gain should stay in order to
ensure stability. The shaded are in Figure 8.5 indicates the forbidden region
yielding from applying the ESAC criterion [18]. The area outside the circle

Figure 8.5 The forbidden

region.
 8.2 Theoretical Interaction Formulation 233

denoted by the radius 1/GM is the forbidden region when the input-lter
design rules [10] are applied. This criterion is usually deemed to be too
conservative [18] and requiring to using extra capacitors for ensuring stability.
Therefore, new advanced criteria are developed to allow impedance overlap
as well.
The system theoretical formulation given above indicates that the instability
would take place, when the Nyquist stability criterion is violated regardless of
the interface, where the impedances are dened [22]. This does not, however,
ensure that the dynamic performance of the converter would be intact or
sufcient.
Figure 8.6 shows an interconnected system, where we have different
interfaces from A1 to An . Let us assume that the interface A1 describes
a typical interface between an EMI lter having output impedance (Zs )
and a regulated converter consisting only of the pure power stage without
any additional passive components added in its input with the closed-loop
input impedance (Zinc ). The closed-loop input impedance (Zinc ) resembles
typically an incremental negative resistor, especially at low frequencies but
usually also at high frequencies. The LC-type input lter has its maximum
impedance value at the resonant frequency. Its phase is also zero at the
resonant frequency. This means that no impedance overlap is usually allowed
at this interface in order to ensure stability. This also means that the input-lter
design rules given in [10] are not conservative when considered in their correct
context. Even if we do not have impedance overlap at the direct input of the
converter, the dynamics of the converter may be degraded according to (8.5),
because the open-loop input impedance would dictate the level of affection.
Therefore, the minor-loop-gain margins should be sufcient as discussed in
[10] in order to ensure proper dynamic performance.
If we assume that the interface A4 is at the direct output of the corresponding
converter, then the impedance overlap would always mean that the voltage-
loop gain would be affected, and may not be desired. Even if we do not have
impedance overlap, the loop gain may be affected because the level of affection
would be determined by the open-loop output impedance as shown in (8.3) [30].

Figure 8.6 Practical interconnected system with several different interfaces.

234 8 Interconnected Systems

If we assume that the load converters at the interfaces A3 and An1 are
equipped with input EMI lters, then the resonant nature of these lters would
naturally cause impedance overlap at those interfaces. The EMI lters do easily
mask the overlap from existing at the pure input of the corresponding convert-
ers and, therefore, the overlap would be allowed. If we assume that the interface
A2 is the direct output of the corresponding converter, then the impedance
overlap would mean that the dynamics of the converter would change.
The engineers designing distributed interconnected systems should be con-
cerned to ensure both stability and dynamic performance of the overall system.
The presented stability-ensuring methods based on the nonconservative for-
bidden regions (Figure 8.5) do not necessarily ensure the dynamic quality of
the system. Some practical evidence is provided in Section 8.4.

8.2.3
Output Voltage Remote Sensing

The output-voltage remote sensing (Figure 8.7) is sometimes used to improve

the static voltage accuracy of the sensitive loads [28] or even to improve the
load transient performance. A stability assessment tool to compute the effect
of the connection impedance (Zcon , Figure 8.7) on the output-voltage loop gain
of the converter is introduced in [28], but no analysis is provided to be able to
conclude what the real dynamical effects of the remote sensing may actually be.
The effect of remote sensing [29] on the converter overall dynamics can be
computed by modeling both the converter and the remote-sensing-impedance
block (Zcon ) by means of their corresponding two-port models as shown
in Figure 8.8. The original dynamics of the converter is naturally activated
when the output-voltage feedback is from the interface A1 (Figure 8.8b). The
remote-sensing-affected dynamics would be activated when the output-voltage
feedback is moved to the interface A2 . The remote-sensing-affected dynamic
representation can be solved by computing io and uinc from Figure 8.8b and
replacing them in the input port of the converter and in the output port of the
impedance block with the computed formulas.
Following the dened procedures, the remote-sensing-affected dynamic
representation of the corresponding open-loop system yields
Yinc Gioo Toio Toic Toio Yinc Gco Toio

 Yino + Gci +
iin 1 + Zoo Yinc 1 + Zoo Yinc 1 + Zoo Yinc
=
uoc Gioc Gioo 1 + Zoo Yinsc
c
Gioc Gco
Zoc
1 + Zoo Yinc 1 + Zoo Yinc 1 + Zoo Yinc
 u 
in
ioc (8.11)
c
 8.2 Theoretical Interaction Formulation 235

Figure 8.7 Application of

output-voltage remote sensing.

Figure 8.8 Output-voltage-

remote-sensing constellation,
where (a) denes two-port
model of the converter and
(b) the two-port model of the
impedance block (Zcon ).

c
where Yinsc is the short-circuit input admittance of the impedance block given
by
Gioc Toic
c
Yinsc = Yinc + (8.12)
Zoc
It may be obvious according to (8.11) that the dynamic effect of the remote
sensing may be much more severe than expected if considering only the
changes in the output-voltage loop gain (i.e., GRSco ) as in [28]:
If we consider that the impedance block consists only of the connection
cabling (Figure 8.9a), which can be modeled by means of a series connection
of an inductor (Lc ) and a resistor (rLc ), then the impedance block model can
be given by


0 1
(8.13)
1 (sLc + rLc )

and the corresponding remote-sensing-affected dynamical models of the

overall system by


  uin 
iin Yino Toio Gci
= ioc (8.14)
uoc Gioo (Zoo + sLc + rLc ) Gco
c
236 8 Interconnected Systems

Figure 8.9 Impedance-block

circuitries: (a) cabling connection
and (b) T-type LCL circuit.

According to (8.14), the dynamic effect may be interpreted to be insignicant,

but it is not true because the adding of an inductive component at the output
would increase the load sensitivity of the converter: the capacitive load would
easily reduce the phase margin of the voltage loop and deteriorate the load
transient performance.
More complicated LC structures may change all the dynamical parameters
and lead to a severe deterioration of the converter dynamics. Example of
such a connection is a T or L1 CL2 connection (Figure 8.9b), which can be
characterized by
s 1+sr C
cc c
Lc1 Lc1 Cc

2 rLc1 + rcc 1 rLc1 + rcc 1
s +s + s +s
2 +
L L c1 Cc Lc1 Lc1 Cc
c1
1 + srcc Cc (rLc1 + s Lc1 )(1 + srcc Cc )


Lc1 Cc Lc1 Cc
(sLc2 + rLc2 + )
rLc1 + rcc 1 rLc1 + rcc 1
s2 + s + s2 + s +
Lc1 Lc1 Cc Lc1 Lc1 Cc
(8.15)
According to (8.15), the converter-side LC circuit may cause peaking at the
corresponding resonant frequency and the lter output-side inductor would
increase the high-frequency output impedance, which implies sensitivity to
capacitive load. The resonant frequency locates typically at high frequencies,
and would increase the loop gain and decrease rapidly the phase. Such
a phenomenon may easily lead to instability. Some practical evidence is
provided in Section 8.4.

8.2.4
Input EMI Filter

The switched-mode converters are typically equipped with an input EMI

lter (Figure 8.10a) in order to suppress the high-frequency current to a level
complying with the relevant EMC standards. The EMI lter may affect the
dynamics of the converter if not carefully designed [10]. The lter effect on
 8.2 Theoretical Interaction Formulation 237

Figure 8.10 Input-EMI-lter

constellation, where part
(a) describes the physical
connection and part (b) the
dynamical model.

the converter dynamics may be analyzed by modeling the EMI lter with its
two-port model (Figure 8.10b), and applying the same procedures as described
in the previous section.
Following the proposed technique, the EMI-lter-affected dynamic open-loop
representation of the converter yields
Giof Toif Yino Toif Toio Toif Gci

 Yinf +
iins 1 + Zof Yino 1 + Zof Yino 1 + Zof Yino
=



uo Giof Gioo 1 + Zof Yinsc 1 + Zof Yin
Zoo Gco
1 + Zof Yino 1 + Zof Yino 1 + Zoo Yino
 u 
ins
io (8.16)
c
where Yin and Yinsc are the ideal and short-circuit input admittances of
the original converter, respectively. The stability of the EMI-lter-converter
system has to be checked by applying the Nyquist stability criterion to the
corresponding minor-loop gain, which usually consists of the lter output
impedance (Zof ) and the closed-loop input impedance of the converter
(Zinc = 1/Yinc ). Sometimes, the converter is used at open loop, especially
in the intermediate bus architectures [5]. In those cases, the minor-loop gain
consists of the lter output impedance and the open-loop input impedance of
the converter (i.e., Zino = 1/Yino ).
According to (8.16), the EMI lter would always affect the dynamic prole
of the associated converter regardless of the value of the internal input-to-
output transfer function (Gioo ). The high input-to-output noise attenuation
at open loop (i.e., Gioo 0) would make the output dynamics of the converter
insensitive to the EMI lter providing that the EMI-lter-converter system
is stable. If high input-to-output noise attenuation does not exist, then
the behavior of the open-loop input impedance and the EMI-lter output
impedance would determine the lter effects. Some experimental evidence is
provided in Section 8.4.
238 8 Interconnected Systems

8.3
Review of Methods to Reduce the Interactions

If the open-loop input-to-output transfer function (Gioo ) and the open-loop

output impedance (Zoo ) can be made very small, then the source and load
interactions would be highly reduced according to the theoretical formulations
derived in Section 8.2.1. The voltage-mode-controlled converters do not have
such features as demonstrated in Chapter 3.
A CCM buck converter under optimally compensated peak-current-mode
(Chapter 4), average-current-mode (Chapter 5), and self-oscillation (Chapter 6)
controls would have high input-to-output noise reduction, which would reduce
the source interactions and prevent the load interactions to propagate through
the converter. The other basic converters do not posses such a feature.
In addition to the above-named control techniques, the high input-noise
attenuation can be accomplished with certain conditions applying input-voltage
feedforward (IVFF) [37, 44, 45].
The control techniques introduced in Chapters 47 tend to increase the
open-loop output impedance and, as a consequence, the load interactions,
of which the reduction of the voltage-loop crossover frequency is the most
common outcome. In order to reduce the basic load interactions, the open-loop
output impedance shall be made small. In practice, this can be accomplished
with certain conditions by applying output-current feedforward (OCFF) [38, 39,
4648].

8.3.1
Input-Voltage Feedforward

The effect of IVFF on the converter dynamics can be analyzed by using the
control engineering block diagrams given in Figure 8.11, which yield the
dynamic description of the converter to be [37]


  uin 
iin Yino Fm qi Gci Toio Fm Gci
= io (8.17)
uo Gioo Fm qi Gco Zoo Fm Gco
uco
It may be obvious according to (8.16) that the original dynamic nature of the
converter would not be changed, that is, if the original converter has resonant
nature, then the IVFF converter would have also the resonant nature. The
IVFF control does not change the output impedance and the output-to-input
transfer function.
The aim of the IVFF is to produce high input-noise attenuation and,
therefore, GIVFF
ioo has to be zero. According to (8.17), this would take place
when Gioo Fm qi Gco = 0, and as a consequence, Fm qi = Gioo /Gco . It may
be obvious that the stated condition can be implemented properly only if Gioo
and Gco have the same zeros as in a buck converter. This means, in practice,
 8.3 Review of Methods to Reduce the Interactions 239

Figure 8.11 Control engineering block diagrams for assessing the effect of input-voltage
feedforward on (a) output dynamics and (b) input dynamics.

that Fm qi D/Uin . The required condition can be accomplished making the

PWM ramp slope directly dependent on the input voltage as illustrated in
Figure 8.12. The comparator equation determining the duty ratio can be given
by uco = kuin dTs according to Figure 8.12 from which the duty-ratio constraints
can be derived by developing the proper partial derivatives as follows:
1
d = (uco kDTs uco ) (8.18)
kUin Ts

which gives Fm qi = D/Uin as required for GIVFF

ioo 0.
The constant k (Figure 8.12) is designed so that the duty ratio is close
to 100% at a certain minimum input voltage (Uinmin ). According to this
assumption, k = VM /Uinmin Ts , where VM = Umax Umin (Figure 8.12). Thus
Fm = Uinmin /VM Uin and qi = DVM /Uinmin . This means that the control-to-
co ) and the control-to-input (Gci
output (GIVFF IVFF
) transfer functions would be
near to independent of the input voltage and, therefore, the output-voltage-loop
dynamics does not depend anymore on the input voltage.

Figure 8.12 Duty-ratio

generation under IVFF
control.
240 8 Interconnected Systems

The scheme of Figure 8.12 is typically implemented in the commercial

PWM modulators by charging the timing capacitor via a resistor from the
input voltage yielding an exponential curve instead of the linear curve. Despite
the deviation from the theoretical ideal, the scheme would produce near to
expected results [37].
It may be obvious from Figure 8.12 that the maximum pulsewidth of the
converter would be limited. This means, in practice, that the load-transient
dynamics of the converter would be deteriorated, because the derivative of the

average inductor current would be limited according to dmax m1 dmax m2 .
More detailed information on the IVFF scheme can be found from [37] and
[45]. Experimental evidence would be provided in Section 8.4.

8.3.2
Output-Current Feedforward

It is well known [46] that the OCFF control would improve the load-transient
dynamics of a converter. The improvement results from the reduction of the
open-loop output impedance. It is claimed in [39] that the open-loop output
impedance can be made zero for every converter, but it is not true because of
the same reasons as discussed in the previous section in conjunction with the
input-to-output transfer function.
The theoretical formulation dening the proper conditions for the zero
open-loop output impedance can be derived from the control engineering
block diagrams as shown in Figure 8.13, where Rs2 is the output-current
sensing resistor,Hi the output-current-loop sensing gain, and Ga the transfer
function from the control voltage to the original control variable of the
converter.

Figure 8.13 Control engineering block diagrams for assessing the effect of
output-current feedforward on (a) output dynamics and (b) input dynamics.
 8.4 Experimental Dynamic Review 241

According to Figure 8.13, the dynamic description of the converter can be

given by


  uin 
iin Yino Toio + Rs2 Hi Ga Gci Ga Gci
= io (8.19)
uo Gioo (Zoo Rs2 Hi Ga Gco ) Ga Gco
uco

which indicates that the OCFF control would change only the open-loop output
impedance and the output-to-input transfer function. The other transfer
functions would stay effectively intact, because the control-related transfer
functions are actually same as in the original converter when the voltage-type
control variable is considered.
The aim of the OCFF control is to produce zero open-loop output impedance.
OCFF
Therefore, Zoo = Zoo Rs2 Hi Ga Gco = 0. As a consequence, the OCFF loop
gain should be designed as

1 Zoo
Hi = (8.20)
Rs2 Ga Gco

It may be obvious that the gain required for Hi in (8.20) may be implemented
in reality only if the control-to-output transfer function (Gco ) is of minimum
phase, that is, it does not contain RHP zero. Thus the buck converter of
the basic converters is the only prospective candidate. It is observed that the
unity-gain OCFF control (i.e.,Hi = 1) would provide highly improved load-
transient response in a hysteretic current-mode [46] and peak-current-mode
[38, 47] controlled buck converters. In order to improve the transient response
of a voltage-mode-controlled buck converter, the OCFF loop gain should be
effective at the resonant frequency of the converter, where the open-loop output
impedance has the highest value. Such a principle has been tried in [48], but
the deciencies in the theoretical treatment have lead to poor performance.
More detailed discussions on the subject can be found from [38]. Experimental
evidence is provided in Section 8.4.

8.4
Experimental Dynamic Review

We show some practical evidence of the interactions, which may take place
in the real interconnected systems comprising of switched-mode converters
in the subsequent sections. The intention is to prove that different converters
have their unique dynamic proles determining how the converter would react
to the external world. The dynamic proles have been already disclosed in the
previous chapters and some discussions made about the possible changes the
external impedances may cause in the dynamic behavior of the converter. We
also demonstrate that the severity of the interactions can be reduced by using
242 8 Interconnected Systems

Figure 8.14 The experimental

buck converter.

proper feedforward control. The experimental material is mainly based on the

buck converter as shown in Figure 8.14 under different control methods.
The output-voltage loop gains of the VMC, PCMC, IVFF, and PCMC-OCF
buck converters were designed to be close to similar at the input voltage of
50 V as shown in Figure 8.15.
The basic load effects are demonstrated connecting at the output of the
converter a series resonant LC having resonant frequency close to that
of the VMC converter. The converters are stable at the resonant load as
shown in Figure 8.16, because the load impedance and the closed-loop output
impedances of the converters do not overlap. The state of the dynamic prole
of the converter cannot be concluded based on Figure 8.16.
The source effects are demonstrated connecting an EMI lter (Figure 8.17)
at the input of the converter. The resonant frequency of the lter was chosen
to be close to that of the VMC converter. The measured closed-loop input

Figure 8.15 Measured frequency responses of the voltage-loop gains of the experimental
converters at the input voltage of 50 V.
 8.4 Experimental Dynamic Review 243

Figure 8.16 The measured closed-loop output impedances of the converters versus the
load impedance.

Figure 8.17 Input EMI lter.

impedances of the converters and the lter output impedance are shown in
Figure 8.18 proving the stability of the cascaded system due to the lack of
impedance overlap. The state of the dynamic prole of the converters cannot
be concluded based on the information provided by Figure 8.18.

8.4.1
Load and Supply Interactions

A series-resonant load (i.e., 500 Hz) was connected at the output of the buck
converters shown in Figure 8.14 at the input voltage of 20 V. The corresponding
open-loop internal impedances (Zoo ) and the load impedance (ZL ) are shown
in Figure 8.19. According to (8.3), the load impedance would affect the voltage-
loop gain if ZL Zoo . From Figure 8.19, it may be obvious that all the other
voltage-loop gains except the loop gain of the PCMC-OCF converter would be
244 8 Interconnected Systems

Figure 8.18 The closed-loop input impedances versus the output impedance of the EMI
lter.

Figure 8.19 The open-loop output impedances versus the 500-Hz resonant load
impedance.

affected. Figure 8.20 conrms the predictions. There are, however, no dynamic
effects because the gain and phase margins are not changed.
The open-loop input impedances (Zino ) and the output impedance of the
EMI lter (Figure 8.17) are shown in Figure 8.21. The impedance overlap
in the VMC converter means that its voltage-loop gain would be affected.
Figure 8.22 conrms the prediction. The other loop gains are intact.
 8.4 Experimental Dynamic Review 245

Figure 8.20 The resonant-load-affected output-voltage loop gains.

Figure 8.21 The open-loop input impedances versus the EMI-lter output impedance.

Figure 8.23 shows the combined effect of the resonant load and source. The
VMC converter is very close to instability. The voltage-loop gains of the other
converters are not changed.
The reason for the observed phenomenon is the behavior of the open-loop
input-to-output transfer function (Gioo ) (Figure 8.24) and the output-to-input
transfer function (Toio ) (Figure 8.25). According to (8.3), the load will affect
the open-loop input admittance if the product Gioo Toio is not zero. According
to Figures 8.24 and 8.25, the product Gioo Toio would amplify the load
interactions especially at the frequencies close to the resonant frequencies
246 8 Interconnected Systems

Figure 8.22 The EMI-lter affected output-voltage loop gains.

Figure 8.23 The resonant-load and EMI-lter-affected output-voltage loop gains.

as demonstrated in Figure 8.23. The PCMC and IVFF converters have rather
high input-to-output noise attenuation, which also means reduced reected
load interactions.
A series resonant load (8 kHz) was connected at the output of the VMC
and PCMC buck converters (Figure 8.14). The minor-loop gain (Zoc /ZL ) was
measured and is shown in Figure 8.26. The cascaded systems are stable because
of the positive phase margins. The minor-loop gain of the VMC converter shows
slight impedance overlap, but the minor-loop gain of the PCMC converter is
always less than unity and, therefore, no impedance overlap does exist. We
 8.4 Experimental Dynamic Review 247

Figure 8.24 The input-to-output transfer functions (Gioo ).

Figure 8.25 The output-to-input transfer functions (Toio ).

may conclude that the voltage-loop gain of the VMC converter may be slightly
changed but the voltage-loop gain of the PCMC converter would be intact.
Figure 8.27 shows that the crossover frequencies of the converters are
reduced equally from 11 to 6 kHz despite the information given by the
minor-loop gain. The open-loop output impedances and the resonant-load
impedance are shown in Figure 8.28, which indicates impedance overlap in
both the converters and, consequently, a reduction of loop gain.
248 8 Interconnected Systems

Figure 8.26 The minor-loop gains of the VMC (solid line) and PCMC (dashed line)
converters at 8-kHz resonant load.

Figure 8.27 8-kHz-resonant-load-affected voltage-loop gains of the VMC and PCMC

converters.

The load of the VMC buck converter was considered to be such that
the minor-loop gain (Zoc /ZL ) has a continuous 6-dB gain margin. The
corresponding phase margin was set to 0 and 60 . The original voltage-loop
gain (solid line) and the load-affected loop gains with 0 phase margin (dashed
line) and 60 phase margin (dotted line) are shown in Figure 8.29. At those
frequencies, where the voltage-loop gain is high, the minor-loop-gain-related
margins are also directly reected to the voltage-loop gain. When approaching
 8.4 Experimental Dynamic Review 249

Figure 8.28 The open-loop output impedances versus the 8-kHz resonant-load
impedance.

Figure 8.29 The internal (solid line) and load-affected loop gains
of the VMC buck converter, where the minor-loop gain (Zoc /ZL )
has continuous 6-dB gain margin and 0 (dashed line) or 60
(dash-dotted line) phase margin.

the crossover frequency, the 6-dB gain margin does not naturally hold any
more, and the crossover frequency may either slightly increase (PM = 0 ) or
decrease (PM = 60 ). In both the cases, the voltage-loop-gain phase margin
reduces from the original phase margin. In practice, this means that the minor-
loop-based margins have to be higher than that usually used in the voltage-loop
250 8 Interconnected Systems

gain for robust stability and in order to maintain adequate dynamics of the
converter.
The source impedance of a VMC buck converter was considered to be
such that the minor-loop gain (Zs /Zinc ) has a continuous 6-dB gain margin.
The corresponding phase margins were set to 0 and 60 . The original
voltage-loop gain (solid line) and the source-affected voltage-loop gains with
0 (dashed line) and 60 phase margin (dash-dotted line) are shown in
Figure 8.30. At the low frequencies, the 6-dB minor-loop-gain-related margin
would ensure intact dynamics. At the higher frequencies, the gain margin
should be higher than the usual 6 dB in order to ensure robust stability and
performance.
The closed-loop output impedance (Zoc ) of the converter can be used to
estimate the load sensitivity by considering that the instability boundary can
be given according to the minor-loop gain (Zoc /ZL ) as ZL = Zoc . This
means that ZL = |Zoc | Zoc 180 [30]. According to this procedure, the
load-impedance instability boundary can be given as shown in Figure 8.31
for VMC and PCM buck converters (Figure 8.14). It shows that the VMC
converter is sensitive to the capacitive load at the frequencies lower than or
equal to its resonant frequency (i.e., the phase is equal to or less than 90 ).
The PCMC converter can be unstable due to the capacitive load only at the
low frequencies. The situation with the other types of loads can be similarly
considered.

Figure 8.30 The internal (solid line) and source-affected loop

gains of the VMC buck converter, where the minor-loop gain
(Zs /Zinc ) has continuous 6-dB gain margin and 0 (dashed line) or
60 (dotted line) phase margin.
 8.4 Experimental Dynamic Review 251

Figure 8.31 The instability boundary of the VMC and PCMC buck converters (Figure 8.14).

8.4.2
Remote Sensing

The impedance block (Figure 8.32) mimicking the impedance of a cabling was
connected at the output of the VMC converter (Figure 8.14). The original and
the cable-affected (RS) open-loop output impedances are shown in Figure 8.33.
The inductive nature of the high-frequency output impedance implies
increased sensitivity to the capacitive load when the remote sensing is
connected. The measured output-voltage loop gains at ideal constant current
load (L(s), solid line) and at capacitive load (i.e., RL = 4, CL = 110 F, ESR =
100 m) with the remote sensing connected (LRS (s), dash-dot line) and without
it (LC (s), dashed line) are shown in Figure 8.34: the remote sensing has clearly
increased the phase lag and may easily lead to instability if the gain would
increase.
The LCL circuit shown in Figure 8.36 was connected at the output of the
synchronous buck converter shown in Figure 8.35. The measured original and
the connection-block-affected output impedances are shown in Figure 8.37.

Figure 8.32 The cabling impedance.

252 8 Interconnected Systems

Figure 8.33 The original and cabling-affected (RS) open-loop output impedances.

Figure 8.34 The measured internal (L(s), solid line) the

capacitive-load affected (LC (s), dashed lie) without remote
sensing, and the capacitive-load-affected (LRS (s), dash-dot line)
with remote sensing output-voltage loop gains.

The inductive nature of the high-frequency output impedance implies in-

creased sensitivity to capacitive load: the measured voltage-loop gains in
Figure 8.38 conrm the prediction, where the internal loop gain is denoted
by means of solid line, the 4-F-capacitive-load-affected loop gain (LC (s)) by
dashed line, and the 4-F-capacitive-load-affected with remote sensing (LRS (s))
 8.4 Experimental Dynamic Review 253

Figure 8.35 Experimental

synchronous buck converter.

Figure 8.36 T-type connection impedance block.

Figure 8.37 The measured internal (Zoo ) and the connection-block-affected (Zoo
RS
)
output impedances.

by dash-dot line. It may be obvious that a slight increase in the loop gain would
lead to instability.
The effect of the connection impedance (Figure 8.36) could have been
more severe as discussed in Section 8.2.3, but the capacitor (Cc ) had to
be selected to be an electrolytic capacitor with rather high ESR in order
to keep the converter stable. Therefore, the resonant peaking in the LCL
circuit is very low as shown in Figure 8.39. Consequently, the interactions
are also reduced. In the practical applications, the resonant peaking could
254 8 Interconnected Systems

Figure 8.38 The measured internal (solid line), the

capacitive-load-affected (LC (s), dashed line), and the
capacitive-load-affected loop gain with remote sensing (LCRS (s),
dash-dot line).

Figure 8.39 The internal input impedance (Zinc ) and the input-to-output transfer
function (Gioc ) of the LCL circuit.

be high and the interactions also more severe due to the nature of the
circuitry.
The synchronous buck converter was subjected to a constant-current-type
load change from 1 to 5 A with slew rate of 2.5A/s. The corresponding
 8.4 Experimental Dynamic Review 255

Figure 8.40 The output-voltage transient (a) original response, (b) response with the
responses of the synchronous buck LCL circuit connected between the converter
converter, when the constant-current-type and load, (c) response, when the remote
load changes from 1 to 5 A (2.5 A/s): sensing is connected over the LCL circuit.

output-voltage transient responses without the LCL circuit (a), with the LCL
connected between the converter and load (b), and applying remote sensing
over the LCL circuit (c) are shown in Figure 8.40: the remote sensing obviously
improves the static accuracy but not the transient response.

8.4.3
System Stability

A small-scale IBA system was constructed based on nonisolated synchronous

buck converters as shown in Figure 8.41. There are numerous different
interfaces (Ai ) from which the stability information can be extracted in terms
of different minor-loop gains. The stated connection impedances are cabling
impedances. The buck converters have to be equipped with an input capacitor
in order to work well. Therefore, the plain input of the converter is never
measurable.
Figure 8.42 shows the measured minor-loop gains at the interfaces AP4 and
AP3 as well as the computed minor-loop gain existing at the direct input of the
corresponding buck converter. The minor-loop gain at AP4 (solid line) shows
that the system is stable although slight impedance overlap exists. The minor-
loop gain at AP3 indicates high impedance overlap and phase close to 180 .
The Nyquist plot shows that the minor-loop gain at AP3 also indicates stable
system, because it does not encircle the point (1, 0) in clockwise direction.
The information the minor-loop gain provides about the robust stability is,
256 8 Interconnected Systems

Figure 8.41 Experimental IBA system.

Figure 8.42 Measured minor-loop gains at the interface AP4 (solid line) and AP3 as well
as the computed minor-loop gain at AP2 .

however, very vague. The computed minor-loop gain at AP2 conrms the situ-
ation. The message is that every interface has its own minor-loop gain giving
information, which may or may not be useful. The dynamic performance of
the system can be inferred only if the minor-loop gain represents the direct
input or output of the converter.
The time-domain information may often be very confusing as shown in
Figure 8.43, where two different voltage measurements are shown at the
 References 257

Figure 8.43 The voltage of the main bus (AM1 ): (a) the beat
frequency of the bus converter and the rst POL converter, and
(b) the beat frequency of the bus converter and the second POL
converter.

interface AM1 (i.e., the main bus). The curves show sinusoidal signals at differ-
ent frequencies depending on the loading condition of the system and the POL
converters. Figure 8.43a shows a sinusoidal signal of 8 kHz, which is actually
the switching-frequency difference (beat frequency) of the bus converter and
the rst POL converter. Figure 8.43b shows similar beat-frequency signal of the
bus converter and the second POL converter. These signals can be easily inter-
preted as a sign of instability when they appear along the increase of the loading.

References

1. S. Luo and I. Battarseh, A review of Telecommunications Energy Conf., 2004,

distributed power systems Part I: DC pp. 234240.
distributed power system, IEEE 5. H. Huang, Coordination of design
Aerosp. Electron. Syst. Mag., vol. 20, no. issues in the intermediate bus
8, 2005, pp. 515. architecture, in Proc. IEEE Applied
2. S. Luo and I. Battarseh, A review of Power Electronics Conf., 2005, pp.
distributed power systems Part II: 169175.
High frequency AC distributed power 6. J.R. Lee, B.H. Cho, S.J. Kim, and F.C.
systems, IEEE Aerosp. Electron. Syst. Lee, Modeling and simulation of
Mag., vol. 21, no. 6, 2006, pp. 514. spacecraft power systems, IEEE
3. L. Brush, Distributed power Trans. Aerosp. Electron. Syst., vol. 24,
architecture demand characteristics, no. 3, 1988, pp. 295303.
in Proc. IEEE Applied Power Electronics 7. A. Emadi, A. Khaligh, C.H. Rivetta,
Conf., 2004, pp. 342345. and G.A. Williamson, Constant power
4. R.V. White, Using on-board power loads and negative impedance
systems, in Proc. IEEE International instability in automotive systems:
258 8 Interconnected Systems

Denition, modeling, stability, and Trans. Power Electron., vol. 10, no. 3,
control of power electronic converters 1995, pp. 280285.
and motor drives, IEEE Trans. Veh. 18. S.D. Sudhoff, S.F. Glover, P.T.
Technol., vol. 55, no. 4, 2006, pp. Lamm, D.H. Schmucker, and D.E.
11121125. Delisle, Admittance space stability
8. C.D. Davidson and R. Scasz, analysis of power electronic systems,
Compatibility of switched-mode IEEE Aerosp. Electron. Syst., vol. 36, no.
rectiers with engine generators, in 3, 2000, pp. 965973.
Proc. IEEE International 19. J.M. Zhang, X.G. Xie, D.Z. Jia,
Telecommunications Energy Conf., 2000, and Z. Qian, Stability problems and
pp. 626631. input impedance improvement for
9. C.M. Hoff and S. Mulukutla, Analysis cascaded power electronic systems, in
of the instability of PFC power Proc. IEEE Applied Power Electronic
supplies with various AC sources, in Conf., 2004, pp. 10181024.
Proc. IEEE Applied Power Electronics 20. S. Abe, H. Nakagawa, M. Hirokawa,
Conf., 1994, pp. 696702. T. Zaitsu, and T. Ninomiya, System
10. R.D. Middlebrook, Input lter stability of full-regulated bus converter
considerations in design and in distributed power system, in Proc.
application of switching regulators, in IEEE International Telecommunications
Proc. IEEE Industry Applications Society Energy Conf., 2005, pp. 563568.
Annual Meeting, 1976, pp. 366382. 21. K. Zenger, A. Altowati, and T. Suntio,
11. R.D. Middlebrook, Design techniques Dynamic properties of interconnected
for preventing input-lter oscillations power systems A system theoretic
in switched-mode regulators, in Proc. approach, in Proc. IEEE Industrial
National Solid-State Power Conversion Electronics and Applications Conf.,
Conf. 1978, pp. A3.1A3.16. 2006, pp. 835840.
12. F.C. Lee and Y. Yu, Input-lter design 22. K. Zenger, A. Altowati, and T. Suntio,
for switching regulators, IEEE Trans. Stability and performance analysis of
Aerosp. Electron. Syst., vol. AES-15, no. regulated converter systems, in Proc.
5, 1979, pp. 627634. IEEE Industrial Electronics Society
13. S.Y. Erich and W.M. Polivka, Input Annual Conf., 2006, pp. 19751980.
lter design criteria for current- 23. P. Li and B. Lehman, Performance
programmed regulators, IEEE Trans. prediction of DCDC converters with
Power Electron., vol. 7, no. 1, 1992, pp. impedances as loads, IEEE Trans.
143151. Power Electron., vol. 19, no. 1, 2004, pp.
14. Y. Jang and R.W. Erickson, Physical 201209.
origin of input lter oscillations in 24. B. Choi, J. Kim, B.H. Cho, S. Choi,
current programmed converters, and C.M. Wildrick, Designing control
IEEE Trans. Power Electron., vol. 7, no. loop for DC-to-DC converters with
4, 1992, pp. 725733. unknown AC dynamics, IEEE Trans.
15. S. Ang and A. Oliva, Power-Switching Indust. Electron., vol. 49, no. 4, 2002,
Converters, Taylor & Francis, Boca pp. 925932.
Raton, FL, USA, 2005, 2nd Edition. 25. B. Choi, B.H. Cho, and S.-S. Hong,
16. T. Suntio, I. Gadoura, and K. Zenger, Dynamics and control of DC-to-DC
Input lter interactions in converters driving other converters
peak-current-mode controlled buck down stream, IEEE Trans. Circuits
converter operating in CICM, IEEE Syst. I: Fundam. Theory Appl., vol. 46,
Trans. Indust. Electron., vol. 49, no. 1, no. 10, 1999, pp. 12401248.
2002, pp. 7686. 26. D. Lee, B. Choi, J. Sun, and B.H. Cho,
17. C.M. Wildrick, F.C. Lee, B.H. Cho, Interpretation and prediction of loop
and B. Choi, A method of dening gain characteristics for switching
the load impedance specication for a power converters loaded with general
stable distributed power system, IEEE load subsystem, in Proc. IEEE Power
 References 259

Electronic Specialist. Conference, 2005, regulated converters to ensure stability

pp. 10241029. and performance in distributed power
27. P. Li and B. Lehman, Accurate loop supply system, in Proc. IEEE
gain prediction for DCDC converters International Telecommunications
due to impact of source/input lter, Energy Conf., 2005, pp. 533538.
IEEE Trans. Power Electron., vol. 20, 37. M. Karppanen, T. Suntio,
no. 4, 2005, pp. 7541017. and M. Sippola, Dynamical
28. C. Gezgin, W.C. Bowman, and V.J. characterization of
Thottuvelil, A stability assessment input-voltage-feedforward-controlled
tool for DC-DC converters, in Proc. buck converter, IEEE Trans. Indust.
IEEE Applied Power Electronics Conf., Electron., vol. 54, no. 2, 2007, pp.
2002, pp. 367373. 10051013.
29. M. Karppanen, T. Suntio, 38. M. Karppanen, M. Hankaniemi,
and M. Sippola, Impact of T. Suntio, and M. Sippola, Dynamical
output-voltage remote sensing on characterization of peak-current-mode
converter dynamics, Int. Rev. Electr. controlled buck converter with
Eng., vol. 2, no. 2, 2007, pp. 196202. output-current feedforward, IEEE
30. M. Hankaniemi, M. Karppanen, Trans. Power Electron., vol. 22, no. 2,
and T. Suntio, Load imposed 2007, pp. 444451.
instability and performance 39. L.D. Varga and N.A. Losic, Synthesis
degradation in a regulated converter, of zero-impedance converter, IEEE
IEE Proc. Electr. Power Appl., vol. 153, Trans. Power Electron., vol. 7, no. 1,
no. 6, 2006, pp. 781786. 1992, pp. 152170.
31. M. Karppanen, M. Sippola, 40. X. Wang, R. Yao, and F. Rao,
and T. Suntio, Source-imposed Subsystem-interaction restraint in the
instability and performance two-stage DC distributed power
degradation in a regulated converter, systems with
in Proc. IEEE Power Electronics decoupling-controlled-integration
Specialists Conf., 2007, pp. 194200. structure, IEEE Trans. Indust.
32. M. Hankaniemi, M. Karppanen, Electron., vol. 52, no. 6, 2005, pp.
T. Suntio, A. Altowati, and K. Zenger, 15551563.
Source-reected load interactions in a 41. P.G. Maranesi, V. Tavazzi,
regulated converter, in Proc. IEEE and V. Varoli, Two-port modeling of
Industrial Electronics Society Annual PWM voltage regulators at low
Conf., 2006, pp. 28932898. frequencies, IEEE Trans. Indust.
33. T. Suntio, K. Kostov, T. Tepsa, Electron., vol. 35, no. 3, 1988, pp.
and J. Kyyra, Using input invariance 444450.
as a method to facilitate system design 42. R.D. Middlebrook, Null double
DPS applications, J. Circuits, Syst. injection and the extra element
Comput., vol. 13, no. 4, 2004, pp. theorem, IEEE Trans. Edu., vol. 32,
707723. no. 3, 1989, pp. 167180.
34. T. Suntio, Input invariance as a 43. T. Suntio, and I. Gadoura, Use of
method to reduce EMI lter unterminated two-port modeling
interactions in telecom DPS systems, technique in analysis of input lter
in Proc. IEEE International interactions in telecom DPS systems,
Telecommunications Energy Conf., 2003, in Proc. IEEE International
pp. 592597. Telecommunications Energy Conf., 2002,
35. M. Karppanen, M. Hankaniemi, pp. 560565.
and T. Suntio, Load and supply 44. D.V. Otto, Reduction of switching
invariance in a regulated converter, in regulator audiosusceptibility to zero,
Proc. IEEE Power Electronics Specialist IEE Electron. Lett., vol. 22, no. 8, 1986,
Conference, 2006, pp. 26632668. pp. 441442.
36. M. Hankaniemi, T. Suntio, 45. J.P. Sjoroos, T. Suntio, K. Kostov,
and M. Sippola, Characterization of and J. Kyyra, Dynamic performance
260 8 Interconnected Systems

of the buck converter with input 47. G.K. Schoneman and D.M. Mitchell,
voltage feed-forward control, in Proc. Output impedance considerations for
European Power Electronics and switching regulators with current-
Applications Conf., 2005, paper no. 630, injected control, IEEE Trans. Power
pp. 110. Electron., vol. 4, no. 1, 1989, pp. 2535.
46. R. Redl, and N.O. Sokal, 48. S. Kanemaru, T. Hamada,
Near-optimum dynamic regulation of T. Nabeshima, T. Sato, and T. Nakano,
DCDC converters using feedforward Analysis and optimum design of a
of output current and input voltage buck-type DC-to-DC converter
with current-mode control, IEEE employing load current feedforward,
Trans. Power Electron., vol. PE-1, no. 3, in Proc. IEEE Power Electronic Specialist
1986, pp. 181191. Conf., 1998, pp. 309314.
 261

9
Control Design Issues

9.1
Introduction

The switched-mode converters are usually used to supply power for different
electronic loads, where the transient-performance requirements can be very
stringent due to low supply voltages and rapidly changing load currents as
in powering the microcomputers [13]. Such applications are usually based
on the use of the intermediate bus architecture (IBA) shown in Figure 9.1a
[4], where the converters operate as voltage-output converters. Although the
load-transient requirements are generally stringent, a part of the converters
(i.e., the bus converter) may be operating even without feedback from the
output voltage for reducing the costs of implementation.
In some applications, the system may incorporate elements, which would
require the use of output-current limiting to protect the converters from
damage as in Telecom uninterruptible power supply (UPS) systems shown in
Figure 9.1b [5]. In such a system, the rectiers providing the recharging of
the storage batteries have to be able to operate both as a voltage-output and
current-output converter requiring the use of multiloop-control arrangement
as depicted in Figure 9.2 [6]. The cascaded nature of the control as well as the
varying dynamic features of the voltage-output and current-output operations
would complicate the design of the required controllers as we have discussed
in Chapter 7.
Basically it is always the question of maintaining robust stability and
achieving adequate transient dynamics [720] within the certain constraints
stipulated by the application. In practice, the internal dynamic prole of the
converter would determine what the real transient dynamics would be in the
operational environment [21]. The practical dynamic behavior of the converter
would depend on the voltage and/or current-loop design but also on the other
factors related to the internal dynamics and external interactions [2233] as
we have discussed especially in Chapters 2 and 8. Of course, the fact is that a
poor control-loop design would yield poor performance but an excellent design
would not necessarily yield the opposite.

Dynamic Prole of Switched-Mode Converter. Teuvo Suntio

2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
ISBN: 978-3-527-40708-8
262 9 Control Design Issues

Figure 9.1 Typical distributed power architectures: (a) IBA system and (b) DC UPS
system.

Figure 9.2
Output-voltage-current
characteristics of a converter.

Excellent input and output transient performance can be achieved without

feedback control by providing proper feedforward from the input voltage and
output current as implied and demonstrated in [34]. Actually the feedforward
schemes were applied in [23] and [35] as introduced in Chapter 8, Section 8.3,
but the goals of the feedforward design were to obtain invariance to the input
 9.1 Introduction 263

and output interactions. Figure 9.3a shows the control block diagram for the
output dynamics including the feedforward gains (qi , qo ). The output voltage
can be solved to be



uo = Gioo + qi Ga Gco uin Zoo qo Ga Gco io (9.1)

which is zero when the feedforward gains are chosen to be

Gioo
qi =
Ga Gco
(9.2)
Zoo
qo =
Ga Gco
The corresponding control block diagram is presented in Figure 9.3b from
which we can solve the input current with the dened feedforward gains (9.2)
to be as follows:
   
Gioo Gci Zoo Gci
iin = Yino uin + Toio + io (9.3)
Gco Gco
Equation (9.3) can also be given as

iin = Yin uin + Toi io (9.4)

where Yin is the familiar ideal input admittance dened earlier in

Chapter 2 and discussed throughout the book. Toi is actually also a special
parameter ideal reverse transfer ratio not observed earlier but actually
characterizing the low frequency behavior of the closed-loop Toic as will be
shown later. In the case of the second-order converters, the ideal reverse
transfer ratio corresponds to M(D) (see Chapter 3, Section 3.4.3). Both of these
special parameters are independent of control mode, load, and state of output
feedback. They can be always solved according to (9.3), when the transfer
functions are known. When Zoo 0 then Toio = Toic = Toi , which will
change the propagation of the output-current noise into the input current
profoundly.
The feedback-amplier design is most often based on the use of proportional-
integral-derivative (PID) or proportional-integral (PI) controls [36, 37]. More
advanced linear and nonlinear control methods [3844] are available and
suited well for the multivariable control applications [41] especially when
digital controllers are in use [45]. Usually the PID-type controllers would
provide adequate dynamic performance and robustness of stability. In analog
control, the constraints of operational ampliers such as the limited gain-
bandwidth (GBW) product would usually dictate the maximum high-frequency
gain of the controller [20] in low-cost applications but the constraints can be
somewhat relaxed by using several discrete operational ampliers in cascade
as demonstrated in [10].
The control systems in the extremely low-cost and high-volume switched-
mode converters, for example in the mobile-phone-battery-charger applications
264 9 Control Design Issues

Figure 9.3 (a) Output dynamics and (b) input dynamics with feedforward gains.
 9.2 Feedback-Loop-Design Constraints 265

[55], are implemented by using a shunt regulator TL 431 [5362] designed

originally for providing a stable voltage reference [56]. The overall low cost has
made the shunt regulator a popular building block of a simple control system.
Its dynamic properties are not usually what the designer assumes [58, 60].
Therefore, the dynamics and stability of the associated converters may not be
satisfactory [59].
The intention of the chapter is not to provide comprehensive control-design
methods but to highlight the constraints the design would incorporate as well
as to introduce an effective way to design the required loop characteristics.
In addition, we will discuss the problems associated with the simple control
systems based on the use of shunt regulators. The voltage-output converters
are treated only. The same methods and constraints also apply to the current-
output converters treated in Chapter 7. The comprehensive list of the references
would provide hints for further reading if more thorough knowledge and more
advanced control methods are of interest.

9.2
Feedback-Loop-Design Constraints

An optimum feedback-loop design according to [8] is the one with the highest
gain below the loop crossover frequency and lowest gain above the crossover
frequency. According to [9], the optimal controller would be quickly responsive
to substantial changes in output voltage and also provide precise steady-state
control. These properties are also inherent in the denition provided by [8].
The rst denition of the optimum design [8] is actually in line with the
requirements of robust stability and fast transient response even if the external
interactions were quite unknown during the writing of [8]: The optimum
design might be expressed as illustrated in Figure 9.4, where the loop gain is
as high as possible up to the loop crossover and as small as possible after the
loop crossover with constant phase margin for all the frequencies.
The reasoning behind the optimality is the fact that such a converter would
be invariant to load and would have fast transient response: The high-loop
gain means that the closed-loop output impedance is extremely low up to the
loop crossover frequency, which means that the load impedance would not
easily affect the internal dynamics [22] and the transient setup time is very
short [2933]. In practice, the load transient is dictated by the properties of
the output capacitors and the other parasitic circuit elements at the output
of the converter and along the path between the load and the converter.
The load impedance does not affect the input impedance of the converter
due to the extremely high attenuation capability of the closed-loop input-to-
output transfer function [22]. Therefore, the stability of the converter due to
the source interactions would be dictated by the ideal input impedance in
such a way that no impedance overlap is tolerated at the direct input of the
converter.
266 9 Control Design Issues

Figure 9.4 Optimal loop shape.

The reality is, however, that such a loop behavior cannot be accomplished
due to the practical constraints involved in the components such as operational
ampliers, and A/D converters as well as to the duty-ratio generation.
The same constraints would also dictate that the internal disturbance-input
transfer functions (i.e., Zoo , Gioo ) would actually dominate the transient
responses [4852] as illustrated in [22] and, therefore, the controllers cannot
be tuned based on the time-domain transient responses even if stated so for
example in [1517].

9.2.1
Phase and Gain Margins

The output-voltage transient (
uo ) can be mathematically presented in

frequency domain [2932] by

uo (s) = Zoc
io (s) (9.5)

where Zoc is the closed-loop output impedance of the converter and

io (s) the Laplace transformation of the excitation signal (i.e., the shape
of output current change). The time-domain responses can be naturally
computed by means of (9.5). Similar formulation can also be derived for the
input-voltage transient behavior. The value of the output impedance at the
voltage-loop crossover frequency corresponds to the amount of undershoot
in the output-voltage transient [29]. This means that a low value of phase
 9.2 Feedback-Loop-Design Constraints 267

Figure 9.5 The peak value of the sensitivity function (a) at the
gain crossover frequency (fgco ) as a function of phase margin and
(b) at the phase crossover frequency (fphco ) as a function of gain
margin.

(PM) or gain (GM) margin would increase the undershoot by causing peaking
in the output impedance or the input-to-output transfer function due to the
sensitivity function S (i.e., Zoc = Zoo S, Gioc = Gioo S; S = (1 + L(s))1 ):
The magnitude of the sensitivity
function can be given at the loop crossover
frequency by |S||L|=1 = 1/ 2(1 cos(PM)) and at the phase crossover
frequency by |S|=180 = |GM/(GM 1)|). Therefore, the output impedance
or the input-to-output transfer function would experience peaking if PM < 60
268 9 Control Design Issues

(see Figure 9.5a) or GM < 18 dB (see Figure 9.5b). The typically used gain
margin of 6 dB would actually cause the peaking of 6 dB. As a consequence,
the loop-gain related margins should be maintained adequate for preventing
the excess peaking and the deterioration of transient performance.

9.2.2
RHP Zeros and Poles

The boost and buckboost converters would exhibit nonminimum-phase

behavior, which means that their control-to-output transfer functions contain
right-half-plane (RHP) zero. This also applies for the fourth-order and higher
order converters. The existence of the RHP zero means actually [46] that
the maximum control bandwidth has to be limited to the frequency of the
RHP zero and usually below it. The situation is worst in the CCM converters
where the RHP zero locates at low frequencies and thereby also effectively
deteriorates the corresponding transient performance. The control-bandwidth
limitation cannot be removed even if such claims have been presented in
[47]. In the DCM converters, the RHP zero typically locates at a rather high
frequencies and does not deteriorate the transient performance similarly to
the CCM mode but it would limit the maximum achievable control bandwidth.
Sometimes, the open-loop converter may even contain an unstable pole (i.e.,
RHP pole) or the load may impose it. This means that the control bandwidth has
to be designed to be higher than the RHP pole in order to maintain stability
[46]. An illustrative example is the PCM-controlled superboost converter
(Figure 9.6a) [63] sometimes used in the space applications. Such a fourth-
order converter contains usually a resonant RHP zero as shown in Figure 9.6b
(solid line) in addition to the resonant LHP pole. It may be obvious that
the control bandwidth has to be designed shorter than the RHP zero. When
the input voltage is further decreased, the resonant LHP pole moves to the
RHP (dashed-line) in such a way that it locates at the higher frequency than
the RHP zero. As a consequence, the converter cannot be controlled but the
duty ratio has to be limited in order that the described situation does not
activate. There are also methods to design the converter in such a way that
the appearance of the RHP pole is shifted to higher duty ratios. Figure 9.6c
shows the output-voltage waveform when the RHP pole is present at open
loop, proving that the unstable pole really exists.

9.2.3
Minimum and Maximum Loop Crossover Frequencies

In practice, the usable loop crossover frequencies (fgco ) without the limitations
caused by the RHP zeros and poles are dened according to [18] to be as
follows:
 9.2 Feedback-Loop-Design Constraints 269

1. fgco > 3fn , where fn is the resonant frequency of the converter: The
resonant pole will cause ringing and the control has to eliminate this
ringing. For doing so the control has to have adequate gain.
f
2. fgco < 5s , where fs is the switching frequency: The high gain at the high
frequencies would also amplify noise, which may affect the pulsewidth
generation and lead to instability (i.e., ripple effects, see e.g., [19]).

Figure 9.6 PCM-controlled superboost converter: (a) schematics, (b) control-to-output

transfer function, and (c) output-voltage waveform due to the unstable pole at open loop.
270 9 Control Design Issues

Figure 9.6 (Continued).

3. fgco < fCrC , where fCrC is the zero caused by the output capacitor: The
output capacitor usually dominates the high-frequency transient
behavior and, therefore, there is no benet raising the crossover
frequency above the output-capacitor ESR frequency.

9.2.4
Internal Gain of an Operational Amplier

The internal gain of the operational amplier (op amp) [18, 20] may also limit
the achievable crossover frequency due to limiting the gain and phase of the
controller. Typical transfer function of an op amp with the gain-bandwidth
(GBW) product of 1.5 MHz is shown in Figure 9.7.
If the gain of the controller exceeds the internal gain of the op amp, the
controller gain and phase would follow the gain and phase of the op amp with
surprising result. This can be easily concluded from (9.6), where the open-loop
gain of the op amp is denoted by GOPA and the ideal gain of the controller by
Zf /Zin (see Figure 9.8b).
Zf GOPA
GOPA
cc = (9.6)
Zin Zf Zf
1+ + GOPA +
Zin Rb
At low frequencies, where the controller gain is usually very high due to
the integral-control action, the phase of the actual controller would not start
from 90 but from zero. This would sometimes remove the conditional-
stability condition to exist even if the theory predicts that to exist.
 9.3 Controller Implementations 271

Figure 9.7 The open-loop gain of an op amp with 1.5-MHz GBW-product.

Figure 9.8 Simple analog PID controller: (a) detailed circuitry and (b) general circuitry.

9.3
Controller Implementations

The most common feedback-loop controller is a modied proportional-integral-

derivative (PID) controller [8, 17, 36, 37], which usually yields sufcient
dynamic properties. The simple analog solution of such a controller also known
as type-3 controller [8] is shown in Figure 9.8a. The resistor Rb contributes
only to the steady-state value of the output voltage when the reference is not
used for control purposes (i.e., ur ). The corresponding transfer function (i.e.,
uco /uo ) can be given by


1 + s Rf Cf (1 + s (Rin + Rd )Cd )
Gcc =   (9.7)

 R f Cf Cp
Rin Cf + Cp s (1 + s Rd Cd ) 1 + s
C f + Cp
272 9 Control Design Issues

where Cf is typically much larger than Cp . As a consequence, the transfer

function simplies to


1 + s Rf Cf (1 + s (Rin + Rd ) Cd )
Gcc =
 (9.8)
Rin Cf s (1 + s Rd Cd ) 1 + s Rf Cp
The PI-type or type-2 controller can be obtained by removing Rd and Cd from
Figure 9.8a. The corresponding transfer function can be obtained from (9.7)
by setting them to zero.
The transfer function from the reference to the control voltage (i.e., uco /ur )
can be given by
 
Zin
Gccr = 1 + 1 + Gcc (9.9)
Rb
where Zin is dened in Figure 9.6b and Gcc in (9.7). As a consequence, the
controller can be dened by
 

 Zf
uco = Gcc ur uo + 1 + ur (9.10)
Rb
where Zf is dened in Figure 9.7b. The reference is usually embedded in the
corresponding PWM microcircuit and, therefore, it is not available and does
not contribute to the controller behavior. If the controller is implemented by
using discrete components then the reference section should be noise free due
to the amplifying effects of Gccr causing easily unexpected noise problems.
For the control-design purposes, the controller transfer function (Gcc ) is
most convenient to be transformed into the form
K (1 + s/z1 ) (1 + s/z2 )
Gcc =

 (9.11)
s 1 + s/p1 1 + s/p2
The contributions of the gain factor (K), the different zeros (zi ) and poles
(pi ) to the physical circuitry in Figure 9.8a can be easily deduced comparing
(9.7) and (9.11) to each other. The typical frequency responses of the PID
and PI controllers are shown in Figures 9.9 and 10, respectively, where fzi and
fpi are the frequencies in Hz corresponding to the zeros and poles given in
angular frequencies (rad/s) in (9.11).

9.4
Optocoupler Isolation

The feedback loop of the transformer isolated converters has to be also

isolated. The most common isolating component is an optocoupler [53, 54].
The optocoupler would be always a critical component in the loop, because its
 9.4 Optocoupler Isolation 273

Figure 9.9 The frequency response of a PID controller.

Figure 9.10 The frequency response of a PI controller.

274 9 Control Design Issues

current transfer ratio (CTR) can vary substantially depending on temperature,

the level of diode current, the type and manufacturer, it may have a delay,
and an RC-circuit-like behavior highly dependent on the collector/emitter
resistance. The transfer function describing its dynamical behavior [55] can be
given by
 
td
1s
2
Gopto CTR   (9.12)
td
(1 + s/3 dB ) 1 + s
2
where the effect of the delay (td ) is approximated by using the rst-order
Pade approximation of estd . In theory, the delay would only increase the


phase lag by td but does not affect the magnitude (i.e., estd =1 td ).
The cut-off frequency (3 dB ) can be found from frequency responses
typically given by the manufacturer at different collector resistors or
from the time-domain step response given at a certain collector resistor
(Rc1 ): The rise time (tr ) and the corresponding cut-off frequency are
related by 3 dB 2.2/tr , from which the equivalent Copto tr /2.2Rc1 . The
corresponding 3 dB at the used collector resistor (Rc2 ) equals Rc1 /Rc2 2.2/tr ,
respectively.

9.5
Shunt-Regulator-Based Control Systems

The control system of a low-cost converter is usually built around the shunt
regulator TL431 developed in the late 1970s for providing stable reference
voltage [56]. In principle, the shunt regulator is an excellent device for
implementing a simple control system due to its ability to drive rather
large currents [5361] when the optocoupler is needed in the control loop as
illustrated in Figure 9.11.

9.5.1
Dynamic Model

The shunt regulator is, however, quite a complicated device dynamically and
its properties are not usually understood [58, 60]. The rst mistake is to
assume that the shunt regulator provides the same properties as an op amp
[60, 62]. The reality is, however, that it is a transconductance amplier (i.e.,
voltage-to-current amplier) having current as an output signal and no negative
feedback present as shown in Figure 9.12a [59]. The corresponding dynamic
equivalent circuit is shown in Figure 9.12b. The op-amp-like features are valid
only if the transconductance gain (gm ) and the output impedance (Zo ) are high
(Figure 9.12b).
 9.5 Shunt-Regulator-Based Control Systems 275

Figure 9.11 Application of TL431 in controlling self-oscillating yback converter with

(a) combined feedback loops and (b) separate feedback loops.

The shunt regulator also provides two feedback loops (Figure 9.12c) one via
the reference input (slow loop) and the other via the cathode (fast loop) [57]. The
loops may be connected together (Figure 9.11a) or separately (Figure 9.11b).
In the case of Figure 9.11b, the overall loop gain is difcult to measure and
analyze. Therefore, the effect of the fast loop in those cases is most often just
forgotten but may have profound effect on the loop behavior.
The dynamic properties of TL431 are usually dened by means of the slow-
loop open-loop frequency response at the bias current of 10 mA by using the
measurement setup shown in Figure 9.13a and based on the cathode voltage
(i.e., uca /uRef ) [56]. The DC specications are stated to be valid down to the
collector current of 1 mA. The open-loop output impedance is not usually
shown in the manufacturers specications but can be measured using the
measurement setup shown in Figure 9.13b.
It should be noted that the open-loop response incorporates the effects of
the transconductance gain (gm ) and the output impedance (Zo ) as follows:
uca Rd Zo
= gm (9.13)
uRef Rd + Zo
The real output signal is, however, the cathode current (ica ) and, therefore, the
desired transfer function can be given by
ica Zo
= gm (9.14)
uRef Rd + Zo
276 9 Control Design Issues

Figure 9.12 Shunt regulator TL431: (a) high-level equivalent circuit, (b) dynamical
equivalent circuit, and (c) the two feedback loops.

Figure 9.13 The frequency-response measurement setups for (a) open-loop gain and
(b) open-loop output impedance.

which shows the inuence of different circuit elements as well as the existence
of the positive feedback. It is obvious that the transfer functions in (9.13) and
(9.14) equals Rd gm and gm , when Zo >> Rd . Typically gm is assumed to be
constant and Zo a capacitor [58, 59], which can be estimated from the given
 9.5 Shunt-Regulator-Based Control Systems 277

frequency responses: the low-frequency gain corresponds to Rd gm and the

cut-off frequency to 1/2Rd Co .
The fast and slow-loop transfer functions can be given from Figure 9.12c as
follows:

iopto 1
=
 Zin + Zf + Zo
uof Zin + Zf (Rd + Zo ) + Rd Zo

 
Zin Zo2 Zin + Zf gm
+

 (9.15)
Zin + Zf (Rd + Zo ) + Rd Zo 1 + Zin gm


iopto Ge Zo Zf gm 1
=


uos Zin + Zf (Rd + Zo ) + Rd Zo 1 + Zin gm

where the subscript extensions f and s denote fast and slow, respectively, and
the termination resistor Rd contains the effect of the corresponding discrete
resistor and the dynamical resistance of the diode, which may be large at the
low diode current. It is also assumed that the bias resistor (Rb ) is much larger
than Rd . If gm >> 1 then (9.15) becomes
iopto 1
=
uof Rd
(9.16)
iopto Ge Zf
=
uos Rd Zin

which is the form typically used in the analyses as, for example, in [60] except
that there is clearly no negative feedback. If we consider the transfer functions
in (9.16) with respect to the cathode voltage (uca ) then the negative sign would
appear but the cathode voltage is not the signal transforming the control
information through the optocoupler.
The dynamic parameters of the shunt regulator (i.e., gm and Zo ) can be
extracted by measuring the open-loop
 response (4) from Figure 9.13a, and the
output impedance Zo = uca
from Figure 9.13b: The transconductance (gm )
ica
can be computationally solved by means of the measured responses from
 
Rd ica
gm = 1 + (9.17)
Zo uRef
A certain shunt regulator was analyzed experimentally by measuring its open-
loop gain (ica /uRef ) and the output impedance (Zo ) as well as extracting the
transconductance gain computationally from the measured responses from
(9.17).
It was noticed that the shunt regulator does not function dynamically properly
at the specied minimum cathode current of 1 mA: The measured cathode
voltage (uca ) is shown in Figure 9.14 during the measurement of the open-loop
gain indicating clearly that the cathode current of 1 mA is not sufcient for
the proper dynamic operation. The proper operation was recovered when the
278 9 Control Design Issues

Figure 9.14 The corrupted cathode-voltage response at the cathode current of 1 mA.

cathode current was increased to 2 mA although the excitation level had to be

kept rather low compared to the excitation at the cathode current of 10 mA in
order to avoid the corrupted response.
The measured open-loop gain and the output impedance at the cathode
current of 10 mA (solid line) and 2 mA (dashed line) are shown in
Figures 9.15 and 16, respectively. Figure 9.16 shows that the output impedance
corresponds to a capacitor, which is expected due to the stray capacitances
in the microelectronic circuit (Figure 9.12a) [56]. The level of cathode current
affects the ESR of the capacitor. The high-frequency responses denes the
value of the ESR (Ro ) and the corresponding value of the capacitor (Co ) can be
solved from the corner frequency (f3 dB ) by Co = 2f31 dB Ro .
The transconductance gain (gm ) was computationally solved by means of the
responses in Figures 9.15 and 16. The corresponding responses are shown in
Figure 9.17 indicating that the high-frequency gain depends on the level of
cathode current.
The shape of gm implies that it can be modeled by means of a second-order
transfer function as
gmo (1 + s/z1 )
gm 2 (9.18)
s + s 2 n + n2
where gmo is the low-frequency transconductance, the damping factor, and
n the undamped natural frequency. At the low cathode current, the resonant
behavior is well damped. The zero (z1 ) can be found at the frequency where
 9.5 Shunt-Regulator-Based Control Systems 279

Figure 9.15 The open-loop gain (ica /uRef ) at the cathode current of 10 mA (solid line)
and 2 mA (dashed line).

Figure 9.16 The open-loop output impedance (Zo ) at the cathode current of 10 mA (solid
line) and 2 mA (dashed line).

the phase is close to 45 and the resonant pole (n ) at the frequency where the
phase is close to 0 .
The actual behavior of such a control system shown in Figure 9.18 was
measured by injecting the excitation signal into both of the loops (see
Figure 9.12). The bias current was adjusted by means of the bias resistor
(Rb ) to 2 and 10 mA. The corresponding responses are shown in Figure 9.19.
280 9 Control Design Issues

Figure 9.17 The computed transconductance gain (gm ) at the cathode current of 10 mA
(solid line) and 2 mA (dashed line).

Figure 9.18 A yback converter with the TL-431-based control system.

The total response of the control system is the sum of the slow- and
fast-loop responses [59] dened in (9.15). The slow-loop response dominates
at the low frequencies and the fast-loop response at the high frequencies.
From Figure 9.18, the slow loop forms basically an integral controller where
the maximum gain is dened by the transconductance gain. Therefore, it
is natural that the low-frequency gain of the control system is very low due
to the behavior of gm shown in Figure 9.17. It may be obvious that the
steady-state voltage accuracy can be poor. The high-frequency response should
correspond to 1/Rd (i.e., 20 dB) but is actually lower. The reason may be
that the dynamic resistance of the optocoupler diode increases the value
of Rd .
The models shown may not be generalized because the properties of various
shunt regulators are known to vary signicantly [59]. The methods to extract
the models are naturally generally applicable. The important message is
 9.5 Shunt-Regulator-Based Control Systems 281

Figure 9.19 The control-system frequency responses at the cathode current of 10 mA

(solid line) and 2 mA (dashed line).

that the properties of the shunt regulator and the control systems based
on them are not what usually are assumed based on the information the
open literature has provided. In addition, the minimum bias current should
be higher than the specied minimum, and preferably close to 10 mA.
If the overall current consumption has to be kept low then the shunt
regulators having much lower minimum currents should be used as discussed
in [61].

9.5.2
Two-Loop Control System

The slow and fast loops of the shunt regulator can be connected into the output
of the converter with several ways:

1. Both of the loops are connected together at the direct output of the
converter as illustrated in Figure 9.11a. The overall output-voltage loop
gain of the converter can be easily measured injecting the excitation
signal into both of the loops simultaneously. The overall control-system
transfer function is the sum of the slow- and fast-loop transfer
functions dened in (9.15).
2. The fast loop may be connected directly to an auxiliary voltage supply
and the slow loop at the output of the converter as shown in [58]. The
fast loop does not contribute dynamically and, therefore, the overall
control-system transfer function is the slow-loop transfer function
dened in (9.15).
282 9 Control Design Issues

3. The converters may be equipped with an extra output lter. It is

customary to connect the fast loop before the lter and the slow loop
directly at the output of the converter as illustrated in Figure 9.11b. As a
consequence, the voltage-loop gain is difcult to measure by means of a
single measurement but it can be found by using the methods dened
below [57].

In the case of separately connected fast and slow loops, the overall
voltage-loop gain can be found by measuring separately the responses of
those loops when the other loop is connected. If we denote the measured fast
f (s) and the measured slow loop by Ls (s) then the overall voltage-loop
loop by LM M

gain (L(s)) can be given by

f (s) + Ls (s) + 2Lf (s)Ls (s)

LM M M M
L(s) = (9.19)
1 LM
f (s)Ls (s)
M

where LM f (s) and Ls (s) can be given as a function of the original fast (Lf (s))
M

and slow (Ls (s)) loops, respectively, by

Lf (s)
f =
LM
1 + Ls (s)
Ls (s) (9.20)
Ls =
M
1 + Lf (s)
The validity of (9.19) assumes that the measured loops are given extracting
the effect of negative feedback as is usual in control engineering. If the
measurement data are the authentic data from the frequency-response analyzer
(i.e., the phase margin is usually observable with respect to zero) then the sign
of the product of the loop gains in the numerator has to be changed to minus
for obtaining correct overall loop gain [57].
In order to analyze the dynamics of the two-loop system shown in
Figure 9.11b, the corresponding transfer functions containing the effect of the
extra lter have to be determined: Those transfer functions can be found by
utilizing the linear two-port model of the converter shown in Figure 9.20, where
the open-loop parameters constitute the G-parameter set dened in (9.21).


 uin 
iin Yino Toio Gci
= io1 (9.21)
uo Gioo Zoo Gco
c

Figure 9.20 Two-port model of a

converter with the added secondary
lter.
 9.5 Shunt-Regulator-Based Control Systems 283

The necessary transfer functions can be solved from Figure 9.20 by means of
pure circuit theory yielding the control-block diagram shown in Figure 9.21,
which represents the output dynamics of the converter. The shunt-regulator-
based control system is also added into the block diagram, where Gccf
and Gccs are the fast and slow-loop transfer functions dened in (9.15),
respectively, ZLF and ZCF the inductive and capacitive impedances of the lter,
respectively, and ZSF = ZLF + ZCF , Gai the combined transfer function of
the other dynamical elements along the path from the shunt regulator to
the control variable (c) as well as Gioo , Zoo , and Gco the internal transfer
functions of the original converter dened in the bottom row of (9.21).
From Figure 9.21, the fast (Lf (s)) and slow (Ls (s)) loops can be, respectively,
given by
Gco
Lf (s) = Gccf Gai
Zoo
1+
ZSF (9.22)
Gco ZCF
Ls (s) = Gccs Gai
Zoo ZSF
1+
ZSF
and the overall loop gain (L(s)) by
L(s) = Lf (s) + Ls (s) (9.23)
This means that the closed-loop input-to-output transfer function (Gioc ) and
the closed-loop output impedance (Zoc ) can be given by
ZCF
Zoo + ZSF
Gioc = Gioo
1 + Lf (s) + Ls (s) (9.24)
(Zoo + ZLF ) ZCF
Zoc =

(Zoo + ZSF ) 1 + Lf (s) + Ls (s)

Figure 9.21 The control-block diagram for the overall system representing the output
dynamics.
284 9 Control Design Issues

It may be obvious that the two-port model of a converter is very useful tool in
the analysis of the complicated feedback arrangements.

9.6
Simple Control-Design Method

The internal open-loop dynamics of the voltage-output converter can be

represented by means of a set of G-parameters by


 uin 
iin Yino Toio Gci
= io (9.25)
uo Gioo Zoo Gco
c
where the upper row represents the input dynamics and the bottom row the
output dynamics. The matrix form can be equally given as the linear two-port
model shown in Figure 9.22.
The closed-loop dynamics can be most conveniently found by utilizing the
control block diagrams shown in Figure 9.23. The voltage-loop gain (L(s)) is
dened from the variables shown in Figure 9.23a by
L(s) = Gse Gcc Ga Gco (9.26)
where Gse is the output-voltage sensing gain, Gcc the controller transfer
function, Ga the transfer function of all the other dynamical elements within
the loop from the controller to the control input (c), and the control-to-output
transfer function (Gco ). The inputs where the input variables (uin , io ) are
connected (Figure 9.23a) are known as disturbance inputs.
The closed-loop dynamic representation can be solved from Figure 9.23 to
be as follows:

iin
=
uo
Yino L(s) Toio L(s) Gci L(s)
+ Yin + Toi
1 + L(s) 1 + L(s) 1 + L(s) 1 + L(s) Gse Gco 1 + L(s)

Gioo Zoo 1 L(s)

1 + L(s) 1 + L(s) Gse 1 + L(s)
 
uin
io (9.27)
ur

Figure 9.22 Two-port model of a

voltage-output converter.
 9.6 Simple Control-Design Method 285

Figure 9.23 Control block diagrams representing closed-loop (a) output dynamics and
(b) input dynamics.

where Yin is the ideal input admittance (i.e., Yino Gioo Gci /Gco )
and Toi is the ideal reverse transfer ratio (i.e., Toio + Zoo Gci
Gco
). The
goal of the control design is to obtain robust stability in terms of
the phase (PM) and gain (GM) margins as well as adequate transient
Zoo
performance associated with the load-current (i.e., 1+L(s) ) and input-
Gioo
voltage (i.e., 1+L(s) ) disturbances as discussed earlier in Section 9.2. The
control design examples in the subsequent subsections are given based
on the buck and boost converters dened in Figure 9.24 and operating in
CCM.

9.6.1
Control Design Example: VMC Buck Converter

The buck converter under VM control is shown in Figure 9.25 and the
associated circuit elements are dened in Figure 9.24a. The dynamic modeling
and characterization of the VMC buck converter is previously presented in
Chapter 3. The goal of the control design is to limit the maximum loop
crossover frequency to 10 kHz and obtain the phase margin of 60 . The
resonant nature of the converter requires the use of PID controller in order to
obtain enough phase boost for the stable operation.
The output-voltage loop gain (L(s)) can be given by
1
L(s) = Gcc Gco (9.28)
VM
where VM is the peak-to-peak voltage of the PWM ramp (Figure 9.25), Gcc
the error-amplier transfer function, and Gco the control-to-output transfer
function of the buck converter. It is obvious that VM (3 V) and Gco (Chapter 3)
are known, and we have to choose Gcc such that the goals are met. Therefore,
286 9 Control Design Issues

Figure 9.24 The schematics

of (a) buck converter and
(b) boost converter.

Figure 9.25 VM-controlled buck converter.

the loop gain (9.28) can be given by

1 K (1 + s/z1 ) (1 + s/z2 )
L(s) =


3 s 1 + s/p1 1 + s/p2
(Uin + UD + (rd rds1 ) IL ) (1 + srC C)
LC (9.29)
rL + Drds1 + D rd + rC 1
s2 + s +
L LC
Gco of the VMC buck converter has the highest gain at the maximum input
voltage, which is 50 V. Therefore, we rst plot the frequency response of the
known part of the loop gain at the input voltage of 50 V as shown in Figure 9.26
(solid line). The next task is to shape the phase of the loop such that the phase
 9.6 Simple Control-Design Method 287

Figure 9.26 The rst phase of the control design: VMC buck converter.

margin of 60 is obtained. This actually means that we assume K = 1. The

maximum phase boost is obtained if the zeros (zi ) are placed at the same
frequency. The resonant nature of the converter (fn 900 Hz) requires that
the zeros have to be placed before the resonant frequency. It is usually advised
that the last pole (p2 ) should be placed to compensate the ESR zero of the
output capacitor but due to the high uncertainty of the value of the zero the
last pole is recommended to be placed at half the switching frequency or
closer. The frequency response of the controller (K = 1) with the zero-pole
placements dened in (9.30) is shown in Figure 9.26 (dashed line).
1
z1,2 = 0.5
LC
p1 = 30 z1 (9.30)
2fs
p2 =
2
In the next phase, the voltage-loop gain with K = 1 should be plotted as
shown in Figure 9.27. According to it, the gain of the controller shall be
increased by 61 dB in order to obtain the crossover frequency of 10 kHz,
which gives K = 1122. The resulting predicted loop gains at the input voltages
of 50 V (solid line) and 20 V (dashed line) are shown in Figure 9.28. The
reduction of the crossover frequency due to the decrease of the input voltage
is typical to the voltage-mode control. The crossover-frequency reduction
may be eliminated by using several cascaded controllers as in [10] to
make the slope of the loop gain steeper in the vicinity of the crossover
frequency.
The same design procedures were also applied to the experimental converter
shown in Figure 9.24a. The measured output-voltage loop gains are shown
288 9 Control Design Issues

Figure 9.27 The second phase of the control design: VMC buck converter.

Figure 9.28 The predicted voltage-loop gains of the VMC buck converter at the input
voltages of 50 V (solid line) and 20 V (dashed line).

in Figure 9.29. The reduction of the phase margins from 60 is caused by

the higher phase lag in the real converter compared to the predictions. More
accurate design would have been obtained if the measured control-to-output
transfer function had been used as the base of the design.
The open-loop (dashed line) and closed-loop (solid line) output impedances of
the practical converter at the input voltage of 50 V are shown in Figure 9.30. The
magnitude of the closed-loop output impedance at the voltage-loop crossover
 9.6 Simple Control-Design Method 289

Figure 9.29 The measured output-voltage loop gains at the input voltages of 50 V
(solid line) and 20 V (dashed line).

Figure 9.30 The measured open-loop (dashed line) and closed-loop (solid line) output
impedances at the input voltage of 50 V.

frequency (11 kHz) is approximately 80 m. The converter was subjected to a

constant-current-type load change with the slew rate of 250 mA/s from 0.2
to 2.5 A. The corresponding output-voltage response is shown in Figure 9.31
where the voltage dip of 170 mV corresponds well with the magnitude of the
load change and the magnitude of the closed-loop output impedance at the
290 9 Control Design Issues

voltage-loop crossover frequency. The setup time is also rather quick due to
the small output impedance at the low frequencies.

9.6.2
Control Design Example: PCMC Buck Converter

The buck converter under PCM control is shown in Figure 9.32 and the
associated circuit elements are dened in Figure 9.24a. The equivalent
inductor-current sensing resistor (Rs ) is 75 m. The dynamic modeling
and characterization of the PCMC buck converter is previously presented
in Chapter 4. The goal of the control design is to limit the maximum
voltage-loop crossover frequency to 10 kHz and obtain the phase margin
of 60 . The resonant-free nature of the converter allows the use of PI
controller. The PWM modulator used in the experimental converter is
UCC 3842, where the error-amplier section contains an extra gain
of 1/3.
The output-voltage loop gain (L(s)) can be given by
1
L(s) = Gcc Gco (9.31)
3Rs
where Rs (Figure 9.32) is the equivalent inductor-current sensing resistor, Gcc
the error-amplier transfer function, and Gco the control-to-output transfer
function of the PCMC buck converter. It is obvious that Rs (75 m) and Gco
(Chapter 4) are known and we have to choose Gcc such that the goals are met.

Figure 9.31 The output-voltage response to a constant-current step change with slew
rate of 250 mA/s at the input voltage of 50 V.
 9.6 Simple Control-Design Method 291

Figure 9.32 The PCM-controlled buck converter.

Therefore, the loop gain (9.31) can be given by

1 K (1 + s/z1 )
L(s) =

3 75 103 s 1 + s/p2
Fm UE (1 + srC C)
LC
rE + Fm qc UE + rC 1
s2 + s +
L LC
UE = Uin + UD + (rd rds1 ) IL , rE = rL + Drds1 + D rd (9.32)
The converter is supposed to be compensated such that Mc 1.45 Uo /2L
(see Chapter 4, Section 4.5.1). Gco of the PCMC buck converter has the highest
phase lag at the highest input voltage but its magnitude stays virtually intact
when the input voltage changes as demonstrated in Chapter 4. Therefore, we
plot the known part of the loop gain as shown in Figure 9.33 (solid line) at the
input voltage of 50 V. The next task is to shape the phase of the loop such that
the phase margin of 60 is obtained. Similarly to the VMC converter the zero
(z1 ) can be placed before the virtual resonant frequency (fn 900 Hz) and
the last pole (p2 ) to obtain the phase margin of 60 . The frequency response
of the controller (K = 1) with the zero-pole placements dened in (9.33) is
shown in Figure 9.34 (dashed line) indicating that the phase margin would
be 68 . In order to obtain 60 , the last pole should be placed at slightly lower
frequencies than dened in (9.33). In the practical converter, the phase margin
would be lower than predicted and, therefore, we do not make the correction.

1
z1 = 0.5
LC
(9.33)
2fs
p2 =
8
292 9 Control Design Issues

Figure 9.33 The rst phase of the control design: PCMC buck converter.

Figure 9.34 The second phase of the control design: PCMC buck converter.

In the next phase, the voltage-loop gain with K = 1 should be plotted as

shown in Figure 9.34. According to it, the gain of the controller shall be
increased by 83 dB to obtain the crossover frequency of 10 kHz, which gives
K = 14125. The resulted predicted loop gains at the input voltages of 50 V
(solid line) and 20 V (dashed line) are shown in Figure 9.35 indicating that the
magnitude of the voltage-loop gain does not change, when the input voltage
changes.
 9.6 Simple Control-Design Method 293

Figure 9.35 The predicted voltage-loop gains at the input voltages of 50 V (solid line)
and 20 V (dashed line).

Figure 9.36 The measured voltage-loop gain of the PCMC buck converter at the input
voltage of 20 V.

The similar control-design procedures were applied to the experimental

converter. The measured voltage-loop gain at the input voltage of 20 V is shown
in Figure 9.36. The reduction of the phase margin from the predicted phase
margin is caused by the higher phase lag in the real converter compared to the
predictions. More accurate design would have been obtained if the measured
control-to-output transfer function had been used as a base of the design.
The open-loop (dashed line) and closed-loop (solid line) output impedances
at the input voltage of 50 V are shown in Figure 9.37. The magnitude of
294 9 Control Design Issues

Figure 9.37 The measured open-loop (dashed line) and closed-loop (solid line) output
impedances at the input voltage of 50 V.

Figure 9.38 The output-voltage response to a constant-current-type step load change

from 0. 2 to 2.5 A with the slew rate of 250 mA/s at the input voltage of 50 V.

the closed-loop output impedance at the voltage-loop crossover frequency

(13 kHz) is approximately 80 m. The converter was subjected to a constant-
current-type load change from 0.2 to 2.5 A with the slew rate of 250 mA/s.
The corresponding output-voltage response is shown in Figure 9.38, where
the voltage dip of 180 mV corresponds well to the magnitude of the load
change and the magnitude of the closed-loop output impedance at the voltage-
loop crossover frequency. The setup time is rather sluggish compared to the
 9.6 Simple Control-Design Method 295

response of the VMC converter (Section 9.6.1, Figure 9.31) due to the larger
output impedance at the low frequencies.

9.6.3
Control Design Example: VMC Boost Converter

The boost converter under VM control is shown in Figure 9.39 and the
associated circuit elements are dened in Figure 9.24b. The dynamic modeling
and characterization of the VMC boost converter is previously presented in
Chapter 3. Gco of the VMC boost converter contains a RHP zero, which
automatically limits the achievable control bandwidth to the frequency of the
RHP zero or below it. The goal of the control design is to maximize the
crossover frequency with a satisfactory phase margin. The resonant nature of
the converter requires the use of PID controller in order to obtain enough
phase boost for the stable operation.
Similarly to the VMC buck converter, the output-voltage loop (L(s)) can be
given by
1
L(s) = Gcc Gco (9.34)
VM
where VM (3 V) and Gco (Chapter 3) are known and we have to choose Gcc such
that the design goals are met. Therefore, the loop gain (9.34) can be given by
1 K (1 + s/z1 ) (1 + s/z2 )
L(s) =


3 s 1 + s/p1 1 + s/p2


 
D (Uo + UD ) rL + rds1 + D2 rc IL sLIL (1 + srC C)
LC (9.35)
rL + Drds1 + D (rd + rC ) D2
s +s
2 +
L LC

Figure 9.39 VM-controlled boost converter.

296 9 Control Design Issues
 2
The RHP zero (i.e., D (Uo +UD )(rLILL+rds1 +D rc )IL ) locates closest to the origin when
the input voltage is lowest and the output load highest. It should be also
observed that the resonant frequency (i.e., D / LC) moves toward the origin
when the input voltage decreases. Therefore, the known part of the loop gain
has to be plotted at the input voltage of 20 V and at the maximum load power
as shown in Figure 9.40 (solid line). The RHP zero locates at 1.369 kHz at
the dened condition and the resonant frequency at 121 Hz. According to the
phase behavior of Gco , it may be obvious that the crossover frequency has
to be limited to approximately half the frequency of the RHP zero. The rst
zero (z1 ) has to be placed before the resonant frequency. The phase boost is
maximized when the zeros (z1,2 ) are placed at the same frequency. The last
zero should be placed below half the switching frequency in order to maintain
sufcient gain margin due to the shape of the high-frequency magnitude of
Gco . The frequency response of the controller (K = 1) is shown in Figure 9.40
(dashed line) with the zero-pole placements dened in (9.36).
D
z1,2 = 0.5
LC
p1 = 30 z1 (9.36)
2fs
p2 =
2
In the next design phase, the voltage-loop gain with K = 1 should be plotted
as shown in Figure 9.41. The phase behavior of the loop gain reveals that the
worst-case phase margin has to be chosen rather low in order to obtain a high
enough crossover frequency: the crossover frequency of 750 Hz means that
the gain of the controller has to be increased by 21.3 dB. This gives K = 11.6

Figure 9.40 The rst phase of the control design: VMC boost converter.
 9.6 Simple Control-Design Method 297

Figure 9.41 The second phase of the control design: VMC boost converter.

Figure 9.42 The open (solid line) and closed-loop (dashed line) output impedances of
the VMC boost converter at the input of 20 V.

and the phase margin of 25 . The corresponding open- and closed-loop output
impedances are shown in Figure 9.42 explaining clearly the reason for the
known inferior load-transient response of a VMC boost converter operating in
CCM.
The predicted voltage-loop gains at the input voltage of 20 V (solid line) and
50 V (dashed line) are shown in Figure 9.43. It may be obvious from Figure 9.43
that the controller design at the high input voltage would have yielded unstable
298 9 Control Design Issues

Figure 9.43 The predicted voltage-loop gains of the VMC boost converter at the input
voltages of 20 V (solid line) and 50 V (dashed line).

operation at the low input voltage. Even if the crossover frequency has doubled
due to the increase of the input voltage, the load-transient response would not
improve because of the output impedances (i.e., the phase margin is still low
causing peaking in the sensitivity function, the resonant frequency has moved
and the damping has reduced).

9.6.4
Control Design Example: PCMC Boost Converter

The boost converter under PCM control is shown in Figure 9.44 and the
associated circuit elements are dened in Figure 9.24b. The equivalent
inductor-current sensing resistor (Rs ) is 150 m. The dynamic modeling
and characterization of the PCMC boost converter is previously presented in
Chapter 4. Gco of the PCMC boost converter contains the same RHP zero
as the corresponding VMC boost converter, which automatically limits the
achievable control bandwidth to the frequency of the RHP zero or below it.
The goal of the control design is to maximize the crossover frequency with a
satisfactory phase margin.
The resonant-free nature of the converter allows the use of PI controller.
The PWM modulator used in the experimental converter is UCC 3842, where
the error-amplier section contains an extra gain of 1/3.
The output-voltage loop gain (L(s)) can be given by
1
L(s) = Gcc Gco (9.37)
3Rs
 9.6 Simple Control-Design Method 299

Figure 9.44 The PCM-controlled boost converter.

similarly to the corresponding PCM-controlled buck converter. It is obvious

that Rs (150 m) and Gco (Chapter 4) are known, and we have to choose Gcc
such that the design goals are met. Therefore, the loop gain (9.37) can be given
by
1
L(s) = Gcc
3 0.15
sp
Fm (D UE IL (rE +D2 rC )) LIL
(1s  )(1 + s rC C)
LC D UE IL (rE + D2 rC )



sp sp sp sp
rE +D2 rC +Fm UE qc Fm IL qo

= s2 +s ( )+
L C
sp sp sp sp sp
D2 + D Fm (UE qo + IL qc ) Fm IL qo (rE + D2 rC )
LC
UE = Uo + UD + (rd + rC rds1 )IL rC Io
rE = rL + Drds1 + D (rd + DrC ) (9.38)

The converter is supposed to be compensated such that Mc Uo /2L for

providing the duty ratio of 100% (see Chapter 4, Section 4.5.2). The design
of the controller shall be carried out at the minimum input voltage and
maximum output power similarly to the corresponding VMC converter.
Therefore, we plot the known part of the loop gain as shown in Figure 9.45
at the input voltage of 20 V. The next task is to shape the phase of
the loop gain, which can be done similarly to the corresponding VMC
converter (i.e., z1 < 2fn , p1 < 0.5 2fs ). The resulting frequency responses
of the controller (K = 1) with the pole-zero placements dened in (9.35)
300 9 Control Design Issues

Figure 9.45 The rst phase of the control design: PCMC boost converter.

Figure 9.46 The second phase of the controller design: PCMC boost converter.

(dashed line) and the corresponding loop gain (solid line) are shown in
Figures 9.45 and 9.46, respectively.
D
z1 = 0.5
LC
(9.39)
2fs
p2 =
8
The high-frequency shape of the loop gain (K = 1) in Figure 9.46 indicates that
the limiting factor in the design is maintaining of the adequate gain margin.
 9.6 Simple Control-Design Method 301

Figure 9.47 The open (solid line) and closed-loop (dashed line) output impedances of
the PCMC boost converter at the input voltage of 20 V.

Therefore, the controller gain can be increased only by 65.9 dB GM. If GM of

6 dB is adequate then the maximum gain increase is 59.9 dB corresponding to
K = 988, the crossover frequency equals 804 Hz, and the phase margin equals
55 . It may be obvious that the PCM control does not remove the effect of the
RHP zero as claimed in [47].
The open- and closed-loop output impedances of the PCMC boost
converter at the input voltage of 20 V are shown in Figure 9.47. The
closed-loop output impedance implies that the load-transient response would

Figure 9.48 The voltage-loop gains at the input voltages of 20 and 50 V.

302 9 Control Design Issues

be quite similar to the corresponding VMC boost converter discussed

earlier.
The predicted voltage-loop gains at the input voltages of 20 V (solid line)
and 50 V (dashed line) are shown in Figure 9.48. The shape of the loop gains
implies that the load-transient behavior would be slightly improved when the
input voltage increases compared to the low-line response.

9.7
Conclusions

Basically the control design is trivial when knowing the behavior of the dynamic
elements contributing to the loop gain and having an ability to interpret the
information incorporated into the frequency responses. In practice, the control
design is difcult because of the uncertainties incorporated into different
passive and active circuit elements. The possible effect of the uncertainties can
be naturally studied effectively only if the analytical models in the symbolic
form are available. The most accurate control design would be achieved if the
control-to-output transfer function is experimentally measured at the operating
point most crucial to the design but naturally the other uncertainties will stay.
As high loop gain as possible before the crossover frequency and as low loop
gain as possible after the crossover frequency would provide the best condition
the control design can ever provide. The single-op-amp-based controller cannot
provide such features but several cascaded controllers or digital control has to
be used for obtaining such a goal.
The simple shunt-regulator-based controllers frequently utilized especially
in the low-cost and high-volume applications deserve much greater concern
than typically given in order to achieve high product quality.
An excellent control design would not necessarily yield robust stability and
excellent transient response, because the external interactions can change the
dynamics of the converter profoundly. The peak-current-mode control and
input-current-feedforward in a buck-type converter would make the converter
quite invariant to those external interactions and also improve substantially
the input-voltage and load transient responses.

References

1. A. Lidow, D. Kinzer, G. Sheridan, delivery, in P roc. IEEE Applied Power

and D. Tam, The semiconductor Electronics Conf., 2003, pp. 39.
roadmap for power management in 3. R. Miftakhutdinov, An analytical
the new millennium, Proc. IEEE, vol. comparison of alternative control
89, no. 6, 2001, pp. 803812. techniques for powering next
2. A. Lidow and G. Sheridan, Dening generation microprocessors, in Proc.
future for microprocessor power Power Supply Design Seminar (SEM
304 9 Control Design Issues

International Telecommunications control, IEEE Trans. Power Electron.,

Energy Conf., 2005, pp. 533538. vol. PE-1, no. 3, 1986, pp. 186191.
25. M. Hankaniemi, M. Karppanen, 35. M. Karppanen, T. Suntio
and T. Suntio, Load-imposed and M. Sippola, Dynamical
instability and performance characterization of
degradation in a regulated converter, input-voltage-feedforward-controlled
IEE Proc. Electr. Power Appl., vol. 153, buck converter, IEEE Trans. Indust.
no. 6, 2006, pp. 781786. Electron., vol. 54, no. 2, 2007, pp.
26. M. Karppanen, M. Sippola, 10051013.
and T. Suntio, Source-imposed 36. K.J. Astrom and T. Hagglund,
instability and performance Advanced PID Control,
degradation in a regulated converter, Instrumentation, Systems, and
in Proc. IEEE Power Electronics. Automation Society (ISA), Research
Specialists Conf., 2007, pp. 194200. Triangle Park, NC, USA, 2006.
27. M. Karppanen, T. Suntio, 37. Y. Li, K.H. Ang and G.C.Y. Chong,
and M. Sippola, Impact of PID control system analysis and
output-voltage remote sensing on design, IEEE Control Syst. Mag., vol.
converter dynamics, Int. Rev. Electr. 26, no. 1, 2006, pp. 3241.
Eng., vol. 2, no. 2, 2007, pp. 196202. 38. H. Kwakernaak, Robust control and
28. C. DeVries, A novel technique for H -optimization Tutorial paper,
compensating POL converters with Automatica, vol. 29, no. 2, 1993, pp.
variable output capacitance, in Proc. 255273.
IEEE Applied Power Electronics Conf., 39. K. Zhou and J.C. Doyle, Essentials of
Robust Control, Prentice-Hall, Upper
2006, pp. 474479.
Saddle River, NJ, USA, 1998.
29. B. Choi, Step load response of a
40. I. Gadoura and T. Suntio, Practical
current-mode-controlled DC-to-DC
robust control of paralleled DC/DC
converter, IEEE Trans. Aerosp.
converters using H loop-shaping
Electron. Syst., vol. 33, no. 4, 1997, pp.
technique, in Proc. Power Conversion
11151121.
and Intelligent Motion Conf. (Europe),
30. C. Gezgin, Predicting load transient
2002, pp. 335340.
response of output voltage in DCDC 41. J.M. Maciejowski, Multivariable
converters, in Proc. IEEE Applied Feedback Design, Addison-Wesley,
Power Electronics Conf., 2004, pp. Reading, MA, USA, 1989.
13391343. 42. F.H. Leung, P.K.S. Tam, and C.K. Li,
31. D. Morrison, Modeling DCDC The control of switching dcdc
converter transient response, Power converters A general LQR problem,
Electronics Technology, 2004, p. 56 IEEE Trans. Indust. Electron., vol. 38,
(online: www.powerelectronics.com). no. 1, 1991, pp. 6571.
32. J. Betten and R. Kollman, Easy 43. G. Escobar. R. Ortega, H. Sira-
calculation yields load transient Ramirez, J.-P. Villain, and I. Zein, An
response, Power Electronics Technology, experimental comparison of several
2005, pp. 4048 (online: nonlinear controllers for power
www.powerelctronics.com). converters, IEEE Control Syst. Mag.,
33. R. Redl, B.P. Erisman, and Z. Zansky, vol. 19, no. 1, 1999, pp. 6682.
Optimizing the load transient 44. P. Mattavelli, L. Rossetto,
response of the buck converter, in and G. Spiazzi, Small-signal analysis
Proc. IEEE Applied Power Electronics of DCDC converters with sliding
Conf., 1998, pp. 170176. mode control, IEEE Trans. Power
34. R. Redl and N.O. Sokal, Electron., vol. 12, no. 1, 1997, pp.
Near-optimum dynamic performance 96102.
of the buck converter using 45. H. Al-Atrash and I. Batarseh, Digital
feed-forward of output current and controller design for a practicing
input voltage with current-mode power electronics engineer, in Proc.
 References 305

IEEE Applied Power Electronics Conf., design of isolated power supplies with
2007, pp. 3441. optocoupler feedback, IEEE Trans.
46. J.S. Freudenberg and D.P. Looze, Power Electron., vol. 20, no. 4, 2005, pp.
Right half plane poles and zeros and 823832.
design tradeoffs in feedback systems, 55. J. Lempinen and T. Suntio,
IEEE Trans. Autom. Cont., vol. AC-30, Small-signal modeling for design of
no. 6, 1985, pp. 555565. robust variable-frequency yback
47. D.M. Mitchell, Tricks of the Trade: battery charger, in Proc. IEEE Applied
Understanding the right-half-plane Power Electronics Conf., 2001, pp.
zero in small-signal DC-DC converter 548554.
models, IEEE Power Electronics Society 56. Texas Instruments, Inc., Dallas (2003,
Newsletter, 2001, pp. 56. February.), Precision adjustable shunt
48. K.D.T. Ngo, S. Kirachaiwanich, regulator, fo-
and M. Walters, Buck modulator with cus.ti.com/lit/ds/symlink/tl431a.pdf.
improved large-power bandwidth, 57. D. Venable, Testing and stabilizing
IEEE Trans. Aerosp. Electron. Syst., vol. feedback loops in todays power
38, no. 4, 2002, pp. 13351343. supplies, Technical Report 17,
49. K. Yao, Y. Meng, and F.C. Lee, Venable Industries, Inc. Austin, TX,
Control bandwidth and transient USA, (online: www.venable.biz).
response of buck converters, in Proc.
58. R. Kollman and J. Betten, Closing the
IEEE Power Electronics Specialists Conf.,
loop with a popular shunt regulator,
2002, pp. 137142.
Power Electronics Technology, 2003, pp.
50. Y. Qiu, J. Sun, M. Xu, K. Lee, and F.C.
3036 (online:
Lee, High-bandwidth designs for
www.powerelctronics.com).
voltage regulators with peak-current
59. T. Tepsa and T. Suntio, Adjustable
control, in Proc. IEEE Applied Power
shunt regulator based control
Electronics Conf., 2006, pp. 2430.
51. S.A. Chickamena- systems, IEEE Power Electron. Lett.,
halli, S. Mahadevan, E. Standford, vol. 1, no. 4, 2003, pp. 9396.
and K. Merley, Effect of target 60. R. Ridley, Designing with the TL431,
impedance and control loop design on Switching Power Magazine, vol. 5, no.
VRM stability, in Proc. IEEE Applied 2, 2004, pp. 2026 (online:
Power Electronics Conf., 2002, pp. www.switchingpowermagazine.com).
196202. 61. C. Basso, Biasing the TL431 for
52. K. Yao, M. Xu, Y. Meng, and F.C. Lee, improved output impedance, Power
Design considerations for VRM Electronics Technology, 2005, pp. 5657
transient response based on the (online: www.powerelectronics.com).
output impedance, IEEE Trans. Power 62. B.T. Irving and M.M. Jovanovic,
Electron., vol. 18, no. 6, 2003, pp. Analysis and design of self-oscillating
12701277. yback converter, in Proc. IEEE
53. R. Mammano, Isolating the control Applied Power Electronics Conf., 2002,
loop, in Proc. Power Supply Design pp. 897903.
Seminar (SEM-1000), Unitrode Corp, 63. F. Tonicello, The control problem of
Merrimack, NH, USA, 1994, pp. maximum point power tracking in
C21C215. power systems, in Proc. 7th European
54. Y. Panov and M.M. Jovanovic, Space Power Conf., 2005, CD-ROM
Small-signal analysis and control publication, pp. 7.
 307

10
The Fourth-Order Converter Superbuck

10.1
Introduction

The second-order converters buck, boost, and buckboost treated in

Chapters 36 form the basis for implementing majority of the converters
used in the practical applications [1]. In some applications such as power-
factor correction [2, 3] and interfacing of solar arrays [48], a highly desired
feature of the converter is the continuous input current. The continuous
current means also reduced EMI noise and current stresses in the associated
capacitors. Only the boost converter of the basic converters can offer such a
feature but its other properties may not be always proper for the intended
application. The continuous input current exists only in the converters having
an inductor placed at the input current path. This means the necessity to use
higher order converters such as introduced in [9].
The Cuk converter [10] (Figure 10.1a) invented in the late 1970s is one
of the best known higher order converters along with the SEPIC converter
(Figure 10.1b) introduced in [1]. From Figure 10.1, the input current in both
of the converters is continuous due to the placement of L1 but the output
capacitor is supplied with continuous current only in the Cuk converter due
to the placement of L2 . The inverse polarity of the output voltage in the
Cuk converter makes its direct use limited and questionable as an optimal
converter.
Several other topological alternatives to implement the desired feature are
available as described in [9] of which the buck-like converter known as step-
down, current-sourced buck, two-inductor buck or superbuck has become a
popular choice especially in the solar energy applications [4, 5, 11] (Figure 10.2).
Its static and dynamic features greatly resemble those of the conventional buck
converter introduced earlier in Chapters 3 and 4. It may, however, have
certain dynamic anomalies such as RHP zeros and poles not present in the
conventional converter. The appearance of those anomalies would greatly
affect the controller design and obtainable transient dynamics [16, 17], but
they can be controlled by means of the power-stage design.

Dynamic Prole of Switched-Mode Converter. Teuvo Suntio

2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
ISBN: 978-3-527-40708-8
308 10 The Fourth-Order Converter Superbuck

Figure 10.1 Fourth-order

converters known as (a) Cuk
and (b) SEPIC.

Figure 10.2 Superbuck

converter.

The input-current ripple can be further reduced by using coupled-inductor

design [1215] as demonstrated in [24]. The coupling of inductors may
further complicate the dynamics of the converter by either introducing
anomalies such as RHP poles and zeros or advancing their appearance.
Usually the simultaneous appearance of RHP poles and zeros makes the
converter inapplicable as demonstrated in [18] (see Chapter 9, Section 9.2.2,
Figure 9.6).
Actually the superbuck converter incorporates features making it a good
candidate for an optimal converter.
The rest of the chapter is dedicated for the modeling and analysis of
the superbuck converter as a representative of the fourth-order converters,
as an example of the anomalies they can introduce, for demonstrating
the applicability of the methods introduced in the previous chapters, and
for illustrating the extent of analysis needed for understanding fully the
converter dynamics. Experimental evidence is provided from two different
PCM-controlled superbuck designs. The study of the existence of the
RHP poles requires the use of RouthHurwitz analysis methods [19],
which are shortly introduce. The methods for analyzing the dynamics
 10.2 Basic Dynamics 309

Figure 10.3 Superbuck converter with the direction of currents and the polarity of
voltages.

related to the coupled-inductor design [20] are also shortly introduced and
applied to the superbuck converter under VM and PCM control. Design
considerations are provided for avoiding the appearance of the RHP
zeros and poles and designing the optimal compensation for the PCM
control.

10.2
Basic Dynamics

The basic dynamics of a certain converter is related to the direct-duty ratio or

voltage-mode control as stated in Chapter 2, which has to be modeled rst:
The direction of the currents and polarity of the voltages constituting the state
variables in a superbuck converter are shown in Figure 10.3. During the on-
time, the main switch (MOSFET) is controlled on. As a consequence, the circuit
structure of the converter becomes as shown in Figure 10.4a. During the off-
time, the main switch is turned off and the diode conducts. As a consequence,
the circuit structure of the converter becomes as shown in Figure 10.4b.
From Figure 10.4, we can construct the state-space representations for the
on- and off-times from which the average and small-signal state spaces and
corresponding models can be solved by applying the methods introduced in
Chapter 3. During the on-time the state space can be given by
diL1 rL1 + rds + rC2 rds + rC2 uC2 uin rC2
= iL1 iL2 + + io
dt L1 L1 L1 L1 L1
diL2 rds + rC2 rL2 + rds + rC1 + rC2 uC1 uC2 rC2
= iL1 iL2 + + io
dt L2 L2 L2 L2 L2
duC1 iL2
=
dt C1
duC2 iL1 i2 io
= +
dt C2 C2 C2
310 10 The Fourth-Order Converter Superbuck

iin = iL1
uo = rC2 iL1 + rC2 iL2 + uC2 rC2 io (10.1)

and during the off-time by

diL1 rL1 + rC1 + rd + rC2 rd + rC2 uC1
= iL1 iL2
dt L1 L1 L1
uC2 uin rC2 UD
+ + io
L1 L1 L1 L1
diL2 rd + rC2 rL2 + rd + rC2 uC2 rC2 UD
= iL1 iL2 + io
dt L2 L2 L2 L2 L2
duC1 iL1
= (10.2)
dt C1
duC2 iL1 iL2 io
= +
dt C2 C2 C2
iin = iL1
uo = rC2 iL1 + rC2 iL2 + uC2 rC2 io

Figure 10.4 Superbuck structures during the (a) on time and (b) off time.
 10.2 Basic Dynamics 311

10.2.1
Averaged Models

10.2.1.1 Averaged State Space

The averaged state space can be obtained by multiplying (10.1) with d and
(10.2) with d
and summing them together. These procedures yield
diL1  rL1 + rC2 + drds + d
rd + d
rC1 rC2 + drds + d
rd
= iL1  iL2 
dt L1 L1
d
uC1  uC2  uin  rC2 d
UD
+ + io 
L1 L1 L1 L1 L1
diL2  rC2 + drds + d
rd rL2 + rC2 + drds + drC1 + d
rd
= iL1  iL2 
dt L2 L2
duC1  uC2  rC2 d
UD
+ + io  (10.3)
L2 L2 L2 L2
duC1  d
iL1  diL2 
=
dt C1 C1
duC2  iL1  iL2  io 
= +
dt C2 C2 C2
iin  = iL1 
uo  = rC2 iL1  + rC2 iL2  + uC2  rC2 io 

10.2.1.2 Steady-State Operating Point

The steady-state operating point can be solved from the averaged state space
(10.3) by setting the derivatives to zero. These procedures yield


Uo = DUin D
UD Drds + DD
rC1 + D
rd + D2 rL1 + D
2 rL2 Io
UC2 = Uo


UC1 = Uin D
UD DrL1 D
RL2 Io
(10.4)
IL1 = DIo
IL2 = D
Io
Iin = DIo

The duty ratio as a function of the input and output voltages, the load current,
and the parasitic elements can be solved from
Uin + UD + (rC1 + rds rd 2rL2 )Io Uo + UD + (rL2 + rd )Io
D2 D+ =0
(rL1 + rL2 rC1 )Io (rL1 + rL2 rC1 )Io
(10.5)
312 10 The Fourth-Order Converter Superbuck

10.2.1.3 Boundary Conduction Mode

The peak-to-peak ripple currents of L1 and L2 can be computed according to
the on-time derivatives of the corresponding currents from (10.1) at a certain
operating point (10.4) to be as follows:
DTs (Uin Uo (DrL1 + rds ) Io )

iL1pp =
L1


  (10.6)
DTs Uin Uo DrL1 + D
rC1 Io

iL2pp =
L2
The mode boundary between the CCM and DCM conduction modes would
take place [4], when the output current equals the half the sum of the
inductor-current peak-to-peak ripples (10.6) or (
iL1pp +
iL2pp )/2 = Io ,
because the sum of the inductor currents is supplied to the output capacitor
(Figure 10.4b) via a diode during the off-time. The critical K-value based
on the sum of inductor currents yields Kcrit = D
, which is the same
as dened for the conventional buck converter in [1]. The K-value for
analyzing the operation mode (i.e., CCM: K > Kcrit , DCM: K < Kcrit , and
BCM: K = Kcrit ) can be given by 2Lp /Ts Req , where Lp = L1 L2 /(L1 + L2 ) and
Req = Uo /Io .

10.2.2
Small-Signal Models

10.2.2.1 Small-Signal State Space

The small-signal state space can be obtained from (10.3) by developing
the proper partial derivatives as introduced in Chapter 2. This procedure
yields when the steady-state operating point (10.4) is also taken into
account:

diL1 R1 R2 D
1 1 rC2 U1
= iL1 iL2 uC1 uC2 + uin + io + d
dt L1 L1 L1 L1 L1 L1 L1

diL2 R2 R3 DuC1 uC2 rC2 U2

= iL1 iL2 + + io + d
dt L2 L2 L2 L2 L2 L2
duC1 D
D Io
= iL1 iL2 d
dt C1 C1 C1
(10.7)
duC2 1 1 1
= iL1 + iL2 io
dt C2 C2 C2

iin = iL1
duC2
uo = uC2 + rC2 C2
dt
 10.2 Basic Dynamics 313

which can also be given in a matrix form as

R1 R2 D
1 1 rC2 U1
diL1

dt L1 L1 L1 L1 L L1 L1
1
di R2 R3 D 1 iL1 rC2 U2
L2 i uin
L2 0
dt = L2 L2 L2 L2

+
L2 L2 io
D
D Io
duC1 0
uC1 0
0 0 d
dt C1 C1 uC2 C 1
1
duC2 1 1
0 0 0 0
dt C2 C2 C2

iL1

1 0 0 0

i 0 0 0
uin
iin L2 io
= d +
uo 0 0 0 1 + rC2 C2 uC1 0 0 0
dt d
uC2
(10.8)
where

R1 = rL1 + rC2 + Drds + D
(rd + rC1 )

R2 = rC2 + Drds + D
rd
R3 = rL2 + rC2 + D(rds + rC1 ) + D
rd
U1 = Uin + UD + (rd rds + DrC1 DrL1 + D
rL2 )Io
U2 = U1 rC1 Io

10.2.2.2 Transfer Functions

The corresponding transfer functions (i.e., G-parameters) can be solved from
(10.8) most conveniently by using proper software packages such as Matlab
with Symbolic Toolbox yielding
Input dynamics:

s R3 C1 + D 2 C2

Yino = s +s
2
+
L1 L2 L 2 C1 C2

1 R3 R2 D

Toio = s2 + s + (1 + srC2 C2 )
L 1 C2 L2 L 2 C1
(10.9)
U1 3 D
Io L2 + (R3 U1 R1 U2 ) C1

Gci = s + s2
L1 U1 L2 C1

2 
(U1 U2 ) C1 + D U1 + DD
U2 + (DR2 + D
R3 )Io C2
+s
U1 L2 C1 C2

Io
+
U1 L2 C1 C2
 314 10 The Fourth-Order Converter Superbuck

Output dynamics:

1 R3 R2 D

Gioo = s2 + s + (1 + srC2 C2 )
L 1 C2 L2 L 2 C1

1 (R3 rC2 )L1 + (R1 rC2 )L2

Zoo = s3 + s2
C2 L1 L2
D2 L1 + D
2 L2 + (R1 R3 R22 rC2 (R1 2R2 + R3 ))C1
+s
L 1 L 2 C1
 (10.10)


2
D R1 + 2DD R2 + D R3 rC2
2
+ (1 + srC2 C2 )
L 1 L 2 C1


(U1 L2 + U2 L1 ) 2 C1 (U1 (R3 R2 ) + U2 (R1 R2 )) + D
L2 DL1 Io

Gco = s +s
L 1 L 2 C2 C1 (U1 L2 + U2 L1 )

 
DU1 + D
U2 D(R1 R2 ) + D
(R2 R3 ) Io
+ (1 + srC2 C2 )
C1 (U1 L2 + U2 L1 )

where the determinant
is



+ R 1 L2
3 R 3 L1
(L1 + L2 )C1 + D2 L1 + D
2 L2 C2 + R1 R3 R22 C1 C2
s +s
4
+s2
L1 L2 L 1 L 2 C1 C2

2 
(R1 2R2 + R3 )C1 + D R1 + 2DD R2 + D
2 R3 C2


1
+s +
L 1 L 2 C1 C2 L 1 L 2 C1 C2
(10.11)

The determinant has two complex conjugate or resonant LHP roots

approximately at

1 1
f1
2 (L1 + L2 )C1
 (10.12)
1 L1 + L2
f2
2 L 1 L 2 C2

which are not exactly the real resonant frequencies but close enough and also
consistent with the circuit schematics shown in Figure 10.2.
The ideal input impedance Zin can be computed as

s2 (U1 L2 + U2 L1 ) C1 + sE1 + E2
Zin =
sU2 C1 DIo

 (10.13)
E1 = U1 (R3 R2 ) + U2 (R1 R2 )C1 + Io D
L2 DL1


E2 = DU1 + D
U2 + D(R2 R1 ) + D
(R3 R2 ) Io
 10.2 Basic Dynamics 315

where the numerator term is the same as in the control-to-output transfer

function (Gco ) shown in (10.10). Its low-frequency value is approximately
equal to DI
Uin
o
, which is the same as in the corresponding conventional buck
converter but the resonant nature of Zin makes the superbuck converter
more sensitive to source interactions. The resonant nature can be reduced
providing parallel damping to C1 . The ideal impedance is symbolically the
same and characteristic to the superbuck topology regardless of the control
mode when the inductors have no coupling between them.
The short-circuit input impedance Zinsc can be computed as

s3 L1 L2 C1 + s2 C1 (L1 (R3 rC2 ) + L2 (R1 rC2 )) + sE1 + E2

Zinsc =
s2 L2 C1 + s(R3 rC2 )C1 + D2
(10.14)
E1 = D2 L1 + D
2 L2 + (R1 R3 R22 (R1 2R2 + R3 )rc2 )C1
E2 = D2 R1 + 2DD
R2 + D
2 R3 rC2

The resonant nature of Zinsc , i.e., a resonant zero pair approximately
 
D2 L1 +D
2 L2
at f 2
1
L1 L2 C1
and a resonant pole pair at f D
2 L2 C1
makes the
superbuck converter more sensitive to the source interactions than the
conventional buck converter. It is obvious that the parallel damping (C1 )
proposed above damps also the resonances in Zinsc , because the resonances
are dependent on C1 .

10.2.3
RHP Poles

The determinant
(10.11) contains the poles of the converter, which may
lie either in LHP or RHP. If an RHP pole exists, it means that the control
bandwidth has to be designed to be higher than the location of the pole. The
roots of the fourth-order polynomial cannot be solved easily in the symbolic
form to verify their locations in the complex plane. Therefore, other methods
such as RouthHurwitz test [19] to study the existence and number of the
RHP roots of a polynomial have to be applied. According to the method,
the RouthHurwitz array is to be constructed based on the coefcients of
the polynomial under considerations

(s) = an sn + an1 sn1 + an2 sn2 + + a1 s + a0 (10.15)

as follows:
sn : an an2 an4
sn1 : an1 an3 an5
sn2 : b1 b2 b3
(10.16)
316 10 The Fourth-Order Converter Superbuck

sn3 : c1 c2 c3
sn4 : d1 d2 d3
:
s0 :

where the rst row (sn ) starts with the highest-order coefcient an , the second
row (sn1 ) with the second-highest-order coefcient an1 , and the next elements
within the rows are as dened in (10.17), the elements of the third and
subsequent rows follow the algorithm dened as
   
 a an2   a an4 
 n  n
an1 an3  an1 an5 
b1 = b2 =
an1 an1
   
 an1 an3   an1 an5 
   
b1 b2   b1 b3 
c1 = c2 = (10.17)
b1 b1
   
 b b2   b b3 
 1   1 
c1 c2  c1 c3 
d1 = d2 =
c1 c1

The number of RHP roots of
(s) is the number of algebraic sign changes in
the elements of the left column of the array (10.16) proceeding from top to
bottom. The rst or second-order polynomial has all roots in LHP if and only
if all the coefcients have the same algebraic sign. In the case of a higher order
polynomial, the same does not anymore guarantee the absence of the RHP
roots but the RouthHurwitz test has to be applied to conrm the situation. The
sign changes in the polynomial coefcients indicate, however, the existence of
at least one RHP root. The missing of one or several of the coefcients means
that the polynomial has either complex imaginary-axis roots or RHP roots or
both.
According to the principles laid down above, the RouthHurwitz array for
the fourth-order polynomial of interest can be given by

s4 : a4 a2 a0
s3 : a3 a1 0
s2 : b1 b2 0 (10.18)
s1 : c1 0 0
s0 : d1 0 0

where b1 , b2 , c1 , and d1 are dened by

a1 a4
b1 = a2 b2 = a0
a3
(10.19)
a0 a3
c1 = a1 d1 = ao
b1
 10.2 Basic Dynamics 317

According to the determinant (10.11), the polynomial coefcients are as

follows:
a4 = 1
R0 (L1 + L2 )
a3 =
L1 L2


(L1 + L2 )C1 + D2 L1 + D
2 L2 C2
a2 = (10.20)
L 1 L 2 C1 C2
R0
a1 =
L 1 L 2 C1
1
a0 =
L 1 L 2 C1 C2
where the parasitic resistances R1 , R2 , and R3 are assumed to be the same
and denoted by R0 . (Note: in practice, R1 and R3 are always greater than
R2 . Therefore, the differences associated to those parasitics in (10.11) are
always positive numbers and the equalling of them would give the worst case
condition.) All the polynomial coefcients are positive real numbers. The array
coefcients b1 , b2 , c1 , and d1 can be given by
(L1 + L2 )(C1 + C2 )
b1 =
L 1 L 2 C1 C2
1
b2 =
L 1 L 2 C1 C2
(10.21)
R 0 C2
c1 =
L1 L2 C1 (C1 + C2 )
1
d1 =
L 1 L 2 C1 C2
which are also positive real numbers. Therefore, all the roots are LHP roots.

10.2.4
Design Considerations

According to (10.10), the control-to-output transfer

 function (Gco ) incorporates
a resonant zero pair approximately at f 2 1 1
(L1 +L2 )C1
, which will move to
RHP, when


C1 (U1 (R3 R2 ) + U2 (R1 R2 )) + D
L2 DL1 Io < 0 (10.22)

Its appearance is dependent on the input voltage and output current as well
as on the sizing of L1 , L2 , and C1 as implied by (10.22). Usually the inductors
are designed to be the same for the logistic and cost reasons. The appearance
of the RHP zero sets an absolute limit on the maximum control bandwidth
318 10 The Fourth-Order Converter Superbuck

and, therefore, the appearance may not be desirable. It may be obvious (10.22)
that the appearance of the RHP zero pair can be eliminated if the inductors
are designed such that L2 D max
D
max
L1 , where Dmax is the maximum duty ratio
the converter may have during the operation. In the applications, where the
input voltage is rather high and the output current rather low, the RHP zero
pair may not appear even if the inductors are equal.

10.3
Coupled-Inductor Superbuck

The coupled-inductor design (Figure 10.5) is usually used in order to reduce

the number of separate inductors and the input or output current ripple
resulting also in the reduction of the overall costs. The application of the
coupled-inductor technique may, however, change profoundly the internal
dynamics of the converter but those effects are not very well known because of
the complicity of the analysis and errors usually made in it as in [2] and [3].
The analysis of the coupled-inductor effects is based on the basic terminal
equations of a nonideal transformer [20] given by
di1 di2
u1 = L1 +M
dt dt
(10.23)
di1 di2
u2 = M + L2
dt dt

Figure 10.5 Coupled-inductor

superbuck converter.

Figure 10.6 Denition of (a) a two-winding transformer and (b) its equivalent circuit.
 10.3 Coupled-Inductor Superbuck 319

where L1 and L2 are the self-inductances measured from the primary (u1 )
and secondary (u2 ) ports, respectively, when the other port is an open circuit,
and M is the mutual inductance governing the transfer of energy between
the primary and secondary as dened in Figure 10.6a. The equivalent circuit
of such a transformer is shown in Figure 10.6b, where Ll1 and Ll2 are the
primary and secondary leakage inductances, respectively, LM is the primary
magnetizing inductance, and n1 and n2 are the primary and secondary number
of turns, respectively, of the ideal transformer.
From (10.23), we may dene that

L1 = Ll1 + LM
2
n2
L2 = Ll2 + LM (10.24)
n1
n2
M = LM
n1

The coupling coefcient (k) of a transformer is usually dened as k = LML

1 2
based on the total energy stored in the transformer [20]. Perfect coupling
between the primary and secondary (i.e., k = 1) means that the leakage
inductances (Lli ) are zero.

10.3.1
Small-Signal Models

The small-signal representation of a coupled-inductor converter can be

constructed from the state space of the original converter dened with the
separate inductors L1 and L2 taking into account that the voltages across the
inductors do not change. Therefore, if we denote the original small-signal
derivatives of the inductor currents by DL1 and DL2 and apply the information
given in (10.23) then the couple-inductor-affected derivatives can be given by
M M
DL1 DL2 DL1 DL2
diL1 L1 L1
= =
dt M2 1 k2
1
L1 L2
(10.25)
M M
DL2 DL1 DL2 DL1
diL2 L2 L2
= =
dt M2 1 k2
1
L1 L2
From (10.25), we may concluded that the perfect coupling (k = 1) would lead
to innite derivatives indicating the existence of high current spikes during
the transients.
 320 10 The Fourth-Order Converter Superbuck

From (10.8) and (10.25), the small-signal coupled-inductor VMC state space
can be given by
L R MR L2 R2 MR3 L2 D
+ MD L2 M
diL1
2 1 2

dt L1 L2 M2 L1 L2 M 2 L1 L2 M 2 L1 L2 M 2
i

diL2 L1 R2 MR1 L1 R3 MR2 DL1 + MD
L1 M L1

dt L1 L2 M2 L1 L2 M 2 L1 L2 M 2 L1 L2 M 2 iL2
=

duC1 D
D uC1
0 0
dt C1 C1
uC2
1 1
duC2
0 0
dt C2 C2

L2 (L2 M)rC2 L2 U1 MU2

1 2 M2
L L L1 L2 M 2 L1 L2 M 2

M (L1 M)rC2 L1 U2 MU1
uin
L L M2 L1 L2 M 2 L1 L2 M 2
1 2
+ io
Io
0 0
C1 d


1
0 0
C2

iL1

 
1 0 0 0 i 0
uin
iin L2 0 0
io
= d + (10.26)
uo 0 0 0 1 + rC2 C2 uC1 0 0 0
dt d
uC2
The transfer functions constituting the G-parameter set can be solved from
(10.26) by using a proper software package yielding.
Input dynamics:

sL2 R3 C1 + D 2 C2

Yino = s2
+ s +
L1 L2 M 2 L2 L 2 C1 C2

1 D

Toio = s2
(L 2 M) + s(R 3 R 2 ) + (1 + srC2 C2 )
(L1 L2 M2 )C2 C1


1 (R3 U1 R1 U2 )C1 + D
L2 + M Io

Gci = s (U1 L2 MU2 ) + s
3 2
L1 L2 M 2 C1

 
(U1 U2 ) C1 + D2 U1 + DD
U2 + (DR2 + D
R3 )Io C2 Io
+s +
C1 C2 C1 C2
(10.27)
 10.3 Coupled-Inductor Superbuck 321

Output dynamics:

1 D

Gioo = s 2
(L 2 M) + s(R 3 R 2 ) + (1 + srC2 C2 )
(L1 L2 M2 )C2 C1

1 3 (R3 rC2 )L1 + (R1 rC2 )L2 2(R2 rC2 )M

Zoo = s + s2
C2 L1 L2 M 2
(R1 R3 R22 (R1 2R2 + R3 )rC2 )C1 + D2 L1 + 2DD
M + D
2 L2
+s
(L1 L2 M2 )C1


2 
D R1 + 2DD R2 + D R3 rC2
2
+ (1 + srC2 C2 )
(L1 L2 M2 )C1


2 (L2 M)U1 + (L1 M)U2 E1 + D
(L2 M) D(L1 M) Io

Gco = s +s
(L1 L2 M2 )C2 (L1 L2 M2 )C1 C2

E2
+ (1 + srC2 C2 )
(L1 L2 M2 )C1 C2
E1 = ((R3 R2 )U1 + (R1 R2 )U2 )C1



E2 = DU1 + D
U2 D(R1 R2 ) + D
(R2 R3 ) Io (10.28)

where the determinant (
) is

R1 L2 + R3 L1 2MR2 1

= s4 + s3 + s2 a2 + sa1 +
L1 L2 M 2 (L1 L2 M2 )C1 C2


(R1 R3 R22 )C1 C2 + (L1 + L2 2M)C1 + D2 L1 + D
2 L2 + 2DD
M C2
a2 =
(L1 L2 M2 )C1 C2

2 
(R1 2R2 + R3 )C1 + D R1 + 2DD
R2 + D
2 R3 C2
a1 = (10.29)
(L1 L2 M2 )C1 C2
Thus the special input impedances (Zin and Zinsc ) can be given by


s2 ((L2 M)U1 + (L1 M)U2 )C1 + s(E1 + D
(L2 M) D(L1 M) Io ) + E2
Zin =
sC1 U2 DIo
s3 (L1 L2 M2 )C1 + s2 C1 (L1 (R3 rC2 ) + L2 (R1 rC2 ) 2(R2 rC2 )M) + sE3 + E4
Zinsc =
s2 L2 C1 + s(R3 rC2 )C1 + D2
E1 = ((R3 R2 )U1 + (R1 R2 )U2 )C1



E2 = DU1 + D
U2 D(R1 R2 ) + D
(R2 R3 ) Io
E3 = D2 L1 + D
2 L2 + 2DD
M + ((R1 R3 R22 (R1 2R2 + R3 )rc2 )C1
E4 = D2 R1 + 2DD
R2 + D
2 R3 rC2 (10.30)

According to the above-presented transfer functions (10.27)(10.30), we may

conclude that the coupling of the inductors will change the location and
damping of the resonances, which will affect the appearance of the resonant
322 10 The Fourth-Order Converter Superbuck

RHP zero pair in the control-to-output transfer function (Gco ), and even change
the behavior of the special input impedances.

10.3.2
RHP Poles

According to the determinant (10.29), the polynomial coefcients are as

follows, when the parasitic resistances R1 , R2 , and R3 are assumed to be equal
and denoted by R0 :

a4 = 1
(L1 + L2 2M)R0
a3 =
L1 L2 M 2


(L1 + L2 2M)C1 + D2 L1 + D
2 L2 + 2DD
M C2
a2 = (10.31)
(L1 L2 M2 )C1 C2
R0
a1 =
(L1 L2 M2 )C1
1
a0 =
(L1 L2 M2 )C1 C2

The coupling coefcient k = M/ L1 L2 is always less than unity. Thus
all the polynomial coefcients in (10.31) are positive real numbers. The
RouthHurwitz array coefcients b1 , b2 , c1 , and d1 (see (10.20)) can be given by

L1 + L2 2M (L1 L2 )2 + 2DD
(L1 + L2 2M) + M2

b1 = +
(L1 L2 M )C2
2 (L1 L2 M2 )(L1 + L2 2M)C1
1
b2 =
(L1 L2 M2 )C1 C2
R0 (L1 L2 )2 + 2DD
(L1 + L2 2M) + M2
c1 =
L1 L2 M2 (L1 + L2 2M)2 C1 + ((L1 L2 )2 + 2DD
(L1 + L2 2M) + M2 )C2
1
d1 = (10.32)
(L1 L2 M2 )C1 C2
which are real positive numbers, because k < 1. Therefore, all the roots of the
determinant are LHP roots regardless of the claims presented in [2] and [3].

10.3.3
Input-Current-Ripple Reduction

Applying the coupled-inductor algorithms (10.25) to the on-time and off-time

inductor-current derivatives (see Section 10.2), the peak-to-peak ripple values
 10.3 Coupled-Inductor Superbuck 323

become

(L2 M) (Uin Uo (DrL1 + rds ) Io ) MD
(rC1 Io + UD )

iL1pp = + DTs
L1 L2 M 2 L1 L2 M 2

(L1 M) (Uin Uo (DrL1 + rds ) Io ) L1 D
(rC1 Io + UD )

iL2pp = DTs
L1 L2 M 2 L1 L2 M 2
(10.33)
According to (10.33), the close-to-zero-input
 ripple or
iL1pp 0 can be
obtained, when M = L2 or k = LL21 . These values are also the commonly
dened zero-input-ripple conditions [24]. As a consequence, the residual
peak-to-peak ripple values are
rC1 Io + UD

iL1pp = DD
Ts
L1 L2
Uin Uo (DrL1 + rds ) Io rC1 Io + UD DD
Ts L1

iL2pp = DTs (10.34)
L2 L1 L2 L2
which imply that L1 should be sufciently higher than L2 for obtaining optimal
ripple reduction.
Substituting M and k in (10.26) with the above given values yields

diL1 R1 R2 R2 R3 1
0
dt L1 L2 L1 L2 L1 L2
iL1
R L R L R3 L1 R2 L2 DL1 + D L2

1
diL2

2 1 1 2

dt (L1 L2 )L2 (L1 L2 )L2 (L1 L2 )L2 L2
iL2
=
du
uC1
C1


D D
0 0
dt C1 C1 uC2

1 1
duC2 0 0
dt C2 C2
1 U1 U2
0
L1 L2 L 1 L2

U2 L1 U1 L2
1 rC2 uin
L L
(L1 L2 )L2
1 2 L2
+ io
Io
0 0 d
C1

1
0 0
C2

 iL1

1 0 0
uin
iin 0 iL2 0 0 0
= d uC1 + 0 0 0
io
uo 0 0 0 1 + rC2 C2
dt d
uC2
(10.35)
324 10 The Fourth-Order Converter Superbuck

As a consequence, the G-parameter-set of transfer functions can be given by

Input dynamics:

s R3 C1 + D 2 C2

Yino = s +s
2
+
L1 L2 L2 L 2 C1 C2

1 R3 R2 D

Toio = s + (1 + srC2 C2 )
(L1 L2 )C2 L2 L 2 C1

1 (R3 U1 R2 U2 )C1 + L2 Io

Gci = s3 (U1 U2 ) + s2
L1 L2 L 2 C1

2  
(U1 U2 ) C1 + D U1 + DD
U2 + (DR2 + D
R3 )Io C2 Io
+s +
L 2 C1 C2 L 2 C1 C2
(10.36)
Output dynamics:

1 R3 R2 D

Gioo = s + (1 + srC2 C2 )
(L1 L2 )C2 L2 L 2 C1

1 3 (R3 rC2 )L1 + (R1 2R2 + rC2 )L2

Zoo = s + s2
C2 (L1 L2 )L2
(R1 R3 R22 (R1 2R2 + R3 )rC2 )C1 + D2 L1 + D
(1 + D)L1
+s
(L1 L2 )L2


2 
D R1 + 2DD R2 + D R3 rC2
2
+ (1 + srC2 C2 )
(L1 L2 )L2

U2 ((R3 R2 )U1 + (R1 R2 )U2 )C1 D(L1 L2 )Io

Gco = s2 + s
L 2 C2 U2 (L1 L2 )C1

 
DU1 + D
U2 + D(R2 R1 ) + D
(R3 R2 ) Io
+ (1 + srC2 C2 )
U2 (L1 L2 )C1
(10.37)
where the determinant (
) is
R3 L1 + (R1 2R2 )L2 1

= s4 + s3 + s2 a2 + sa1 +
(L1 L2 )L2 (L1 L2 )L2 C1 C1

2 
(R1 R3 R2 )C1 C2 + (L1 L2 )C1 + D L1 + D
(1 + D)L2 C2
2
a2 = (10.38)
(L1 L2 )L2 C1 C2


(R1 2R2 + R3 )C1 + D2 R1 + 2DD
R2 + D
2 R3 C2
a1 =
(L1 L2 )L2 C1 C2
 10.4 PCM-Controlled Superbuck 325

Thus the special input impedances (Zin and Zinsc ) can be given by

s2 (L1 L2 )U2 C1 + s(((R3 R2 )U1 + (R1 R2 )U2 )C1 D(L1 L2 )Io ) + E1

Zin =
U2 C1 s DIo
s3 (L1 L2 )L2 C1 + s2 C1 (L1 (R3 rC2 ) + L2 (R1 2R2 + rC2 )) + sE2 + E3
Zinsc =
s2 L2 C1 + s(R3 rC2 )C1 + D2


E1 = DU1 + D
U2 + D(R2 R1 ) + D
(R3 R2 ) Io (10.39)
E2 = D2 L1 + D
(1 + D)L2 + ((R1 R3 R22 (R1 2R2 + R3 )rc2 )C1
E3 = D2 R1 + 2DD
R2 + D
2 R3 rC2

According to the above given transfer functions, we may conclude that the
coupling of the inductors yielding input-current ripple reduction damps the
resonances in the input-to-output (Gioo ) and output-to-input (Toio ) transfer
functions making them close to rst-order transfer functions and accelerates
the appearance of the RHP zero in the control-to-output transfer function.

10.3.4
Design Considerations

According to (10.37), the control-to-outputtransfer function (Gco ) incorporates

a RHP zero pair approximately at f 1
2
1
(1k2 )L1 C1
, when

((R3 R2 )U1 + (R1 R2 )U2 )C1 D(1 k2 )L1 Io < 0 (10.40)

where k is the coupling coefcient. The appearance of the RHP zero is most
likely in the applications where the input voltage is rather low and the output
current rather high. Equation (10.40) gives explicit suggestions for the possible
design actions for preventing the appearance of the RHP zero if possible.

10.4
PCM-Controlled Superbuck

In order to reduce the resonant behavior, the PCM control is applied as

shown in Figure 10.7, where the pulsewidth modulation is based on the sum
of the inductor currents measured by means of a current transformer from
the source of the MOSFET. As a consequence, the resonant behavior would
be removed from the output dynamics but it would be still left at the input
dynamics, because only the sum of the inductor currents is tightly regulated
and the individual currents can still freely vary within the sum.
326 10 The Fourth-Order Converter Superbuck

Figure 10.7 PCM-controlled

superbuck.

10.4.1
Small-Signal Models

The modeling of the PCM-controlled superbuck converter [11] can be carried

out similarly to the modeling of the conventional buck   converter introduced in
Chapter 4 by substituting the perturbed duty ratio d in the VMC state space
(10.8) with the duty-ratio constraints given by
 
d = Fm ico qL1 iL1 qL2 iL2 qC1 uC1 qC2 uC2 qin uin qo io (10.41)

The coefcients of (10.41) can be solved by linearizing the comparator equation

(i.e., the equation governing the reset control of the RS-ip-op in Figure 10.7),
which is constructed based on the combined inductor-current waveforms
shown in Figure 10.8.

ico mc dTs = iL1  + iL2  +
iL (10.42)

Figure 10.8 The combined

inductor-current waveforms
constituting the comparator control.
 10.4 PCM-Controlled Superbuck 327

dd
Ts n
According to Chapter 4,
iL can be given by 2 i=1 (mi1 + mi2 ), and
consequently, the comparator equation becomes
dd
Ts
ico mc dTs = iL1  + iL2  + (m11 + m12 + m21 + m22 ) (10.43)
2
Computing the inductor-current slopes from the actual circuit (i.e., m11 and
m21 can be obtained from (10.1) and m12 and m22 from (10.2) given in Section
10.2) and substituting them in (10.43) yields

dd
Ts u1  u2 
ico mc dTs = iL1  + iL2  + + (10.44)
2 L1 L2

where u1  and u2  are dened as

u1  = uC1  + UD + (rC1 + rd rds )iL1  + (rd rds )iL2 

(10.45)
u2  = uC1  + UD + (rd rds )iL1  + (rd rC1 rds )iL2 

The coefcient of the duty-ratio constraints (10.40) can be solved from (10.43)
by developing the proper partial derivatives. This procedure yields
1
Fm =

 
D D Ts U1 U2
Ts Mc + +
2 L1 L2


DD Ts rC1 + rd rds rd rds
qL1 =1+ +
2 L1 L2


DD Ts rd rds rd rC1 rds
qL2 =1+ +
2 L1 L2
DD
Ts (L1 + L2 )
qC1 = (10.46)
2L1 L2
qC2 = qin = qo = 0
U1 = Uin + UD + (rd rds + DrC1 DrL1 + D
rL2 )Io
U2 = U1 rC1 Io

The formula of the duty-ratio gain (Fm ) in (10.46) indicates that Fm would
become innite (i.e., the denominator will become zero), when the maximum
duty ratio (Dmax ) is reached

Dmax = 0.5 + Mc L1 L2 /(L2 U1 + L1 U2 ) (10.47)

When the compensation (Mc ) is set to zero, the maximum duty ratio
corresponds to 0.5 as in the conventional buck converter. An increase in the
duty ratio beyond 0.5 will force the converter to enter into the second-harmonic
mode as explained in detail in Chapter 4, Section 4.3.3.
 328 10 The Fourth-Order Converter Superbuck

The corresponding PCM  state space can be obtained from (10.8) by replacing
the perturbed duty ratio d with (10.41). This gives
R +F q U R2 + Fm qL2 U1 D
+ Fm qC1 U1 1
diL1 1 m L1 1

dt L1 L1 L1 L1

di R2 + Fm qL1 U2 R3 + Fm qL2 U2 D Fm qC1 U2 1 iL1
L2
L2 iL2
dt = L2 L2 L2



uC1

duC1 D + Fm qL1 Io D Fm qL2 Io Fm qC1 Io
0
dt C1 C1 C1 uC2

1 1
duC2
0 0
dt C2 C2
1 rC2 Fm U1
1
L L 1 L1
U2
rC2 F m uin
0
L2 L2
+ io
Fm Io
0 0 ico
C1

1
0 0
C2


  iL1
 uin 
iin 1 0 0 0 iL2 0 0 0
= d + io (10.48)
uo 0 0 0 1 + rC2 C2 uC1 0 0 0
dt ico
uC2
The corresponding transfer functions can be solved from (10.48) most
conveniently by using proper software packages such as Matlab with Symbolic
Toolbox. The resulting symbolic transfer functions with the parasitic elements
are extremely long and do not easily offer the desired information. From the
dynamic point of view, the parasitics are, however, essential and cannot be
neglected, because they will provide the damping of the resonances and also
shift the appearance of the possible RHP zeros and poles. We give here only
the nonparasitic transfer functions for introducing the effect of PCM control
in the dynamics of the converter. The effect of the parasitics will be discussed
when appropriate.
Input dynamics:



 
1 F U2 C1 + I1 qC1 L2
2 m
C1 + D D Fm U2 qC1 + Io C2 Fm Io qC1

Yino = s +s
3
+s
L1 L 2 C1 L 2 C1 L 2 C1 C2



1 Fm (U2 U1 ) C1 + Io qC1 L2 D + Fm (U1 U2 ) qC1 Io

Toio = s2 + s +
L 1 C2 L 2 C1 L 2 C1

2  
Fm U1 3

2 D Io (U1 U2 ) C1 + D U1 + DD
U2 C2 Io

Gci = s +s +s +
L1 U1 C1 U1 L2 C1 C2 U1 L2 C1 C2
(10.49)
 10.4 PCM-Controlled Superbuck 329

Output dynamics:

1 Fm Io qC1 D Fm U2 qC1

Gioo = s s
2
+
L 1 C2 C1 L 2 C1


1 2 m (U2 1
F L + U1 L2 ) C1 Io qC1 L1 L2

Zoo = s +s
3
C2 L 1 L 2 C1



D2 L1 + D
2 L2 + Fm (qC1 U1 D
L2 U2 DL1 + Io D
L2 DL1 )
+s
L 1 L 2 C1




Fm DU1 + D U2
+
L 1 L 2 C1


 
Fm (U1 L2 + U2 L1 ) 2 D L2 DL1 Io DU1 + D
U2

Gco = s +s +
L 1 L 2 C2 (U1 L2 + U2 L1 ) C1 (U1 L2 + U2 L1 ) C1
(10.50)
where the determinant (
) is


Fm (U1 L2 + U2 L1 ) C1 Io qC1 L1 L2

= s4 + s3
L 1 L 2 C1



(L1 + L2 )(C1 + D
2 C2 ) + Fm (qC1 U1 D
L2 U2 DL1 + Io D
L2 DL1 )C2
+s 2
L 1 L 2 C1 C2
Fm ((DU1 + D
U2 )C2 Io qC1 (L1 + L2 )) 1 + Fm qC1 (U1 U2 )
+s +
L 1 L 2 C1 C2 L 1 L 2 C1 C2
(10.51)
The complexity of the fourth-order determinant is such that it is not anymore
possible to nd the approximate symbolic solutions for its roots of which two
are usually real and one is a complex conjugate root.
The ideal input impedance (Zin ) is the same as dened in (10.13). The
nonparasitic short-circuit input impedance (Zinsc ) can be computed to be

s3 L1 L2 C1 + s2 A + s B + Fm (DU1 + D
U2 )
Zinsc =



L2 C1 + s Fm U2 C1 Io qC1 L2 + D D Fm U2 qC1 + Io
s2


A = Fm (U1 L2 + U2 L1 ) C1 Io qC1 L1 L2


B = (D2 L1 + D
L2 + Fm (qC1 U1 D
L2 U2 DL1 + Io (D
L2 DL1 )))
(10.52)
Ref. [4] actually claims that the dynamics of the PCM-controlled superbuck
converter can be obtained from the dynamic representation of the conventional
buck converter by replacing the output inductor with the parallel inductor
Lp = L1 L2 /(L1 + L2 ), but this claim does not hold because of the resonant
behavior in the input dynamics and the possible appearance of the RHP poles
and zeros.
330 10 The Fourth-Order Converter Superbuck

10.4.2
Design Considerations

The PCMC converter incorporates the same RHP zero pair in its control-
to-output transfer function (Gco ) as the VMC converter discussed in Section
10.2.4. Therefore, the inductors should be designed such that L2 D max
L ,
D
max 1
where Dmax corresponds to the maximum duty ratio of the converter operation.
High input voltage and low output power may allow the inductors to be the
same even if the maximum duty ratio is higher than 0.5. The inductor-current
feedback in the PCMC converter also requires considering the selection of
the inductor-current slope compensation (Mc ) and the possible appearance of
RHP poles and their elimination.

10.4.2.1 Inductor-Current-Feedback Compensation

The zeroth-order term of the numerator of the input-to-output transfer
function (Gioo in (10.50)) a0 = D Fm (U2 qC1 + Io (D(qL1 qL2 ) + (R2
R3 )qC1 )) implies that the low-frequency gain of Gioo can be made small
by designing the compensation Mc such that the zeroth-order term (a0 ) will
be zero similarly to the conventional buck converter. a0 is clearly dependent
on the load current and parasitics and, therefore, the perfect compensation
DD
Uin
is not possible. When neglecting the parasitics, a0 D Mc Lp +(D
D)U , and
in
thus the close-to-optimal compensation becomes
Uo L1 L2
Mc Lp = (10.53)
2Lp L1 + L2
which has the same form as dened for the conventional buck converter, when
its inductor is substituted with the parallel inductor (Lp ).

10.4.2.2 Avoiding RHP Poles

According to the determinant (
) (10.51), the polynomial coefcients are as
follows:
a4 = 1

DD
Ts Io
Fm Uin
2C1
a3 =
Lp

DD
Ts
(L1 + L2 )(C1 + D
2 C2 ) + Fm + Io (D
L2 DL1 )C2
2Lp
a2 =
L 1 L 2 C1 C2
 
DD
Ts Io L1 L2 1
a1 = Fm Uin
2L2p C2 L 1 L 2 C1
1
a0 = (10.54)
L 1 L 2 C1 C2
 10.5 Coupled-Inductor PCM-Controlled Superbuck 331

where the parasitics are neglected. The coefcients a0 and a4 are always
positive and, therefore, all the other polynomial coefcients and the left-column
elements of the RouthHurwitz array have to be also positive. The positive signs


of the coefcients a3 , a2 , and a1 require that C1 > DD Ts I o
2Uin
= C1min , LL21 DD
,


and C2 > DD 2Uin
L2 or C2 > LL1 L2 2 C1min , respectively. If the converter is
Ts Io L1 L2
p p

designed such that L2 = D max

D
max
L1 , where Dmax is the dened maximum duty
ratio of the converter operation as well as that C1 and C2 are chosen to be
much larger than their above-dened minimum values then the array element
b1 can be given by
 
1 1 Lp 1 1 1 1
+ (D
2 ) = + D
max (D
max Dmax )
L p C2 (L1 + L2 ) C1 L p C2 C1

which is a positive number if C1 /C2 > D
max (Dmax D
max ). This requirement
is easily met in the practical design, because |D
max (Dmax D
max )|max = 0.125.
Similarly, the array element c1 can be given by

Fm Uin 1 1
= Fm Uin 1 1


L 1 L 2 C1 Lp C2 L1 L2 C1





 C2
1 + D
2 1 + Dmax Dmax Dmax
L1 + L2 C1 C1

which indicates a negative sign to appear if Dmax > 0.5. In reality, the effect of
the parasitics is such that the negative sign may not appear but the situation
should be carefully checked. The sign of the elements b2 and d1 is always
positive because of equaling the zeroth-order polynomial coefcient.
The dynamic issues covered above gave some basic guidelines for the
design, which with the peak-to-peak inductor-current ripple and load response
denitions would led to an optimal design of the PCM-controlled converter.

10.5
Coupled-Inductor PCM-Controlled Superbuck

10.5.1
Small-Signal Models

The small-signal modeling of the coupled-inductor PCM-controlled superbuck

converter can be carried out similarly to the modeling of the basic PCM
control presented in Section 10.4.1. We assume herethat the input-ripple-
reduction conditions are in effect (i.e., M = L2 and k = L2 /L1 ). The coupling
of inductors and the ripple-reduction conditions naturally change the ripple
of the inductor-current feedback (i.e.,
iL ). The coupled-inductor-affected up
(mi1 ) and down (mi2 ) slopes dening
iL (see Section 10.4.1) can be obtained
from the basic VMC on- and off-time state spaces by applying the algorithm
dened in Section 10.3 (10.25). The application of the algorithms yields
 332 10 The Fourth-Order Converter Superbuck
 
dd
Ts rd rds rL2 + rd rds uC1  + UD

iL = iL1  + iL2  + (10.55)
2 L2 L2 L2
We dene the required duty-ratio constraints by linearizing the comparator
equation
ico  mc dTs = iL1  + iL2  +
iL (10.56)
which yields
1
Fm = 
(D
D)

Ts Mc + (Uin + DUD + (rd rds DrL1 + 2D rL2 )Io )
2L2
DD
Ts
qL1 = 1 + (rd rds )
2L2
DD
Ts
qL2 = 1 + (rd rds + rL2 ) (10.57)
2L2
DD
Ts
qC1 =
2L2
qC2 = qin = qo = 0
The main difference in the duty-ratio constraints in (10.57) compared to those
presented in Section 10.4.1 (Eq. (10.46)) is that they are only dependent on L2 .
The coupled-inductor PCMC state space can be obtained from the
corresponding VMC state  space (Section 10.3.3, Eq. (10.35)) by substituting
the perturbed duty ratio d by means of
 
d = Fm ico qL1 iL1 qL2 iL2 qC1 uC1 (10.58)

which yields

R + F q
U
R2 + Fm qL2
U1 1 + Fm qC1
U1

1 m L1 1
0
diL1

L12
L12
L12
dt
L1
D + D
Fm qC1
U2 iL1
diL2
R3 + Fm qL1
U2
R4 + Fm qL2
U2 L2 1
iL2
dt =
L12
L12
L12 L2
u
duC1 D
+ Fm qL1 Io D Fm qL2 Io Fm qC1 Io C1
0
dt uC2
C1 C1 C1
duC2 1 1
dt 0 0
C2 C2
1 Fm
U1
0

L12
L12

1 rC2 Fm
U2
uin

L12 L
L
+
2 12 io
0 F m Io

0 d
C1
1
0 0
C2
 10.5 Coupled-Inductor PCM-Controlled Superbuck 333
i
  L1


uin
1 0 0 0 iL2
iin
= d + 0 0 0
io (10.59)
uo 0 0 0 1 + rC2 C2 uC1 0 0 0
dt d
uC2

where
R1
R4 ,
L12 ,
U1 , and
U2 are as follows:

R1 = R1 R2 ,
R2 = R2 R3
L1 L1

R3 = R2 R1 ,
R4 = R3 R2
L2 L2
(10.60)

L12 = L1 L2 ,
L1

U1 = U1 U2 ,
U2 = U2 U1
L2

and R1 R3 , U1 and U2 are as dened in (10.8) (Section 10.2.2).

The corresponding transfer functions can be solved most conveniently from
(10.59) by using proper software packages such as Matlab with Symbolic
Toolbox. The resulting symbolic transfer functions with the parasitic elements
are extremely long and do not easily offer the desired information. From the
dynamic point of view, the parasitics are, however, essential and cannot be
neglected, because they will provide the damping of the resonances and also
shift the appearance of the possible RHP zeros and poles. We give here only the
nonparasitic transfer functions for introducing the combined effect of PCM
control and coupled-inductor technique in the dynamics of the converter. The
effect of the parasitics will be discussed when appropriate.
Input dynamics:


1 Fm Uin C1 qC1 Io L2

Yino = s3 + s2
L 1 L2 L 2 C1

C1 + (D2 DFm (qC1 Uin + Io ))C2 Fm qC1 Io
+s
L 2 C1 C2 L 2 C1 C2
D Fm Io

Toio = (10.61)
(L1 L2 )L2 C1 C2

Fm Io DUin 1

Gci = s2 + s +
(L1 L2 )C1 I o L2 L 2 C2

Output dynamics:

D Fm qC1 Uin

Gioo =
(L1 L2 )L2 C1 C2


1 F Uin C1 qC1 Io L2
2 m

Zoo = s +s
3
C2 L 2 C1
334 10 The Fourth-Order Converter Superbuck

D2 L1 + D
(1 + D)L2 DFm (qC1 Uin + Io )(L1 L2 )

+s
(L1 L2 )L2 C1

Fm Uin
+
(L1 L2 )L2 C1

Fm Uin 2 DIo 1

Gco = s s + (10.62)
L 2 C2 Uin C1 (L1 L2 )C1
where the determinant (
) is


3 Fm Uin C1 qC1 Io L2

=s +s4
L 2 C1
(C1 + (D2 DFm (qC1 Uin + Io ))C2 )(L1 L2 ) + L2 C2
+ s2
(L1 L2 )L2 C1 C2


Fm Uin C2 qC1 Io (L1 L2 ) 1
+s + (10.63)
(L1 L2 )L2 C1 C2 (L1 L2 )L2 C1 C2
Thus the special input impedances (Zin and Zinsc ) are
s2 (L1 L2 )Uin C1 s DIo (L1 L2 ) + Uin
Zin =
s Uin C1 DIo
s3 (L1 L2 )L2 C1 + s2 E1 + s E2 + Fm Uin
Zinsc =

s2 L2 C1 + s Fm Uin C1 qC1 Io L2 + D(D Fm (qC1 Uin + Io ))


E1 = Fm Uin C1 qC1 Io L2 (L1 L2 )
E2 = (D2 DFm (qC1 Uin + Io ))(L1 L2 ) + L2 (10.64)

where Zin is the same as presented in Section 10.3.3, Eq.(10.39).

Comparing the above given transfer functions in (10.61)(10.64) to the cor-
responding VMC transfer functions given in Section 10.3.3 ((10.36)(10.39)),
we can conclude that the PCM control
1. changes the low-frequency value of the open-loop input impedance
(Zino ) to the corresponding negative incremental resistor dened as
Fm qC1 Io ,
2. increases the low-frequency gain of the control-to-input-current and
control-to-output-voltage transfer functions and the output impedance,
3. attenuates the input-voltage noise, which is maximized when the
compensation (Mc ) is designed to be such that D Fm qC1 Uin 0 or
Mc Uo /2L2 . As a consequence of this, the low-frequency input
impedance corresponds to Uin /DIo .
The described changes are similar to the changes in the conventional buck
converter discussed in Chapter 4.
When the optimal compensation Mc Uo /2L2 is applied, the nonparasitic
dynamic description can be given as follows:
 10.5 Coupled-Inductor PCM-Controlled Superbuck 335

Input dynamics:

DIo
s
1 Uin C1
Yino =
L1 L2 DIo 1
s2 s +
Uin C1 (L1 L2 )C1
2L2 Io
D(1 )
Ts D
Uo

Toio = (10.65)
(L1 L2 )L2 C1 C2

2L2 Io DUin 1

Gci = s +s
2
+
Ts D
Uo (L1 L2 )C1 I o L2 L 2 C2

Output dynamics:

Gioo = 0

1 3 2Uin C1 DD
Ts Io

Zoo = s + s2
C2 D
Uin Ts C1
2L2 DIo 
D2 L1 + D
(1 + D)L2 (L1 L2 )
Ts D
Uin 2
+s +

(L1 L2 )L2 C1 D Ts (L1 L2 )C1

2 DIo 1

Gco = s2 s + (10.66)
Ts D
C2 Uin C1 (L1 L2 )C1

where the determinant (
) is

 
2 1 DIo 1

= s2 + s
+ s2 s + (10.67)
D Ts L 2 C2 Uin C1 (L1 L2 )C1

Thus Zin , Zinsc , and Zino (see (10.65)) becomes equal, because Gioo = 0,
and can be naturally given by

DIo 1
s2 s +
Uin C1 (L1 L2 )C1
Zin,sc,o = (L1 L2 ) (10.68)
DIo
s
Uin C1

The equality of Zin and Zinsc may be difcult to determine from (10.64)
even if the optimal-compensation conditions are substituted in them. It is,
however, possible to notice that the common factor of the numerator and
L2
denominator of Zinsc is s + Fm . Same difculty also applies with the
Uin
open-loop input impedance in (10.61) but the corresponding common factor
is the rst polynomial multiplier of the determinant given in (10.67).
336 10 The Fourth-Order Converter Superbuck

10.5.2
Design Considerations

The PCM-controlled converter has the same RHP zeros as the corresponding
VMC converter as introduced in Section 10.3.4. In order to avoid the appearance
of the zeros, the design of the converter should be made as implied by (10.40)
in Section 10.3.4.
According to the determinant (
) in (10.63), its polynomial coefcients are
as follows:

a4 = 1


Fm Uin C1 qC1 Io L2
a3 =
L 2 C1
(C1 + (D2 DFm (qC1 Uin + Io ))C2 )(L1 L2 ) + L2 C2
a2 = (10.69)
(L1 L2 )L2 C1 C2


Fm Uin C2 qC1 Io (L1 L2 )
a1 =
(L1 L2 )L2 C1 C2
1
a0 =
(L1 L2 )L2 C1 C2
The coefcients a0 and a4 are positive real numbers, which means that all the
other coefcients have to be also positive real numbers. The positive sign of a3


requires that C1 DD Ts I o
2Uin
. The positive sign of a2 can be ensured by choosing
DD
Ts Uin
L1 2Io (1k2 ) , which can be usually easily met. It is obvious that the low input
voltage and high output current may introduce the worst case in design. The


positive sign of a1 can be ensured by choosing C2 DD 2Uin
Ts I o
(1 k2 )k2 , which
is usually easy to meet. The worst case naturally coincides with the low input
voltage and high output current.
The RouthHurwitz-array coefcients b1 , b2 , c1 , and d1 (see Section 10.2.3)
can be given by
1
b1
L 2 C2
1
b2
(L1 L2 )L2 C1 C2
(10.70)
c1 0
1
d1
(L1 L2 )L2 C1 C2
where we assume that C1 and C2 are much larger than their above-dened
minimum values and L1 close to its. The coefcients b1 , b2 , and d1 are clearly
real positive numbers, because L2 < L1 . The coefcient c1 is, however, equal
to zero and this may imply the existence of the RHP poles. The nonparasitic
 10.6 Dynamic Review 337

determinant (
) of the optimally compensated converter

 in (10.67) shows that
a resonant RHP-pole pair may exist at fpRHP = 2 1 1
(L1 L2 )C1
. The control-to-
output (Gco ) transfer function in (10.66) has, however, a resonant RHP-zero
pair at the same frequency, which may cancel the RHP-pole pair but it is difcult
to conrm by means of the symbolic expression with the parasitics included
due to their complexity. The analysis is, however, quite straightforward when
the component values are xed.

10.6
Dynamic Review

Two different superbuck converters were designed as shown in Figure 10.9.

The converter in Figure 10.9a has double the input voltage and one fourth of
the output current of the converter as shown in Figure 10.9b. The switching
frequencies are close to 400 kHz. Basically the parallel inductances were
dened giving the peak-to-peak output ripple current of 40% of the average
inductor current at the maximum output power (see Section 10.2.1, Eq. (10.6)).
The parallel inductance of the rst converter was computed to be 7.5 H. The
inductors were chosen to be equal and approximately of 15 H. The parallel
inductor of the second converter was computed to be 1.2 H. It was noticed
that the inductors cannot be anymore equal due to the appearance of RHP
zeros and poles as discussed in Section 10.4.2.
Therefore, L2 D max
D
max
L1 at Dmax = 0.75. According to this, L1 and L2 were
chosen to be 1.6 and 5.0 H, respectively. The approximated practical values
are shown in Figure 10.9b. All the capacitors are multilayer ceramics and,
therefore, their real capacitance values are dependent on the voltages across
them. Similarly, the inductance values are dependent on the current owing
through them. This means that the component values are highly dependent
on the operating point, which should be carefully considered when analyzing
the dynamics of the converter.
The inductor-current feedback was compensated by using the optimal
compensation dened in Section 10.4.2, but the actual level of compensation
would deviate from the desired value.
The current-mode-controlled converters cannot operate at open loop at the
constant-current load due to their current-output nature. Therefore, all the
other frequency responses than the output impedance have to be measured
at resistive load as depicted in Figure 10.9. The internal dynamics has to be
computed according to the load-affected transfer functions dened in Chapter
2, Section 2.4.1 as discussed in Chapter 4.

10.6.1
Superbuck I: 1520 V/10 V/2.5 A

The dynamics of the converter were measured at the input voltages of 15 and
20 V at the maximum load power, where the corresponding duty ratios are
338 10 The Fourth-Order Converter Superbuck

Figure 10.9 PCM-controlled

superbuck converters: (a) High
input voltage and low output
current and (b) Low input
voltage and high output
current.

0.688 and 0.344, respectively. Figure 10.10 shows the measurement-data-based

computed internal control-to-output (Gco ) transfer functions. The RHP zero
and/or pole should be present if existing, when the duty ratio is greater than
0.5 but Figure 10.10 proves that the parasitics have moved the appearance
of the zeros and poles to higher duty ratios. The control-to-output transfer
function is clearly of rst order and resembles greatly the control-to-output
transfer function of the conventional buck converter analyzed in Chapter 4,
Section 4.5.1.
 10.6 Dynamic Review 339

Figure 10.10 The computed internal control-to-output transfer functions at the input
voltage of 15 V (dashed line) and 20 V (solid line).

The measured and predicted output impedance (Zoo ) at the input voltage
of 20 V are shown in Figure 10.11. It is clearly of rst order and quite similar
to the output impedance of the conventional converter shown in Chapter 4,
Section 4.5.1. The behavior of the pure output dynamics is quite expected,
because the sum of the inductor currents is tightly regulated and supplied
during the on- and off-times to the output section.

Figure 10.11 The measured (solid line) and predicted (dashed line) internal output
impedance at the input voltage of 20 V.
340 10 The Fourth-Order Converter Superbuck

Figure 10.12 The measured load-affected (GRioo , solid line), the

computed internal (Gioo , dashed line), and predicted internal
(Gioo , dash-dot line) input-to-output transfer functions at the
input voltage of 20 V.

The measured load affected, the computed and predicted internal input-
to-output transfer functions (Gioo ) are shown in Figure 10.12. The internal
dynamics contains clearly the resonant behavior as discussed earlier. The
measured load-affected transfer function (GRioo ) has clearly much higher
attenuation than the real internal attenuation (Gioo ) because of the voltage-
dividing effect of the output impedance and the load resistor.
The state of the inductor-current compensation can be actually concluded
based on the behavior of Gioo : The under-compensation is reected as the
phase starting from 180 and also increased damping of the resonances.
The over-compensation is reected as the phase starts from zero. In both of
the cases, the attenuation is naturally reduced from the theoretical optimum
compensation as shown in Figure 10.13. According to the gure, the actual
converter is slightly over compensated when Mc 0.835 A/s.
The measured and predicted open-loop input impedances are shown in
Figure 10.14 at the input voltage of 20 V, when the converter is equipped
with an input capacitor (i.e., the high-frequency behavior is due to the input
capacitor). The shape of the input impedance resembles the shape of the open-
loop input impedance of the VMC converter analyzed in Chapter 3, Section
3.6.1. Its behavior does not, however, mean similar sensitivity to the input
lter as in the VMC converter because of the high-input-noise attenuation (see
Figure 10.12) as discussed in Chapter 8.
The shape of the control-to-output transfer function implies (see Chapter 9)
that a proportional-integral (PI) controller with an additional high-frequency
pole can be used to obtain the desired voltage-loop crossover frequency, phase
 10.6 Dynamic Review 341

Figure 10.13 Predicted input-to-output transfer function at the

input voltage of 20 V with varying level of compensation:
Mc = 0.835 A/s; solid line, Mc = 0.5 A/s; dashed line, and
Mc = 0.7 A/s; dash-dot line.

Figure 10.14 Measured (solid line) and predicted (dashed line) internal open-loop input
impedances at the input voltage of 20 V with an added input capacitor.

and gain margins. Usually the absolute maximum crossover frequency is

considered to be one fth of the switching frequency, which would limit the
crossover frequency to 88 kHz in this application. Figure 10.10 shows that the
phase lack starts rapidly increasing at the high frequencies and, therefore,
342 10 The Fourth-Order Converter Superbuck

the crossover frequency should be chosen less than the dened absolute
maximum value for maintaining robustness of the stability and performance:
The measured voltage-loop gains at the input voltages of 15 and 20 V are
shown in Figure 10.15 indicating the crossover frequency of 63 kHz and phase
margin of 66 . The loop gain stays almost intact regardless of the changes in
the input voltage, which is characteristic to the PCM-controlled buck converter.
The closed-loop input impedance was measured yielding the response
shown in Figure 10.16 (solid line) compared to the open-loop input impedance
(dashed line). It shows that the closed-loop and open-loop input impedances
are the same and, therefore, the ideal and short-circuit impedances are also
the same and equal to the impedances shown in Figure 10.16. The resonant
nature of the closed-loop input impedance may make the superbuck more
sensitive to EMI lter instability than the conventional buck converter, where
the closed-loop input impedance is a frequency-independent constant (see
Chapter 4, Section 4.5.1).
The converter was equipped with an input lter causing impedance violation
as shown in Figure 10.17. The Nyquist plot of the minor-loop gain (i.e., Zs /Zinc )
in Figure 10.18 proves that the converter is stable at the input voltage of 15 V
but the instability takes place, when the input voltage is lowered to 11.5 V.
The voltage-loop gains shown in Figure 10.15 are actually measured when
the input lter was connected. The small deviation in the phase curve at the
input-lter resonant frequency of 13 kHz is the only sign of the interaction
proving the existence of high-source insensitivity.
The input-side resonance can be damped by connecting a series RC circuit
in parallel with the capacitor C1 (Figure 10.9) in order to further reduce the

Figure 10.15 The measured output-voltage-loop gains at the input voltage of 20 V (solid
line) and 15 V (dashed line).
 10.6 Dynamic Review 343

Figure 10.16 The measured closed-loop (solid line) and

open-loop (dashed line) input impedances at the input voltage of
15 V from which the effect of the input capacitor is removed.

Figure 10.17 The measured input-lter output impedance (dashed line) and the
closed-loop input impedance (solid line) of the converter at the input voltage of 15 V.

sensitivity to input-lter instability as shown in Figure 10.19 (i.e., the damping

composes of series connection of an 1- resistor and a 40-F capacitor).
The measured open-loop (dashed line) and closed-loop (solid line) output
impedances of the converter are shown in Figure 10.20, where the magnitude
of the closed-loop impedance at the loop crossover frequency (63 kHz) is
approximately 100 m.
344 10 The Fourth-Order Converter Superbuck

Figure 10.18 The Nyquist plots the impedance-based minor-loop gains (i.e., Zs /Zinc ) at
the input voltage of 15 V (solid line) and 11.5 V (dashed line).

Figure 10.19 The original (solid line) and damped (dashed line) input impedances at the
input voltage of 15 V.

The converter was subjected to a constant-current type load change from 0.5
to 2.5 A with a slew rate of 2.5 A/s. The resulting output-voltage response
is shown in Figure 10.21: the observed voltage dip is approximately 200 mV,
which corresponds closely to the step change in the load current and the
magnitude of the output impedance (100 m) at the voltage-loop crossover
frequency (Figure 10.20). The setup time of the transient is typical to the
current-mode-controlled converter (i.e., slow) due to the magnitude of the
low-frequency closed-loop output impedance (Figure 10.20).
 10.6 Dynamic Review 345

Figure 10.20 The measured open-loop (dashed line) and closed-loop (solid line) output
impedances.

Figure 10.21 Output-voltage response to the constant-current-type step change in the

load current from 0.5 A to 2.5 A with slew rate of 2.5 A/s.

10.6.2
Superbuck II: 69 V/3.4 V/12 A

The maximum allowed crossover frequency of the superbuck II (Figure 10.9b)

is approximately 72 kHz due to the switching frequency of 360 kHz. The phase
of the control-to-output transfer function (Gco ) decreases rapidly at the higher
346 10 The Fourth-Order Converter Superbuck

Figure 10.22 The measured voltage-loop gains at the input voltage of 6 V (solid line) and
9 V (dashed line).

Figure 10.23 The measured open (solid line) and closed-loop (dashed line) input
impedances at the input voltage of 6 V and at the output current of 12 A.

frequencies and, therefore, the controller was designed to give approximately

30 kHz of the crossover frequency as shown in Figure 10.22 at the input
voltage of 6 V and 9 V, respectively. The phase margin is approximately 70 .
The open- and closed-loop input impedances at the input voltage of 6 V and
at the output current of 12 A are given in Figure 10.23. The high-frequency
behavior of the impedances is due to the input capacitor, which is 1.5 F
 10.6 Dynamic Review 347

Figure 10.24 The measured open (dashed line) and closed-loop (solid line) output
impedances at the input voltage of 6 V.

Figure 10.25 The output-voltage response to the constant-current load change from 2 to
12 A at the slew rate of 2.5 A/s at the input voltage of 6 V.

at open loop and 10 F at closed loop. The impedances are quite the same
and, therefore, the ideal and short-circuit input impedances are the same and
as shown in Figure 10.23 due to the high attenuation of the input-to-output
transfer function.
The measured open (dashed line) and closed-loop (solid line) output
impedances are shown in Figure 10.24, where the magnitude of the closed-loop
348 10 The Fourth-Order Converter Superbuck

output impedance at the voltage-loop crossover frequency (32 kHz) is

approximately 48 m. The open-loop output impedance is clearly of rst
order.
The converter was subjected to a constant-current step load change from 2 to
12 A at the slew rate of 2.5 A/s at the input voltage of 6 V. The corresponding
output-voltage response is shown in Figure 10.25. The output-voltage dip is
approximately 0.48 V, which corresponds to the magnitude of the closed-loop
output impedance at the voltage-loop crossover frequency well.
The steady-state duty ratio of the converter is approximately 0.64. The
controller increases naturally the duty ratio during the transient state. The
time-domain response is quite neat and indicates that the RHP poles are not
activated due to the design of the inductors.

10.7
Summary

The dynamics of the higher order converters tends to be much more

complicated than the industrial designers and even the academic researchers
usually assume as implied also in [17]. The book provides the necessary
understanding of the dynamic processes inside the converter and the
methods to extract the dynamic models, which can be successfully utilized as
demonstrated in this chapter. It may be obvious that the sum of certain vices
and virtues is constant: If we add some benecial circuit structures in order
to remove or reduce something, we can be sure that we have added features,
which may somehow worsen the quality of the converter as a consequence.
Knowing the consequences would minimize the risks.

References

1. R.W. Erickson and D. Maksimovic, J.C. Marpinard, A versatile zero ripple

Fundamentals of Power Electronics, topology, in Proc. IEEE Power
Kluwer, Norwell, MA, USA, 2001, 2nd Electronics Specialists Conf., 1988,
Edition. pp. 133141.
2. V. Grigore and J. Kyyra, A step-down 5. F. Tonicello, The control problem of
converter with low ripple input maximum point power tracking in
current for power factor correction, in power systems, in Proc. 7th European
Proc. IEEE Applied Power Electronics Space Power Conf., 2005, p. 7.
Conf., 2000, pp. 188194. 6. E. Karapete, M. Boztepe, and
3. W.W. Weaver and P.T. Krein, M. Colak, Development of a suitable
Analysis and application of model for characterizing photovoltaic
current-sourced buck converter, in arrays with shaded solar cells, Solar
Proc. IEEE Applied Power Electronics Energy, vol. 81, no. 8, 2007,
Conf., 2007, pp. 16641670. pp. 977992.
4. A. Capel, H. Spruyt, A. Windberg, 7. S. Liu and R. Dougal, Dynamic
D. OSullivan, A. Crausaz, and multiphysics model for solar array,
 351

Index

a state space (self-osc) 196

Average-current-mode control (ACMC) step-up converter 66
13, 169 Boundary conduction mode (BCM) 56
current-loop amplier 169, 170 critical mode 56
current-loop-amplier transfer function transition mode 56
171 Buck Converter 13, 6466, 8792, 126
duty-ratio generation 171 CCM state space 74, 76, 79
general input-to-output transfer DCM state space 88
functions 173 Duty-ratio constraints (PCMC, CM)
PI-type controller 170, 171 126
Averaged state space 63 Duty-ratio constraints (PCMC, DCM)
time averaged output current 64 141
time-averaged capacitor voltage 63 input-to-output transfer function 75,
time-averaged inductor current 63 91
time-averaged input current 64 input-to-state transfer function
75, 77
b input-to-output transfer function
Bode plot 46, 47 ff., 51 ff. (current-output mode) 214
phase margin (PM) 48, 51 ff. input-to-output transfer function
gain margin 48 ff. (PCMC, CCM) 132
Boost Converter 6668, 126128, input-to-output transfer function
133134, 142 (self-osc) 195
DCM state space 93 operating point 88
Duty-ratio constraints (PCMC, CCM) state space (current-output mode)
126 214
Duty-ratio constraints (PCMC, DCM) state space (PCMC, CCM) 132
142 state space (PCMC, DCM) 144
input-to-output transfer function state space (self-osc) 195
78 ff., 93 BuckBoost Converter 6870, 9495
CCM state space 77 ff. 128131
input-to-output transfer function CCM state space 80 ff.
(PCMC, CCM) 133 DCM state space 94
input-to-output transfer function Duty-ratio constraints (PCMC, CCM)
(self-osc) 196 128
operating point 78, 93 Duty-ratio constraints (PCMC, DCM)
RHP zero 196 143
state space (PCMC, CCM) 133, 144 yback converter 68

Dynamic Prole of Switched-Mode Converter. Teuvo Suntio

2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
ISBN: 978-3-527-40708-8
352 Index

BuckBoost Converter (contd.) Open-loop G-parameters 284

input-to-output transfer function PCMC boost converter 298
80 ff., 95 PCMC buck converter 290
input-to-output transfer function VMC boost converter 295
(PCMC, CCM) 135 VMC buck converter 285
input-to-output transfer functions voltage-loop gain (L(s)) 284
(self-osc) 197 zero pole placement 287, 291, 296,
operating point 94, 80 300
RHP zero 197 Control dynamics 17
state space (PCMC, CCM) 135 Control variable (c) 227
state space (PCMC, DCM) 144 reference voltage (ur ) 227
state space (self-osc) 197 Control-block diagram
step-up/down converter 68 closed-loop converter 25
current-output converter 23 ff.
c input dynamics 23 ff., 285
Canonical equivalent circuit 4, 5 ff., 8, output dynamics 23 ff., 285
9 ff. voltage-output converter 23 ff.
G-parameters 8 Coupled-induction superbuck converter
Y-parameters 8 (PCMC)
current-output converter 8, 9 ff. duty-ratio constraints coefcient 332
nonideal load 9 ff. ideal input impedance 334
nonideal source 9 ff. input dynamics 333
Norton equivalent circuit 8 optimal inductor current compensation
Thevenin equivalent circuit 8 334
two-port model 8 output dynamics 333
voltage-output converter 8, 9 ff. short-circuit input impedance 334
Characteristic polynomial 46 state space 332
Closed-loop converter 23, 24 ff. Coupled-inductor superbuck converter
control-block diagram 25 (PCMC)
current loop 24 RHP pole 336
current-output converter 24, Routh-Hurwitz array 336
25 ff. Coupled-inductor superbuck converter
(VMC)
external feedback 23
ideal input impedance 321
internal loop gain 24
input dynamics 320
loop gain 25
input-current-ripple reduction 323
multiloop control system 24
output dynamics 321
sensor gain Gse 25
RHP pole 322
voltage loop 24
RouthHurwitz array 322
voltage-output converter 25 ff.
Complementary sensitivity function T(s) short-circuit input impedance 321
46 state space 320
Conditional stability 50 Coupled-inductor technique
Continuous conduction mode (CCM) 4, coupling coefcient 319
55 leakage inductance 319
Control bandwidth 49, 52 ff. magnetizing inductance 319
complementary sensitivity function modeling algorithm 319
(T(s)) 49 mutual inductance 319
loop crossover frequency (fgco ) 52 self-inductances 319
Control design 261 terminal equation 318
Closed-loop G-parameters 284 Cuk converter 308
control block diagram 285 Current-Output Converter 2728, 213 ff.
 Index 353

cascaded control system 218 voltage-mode control (VMC) 59

closed-loop dynamic prole 27 Discontinuous conduction mode (DCM)
duality transform 215 5, 56
general load 213 Dynamic analysis 17
general transfer function 215 Dynamic prole 1, 2, 4
(ideal input admittance) 32 analog control 4
Load effect 217 digital control 4
loop crossover frequency 211 internal dynamics 3
nonideal load 217 internal models 2
nonideal source 217 resonant behavior 3
(open-circuit input admittance) 32 two-port network 2
Source effect 217 two-port parameters G, Y, H, Z, 2
state space averaging 213
Current-sourced converter 2 ff. f
current-output mode 2 ff. Fixed-frequency operation 5, 60
voltage-output mode 2 ff. Forbidden region 6
Cycle-time constraints 193
g
d Gain margin (GM) 6, 48
DC UPS system 212 ff., 261 General transfer function (VMC, CCM)
Determinant 21 boost converter 83
Diode-Switched Boost Converter 67, buck converter 83
7778 buck-boost converter 83
ideal input admittance 78 CCM 83
input-to-output transfer function xed frequency operation 83
78 ideal input admittance 85
operating point 78 RHP zero 83
(short-circuit input admittance) short-circuit input admittance 85
78
state space 77 i
Diode-Switched Buck 77 Ideal input admittance 30, 32
ideal input admittance 77 Ideal output dynamics
operating point 77 control block diagram 263
short-circuit input admittance 77 ideal input admittance 263
Diode-Switched Buck Converter 65, 76 ideal reverse transfer gain (Toi )
state space 76 263
Diode-Switched BuckBoost Converter input voltage feedforward gain (qi )
69, 8081 263
ideal input admittance 81 output current feedforward gain (qo )
input-to-output transfer function 80 263
operating point 80 Input EMI lter
short-circuit input admittance 81 converter input impedance 237
state space 80 lter output impedance 237
Direct-duty-ratio control 5 input-to-output transfer function 237
Direct-on-time control 12, 59, 60 matrix model 237
boost converter 68 minor-loop gain 237
buck converter 66 two-port model 237
buck-boost converter 70 Input variable
comparator equation 61 input vector 20
modulator gain 61 Input voltage feed forward (IVFF)
time-averaged variables 59 control block diagram 239
354 Index

Input voltage feed forward (IVFF) (contd.) Load transient 266

duty-ratio constraints 239 Loop crossover frequency (fgco ) 49
duty-ratio gain (FM ) 239
Duty-ratio generation 239 ff. m
feed forward gain (qi ) 239 Minor-loop gain 6, 7, 226
matrix model 238 impedance overlap 7
Input-output stability 230 stability 7
Input-output transfer matrix transient performance 7
audiosusceptibility 21 Mode limit, CCM 136
control-to-input transfer function 21 Mode limit, DCM 145
control-to-output transfer function 21 Modulator gain (Ga ) 62
input admittance 21 Multiloop control system 24
line-to-output transfer function 21 internal loop gain 24
output admittance 21 overload protection 24
output impedance 21
output-to-input transfer function 21
n
Input-voltage-feedforward (IVFF) 19, 238
Nyquist 51 ff.
Interconnected system 233 ff.
Nyquist plot
beat frequency 257
gain margin (GM) 51 ff.
cascaded system 231 ff.
characteristic polynomial 231
ESAC criterion 6 o
forbidden region 6, 232 On-time constraints 195
gain margin 6 Open-circuit input admittance 32
Open-loop converter 17
ideal input admittance 229
average-current-mode control 18
input impedance 232
current-output converter 17
input-output stability 231
external feedback 17
Internal stability 225, 231
internal feedback 18
Load effect 228
internal feedforward 18
load subsystem 225
peak-current-mode control 18
minor-loop gain 6, 225, 232
voltage-output converter 17
Nyquist stability criterion 6, 232, 233
Operational amplier
short-circuit input admittance 229
GBW product 270
Source effect 229
internal gain 270
source impedance 232
Optocoupler
stability analysis 225
current transfer ratio (CTR)
supplying subsystem 225 274
system interface 233 cut-off frequency 274
Intermediate-bus-architecture (IBA) 261,
delay (td ) 274
228 ff.
rise time (tr ) 274
Internal stability 226, 230
transfer function 274
Output variable
j output vector 20
Jacobian matrix 38 Output voltage feedback loop 265
bias resistor (Rb ) effect 272
l gain margin peaking 266
Linearization 3738 load transient 266
Jacobian matrix 38 loop crossover frequency 268
Load effect 28 operational amplier gain 270
current-output converter 32 Optimal loop shape 266 ff.
voltage-output converter 30 optocoupler 274
 Index 355

phase margin peaking 266 PCM converter 10

PI controller 272 push-pull converter 125
PID controller 270 subharmonic mode 121
RHP pole 268 Transformer isolation 124
RHP zero 268 Phase margin (PM) 48
Output voltage remote sensing 7 PI controller
cabling impedance 235 transfer function 272
impedance block (Zcon ) 234, 235 ff. PID controller
LCL impedance 236 Bode plot 273
load transient performance 234 transfer function 270
short-circuit input admittance 235 Polar plot, 46 47 ff.
static voltage accuracy 234 Pulsewidth modulation (PWM) 5
two-port model 234 ff.
Output-current-feedforward (OCF) 10 r
buck converter 241 RHP zero 50, 52 ff.
control block diagram 240
matrix model 241 s
output current loop gain (Hi ) 240 Second-Order Transfer Function 4043
Overload protection critically damped 41
constant-current limiting 211 damped natural frequency 42
modied constant-power limiting damping factor 40
211 oscillatory 41
overdamped 41
p quality factor 40
PCMC boost converter undamped natural frequency 40
control design 298 underdamped 41
RHP zero 298 Self-oscillation control 13, 189
voltage loop gain 298 cycle-time constraints 193
zero pole placement 300 general averaged state space 191
PCMC buck converter on-time constraints 195
control design 290 steady-state cycle time 193
voltage loop gain 290 steady-state equivalent circuit 191 ff.
zero pole placement 291 variable-frequency operation 13, 189
Peak-current-mode control (PCMC) 12 Sensitivity function S(s) 46
active reset forward converter 125 SEPIC converter 308
averaged inductor current 124 Short-circuit input admittance 30
comparator equation 123 Shunt regulator TL431 13, 274
compensation ramp (Mc ) 122 combined loops 282
control current (ico ) 122 control block diagram 283
duty-ratio constraints 123 control system frequency response
duty-ratio gain (FM) 123 281
feed forward gain 123 dynamic model 275
feedback gain 123 fast loop 283
forward converter 125 fast loop frequency response 277
full-bridge converter 125 frequency response measurement
general input-to-output transfer 275
function 129 ideal frequency response 277
half-bridge converter 125 open loop frequency response 275
mode limit CCM, 136 output impedance (Zo ) 275
mode limit DCM, 145 slow loop 275, 283
multi-inductor-current feedback 124 slow loop frequency response 277
356 Index

Shunt regulator TL431 (contd.) two-inductor buck 13

transconductance (gm ) 275, 277 two-inductor buck converter 307
transfer function (gm ) 278 Superbuck converter (PCMC) 338
two loop control system 281 comparator equation 326
two-port model 282 determinant 329
Single-pole transfer function 40 duty-ratio constraints 326
left-half-plane (LHP) pole 40 duty-ratio constraints coefcient 327
right-half-plane (RHP) pole 40 ideal input impedance 329
Single-zero transfer function 39 input dynamics 328
left-half-plane (LHP) zero 39 maximum duty ratio 327
right-half-plane (RHP) zero 39 optimal inductor-current compensation
Small-signal equivalent circuit 84 ff. 330
boost converter 83 output dynamics 329
buck converter 83 RHP pole 330
buck-boost converter 83 RHP zero 330
CCM 83 state space 328
xed-frequency operation 83 Superbuck converter (VMC)
Source effect 28 Averaged state space 311
current-output converter 32 determinant 314
voltage-output converter 30 ideal input impedance 314
Stability 4648 input dynamics 313
Bode plot 46 off-time state space 310
characteristic polynomial 46 on-time state space 310
conditional stability 48 output dynamics 314
gain margin 48 resonant frequency 314
marginal stability 46 RHP pole 316
phase margin 48 RHP zero 317
Polar plot 46 RouthHurwitz array 316
State space RouthHurwitz test 315
input variable 19 short-circuit input impedance 315
output variable 19 small-signal state space 312
state variable 19 steady-state operating point 311
transfer matrix 20 Switched-mode converters 1
State variable current-to-current converter 1
state-variable vector 20 current-to-voltage converter 1
State-space-averaging (SSA) 4, 59 voltage-to-current converter 1
state-transition matrix 21 voltage-to-voltage converter 1
Steady-state equivalent circuit 82 ff. Synchronous Boost Converter 67, 7980
boost converter 82 ideal input admittance 80
buck converter 82 input-to-output transfer function 80
buck-boost converter 82 operating point 80
CCM 82 short-circuit input admittance 80
xed-frequency operation 82 state space 80
Superboost converter Synchronous buck converter 65, 71 ff.
RHP pole 269 determinant 75
RHP zero 269 ideal input admittance 76
Superbuck converter 13 input-to-output transfer function 75
current-sourced buck 13 input-to-state transfer function 75
current-sourced buck converter 307 operating point 76
fourth-order converter 13 short-circuit input admittance 76
step-down buck converter 307 state space 74
 Index 357

Synchronous BuckBoost Converter 69, v

8182 Variable-frequency operation 5, 60
ideal input admittance 82 VMC boost converter
input-to-output transfer function 81 control design 295
operating point 81 resonant frequency 296
short-circuit input admittance 82 RHP zero 296
state space 81 voltage loop gain 295
zero pole placement 296
VMC buck converter
t
control design 285
Transfer function 38
resonant frequency 287
angular frequency 38
voltage loop gain 285
denominator polynomial 38
zero pole placement 287
frequency 38
Voltage-mode control (VMC) 59
numerator polynomial 38
direct-duty-ratio control 59
pole 39
VMC converter 5
zero 39
VMC operation 5
Two-port model 22 ff., 227
Voltage-Output Converter 2627,
closed loop 227
213 ff.
current-output converter 22 ff.
closed-loop dynamic prole 26
nonideal load 29 ff.
general load 213
nonideal source 29 ff.
ideal input admittance 30
Norton equivalent circuit 21
short-circuit input admittance 30
open loop 227
Voltage-sourced converter 2 ff.
Thevenin equivalent circuit 22
current-output mode 2 ff.
two-port network 21
voltage-output mode 2 ff.
voltage-output converter 22 ff., 227