Improved Models for DC-DC

Converters

Bengt Johansson

Licentiate Thesis
Department of Industrial Electrical Engineering
and Automation
Department of
Industrial Electrical Engineering and Automation
Lund University
P.O. Box 118
SE-221 00 LUND
SWEDEN

http://www.iea.lth.se

ISBN 91-88934-29-2
CODEN:LUTEDX/(TEIE-1037)/1-365/(2003)

©Bengt Johansson, 2003

Printed in Sweden by Media-Tryck
Lund University
Lund, 2003

ii
Abstract

To obtain high performance control of a dc-dc converter, a good model

of the converter is needed. It is suitable to consider the load to be included in
the converter model since it usually affects the dynamics. The load is often
the most variable part of this system. If the load current and the output
voltage are measured there are good possibilities to obtain a good model of
the load on-line. Adaptive control can then be applied to improve the control.
In peak current-mode control, the output voltage and the inductor
current are measured and utilized by the controller. This thesis analyses some
properties that can be obtained if the load current is also measured and
utilized for control. Accurate expressions for the control-to-output transfer
function, the output impedance, and the audio susceptibility are derived for
the buck, boost, and buck-boost converters operated in continuous
conduction mode in the case where the load is a linear resistor. If the
measured load current is utilized properly by the controller, the output
impedance becomes low and the control-to-output transfer function becomes
almost invariant for different loads. The use of load current acts as a
feedforward term if the load is a current source. However, if the load is a
resistor the load current is influenced by changes in the output voltage and
the stability is affected. Therefore, the use of load current is not a feedforward
action in this case. Instead it can be seen as gain scheduling, which can be
considered a special case of adaptive control.
In the thesis it is also shown that the two published models for current-
mode control, Ridley (1991) and Tan and Middlebrook (1995), give accurate
expressions for the control-to-output transfer function and the output
impedance but not for the audio susceptibility. A novel model for the audio
susceptibility is presented and it is used to improve the two published models.
Most of the results in the thesis are validated by comparing the frequency
responses predicted by the expressions and switched large-signal simulation
models.

iii
iv
Acknowledgements

First of all, I would like to thank my supervisors Professor Gustaf Olsson

and Dr Matz Lenells for guidance during my work and for comments on the
draft of this thesis. Thanks also to Dr Per Karlsson for reading and
commenting the draft.
Furthermore, I would like to thank Johan Fält, Ericsson Microwave
Systems AB, and Mikael Appelberg, Ericsson Power Modules AB, for
discussions about practical problems with dc-dc converters and for helping
me with measurements.
I would also like to thank the staff at the Department of Technology,
University College of Kalmar, and the staff at the Department of Industrial
Electrical Engineering and Automation (IEA), Lund University, who have
helped me in many ways.
This work has been financially supported by The Knowledge Foundation,
University College of Kalmar, and Ericsson Microwave Systems AB.
Finally, I would like to thank my family for supporting me.

Kalmar, October 2003

Bengt Johansson

v
vi
Notation

Frequently used signals and parameters are presented with name and
description in the list below. Signals and parameters that only appear in one
of the chapters are not included in the list. The names of signals consist of
lower-case letters. However, exceptions are made for the subscript part of the
names. The names of the signals are also used to denote their dc values but
capital letters are used in this case. However, the letters in the subscript part
are not changed. The dc value names are not included in the list.

Name Description

C Capacitance of the capacitor

d Duty cycle
d' d'=1− d
ic Current reference
icap Capacitor current
ie External ramp used for slope compensation
iinj Current injected into the output stage
iL Inductor current
iload Load current
kf Input voltage feedforward gain (see Section 3.4)
kr Output voltage feedforward gain (see Section 3.4)
L Inductance of the inductor
m1 Slope of the inductor current while the transistor in on
− m2 Slope of the inductor current while the transistor in off
mc Relative slope of the external ramp, m c = 1 + M e M 1
Me Slope of the external ramp
R Resistance of the load resistor
Rc Equivalent Series Resistance (ESR) of the capacitor
Ts Switching period

vii
vg Input voltage
vo Output voltage
v ref Voltage reference
δ Control signal of the transistor driver
ωn Half the switching frequency, ω n = π Ts

Signals are often divided into a dc part and an ac part. The ac part is
denoted by using the hat-symbol (^). As mentioned earlier, the dc part is
denoted by using capital letters. To explicitly denote that a signal is a function
of time, the variable t is added to the name, i.e. signalname(t ) . The
sampled version of a continuous-time signal is denoted by replacing the
variable t with n . The Laplace transform of a continuous-time signal is
denoted by replacing the variable t with s . The Z-transform of a discrete-
time signal is denoted by replacing the variable n with z .
The notation is to some extent chosen such that it is compatible with the
one used by Ridley (1991).

viii
Contents

CHAPTER 1 INTRODUCTION........................................................1
1.1 BACKGROUND ..........................................................................1
1.2 MOTIVATION FOR THE WORK ..................................................5
1.3 MAIN CONTRIBUTIONS.............................................................7
1.4 A GUIDE FOR THE READER AND THE OUTLINE OF THE
THESIS ......................................................................................9
1.5 PUBLICATIONS........................................................................10
CHAPTER 2 STATE-SPACE AVERAGING ................................11
2.1 INTRODUCTION .......................................................................11
2.2 OPERATION OF THE BUCK CONVERTER .................................12
2.3 MODEL OF THE BUCK CONVERTER ........................................15
2.4 SIMULATION OF A BUCK CONVERTER ...................................33
2.5 OPERATION OF THE BOOST CONVERTER ...............................39
2.6 MODEL OF THE BOOST CONVERTER ......................................44
2.7 SIMULATION OF A BOOST CONVERTER ..................................55
2.8 OPERATION OF THE BUCK-BOOST CONVERTER .....................58
2.9 MODEL OF THE BUCK-BOOST CONVERTER ...........................61
2.10 SIMULATION OF A BUCK-BOOST CONVERTER .......................71
2.11 SUMMARY AND CONCLUDING REMARKS ..............................74
CHAPTER 3 CURRENT-MODE CONTROL ...............................77
3.1 INTRODUCTION .......................................................................77
3.2 OPERATION OF CURRENT-MODE CONTROL ...........................78
3.3 AN ACCURATE CONTROL-TO-CURRENT TRANSFER
FUNCTION ...............................................................................83
3.4 THE RIDLEY AND TAN MODELS APPLIED TO THE BUCK
CONVERTER ...........................................................................87

ix
 3.5 A COMPARISON OF THE TWO MODELS AND THE
SIMULATION RESULTS ......................................................... 109
3.6 THE RIDLEY MODEL APPLIED TO THE BOOST CONVERTER 115
3.7 THE RIDLEY MODEL APPLIED TO THE BUCK-BOOST
CONVERTER ......................................................................... 125
3.8 SUMMARY AND CONCLUDING REMARKS ............................ 136
CHAPTER 4 A NOVEL MODEL ................................................. 139
4.1 CHAPTER SURVEY ............................................................... 139
4.2 A NOVEL MODEL FOR THE AUDIO SUSCEPTIBILITY............ 139
4.3 AUDIO SUSCEPTIBILITY OF THE BUCK CONVERTER............ 154
4.4 AUDIO SUSCEPTIBILITY OF THE BOOST CONVERTER .......... 161
4.5 AUDIO SUSCEPTIBILITY OF THE BUCK-BOOST CONVERTER 175
4.6 SUMMARY AND CONCLUDING REMARKS ............................ 183
CHAPTER 5 IMPROVED MODELS........................................... 185
5.1 CHAPTER SURVEY ............................................................... 185
5.2 IMPROVED EXPRESSIONS FOR THE BUCK CONVERTER ....... 185
5.3 IMPROVED EXPRESSION FOR THE BOOST CONVERTER ....... 188
5.4 IMPROVED EXPRESSION FOR THE BUCK-BOOST
CONVERTER ......................................................................... 196
5.5 SUMMARY AND CONCLUDING REMARKS ............................ 205
CHAPTER 6 APPROXIMATIONS OF OBTAINED
EXPRESSIONS............................................................................... 207
6.1 CHAPTER SURVEY ............................................................... 207
6.2 APPROXIMATE MODEL FOR THE BUCK CONVERTER ........... 207
6.3 APPROXIMATE MODEL FOR THE BOOST CONVERTER ......... 229
6.4 APPROXIMATE MODEL FOR THE BUCK-BOOST
CONVERTER ......................................................................... 241
6.5 SUMMARY AND CONCLUDING REMARKS ............................ 258
CHAPTER 7 USING LOAD CURRENT FOR CONTROL....... 259
7.1 CHAPTER SURVEY ............................................................... 259
7.2 A REVIEW ............................................................................ 259
7.3 PRINCIPAL PROPERTIES ....................................................... 262
7.4 PROPERTIES OF THE BUCK CONVERTER .............................. 272
7.5 PROPERTIES OF THE BOOST CONVERTER ............................ 298
7.6 PROPERTIES OF THE BUCK-BOOST CONVERTER ................. 325
7.7 SUMMARY AND CONCLUDING REMARKS ............................ 336

x
CHAPTER 8 SUMMARY ..............................................................341
8.1 RESULTS ...............................................................................341
8.2 FUTURE WORK .....................................................................344
CHAPTER 9 ERRATA FOR THREE PAPERS ..........................345
9.1 PAPER 1 ................................................................................346
9.2 PAPER 2 ................................................................................347
9.3 PAPER 3 ................................................................................348
CHAPTER 10 REFERENCES .......................................................351

xi
xii
 Chapter 1 Introduction

This thesis is concerned with the modeling and control of dc-dc

converters with current-mode control. Special focus is on using load current
measurements for control.
In this first chapter, the background of the problem is described, the
motivation for the work is presented and the contributions of the thesis are
outlined.

1.1 Background

DC-DC Converters

The principal schematic of a dc-dc converter is shown in Figure 1.1. It

converts a dc input voltage, v g (t ) , to a dc output voltage, v o (t ) , with a
magnitude other than the input voltage (Erickson and Maksimovic, 2000,
Section 1.1). The converter often includes one (or several) transistor in order
to control the output voltage, using the control signal δ (t ) .
It is desirable that the conversion be made with low losses in the
converter. Therefore, the transistor is not operated in its linear interval.
Instead, it is operated as a switch and the control signal is binary. While the
transistor is on, the voltage across it is low which means that the power loss in
the transistor is low. While the transistor is off, the current through it is low
and the power loss is also low. To obtain low losses, resistors are avoided in
the converters. Capacitors and inductors are used instead since they ideally do
not have any losses.
The electrical components can be combined and connected to each other
in different ways, called topologies, each one having different properties. The
buck, boost, and buck-boost converters are three basic converter topologies.

1
2 Chapter 1. Introduction

iload(t)

Dc-dc
vg(t) power Load vo(t)
converter

δ (t)
Figure 1.1: A dc-dc converter.

The buck converter has an output voltage that is lower than the input voltage.
The boost converter has an output voltage that is higher than the input
voltage (in steady state). The buck-boost converter is able to have an output
voltage magnitude that is higher or lower than the input voltage magnitude.
Figure 1.2 shows the buck converter with two controllers. Here it is
assumed that all components are ideal. The load consists of a resistor with
resistance R . The converter has an output low-pass filter consisting of an
inductor with inductance L and a capacitor with capacitance C . While the
transistor is on, the inductor current, i L (t ) , increases since the input voltage
is higher than the output voltage in the buck converter. As the transistor is
turned off, the diode must start to conduct since the inductor current cannot
stop instantly. The voltage across the diode is zero when it is conducting and
the inductor current will decrease. Figure 1.3 shows the waveforms of the
control signal and the inductor current. The converter is usually designed so
that the magnitude of the ripple in the output voltage becomes small. That is
why the inductor current waveform in Figure 1.3 increases and decreases
almost linearly. The voltage across the diode is equal to the input voltage or
equal to zero. The output filter of the converter filters this voltage waveform
and the magnitude of the ripple in the output voltage depends on the filter
design. If the inductor current becomes zero before the transistor is turned
on, it will remain at zero until the transistor is turned on since the diode only
can conduct in one direction. If the converter is operated so that the inductor
current is zero during some part of the switching period, it is said to be
Chapter 1. Introduction 3

iL(t) L iload(t)

vg(t) C R vo(t)

Driver
Vref Voltage ic(t) Current δ (t)
controller controller

Figure 1.2: The buck converter with a current controller and a voltage
controller.

δ (t)

t
0 Tsd(t) Ts
iL(t)

t
Figure 1.3: The waveforms of the control signal and the inductor current.

operated in discontinuous conduction mode. Otherwise, it is operated in

continuous conduction mode.
The switching period, Ts , of the converter is determined by the control
signal δ (t ) , as shown in Figure 1.3. In this figure, the switching period is
held constant. The average output voltage is controlled by changing the width
of the pulses. In Figure 1.3, the falling edge is controlled i.e. when the
4 Chapter 1. Introduction

transistor should turn off. The duty cycle, d (t ) , is a real value in the interval
0 to 1 and it is equal to the ratio of the width of a pulse to the switching
period. The duty cycle is actually a discrete-time signal.

State-Space Averaging

The converter acts as a time-invariant system while the transistor is on.

While the transistor is off the converter acts as another time-invariant system
and if the inductor current reaches zero, the converter acts as yet another
time-invariant system. If the transistor is controlled as described previously,
the converter can be described as switching between different time-invariant
systems during the switching period. Consequently, the converter can be
modeled as a time-variant system. State-space averaging (Middlebrook and
Cuk, 1976) is one method to approximate this time-variant system with a
linear continuous-time time-invariant system. This method uses the state-
space description of each time-invariant system as a starting point. These
state-space descriptions are then averaged with respect to their duration in the
switching period. The averaged model is nonlinear and time-invariant and has
the duty cycle, d (t ) , as the control signal instead of δ (t ) . This model is
finally linearized at the operating point to obtain a small-signal model. From
the model we will extract three major transfer functions:

• The control-to-output transfer function describes how a change in the

control signal affects the output voltage.
• The output impedance describes how a change in the load current affects
the output voltage.
• The audio susceptibility describes how a change in the input voltage
affects the output voltage.

Current-Mode Control

Figure 1.2 shows the buck converter controlled by two control loops. The
inductor current is fed back to the current controller in the inner loop and
the output voltage is fed back to the voltage controller in the outer loop. This
control method is called current-mode control. (The name current controller
is used instead of current modulator in this thesis, see Section 3.2.) Assume
that the outer loop is not present. The system is then a closed loop system
since the inductor current is fed back. If the outer loop is added, a new closed
loop is obtained. The control signal from the outer loop acts as the reference
Chapter 1. Introduction 5

signal for the current controller. The three transfer functions mentioned
above will in general be different for the new closed loop system.
The current controller controls the inductor current. This can be made in
different ways. One way is to control the peak value of the inductor current
in each switching period. Ridley (1991) and Tan and Middlebrook (1995)
have presented two models for current-mode control. (The voltage controller
is actually excluded.) The main difference between the two models is the
modeling of the current loop gain.
The output voltage is fed back to the voltage controller so that the output
voltage is kept near the voltage reference signal Vref (see Figure 1.2). The
voltage controller controls the reference signal of the current controller, ic (t ) .
An alternative is to let the voltage controller control the duty cycle directly.
This means that the measurement of the inductor current and the current
controller are not needed. This control method is called voltage-mode
control.

1.2 Motivation for the Work

Many aspects must be considered in the case where a converter is to be
designed. One such aspect is keeping the output voltage in the specified
voltage interval. Here are some examples of changes that can decrease the
variation of the output voltage:

• Change the properties of some of the components in the converter, e.g.

increase the capacitance of the capacitor.
• Change the converter topology.
• Change to a more advanced controller.
• Increase the number of signals that are measured and used by the
controller.

Each one of these changes has one or several disadvantages such as:

• Higher cost.
• Increased weight and volume.
• Lower reliability.
• Lower efficiency (see Poon, Tse, and Liu (1999)).

Therefore, the change or changes that are most suitable depend to a large
extent on the converter specification at hand.
6 Chapter 1. Introduction

Converters can be made better in some sense as better components are

developed and more knowledge is available. This motivates research in the
areas of components, converter topologies and controllers for example.
To obtain high performance control of a system, a good model of the
system is needed. A model of a system can be derived by using the laws of
physics and/or by using measurements of the system, i.e. system identification
(Ljung, 1999). When the system is changed during the time it is in use, it is
an advantage to apply system identification that can be used on-line for
updating the model. The adjusted model is then used to adjust the parameters
of the controller, which is the essence of adaptive control (Åström and
Wittenmark, 1995). An adaptive controller can perform better control than a
non-adaptive controller, which must be designed for the worst case.
A difficulty with adaptive control is making the identification such that
the model adjusts sufficiently fast during a system change without making the
identification sensitive to measurement noise. If the adjustment is slow, the
controller must be designed to be cautious and there will be no significant
improvement compared to a non-adaptive controller.
The adjustment can, in general, be made faster if the number of
parameters to be estimated in a system is fewer. One way to achieve this is to
fix the parameters whose values are known with great precision and vary only
slightly. Another way is to measure a larger number of signals in the process
and the reason for this is explained as follows. A way to decrease the number
of parameters to be estimated is to simply identify a part of the system. To
identify this subsystem, its input and output signals must be measured. If a
larger number of signals in the process are measured, it may be possible to
divide the process into different parts. Note that the time for the sampling
and computation are not considered in this discussion.
It is suitable to consider the load to be included in the converter model
since it (usually) affects the dynamics. If a measurement of the load current,
iload (t ) , (see Figure 1.2) is introduced, it is possible to consider the load as
one part to be identified. The output voltage is then regarded as the input
signal and the load current the output signal of this part. If adaptive control is
to be introduced, a suitable first step may be to only identify the load. Often
this is the most variable part of the converter. This first step may be sufficient
to obtain a controller that meets the performance specifications. As a second
step, identification of the rest of the converter may further improve the
control. Then computational time is one price to pay. This second step may
be more expensive than other solutions to improve the performance of the
Chapter 1. Introduction 7

closed loop system. This discussion motivates the research in identification of

the load.
As mentioned above, the output voltage and the load current should be
measured to obtain fast load identification. There are several papers that
suggest that the load current should be measured and utilized for control of
the converter and they show what properties are obtained. Two of these
papers are mentioned here. The output voltage and the inductor current are
assumed to be measured besides the load current in these two papers.
Redl and Sokal (1986) show that the transient in the output voltage due
to a step change in the load can be much reduced. They call the use of the
measured load current feedforward. For a definition of feedforward, see
Åström and Hägglund (1995, Section 7.3). Redl and Sokal also show that the
control-to-output transfer function does not change when this feedforward is
introduced.
The dc gain of the control-to-output transfer function normally depends
on the load. Hiti and Borojevic (1993) use the measured load current to make
the control-to-output transfer function invariant for different loads at dc for
the boost converter. Hiti and Borojevic thus show that the control-to-output
transfer function changes when the use of measured load current is
introduced. The control Hiti and Borojevic use turns out to be exactly the
same as the one Redl and Sokal propose for the boost converter.
To summarize, Redl and Sokal show that the control-to-output transfer
function does not change when the use of measured load current is
introduced while Hiti and Borojevic show that it does change. It thus seems
to be a contradiction. Since the output voltage and the load current are
assumed to be measured in the two papers, the analysis may be connected to
identification of the load in some way. Therefore, it is motivated to
investigate this possible connection and contradiction before the work with
identification of the load starts.

1.3 Main Contributions

Some of the properties that can be obtained using measured load current
for control are analyzed in this thesis. The analysis is only made for the case
where current-mode control is used. An accurate model is used in the case
where the load is a linear resistor.

1. The analysis confirms that low output impedance can be obtained.

8 Chapter 1. Introduction

2. The analysis shows that in the case where the load is a current source, i.e.
the load current is independent of the output voltage, the following
properties are obtained:
• The use of measured load current for control is feedforward.
• The control-to-output transfer function does not change when this
feedforward is introduced.

3. The analysis shows that in the case where the load is a linear resistor, the
following properties are obtained:
• The control-to-output transfer function can change when the
measured load current is introduced for control.
• The converter can become unstable when the measured load current
is introduced for control.
• The control-to-output transfer function can be almost invariant for
different linear resistive loads if the measured load current is used for
control. This is especially the case for the buck converter.
• The use of measured load current for control is not feedforward. It
can instead be seen as gain scheduling, which can be considered a
special case of adaptive control (Åström and Wittenmark, 1995,
Chapter 9).

In the thesis it is also shown that the two published models for current-
mode control, Ridley (1991) and Tan and Middlebrook (1995), give accurate
expressions for the control-to-output transfer function and the output
impedance but not for the audio susceptibility. A novel model for the audio
susceptibility is presented and it is used to improve the Ridley and Tan
models. The novel model is in some cases inaccurate at low frequencies but
the improvements are made in such a way that this shortcoming is not
transferred to the improved models. The improved models are accurate.
Accurate (continuous-time) expressions for the control-to-output transfer
function, the output impedance, and the audio susceptibility are in this thesis
derived for dc-dc converters that meet the following specifications:

• The converter topology is buck, boost or buck-boost.

• The converter is operated in continuous conduction mode.
• Current-mode control with constant switching frequency and peak-
current command is used.
• The load is a linear resistor.
Chapter 1. Introduction 9

1.4 A Guide for the Reader and the Outline of the

Thesis

A Guide for the Reader

If the reader is a designer of dc-dc converters and is about to use

measured load current for the controller, it is recommended that Chapter 7 is
read since it will increase the understanding of what properties can be
expected. However, if the converter topology is buck (-derived), Section 7.5
and Section 7.6 can be omitted and if the converter topology is boost (-
derived), Section 7.6 can be omitted. Section 7.6 should only be read if the
converter topology is buck-boost (-derived).
The accurate transfer functions for current-mode control that are used as
a basis for the analysis in Chapter 7 are derived throughout Chapters 2-6.
This derivation is mainly of academic interest. In most of these chapters, the
buck, boost, and buck-boost converters are treated in sequence. The
derivations for the buck-boost converter are very similar to the derivations for
the boost converter. Therefore, all the sections treating the buck-boost
converter can be omitted without missing any information of principal
interest.
For those readers that are familiar with the models presented by Ridley
(1991) and Tan and Middlebrook (1995) and are interested in the
derivations of the accurate expressions for the audio susceptibility, it is
recommended to read Section 3.5, Chapter 4 (except Section 4.5), and
Section 5.3. Note that the inaccuracy of the Ridley and Tan models in the
case where the audio susceptibility is considered is small in most cases and
therefore of little practical interest. However, the accurate expressions for the
audio susceptibility can be of academic interest since it is easier to draw
conclusions from an analysis if it is known that the error in the model that is
used as a starting point is small.

Outline of the Thesis

In Chapter 2, the operation of the buck, boost, and buck-boost

converters are described. State-space averaging is used to derive models of the
converters. These models are compared with results from simulations of
switched models.
10 Chapter 1. Introduction

Current-mode control is explained in Chapter 3. The Ridley and Tan

models are used to obtain models of the buck converter with current-mode
control. These models are compared with results from simulations of a buck
converter. The results of the comparison are explained. The Ridley model is
also used to obtain models of the boost and buck-boost converters with
current-mode control. These models are also compared with simulation
results.
Chapter 4 presents the novel model for the audio susceptibility. The
model is applied to the three converter topologies and the obtained
expressions are compared with the corresponding ones in Chapter 3.
In Chapter 5, the Ridley and Tan models are improved by using the
results in Chapter 4.
Chapter 6 shows some approximations of the models for current-mode
control presented in the previous chapters.
Chapter 7 analyzes some properties that can be obtained when using load
current measurements to control the converter. The results of this analysis are
compared with simulation results.
A summary is presented in Chapter 8.

1.5 Publications
The author has published the following conference papers:

1. Johansson, B. and Lenells, M. (2000), Possibilities of obtaining

small-signal models of DC-to-DC power converters by means of
system identification, IEEE International Telecommunications Energy
Conference, pp. 65-75, Phoenix, Arizona, USA, 2000.
2. Johansson, B. (2002a), Analysis of DC-DC converters with current-
mode control and resistive load when using load current
measurements for control, IEEE Power Electronics Specialists
Conference, vol. 1, pp. 165-172, Cairns, Australia, 2002.
3. Johansson, B. (2002b), A comparison and an improvement of two
continuous-time models for current-mode control, IEEE
International Telecommunications Energy Conference, pp. 552-559,
Montreal, Canada, 2002.

Paper 1 is not included in this thesis. Paper 2 contains parts of Chapter 7.

Paper 3 contains parts of Chapters 3-6.
 Chapter 2 State-Space
Averaging

In this chapter we derive small-signal models for three basic converter

topologies by means of state-space averaging. To find out if the frequency
functions predicted by the derived models are accurate, they are compared
with simulation results. Switched (large-signal) simulation models are utilized.
The derived models will be utilized in Chapter 3 where models for converters
with current-mode control are considered.

2.1 Introduction
The converter can be described as switching between different time-
invariant systems during each switching period and is subsequently a time-
variant system. There are several methods that approximate this time-variant
system with a linear continuous-time time-invariant system. State-space
averaging (Middlebrook and Cuk, 1976), circuit averaging (Wester and
Middlebrook (1973) and Vorperian (1990)), and the current-injected
approach (Clique and Fossard, 1977) are some of them. State-space averaging
is used in this chapter to derive models for the buck, boost, and buck-boost
converters.
The operation of the buck converter is explained in Section 2.2. State-
space averaging is used in Section 2.3 to derive a model of the buck converter.
The control-to-output transfer function, the output impedance, and the
audio susceptibility are extracted from this model. The method of state-space
averaging is included in Section 2.3 and it is presented in a little different way
compared to the traditional one. In Section 2.4, a switched simulation model
of the buck converter is presented. It is shown how the frequency functions of
the converter are obtained from this simulation model. The frequency

11
12 Chapter 2. State-Space Averaging

functions are presented and compared with the three transfer functions
derived in Section 2.3.
The operation of the boost converter is explained in Section 2.5. State-
space averaging is applied to the boost converter in Section 2.6 and the result
is compared with simulation results in Section 2.7. The corresponding work
is made for the buck-boost converter in Section 2.8, 2.9, and 2.10. A
summary and concluding remarks are presented in Section 2.11.

2.2 Operation of the Buck Converter

The circuit and operation of the buck converter are presented in this
section. Numerous notations are introduced and some design considerations
are presented.
The components of a converter are not ideal and some of these non-
idealities can be considered in a model. Only one non-ideality is considered in
the model that will be used. The capacitor is modeled as an ideal capacitor in
series with an ideal resistor with resistance Rc . The resistance Rc is called the
Equivalent Series Resistance (ESR) of the capacitor. The ESR is used to
represent all the power losses in the capacitor. The dependency of power loss
on frequency is not addressed here. Figure 2.1 shows the circuit that will be
used for the buck converter.
We assume that steady state is reached. The control signal, δ (t ) , then
consists of pulses with constant width. The waveforms of the signals in the
circuit are as shown in Figure 2.2. In Section 2.4, a simulation model will be
presented and this model is used to obtain the presented waveforms. The time
intervals where the control signal δ (t ) is high are called t on and the once
where δ (t ) is low are called t off . The switching period, Ts , is the time
between two successive positive flanks of δ (t ) and hence equal to the sum of
t on and t off . The ratio of t on to Ts is called the duty cycle or the duty ratio
and it is denoted by d (t ) . The duty cycle is constant in steady state. During
t on the transistor operates in the on state and during t off the transistor
operates in the off state. The voltage across the diode, v diode (t ) , is equal to
the input voltage, v g (t ) , during t on . The input voltage is held constant in
the simulation. During t off the diode voltage is equal to zero since we assume
that the converter is operated in continuous conduction mode (see Section
1.3). The diode voltage is filtered by the output low-pass LC-filter. The
corner frequency of this filter is chosen to be much lower than the switching
Chapter 2. State-Space Averaging 13

itrans(t) iL(t) L iload(t)

vL(t)
vESR(t) Rc
vg(t) Driver vdiode(t) R vo(t)
v(t) C
idiode(t) icap(t)

δ (t)

Figure 2.1: The circuit of the buck converter.

frequency to obtain small magnitude of the ripple in the output voltage,

v o (t ) . Consequently, this voltage is approximately equal to the mean value of
the diode voltage. The mean value of v diode (t ) is lower than v g (t ) .
Therefore, v o (t ) is lower than v g (t ) in steady state.
The voltage across the inductor, v L (t ) , is equal to the difference between
v diode (t ) and v o (t ) . The inductor current, i L (t ) , is proportional to the
integral of v L (t ) . Therefore, i L (t ) increases during t on and decreases during
t off . During each time interval, the slope of i L (t ) is almost constant since
v L (t ) is almost constant.
The inductor current is equal to the sum of the transistor current,
itrans (t ) , and the diode current, idiode (t ) . The transistor current is equal to
i L (t ) during t on since idiode (t ) is zero. The diode current is equal to i L (t )
during t off since itrans (t ) is zero. The load current, iload (t ) , is almost
constant since v o (t ) is almost constant. The capacitor current, icap (t ) , is
equal to the difference between i L (t ) and iload (t ) . The mean value of
icap (t ) is zero in steady state. Consequently, iload (t ) and i L (t ) have the
same mean value.
The voltage across the (ideal) capacitor, v(t ) , is proportional to the
integral of icap (t ) . The voltage across the capacitor’s ESR, v ESR (t ) , is
proportional to icap (t ) . The output voltage, v o (t ) , is equal to the sum of
v(t ) and v ESR (t ) .
Table 2.1 shows the parameter values used in the simulation. These are
also used by Ridley (1991). The switching frequency, f s , is equal to 50 kHz
(the inverse of Ts ). If the ESR in the capacitor is negligible, the corner
frequency of the LC filter is
14 Chapter 2. State-Space Averaging

2
δ (t ) 1
0
20
vdiode (t ) 10
0
10
v L (t ) 0
-10
6
i L (t ) 5
4
10
itrans (t ) 5
0
10
idiode (t ) 5
0
5.05
iload (t ) 5
4.95
1
icap (t ) 0
-1
5.01
v(t ) 5.005
5
0.02
v ESR (t ) 0
-0.02
5.05
vo (t ) 5
4.95
0 0.5 1 1.5 2 2.5 3 3.5 4
t on t off t on t off -5
x 10

t (s)

Figure 2.2: The waveforms of the signals in steady state for a buck converter.
The unit of the voltages is Volt and the unit of the currents is
Ampere.
Chapter 2. State-Space Averaging 15

1
f0 = = 1.3 kHz. (2.1)
2π LC

f 0 is thus much lower than f s , which means that the magnitude of the
ripple in v o (t ) is small. The magnitude is decreased if L , C , or f s is
increased. However, there are disadvantages by doing so, for example:

• If L is increased, it will take longer time for the inductor current to reach
a new average level. This is needed when a step change occurs in the load
current.
• If L or C is increased, the volume, weight, and cost of the converter are
increased.
• If f s is increased, the switching losses in the transistor are increased.

The ESR of the capacitor also contributes to the ripple in v o (t ) since this
voltage is equal to the sum of v(t ) and v ESR (t ) . Furthermore, it causes a step
change in v o (t ) when a step change occurs in the load current. This is one of
the reasons for the use of capacitors with low ESR in converters.

Table 2.1: The parameter values used in the simulation of the buck converter.

Parameter Value
L 37.5 µH
C 400 µF
Rc 14 mΩ
R 1Ω
v g (t ) 11 V
d (t ) 0.455
Ts 20 µs

2.3 Model of the Buck Converter

In this section, a linear time-invariant model of the buck converter is
derived by means of state-space averaging. The converter can be described as
switching between different time-invariant systems and the state-space
16 Chapter 2. State-Space Averaging

description of each one of these systems is first derived. These state-space

descriptions are used as a starting point in the method of state-space
averaging. This method is presented and then applied to the buck converter.
The result is a linear time-invariant model in state-space description. Finally,
several transfer functions are extracted from this model.

State-Space Description for Each Time Interval

Since it is assumed that the converter is operated in continuous

conduction mode, two different systems must be considered. The state-space
description of each one of these two systems is derived in this subsection.
While the transistor is on, the voltage across the diode is equal to the
input voltage. The circuit in Figure 2.3 can therefore be used as a model of
the buck converter during t on . In the figure, a current source is added. It
injects the current iinj (t ) into the output stage of the converter. This current
is an input signal and is needed to determine the output impedance.
From Figure 2.3, the following equations are obtained:

di L (t ) 1
dt
(
= v g (t ) − v o (t ) ,
L
) (2.2)

dv(t ) 1  v (t ) 
=  i L (t ) − o − iinj (t )  , (2.3)
dt C R 

 v (t ) 
v o (t ) = v(t ) + Rc  i L (t ) − o − iinj (t )  . (2.4)
 R 

(2.4) is rearranged to:

v o (t ) +
Rc
R
(
v o (t ) = v(t ) + Rc i L (t ) − iinj (t ) , ) (2.5)

(
v(t ) + Rc i L (t ) − iinj (t ) )
v o (t ) = , (2.6)
1 + Rc R
Chapter 2. State-Space Averaging 17

iL(t) L iinj(t)

Rc
vg(t) R vo(t)
v(t) C

Figure 2.3: The circuit of the buck converter during t on .

RRc R RRc
v o (t ) = i L (t ) + v(t ) − iinj (t ) . (2.7)
R + Rc R + Rc R + Rc

(2.7) is used to substitute v o (t ) in (2.2) and (2.3):

di L (t ) RRc R
=− i L (t ) − v(t ) +
dt (R + Rc )L (R + Rc )L
(2.8)
1 RRc
v g (t ) + i (t ) ,
L (R + Rc )L inj
dv(t ) 1 Rc 1
= i L (t ) − i L (t ) − v(t ) +
dt C (R + Rc )C (R + Rc )C
(2.9)
Rc 1
iinj (t ) − iinj (t ) .
(R + Rc )C C

(2.9) is simplified:

dv(t ) R 1 R
= i L (t ) − v(t ) − i (t ) . (2.10)
dt (R + Rc )C (R + Rc )C (R + Rc )C inj
The circuit in Figure 2.3 is a second order system. Let i L (t ) and v(t ) be
chosen as the state variables. Regard v g (t ) and iinj (t ) as the input signals
and v o (t ) as the output signal. By using (2.8), (2.10), and (2.7), the
following state-space system is obtained:
18 Chapter 2. State-Space Averaging

 dx(t )
 = A 1 x(t ) + B 1u(t )
 dt (2.11)
 y (t ) = C 1 x(t ) + E 1u (t )

where

i (t )
x(t ) =  L  , (2.12)
 v(t ) 

 v g (t ) 
u(t ) =  , (2.13)
iinj (t )

y (t ) = vo (t ) , (2.14)

 RRc R 
− (R + R )L −
(R + Rc )L  ,
A1 =  c
(2.15)
 R
−
1 
 (R + Rc )C (R + Rc )C 

1 RRc 
L (R + Rc )L  ,
B1 =  (2.16)
0 −
R 
 (R + Rc )C 

 RRc R 
C1 =  , (2.17)
 R + Rc R + Rc 

 RRc 
E 1 = 0 − . (2.18)
 R + Rc 

While the transistor is off, the voltage across the diode is equal to zero.
The circuit in Figure 2.4 can therefore be used as a model of the buck
converter during t off . The circuit in Figure 2.3 and Figure 2.4 are the same if
Chapter 2. State-Space Averaging 19

iL(t) L iinj(t)

Rc
R vo(t)
v(t) C

Figure 2.4: The circuit of the buck converter during t off .

v g (t ) is zero. Therefore, a state-space model for the circuit in Figure 2.4 can
be obtained by setting all the coefficients for v g (t ) to zero in (2.11):

 dx(t )
 = A 2 x(t ) + B 2 u(t )
 dt , (2.19)
 y (t ) = C 2 x(t ) + E 2 u(t )

where

A 2 = A1 , (2.20)

 RRc 
0 (R + R )L 
B2 =  c , (2.21)
0 − R 
 (R + Rc )C 

C 2 = C1 , (2.22)

E 2 = E1 . (2.23)

The Method of State-Space Averaging

The converter behaves like switching between the two different linear
time-invariant systems (2.11) and (2.19) during the switching period, so it
looks like a time-variant system. State-space averaging will be used in the next
20 Chapter 2. State-Space Averaging

subsection to approximate this time-variant system with a linear continuous-

time time-invariant system. In this subsection, the method of state-space
averaging is presented. The first step is calculating a nonlinear time-invariant
system by means of averaging and the second step is linearizing this nonlinear
system. The presentation is a little different compared to the traditional one.
The two linear systems are first averaged with respect to their duration in
the switching period:

 dx(t )
 = (d (t ) A 1 + (1 − d (t ) )A 2 )x(t ) + (d (t )B 1 + (1 − d (t ) )B 2 )u(t )
 dt (2.24)
 y (t ) = (d (t )C1 + (1 − d (t ) )C 2 )x(t ) + (d (t )E 1 + (1 − d (t ) )E 2 )u(t )

(2.24) is an approximation of the time-variant system and new variable

names should formally have been used. To limit the number of variable
names, this is not made. The duty cycle, d (t ) , is an additional input signal in
(2.24). A new input vector is therefore defined:

u(t ) 
u' (t ) =  . (2.25)
 d (t ) 

This is not made in traditional presentations of state-space averaging, where

the control signal d (t ) is kept separate from the disturbance signals v g (t )
and iinj (t ) . However, in system theory, all control signals and disturbance
signals are put in an input vector.
Since the duty cycle can be considered to be a discrete-time signal with
sampling interval Ts , one cannot expect the system in (2.24) to be valid for
frequencies higher than half the switching frequency.
The system in (2.24) is a nonlinear time-invariant system. It is nonlinear
since there are products of two input signals and it is time-invariant since all
the coefficients are independent of time.
A nonlinear time-invariant system with state-vector x(t ) , input vector
u' (t ) , and output vector y (t ) , are written as

 dx(t )
 = f (x(t ), u' (t ) )
 dt , (2.26)
 y (t ) = g (x(t ), u' (t ) )
Chapter 2. State-Space Averaging 21

A straight-forward linearization is applied, where we define the deviations

from an operating point as follows:

x(t ) = X + xˆ (t ) , (2.27)

u' (t ) = U'+uˆ ' (t ) , (2.28)

y (t ) = Y + yˆ (t ) . (2.29)

Capital letters denote the operating-point (dc, steady-state) values and the
hat-symbol (^) denotes perturbation (ac) signals. Assume that the operating
point is an equilibrium point, i.e.

f (x(t ), u' (t ) ) = 0.
x (t ) = X (2.30)
u'(t ) = U'

The operating point output values are

Y = g (x(t ), u' (t ) ) .
x (t ) = X (2.31)
u'(t ) = U'

The following linearized (ac, small-signal) system can now be obtained

from (2.26) (Goodwin, Graebe and Salgado, 2001, Section 3.10):

 dxˆ (t )
 = A' xˆ (t ) + B' uˆ ' (t )
 dt , (2.32)
 yˆ (t ) = C' xˆ (t ) + E' uˆ ' (t )

where

 ∂f 
A' =   ,
 ∂x  x (t ) = X
(2.33)
u'(t ) = U'
22 Chapter 2. State-Space Averaging

 ∂f 
B' =   ,
 ∂u'  x (t ) = X
(2.34)
u'(t ) = U'

 ∂g 
C' =   ,
 ∂x  x (t ) = X
(2.35)
u'(t ) = U'

 ∂g 
E' =   .
 ∂u'  x (t ) = X
(2.36)
u'(t ) = U'

(2.32) is an approximation of the nonlinear system and new variable names

should formally have been used. To limit the number of variable names, this
is not made.
(2.24) is a special case of (2.26). The equations (2.28) and (2.30)-(2.36)
will now be rewritten for this special case. The following equation is obtained
if (2.28) is applied to (2.25):

 u(t )   U  uˆ (t ) 
u' (t ) =  = +ˆ , (2.37)
 d (t )  D  d (t )

The following variables are now defined:

d ' (t ) = 1 − d (t ) , (2.38)

D' = 1 − D . (2.39)

The variable d ' (t ) is equal to the fraction of the time the transistor is off. D '
is the operating-point value of d ' (t ) . (2.30) and (2.31) are rewritten by using
(2.24):

 0 = AX + BU
 , (2.40)
Y = CX + EU

where
Chapter 2. State-Space Averaging 23

A = DA 1 + D' A 2 , (2.41)

B = DB 1 + D ' B 2 , (2.42)

C = DC1 + D ' C 2 , (2.43)

E = DE 1 + D ' E 2 . (2.44)

(2.40) is rewritten:

 X = − A −1BU

(
 Y = − CA −1B + E U
.
) (2.45)

(2.33)-(2.36) are reformulated by using (2.24):

 ∂f 
A' =   = (d (t ) A 1 + (1 − d (t ) )A 2 ) =A,
 ∂x  x (t ) = X x (t ) = X
(2.46)
u'(t ) = U' u'(t ) = U'

 ∂f   ∂f ∂f 
B' =   = =
 ∂u'  x (t ) = X  ∂u ∂d  x (t ) = X
u'(t ) = U' u'(t ) = U'

[d (t )B 1 + (1 − d (t ))B 2 (A 1 − A 2 )x(t ) + (B 1 − B 2 )u(t )] x (t ) = X

= (2.47)
u'(t ) = U'

[B (A 1 − A 2 )X + (B 1 − B 2 )U] ,

 ∂g 
C' =   = (d (t )C 1 + (1 − d (t ) )C 2 ) =C,
 ∂x  x (t ) = X x (t ) = X
(2.48)
u'(t ) = U' u'(t ) = U'
24 Chapter 2. State-Space Averaging

 ∂g   ∂g ∂g 
E' =   = =
 ∂u'  x (t ) = X  ∂u ∂d  x (t ) = X
u'(t ) = U' u'(t ) = U'

[d (t )E 1 + (1 − d (t ))E 2 (C1 − C 2 )x(t ) + (E1 − E 2 )u(t )] x (t ) = X

= (2.49)
u'(t ) = U'

[E (C1 − C 2 )X + (E1 − E 2 )U] ,

(2.32) can now be rewritten:

 dxˆ (t )
 = Axˆ (t ) + B' uˆ ' (t )
 dt . (2.50)
 yˆ (t ) = Cxˆ (t ) + E' uˆ ' (t )

B' and E' are defined as

B' = [B B d ] , (2.51)

E' = [E E d ] , (2.52)

where

B d = (A 1 − A 2 )X + (B 1 − B 2 )U , (2.53)

E d = (C1 − C 2 )X + (E 1 − E 2 )U . (2.54)

The results of state-space averaging method are the dc model (2.45) (or
(2.40)) and the ac model (2.50).

Applying State-Space Averaging

The method of state-space averaging is applied to the buck converter in

this subsection. The approximation made in the linearization is also
considered.
The following equations are obtained if (2.27)-(2.29) are applied to
(2.12)-(2.14):
Chapter 2. State-Space Averaging 25

i (t )  I  iˆ (t )
x(t ) =  L  =  L  +  L  , (2.55)
 v(t )   V   vˆ(t ) 

 v g (t )  V g   vˆ g (t ) 
 u(t )   U  uˆ (t )      ˆ 
u' (t ) =   =  D  + dˆ (t ) = iinj (t ) =  0  + iinj (t ) , (2.56)
 d (t )       d (t )   D   ˆ 
     d (t ) 

y (t ) = vo (t ) = Vo + vˆo (t ) . (2.57)

Note that the dc value of iinj (t ) is set to zero in (2.56) so that only the load
resistor determines the dc load current.
(2.41)-(2.44) can easily be expanded since D + D ' = 1 :

 RRc R 
− (R + R )L −
(R + Rc )L  ,
A= c
(2.58)
 R
−
1 
 (R + Rc )C (R + Rc )C 

 1 RRc 
D L (R + Rc )L  ,
B= (2.59)
 0 −
R 
 (R + Rc )C 

 RRc R 
C= , (2.60)
 R + Rc R + Rc 

 RRc 
E = 0 − . (2.61)
 R + Rc 

The dc equations will now be derived. The following equations are

obtained if (2.40) is expanded:
26 Chapter 2. State-Space Averaging

RRc R D
0=− IL − V + Vg , (2.62)
(R + Rc )L (R + Rc )L L

R 1
0= IL − V, (2.63)
(R + Rc )C (R + Rc )C
RRc R
Vo = IL + V. (2.64)
R + Rc R + Rc

(2.63) is simplified to:

V = RI L . (2.65)

(2.65) is inserted into (2.64):

Rc R
Vo = V+ V =V . (2.66)
R + Rc R + Rc

(2.65) is inserted in (2.62):

Rc R
0=− V− V + DV g , (2.67)
(R + Rc ) (R + Rc )
V
=D. (2.68)
Vg

The dc current to the capacitor is zero and the dc voltage across the
capacitor’s ESR is zero. This explains the results in (2.65) and (2.66).
Equation (2.68) shows the dc amplification of the buck converter. The
voltage across the diode is equal to V g during the fraction D of the time and
equal to zero otherwise. V (= Vo ) is equal to the mean value of the voltage
across the diode ( DV g ).
(2.53) and (2.54) are expanded and written on an explicit form:
Chapter 2. State-Space Averaging 27

1 L 0 V L 
B d = 0X +   U= g , (2.69)
 0 0  0 

E d = 0X + 0U = 0 . (2.70)

(2.51) and (2.52) are expanded:

D RRc Vg 
 
B' =  L
(R + Rc )L L ,
(2.71)
 R 
0 − 0 
 (R + Rc )C 

 RRc 
E' = 0 − 0 . (2.72)
 R + Rc 

All the coefficient matrices in the ac model (2.50) are now available.
The approximation made in the linearization is now considered. The
following equation is obtained from the nonlinear system (2.24):

di L (t ) RRc R
=− i L (t ) − v(t ) +
dt (R + Rc )L (R + Rc )L
(2.73)
1 RRc
d (t )v g (t ) + i (t ) .
L (R + Rc )L inj
(2.73) is rewritten by using (2.55) and (2.56):

( )
d I L + iˆL (t )
=−
RRc
(
I + iˆ (t ) − ) R
(V + vˆ(t ) ) +
dt (R + Rc )L L L (R + Rc )L
1
L
( )( )
D + dˆ (t ) V g + vˆ g (t ) +
RRc
iˆ (t ) .
(R + Rc )L inj
(2.74)

(2.74) is rewritten by using (2.62):

28 Chapter 2. State-Space Averaging

d iˆL (t ) RRc R
=− iˆL (t ) − vˆ(t ) +
dt (R + Rc )L (R + Rc )L
(2.75)
1 1 1 RRc
Dvˆ g (t ) + V g dˆ (t ) + dˆ (t )vˆ g (t ) + iˆ (t ) .
L L L (R + Rc )L inj
The following equation is obtained from the linearized system in (2.50):

d iˆL (t ) RRc R
=− iˆL (t ) − vˆ(t ) +
dt (R + Rc )L (R + Rc )L
(2.76)
D RRc Vg
vˆ g (t ) + iˆinj (t ) + dˆ (t ) .
L (R + Rc )L L

The difference between (2.75) and (2.76) is the term dˆ (t )vˆ g (t ) L , which is a
scaled product of two perturbation signals. Thus, this product is neglected in
the linearization.

Extracting the Transfer Functions

The control-to-output transfer function, the output impedance and the

audio susceptibility will now be derived from the linearized system in (2.50).
Assume that the state is zero initially. The Laplace transform of (2.50) is

sxˆ ( s ) = Axˆ ( s) + B' uˆ ' ( s )

 . (2.77)
 yˆ ( s ) = Cxˆ ( s ) + E' uˆ ' ( s)

To be spared from introducing new variable names, the Laplace transform of

a signal is denoted by the same name as the signal, e.g. L{v(t )} = v( s ) , even if
this is not a formally correct notation.
(2.77) is rewritten:

xˆ ( s ) = (sI − A )−1 B' uˆ ' ( s)

 . (2.78)
 yˆ ( s ) = Cxˆ ( s ) + E' uˆ ' ( s )

The first equation in (2.78) is expanded:

Chapter 2. State-Space Averaging 29

−1
 RRc R 
 (R + R )L + s (R + Rc )L 
xˆ ( s ) =  c •
− R 1
+ s
 (R + Rc )C (R + Rc )C 
(2.79)
D RRc Vg 
 
L (R + Rc )L L ˆ
u' ( s ) .
 R 
0 − 0 
 (R + Rc )C 

The matrix inversion in (2.79) is calculated:

1
xˆ ( s ) = •
 RRc  1  R2
 + s  + s +
 (R + R )L   (R + R )C 
 (R + Rc ) LC
2
 c  c

 1 R  D RRc Vg  (2.80)
 (R + R )C + s −
(R + Rc )L   L 
 c (R + Rc )L L  uˆ ' ( s ) .
 + s 0 
R RRc R
− 0 
 (R + Rc )C (R + Rc )L   (R + Rc )C 

(2.80) is simplified:
30 Chapter 2. State-Space Averaging

1
xˆ ( s ) = •
RRc + R 2
L + RRc C
+s + s2
(R + Rc ) LC (R + Rc )LC
2

 D D RRc RRc R2
 + s + s +
 (R + Rc )LC L (R + Rc )2 LC (R + Rc )L (R + Rc )2 LC
 RD R 2 Rc R 2 Rc R (2.81)
 − −s
 (R + Rc )LC (R + Rc ) LC (R + Rc ) LC (R + Rc )C
2 2

Vg Vg 
+s 
(R + Rc )LC L 
uˆ ' ( s ) ,
RV g 
(R + Rc )LC 

1
xˆ ( s ) = •
R + s (L + RRc C ) + s 2 (R + Rc )LC
(2.82)
 D(1 + s (R + Rc )C ) R(1 + sRc C ) V g (1 + s (R + Rc )C )
 RD − sRL RV uˆ ' ( s) .
 g 

Six transfer functions are obtained from (2.82):

iˆL ( s ) V g (1 + s (R + Rc )C )
= , (2.83)
dˆ ( s ) R + s(L + RRc C ) + s 2 (R + Rc )LC

vˆ( s ) RV g
= , (2.84)
dˆ ( s ) R + s(L + RRc C ) + s 2 (R + Rc )LC

iˆL ( s ) R (1 + sRc C )
= , (2.85)
iˆinj ( s) R + s (L + RRc C ) + s 2 (R + Rc )LC

vˆ( s) − sRL
= , (2.86)
ˆiinj ( s) R + s (L + RR C ) + s 2 (R + R )LC
c c
Chapter 2. State-Space Averaging 31

iˆL ( s ) D(1 + s (R + Rc )C )
= , (2.87)
vˆ g ( s ) R + s (L + RRc C ) + s 2 (R + Rc )LC

vˆ( s ) RD
= . (2.88)
vˆ g ( s ) R + s (L + RRc C ) + s 2 (R + Rc )LC

The second equation in (2.78) is now expanded:

 RRc R   RRc 
yˆ ( s ) =   xˆ ( s ) + 0 − 0 uˆ ' ( s ) . (2.89)
 R + Rc R + Rc   R + Rc 

The control-to-output transfer function is obtained by combining (2.89),

(2.83), and (2.84):

vˆo ( s ) RRc iˆL ( s ) R vˆ( s)

= + =
dˆ ( s ) R + Rc dˆ ( s) R + Rc dˆ ( s )
RRcV g (1 + s(R + Rc )C ) + R 2V g
=
(R + Rc )(R + s(L + RRc C ) + s 2 (R + Rc )LC )
(2.90)
RV g (R + Rc ) + sRRcV g (R + Rc )C
=
(R + Rc )(R + s(L + RRc C ) + s 2 (R + Rc )LC )
RV g (1 + sRc C )
.
R + s (L + RRc C ) + s 2 (R + Rc )LC

The output impedance is obtained by combining (2.89), (2.85), and (2.86):

32 Chapter 2. State-Space Averaging

vˆo ( s )
Z out ( s ) = − =
iˆinj ( s )
RRc iˆL ( s ) R vˆ( s ) RRc
− − + =
R + Rc ˆiinj ( s ) R + Rc iˆinj ( s ) R + Rc
− RRc R (1 + sRc C ) + RsRL
+
(R + Rc )(R + s(L + RRc C ) + s 2 (R + Rc )LC )
RRc (R + s(L + RRc C ) + s 2 (R + Rc )LC )
= (2.91)
(R + Rc )(R + s(L + RRc C ) + s 2 (R + Rc )LC )
RsRL + RRc (sL + s 2 (R + Rc )LC )
=
(R + Rc )(R + s(L + RRc C ) + s 2 (R + Rc )LC )
sRL(R + Rc (1 + s(R + Rc )C ))
=
(R + Rc )(R + s(L + RRc C ) + s 2 (R + Rc )LC )
sRL(1 + sRc C )
.
R + s (L + RRc C ) + s 2 (R + Rc )LC

The output impedance is the impedance of the converter with respect to the
output terminals. The load resistance, R , is here defined to be included in
the output impedance. According to Erickson and Maksimovic (2000,
preamble of Chapter 8), the load resistance can either be included or
excluded. The minus sign in the definition of Z out (s ) in (2.91) is due to the
definitions of vo (t ) and iinj (t ) in Figure 2.3. The audio susceptibility is
derived by combining (2.89), (2.87), and (2.88):
Chapter 2. State-Space Averaging 33

vˆo ( s) RRc iˆL ( s ) R vˆ( s )

= + =
vˆ g ( s ) R + Rc vˆ g ( s ) R + Rc vˆ g ( s )
RRc D (1 + s(R + Rc )C ) + R 2 D
=
(R + Rc )(R + s(L + RRc C ) + s 2 (R + Rc )LC ) (2.92)
RD(R + Rc ) + sRRc D(R + Rc )C
=
(R + Rc )(R + s(L + RRc C ) + s 2 (R + Rc )LC )
RD(1 + sRc C )
.
R + s (L + RRc C ) + s 2 (R + Rc )LC

The audio susceptibility is also called the line-to-output transfer function and
the input-to-output transfer function.
The zero (1 + sRc C ) is added to the transfer function in the case where
vˆo ( s ) is the output compared to the case where vˆ( s ) is the output. This is
apparent by comparison of (2.92), (2.88), (2.91), (2.86), (2.90), and (2.84).
A short review of this section is now presented. First, a state-space
description of the buck converter for the case where the transistor operates in
the on-state was derived. After that, the same was made for the case where the
transistor operates in the off-state. These two state-space descriptions were
used to obtain a linear time-invariant model by means of state-space
averaging. Finally, several transfer functions were extracted from this model.

2.4 Simulation of a Buck Converter

In this section, a switched (large-signal) simulation model is presented.
The frequency functions predicted by the transfer functions derived in
Section 2.3 are compared with simulation results.

Simulation Model

In this subsection, it is shown how the simulation model is built to obtain

different frequency functions.
We simulate a buck converter by using the software
MATLAB/SIMULINK including Power System Blockset. Table 2.2 shows
the versions of the software. Figure 2.5 shows the complete simulation model.
The simulation model of the buck converter is put into a subsystem. The
34 Chapter 2. State-Space Averaging

subsystem has three input signals and three output signals and they are
described in Table 2.3.
The subsystem is shown in Figure 2.6. The input and output signals are
used as interface signals to the electrical part of the simulation model. The
input signals vg and iinj, control a voltage and a current source, respectively.
However, iinj is first multiplied by -1 to obtain a direction of the injected
current that agrees with the one defined in Figure 2.3. The input signal delta,
controls the transistor. A controllable switch emulates the diode. The inverse
of delta is used to control this switch since the diode should conduct when the
transistor in not conducting. To be able to start the simulation, a dummy
resistor is included in the model. The resistance of this resistor is set to 1 MΩ
and its effect on the simulation result is negligible. The output signals vo,
iload, and iL are measurements of the output voltage, load current, and
inductor current, respectively.
The input and output signals of the converter are connected as shown in
Figure 2.5, to obtain the frequency functions of the converter. The input

Table 2.2: The versions of the software.

Software Version
MATLAB 5.3
SIMULINK 3.0
Power System Blockset 2.0

Table 2.3: The input and output signals of the subsystem.

Name Type Description

vg Input Input voltage, v g (t )
iinj Input Injected current, iinj (t )
delta Input Control signal, δ (t )
vo Output Output voltage, v o (t )
iload Output Load current, iload (t )
iL Output Inductor current, i L (t )
Chapter 2. State-Space Averaging 35

magnitude
signal
Vg angle
vg

vghat
vg vo
vo
Iinj iinj iload
iinj iload
delta iL
iL
Buck
iinjhat S Q
delta converter

D R !Q
d

dhat sawtooth

0.5

Figure 2.5: The complete simulation model.

3 2 1
iL iload vo
i + -i
1 2 + - +
v
g m L -
Transistor signal
+ 2 m Rc
signal Diode
- Rdummy R
emulator
1 g
C

NOT -1
1 3 2
vg delta iinj

Figure 2.6: The buck converter subsystem.

voltage, vg, is the sum of its dc value, Vg, and its ac value vghat. The injected
current, iinj, and the duty cycle, d, are implemented in a corresponding way.
The dc value of iinj, i.e. Iinj, is equal to zero in all the simulations. Only one
signal generator at the time is activated. Specifying the amplitude to be equal
to zero inactivates a signal generator.
The pulse width modulation (PWM) makes use of a saw-tooth signal.
The signal sawtooth is increasing linearly from 0 to 1 as shown in Figure 2.7.
When the signal reaches 1, it is instantly set to 0 again. The period of the
signal is equal to Ts , i.e. the switching period. When the signal sawtooth
36 Chapter 2. State-Space Averaging

0.8
sawtooth, d

0.6

0.4

0.2

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
t (s) -4
x 10

1
0.8
delta

0.6
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
t (s) -4
x 10
Figure 2.7: The waveforms of the signals in a pulse-width modulator.

becomes greater than the duty cycle, d, the SR-latch is reset. A relay block is
used to generate the reset signal. The output signal of the relay block is equal
to 1 if the input signal is positive and equal to 0 if the input signal is negative.
The hysteretic property in the relay block is thus not utilized. A pulse
generator sets the SR-latch. The period of the set signal is equal to Ts and the
duration of each pulse is 1 percent of Ts . The set signal is synchronized with
the sawtooth signal so that the SR-latch is set each time the sawtooth signal
goes from 1 to 0. The output signal of the SR-latch, delta, is a pulse train and
the width of each pulse is determined by the duty cycle signal d. Note that it
is possible to obtain PWM without using a SR-latch by using the inverse of
the reset signal instead.
Consider the output signal vo of the converter which represents the
output voltage. This signal has a Fourier component with a frequency equal
to the frequency of the signal from the active signal generator. There exist
other Fourier components in the output voltage (Erickson and Maksimovic,
2000, Section 7.1). The linearized model derived in Section 2.3 will be
compared with the simulation results in the next subsection. The output of a
linear system only consists of one Fourier component if the input is a
Chapter 2. State-Space Averaging 37

sinusoidal. The frequency of this component is the same as the frequency of

the input sinusoidal. To be able to compare the simulation results with the
linearized model, only the Fourier component in the output voltage with a
frequency equal to the frequency of the signal from the active signal generator
is considered. A network analyzer also just considers this Fourier component
(Erickson and Maksimovic, 2000, Section 8.5).
A Fourier component of a continuous-time periodic signal with period T
is calculated by solving an integral where the length of the integration interval
is equal to T (or a multiple of T ). Therefore, the period of the output
voltage is now considered and let it be denoted by T . If none of the signal
generator is active, then T is equal to the switching period Ts . If one signal
generator is active and the period of its signal is equal to Ts , then T is still
equal to Ts . If the period of the signal from the signal generator is 2Ts , then
T = 2Ts , and so forth. However, if the period of the signal from the signal
generator is 1.5Ts , then T is equal to 3Ts , not 1.5Ts . The relative difference
between the periods of the two signals can be large if the period of the signal
from the signal generator is chosen “badly”. In a simulation, the Fourier
integral is solved numerically and the length of the integration interval is
equal to T to obtain fast simulation. Furthermore, the period of the signal
from the active signal generator should be equal to a multiple of the switching
period Ts , to avoid unnecessary large T . We follow these recommendations
in all the simulations where the frequency functions of the converter is
searched.
The switching frequency, f s , is equal to 50 kHz (the inverse of Ts ). We
evaluate the frequency functions at the frequencies 16666.6667 Hz ( ≈ f s 3 ),
10000 Hz ( = f s 5 ), 5000 Hz, 2500 Hz, 1000 Hz, 500 Hz, 250 Hz, 100
Hz, and 50 Hz. To evaluate a frequency function at a specific frequency, the
active signal generator is set to generate a sinusoidal with this frequency. The
specific frequency is also set in the Fourier analysis block. This block has the
output voltage vo as the input signal. It analyzes the Fourier component with
the specific frequency by using a part of the signal vo. The part has an interval
length equal to the inverse of the specific frequency. The interval length is
equal to the period of the signal vo for each evaluated frequency and the
Fourier component is therefore calculated correctly according to the previous
discussion.
The result of the Fourier analysis is the magnitude and the phase of the
component. The analysis is repeated during the simulation and the results are
viewed by using the oscilloscope block. At the start of each simulation, the
result of the Fourier analysis changes considerably since the inductor current
38 Chapter 2. State-Space Averaging

and the capacitor voltage are far from the final dc values. The simulation is
stopped when the changes in the result of the Fourier analysis is negligible.
The result of a frequency function at a specific frequency is a magnitude
and the phase. The magnitude is equal to the ratio of the magnitude of the
output voltage, obtained from the Fourier analysis, to the magnitude of the
signal from the active signal generator. The phase is equal to the phase of the
output voltage, obtained from the Fourier analysis, since the phase of the
signal from the active signal generator is zero.
The magnitude of the signal from the active signal generator should be
small to avoid nonlinear properties of the converter. However, it should not
be too small since this can result in numerical problems in the simulator. To
obtain confidence that a suitable magnitude is chosen, we conduct an extra
simulation with another magnitude. The magnitude is changed by least a
factor of 2. If the change of the result of the frequency function at the specific
frequency is negligible, the originally chosen magnitude is considered suitable.
A suitable solver algorithm and a small step size must be used in the
simulation to obtain accurate results. We use the settings shown in Table 2.4
for the simulator solver. To obtain confidence that these settings are suitable,
extra simulations have been conducted with other settings. The “Max step
size” and the “Relative tolerance” have been reduced to one tenth of their
values in the table but the change in the simulation result where negligible.
For each one of the frequencies mentioned previously, we conduct a
simulation. This procedure is repeated three times since there are three
different signal generators.
The parameter values shown in Table 2.1 are used in all the simulations
except for v g (t ) and d (t ) . They are of course not constant in the
simulations. The values in the table are instead their dc values.

Table 2.4: The settings for the simulator solver.

Name Value
Variable-step,
Type
ode23tb (stiff/TR-BDF2)
Max step size 2e-7
Relative tolerance 1e-3
Initial step size auto
Absolute tolerance auto
Chapter 2. State-Space Averaging 39

Simulation Results

In this subsection, the frequency functions predicted by the transfer

functions derived in Section 2.3 are compared with simulation results.
Figure 2.8 shows the Bode plot for the control-to-output transfer
function (2.90). The frequency function obtained in the case where the
magnitude of the signal dhat is non-zero in the simulations is also shown in
the figure. The figure shows that the control-to-output transfer function
derived in Section 2.3 agrees closely with the simulation results.
Figure 2.9 shows the Bode plot for the output impedance (2.91). The
frequency function obtained in the case where the magnitude of the signal
iinjhat is non-zero in the simulations is also shown in the figure. However,
the phase of the frequency function is shifted 180 degrees due to the
definition of the output impedance in (2.91). The figure shows that the
output impedance derived in Section 2.3 agrees closely with the simulation
results.
Figure 2.10 shows the Bode plot for the audio susceptibility (2.92). The
frequency function obtained in the case where the magnitude of the signal
vghat is non-zero in the simulations is also shown in the figure. The figure
shows that the audio susceptibility derived in Section 2.3 agrees closely with
the simulation results.

2.5 Operation of the Boost Converter

In this section, the circuit and operation of the boost converter are
presented. A linear model of the boost converter will be derived in Section 2.6
and compared with simulation results in Section 2.7. The boost converter will
be treated similarly as the buck converter in the previous sections.
Figure 2.11 shows the circuit that will be used for the boost converter.
Figure 2.12 shows the waveforms of the signals in the circuit in steady state.
We obtained the waveforms by using the simulation model that will be
presented in Section 2.7. Table 2.5 shows the parameter values used in the
simulation. The control signal, δ (t ) , contains pulses with constant width in
steady state. The transistor is on during t on and off during t off . The voltage
across the inductor, v L (t ) , is equal to the input voltage, v g (t ) , during t on .
v g (t ) is held constant in the simulation. The inductor current, i L (t ) , is
proportional to the integral of v L (t ) . Therefore, i L (t ) increases during t on .
40 Chapter 2. State-Space Averaging

40

20

0
Phase (deg); Magnitude (dB)

-20

1 2 3 4
10 10 10 10

-50

-100

-150

-200
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 2.8: The control-to-output transfer function of a buck converter. Solid

line: the analytic model. X: the simulation results.

The voltage across the diode, v diode (t ) , is equal to the output voltage, v o (t ) ,
during t on . Consequently, the diode current, idiode (t ) , is zero during t on
and the transistor current, itrans (t ) , is equal to i L (t ) . idiode (t ) is equal to
i L (t ) during t off since itrans (t ) is zero. v diode (t ) is zero during t off since
the diode is conducting. Therefore, v L (t ) is equal to the difference between
v g (t ) and v o (t ) during t off . v L (t ) would be positive if v o (t ) was lower
than v g (t ) and i L (t ) would continue to increase during t off . This cannot
be the case in steady state. v o (t ) must therefore be higher than v g (t ) . v L (t )
is thus negative during t off and it is almost constant if the converter is
reasonable designed, i.e. that v o (t ) exhibits low ripple. Consequently, the
slope of i L (t ) is almost constant during each time interval.
The load current, iload (t ) , is almost constant since v o (t ) is almost
constant. The capacitor current, icap (t ) , is equal to the difference between
idiode (t ) and iload (t ) . The mean value of icap (t ) must be zero in steady
state. The mean value of iload (t ) is therefore equal to the mean value of
idiode (t ) .
Chapter 2. State-Space Averaging 41

-20
Phase (deg); Magnitude (dB)

-40

-60
1 2 3 4
10 10 10 10

100

50

-50

-100
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 2.9: The output impedance of a buck converter. Solid line: the
analytic model. X: the simulation results.

Table 2.5: The parameter values used in the simulation of the boost
converter.

Parameter Value
L 37.5 µH
C 400 µF
Rc 14 mΩ
R 1Ω
v g (t ) 5V
d (t ) 0.382
Ts 20 µs
42 Chapter 2. State-Space Averaging

-20
Phase (deg); Magnitude (dB)

-40

-60
1 2 3 4
10 10 10 10

-50

-100

-150

-200
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 2.10: The audio susceptibility of a buck converter. Solid line: the
analytic model. X: the simulation results.

iL(t) L idiode(t) iload(t)

vL(t)
vdiode(t)
vESR(t) Rc
vg(t) R vo(t)
v(t) C
Driver itrans(t) icap(t)

δ (t)

Figure 2.11: The circuit of the boost converter.

Chapter 2. State-Space Averaging 43

2
δ (t ) 1
0
5
v L (t ) 0
-5
14
i L (t ) 13
12
10
vdiode (t ) 5
0
20
idiode (t ) 10
0
20
itrans (t ) 10
0
8.5
iload (t ) 8
7.5
10
icap (t ) 0
-10
8.1
v(t ) 8
7.9
0.2
v ESR (t ) 0
-0.2
8.5
vo (t ) 8
7.5
0 0.5 1 1.5 2 2.5 3 3.5 4
t on t off t on t off -5
x 10

t (s)

Figure 2.12: The waveforms of the signals in steady state for a boost
converter. The unit of the voltages is Volt and the unit of the
currents is Ampere.
44 Chapter 2. State-Space Averaging

The voltage across the (ideal) capacitor, v(t ) , is proportional to the

integral of icap (t ) . The voltage across the capacitor’s ESR, v ESR (t ) , is
proportional to icap (t ) . The output voltage, v o (t ) , is equal to the sum of
v(t ) and v ESR (t ) .
The capacitor and the load are fed by idiode (t ) . idiode (t ) is discontinuous
and the magnitude of its ripple is high. Therefore, v o (t ) is also discontinuous
(due to the capacitor’s ESR). This is not the case for the buck converter where
the capacitor and the load are fed by the inductor current. The inductor
current is continuous and it is possible to keep the magnitude of its ripple at a
reasonable level. It is therefore more difficult to obtain a low magnitude of
the ripple in v o (t ) for the boost converter than for the buck converter.

2.6 Model of the Boost Converter

A linear time-invariant model of the boost converter is derived by means
of state-space averaging in this section. The methodology is analogous to the
buck converter modeling.

State-Space Description for Each Time Interval

In this subsection, state-space descriptions of the boost converter are

derived for the cases where the transistor operates in the on-state and in the
off-state.
While the transistor is on, the voltage across the transistor is zero and the
diode is not conducting. The circuit in Figure 2.13 can be used as a model of
the boost converter during t on . In the figure, a current source is added.
From Figure 2.13, the following equations are obtained:

di L (t ) 1
= v g (t ) , (2.93)
dt L

dv(t ) 1  v o (t ) 
= − − iinj (t )  , (2.94)
dt C R 

 v (t ) 
v o (t ) = v(t ) + Rc  − o − iinj (t )  . (2.95)
 R 
Chapter 2. State-Space Averaging 45

iL( L iinj(t)

) vL(t)
Rc
vg(t) R vo(t)
v(t) C

Figure 2.13: The circuit of the boost converter during t on .

(2.95) is rearranged to:

Rc
v o (t ) + v o (t ) = v(t ) − Rc iinj (t ) , (2.96)
R

v(t ) − Rc iinj (t )
v o (t ) = , (2.97)
1 + Rc R

R RRc
v o (t ) = v(t ) − iinj (t ) . (2.98)
R + Rc R + Rc

(2.98) is used to substitute v o (t ) in (2.94):

dv(t ) 1 Rc 1
=− v(t ) + iinj (t ) − iinj (t ) . (2.99)
dt (R + Rc )C (R + Rc )C C

(2.99) is simplified:

dv(t ) 1 R
=− v(t ) − i (t ) . (2.100)
dt (R + Rc )C (R + Rc )C inj
By using (2.93), (2.100), and (2.98), the following state-space system is
obtained:
46 Chapter 2. State-Space Averaging

 dx(t )
 = A 1 x(t ) + B 1u(t )
 dt (2.101)
 y (t ) = C 1 x(t ) + E 1u (t )

where

i (t )
x(t ) =  L  , (2.102)
 v(t ) 

 v g (t ) 
u(t ) =  , (2.103)
iinj (t )

y (t ) = vo (t ) , (2.104)

0 0 
A 1 = 0 − 1 , (2.105)
 ( + ) 
 R R c C 

1 
 0 
B1 =  L R , (2.106)
0 − 
 (R + Rc )C 

 R 
C 1 = 0 , (2.107)
 R + Rc 

 RRc 
E 1 = 0 − . (2.108)
 R + Rc 

While the transistor is off, the voltage across the diode is equal to zero.
The circuit in Figure 2.3 can therefore be used as a model of the boost
converter during t off . A state-space model for this circuit is presented in
Section 2.3, (2.11)-(2.18). The state-space model of the boost converter
during t off is thus
Chapter 2. State-Space Averaging 47

 dx(t )
 = A 2 x(t ) + B 2 u(t )
 dt , (2.109)
 y (t ) = C 2 x(t ) + E 2 u(t )

where

 RRc R 
 − (R + R )L −
(R + Rc )L  ,
A2 =  c
(2.110)
 R
−
1 
 (R + Rc )C (R + Rc )C 

1 RRc 
L (R + Rc )L  ,
B2 =  (2.111)
0 −
R 
 (R + Rc )C 

 RRc R 
C2 =  , (2.112)
 R + Rc R + Rc 

 RRc 
E 2 = 0 −  = E1 . (2.113)
 R + Rc 

Applying State-Space Averaging

The method of state-space averaging is applied to the boost converter in

this subsection.
The following equations are obtained if (2.41)-(2.44) are expanded:

 RRc R 
− D' (R + R )L − D'
(R + Rc )L  ,
A= c
(2.114)
 D' R
−
1 
 (R + Rc )C (R + Rc )C 
48 Chapter 2. State-Space Averaging

1 RRc 
L D'
B=
(R + Rc )L  , (2.115)
0 −
R 
 (R + Rc )C 

 RRc R 
C =  D' , (2.116)
 R + Rc R + Rc 

 RRc 
E = 0 − . (2.117)
 R + Rc 

The dc equations will now be derived. The following equations are

obtained if (2.40) is expanded:

RRc D ' RD' 1

0=− I − V + Vg , (2.118)
(R + Rc )L L (R + Rc )L L

RD' 1
0= IL − V, (2.119)
(R + Rc )C (R + Rc )C
RRc D' R
Vo = IL + V. (2.120)
R + Rc R + Rc

(2.119) is simplified to:

V = RD ' I L . (2.121)

(2.121) is used to rewrite (2.120):

Rc R
Vo = V+ V =V . (2.122)
R + Rc R + Rc

(2.121) is used to rewrite (2.118):

Chapter 2. State-Space Averaging 49

Rc R
0=− V− D 'V + V g , (2.123)
(R + Rc ) (R + Rc )
V R + Rc R + Rc 1
= = = .
V g RD'+ Rc R − RD + Rc R (2.124)
1− D
R + Rc

The dc voltage across the capacitor’s ESR is zero and this explains the result
in (2.122). The mean value of the load current is equal to the mean value of
the diode current. The diode current is equal to the inductor current during
the fraction D ' of the time and equal to zero otherwise. The mean value of
the diode current is therefore (approximately) equal to D' I L and this
explains the result in (2.121). (2.124) shows the dc amplification of the boost
converter and it is higher than 1.
(2.53) and (2.54) are expanded and simplified:

 RRc R 
 (R + R )L   RRc 
( + )  0 −
(R + Rc )L  U =
R R L
Bd =  c c X +
− R  
0 0 0 
 (R + Rc )C 
(2.125)
 RRc R 
 (R + R )L I L + (R + R )L V 
 c c ,
 −
R
IL 
 (R + Rc )C 

 RRc  RRc
E d = − 0 X + 0U = − IL . (2.126)
 R + Rc  R + Rc

I L and V in (2.125) and (2.126) are replaced by using (2.121) and (2.124):
50 Chapter 2. State-Space Averaging

 Rc RD'   (RD'+ Rc )V 
 (R + R )LD ' V + (R + R )LD ' V   (R + R )LD ' 
Bd =  c c = c =
 −
1
V   −
V 
 (R + Rc )CD'   (R + Rc )CD ' 
(2.127)
 Vg 
 
 LD ' ,
− Vg 
 (RD'+ R )CD ' 
 c 

Rc R cV g
Ed = − V =− . (2.128)
(R + Rc )D' (RD'+ Rc )D'
(2.51) and (2.52) are expanded:

1 RRc D ' Vg 
 
B' = 
L (R + Rc )L LD ' , (2.129)
 R Vg 
0 − −
 (R + Rc )C (RD'+ Rc )CD' 

 RRc Rc V g 
E' = 0 − − . (2.130)
 R + Rc D ' (RD'+ Rc ) 

All the coefficient matrices in the ac model (2.50) are now available.

Extracting the Transfer Functions

The control-to-output transfer function, the output impedance and the

audio susceptibility will now be extracted from the linearized system. The first
equation in (2.78) is written on the form:
Chapter 2. State-Space Averaging 51

−1
 RRc D ' RD' 
 (R + R )L + s (R + Rc )L 
xˆ ( s ) =  c •
− RD' 1
+ s
 (R + Rc )C (R + Rc )C 
(2.131)
1 RRc D ' Vg 
 
L (R + Rc )L LD ' uˆ ' ( s ) .
 R Vg 
0 − −
 (R + Rc )C (RD'+ Rc )CD' 
The matrix inversion in (2.131) is calculated:

1
xˆ ( s ) = •
 RRc D '  1  R 2 D' 2
 + s  + s +
 (R + R )L   (R + R )C 
 (R + Rc ) LC
2
 c  c

 1 RD' 
 (R + R )C + s −
(R + Rc )L  •
 c
 RD ' RRc D '
+ s
(2.132)
 (R + Rc )C (R + Rc )L 
1 RRc D ' Vg 
 
L (R + Rc )L LD '  uˆ ' ( s ) .
 R Vg 
0 − −
 (R + Rc )C (RD'+ Rc )CD' 
(2.132) is modified to:
52 Chapter 2. State-Space Averaging

1
xˆ ( s ) = •
RRc D '+ R D '
2
L + RRc CD '
2
+s + s2
(R + Rc ) LC
2 (R + Rc )LC
 1 1 RRc D ' RRc D ' R 2 D'
 +s +s +
 (R + Rc )LC L (R + Rc )2 LC (R + Rc )L (R + Rc )2 LC
 RD ' R 2 Rc D ' 2 R 2 Rc D ' R (2.133)
 − −s
 (R + Rc )LC (R + Rc ) LC (R + Rc ) LC (R + Rc )C
2 2

Vg Vg  RD 'V g
+s +
(R + Rc )LCD ' LD' (R + Rc )(RD'+ Rc )LCD ' uˆ ' ( s) ,
RV g RRc D'V g Vg 
− −s
(R + Rc )LC (R + Rc )(RD'+ Rc )LCD' (RD'+ Rc )CD' 

1
xˆ ( s ) = •
RD ' (RD '+ Rc ) (R + Rc ) + s(L + RRc CD ') + s 2 (R + Rc )LC
1 + s(R + Rc )C RD' (1 + sRc C )
 R 2 Rc DD'
 RD' − − sRL
 (R + Rc ) (2.134)
Vg  RD'  
1 + + s ( R + R )C  
D'  
c
RD '+ Rc   uˆ ' ( s ) .

Vg
D ' (RD'+ Rc )
(
R 2 D ' 2 − s (R + Rc )L  )


Six transfer functions are derived from (2.134):

iˆL ( s )
=
dˆ ( s )
Vg   (2.135)
1 + RD ' + s (R + Rc )C 
D'  RD'+ Rc 
,
RD ' (RD '+ Rc ) (R + Rc ) + s(L + RRc CD ') + s 2 (R + Rc )LC
Chapter 2. State-Space Averaging 53

vˆ( s )
=
dˆ ( s )
Vg
D ' (RD '+ R )
(R 2
D ' 2 − s (R + Rc )L ) (2.136)
c
,
RD ' (RD '+ Rc ) (R + Rc ) + s(L + RRc CD ') + s 2 (R + Rc )LC

iˆL ( s )
=
iˆinj ( s)
(2.137)
RD ' (1 + sRc C )
,
RD ' (RD '+ Rc ) (R + Rc ) + s(L + RRc CD ') + s 2 (R + Rc )LC

vˆ( s)
=
iˆinj ( s)
R 2 Rc DD ' (2.138)
− − sRL
(R + Rc )
,
RD ' (RD '+ Rc ) (R + Rc ) + s(L + RRc CD ') + s 2 (R + Rc )LC

iˆL ( s )
=
vˆ g ( s )
(2.139)
1 + s (R + Rc )C
,
RD ' (RD '+ Rc ) (R + Rc ) + s(L + RRc CD ') + s 2 (R + Rc )LC

vˆ( s )
=
vˆ g ( s )
(2.140)
RD'
.
RD ' (RD '+ Rc ) (R + Rc ) + s(L + RRc CD ') + s 2 (R + Rc )LC

The second equation in (2.78) is expanded:

54 Chapter 2. State-Space Averaging

 RRc R 
yˆ ( s ) =  D'  xˆ ( s ) +
 R + Rc R + Rc 
(2.141)
 RRc R cV g 
0 − −
(RD'+ Rc )D' 
uˆ ' ( s ) .
 R + Rc

The control-to-output transfer function is obtained by combining (2.141),

(2.135), and (2.136):

vˆo ( s ) RRc D ' iˆL ( s ) R vˆ( s ) R cV g

= + − =K=
dˆ ( s ) R + Rc dˆ ( s ) R + Rc dˆ ( s ) (RD'+ Rc )D '
Vg
(RD'+ R )D'
(R 2
)
D ' 2 − s (R + Rc )L (1 + sRc C )
(2.142)
c
.
RD ' (RD '+ Rc ) (R + Rc ) + s(L + RRc CD ') + s 2 (R + Rc )LC

The output impedance is obtained by combining (2.141), (2.137), and

(2.138):

vˆo ( s )
Z out ( s ) = − =
iˆinj ( s)
RRc D ' iˆL ( s ) R vˆ( s) RRc
− − + =K=
R + Rc iˆinj ( s ) R + Rc iˆinj ( s ) R + Rc (2.143)
 R Rc DD ' 2 
 + sRL  (1 + sRc C )
 (R + Rc ) 
 
.
RD ' (RD '+ Rc ) (R + Rc ) + s(L + RRc CD ') + s 2 (R + Rc )LC

The audio susceptibility is obtained by combining (2.141), (2.139), and

(2.140):
Chapter 2. State-Space Averaging 55

vˆo ( s) RRc D ' iˆL ( s ) R vˆ( s )

= + =
vˆ g ( s ) R + Rc vˆ g ( s ) R + Rc vˆ g ( s )
1
R + Rc
(
RRc D ' (1 + s(R + Rc )C ) + R 2 D' )
=
RD ' (RD '+ Rc ) (R + Rc ) + s(L + RRc CD ') + s 2 (R + Rc )LC
(2.144)
1
(RD' (R + Rc ) + sRRc D' (R + Rc )C )
R + Rc
=
RD ' (RD '+ Rc ) (R + Rc ) + s(L + RRc CD ') + s 2 (R + Rc )LC
RD ' (1 + sRc C )
.
RD ' (RD '+ Rc ) (R + Rc ) + s(L + RRc CD ') + s 2 (R + Rc )LC

The zero (1 + sRc C ) is added to the transfer function in the case where
vˆo ( s ) is the output compared to the case where vˆ( s ) is the output. This is
apparent by comparison of (2.144), (2.140), (2.143), (2.138), (2.142), and
(2.136). The buck converter exhibits the same property, see Section 2.3. Note
that the control-to-output transfer function, (2.142), has a right half plane
zero and is not a minimum phase system. This is not the case for the buck
converter.

2.7 Simulation of a Boost Converter

In this section, a simulation model for a boost converter is presented. The
frequency functions predicted by the transfer functions derived in Section 2.6
are compared with simulation results.
To simulate a boost converter the simulation model shown in Figure 2.5
is used, except the buck converter subsystem is replaced with the boost
converter subsystem shown in Figure 2.14.
The parameter values shown in Table 2.5 are used in all the simulations
except for v g (t ) and d (t ) . They are of course not constant in the
simulations. The values in the table are instead their dc values.
Figure 2.15 shows the Bode plot for the control-to-output transfer
56 Chapter 2. State-Space Averaging

3 2 1
iL iload vo
i i
+ - 1 2 + - +
v
L g m -
Diode signal
emulator
+ 2 m Rc
signal
- Rdummy Transistor R
1 g
C

NOT -1

1 3 2
vg delta iinj

Figure 2.14: The boost converter subsystem.

30

20

10

0
Phase (deg); Magnitude (dB)

-10

-20
1 2 3 4
10 10 10 10

-50

-100

-150

-200

-250
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 2.15: The control-to-output transfer function of a boost converter.

Solid line: the analytic model. X: the simulation results.
Chapter 2. State-Space Averaging 57

-10

-20
Phase (deg); Magnitude (dB)

-30

-40
1 2 3 4
10 10 10 10

100

50

-50

-100
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 2.16: The output impedance transfer function of a boost converter.

Solid line: the analytic model. X: the simulation results.

function (2.142). The frequency function obtained in the case where the
magnitude of the signal dhat is non-zero in the simulations is also shown in
the figure. The figure shows that the control-to-output transfer function
derived in Section 2.6 agrees closely with the simulation results.
Figure 2.16 shows the Bode plot for the output impedance (2.143). The
frequency function obtained in the case where the magnitude of the signal
iinjhat is non-zero in the simulations is also shown in the figure. However,
the phase of the frequency function is shifted 180 degrees due to the
definition of the output impedance in (2.143). The figure shows that the
output impedance derived in Section 2.6 agrees closely with the simulation
results.
Figure 2.17 shows the Bode plot for the audio susceptibility (2.144). The
frequency function obtained in the case where the magnitude of the signal
vghat is non-zero in the simulations is also shown in the figure. The figure
shows that the audio susceptibility derived in Section 2.6 agrees closely with
the simulation results.
58 Chapter 2. State-Space Averaging

20

-20
Phase (deg); Magnitude (dB)

-40

-60
1 2 3 4
10 10 10 10

-50

-100

-150

-200
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 2.17: The audio susceptibility transfer function of a boost converter.

Solid line: the analytic model. X: the simulation results.

2.8 Operation of the Buck-Boost Converter

In this section, the circuit and operation of the buck-boost converter are
presented. A linear model of the buck-boost converter will be derived in
Section 2.9 and compared with simulation results in Section 2.10. The buck-
boost converter will be treated similarly as the buck and boost converters in
the previous sections.
Figure 2.18 shows the circuit that will be used for the buck-boost
converter. Figure 2.19 shows the waveforms of the signals in the circuit in
steady state. A simulation model will be presented in Section 2.10 and this
model is used to generate the waveforms. The parameter values in Table 2.6
are used in the simulation. The control signal, δ (t ) , is pulse shaped and all
the pulses have the same width in steady state. The transistor is in the on state
during t on and in the off state during t off . The voltage across the inductor,
v L (t ) , is equal to the input voltage, v g (t ) , during t on . In the simulation,
Chapter 2. State-Space Averaging 59

itrans(t) idiode(t) L iload(t)

vdiode(t)
vESR(t) Rc
vg(t) Driver vL(t) R vo(t)
v(t) C
iL(t) icap(t)

δ (t)

Figure 2.18: The circuit of the buck-boost converter.

Table 2.6: The parameter values used in the simulation of the buck-boost
converter.

Parameter Value
L 37.5 µH
C 400 µF
Rc 14 mΩ
R 1Ω
v g (t ) 5V
d (t ) 0.620
Ts 20 µs

v g (t ) does not change. The inductor current, i L (t ) , is proportional to the

integral of v L (t ) and i L (t ) increases during t on . The transistor is not
conducting during t off so the diode must conduct instead and the voltage
across the diode is zero. Therefore, v L (t ) is equal to − vo (t ) , i.e. minus the
output voltage. In steady state, i L (t ) must decrease during t off since it
increases during t on . Consequently, v L (t ) must be negative during t off and
v o (t ) must be positive. Note the definition of the polarity of v o (t ) in Figure
2.18. It was defined the other way round for the buck and boost converters.
60 Chapter 2. State-Space Averaging

2
δ (t ) 1
0
10
v L (t ) 0
-10
22
i L (t ) 21
20
20
vdiode (t ) 10
0
40
idiode (t ) 20
0
40
itrans (t ) 20
0
8.5
iload (t ) 8
7.5
20
icap (t ) 0
-20
8.5
v(t ) 8
7.5
0.2
v ESR (t ) 0
-0.2
8.5
vo (t ) 8
7.5
0 0.5 1 1.5 2 2.5 3 3.5 4
t on t off t on t off -5
x 10

t (s)

Figure 2.19: The waveforms of the signals in steady state for a buck-boost
converter. The unit of the voltages is Volt and the unit of the
currents is Ampere.
Chapter 2. State-Space Averaging 61

v L (t ) is almost constant during t off if the converter is reasonable designed,

i.e. that the magnitude of the ripple in v o (t ) is low. Hence, during each time
interval the slope of i L (t ) is almost constant.
The voltage across the diode, v diode (t ) , is equal to v g (t ) + v o (t ) during
t on . The diode current, idiode (t ) , is therefore zero during t on and the
transistor current, itrans (t ) , is equal to i L (t ) . idiode (t ) is equal to i L (t )
during t off since itrans (t ) is zero.
Since v o (t ) is almost constant, also the load current, iload (t ) , is almost
constant. The capacitor current, icap (t ) , is equal to the difference between
idiode (t ) and iload (t ) . In steady state, the mean value of icap (t ) is zero and
the mean value of iload (t ) is therefore equal to the mean value of idiode (t ) .
The voltage across the (ideal) capacitor, v(t ) , is proportional to the
integral of icap (t ) . The voltage across the capacitor’s ESR, v ESR (t ) , is
proportional to icap (t ) . The output voltage, v o (t ) , is equal to the sum of
v(t ) and v ESR (t ) .
The capacitor and the load are fed by idiode (t ) which is the case also for
the boost converter. See Section 2.5 for a discussion about the consequences.

2.9 Model of the Buck-Boost Converter

A linear time-invariant model of the buck-boost converter is derived by
means of state-space averaging in this section. The methodology is analogous
to the modeling of the buck and boost converters.

State-Space Description for Each Time Interval

In this subsection, state-space descriptions of the buck-boost converter are

derived for the cases where the transistor operates in the on-state and in the
off-state.
While the transistor is on, the voltage across the transistor is zero and the
diode is not conducting. The circuit in Figure 2.13 can therefore be used as a
model of the buck-boost converter during t on . A consequence of using this
circuit is that the injected current, iinj (t ) , is defined in opposite direction
compared to the definition used for the boost converter. This will be
compensated for in the simulation model for the buck-boost converter.
(Compare Figure 2.14 and Figure 2.20.) A state-space model for the circuit in
Figure 2.13 is presented in Section 2.6, (2.101)-(2.108). The state-space
model of the buck-boost converter during t on is thus
62 Chapter 2. State-Space Averaging

 dx(t )
 = A 1 x(t ) + B 1u(t )
 dt , (2.145)
 y (t ) = C 1 x(t ) + E 1u (t )

where

i (t )
x(t ) =  L  , (2.146)
 v(t ) 

 v g (t ) 
u(t ) =  , (2.147)
iinj (t )

y (t ) = vo (t ) , (2.148)

0 0 
A 1 = 0 − 1 , (2.149)
 ( + ) 
 R R c C 

1 
 0 
B1 =  L R , (2.150)
0 − 
 (R + Rc )C 

 R 
C 1 = 0 , (2.151)
 R + Rc 

 RRc 
E 1 = 0 − . (2.152)
 R + Rc 

While the transistor is off, the voltage across the diode is equal to zero.
The circuit in Figure 2.4 can therefore be used as a model of the buck-boost
converter during t off . A state-space model for this circuit is presented in
Section 2.3, (2.19)-(2.23). The state-space model of the buck-boost converter
during t off is thus
Chapter 2. State-Space Averaging 63

 dx(t )
 = A 2 x(t ) + B 2 u(t )
 dt , (2.153)
 y (t ) = C 2 x(t ) + E 2 u(t )

where

 RRc R 
 − (R + R )L −
(R + Rc )L  ,
A2 =  c
(2.154)
 R
−
1 
 (R + Rc )C (R + Rc )C 

 RRc 
0 (R + R )L 
B2 =  c , (2.155)
0 − R 
 (R + Rc )C 

 RRc R 
C2 =  , (2.156)
 R + Rc R + Rc 

 RRc 
E 2 = 0 −  = E1 . (2.157)
 R + Rc 

Applying State-Space Averaging

The method of state-space averaging is applied to the buck-boost

converter in this subsection.
The following equations are obtained if (2.41)-(2.44) are expanded:

 RRc R 
− D' (R + R )L − D'
(R + Rc )L  ,
A= c
(2.158)
 D' R
−
1 
 (R + Rc )C (R + Rc )C 
64 Chapter 2. State-Space Averaging

 1 RRc 
D L D'
(R + Rc )L  ,
B= (2.159)
 0 −
R 
 (R + Rc )C 

 RRc R 
C =  D' , (2.160)
 R + Rc R + Rc 

 RRc 
E = 0 − . (2.161)
 R + Rc 

The dc equations will now be derived. The following equations are

obtained if (2.40) is expanded:

RRc D ' RD ' D

0=− I − V + Vg , (2.162)
(R + Rc )L L (R + Rc )L L

RD' 1
0= IL − V, (2.163)
(R + Rc )C (R + Rc )C
RRc D' R
Vo = IL + V. (2.164)
R + Rc R + Rc

(2.163) is rewritten:

V = RD ' I L . (2.165)

(2.165) is used to rewrite (2.164):

Rc R
Vo = V+ V =V . (2.166)
R + Rc R + Rc

(2.165) is used to rewrite (2.162):

Chapter 2. State-Space Averaging 65

Rc R
0=− V− D'V + DV g , (2.167)
R + Rc R + Rc

V D (R + Rc ) D (R + Rc ) D
= = = .
Vg RD'+ Rc R − RD + Rc R (2.168)
1− D
R + Rc

The dc voltage across the capacitor’s ESR is zero and this explains the result
in (2.166). The mean value of the load current is equal to the mean value of
the diode current. The diode current is equal to the inductor current during
the fraction D ' of the time and equal to zero otherwise. The mean value of
the diode current is therefore (approximately) equal to D' I L and this
explains the result in (2.165). Eq. (2.168) shows the dc amplification of the
buck-boost converter. It is higher than 1 for D > (R + Rc ) (2 R + Rc ) .
(2.53) and (2.54) is expanded and simplified:

 RRc R 
 (R + R )L  1 RRc 
( + ) X +  L
−
(R + Rc )L  U =
R R L
Bd =  c c
− R  
0 0 0 
 (R + Rc )C 
(2.169)
 RRc R 1 
 (R + R )L I L + (R + R )L V + L V g 
 c c ,
 −
R
IL 
 (R + Rc )C 

 RRc  RRc
E d = − 0 X + 0U = − IL . (2.170)
 R + Rc  R + Rc

I L and V in (2.169) and (2.170) is replaced by using (2.165) and (2.168):

66 Chapter 2. State-Space Averaging

 Rc RD' 1 
 (R + R )LD ' V + (R + R )LD ' V + L V g 
Bd =  c c =
 −
1
V 
 (R + Rc )CD' 
(2.171)
 (RD '+ Rc )V V g   DV g D'V g   Vg 
 +   +   
 (R + Rc )LD ' L  =  LD ' LD '  =  LD ' ,
 V  − DV g   DV g 
 − (R + R )CD '   (RD '+ R )CD '  − (RD'+ R )CD ' 
 c   c   c 

Rc Rc DV g
Ed = − V =− . (2.172)
(R + Rc )D' (RD'+ Rc )D'
(2.51) and (2.52) are expanded:

D RRc D ' Vg 
 
B' = 
L (R + Rc )L LD ' , (2.173)
 R DV g 
0 − −
 (R + Rc )C (RD'+ Rc )CD' 

 RRc Rc DV g 
E' = 0 − −
(RD'+ Rc )D' 
. (2.174)
 R + Rc

All the coefficient matrices in the ac model (2.50) are now available.

Extracting the Transfer Functions

The control-to-output transfer function, the output impedance and the

audio susceptibility will now be extracted from the linearized system. The first
equation in (2.78) is expanded:
Chapter 2. State-Space Averaging 67

−1
 RRc D ' RD ' 
 (R + R )L + s (R + Rc )L 
xˆ ( s ) =  c •
− RD ' 1
+ s
 (R + Rc )C (R + Rc )C 
(2.175)
D RRc D ' Vg 
 
L (R + Rc )L LD ' uˆ ' ( s) .
 R DV g 
0 − −
 (R + Rc )C (RD'+ Rc )CD' 
The matrix inversion in (2.175) is calculated:

1
xˆ ( s ) = •
 RRc D '  1  R 2 D' 2
 + s  + s +
 (R + R )L   (R + R )C 
 (R + Rc ) LC
2
 c  c

 1 RD ' 
 (R + R )C + s −
(R + Rc )L  •
 c
 RD' RRc D '
+ s
(2.176)
 (R + Rc )C (R + Rc )L 
D RRc D' Vg 
 
L (R + Rc )L LD '  uˆ ' ( s ) .
 R DV g 
0 − −
 (R + Rc )C (RD'+ Rc )CD' 
(2.176) is modified to:
68 Chapter 2. State-Space Averaging

1
xˆ ( s ) = ×
RRc D'+ R D'
2 2
L + RRc CD'
+s + s2
(R + Rc ) LC
2 (R + Rc )LC
 D D RRc D' RRc D' R 2 D'
 +s +s +
 (R + Rc )LC L (R + Rc )2 LC (R + Rc )L (R + Rc )2 LC
 RDD' R 2 Rc D ' 2 R 2 Rc D ' R (2.177)
 − −s
 (R + Rc )LC (R + Rc ) LC (R + Rc ) LC (R + Rc )C
2 2

Vg Vg 
RDD'V g
+s +
(R + Rc )LCD ' LD' (R + Rc )(RD'+ Rc )LCD' uˆ ' (s) ,
RV g RRc DD 'V g DV g 
− −s
(R + Rc )LC (R + Rc )(RD'+ Rc )LCD ' (RD'+ Rc )CD' 

1
xˆ ( s ) = •
RD ' (RD '+ Rc ) (R + Rc ) + s(L + RRc CD ') + s 2 (R + Rc )LC
 D(1 + s (R + Rc )C ) RD' (1 + sRc C )
 R 2 Rc DD'
 RDD' − − sRL
 (R + Rc ) (2.178)
Vg  RDD' 
1 + + s ( R + R )C 
D '  
c
RD '+ Rc  uˆ ' ( s ) .
V g (R + Rc ) 
D' (RD'+ Rc )
(
RD ' 2 − sLD )


Six transfer functions are now derived from (2.178):

iˆL ( s )
=
dˆ ( s )
Vg   (2.179)
1 + RDD' + s(R + Rc )C 
D'  RD '+ Rc 
,
RD ' (RD '+ Rc ) (R + Rc ) + s(L + RRc CD ') + s 2 (R + Rc )LC
Chapter 2. State-Space Averaging 69

vˆ( s )
=
dˆ ( s )
V g (R + Rc )
D' (RD'+ R )
(RD' 2
− sLD ) (2.180)
c
,
RD ' (RD '+ Rc ) (R + Rc ) + s(L + RRc CD ') + s 2 (R + Rc )LC

iˆL ( s )
=
iˆinj ( s)
(2.181)
RD ' (1 + sRc C )
,
RD ' (RD'+ Rc ) (R + Rc ) + s(L + RRc CD ') + s 2 (R + Rc )LC

vˆ( s)
=
iˆinj ( s)
R 2 Rc DD ' (2.182)
− − sRL
(R + Rc )
,
RD ' (RD '+ Rc ) (R + Rc ) + s(L + RRc CD ') + s 2 (R + Rc )LC

iˆL ( s )
=
vˆ g ( s )
(2.183)
D(1 + s (R + Rc )C )
,
RD ' (RD '+ Rc ) (R + Rc ) + s(L + RRc CD ') + s 2 (R + Rc )LC

vˆ( s )
=
vˆ g ( s )
(2.184)
RDD'
.
RD ' (RD '+ Rc ) (R + Rc ) + s(L + RRc CD ') + s 2 (R + Rc )LC

The second equation in (2.78) is expanded:

70 Chapter 2. State-Space Averaging

 RRc R 
yˆ ( s ) =  D'  xˆ ( s ) +
 R + Rc R + Rc 
(2.185)
 RRc Rc DV g 
0 − −
(RD'+ Rc )D' 
uˆ ' ( s ) .
 R + Rc

The control-to-output transfer function is obtained by combining (2.185),

(2.179), and (2.180):

vˆo ( s ) RRc D ' iˆL ( s ) R vˆ( s ) Rc DV g

= + − =K=
dˆ ( s ) R + Rc dˆ ( s ) R + Rc dˆ ( s ) (RD'+ Rc )D '
V g (R + Rc )
(RD'+ Rc )D'
( )
RD' 2 − sLD (1 + sRc C )
(2.186)

.
RD ' (RD '+ Rc ) (R + Rc ) + s(L + RRc CD ') + s 2 (R + Rc )LC

The output impedance is derived by combining (2.185), (2.181), and

(2.182):

vˆo ( s )
Z out ( s ) = − =
iˆinj ( s)
RRc D ' iˆL ( s ) R vˆ( s) RRc
− − + =K=
R + Rc iˆinj ( s ) R + Rc iˆinj ( s ) R + Rc (2.187)
 R Rc DD '
2 
 + sRL  (1 + sRc C )
 (R + Rc ) 
 
.
RD ' (RD '+ Rc ) (R + Rc ) + s(L + RRc CD ') + s 2 (R + Rc )LC

The audio susceptibility is calculated by combining (2.185), (2.183), and

(2.184):
Chapter 2. State-Space Averaging 71

vˆo ( s) RRc D ' iˆL ( s ) R vˆ( s )

= + =
vˆ g ( s ) R + Rc vˆ g ( s ) R + Rc vˆ g ( s )
1
R + Rc
(
RRc DD ' (1 + s(R + Rc )C ) + R 2 DD' )
=
RD ' (RD '+ Rc ) (R + Rc ) + s(L + RRc CD ') + s 2 (R + Rc )LC
(2.188)
1
(RDD' (R + Rc ) + sRRc DD' (R + Rc )C )
R + Rc
=
RD ' (RD '+ Rc ) (R + Rc ) + s(L + RRc CD ') + s 2 (R + Rc )LC
RDD' (1 + sRc C )
.
RD ' (RD '+ Rc ) (R + Rc ) + s(L + RRc CD ') + s 2 (R + Rc )LC

The zero (1 + sRc C ) is added to the transfer function in the case where
vˆo ( s ) is the output compared to the case where vˆ( s ) is the output. This is
apparent by comparison of (2.188), (2.184), (2.187), (2.182), (2.186), and
(2.180). The buck and boost converters exhibit the same property, see Section
2.3 and 2.6. Note that the control-to-output transfer function, (2.186), has a
right half plane zero and is not a minimum phase system. This is also the case
for the boost converter but not for the buck converter.

2.10 Simulation of a Buck-Boost Converter

In this section, a simulation model for a buck-boost converter is
presented. The frequency functions predicted by the transfer functions
derived in Section 2.9 are compared with simulation results.
To simulate a buck-boost converter the simulation model shown in
Figure 2.5 is used, except the buck converter subsystem is replaced with the
buck-boost converter subsystem shown in Figure 2.20.
The parameter values shown in Table 2.6 are used in all the simulations
except for v g (t ) and d (t ) . They are of course not constant in the
simulations. The values in the table are instead their dc values.
Figure 2.21 shows the Bode plot for the control-to-output transfer
function (2.186). The frequency function obtained in the case where the
magnitude of the signal dhat is non-zero in the simulations is also shown in
the figure. The figure shows that the control-to-output transfer function
derived in Section 2.9 agrees closely with the simulation results.
72 Chapter 2. State-Space Averaging

3 2 1
iL iload vo
i -1
1 2 2 1 - + +
v
g m m g -
Transistor Diode signal
+ emulator
+ i - Rc
signal
Rdummy R
-
L C

NOT

1 3 2
vg delta iinj

Figure 2.20: The buck-boost converter subsystem.

40

30

20

10
Phase (deg); Magnitude (dB)

-10
1 2 3 4
10 10 10 10

-50

-100

-150

-200

-250
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 2.21: The control-to-output transfer function of a buck-boost

converter. Solid line: the analytic model. X: the simulation
results.
Chapter 2. State-Space Averaging 73

-10

-20
Phase (deg); Magnitude (dB)

-30

-40
1 2 3 4
10 10 10 10

100

50

-50

-100
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 2.22: The output impedance transfer function of a buck-boost

converter. Solid line: the analytic model. X: the simulation
results.

Figure 2.22 shows the Bode plot for the output impedance (2.187). The
frequency function obtained in the case where the magnitude of the signal
iinjhat is non-zero in the simulations is also shown in the figure. However,
the phase of the frequency function is shifted 180 degrees due to the
definition of the output impedance in (2.187). The figure shows that the
output impedance derived in Section 2.9 agrees closely with the simulation
results.
Figure 2.23 shows the Bode plot for the audio susceptibility (2.188). The
frequency function obtained in the case where the magnitude of the signal
vghat is non-zero in the simulations is also shown in the figure. The figure
shows that the audio susceptibility derived in Section 2.9 agrees closely with
the simulation results.
74 Chapter 2. State-Space Averaging

-20
Phase (deg); Magnitude (dB)

-40

-60

1 2 3 4
10 10 10 10

-50

-100

-150

-200
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 2.23: The audio susceptibility transfer function of a buck-boost

converter. Solid line: the analytic model. X: the simulation
results.

2.11 Summary and Concluding Remarks

The converters are time-variant systems. State-space averaging was used in
this chapter to derive linear continuous-time time-invariant models for the
buck, boost, and buck-boost converters. The control-to-output transfer
function, the output impedance, and the audio susceptibility were extracted
from each one of these models. These results were compared with the
frequency functions of the converters obtained from simulations of switched
models. Only one of the Fourier components in the output voltage was
considered when these comparisons were made. It was the Fourier component
that had the same frequency as the input signal. A network analyzer also just
considers this Fourier component. Evaluation of a converter by means of a
network analyzer is common and this is one of the reasons for the interest in
models that can predict the frequency functions.
Chapter 2. State-Space Averaging 75

The conclusion of the comparison is that the models derived by using

state-space averaging agree closely with the simulation results at the tested
frequencies ( f s 1000 , f s 500 , K , f s 3 ). However, it should be noted that
the linearized models are valid only for small signals.
 Chapter 3 Current-Mode
Control

In this chapter we consider models for converters with current-mode

control. Various transfer functions are derived and subsequently compared
with simulation results. Switched (large-signal) simulation models are utilized.
In Chapter 6, some of the derived transfer functions will be approximated
and then, in Chapter 7, used to analyze some properties that can be obtained
when load current measurements are utilized for control.

3.1 Introduction
In current-mode control, the inductor current is feed back in an inner
control loop and the output voltage is feed back in an outer control loop.
Current-mode control is also called current programmed control and current-
injected control. Descriptions of current-mode control can be found in e.g.
Kislovski, Redl and Sokal (1991, Chapter 5), Erickson and Maksimovic
(2000, Chapter 12), and Mitchell (1988, Chapter 6).
A large number of continuous-time models for current-mode control
have been presented during the years. Some of these models are intended to
be accurate also at high frequencies. The models presented by Ridley (1991),
Tan and Middlebrook (1995), and Tymerski and Li (1993) are designed to
be accurate from dc to half the switching frequency. The main difference
between the Ridley and Tan models is the modeling of the current loop gain.
Tymerski and Li (1993) presents a state-space model while the Ridley and
Tan models uses the PWM switch model (Vorperian, 1990).
In Tymerski (1994), time-varying system theory is used to derive models
for the frequency function and they are claimed to be exact for all frequencies.
Only the control-to-output frequency function is derived (so far) in the case
where current-mode control is utilized. The model is more complicated than

77
78 Chapter 3. Current-Mode Control

the previously mentioned models. When the control-to-output frequency

function of a converter is used to design a controller, the frequency interval dc
to half the switching frequency is the most interesting. The previously
mentioned models may therefore be good enough when designing a
controller.
When an accurate model of current-mode control has been needed, the
one presented by Ridley (1991) often has been chosen. An example of this is
Lo and King (1999), where the choice was between the Ridley and Tan
models. In Lo and King (1999), the Tan model is considered suspect and
some other authors have also expressed this opinion.
The operation of current-mode control is explained in Section 3.2. An
accurate control-to-current transfer function is reviewed in Section 3.3. In
Section 3.4, the Ridley and Tan models are reviewed and compared. These
two models utilizes the accurate control-to-current transfer function reviewed
in Section 3.3. The Ridley and Tan models are used to obtain the control-to-
output transfer function, the output impedance, and the audio susceptibility
of the buck converter with current-mode control and these derivations are
also presented in Section 3.4. These transfer functions are compared with
results from simulations of a buck converter in Section 3.5. The results of the
comparison are also explained in this section. In Section 3.6, the Ridley
model is used to obtain the transfer functions for the boost converter with
current-mode control. These transfer function are also compared with results
from simulations of a boost converter in Section 3.6. The corresponding
work is made for the buck-boost converter in Section 3.7. A summary and
concluding remarks are presented in Section 3.8.

3.2 Operation of Current-Mode Control

The operation and implementation of current-mode control are discussed
in this section.
In current-mode control, two control loops are used (Redl and Sokal,
1986). See Figure 3.1. The inner loop is fast and controls the inductor
current, i L (t ) . The outer loop is slower and controls the output voltage,
v o (t ) . The inductor current is fed back via the current controller in the inner
loop while the output voltage is fed back via the voltage controller in the
outer loop. The voltage controller has the reference signal v ref (t ) . The
voltage controller tries to get v o (t ) equal to v ref (t ) by changing its control
Chapter 3. Current-Mode Control 79

vo(t)
vref(t) Voltage ic(t) Current δ (t)
Converter iL(t)
controller controller

Figure 3.1: Current-mode control.

vo(t)
vref(t) Voltage δ (t) Converter
controller iL(t)

Figure 3.2: Voltage-mode control.

signal, ic (t ) . This signal is subsequently used as the reference signal for the
current controller. The current controller aims at getting i L (t ) equal to ic (t )
(in a sense) by changing its control signal, δ (t ) , which is the input (control)
signal of the converter. Thus, current-mode control is an application of
cascade control (Goodwin, Graebe and Salgado, 2001, Section 10.7).
In the case of current-mode control, the control signal of the voltage
controller is analog and the control signal of the current controller is digital
(binary).
In the case of voltage-mode control, see Figure 3.2, the control signal of
the voltage controller is digital using δ (t ) . There is no current controller and
the inductor current does not need to be measured. A voltage controller is
shown in Figure 3.3. The first (left) part is usually a voltage-error amplifier
and its output signal, vc (t ) , is analog. The second (right) part of the
controller is a pulse width modulator (compare with the circuit shown in
Figure 2.5). The duty cycle, d (t ) , depends linearly on the control signal
vc (t ) . The gain of this linearity depends on the peak-to-peak value of the
saw-tooth signal. The gain is equal to 1 in Figure 2.5 and that is why the duty
cycle, d (t ) , is equal to the signal d. Voltage-mode control is also called duty
ratio control.
80 Chapter 3. Current-Mode Control

Sawtooth and
pulse generator
δ (t)
S Q

vref(t) vc(t) R

vo(t)
Figure 3.3: A voltage controller in voltage-mode control.

Sawtooth and
pulse generator
δ (t)
ie(t) S Q

ic(t) R

iL(t)
Figure 3.4: A current controller in current-mode control.

A typical current controller in current-mode control is implemented as

shown in Figure 3.4. The peak inductor current is controlled and the control
method is therefore called peak current-mode control. This is the most
common type of current-mode control and the word “peak” is often left out.
If Figure 3.4 and Figure 3.3 are compared, the current controller seems to
consist only of a modulator. Kislovski, Redl and Sokal (1991, Chapter 5) use
the name current modulator instead of current controller. Average current-
mode control is another type of current-mode control (Kislovski, Redl and
Chapter 3. Current-Mode Control 81

δ (t)

t
0 Tsd(t) Ts
iL(t)+ie(t), ic(t)

Figure 3.5: The waveforms of the signals in a current controller in (peak)

current-mode control.

Sokal, 1991, Chapter 5). The first part of the current controller in average
current-mode control is a current-error amplifier. It may in this case not be
suitable to call the current controller a current modulator since one may
consider it consist of more than a modulator (compare with the voltage
controller in voltage-mode control). The name current controller may
therefore be seen as more general and it is used in this thesis. The modulator
is seen as a (large or small) part of the current controller.
The operation of the current controller in (peak) current-mode control
shown in Figure 3.4 will now be explained. For a moment assume that the
saw-tooth signal, ie (t ) , is not present. The period of the signal from the pulse
generator is equal to Ts and the signal sets the SR-latch. Each time this
occurs, the transistor is turned on and the inductor current, i L (t ) , starts to
increase as shown in Figure 3.5. When i L (t ) becomes greater than the signal
ic (t ) , the SR-latch is reset and i L (t ) then decreases until a new set pulse is
generated. This is the same function as the pulse width modulator in Figure
2.5 and Figure 3.3 except the inductor current, i L (t ) , replaces the saw-tooth
signal. Compare the waveforms shown in Figure 3.5 and Figure 2.7. The
signal ic (t ) is the reference signal of the current controller. The current
controller tries to get i L (t ) equal to ic (t ) in the sense that it is the peak value
of i L (t ) that is of interest. In average current-mode control, it is the average
value of i L (t ) that is of interest. The current controller in (peak) current-
mode control is fast since it manages to get the peak value of i L (t ) equal to
82 Chapter 3. Current-Mode Control

ic (t ) directly. The inner closed loop system in (peak) current-mode control

can therefore be seen as a current source.
To be compatible with the definitions made by Ridley (1991), the saw-
tooth signal, ie (t ) , is from now on called the external ramp.
The feedback of i L (t ) can cause instability (Erickson and Maksimovic,
2000, Section 12.1). The control of the inductor current is unstable if the
steady-state duty cycle, D , is greater than 0.5. It is unstable in the sense that
the duty cycle, d (t ) , never reaches a constant level even if ic (t ) is constant.
However, it is stable in the sense that the peak value of i L (t ) is equal to
ic (t ) .
It is possible to obtain stability also in the case where D is greater than
0.5 if slope compensation is utilized. With slope compensation, ic (t ) is
compared with the sum of i L (t ) and an external ramp, ie (t ) . The slope of
the sum is greater than the slope of i L (t ) alone. The characteristic value α is
now defined as

M2 − Me
α= , (3.1)
M1 + M e

where M e is the slope of ie (t ) , M 1 is the slope of i L (t ) while the transistor

is on and − M 2 is the slope of i L (t ) while the transistor is off. None of M e ,
M 1 , and M 2 is negative with these definitions. M e must be chosen such
that α < 1 to obtain stability.
The inductor current is controlled in current-mode control and it should
be fed back. However, it is not always necessary to measure the inductor
current directly. It may for instance be possible to measure and feed back the
transistor current instead, as it has the same waveform as the inductor current
during a certain part of the switching period.
This section is ended by considering some properties that can be obtained
by using measured inductor current.

• In peak current-mode control, current limiting is obtained automatically

since the peak current is controlled.
• It is easy to connect several converters in parallel.
• The voltage controller is easier to design since the process to be controlled
is approximately of order 1 instead of order 2 (compare Figure 3.11 and
Figure 2.8).
Chapter 3. Current-Mode Control 83

• The audio susceptibility can become low for a buck converter by

choosing proper slope compensation (compare Figure 4.14 and Figure
2.10).
• A disadvantage is that the output impedance is high (compare Figure
3.12 and Figure 2.9).

3.3 An Accurate Control-to-Current Transfer

Function
The high-frequency extensions in the Ridley and Tan models are based
on the accurate control-to-current transfer function that is reviewed in this
section.
In Figure 3.5, ie (t ) is added to i L (t ) but the same function is obtained
if ie (t ) is subtracted from ic (t ) and this is used in Figure 3.6(a). Figure
3.6(a) shows the waveforms of the signals ic (t ) , ic (t ) − ie (t ) , and two
different versions of the inductor current. The first version (solid line) shows
the inductor current waveform in steady state, i.e. in the case where there are
no perturbations of ic (t ) . The second version (dashed line) shows the
inductor current waveform in the case where there is a step perturbation in
ic (t ) as shown in Figure 3.6(a). The transistor is assumed to turn on at the
points t = nTs , where n is an integer. The transistor will then turn off at the
points t = (n + D )Ts in steady state. It is assumed that the changes in the
input voltage and the output voltage are negligible so that the slopes of the
inductor current can be considered constant.
In Figure 3.6(a), ic (t ) was perturbed, and this perturbation, iˆc (t ) , is
shown in Figure 3.6(b). This perturbation results in a perturbation in the
inductor current that is equal to the difference between the two versions of
the inductor current in Figure 3.6(a). The perturbation in the inductor
current, iˆL (t ) , is shown in Figure 3.6(c).
An approximation of iˆL (t ) is shown in Figure 3.6(d). This
approximation causes an error but the integral of this error is small compared
to the integral of iˆL (t ) if the amplitude of iˆL (t ) is small. The reason for this
is that the time needed to change from one level to the next is short compared
to Ts in this case. If we search for a linearized model, the use of the waveform
in Figure 3.6(d) instead of the waveform in Figure 3.6(c) does not cause an
approximation since:
84 Chapter 3. Current-Mode Control

ic(t)
ic(t)-ie(t)

iL(t)
t
Ts(k-1+D) Ts(k+1) Ts(k+1+D)
(a)
^i (t)
c

t
Ts(k-1+D) Ts(k+D) Ts(k+1+D)
(b)
^i (t)
L

t
Ts(k-1+D) Ts(k+D) Ts(k+1+D)
(c)
i^ (t) (app.)
L

t
Ts(k-1+D) Ts(k+D) Ts(k+1+D)
(d)
^i (n)
c

n
k-1 k k+1
(e)
^i (n)
L

n
k-1 k k+1
(f)

Figure 3.6: Different versions of the currents in a current controller.

Chapter 3. Current-Mode Control 85

∆t

-Me
^i (k+1)
c

M1

^i (k) -M2
L

M1 -M2 ^i (k+1)
L

t
Ts(k+1+D)
Figure 3.7: An enlargement of a part of Figure 3.6(a).

lim
∫ error (t ) dt = 0 . (3.2)
max( iˆ (t ) )→0
L
∫ iˆL (t ) dt
The signal ic (t ) affects the waveform of the inductor current. If the
changes in ic (t ) are small, the value of ic (t ) is important only in a small
surrounding of the points t = (n + D )Ts If the changes of ic (t ) in these
surroundings are slow, only a sampled version of ic (t ) (or iˆc (t ) ) is needed to
obtain an accurate model. A sampled version of iˆc (t ) is shown in Figure
3.6(e) and it is denoted iˆc (n) . To be spared from introducing new variable
names, the sampled (discrete-time) version of a continuous-time signal is
denoted by the same name as the continuous-time signal, which is not a
formally correct notation.
A sampled version of the approximate iˆL (t ) is shown in Figure 3.6(f) and
it is denoted iˆL (n) . An expression for iˆL (n) will now be derived. A small
part of Figure 3.6(a) is enlarged and it is shown in Figure 3.7. The following
two equations are obtained from Figure 3.7:

iˆL (k ) − iˆc (k + 1) = M 1∆t + M e ∆t , (3.3)

iˆL (k ) − iˆL (k + 1) = M 1∆t + M 2 ∆t . (3.4)

86 Chapter 3. Current-Mode Control

We first solve from (3.3):

∆t =
1
M1 + M e
(
iˆL (k ) − iˆc (k + 1) t ,) (3.5)

and insert the result into (3.4):

iˆL (k + 1) = iˆL (k ) − (M 1 + M 2 )
1
M1 + M e
(
iˆL (k ) − iˆc (k + 1) =)
 M1 + M e M1 + M 2 ˆ
  i L (k ) +
M +M − M +M  (3.6)
 1 e 1 e 
 M + M e M1 + M 2 ˆ
1 − 1 +  ic (k + 1) = −αiˆL ( k ) + (1 + α )iˆc (k + 1) .
 M1 + M e M1 + M e 
 

The variable k is substituted by n in (3.6):

iˆL (n + 1) = −αiˆL (n) + (1 + α )iˆc (n + 1) . (3.7)

where α is defined in (3.1). The difference equation (3.7) is stable if α < 1 ,

which is in agreement with the condition presented in Section 3.2.
Remember that (3.7) is valid only for small iˆc (t ) and iˆL (n) , so (3.7) can not
be used to claim that iˆL (n) approaches infinity if α > 1 . The dc gain of
(3.7) is equal to 1, which is evident by setting iˆL ( n + 1) equal to iˆL (n) . This
result should be used with some care. The reason is that a constant iˆc (t ) ≠ 0
may affect the output voltage such that its change is not negligible, which is
an assumption in the derivation of (3.7).
The following discrete-time transfer function is obtained from the Z-
transform of (3.7):

iˆL ( z ) (1 + α ) z
H ( z) = = (3.8)
iˆc ( z ) z +α

iˆL ( z ) and iˆc ( z ) denote the Z-transform of iˆL (n) and iˆc (n) . To be
spared from introducing new variable names, the Z-transform of a signal is
denoted by the same name as the signal, e.g. Z {i (n)} = i ( z ) .
Chapter 3. Current-Mode Control 87

The approximate perturbed inductor current shown in Figure 3.6(d) is

reconstructed from iˆL (n) by using a zero-order-hold circuit. The
continuous-time control-to-current transfer function is therefore (Åström and
Wittenmark, 1997, Section 7.7):

Fh ( s ) = =
(
iˆL ( s) 1 (1 + α ) e sTs 1 − e − sTs ) (3.9)
iˆc ( s ) Ts e sTs + α s

The first fraction in (3.9) represents the sampling of iˆc (t ) . The second
fraction is H (z ) transformed into continuous-time domain by substituting
z with e sTs . The third fraction is the transfer function of a zero-order-hold
circuit. Note that there are other frequency components in iˆL (t ) than (3.9)
predicts (Perreault and Verghese, 1997). If iˆc (t ) is a sinusoidal, the model
only predicts the Fourier component that is of the same frequency as iˆc (t ) .
The same conclusion was made about the models derived in Chapter 2 (see
Section 2.4).
Once again, note that the derivation of (3.9) is made with the assumption
that the changes in the input and output voltage are negligible. The high-
frequency extensions in the Ridley and Tan models are based on the accurate
control-to-current transfer function (3.9).

3.4 The Ridley and Tan Models Applied to the Buck

Converter
In this section, the Ridley and Tan models are reviewed and compared.
One block diagram is here used for both of these models. However, the
contents in the blocks are (of course) not the same for the two models.
General expressions for the control-to-output transfer function, the output
impedance, and the audio susceptibility are derived from the block diagram.
These general expressions are then applied to the buck converter. The
obtained expressions are finally used to derive expressions according to the
Ridley and Tan models.

A Brief Review

Both the Ridley and Tan models are unified models, i.e. they can be
applied to different types of converter topologies. The block diagram in
88 Chapter 3. Current-Mode Control

^vg(s)
v^ o(s)
Converter i^L(s)
^
d(s)
kf Fm(s) kr Ri

He(s)

v^ c(s)
Ri

i^c(s)

Figure 3.8: A small-signal model of the current controller and the converter.

Figure 3.8 is used to compare the Ridley and Tan models. Both are small-
signal models and, therefore, the linearized model of the converter is included
in Figure 3.8.
The model of the current controller consists of six blocks. The Ri blocks
will be explained later in this subsection. Fm ( s ) is the transfer function of the
modulator. Changes in the input and output voltages affect the control and
this effect are modeled with the feedforward gains k f and kr . Note that the
input and output voltages are not fed forward in Figure 3.1. The reason why
they are needed in Figure 3.8 is that there are Fourier components missing in
{ }
the signal L−1 iˆL ( s) compared the signal i L (t ) in Figure 3.1 (see Section
2.4). It is not just the dc component that is missing. The input and output
voltages affect the slopes of the inductor current in each switching period,
which is an important factor in the current controller. The use of vˆ g ( s ) and
vˆo ( s ) in the small-signal model of the current controller therefore
complements iˆL ( s ) so that the waveform of i L (t ) is better known. The
feedforward gains are in the Ridley and Tan models calculated in a way that
makes the amplification of the closed loop system correct at dc. In Section
3.5, it will be shown that this way introduces a modeling error at high
frequencies.
Chapter 3. Current-Mode Control 89

The high-frequency extensions in the Ridley and Tan models are based
on the accurate control-to-current transfer function (3.9). H e (s) is used to
include the high-frequency extension in the Ridley model and it is calculated
to be:

sTs
H e ( s) = . (3.10)
e sTs − 1

The Ridley model utilizes an approximation of (3.10). The approximation:

s s2
1+ + 2
ω n Qz ω n
e − sTs ≈ , (3.11)
s s2
1− + 2
ω n Qz ω n

where

−2
Qz = , (3.12)
π

π
ωn = . (3.13)
Ts

is used both by Ridley and Tan to replace the exponential functions. The
approximation error of (3.11) is zero at dc and half the switching frequency,
ωn :

s s2
1+ + 2
ω n Qz ω n
= 1 = e − sTs ,
2
s s s = j0
1− + 2
ω n Qz ω n
s = j0
90 Chapter 3. Current-Mode Control

s s2 j
1+ + 2 1+ −1
ω n Qz ω n Qz
= = − 1 = e − sTs .
s s2 1−
j
−1 s = jω n
1− + 2
ω n Qz ω n Qz
s = jω n

By combining (3.10) and (3.11), the approximate H e (s) is:

sTs
H e (s) = =
s s2 s s2
1− + 2 1+ + 2
ω nQz ω n ω nQz ω n
−
s s2 s s2
1+ + 2 1+ + 2
ω n Qz ω n ω nQz ω n (3.14)
 s2 
sTs 1 + 
s
+ 2 
 ω n Qz ω n  s s2
=1+ + 2 ,
s ω n Qz ω n
−2
ω nQz

By combining (3.9) and (3.11), an approximate expression for Fh (s ) is

obtained:
Chapter 3. Current-Mode Control 91

s s2
1+ + 2
ω n Qz ω n
1−
s s2
1− + 2
iˆL ( s ) 1 1+α ω n Qz ω n
Fh ( s ) = = =
iˆc ( s ) Ts s s2 s
1+ + 2
ω n Qz ω n
1+α
s s2
1− + 2
ω n Qz ω n (3.15)
s
−2
1+α ω n Qz
=
sTs s s2  s s 2 
1− + + α 1 + + 
ω n Q z ω n2  ω nQz ω 2 
 n 
1+α 1
= .
s s 2
s 1−α s2
1+α − (1 − α ) + 2 (1 + α ) 1 − +
ω n Qz ωn ω n Q z 1 + α ω n2

To rewrite (3.15) further, the following results are needed: In steady state, the
inductor current is increasing with the slope M 1 during DTs and decreasing
with the slope − M 2 during D' Ts each switching period. The increase must
be equal to the decrease:

M 1 DTs = M 2 D ' Ts , (3.16)

D
M 2 = M1 . (3.17)
D'

By using (3.17), the following expression is rewritten:

92 Chapter 3. Current-Mode Control

M1 + Me M2 − Me
−
1 − α M1 + M e M 1 + M e M 1 − M 2 + 2M e
= = =
1 + α M1 + Me M2 − Me M1 + M 2
+
M1 + M e M1 + M e
D
M1 − M1 + 2M e (3.18)
D' M D '− M 1 D + 2 M e D '
= 1 =
D M 1 (D '+ D )
M1 + M1
D'
M M  M 
D '− D + e 2 D ' = D '−(1 − D ') + e 2 D' = 1 + e  2 D'−1 ,
M1 M1  M1 

(3.15) is now rewritten by using (3.18):

iˆL ( s) 1
Fh ( s ) = = =
iˆc ( s ) πs  M   2
1+  1 + e  2 D '−1 + s
2ω n  M 1   ω2
  n
(3.19)
1
,
s s2
1+ +
ω nQ ω n2

where

1
Q= , (3.20)
π (mc D'−0.5)

Me
mc = 1 + . (3.21)
M1

The high-frequency extension in the Tan model is obtained by including

a pole in Fm (s ) . Tan and Middlebrook (1995) present the following model
for the buck converter:
Chapter 3. Current-Mode Control 93

1
Fm ( s ) = ,
 (D'− D )V g   s  (3.22)
Me +  T s 1 + 
 2L   ω 
   p 

DD' Ts
kf =− , (3.23)
2L

kr = 0 , (3.24)

H e ( s) = 1 , (3.25)

Ri = 1 Ω , (3.26)

ωn
ωp = . (3.27)
Q

In a practical current-mode controller, the inductor current is measured

and transformed to a voltage signal. Voltage signals also represent the control
signal and the external ramp signal. Ridley models this by including a gain,
Ri , in the inductor current feedback loop. Tan does not model this and Ri is
therefore set to 1 in (3.26). The following variables are used in the Ridley
model:

vˆc ( s ) = Ri iˆc ( s) , (3.28)

S n = Ri M 1 , (3.29)

S f = Ri M 2 , (3.30)

Se = Ri M e . (3.31)

Ridley (1990a) presents the following model for the buck converter:

1 1
Fm ( s) = Fm = = , (3.32)
(S n + Se )Ts mc S nTs
94 Chapter 3. Current-Mode Control

DTs Ri  D
kf = − 1 −  , (3.33)
L  2

Ts Ri
kr = , (3.34)
2L

s s2
H e ( s) = 1 + + 2. (3.35)
ω n Qz ω n

General Expressions for the Transfer Functions

In this subsection, general expressions for the control-to-output transfer

function, the output impedance, and the audio susceptibility are derived from
the block diagram presented in the previous subsection.
The duty cycle perturbation is obtained from Figure 3.8:

( )
dˆ ( s ) = Fm ( s ) k f vˆ g ( s ) + k r vˆo ( s ) + Ri iˆc ( s) − H e ( s ) Ri iˆL ( s ) . (3.36)

The model of the converter is linear and the outputs are therefore

vˆo ( s) ˆ vˆ ( s ) ˆ vˆ ( s)
vˆo ( s ) = d ( s) + o iinj ( s ) + o vˆ g ( s ) , (3.37)
ˆ
d (s) ˆ
iinj ( s ) vˆ g ( s)

iˆ ( s ) ˆ iˆ ( s ) ˆ iˆ ( s )
iˆL ( s ) = L d (s) + L iinj ( s ) + L vˆ g ( s ) . (3.38)
dˆ ( s ) iˆinj ( s ) vˆ g ( s )

Note that fractions in (3.37) and (3.38) must be regarded as transfer

functions and dˆ ( s ) , iˆinj ( s ) , and vˆ g ( s ) cannot be canceled. (3.36) is
rewritten by using (3.37) and (3.38):
Chapter 3. Current-Mode Control 95

dˆ ( s ) Fm−1 ( s ) = k f vˆ g ( s ) +
 vˆ ( s ) vˆ ( s) ˆ vˆ ( s) 
kr  o dˆ ( s ) + o iinj ( s ) + o vˆ g ( s )  +
 dˆ ( s) iˆinj ( s ) vˆ g ( s ) 
 
(3.39)
Ri iˆc ( s ) −
 iˆ ( s ) iˆL ( s) ˆ iˆ ( s) 
H e ( s ) Ri  L dˆ ( s ) + iinj ( s ) + L vˆ g ( s )  .
 dˆ ( s ) iˆinj ( s ) vˆ g ( s ) 
 

All the terms containing dˆ ( s ) in (3.39) are moved to the left:

 −1 vˆ ( s ) iˆ ( s ) 
 Fm ( s) − k r o + H e ( s ) Ri L  dˆ ( s ) =
 dˆ ( s) dˆ ( s ) 

 ˆ ˆ 
 k f + k r v o ( s ) − H e ( s ) Ri i L ( s )  vˆ g ( s) + (3.40)
 vˆ g ( s) vˆ g ( s ) 

 vˆ ( s ) iˆ ( s )  ˆ
k o − H e ( s ) Ri L i ( s ) + Ri iˆc ( s ) .
 iˆinj ( s )
r
ˆinj ( s )  inj
i
 

(3.40) is rewritten:

−1
 vˆ ( s ) iˆ ( s ) 
dˆ ( s ) =  Fm−1 ( s) − k r o + H e ( s ) Ri L  •
 dˆ ( s) dˆ ( s ) 

 ˆ 
  k + k vˆo ( s ) − H ( s ) R i L ( s )  vˆ ( s ) + (3.41)
 f r
vˆ g ( s )
e i
vˆ g ( s ) 
g

 vˆ ( s ) iˆL ( s )  ˆ 
k o − H e ( s ) Ri iinj ( s ) + Ri iˆc ( s )  .
 r iˆinj ( s ) iˆinj ( s)  
 

(3.37) is now modified using (3.41):

96 Chapter 3. Current-Mode Control

−1
vˆ ( s)  −1 vˆ ( s) iˆ ( s ) 
vˆo ( s ) = o  Fm ( s) − k r o + H e ( s ) Ri L  •
dˆ ( s )  dˆ ( s ) dˆ ( s ) 
 ˆ 
  k + k vˆo ( s) − H ( s) R i L ( s )  vˆ ( s ) +
 f r
vˆ g ( s )
e i
vˆ g ( s ) 
g
 (3.42)
 vˆ ( s ) iˆ ( s )  ˆ 
k o − H e ( s ) Ri L iinj ( s ) + Ri iˆc ( s )  +
 iˆinj ( s )
r
iˆinj ( s )  
 
vˆo ( s ) ˆ vˆ ( s )
i (s) + o vˆ g ( s ) .
ˆiinj ( s) inj vˆ g ( s )

The control-to-output transfer function of the closed loop system, which

includes the converter and the current controller, is obtained from (3.42):

−1
vˆo ( s) vˆo ( s )  −1 vˆ ( s) iˆL ( s ) 
=  Fm ( s ) − kr o + H ( s ) R  Ri =
dˆ ( s )  dˆ ( s ) 
e i
iˆc ( s ) dˆ ( s )
Ri (3.43)
−1 −1
.
 vˆ ( s )  iˆL ( s )  vˆo ( s ) 
Fm−1 ( s ) o  − kr + H e ( s ) Ri  
ˆ
 d (s)  dˆ ( s )  dˆ ( s) 

The subscript ol will be used for the converter transfer functions, i.e. for the
open loop system. When otherwise obvious the subscript will be excluded.
The output impedance of the closed loop system is obtained from (3.42):

vˆo ( s )
Z out ( s ) = − =
ˆiinj ( s)

 vˆ ( s )   ˆ 
kr  o  − H (s) R  i L (s) 
 iˆinj ( s)  e i
 iˆinj ( s )   vˆ ( s )  (3.44)
  ol   ol
− − o  .
 
−1 −1  iˆinj ( s ) 
vˆ ( s ) iˆ ( s )  vˆo ( s )    ol
Fm−1 ( s) o  − k r + H e ( s ) Ri L
 ˆ  dˆ ( s )  dˆ ( s ) 
 d ( s) 
Chapter 3. Current-Mode Control 97

The audio susceptibility of the closed loop system can be obtained from
(3.42):

vˆo ( s)
=
vˆ g ( s )
 vˆ ( s)  ˆ 
k f + kr  o  − H e ( s ) Ri  i L ( s )  (3.45)
 vˆ g ( s )   vˆ g ( s )   vˆ ( s ) 
  ol   ol
+ o  .
−1 −1  vˆ g ( s) 
 vˆ ( s ) iˆ ( s )  vˆo ( s )    ol
Fm−1 ( s ) o  − k r + H e ( s ) Ri L
 ˆ  dˆ ( s )  dˆ ( s ) 
 d (s) 

Transfer Functions for the Buck Converter

The general expressions for the control-to-output transfer function, the

output impedance, and the audio susceptibility derived in the previous
subsection are in this subsection applied to the buck converter.
For the buck converter, (3.43) is rewritten by using (2.83) and (2.90):

vˆo ( s )
=
iˆc ( s )
Ri
=
−1 den ( s ) 1 + s (R + Rc )C
Fm ( s ) ol
− k r + H e ( s ) Ri
RV g (1 + sRc C ) R(1 + sRc C )
(3.46)
R(1 + sRc C )
=
Fm−1 ( s ) denol ( s ) k r
− R (1 + sRc C ) + H e ( s )(1 + s(R + Rc )C )
Ri Vg Ri
R (1 + sRc C )
,
den( s)

where
98 Chapter 3. Current-Mode Control

denol ( s ) = R + s(L + RRc C ) + s 2 (R + Rc )LC =

(3.47)
R (1 + sRc C ) + sL(1 + s (R + Rc )C ) ,

den( s ) =
Ri−1 Fm−1 ( s )V g−1 denol − Ri−1k r R (1 + sRc C ) + H e ( s)(1 + s(R + Rc )C ) . (3.48)

denol (s ) is the denominator in the open loop transfer functions and den(s )
is the denominator in the closed loop transfer functions. For the buck
converter, (3.44) can be rewritten by using (2.83), (2.85) (2.90), and (2.91):

vˆo ( s )
Z out ( s ) = − =
iˆinj ( s)
 sRL(1 + sRc C ) R(1 + sRc C ) 
Ri−1 R (1 + sRc C ) k r + H e ( s ) Ri 
 denol ( s ) denol ( s ) 
+
den( s )
(3.49)
sRL(1 + sRc C )
=
denol ( s )
R (1 + sRc C ) 1
•
den( s) denol ( s )
(R −1
(
i k r sRL 1 + )
sRc C ) + H e ( s ) R(1 + sRc C ) + sLden( s ) .

For the buck converter, (3.45) is modified by using (2.83), (2.87), (2.90), and
(2.92):
Chapter 3. Current-Mode Control 99

vˆo ( s)
= Ri−1 R (1 + sRc C ) •
ˆv g ( s )

 k f + k r RD(1 + sRc C ) − H e ( s) Ri D(1 + s (R + Rc )C ) 

 
 
 denol ( s ) denol ( s ) 
+
den( s )
(3.50)
RD (1 + sRc C )
=
denol ( s )
R (1 + sRc C ) 1
den( s) denol ( s )
(
Ri−1k f denol ( s) +

Ri−1k r RD(1 + sRc C ) − H e ( s) D (1 + s(R + Rc )C ) + Dden( s) . )

Transfer Functions Obtained by Applying the Ridley Model
to the Buck Converter

In this subsection, the transfer functions for the buck converter with
current-mode control are derived according to the Ridley model.
For the buck converter, the slope of the inductor current while the
transistor is on is

V g − Vo
M1 = . (3.51)
L

The first term in (3.48) is modified by using (3.32), (3.29), (3.51), (2.66),
(2.68), (2.39), and (3.47):
100 Chapter 3. Current-Mode Control

Ri−1 Fm−1 ( s )V g−1denol ( s ) = Ri−1mc S n TsV g−1denol ( s ) =

V g − Vo Ts
Ri−1m c Ri M 1TsV g−1denol ( s ) = mc denol ( s ) =
L Vg
 Vo  (3.52)
Ts m c 1 − den ( s ) = Ts mc (1 − D )denol ( s ) =
L  V g  ol L
 
Ts m c D '
R (1 + sRc C ) + Ts mc D' s (1 + s(R + Rc )C ) .
L

(3.48) is rewritten using (3.52) and (3.34):

Ts m c D '
den( s ) = R (1 + sRc C ) + Ts mc D ' s (1 + s(R + Rc )C ) −
L
Ts Ri
Ri−1 R(1 + sRc C ) + H e ( s )(1 + s (R + Rc )C ) = (3.53)
2L
RT
(1 + s(R + Rc )C )(H e ( s) + sTs mc D') + s (mc D'−0.5)(1 + sRc C ) .
L

The following is obtained from (3.19):

s s2 sπ (mc D '−0.5) s2
Fh−1 ( s ) = 1 + + =1 + + =
ω nQ ω n2 ωn ω n2
sπ sπ s2 s s2 sπ (3.54)
1+ + m c D'+ 2 = 1 + + 2 + mc D ' =
− 2ω n ω n ωn ω n Q z ω n π Ts
H e ( s ) + sTs m c D' ,

where H e (s ) is defined in (3.14) and (3.35). (3.53) is rewritten by using

(3.54):

RTs
den( s ) = (1 + s(R + Rc )C )Fh−1 ( s ) + (mc D'−0.5)(1 + sRc C ) . (3.55)
L

The large parenthesis in (3.49) is rewritten by using (3.34), (3.55),

(3.54), and (3.47):
Chapter 3. Current-Mode Control 101

Ri−1 k r sRL(1 + sRc C ) + H e ( s ) R(1 + sRc C ) + sLden( s ) =

 −1 Ts Ri 
 Ri sL + H e ( s)  R(1 + sRc C ) +
 2 L 

(mc D'−0.5)(1 + sRc C ) =
RTs
sL (1 + s(R + Rc )C )Fh−1 ( s ) +
 L  (3.56)
 Ts 
s + H e ( s ) + sTs (mc D '−0.5) R(1 + sRc C ) +
 2 
sL(1 + s (R + Rc )C )Fh−1 ( s ) =
Fh−1 ( s) R (1 + sRc C ) + sL(1 + s (R + Rc )C )Fh−1 ( s ) = Fh−1 ( s )denol ( s ) .

The large parenthesis in (3.50) is rewritten by using (3.34), (3.55),

(3.54), and (3.47):
102 Chapter 3. Current-Mode Control

Ri−1 k f denol ( s) + Ri−1k r RD (1 + sRc C ) −

H e ( s ) D(1 + s (R + Rc )C ) + Dden( s ) =
Ts Ri
Ri−1k f denol ( s) + Ri−1 RD(1 + sRc C ) −
2L
H e ( s ) D(1 + s (R + Rc )C ) +
RTs
D(1 + s (R + Rc )C )Fh−1 ( s ) + D (mc D'−0.5)(1 + sRc C ) =
L
T T 
Ri−1k f denol ( s) +  s + s (m c D'−0.5) DR (1 + sRc C ) +
 2 L L 
(− H e ( s) )
+ Fh−1 ( s ) D (1 + s(R + Rc )C ) = (3.57)
Ts
Ri−1k f denol ( s) + mc D ' DR (1 + sRc C ) +
L
sTs m c D' D (1 + s(R + Rc )C ) =
Ts
Ri−1k f denol ( s) + mc D ' D (R(1 + sRc C ) + sL(1 + s(R + Rc )C )) =
L
 −1 T 
 Ri k f + s m c D ' D  denol ( s ) =
 L 
Ts  L 
D mc D '+ k f  denol ( s ) .
L  DTs Ri 

(3.50) is rewritten by using (3.57):

RTs  L 
D mc D'+ k f  (1 + sRc C )
vˆo ( s) L  DTs Ri  (3.58)
= .
ˆv g ( s ) den( s )

The transfer functions obtained by applying the Ridley model to the buck
converter are now summarized. The denominators are the same and given by
(3.55). The control-to-output transfer function of the closed loop system is
given directly by (3.46). The output impedance of the closed loop system is
obtained by combining (3.49) and (3.56). The audio susceptibility of the
Chapter 3. Current-Mode Control 103

closed loop system is obtained by combining (3.58) and (3.33). The results
are:

vˆo ( s ) vˆ ( s ) R(1 + sRc C )

= o = , (3.59)
ˆic ( s ) vˆc ( s ) Ri den( s)

R(1 + sRc C )Fh−1 ( s)

Z out ( s ) = , (3.60)
den( s )

RTs   D 
D mc D '−1 −   (1 + sRc C )
vˆo ( s) L   2  (3.61)
= ,
vˆ g ( s ) den( s )

where

RTs
den( s ) = (1 + s (R + Rc )C )Fh−1 ( s ) + (mc D'−0.5)(1 + sRc C ) . (3.62)
L

and Fh (s) is defined in (3.19).

Transfer Functions Obtained by Applying the Tan Model

to the Buck Converter

In this subsection, the transfer functions for the buck converter with
current-mode control are derived according to the Tan model.
(3.48) is rewritten by using (3.22), (3.24)-(3.27), and (3.47):

 (D'− D )V g   s  −1
den( s ) =  M e +  Ts  1 +
   V g •
 2L   ωn Q  (3.63)
(R(1 + sRc C ) + sL(1 + s(R + Rc )C )) + (1 + s(R + Rc )C ) .
One part of (3.63) is first rewritten by using (3.21), (3.51), (2.66), (2.68),
and (2.39):
104 Chapter 3. Current-Mode Control

 (D'− D )V g  −1  (D'− D )V g 
Me +  TsV g =  M 1 (mc − 1) +  TsV g−1 =
 2L   2L 
   
 V g − Vo (D'− D )V g  −1

 (m c − 1 ) +  TsV g =

 L 2L 
(3.64)
1 V  
 1 − o  (m − 1) + D '− D  T = Ts  m D '− D '+ D'− D  =
 L  Vg  c
2L 
s
L 
c
2 
   
Ts  D'+ D  Ts
 m c D '−  = (mc D '−0.5) .
L  2  L

(3.63) can now be rewritten by using (3.64), (3.20), (3.13), and (3.19):
Chapter 3. Current-Mode Control 105

Ts  s 
den( s ) = (mc D'−0.5) 1 +  •
L  ω nπ (mc D'−0.5) 
(R(1 + sRc C ) + sL(1 + s(R + Rc )C )) + (1 + s(R + Rc )C ) =
 Ts sTs 
 (mc D'−0.5) + •
 L Lπω n 

(R(1 + sRc C ) + sL(1 + s(R + Rc )C )) + (1 + s(R + Rc )C ) =
Ts
(mc D'−0.5)R(1 + sRc C ) + Ts (mc D'−0.5)s(1 + s(R + Rc )C ) +
L
sTs s 2 Ts
R (1 + sRc C ) + (1 + s(R + Rc )C ) + (1 + s(R + Rc )C ) =
Lπω n πω n
 2
Ts 
(1 + s(R + Rc )C ) 1 + Ts (mc D'−0.5)s + s +
 πω n  (3.65)

RTs  s 
 mc D'−0.5 +  (1 + sRc C ) =
 πω n 
L  
 
 2 
(1 + s(R + Rc )C ) 1 + s
+ 2 +
s
π 1 ωn 
 
 Ts π (mc D '−0.5) 
RTs   
 mc D'−0.51 − s 2   (1 + sRc C ) =
  πω n 
L   
RTs   
(1 + s(R + Rc )C )Fh−1 (s) +  mc D '−0.51 − s 2   (1 + sRc C ) .
  πω n 
L   

The large parenthesis in (3.49) is modified by using (3.24), (3.25), (3.65),

(3.14), (3.54), and (3.47):
106 Chapter 3. Current-Mode Control

Ri−1 k r sRL(1 + sRc C ) + H e ( s ) R(1 + sRc C ) + sLden( s ) =

R (1 + sRc C ) + sL(1 + s (R + Rc )C )Fh−1 ( s ) +
RTs   
sL  mc D'−0.51 − s 2  (1 + sRc C ) =
  
πω n
L   
  
1 + sT  m D'−0.51 − s 2    R (1 + sR C ) +
 s  c  πω n   
c
  
(3.66)
sL(1 + s (R + Rc )C )Fh−1 ( s ) =
 2 
1 + s + s + sTs mc D'  R(1 + sRc C ) +
 − 2 Ts πω n Ts 
 
sL(1 + s (R + Rc )C )Fh−1 ( s ) =
(H e (s) + sTs mc D')R(1 + sRc C ) + sL(1 + s(R + Rc )C )Fh−1 (s) =
Fh−1 ( s) R (1 + sRc C ) + sL(1 + s (R + Rc )C )Fh−1 ( s ) = Fh−1 ( s )denol ( s ) .

The large parenthesis in (3.50) is altered by using (3.24)-(3.26), (3.65),

(3.19), (3.20), (3.13), and (3.47):
Chapter 3. Current-Mode Control 107

Ri−1 k f denol ( s) + Ri−1k r RD(1 + sRc C ) −

H e ( s ) D(1 + s (R + Rc )C ) + Dden( s ) =
k f denol ( s) − D(1 + s (R + Rc )C ) + D(1 + s(R + Rc )C )Fh−1 ( s) +
RTs   
D  mc D'−0.51 − s 2   (1 + sRc C ) =
  πω n 
L   
k f denol ( s) +
 2 
 − 1 + 1 + s π (mc D'−0.5) + s  D(1 + s(R + Rc )C ) +
 π Ts ω n2 
 
Ts   2 
D mc D'−0.51 − s   R (1 + sRc C ) =
 πω n 
L   
 s 
k f denol ( s) + sTs  mc D'−0.5 +  D(1 + s(R + Rc )C ) + (3.67)
πω n 
 
Ts   2 
D mc D'−0.51 − s   R (1 + sRc C ) =
L   πω n 

k f denol ( s) +
Ts   2 
D mc D '−0.51 − s   (R (1 + sRc C ) + sL(1 + s (R + Rc )C )) =
L  πω 
 n 
k f denol ( s) +
Ts   2 
D mc D '−0.51 − s   denol ( s ) =
 πω n 
L   
Ts  L 1 2 
D mc D '+ k f − 1 − s   denol ( s) .
 2 πω n 
L  DTs 

By inserting (3.67) into (3.50) we get:

108 Chapter 3. Current-Mode Control

RTs  L 1 2 
D mc D '+ k f − 1 − s   (1 + sRc C )
 2 πω n 
vˆo ( s) L  DTs  (3.68)
= .
vˆ g ( s ) den( s )

The transfer functions obtained by applying the Tan model to the buck
converter are now summarized. The denominators are the same and given by
(3.65). The control-to-output transfer function of the closed loop system is
given directly by (3.46). The output impedance of the closed loop system is
obtained by combining (3.49) and (3.66). The audio susceptibility of the
closed loop system is obtained by combining (3.68) and (3.23). The results
are:

vˆo ( s ) R(1 + sRcC )

= , (3.69)
iˆc ( s ) den( s )

R(1 + sRc C )Fh−1 ( s)

Z out ( s ) = , (3.70)
den( s )

RTs   D s 
 (1 + sRc C )
D m c D '−1 −  + 
vˆ o ( s) L   2  πω n  (3.71)
= ,
vˆ g ( s ) den( s )

where

den( s ) =
RTs    (3.72)
(1 + s(R + Rc )C )Fh−1 (s) +  mc D '−0.51 − s 2   (1 + sRc C ) .
  πω n 
L   

and Fh (s) is defined in (3.19).

The denominator in the Tan model, (3.72), is almost the same as the one
in the Ridley model, (3.62). The difference is often insignificant for
converters that are used in practice. The control-to-output transfer functions
predicted by the Ridley and Tan models are therefore approximately the same
since the numerators in (3.59) and (3.69) are exactly the same. The same is
Chapter 3. Current-Mode Control 109

true for the output impedances since the numerators in (3.60) and (3.70) are
exactly the same. However, the numerator in (3.61) and (3.71) are not the
same. The audio susceptibility predicted by the Tan model includes an extra
zero compared to the Ridley model.

3.5 A Comparison of the Two Models and the

Simulation Results
In this section, a simulation model of a buck converter with current-
mode control is presented. The transfer functions derived in Section 3.4 by
means of the Ridley and Tan models are compared with simulation results.
The results of the comparison are also explained.

Simulation Model

A simulation model of a buck converter with current-mode control is

presented in this subsection.
Figure 3.9 shows the simulation model. The inductor current iL is fed
back and added to the external compensation signal ie and the sum is
compared to reference signal ic. The signal ie is obtained by multiplying the
signal sawtooth with Ts M e . The slope of sawtooth is equal to 1 Ts so the
slope of ie is equal to M e . The reference signal, ic, is the sum of its dc value,
Ic, and its ac value ichat.
If the output of the relay block is connected directly to the reset input of
the SR-latch, the simulation program report an existence of an algebraic loop.
This was not the case in the simulation model shown in Figure 2.5 but in the
simulation model shown in Figure 3.9, the feedback of the inductor current
causes this algebraic loop.
To avoid the algebraic loop, the subsystem shown is Figure 3.10 is
inserted between the relay and the SR-latch. The subsystem breaks the
algebraic loop by inserting a delay in the loop. The delay is one simulation
step. To avoid a significant deterioration of the simulation result, a small step
size must be used. Since the input signal is binary, it is enough to use a small
step size only when the input signal changes. To force the simulator to take a
small step size at these occasions, the original delta signal is first created. The
output signal of the integrator is limited to the interval [-0.1 1.1]. The
absolute value of the input signal of the integrator is great and when the
110 Chapter 3. Current-Mode Control

magnitude
signal
Vg angle
vg

vghat
vg vo
vo
Iinj iinj iload
iinj iload
delta iL
iL
Buck
iinjhat S Q
delta converter

Ic in out R !Q
ic
Avoid
algebraic
ichat loop

Ts*Me
sawtooth ie

0.5

Figure 3.9: The simulation model for current-mode control (without the
voltage controller).

1 1
in out
Step delay

1
S Q 1e12
original delta s
Integrator
R !Q with limited
output

0.5

Figure 3.10: The subsystem that breaks the algebraic loop.

original delta signal changes, the sign of this input signal changes. This means
that the output signal of the integrator changes very fast from one limit value
to the other when a change in the original delta signal occurs. The sign of the
input signal to the relay block will therefore also change after a very short
time and this causes the simulator to take a short step. The solution presented
Chapter 3. Current-Mode Control 111

here is probably not the most optimal one, but it has proven to be good
enough for the purpose here.
The parameters used in the simulation model presented in Section 2.4 are
also used here. Ic is adjusted manually so that the average value of the output
voltage, Vo , is equal to 5 V ( D =0.455). M e is calculated by using (3.21)
and (3.51):

V g − Vo
Me = (mc − 1) , (3.73)
L

where mc is chosen to be 2.

Simulation Results

The transfer functions derived in Section 3.4 by means of the Ridley and
Tan models are compared with simulation results in this subsection.
Figure 3.11 shows the Bode plots for the control-to-output transfer
functions in (3.59) and (3.69) together with the simulation results. Ri is set
to 1 Ω. From the figure it is seen that the control-to-output transfer functions
predicted by the Ridley and Tan models are almost the same and they agree
closely with the simulation results.
Figure 3.12 shows the Bode plots for the output impedances in (3.60)
and (3.70) together with the simulation results. From the figure it is seen that
the output impedances predicted by the Ridley and Tan models are almost
the same and they agree closely with the simulation results.
Figure 3.13 shows the Bode plots for the audio susceptibilities in (3.61)
and (3.71) together with the simulation results. From the figure it is seen that
the audio susceptibilities predicted by the Ridley and Tan models are not the
same and neither agrees closely with the simulation results at high
frequencies. The Tan model has the largest deviation from the simulation
results in the magnitude while the Ridley model has the largest deviation
from the simulation results in the phase shift.

Explanation of the Comparison Results

In the previous subsection we compared the transfer functions obtained

by utilizing the Ridley and Tan models with simulation results. The results of
this comparison are explained in this subsection.
112 Chapter 3. Current-Mode Control

-10

-20
Phase (deg); Magnitude (dB)

-30

-40
1 2 3 4
10 10 10 10

-50

-100

-150
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 3.11: The control-to-output transfer function of a buck converter with

a current controller. X: the simulation results. Solid line: the
Ridley model. Dashed line: the Tan model. Note that the two
lines almost coincide.

The control-to-current transfer function, which was reviewed in Section

3.3, is derived under the assumption that there are no changes in the input
and output voltages. To cope with changes in these voltages, the feedforward
gains k f and kr are included in the Ridley and Tan models. The
feedforward gains are calculated in a way that makes the amplification of the
closed loop system correct at dc. At high frequencies, the perturbation of the
voltage across the inductor cannot be considered constant during a switching
period. However, the amplitude of the changes in the output voltage is low at
high frequencies due to the output capacitor. This explains why the control-
to-output transfer functions and the output impedances predicted by the
Ridley and Tan models are so accurate. In the case where the audio
susceptibility is considered, also the input voltage changes. Since the input
voltage is the input signal in this case, its amplitude is assumed to be unity.
The perturbation of the voltage across the inductor is therefore not small at
high frequencies and the errors in the Ridley and Tan models are significant.
Chapter 3. Current-Mode Control 113

-10

-20
Phase (deg); Magnitude (dB)

-30

-40
1 2 3 4
10 10 10 10

-20

-40

-60

-80
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 3.12: The output impedance of a buck converter with a current

controller. X: the simulation results. Solid line: the Ridley
model. Dashed line: the Tan model. Note that the two lines
almost coincide.

A more thorough explanation of the comparison results will now be

presented. The perturbation in the inductor current, iˆL (t ) , is affected by
three signals. The control-to-current transfer function shows how iˆL (t )
depends on the signal iˆc (t ) . The inductor current also depends on the input
and output voltages and since these dependencies are modeled at dc, it is
expected that there is a modeling error at higher frequencies. This modeling
error is due to the fact that the voltage across the inductor cannot be
considered constant during a switching period if the period of the voltage is in
the same magnitude as the switching period. To obtain a simple model of the
modeling error, it is reasonable to assume that the modeling error is
proportional to the frequency and the magnitude of the input and output
voltages:

( )
error in iˆL ~ ω vˆo , (3.74)
114 Chapter 3. Current-Mode Control

-20

-30

-40

-50
Phase (deg); Magnitude (dB)

-60

-70
1 2 3 4
10 10 10 10

-50

-100

-150
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 3.13: The audio susceptibility of a buck converter with a current

controller. X: the simulation results. Solid line: the Ridley
model. Dashed line: the Tan model.

( )
error in iˆL ~ ω vˆ g . (3.75)

The control-to-output transfer function is first considered and in this case

vˆ g (t ) is zero. To simplify the discussion, the ESR of the capacitor is assumed
to be negligible. At high frequencies, the impedance of the capacitor and the
load resistor (connected in parallel) is dominated by the impedance of the
capacitor and

vˆo 1
~ . (3.76)
iˆL ω

The following result is obtained if (3.74) and (3.76) are combined:

Chapter 3. Current-Mode Control 115

( ) 1
error in iˆL ~ ω vˆo ~ ω iˆL ~ iˆL .
ω
(3.77)

From (3.77), it is seen that the relative error in iˆL (t ) is independent of

the frequency. The relative error in vˆo (t ) is proportional to the relative error
in iˆL (t ) . The conclusion is that there is no significant increase in the relative
error in vˆo (t ) when the frequency of iˆc (t ) is increased.
The output impedances are now considered and in this case vˆ g (t ) and
ˆic (t ) are zero. The perturbation of the total current, iˆtotal (t ) , to the output
stage is the sum of iˆL (t ) and − iˆinj (t ) . The corresponding result of (3.76) is:

vˆo 1
~ . (3.78)
iˆtotal ω

Assume that iˆinj (t ) has constant amplitude and that the frequency increases.
The changes in the output voltage then decrease according to (3.78). Since
vˆ g (t ) and iˆc (t ) are zero, the changes in the inductor current depends only
on the changes in the output voltage. Therefore, the changes in the inductor
current decrease and the fraction iˆL iˆinj also decreases. iˆtotal (t ) is more and
more dominated by iˆinj (t ) and the relative error in iˆtotal (t ) does not increase
even if the relative error in iˆL (t ) increases.
Finally, the audio susceptibility is considered and now iˆc (t ) is zero. From
Figure 3.13, it is seen that the audio susceptibility can be approximated by a
first order system. This means that the amplitude of iˆL (t ) is almost the same
for all shown frequencies if the amplitude of the input voltage is constant.
Assume that the changes in the input voltage have constant amplitude and
that the frequency increases. From (3.75), it is apparent that the error in
iˆL (t ) increases. Therefore, the relative error in iˆL (t ) also increases.
Consequently, the relative error in vˆo (t ) increases when the frequency of
vˆ g (t ) increases.

3.6 The Ridley Model Applied to the Boost

Converter
In this section, the Ridley model is used to obtain the control-to-output
transfer function, the output impedance, and the audio susceptibility for the
boost converter with current-mode control. These transfer function are also
compared with simulation results.
116 Chapter 3. Current-Mode Control

Transfer Functions

In Section 3.4 we derived general expressions for the control-to-output

transfer function, the output impedance, and the audio susceptibility. In this
subsection, these expressions are used to derive the transfer functions for the
boost converter with current-mode control according to the Ridley model.
The methodology is analogous to the buck converter modeling.
For the boost converter, (3.43) is modified by using (2.135) and (2.142):
Chapter 3. Current-Mode Control 117



vˆo ( s )  denol ( s )
= Ri  Fm−1 ( s ) −
iˆc ( s ) 

Vg
(RD'+ Rc )D'
2 2
(
R D ' − s(R + Rc )L (1 + sRc C ) )

−1
Vg   
1 + RD ' + s(R + Rc )C  
D'  RD'+ Rc  
k r + H e ( s ) Ri  =
Vg
(RD'+ Rc )D'
2 2
(
R D' − s(R + Rc )L (1 + sRc C ) 

)


 −1
(RD'+ Rc )D'V g−1denol ( s)
− kr +
( )
Ri Fm ( s) 2 2
 R D ' − s (R + Rc )L (1 + sRc C )

−1
 
(RD'+ Rc )1 + RD' + s(R + Rc )C  
 RD'+ Rc 
 =
( )
H e ( s ) Ri
R D ' − s (R + Rc )L (1 + sRc C ) 
2 2

 (3.79)

( )
Ri R 2 D ' 2 − s (R + Rc )L (1 + sRc C ) •
(Fm−1 ( s ) ( )
RD'+ Rc D 'V g−1denol ( s ) −
( )
k r R 2 D' 2 − s(R + Rc )L (1 + sRc C ) +
H e ( s ) Ri (RD '+ Rc + RD '+ s(R + Rc )C (RD '+ Rc )))−1 =
 (R + Rc )L   Fm−1 ( s ) RD'+ Rc denol ( s )
RD' 1 − s  (1 + sR c C ) • 
 Ri
−
 R 2 D' 2   R Vg
kr  (R + Rc )L 
RD' 1 − s  (1 + sRc C ) +
Ri  R 2 D' 2 
−1
 2 RD'+ Rc RD'+ Rc 
H e ( s ) + s (R + Rc )C   =
 RD' RD' 
 (R + Rc )L 
RD ' (1 + sRc C ) 1 − s
 R 2 D' 2 
,
den( s)
118 Chapter 3. Current-Mode Control

where

denol ( s ) =
RD' (RD'+ Rc ) (R + Rc ) + s (L + RRc CD ') + s 2 (R + Rc )LC = (3.80)
RD' ((RD '+ Rc ) (R + Rc ) + sRc C ) + sL(1 + s (R + Rc )C ) ,

den( s ) = Ri−1 Fm−1 ( s )V g−1denol ( s) (RD '+ Rc ) R −

( (
Ri−1k r RD' (1 + sRc C ) 1 − s(R + Rc )L R 2 D ' 2 + )) (3.81)
H e ( s )((2 RD'+ Rc ) (RD') + s(R + Rc )C (RD '+ Rc ) (RD ')) .

den ol (s ) is the denominator in the open loop transfer functions and den(s)
is the denominator in the closed loop transfer functions. For the boost
converter, (3.44) is rewritten by using (2.135), (2.137), (2.142), and (2.143):
Chapter 3. Current-Mode Control 119

 (R + Rc )L 
Ri−1 RD' (1 + sRc C ) 1 − s
vˆ ( s )  R 2 D' 2 
Z out ( s ) = − o = •
iˆinj ( s ) den( s )
  R 2 R DD'  
  c
+ sRL  (1 + sR C ) 
  R + Rc 

c
RD ' (1 + sRc C ) 
 kr + H e ( s ) Ri +
 denol ( s ) denol ( s ) 
 
 
 
 R R DD'
2 
 c
+ sRL  (1 + sR C )
 R+R  c
 c  (3.82)
=
denol ( s )
R (1 + sRc C ) 1
•
den( s) denol ( s )
 −1 
 Ri D ' 1 − s
(R + Rc )L   RRc DD' 
 R(1 + sRc C ) +
  k r  + sL 
  R 2 D ' 2   R + Rc 
 (R + Rc )L 
D ' 1 − s  H e ( s ) RD ' (1 + sRc C ) +
 R 2 D' 2 
 RRc DD '  
 + sL  den( s )  .

 R + Rc  

For the boost converter, (3.45) is rewritten by using (2.135), (2.139),

(2.142), and (2.144):
120 Chapter 3. Current-Mode Control

 (R + Rc )L 
Ri−1 RD ' (1 + sRc C ) 1 − s
vˆo ( s)  R 2 D' 2 
= •
vˆ g ( s ) den( s )
 RD ' (1 + sRc C ) 1 + s (R + Rc )C 
 k f + kr − H e ( s ) Ri +
 denol ( s ) 
 denol ( s)
RD ' (1 + sRc C )
=
denol ( s) (3.83)
R (1 + sRc C ) 1  −1  (R + Rc )L 
 Ri D ' 1 − s  k f denol ( s ) +
den( s) denol ( s )   R 2 D' 2 
 (R + Rc )L 
Ri−1 D ' 1 − s  k r RD ' (1 + sRc C ) −
 R 2 D' 2 
 (R + Rc )L  
D ' 1 − s  H e ( s )(1 + s (R + Rc )C ) + D' den( s)  .
 R D' 
2 2


Ridley (1990a) presents the following model for the boost converter:

Ts Ri
kf =− , (3.84)
2L

D ' 2 Ts Ri
kr = , (3.85)
2L

Fm (s) is defined in (3.32), and H e (s ) is defined in (3.35).

For the boost converter, the slope of the inductor current while the
transistor is on is

Vg
M1 = . (3.86)
L

The first term in (3.81) is rewritten by using (3.32), (3.29), (3.86), and
(3.80):
Chapter 3. Current-Mode Control 121

Ri−1 Fm−1 ( s )V g−1denol ( s ) (RD'+ Rc ) R =

Ri−1mc S nTsV g−1denol ( s ) (RD'+ Rc ) R =
Ri−1mc Ri M 1TsV g−1denol ( s) (RD '+ Rc ) R =
V g Ts RD '+ Rc (3.87)
mc denol ( s ) =
L Vg R
Ts mc RD'+ Rc   RD '+ Rc  
 R + R + sRc C  + sL(1 + s (R + Rc )C ) .
 RD '   

L R   c  

(3.81) is rewritten by using (3.87) and (3.85):

den( s ) =
Ts mc RD'+ Rc   RD '+ Rc  
 RD '  + sR C  + sL(1 + s(R + Rc )C ) −
  R+R c  
L R   c  
RD '3 Ts (R + R )L (3.88)
(1 + sRc C ) 1 − s 2 c2  +
2L  R D' 
 2 RD'+ Rc RD'+ Rc 
H e ( s ) + s(R + Rc )C .
 RD' RD ' 

The large parenthesis in (3.82) is rewritten by using (3.85), (3.88), (3.35)

and (3.80):
122 Chapter 3. Current-Mode Control

 (R + Rc )L   RRc DD' 
Ri−1 D ' 1 − s  kr  + sL  R (1 + sRc C ) +
 R D '   R + Rc
2 2

 (R + Rc )L 
D ' 1 − s  H e ( s ) RD' (1 + sRc C ) +
 R 2 D' 2 
 RRc DD ' 
 + sL  den( s ) = K =
 R + Rc  (3.89)
  T m DD' 2 R  R RD'+ R Rc2 D
 1 + s c c  c
+ +
 L  R + Rc RD ' RD ' (R + Rc )
 
 s s2  RD'+ Rc D RD'+ Rc 
 + 2  + sTs mc D '  den ( s ) .
 ω nQz ω  RD ' RD '  ol
 n  

The large parenthesis in (3.89) is approximately equal to Fh−1 ( s ) in (3.54)

for small Rc .
The large parenthesis in (3.83) is rewritten by using (3.85), (3.88), (3.35)
and (3.80):

 (R + Rc )L 
Ri−1 D ' 1 − s  k f denol ( s ) +
 R 2 D' 2 
 (R + Rc )L 
Ri−1 D ' 1 − s  k r RD ' (1 + sRc C ) −
 R 2 D' 2 
 (R + Rc )L 
D ' 1 − s  H e ( s )(1 + s (R + Rc )C ) + D ' den( s ) = K =
 R 2 D' 2  (3.90)
T  
 s D '  m RD '+ Rc + L k  + R + Rc
2
1 + s −
 L  c R T s Ri
f 
 R D'
2  ω2 
  n 
T  R  1 L 
s s 1 + c   + k f   denol ( s ) .
RD '  R   2 T s Ri 


(3.83) is rewritten by using (3.90):

Chapter 3. Current-Mode Control 123

vˆo ( s)  RTs  RD'+ Rc L  R + Rc 

1 + s
2 
−
= D '  mc + k f  +
ˆv g ( s )  L  ω2 
  R T s Ri  RD'  n 
(3.91)
T  R  1 L   (1 + sRc C )
s s 1 + c   + k f   .
D'  R   2 T s Ri 
  den( s )

The transfer functions obtained by applying the Ridley model to the

boost converter are now summarized. The denominators are the same and
given by (3.88). The control-to-output transfer function of the closed loop
system is given directly by (3.79). The output impedance of the closed loop
system is obtained by combining (3.82) and (3.89). The audio susceptibility
of the closed loop system is obtained by combining (3.91) and (3.84). The
results are:

 (R + Rc )L 
RD ' (1 + sRc C ) 1 − s 
vˆo ( s )
=
vˆo ( s )
=  R 2 D' 2 
,
(3.92)
iˆc ( s ) vˆc ( s ) Ri den( s )

R(1 + sRc C )
Z out ( s ) = •
den( s )
  T m DD ' 2 R  R RD '+ R Rc2 D
 1 + s c c  c
+ +
 L  R + Rc RD' RD' (R + Rc ) (3.93)
 
 s s2  RD'+ Rc D RD'+ Rc 
 + 2  + sTs m c D ' ,
 ω nQz ω  RD ' RD' 
 n  

vˆo ( s)
=
vˆ g ( s )
 RT   
 s D '  m RD'+ Rc − 0.5  + R + Rc
2 (3.94)
1 + s  (1 + sRc C )
 L 
c
R  RD'  ω2 
  n 
,
den( s )

where
124 Chapter 3. Current-Mode Control

den( s ) =
Ts mc RD'+ Rc   RD '+ Rc  
 RD ' 
 + sRc C  + sL(1 + s(R + Rc )C ) −
 
L R   R + Rc  
RD '3 Ts (R + R )L
(1 + sRc C ) 1 − s 2 c2  + (3.95)
2L  R D' 
 2 RD'+ Rc RD'+ Rc 
H e ( s ) + s(R + Rc )C ,
 RD' RD ' 

and H e (s ) is defined in (3.35).

Simulation Results

In this section, a simulation model of a boost converter with current-

mode control is presented. The transfer functions derived in the previous
subsection are compared with simulation results.
The simulation model shown in Figure 3.9 is used, except the buck
converter subsystem is replaced with the boost converter subsystem shown in
Figure 2.14. The parameters used in the simulation model presented in
Section 2.7 are also used here. Ic is adjusted manually so that the average
value of the output voltage, Vo , is equal to 8 V ( D =0.382). M e is calculated
by using (3.21) and (3.86):

Vg
Me = (mc − 1) , (3.96)
L

where mc is chosen to be 2.
Figure 3.14 shows the Bode plot for the control-to-output transfer
function in (3.92) together with the simulation results. Ri is set to 1 Ω. From
the figure it is seen that the control-to-output transfer function predicted by
the Ridley model agrees closely with the simulation results.
Figure 3.15 shows the Bode plot for the output impedance in (3.93)
together with the simulation results. From the figure it is seen that the output
impedance predicted by the Ridley model agrees closely with the simulation
results.
Chapter 3. Current-Mode Control 125

-10

-15
Phase (deg); Magnitude (dB)

-20

-25
1 2 3 4
10 10 10 10

-50

-100

-150

-200

-250
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 3.14: The control-to-output transfer function of a boost converter

with a current controller. X: the simulation results. Solid line:
the Ridley model.

Figure 3.16 shows the Bode plot for the audio susceptibility in (3.94)
together with the simulation results. From the figure it is seen that the audio
susceptibility predicted by the Ridley model does not agree closely with the
simulation results at high frequencies.
The conclusions about the agreement between the simulation results and
the transfer functions obtained from the Ridley model are thus the same as
for the buck converter.

3.7 The Ridley Model Applied to the Buck-Boost

Converter
In this section, the Ridley model is used to obtain the control-to-output
transfer function, the output impedance, and the audio susceptibility for the
buck-boost converter with current-mode control. These transfer function are
also compared with simulation results.
126 Chapter 3. Current-Mode Control

-10

-20
Phase (deg); Magnitude (dB)

-30

-40
1 2 3 4
10 10 10 10

-20

-40

-60

-80
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 3.15: The output impedance of a boost converter with a current

controller. X: the simulation results. Solid line: the Ridley
model.

Transfer Functions

In Section 3.4 we derived general expressions for the control-to-output

transfer function, the output impedance, and the audio susceptibility. In this
subsection, these expressions are used to derive the transfer functions for the
buck-boost converter with current-mode control according to the Ridley
model. The methodology is analogous to the modeling of the buck and boost
converters.
For the buck-boost converter, (3.43) is rewritten by using (2.179) and
(2.186):
Chapter 3. Current-Mode Control 127



vˆo ( s )  denol ( s )
= Ri  Fm−1 ( s ) −
V g (R + Rc )
iˆc ( s ) 
 (RD'+ Rc )D'
2
(
RD ' − sLD (1 + sRc C ) )

−1
Vg   
1 + RDD ' + s(R + Rc )C  
D'  RD'+ Rc  
k r + H e ( s ) Ri  =
V g (R + Rc )
(RD'+ Rc )D'
2
( )
RD ' − sLD (1 + sRc C ) 



 −1
(RD'+ Rc )D'V g−1denol (s)
− kr +
( )
Ri Fm ( s )
 ( R + R ) RD ' 2
− sLD (1 + sR C )
 c c
−1
 
(RD'+ Rc )1 + RDD' + s(R + Rc )C  
 RD'+ Rc 
 =
( )
H e ( s ) Ri
(R + Rc ) RD' −sLD (1 + sRc C ) 
2

 (3.97)

(
Ri (R + Rc ) RD' 2 − sLD (1 + sRc C ) •)
(Fm−1 ( s ) ( )
RD'+ Rc D 'V g−1 denol ( s ) −
( )
k r (R + Rc ) RD' 2 − sLD (1 + sRc C ) +
H e ( s ) Ri (RD '+ Rc + RDD '+ s (R + Rc )C (RD '+ Rc )))−1 =
 LD   Fm−1 ( s ) RD '+ Rc denol ( s )
RD' 1 − s  (1 + sR c C ) 
 Ri
−
 RD' 2   R + Rc Vg
kr  LD 
RD' 1 − s  (1 + sRc C ) +
Ri  RD' 2 
−1
 RD '+ RDD'+ Rc RD'+ Rc 
H e ( s ) + sC   =

 (R + Rc )D' D' 
 LD 
RD ' (1 + sRc C ) 1 − s 
 RD ' 2 
,
den( s )
128 Chapter 3. Current-Mode Control

-20
Phase (deg); Magnitude (dB)

-40

-60
1 2 3 4
10 10 10 10

-50

-100

-150
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 3.16: The audio susceptibility of a boost converter with a current

controller. X: the simulation results. Solid line: the Ridley
model.

where

denol ( s) =
RD' (RD '+ Rc ) (R + Rc ) + s(L + RRc CD ') + s 2 (R + Rc )LC = (3.98)
RD' ((RD'+ Rc ) (R + Rc ) + sRc C ) + sL(1 + s (R + Rc )C ) ,

den( s ) = Ri−1 Fm−1 ( s )V g−1denol ( s ) (RD'+ Rc ) (R + Rc ) −

( (
Ri−1k r RD' (1 + sRc C ) 1 − sLD RD' 2 + )) (3.99)
H e ( s )((RD ' (1 + D ) + Rc ) ((R + Rc )D') + sC (RD'+ Rc ) D ') .

den ol (s ) is the denominator in the open loop transfer functions and den(s)
is the denominator in the closed loop transfer functions. For the buck-boost
converter, (3.44) is rewritten by using (2.179), (2.181), (2.186), and (2.187):
Chapter 3. Current-Mode Control 129

 LD 
Ri−1 RD' (1 + sRc C ) 1 − s 
vˆ ( s )  RD' 2 
Z out ( s ) = − o = •
iˆinj ( s ) den( s )
  R 2 R DD'  
  c
+ sRL  (1 + sR C ) 
  R + Rc 

c
RD ' 1 + sRc C 
( )
 kr + H e ( s ) Ri +
 denol ( s ) denol ( s ) 
 
 
 
 R 2 R DD' 
 
 R + R + sRL  (1 + sRc C )
c

 c  (3.100)
=
denol ( s)
R (1 + sRc C ) 1
•
den( s) denol ( s )
 −1   RR DD ' 
 Ri D ' 1 − s LD  k r  c
 R + R + sL  R(1 + sRc C ) +

  RD '  
2
 c 
 LD 
D ' 1 − s  H e ( s) RD' (1 + sRc C ) +
 RD' 2 
 RRc DD '  
  
 R + R + sL  den( s )  .
 c  

For the buck-boost converter, (3.45) is rewritten by using (2.179), (2.183),

(2.186), and (2.188):
130 Chapter 3. Current-Mode Control

 LD 
Ri−1 RD ' (1 + sRc C ) 1 − s 
vˆo ( s)  RD ' 2 
= •
vˆ g ( s) den( s)
 RDD' (1 + sRc C ) D (1 + s(R + Rc )C ) 
 k f + k r − H e ( s ) Ri  +
 denol ( s ) denol ( s ) 
RDD ' (1 + sRc C )
=
denol ( s) (3.101)
R (1 + sRc C ) 1  −1  LD 
 Ri D' 1 − s  k f denol ( s ) +
den( s) denol ( s )   RD' 2 
 LD 
Ri−1 D ' 1 − s  k r RDD' (1 + sRc C ) −
 RD ' 2 
 LD  
D ' 1 − s  H e ( s ) D (1 + s ( R + R c )C ) + DD ' den ( s )  .
 RD ' 2  

Ridley (1990a) presents the following model for the buck-boost

converter:

DTs Ri  D
kf =− 1 −  , (3.102)
L  2

D ' 2 Ts Ri
kr = , (3.103)
2L

Fm (s) is defined in (3.32), and H e (s ) is defined in (3.35).

For the buck-boost converter, the slope of the inductor current while the
transistor is on is

Vg
M1 = . (3.104)
L

The first term in (3.99) is rewritten by using (3.32), (3.29), (3.104), and
(3.98):
Chapter 3. Current-Mode Control 131

Ri−1 Fm−1 ( s )V g−1denol ( s ) (RD '+ Rc ) (R + Rc ) =

Ri−1mc S nTsV g−1denol ( s ) (RD '+ Rc ) (R + Rc ) =
Ri−1mc Ri M 1TsV g−1denol ( s ) (RD '+ Rc ) (R + Rc ) =
V g Ts RD '+ Rc (3.105)
mc denol ( s ) =
L Vg R + Rc
Ts mc RD'+ Rc   RD '+ Rc  
 RD ' 
 + sRc C  + sL(1 + s(R + Rc )C ) .
L R + Rc   
 R + Rc  

(3.99) is rewritten by using (3.105) and (3.103):

den( s ) =
Ts mc RD'+ Rc   RD '+ Rc  
 RD ' 
 R+R + sR C 
 + sL (1 + s ( R + R )C )−
L R + Rc  
c c
 c  
RD '3 Ts
(1 + sRc C ) 1 − s LD2  +
(3.106)
2L  RD' 
 RD ' (1 + D ) + Rc RD'+ Rc 
H e ( s ) + sC .

 (R + Rc )D' D' 

The large parenthesis in (3.100) is rewritten by using (3.103), (3.106),

(3.35) and (3.98):
132 Chapter 3. Current-Mode Control

 LD   RRc DD ' 
Ri−1 D ' 1 − s  k r  + sL  R (1 + sRc C ) +
 RD '   R + Rc
2

 LD 
D ' 1 − s  H e ( s ) RD ' (1 + sRc C ) +
 RD ' 2 
 RRc DD ' 
 + sL  den( s ) = K = (3.107)
 R + Rc 
 R(R + Rc D ') − Rc2 D Ts mc DD ' RRc (RD '+ Rc ) Rc
 + + +

 (R + Rc ) 2
L (R + Rc ) 2 (R + Rc )D'
 s
 s 2  RD '+ Rc (1 − DD ') RD '+ Rc 
+ 2 + sTs mc D '  denol ( s ) .
 ω n Q z ω  (R + Rc )D ' ( + )
 n 
R R c D ' 

The large parenthesis in (3.107) is approximately equal to Fh−1 ( s ) in (3.54)

for small Rc .
The large parenthesis in (3.101) is rewritten by using (3.103), (3.106),
(3.35) and (3.98):

 LD 
Ri−1 D ' 1 − s  k f denol ( s ) +
 RD ' 2 
 LD 
Ri−1 D ' 1 − s  k r RDD' (1 + sRc C ) −
 RD ' 2 
 LD 
D ' 1 − s  H e ( s ) D (1 + s(R + Rc )C ) + DD ' den( s) = K =
 RD ' 2  (3.108)
T  2 
 s D '  m RD'+ Rc + L k  + D 1 + s  −
 L  c
R + Rc
f  RD '  2 

DTs Ri   ωn 
Ts  D L 
s  + k f   Ddenol ( s ) .
RD '  2 Ts Ri 


(3.101) is rewritten by using (3.108):

Chapter 3. Current-Mode Control 133

vˆo ( s)  RTs  RD'+ Rc L  D s2 

= D '  mc + k f  + 1 + 2 −
vˆ g ( s )  L R + Rc  
  DTs Ri  D'  ω n 
(3.109)
T D L   D (1 + sRc C )
s s  + k f   .
D '  2 Ts Ri 
  den( s )

The transfer functions obtained by applying the Ridley model to the

buck-boost converter are now summarized. The denominators are the same
and given by (3.106). The control-to-output transfer function of the closed
loop system is given directly by (3.97). The output impedance of the closed
loop system is obtained by combining (3.100) and (3.107). The audio
susceptibility of the closed loop system is obtained by combining (3.109) and
(3.102). The results are:

 LD 
RD ' (1 + sRc C ) 1 − s 
ˆv o ( s ) ˆvo ( s )  RD ' 2  (3.110)
= = ,
iˆc ( s ) vˆc ( s ) Ri den( s )

R(1 + sRc C )
Z out ( s ) = •
den( s )
 R(R + Rc D ') − Rc2 D Ts mc DD ' RRc (RD '+ Rc ) Rc
 + + +

 (R + Rc ) 2
L (R + Rc ) 2 (R + Rc )D' (3.111)

 s
 s 2  RD'+ Rc (1 − DD ') RD'+ Rc 
+ 2 + sTs mc D ' ,
 ω n Q z ω  (R + Rc )D '
 n 
(R + Rc )D' 
134 Chapter 3. Current-Mode Control

vˆo ( s)  RTs  RD'+ Rc  D  D  s2 

= D '  mc − 1 −   + 1 + 2 −
vˆ g ( s )  L  R + Rc  2   D '  ω n 


Ts  D  D    D(1 + sRc C )
s  − D1 −    =
D'  2  2    den( s )
(3.112)
 RT  
 s D '  m RD'+ Rc − 1 − D   + D 1 + s
2
+
 L  c R+R  2   D '  ω n2 
  c 
T D  D(1 + sRc C )
s s  ,
2  den( s )

where

den( s ) =
Ts mc RD'+ Rc   RD '+ Rc  
 RD ' 
 R+R + sR C 
 + sL (1 + s ( R + R )C )−
L R + Rc  
c c
 c  
RD '3 Ts
(1 + sRc C ) 1 − s LD2  +
(3.113)
2L  RD' 
 RD ' (1 + D ) + Rc RD'+ Rc 
H e ( s ) + sC .

 (R + Rc )D' D' 

and H e (s ) is defined in (3.35).

Simulation Results

In this section, a simulation model of a buck-boost converter with

current-mode control is presented. The transfer functions derived in the
previous subsection are compared with simulation results.
The simulation model shown in Figure 3.9 is used, except the buck
converter subsystem is replaced with the buck-boost converter subsystem
shown in Figure 2.20. The parameters used in the simulation model
presented in Section 2.10 are also used here. Ic is adjusted manually so that
the average value of the output voltage, Vo , is equal to 8 V ( D =0.620). M e
is calculate by using (3.21) and (3.104):
Chapter 3. Current-Mode Control 135

-10

-12

-14

-16
Phase (deg); Magnitude (dB)

-18

-20
1 2 3 4
10 10 10 10

-50

-100

-150

-200

-250
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 3.17: The control-to-output transfer function of a buck-boost

converter with a current controller. X: the simulation results.
Solid line: the Ridley model.

Vg
Me = (mc − 1) , (3.114)
L

where mc is chosen to be 2.
Figure 3.17 shows the Bode plot for the control-to-output transfer
function in (3.110) together with the simulation results. Ri is set to 1 Ω.
From the figure it is seen that the control-to-output transfer function
predicted by the Ridley model agrees closely with the simulation results.
Figure 3.18 shows the Bode plot for the output impedance in (3.111)
together with the simulation results. From the figure it is seen that the output
impedance predicted by the Ridley model agrees closely with the simulation
results.
136 Chapter 3. Current-Mode Control

-10

-20
Phase (deg); Magnitude (dB)

-30

-40
1 2 3 4
10 10 10 10

-20

-40

-60

-80
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 3.18: The output impedance of a buck-boost converter with a current

controller. X: the simulation results. Solid line: the Ridley
model.

Figure 3.19 shows the Bode plot for the audio susceptibility in (3.112)
together with the simulation results. From the figure it is seen that the audio
susceptibility predicted by the Ridley model does not agree closely with the
simulation results at high frequencies.
The conclusions about the agreement between the simulation results and
the transfer functions obtained from the Ridley model are thus the same as
for the buck converter and the boost converter.

3.8 Summary and Concluding Remarks

The Ridley and Tan models were in this chapter used to obtain the
control-to-output transfer function, the output impedance, and the audio
susceptibility of the buck converter with current-mode control. These transfer
functions were compared with results from simulations of a buck converter.
The main results of the comparison are:
Chapter 3. Current-Mode Control 137

-10

-20
Phase (deg); Magnitude (dB)

-30

-40
1 2 3 4
10 10 10 10

-20

-40

-60

-80
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 3.19: The audio susceptibility of a buck-boost converter with a current

controller. X: the simulation results. Solid line: the Ridley
model.

• The control-to-output transfer functions predicted by the Ridley and

Tan models are almost the same and they agree closely with the
simulation results.
• The output impedances predicted by the Ridley and Tan models are
almost the same and they agree closely with the simulation results.
• The audio susceptibilities predicted by the Ridley and Tan models
are not the same and neither agrees closely with the simulation results
at high frequencies.

The reason why the control-to-output transfer functions and the output
impedances are so accurate at high frequencies is that the perturbation in the
output voltage is small at high frequencies. In the case where the audio
susceptibility is considered, the perturbation in the input voltage is not small
at high frequencies while the frequency function is evaluated.
In this chapter, the Ridley model was also used to obtain the transfer
functions for the boost and buck-boost converters with current-mode control.
138 Chapter 3. Current-Mode Control

These transfer function were compared with results from simulations. The
result of this comparison is that the audio susceptibilities predicted by the
Ridley model do not agree closely with the simulation results at high
frequencies, i.e. the same result as was obtained for the buck converter.
 Chapter 4 A Novel Model

Models for converters with current-mode control were considered in

Chapter 3. We showed that the way the changes in the input and output
voltages are treated in the Ridley and Tan models introduces a modeling error
at high frequencies. We also showed that this modeling error is significant for
the audio susceptibility. To obtain an accurate model for the audio
susceptibility, the changes in the input and output voltages must be treated in
a more refined way. In this chapter, a novel model for the audio susceptibility
is derived and compared with simulation results. This model will be utilized
in Chapter 5 to improve the Ridley and Tan models.

4.1 Chapter Survey

A novel model for the audio susceptibility is derived in Section 4.2. In
Section 4.3, this model is applied to the buck converter and the result is
compared with simulation results and the Ridley and Tan models. In Section
4.4, the novel model is applied to the boost converter and the result is
compared with simulation results and the Ridley model. The corresponding
work is made for the buck-boost converter in Section 4.5. A summary and
concluding remarks are presented in Section 4.6.

4.2 A Novel Model for the Audio Susceptibility

In this section, a novel model for the audio susceptibility is derived. It is
first shown that it is easy to obtain a model if the perturbation in the duty
cycle, dˆ ( s ) , is known. As an intermediate step, a model for dˆ ( z ) is derived.
To be able to motivate that the novel model is reasonable, the spectrum of
dˆ (n) and vˆo (t ) are examined. Finally, the novel model is presented.

139
140 Chapter 4. A Novel Model

Introduction

The audio susceptibility describes how a perturbation in the input voltage

affects the output voltage. Figure 4.1 shows the converter and the current
controller. In the case where the audio susceptibility is considered, the
reference signal, ic (t ) , is constant and the input voltage, v g (t ) , is perturbed.
If the current controller is not present, a perturbation in v g (t ) causes a
perturbation in the output voltage and the inductor current according to the
transfer functions

 vˆo ( s ) 
  , (4.1)
 vˆ g ( s ) 
  ol

 iˆL ( s) 
  , (4.2)
 vˆ g ( s ) 
  ol

which were derived for the buck, boost, and buck-boost converters in
Chapter 2. In the case where the current controller is present, a perturbation
in v g (t ) causes a perturbation in the duty cycle of δ (t ) since the inductor
current is fed back to the current controller. The perturbation in the duty
cycle of δ (t ) causes a perturbation in the output voltage and the inductor
current according to the transfer functions

vˆo ( s )
, (4.3)
dˆ ( s )

iˆL ( s )
, (4.4)
dˆ ( s )

that were derived in Chapter 2. Since the rule of superposition holds for the
linearized converter, the Laplace transform of the perturbed output voltage is

 vˆ ( s )  ˆ
vˆo ( s ) =  o  vˆ g ( s ) + vo ( s ) dˆ ( s ) . (4.5)
 vˆ g ( s )  dˆ ( s )
  ol
Chapter 4. A Novel Model 141

vg(t) vo(t)
Converter i (t)
ic(t) Current δ (t) L

controller

Figure 4.1: The converter and the current controller.

If dˆ ( s ) were known, vˆo ( s ) could be calculated with (4.5). A model for

dˆ ( z ) (the Z-transform of the sampled version of dˆ (t ) ) will first be derived.

A Model for d̂ ( z )

In this subsection, a model for dˆ ( z ) is derived. A model for the discrete-

time signal dˆ (n) is first derived by considering the perturbations in the slopes
of the inductor current. The result is then transformed to the frequency
domain.
The voltage across the inductor depends on the input and output voltages
and the topology of the converter according to Table 4.1, see Chapter 2 or
Ridley (1990b, Table 4.2). The positive voltage across the inductor as the
transistor is on is called v on (t ) while the positive voltage across the inductor
as the transistor is off is called voff (t ) . Both these voltages are here defined to
be equal to the expressions in Table 4.1 for all t , i.e. the expressions for
v on (t ) is valid also when the transistor is off and the expressions for voff (t )
is valid also when the transistor is on.

Table 4.1: The positive voltage across the inductor.

Buck Boost Buck-Boost

v on (t ) v g (t ) − vo (t ) v g (t ) v g (t )
voff (t ) v o (t ) v o (t ) − v g (t ) v o (t )
142 Chapter 4. A Novel Model

The slopes of the inductor current are calculated by:

1
m1 (t ) = M 1 + mˆ 1 (t ) = (Von + vˆon (t ) ) = 1 von (t ) , (4.6)
L L

m 2 (t ) = M 2 + mˆ 2 (t ) =
1
L
( 1
)
Voff + vˆoff (t ) = v off (t ) .
L
(4.7)

m1 (t ) is the positive slope of the inductor current while the transistor is on.
Note that it is defined for all t since it is a function of v on (t ) . The negative
slope of the inductor current as the transistor is off corresponds to − m2 (t ) .
This function is also defined for all t since it is a function of voff (t ) . All of
m1 (t ) , M 1 , v on (t ) , Von , m 2 (t ) , M 2 , voff (t ) , and Voff are positive with
these definitions and m1 (t ) , mˆ 1 (t ) , v on (t ) , vˆon (t ) , m 2 (t ) , mˆ 2 (t ) , voff (t ) ,
and vˆoff (t ) are defined for all t with these definitions. The perturbation
signals mˆ 1 (t ) and mˆ 2 (t ) are zero when there are no perturbations of the
input and output voltages.
When the audio susceptibility is considered, ic (t ) is constant and equal
to its dc value I c . Figure 4.2 shows the waveforms of I c , I c minus the
external ramp, ie (t ) , and two different versions of the inductor current. The
first version ( iLss (t ) , solid line) shows how the inductor current waveform is
in steady state, i.e. when there are no perturbations of the inductor current
slopes m1 (t ) and − m2 (t ) . The dashed line shows an example of how the
inductor current waveform if there are perturbations of the inductor current
slopes. The transistor is assumed to turn on at the points t = nTs , where n is
an integer. In steady state the transistor will then turn off at the points
t = (n + D )Ts . To find out how much the inductor current changes, the
inductor current slopes are integrated. The following two equations are
obtained from Figure 4.2:

( )
Ts k
I c − i L (Ts k ) = M eTs D + dˆ ( k − 1) + ∫
m2 (t )dt , (4.8)
(
Ts k −1+ D + dˆ ( k −1) )
(
Ts k + D + dˆ ( k ) )
(
I c − i L (Ts k ) = M eTs D + dˆ ( k ) + ) ∫ m1 (t )dt . (4.9)
Ts k
Chapter 4. A Novel Model 143

^ ^
Tsd(k-1) Tsd(k)
Ic
-Me
-m2(t) Ic-ie(t)
M1

-M2 iL(t)
m1(t) iLss(t)
t
Ts(k-1) Ts(k-1+D) Tsk Ts(k+D) Ts(k+1)

Figure 4.2: The waveforms in steady state and in the case where the input and
output voltages change.

By use of (4.7) the integral in (4.8) is rewritten as:

Ts k Ts k (
Ts k −1+ D + dˆ ( k −1) )
∫
m2 (t )dt = ∫ m2 (t )dt − ∫ m2 (t )dt =
(
Ts k −1+ D + dˆ ( k −1) ) Ts (k −1+ D ) Ts (k −1+ D )
(4.10)
Ts k (
Ts k −1+ D + dˆ ( k −1) )
M 2Ts D '+ ∫ mˆ 2 (t )dt − M 2Ts dˆ (k − 1) − ∫ mˆ 2 (t )dt .
Ts (k −1+ D ) Ts (k −1+ D )

Using (4.6) the integral in (4.9) is rewritten as:

(
Ts k + D + dˆ ( k ) ) Ts (k + D ) (
Ts k + D + dˆ ( k ) )
∫ m1 (t )dt = ∫ m1 (t )dt + ∫ m1 (t )dt =
Ts k Ts k Ts (k + D )
(4.11)
Ts (k + D ) (
Ts k + D + dˆ ( k ) )
M 1Ts D + ∫ mˆ 1 (t )dt + M 1Ts dˆ (k ) + ∫
mˆ 1 (t )dt .
Ts k Ts (k + D )

(4.8) and (4.9) are combined and by using (4.10) and (4.11) the following is
obtained:
144 Chapter 4. A Novel Model

(
M eTs D + dˆ (k ) + )
Ts (k + D ) Ts (k + D + dˆ ( k ) )
M 1Ts D + ∫ m1 (t )dt + M 1Ts d (k ) +
ˆ ˆ
∫ mˆ 1 (t )dt =
Ts k Ts (k + D )

( )
(4.12)
M T D + dˆ (k − 1) +
e s
Ts k (
Ts k −1+ D + dˆ ( k −1) )
M 2Ts D '+ ∫ mˆ 2 (t )dt − M 2Ts dˆ (k − 1) − ∫ mˆ 2 (t )dt .
Ts (k −1+ D ) Ts (k −1+ D )

(4.12) is rewritten:

Ts k Ts (k + D )
(M e + M 1 )Ts dˆ (k ) = ∫ mˆ 2 (t )dt − ∫ mˆ 1 (t )dt +
Ts (k −1+ D ) Ts k

(M e − M 2 )Ts dˆ (k − 1) + (4.13)
T (k −1+ D + dˆ ( k −1) )
s T (k + D + dˆ ( k ) ) s

(M 2 D'− M 1 D )Ts − ∫ mˆ 2 (t )dt − ∫ mˆ 1 (t )dt .

Ts (k −1+ D ) Ts (k + D )

The first term in the last row in (4.13) is zero according to (3.16). The
magnitudes of the second and third term in the last row in (4.13) are:

(
Ts k −1+ D + dˆ ( k −1) ) (
Ts k −1+ D + dˆ ( k −1) )
∫ mˆ 2 (t )dt ≤ ∫ max( mˆ 2 (t ) )dt = max( mˆ 2 (t ) )Ts dˆ (k − 1) . (4.14)
Ts (k −1+ D ) Ts (k −1+ D )

(
Ts k + D + dˆ ( k ) ) ( )
Ts k + D + dˆ ( k )

∫ mˆ 1 (t )dt ≤ ∫ max( mˆ 1 (t ) )dt = max( mˆ 1 (t ) )Ts dˆ (k ) . (4.15)

Ts (k + D ) Ts (k + D )

From (4.14) and (4.15), it is seen that these two terms can be neglected since
each one of them is less than or equal to a scaled product of two perturbation
signals. The same type of approximation is used in state-space averaging (see
Section 2.3). By using these results and (4.13), the following approximate
expression for dˆ (k ) is obtained:
Chapter 4. A Novel Model 145

Ts k Ts (k + D )

∫ mˆ 2 (t )dt − ∫ mˆ 1 (t )dt + (M e − M 2 )Ts dˆ (k − 1) (4.16)

Ts (k −1+ D ) Ts k
dˆ (k ) = .
(M e + M 1 )Ts

The integer variable k is substituted by the integer variable n and (4.16) can
therefore be rewritten as

1
dˆ (n) = (M − M 2 )Ts dˆ (n − 1) +
(M e + M 1 )Ts e
 Ts n Ts (n −1+ D ) Ts (n + D ) Ts n  (4.17)
1  mˆ (t )dt − .
(M e + M 1 )Ts  −∫∞ ∫ ∫ ∫
mˆ (t ) dt − mˆ (t ) dt + mˆ (t ) dt
2 2 1 1

−∞ −∞ −∞ 

The discrete-time signal dˆ (n) is the result of sampling the continuous-

time signal dˆ (t ) . The sampling interval changes if the converter is not in
steady state. Similarly to the discussion in Section 3.3, sampling at the points
t = (n + D )Ts is a good approximation if the magnitude of dˆ (t ) is small and
the changes of dˆ (t ) are slow in the surroundings of these sampling points.
From (4.17), it is seen that dˆ (n) is a sum of a discrete-time part (first term)
and a continuous-time part (last term). To be able to create the discrete-time
signal dˆ (n) , the continuous-time part must deliver its value at t = (n + D )Ts
so that it can be sampled and added to the discrete-time part of (4.17). In the
first integral in (4.17),

Ts n

∫ mˆ 2 (t )dt ,
−∞

the signal mˆ 2 (t ) is integrated up to t = nTs so the value of the integral is

known at t = nTs . Since the value of the integral has to be delivered at
t = (n + D )Ts , the value must be delayed by DTs . In the second integral in
(4.17),

Ts (n −1+ D )

∫ mˆ 2 (t )dt ,
−∞
146 Chapter 4. A Novel Model

the signal mˆ 2 (t ) should be integrated up to t = (n − 1 + D )Ts . Since the value

of the integral is to be delivered at t = (n + D )Ts , the value must be delayed
by Ts . In the third integral in (4.17),

Ts (n + D )

∫ mˆ 1 (t )dt ,
−∞

the signal mˆ 1 (t ) is integrated up to t = (n + D )Ts and therefore it will not be

delayed. In the fourth integral in (4.17),

Ts n

∫ mˆ 1 (t )dt ,
−∞

the signal mˆ 1 (t ) is integrated up to t = nTs and has to be delayed by DTs .

By using these results and (4.17), the signal dˆ (n) can be created as shown in
Figure 4.3.
The block diagram in Figure 4.3 is transformed from the time domain to
the frequency domain and the result is shown in Figure 4.4. A feedback is
included in Figure 4.4 to eliminate one of the input signals. The feedback is
transformed from the discrete-time part to the continuous-time part if z is
substituted with e sTs and the result is shown in Figure 4.5.
An expression for xˆ ( s) in Figure 4.5 can be obtained from Figure 4.5:

(M e − M 2 )Ts e − sT xˆ ( s)
s

xˆ ( s ) = +
(M e + M 1 )Ts
(4.18)
e − sDTs s −1 mˆ 2 ( s) − e − sTs s −1 mˆ 2 ( s) − s −1 mˆ 1 ( s ) + e − sDTs s −1mˆ 1 ( s )
.
(M e + M 1 )Ts

Rearranging (4.18) gives:

e − sDTs − e − sTs e − sDTs − 1

mˆ 2 ( s) + mˆ 1 ( s )
sTs sTs (4.19)
xˆ ( s ) = .
M e + M 1 − (M e − M 2 )e − sTs

The denominator in (4.19) is now rewritten by using (3.21) and (3.17):

Chapter 4. A Novel Model 147

^ (t)
m2
∫ Delay: TsD
Sampling
^
∫ Delay: Ts 1 d(n)
^ (t)
m1
∫ Delay: 0 (Me+M1)Ts

∫ Delay: TsD 1
^
d(n-1) (Me+M1)Ts
(Me-M2)Ts

Figure 4.3: A time-domain model for the duty cycle perturbation signal.

^ 2(s)
m
s-1 e-sTs D
Sampling
^
s-1 e-sTs 1 d(z)
^ 1(s)
m
s-1 (Me+M1)Ts

s-1 e-sTs D 1
^
d(z)z -1 (Me+M1)Ts
(Me-M2)Ts z-1

Figure 4.4: A frequency-domain model for the duty cycle perturbation signal.

^ 2(s)
m
s-1 e-sTs D Sampling
-1 -sTs ^
s e 1 d(z)
^ 1(s)
m
s -1 (Me+M1)Ts
^
x(s)
s -1
e -sTs D 1
(Me+M1)Ts
(Me-M2)Ts e-sTs

Figure 4.5. The feedback is moved to the continuous-time part.

148 Chapter 4. A Novel Model

v^ g(s) v^on(s) ^ 1(s)

m Sampling
1 ^ ^
Table L x(s) d(z)
^ 2(s) (4.21)
v^ o(s) 4.1 v^ off(s) 1 m
L

Figure 4.6: The block diagram with vˆ g (t ) and vˆo (t ) as input signals.

M e + M 1 − (M e − M 2 )e − sTs =
 D
M 1 (m c − 1) + M 1 −  M 1 (m c − 1) − M 1  e − sTs =
 D' 
M
M 1 m c − 1 (m c D'− D '− D )e − sTs = (4.20)
D'
M1
D'
(
m c D'−(m c D'−1)e − sTs = )
M1
D'
( ( )
m c D' 1 − e − sTs + e − sTs . )
(4.19) is rewritten by using (4.20):

e − sDTs − e − sTs e − sDTs − 1

mˆ 2 ( s) + mˆ 1 ( s )
sTs sTs (4.21)
xˆ ( s ) =
(m D' (1 − e )+ e )
− sTs − sTs
.
−1
M 1 D' c

By using (4.6), (4.7), (4.21), and Table 4.1, the block diagram in Figure
4.6 is obtained and it is a model for dˆ ( z ) . Note that this block diagram is
derived without the prerequisite that vˆ g (t ) and vˆo (t ) are sinusoidal.

Spectrum of the Signals

In this subsection, the spectrum of dˆ (n) and vˆo (t ) are examined.

First we simulate a buck converter with a current controller by using the
simulation model presented in Section 3.5. The simulation is conducted in
the way used to evaluate the audio susceptibility at one frequency. The
Chapter 4. A Novel Model 149

-3
x 10
2.5

1.5
|c(k)|

0.5

0
0 2 4 6 8 10 12 14 16 18 20
k

Figure 4.7: The discrete Fourier series for dˆ (n) ( ω m =5000π rad/s).

voltage vˆ g (t ) is thus sinusoidal. The simulation is conducted for the

frequency 2500 Hz ( ω m =5000π rad/s). We record the signal delta during the
simulation. From this result the duty cycle, d (n) , of the signal delta is
manually measured for the last switching periods in the simulation. The
period of d (n) is equal to 20 T s so only 20 values of d (n) are measured.
The mean value of this sequence is subtracted to obtain dˆ (n) . The discrete
Fourier series is calculated for the series dˆ (n) . Figure 4.7 shows the absolute
value of the complex Fourier coefficients, c(k ) . The value c(0) is zero since
the mean value is removed. The signal dˆ (n) has a Fourier component at ω m
which corresponds to k =1. The other Fourier components in the interval dc
to half the switching frequency ( k =10) are negligible. The coefficients
c(11) K c(19) are the complex-conjugate values of c(9) K c(1) . The
sinusoidal signal corresponding to the coefficient c(1) is presented in Figure
4.8 together with the dˆ (n) series. dˆ (n) is almost a sampled version of the
sinusoidal signal since c(2) K c(10) are negligible compared to c(1) .
150 Chapter 4. A Novel Model

-3
x 10
5

-1

-2

-3

-4

-5
0 0.5 1 1.5 2 2.5 3 3.5 4
t (s) -4
x 10
Figure 4.8: The sinusoidal signal corresponding to the coefficient c(1)
together with the dˆ (n) ( ω m =5000π rad/s).

A simulation for the frequency 12500 Hz is also conducted. In this case

the period of d (n) is equal to 4 T s so only 4 values of d (n) are measured.
Figure 4.9 shows the absolute value of the complex Fourier coefficients and
c(1) is dominating also in this case and is presented in Figure 4.10 together
with the dˆ (n) series, where dˆ (n) is almost a sampled version of the
sinusoidal signal.
The conclusion from the simulation results is that dˆ (n) approximately is
obtained by sampling a sinusoidal signal, even when ω m is chosen to be
almost as high as half the switching frequency.
To investigate this phenomenon, the spectrum of vˆo (t ) is first
considered. It is approximated with the one shown in Figure 4.11 (Verghese
and Thottuvelil, 1999). In steady state, the ripple in vˆo (t ) consists of Fourier
components with the frequencies ω s , 2ω s , 3ω s , K , where ω s = 2π Ts , i.e.
the switching frequency.
Chapter 4. A Novel Model 151
-3
x 10
2

1.8

1.6

1.4

1.2
|c(k)|

0.8

0.6

0.4

0.2

0
0 1 2 3 4
k

Figure 4.9: The discrete Fourier series for dˆ (n) ( ω m =25000π rad/s).

In the case where dˆ (n) consist of one Fourier component (in the interval
[0, ω s 2 ]) with the frequency ω m , the Fourier components with the
frequencies ω m , ω s − ω m , ω s + ω m , 2ω s − ω m , 2ω s + ω m , K , are added
to vˆo (t ) . The Fourier component with the frequency ω s − ω m is smaller
than the one with the frequency ω m since the converter has a low-pass
output filter. For ω m almost as high as half the switching frequency, the
Fourier component with the frequency ω s − ω m has a magnitude that is
almost as high as the one with the frequency ω m . However, for low ω m , the
Fourier component with the frequency ω s − ω m is negligible compared to
the one with the frequency ω m .
Also when vˆg (t ) consists of one Fourier component with the frequency
ωm , the Fourier components with the frequencies
ω m , ω s − ω m , ω s + ω m , 2ω s − ω m , 2ω s + ω m , K , are added to vˆo (t ) . A
Fourier component with higher frequency has a lower magnitude also in this
case.
Assume that each one of dˆ (n) and vˆg (t ) consists of one Fourier
component with the frequency ω m and that the name dˆ ( z ) after the
sampler in Figure 4.6 is replaced with xˆ ( z ) . Then vˆo (t ) will approximately
152 Chapter 4. A Novel Model

-3
x 10
4

-1

-2

-3

-4
0 2 4 6 8
t (s) -5
x 10
Figure 4.10: The sinusoidal signal corresponding to the coefficient c(1)
together with the dˆ (n) ( ω m =25000π rad/s).

consist of Fourier components with frequencies shown in Figure 4.11. The

signal xˆ (t ) (the inverse Laplace transform of xˆ ( s) ) in Figure 4.6 will also
consist of Fourier components with frequencies shown in Figure 4.11 since all
the blocks in Figure 4.6 represent linear systems. The input signal vˆg (t ) (the
inverse Laplace transform of vˆ g ( s ) ) in Figure 4.6 contributes only to the
Fourier component in xˆ (t ) with the frequency ω m . If xˆ (t ) is sampled with
the frequency ω s as shown in Figure 4.6, the Fourier components in xˆ (t )
with the frequencies ω s , 2ω s , 3ω s , K only contribute to the dc value of
x(n) , i.e. the ac part, xˆ (n) , is not affected. Since the sampling interval in
fact varies a little, the ac part may be affected. However, this is not taken into
account in the following discussion. The Fourier components in xˆ (t ) with
the frequencies ω s − ω m , ω s + ω m , 2ω s − ω m , 2ω s + ω m , K only
contribute to the Fourier component (in the interval [0, ω s 2 ]) in xˆ (n)
with the frequency ω m . These effects are due to aliasing (Åström and
Chapter 4. A Novel Model 153

ω
0 ωm ωs-ωm ωs ωs+ωm 2ωs-ωm 2ωs 2ωs+ωm

Figure 4.11: The approximate spectrum of vˆo (t ) .

Wittenmark, 1997, Section 7.4). Thus xˆ (n) will only consist of one Fourier
component (in the interval [0, ω s 2 ]) which explains the conclusion made
from the simulation results.
Assume that the input signal vˆo (t ) in Figure 4.6 is replaced with a signal
consisting of just the Fourier component in vˆo (t ) with the frequency ω m .
This introduce an error in xˆ (t ) . For low ω m the relative error in xˆ (t ) is
small since the Fourier components in vˆo (t ) with the frequencies ω s − ω m ,
ω s + ω m , 2ω s − ω m , 2ω s + ω m , K are negligible compared to the one
with the frequency ω m . For ω m almost as high as half the switching
frequency, the Fourier component in vˆo (t ) with the frequency ω s − ω m is
significant compared to the one with the frequency ω m . This would
introduce a large relative error in xˆ (t ) if the input signal vˆg (t ) in Figure 4.6
is zero. In Section 3.5, it was concluded that the magnitude of vˆo (t ) is much
smaller than the magnitude of vˆg (t ) at high frequencies (i.e. almost as high
as ω n ). The relative error in xˆ (t ) is therefore small also at high frequencies.
The important conclusions are now summarized for the case where vˆ g (t )
is a sinusoidal with the frequency ω m :

• The discrete-time signal d (n) consists of just one Fourier

component (approximately) and it has the frequency ω m . Therefore,
d (n) can be obtained by sampling a sinusoidal signal.
• A good approximation of the signal xˆ (t ) in Figure 4.6 is obtained if
the input signal vˆo (t ) is replaced with a signal consisting of just the
Fourier component in vˆo (t ) with the frequency ω m .
154 Chapter 4. A Novel Model

v^ o(s)
^ 1(s) v^ g(s) ol
v^ g(s) v^on(s) m
1 ^ ^
Table L x(s) d(s) v^ o(s) v^ o(s)
^vo(s) 4.1 v^ off(s) m^ 2(s)(4.21) ^
1 d(s)
L

Figure 4.12: A model of the audio susceptibility.

A Novel Model of the Audio Susceptibility

To find a linear model of the audio susceptibility that is accurate from dc

to half the switching frequency it is sufficient to consider just the case where
the perturbation in the input voltage, vˆg (t ) , is a sinusoidal signal with the
frequency ω m . The model should be accurate for any ω m in the frequency
interval. From the summarized conclusions above, it seems reasonable that an
accurate model of the audio susceptibility can be obtained from the block
diagram in Figure 4.12, where the block diagram in Figure 4.6 is combined
with equation (4.5). xˆ ( s) is used as dˆ ( s ) and the output vˆo ( s ) from (4.5) is
fed back to the input vˆo ( s ) in the block diagram in Figure 4.6. The Fourier
component in the output voltage with the frequency ω m (see Section 2.11) is
correctly predicted in (4.5) and it is enough to use this component as the
vˆo ( s ) -input in the block diagram in Figure 4.6. In Section 4.4, it will be
shown that the prediction of the audio susceptibility by means of the block
diagram in Figure 4.12 has a serious shortcoming. At low frequencies, the
prediction is very sensitive to modeling errors in the different blocks in the
diagram.

4.3 Audio Susceptibility of the Buck Converter

In this section, the novel model derived in Section 4.2 is applied to the
buck converter. The obtained expression is compared with simulation results,
experimental result, and the Ridley and Tan models.
Chapter 4. A Novel Model 155

A Novel Expression

The audio susceptibility of the buck converter with current-mode control

is obtained by using Figure 4.12, Table 4.1, (4.21), (2.92), (2.90), and (3.51):

RD(1 + sRc C )
vˆ o ( s ) = vˆ g ( s) +
denol ( s )
 e − sDTs − e − sTs vˆ o ( s ) e − sDTs − 1 vˆ g ( s ) − vˆ o ( s ) 
 +  (4.22)
RV g (1 + sRc C )  sTs L sTs L 
 ,
denol ( s ) 

V g − V o −1
L
(
D' m c D' 1 − e ( − sTs
+e )
− sTs
) 

 

where

denol ( s ) = R + s(L + RRc C ) + s 2 (R + Rc )LC =

(4.23)
R (1 + sRc C ) + sL(1 + s (R + Rc )C ) .

(4.22) is rewritten by using (2.66), (2.68), and (2.39):

denol ( s )
Vg
L
(m D' (1 − e )+ e )vˆ
c
− sTs − sTs
o (s) =

Vg
L
(m D' (1 − e )+ e )RD(1 + sR C )vˆ
c
− sTs − sTs
c g ( s) + (4.24)
 e − sDTs − e − sTs vˆ o ( s ) e − sDTs − 1 vˆ g ( s ) − vˆ o ( s ) 
RV g (1 + sRc C )  + .
 sT L sT L 
 s s 

(4.24) is rewritten:
156 Chapter 4. A Novel Model

vˆ o ( s )  V g
=
vˆ g ( s )  L
( ( ) )
m c D' 1 − e − sTs + e − sTs RD(1 + sRc C ) +

e − sDTs − 1 
RV g (1 + sRc C ) •
sTs L 
(4.25)
 V
( ( ) )
 denol ( s ) g m c D' 1 − e − sTs + e − sTs −
 L

−1
 e − sDTs − e − sTs e − sDTs − 1  
RV g (1 + sRc C )  −  .
 sT L sT L 
 s s 

The following is obtained if the numerator and denominator in (4.25) are

( (
multiplied with Ts V g 1 − e − sTs : ))
Chapter 4. A Novel Model 157

vˆ o ( s )  Ts  − sTs 
=  m c D'+ e  RD(1 + sRc C ) +
vˆ g ( s )  L  1 − e − sTs 
 
e − sDTs − 1 1 
RTs (1 + sRc C ) •
sTs L 1 − e − sTs 

−1
  − sTs  
 den ( s ) Ts  m c D '+ e  − RTs (1 + sRc C ) 1  =
 ol
L  1 − e − sTs  sTs L 
  
RTs  e − sTs e − sDTs − 1 1 
D m c D '+ +  (1 + sRc C ) •
L  1− e s− sT sDTs 1 − e − sTs 
 
  − sTs 
 sL(1 + s (R + R )C ) Ts  mc D '+ e +
 c
L  1 − e − sTs 
  
−1 (4.26)
T  − sTs  
R (1 + sRc C ) s  mc D '+ e  − RTs (1 + sRc C ) 1  =
L  1 − e − sTs  sTs L 
 
RTs  1  sTs 1 − e − sDTs sTs  
D m c D '−  −  (1 + sRc C ) •
L  sT  1 − e − sTs sDT e sTs − 1  
 s  s

  sTs 
 (1 + s(R + Rc )C ) + sTs m c D'  +

  e −1
sTs

−1
RTs  1 1  
 m c D'+ −  (1 + sRc C ) =
 sTs  
L  e − 1 sTs  
RTs
L
( )
D m c D'− F f ( s ) (1 + sRc C )
,
den( s)

where

1  sTs 1 − e − sDTs sTs 

F f ( s) =  − =
sTs  1 − e − sTs sDT e sTs − 1 
 s
(4.27)
 D  (3 − 2 D )DTs
1 −  − s−
(
1 − 2 D + D 2 DTs2 2
s + K,
)
 2 12 24
158 Chapter 4. A Novel Model

den( s ) = (1 + s(R + Rc )C )(H e ( s ) + sTs mc D') +

RTs  1 − H e (s)  (4.28)
 m c D'−  (1 + sRc C ) ,
 
L  sTs 

and H e (s ) is the same as in (3.10). The Taylor series of F f (s ) is also shown

in (4.27). (4.26) is the novel expression for the audio susceptibility of the
buck converter with current-mode control. The denominator in this novel
expression will now be compared with the denominators in the Ridley and
Tan models.
If H e (s ) in (4.28) is approximated by (3.14) and (3.54) is used, the
following approximate denominator is obtained:

den( s ) = (1 + s(R + Rc )C )Fh−1 ( s) +

 s s2 
 1−1− − 2
RTs  ω nQz ω n 
 m c D'−  (1 + sRc C ) =
L  sTs 
 
 
 sTs s 2Ts  (4.29)
 −
RTs  2 πω n 
(1 + s(R + Rc )C )Fh−1 (s) +  mc D '−  (1 + sRc C ) =
L  sTs 
 
 
  
(1 + s(R + Rc )C )Fh−1 (s) + RTs  mc D '−0.51 − s 2   (1 + sRc C ) .

L 
  πω n  

(4.29) is exactly the same as (3.72), i.e. the denominator in the Tan model. In
Section 3.4, it was concluded that the denominator in the Tan model is
almost the same as the denominator in the Ridley model, (3.62). The novel
expression and the audio susceptibility predicted by the Ridley and Tan
models thus have approximately the same denominator but three different
numerators, compare (4.26), (3.61), and (3.71).
Chapter 4. A Novel Model 159

-20

-30

-40

-50
Phase (deg); Magnitude (dB)

-60

-70
1 2 3 4
10 10 10 10

-50

-100

-150
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 4.13: The audio susceptibility of a buck converter with a current

controller ( mc =2). X: the simulation results. Dotted line: the
novel expression. Solid line: the Ridley model. Dashed line: the
Tan model.

A Comparison of the Novel Expression and the Simulation

Results

In this subsection, the novel expression obtained in the previous

subsection is compared with simulation results, experimental result, and the
Ridley and Tan models.
Figure 4.13 shows the Bode plot for the audio susceptibility according to
the novel expression in (4.26) together with the results presented in Figure
3.13, i.e. the simulation results and the audio susceptibilities predicted by the
Ridley and Tan models. From the figure it is seen that the novel expression
agrees closely with the simulation results also at high frequencies.
Figure 4.14 shows the same as Figure 4.13 except mc is changed from 2
to 1.5 and experimental result presented by Ridley is included (copied
manually from plot in Ridley (1991)). From the figure it is seen that the
160 Chapter 4. A Novel Model

-30

-40

-50

-60
Phase (deg); Magnitude (dB)

-70

-80
1 2 3 4
10 10 10 10

-50

-100

-150
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 4.14: The audio susceptibility of a buck converter with a current

controller ( mc =1.5). X: the simulation results. Dotted line: the
novel expression. Solid line: the Ridley model. Dashed line: the
Tan model. Dash-dotted line: the measurement made by
Ridley (the phase shift curve is not available).

novel expression agrees closely with the simulation results also at high
frequencies. The choice mc =1.5 makes the audio susceptibility very small at
dc since there is a subtraction between two almost equal values in the
numerator of transfer function. The modeling errors in the Ridley and Tan
models cause larger relative errors at high frequencies in this case. This is most
evident in the Ridley model. Ridley (1991) explains the difference between
the audio susceptibility predicted by his model and experimental result at
mc =1.5 by saying that the measurements were unreliable due to noise and
grounding problems. The experimental result from Ridley agrees closely (if
we take into consideration that it is an experimental result) with the
simulation results and the novel expression as seen from Figure 4.14. This
indicates that it is Ridley’s measurements that are correct, not his model.
Chapter 4. A Novel Model 161

-20

-30

-40
Phase (deg); Magnitude (dB)

-50

-60
1 2 3 4
10 10 10 10

150

100

50

0
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 4.15: The audio susceptibility of a buck converter with a current

controller ( mc =1). X: the simulation results. Dotted line: the
novel expression. Solid line: the Ridley model. Dashed line: the
Tan model.

Figure 4.15 shows the same as Figure 4.13 except mc is changed to 1.

From the figure it is seen that the novel expression agrees closely with the
simulation results also at high frequencies. The choice mc =1 does not make
the audio susceptibility small at dc. The result of the subtraction in the
numerator of the transfer function is of opposite sign at dc compared to the
case where mc =2. This is seen from the phase shift curves in Figure 4.13 and
Figure 4.15.

4.4 Audio Susceptibility of the Boost Converter

In this section, the novel model derived in Section 4.2 is applied to the
boost converter. The obtained expression is compared with simulation results
and the Ridley model. It is observed that the novel expression makes strange
predictions at low frequencies and the cause of this is examined.
162 Chapter 4. A Novel Model

A Novel Expression

The audio susceptibility of the boost converter with current-mode control

is obtained by using Figure 4.12, Table 4.1, (4.21), (2.144), (2.142), and
(3.86):

RD ' (1 + sRc C )
vˆ o ( s ) = vˆ g ( s ) +
denol ( s )
Vg
(RD'+ Rc )D'
(R 2
)
D' 2 − s(R + Rc )L (1 + sRc C )
•
denol ( s ) (4.30)
e − sDTs
−e − sTs vˆ o ( s ) − vˆ g ( s ) e − sDTs
− 1 vˆ g ( s ) 
 +
 sTs L sTs L 
 ,


Vg
( (
D' −1 m c D' 1 − e − sTs ) +e − sTs
) 

 L 

where

denol ( s ) =
RD' (RD'+ Rc ) (R + Rc ) + s(L + RRcCD ') + s 2 (R + Rc )LC = (4.31)
RD' ((RD'+ Rc ) (R + Rc ) + sRcC ) + sL(1 + s(R + Rc )C ) ,

(4.30) is rewritten:

denol ( s )
Vg
LD '
(m D' (1 − e )+ e )vˆ
c
− sTs − sTs
o (s) =

Vg
LD '
(m D' (1 − e )+ e )RD' (1 + sR C )vˆ
c
− sTs − sTs
c g ( s) +
(4.32)
(RD'+ Rc
Vg
)D'
(R 2
)
D' 2 − s(R + Rc )L (1 + sRc C ) •

 e − sDTs − e − sTs vˆ o ( s ) − vˆ g ( s ) e − sDTs − 1 vˆ g ( s ) 

 + .
 sTs L sTs L 


(4.32) is rewritten:
Chapter 4. A Novel Model 163

vˆ o ( s )  V g
=
vˆ g ( s )  LD '
( ( ) )
m c D ' 1 − e − sTs + e − sTs RD' (1 + sRc C ) +

Vg
(RD'+ Rc )D'
( )
R 2 D' 2 − s(R + Rc )L (1 + sRc C ) •

 e − sDTs − 1 e − sDTs − e − sTs 

 −  •
 sTs L sTs L  (4.33)
 
 V
( (
 denol ( s ) g m c D' 1 − e − sTs + e − sTs −
 LD '
) )

−1
e − sDTs − e − sTs 
Vg
(RD'+ Rc )D'
(
R D' − s(R + Rc )L (1 + sRc C )
2 2
)
sTs L


.


The following is obtained if the numerator and denominator in (4.33) are

( (
multiplied with D' Ts V g 1 − e − sTs : ))
vˆ o ( s )  Ts  − sTs 
=  m c D'+ e  RD' (1 + sRc C ) −
vˆ g ( s )  L  1 − e − sTs 
 
Ts
RD'+ Rc
( )
R 2 D ' 2 − s (R + Rc )L (1 + sRc C )
1 
•
sTs L 
(4.34)
  − sTs 
 den ( s ) Ts  m c D '+ e −
 ol
L  1 − e − sTs 
  
−1
e − sDTs − e − sTs 
Ts
( )
R 2 D ' 2 − s (R + Rc )L (1 + sRc C )
1 

RD'+ Rc sTs L 1 − e − sTs 

The numerator in (4.34) is rewritten:

164 Chapter 4. A Novel Model

Ts  − sTs 
 m c D'+ e  RD' (1 + sRc C ) −
L  − − sTs 
 1 e 
Ts
RD'+ Rc
( )
R 2 D ' 2 − s (R + Rc )L (1 + sRc C )
1
sTs L
=

RTs  e − sTs R 2 D' 2 − s(R + Rc )L 1 

D '  m c D '+ − (1 + sRc C ) =
L 
 1 − e − sTs RD ' (RD'+ Rc ) sTs  (4.35)

 RTs  1 RD' 1  R + Rc 
 D'  mc D '+ sT − +  (1 + sRc C ) =
 L  
  e s − 1 RD'+ Rc sTs  RD'+ Rc 
 RTs   RD'  1  R + Rc 
 D'  mc D '− F f ( s ) + 1 −  +  (1 + sR C ) ,
 L   sT  RD'+ R  c
   RD'+ Rc  s  c 

where

1 1 1 T T3
F f ( s) = − sT = − s s + s s3 + K. (4.36)
sTs e s − 1 2 12 720

The Taylor series of F f (s ) is also shown in (4.36). The denominator in

(4.34) is now rewritten:
Chapter 4. A Novel Model 165

Ts  − sTs 
denol ( s )  m c D '+ e −
L  1 − e − sTs 
 
e − sDTs − e − sTs
Ts
( )
R 2 D ' 2 − s (R + Rc )L (1 + sRc C )
1
=
RD'+ Rc sTs L 1 − e − sTs
T T 1
denol ( s ) s m c D'+ denol ( s ) s sT −
L L e s −1
e sTs e − sDTs − 1 1
Ts
RD'+ Rc
( )
R 2 D ' 2 − s (R + Rc )L (1 + sRc C ) =
sTs L e sTs − 1
T
denol ( s ) s m c D'+
L
  RD'+ Rc   Ts 1
 R + R + sRc C  + sL(1 + s (R + Rc )C ) L sTs
 RD '    −
 e −1
  c  
Ts D'
L(RD'+ Rc )
( )
R 2 D ' 2 − s (R + Rc )L (1 + sRc C )
e sD 'Ts − 1 1
sD ' Ts e sTs − 1
=

Ts T RD ' RD'+ Rc 1
denol ( s ) m c D '+ s + (4.37)
L L R + Rc e s − 1
sT

RD' sT sT
Rc C sT s + (1 + s (R + Rc )C ) sT s −
L e −1s
e s −1
 Ts D' e sD 'Ts − 1 1
 R 2 D' 2 +
 L(RD'+ R ) sD' Ts e sTs − 1
 c

D' e sD 'Ts − 1 sTs

R 2 D' 2 Rc C −
L(RD'+ Rc ) sD' Ts e sTs − 1
D' sD 'Ts
− 1 sTs 
(R + Rc )(1 + sRc C ) e =
RD'+ Rc sD' Ts e sTs − 1 
Ts  RD' 
denol ( s ) m c D '+  1 + Rc C + s (R + Rc )C  H e ( s ) −
L  L 
 D' D ' (R + R c )  sD 'Ts − 1
 R 2 D' 2 Rc C − (1 + sRc C ) e H e (s) +
 L(RD'+ Rc ) RD'+ Rc  sD' Ts
 Ts RD ' RD'+ Rc Ts D' e sD 'Ts − 1  1
 − R 2 D' 2 ,
 L
 R + Rc L(RD'+ Rc ) sD' Ts  e sTs − 1
166 Chapter 4. A Novel Model

where H e (s ) is the same as in (3.10). The novel expression for the audio
susceptibility of the boost converter with current-mode control is obtained by
using (4.35) and (4.37) to rewrite (4.34):

vˆ o ( s )
=
vˆ g ( s )
 RT 
 s D'  m D'− F ( s ) + 1 − RD '  1

 R + Rc
+  (1 + sR C ) (4.38)
 L  c f  RD '+ Rc  sT  RD '+ R  c
    s  c 
,
den( s )

where

den( s ) =
Ts m c   RD'+ Rc  
D'  RD'  + sRc C  + sL(1 + s(R + Rc )C ) −
 
L   R + Rc  
 R D' Rc C (R + Rc )D'
2 3  e s −1
sD 'T

 L(RD'+ Rc )
− (1 + sRc C )
 sD' Ts
H e (s) +
 RD '+ R c  (4.39)
 RD' 
H e ( s )1 + Rc C + s (R + Rc )C  +
 L 
Ts RD'3  RD'+ Rc RD' e sD 'Ts − 1  1
 − H (s) ,
L  RD'+ R D' RD'+ R sD' T  e sD ' T
 c c s  s

F f (s ) is defined in (4.36) and H e (s ) is defined in (3.10).

If (4.38) and (4.39) are compared with (3.94) and (3.95), i.e. the Ridley
model, it is seen that there are some differences both in the numerator and in
the denominator.

A Comparison of the Novel Expression and the Simulation

Results

In this subsection, the novel expression obtained in the previous

subsection is compared with simulation results and the Ridley model. It is
Chapter 4. A Novel Model 167

-20
Phase (deg); Magnitude (dB)

-40

-60
0 1 2 3 4
10 10 10 10 10

-50

-100

-150
0 1 2 3 4
10 10 10 10 10
Frequency (Hz)

Figure 4.16: The audio susceptibility of a boost converter with a current

controller ( Rc =14 mΩ). X: the simulation results. Dotted line:
the novel expression. Solid line: the Ridley model.

observed that the novel expression makes strange predictions at low

frequencies. To understand the reason for this, the structure of the novel
expression is examined as a first step. The cause of the strange predictions will
be examined further in the next subsection.
Figure 4.16 shows the Bode plot for the audio susceptibility according to
the novel expression in (4.38) together with the results presented in Figure
3.16, i.e. the simulation results and the audio susceptibility predicted by the
Ridley model. Note that, contrary to previous Bode plots, a simulation result
for the frequency 10 Hz is included. From the figure it is seen that the novel
expression agrees closely with the simulation results at high frequencies.
However, the novel expression does not agree with the simulation results at
low frequencies.
To understand the reason for the difference between the novel expression
and the simulation results at low frequencies, it is interesting to also consider
the case where the ESR of the capacitor, Rc , is set to zero. To obtain new
168 Chapter 4. A Novel Model

-20
Phase (deg); Magnitude (dB)

-40

-60
0 1 2 3 4
10 10 10 10 10

-50

-100

-150
0 1 2 3 4
10 10 10 10 10
Frequency (Hz)

Figure 4.17: The audio susceptibility of a boost converter with a current

controller ( Rc =0 Ω). X: the simulation results. Dotted line: the
novel expression. Solid line: the Ridley model.

simulation results, Ic is first adjusted manually so that the average value of the
output voltage, Vo , is equal to 8 V ( D =0.375). The Bode plot in Figure 4.17
shows the simulation results. It also shows the audio susceptibility predicted
by the novel expression and the Ridley model where the new Rc and D are
used. From the figure it is seen that the novel expression agrees closely with
the simulation results also at low frequencies.
Consider the case where Rc =0 Ω. The numerator in (4.38) can then be
written:

RTs  1 T T3  1
D '  m c D '− − s s + s s 3 + K  + , (4.40)
L   2 12 720   D'
  

and it is a series in s n where n ≥ 0 . The denominator in (4.38), i.e. (4.39), is

written:
Chapter 4. A Novel Model 169

Ts m c
L
(
D' RD' 2 + sL(1 + sRC ) + )
e sD 'Ts − 1
sD' Ts
H e (s) +
(4.41)
T RD' 3  e sD 'Ts − 1 
H e ( s )(1 + sRC ) + s 1 −  H e (s) 1 .
L  sD' Ts  sD' Ts


The following expressions are written as Taylor series:

e sD 'Ts − 1 D ' Ts D' 2 Ts2 2

=1+ s+ s + K, (4.42)
sD' Ts 2 6

sTs Ts T2
H e (s) = =1− s + s s2 + K, (4.43)
e sTs − 1 2 12

 e sD 'Ts − 1  1 1 D ' Ts D' 2 Ts2 2

1 −  =− − s− s + K. (4.44)
 sD ' T  sD' T 2 6 24
 s  s

The denominator (4.39) is therefore also a series in s n where n ≥ 0 . The

gain of the denominator approaches a specific value as the frequency
decreases. The same is true for the nominator.
Consider now the case where Rc ≥ 0 Ω. The term:

 RD'  1
1 −  (4.45)
 RD'+ Rc  sT
  s

makes the numerator in (4.38) to be a series in s n where n ≥ −1 . The term:

Ts RD'3  RD'+ Rc RD' e sD 'Ts − 1  1

 − H ( s) (4.46)
L  RD'+ R D' RD'+ R sD' T  e sD' Ts
 c c s 

makes the denominator (4.39) also to be a series in s n where n ≥ −1 . Both

the numerator and the denominator in (4.38) therefore contain an s −1 term.
From Figure 4.16, it is seen that the gain starts to increase at approximately
100 Hz as the frequency decreases and this is due to the s −1 term in the
numerator of (4.38). It is also seen that the gain stops increasing at
170 Chapter 4. A Novel Model

approximately 10 Hz and this is due to the s −1 term in the denominator of

(4.38).
In the case where Rc =0 Ω, there are no s −1 terms and the gain does not
change so much at low frequencies.

Explanation of the Comparison Results

The cause of the difference between the predictions made by the novel
expression and the simulation results is examined further in this subsection.
To investigate the accuracy of the predictions made by the novel
expression, some correction blocks are added in Figure 4.18 compared to
Figure 4.12. In Figure 4.12, the audio susceptibility and the control-to-
output transfer function of the open loop are used to determine the output
voltage, vˆo ( s ) . These two transfer functions were derived in Chapter 2 by
means of state-space averaging. The two transfer functions are not exact due
to the averaging process and this motivates the correction blocks N1 ( s ) and
N 2 ( s) in Figure 4.18. The gain of each one of the two correction blocks is
close to unity since the transfer functions derived in Chapter 2 predict the
simulation results closely.
The predictions are close to the simulation results in the sense that if the
input is a sinusoidal signal, the predictions of the amplitude and phase of the
Fourier component of the output signal with the frequency equal to the
frequency of the input signal are satisfactory (see Section 2.11). If this is the
case also for 0 Hz, the transfer functions will predict the (stationary) change
in the average value of the output signal caused by a step change in the input
signal. This was not checked in Chapter 2 but since the correction blocks
N1 ( s ) and N 2 ( s) are included in Figure 4.18, it is assumed that the change
in the average value of the output voltage, vˆo ( s ) , in Figure 4.18 caused by a
step change in the input voltage and the duty cycle will be correct.
To be able to motivate the correction block N 3 ( s) , we first need a
simulation result. We conduct a simulation with a model similar to the one
used to obtain the simulation results presented in Figure 4.17. However, the
change in the input voltage, vghat, is a step signal instead of a sinusoidal
signal. Furthermore, both the converter and the current controller are
duplicated and the duplicate uses a constant input voltage. The output
voltage from each of these two converters is recorded and the difference is
calculated. The difference in steady state is shown in Figure 4.19. The step of
the input voltage is 0.01 V.
Chapter 4. A Novel Model 171

v^ o(s)
N1(s)
^ 1(s) v^ g(s) ol
v^ g(s) v^on(s) m
1 ^ ^
Table L x(s) d(s) v^ o(s) v^ o(s)
^ 2(s) (4.21) N2(s)
4.1 v^ off(s) 1 m ^
d(s)
L
^vo(s)
N3(s)

Figure 4.18: The model of the audio susceptibility with correction blocks.

-3
x 10
8.7

8.65

8.6

8.55

8.5
vohat (V)

8.45

8.4

8.35

8.3

8.25

8.2
0 0.5 1 1.5 2 2.5 3 3.5
t (s) -5
x 10

t on t off t on t off

Figure 4.19: The difference (in steady state) in the output voltage caused by a
step change in the input voltage.
172 Chapter 4. A Novel Model

The correction block N 3 ( s) in Figure 4.18 can now be motivated as

follows. The novel expression for the audio susceptibility should be able to
predict the change in the average output voltage. The change in the voltage
across the inductor during the time the transistor is off, vˆoff (t ) , is equal to
vˆo (t ) − vˆ g (t ) according to Table 4.1. To obtain a good model, vˆoff (t )
should not be calculated by using the average value of vˆo (t ) . Instead, the
average value of vˆo (t ) as the transistor is off should be used. Note from
Figure 4.19 that these two average values are not the same and N 3 ( s) must
be used to correct this deviation.
If Figure 4.18 is used instead of Figure 4.12, the following equation
corresponds to (4.30):

RD ' (1 + sRc C )
vˆ o ( s ) = N 1 ( s)vˆ g ( s ) +
denol ( s )
Vg
(RD'+ Rc )D'
(R 2
)
D' 2 − s(R + Rc )L (1 + sRc C )
N 2 ( s) •
denol ( s ) (4.47)
e − sDTs
−e − sTs N 3 ( s )vˆ o ( s ) − vˆ g ( s ) e − sDTs
− 1 vˆ g ( s ) 
 +
 sTs L sTs L 
 .


Vg
( (
D' −1 m c D ' 1 − e − sTs ) +e − sTs
) 

 L 

By using (4.47) instead of (4.30), the corresponding equations to (4.32),

(4.33), (4.34), (4.35), (4.37), (4.38), and (4.39) have been derived. Only the
corresponding equations to (4.38), and (4.39) are presented here:

vˆ o ( s )  RTs
= D' •
vˆ g ( s )  L
   
 m c D ' N 1 ( s ) − F f ( s ) N 1 ( s ) +  N 1 ( s ) − RD ' N 2 ( s )  1 + (4.48)
  RD'+ Rc  sT 
   s 
R + Rc  (1 + sRc C )
N 2 ( s )  ,
RD'+ Rc  den( s )

where
Chapter 4. A Novel Model 173

den( s ) =
Ts m c   RD '+ Rc  
D '  RD '  + sRc C  + sL(1 + s(R + Rc )C ) −
 
L   R + Rc  
 R 2 D ' 3 Rc C D ' (R + Rc )  e sD 'Ts − 1

 L(RD '+ Rc )
− (1 + sRc C )
 sD ' Ts
•
 RD'+ Rc 
H e (s) N 2 ( s) N 3 (s) +
(4.49)
 RD' 
H e ( s )1 + Rc C + s (R + Rc )C  +
 L 
Ts RD'3  RD'+ Rc RD ' e sD 'Ts − 1 
− N 2 (s) N 3 (s)  •

L  RD'+ Rc D ' RD'+ Rc sD' Ts 

1
H e (s) .
sD' Ts

If N1 ( s ) , N 2 ( s) and N 3 ( s) are equal to unity, (4.48) is the same as (4.38)

and (4.49) is the same as (4.39).
Consider the case where Rc =0 Ω. Then (4.48) and (4.49) can be
simplified:

vˆ o ( s )
=
vˆ g ( s )
RTs  1  1
D '  m c D' N 1 ( s ) − F f ( s ) N 1 ( s ) + (N 1 ( s) − N 2 ( s ) ) +
 N 2 ( s ) (4.50)
L  sT s  D'
den( s )

and
174 Chapter 4. A Novel Model

den( s ) =
Ts m c
L
(
D ' RD ' 2 + sL(1 + sRC ) + )
e sD 'Ts − 1
H e ( s ) N 2 ( s ) N 3 ( s) + H e ( s)(1 + sRC ) + (4.51)
sD ' Ts
Ts RD'3  e sD 'Ts − 1  1
1 − N 2 ( s ) N 3 ( s )  H e ( s) .
L  sD ' Ts  sD' Ts
 

The dc gain predicted by the model in Figure 4.18 is found by letting s tend
to zero in (4.50) and (4.51). If N1 (0) , N 2 (0) and N 3 (0) are equal to
unity, the dc gain is calculated by using (4.36), (4.42), (4.43), and (4.44):

RTs  1 1
D '  mc D '−  +
ˆv o ( s ) L  2  D'
lim = 3
. (4.52)
ˆv g ( s ) Ts mc T RD '
s→0 RD '3 +2 − s
L 2L

The dc gain predicted by the Ridley model, (3.94), is exactly the same as
(4.52). If N1 (0) and N 3 (0) are equal to unity and N 2 (0) is not equal to
unity, the numerator and the denominator must be multiplied with s before
the dc gain is calculated:

RTs 1
D' (1 − N 2 (0) )
vˆ ( s ) L Ts
1
lim o = 3
= . (4.53)
vˆ ( s ) Ts RD' D'
s→0 g (1 − N 2 (0)) 1
L D ' Ts

If N 2 (0) and N 3 (0) are equal to unity and N1 (0) is not equal to unity, the
absolute value of the gain tend to infinity as s tend to zero. If N1 (0) and
N 2 (0) are equal to unity and N 3 (0) is not equal to unity, the gain tend to
zero as s tend to zero. This discussion shows that the predictions made by
the model of the gain at low frequencies is very sensitive to errors in some
parts of the model. From Figure 4.17, it was seen that the model in Figure
4.12 predicts the simulation results closely at low frequencies in the case
where Rc =0 Ω. The conclusion is that this is due to more good luck than
good management.
Chapter 4. A Novel Model 175

From Figure 4.16, it was seen that the model in Figure 4.12 does not
predict the simulation results closely at low frequencies in the case where
Rc =14 mΩ. This may be due to absence of good luck. From Figure 4.16, it
was also seen that the Ridley model predicts the simulation results closely at
low frequencies in the case where Rc =14 mΩ. This seems a little surprising
since Ridley uses an approximate model of the PWM switch when the
feedforward gains are derived. In Vorperian (1990), the PWM switch model
depends on Rc but Ridley uses this model as if Rc =0 Ω.

4.5 Audio Susceptibility of the Buck-Boost

Converter
In this section, the novel model derived in Section 4.2 is applied to the
buck-boost converter. This is made in a way similar to the one used for the
boost converter. The obtained expression is compared with simulation results
and the Ridley model.

A Novel Expression

The audio susceptibility of the buck-boost converter with current-mode

control is obtained by using Figure 4.12, Table 4.1, (4.21), (2.188), (2.186),
and (3.104):

RDD' (1 + sRc C )
vˆ o ( s ) = vˆ g ( s ) +
denol ( s )
V g (R + R c )
(RD'+ Rc )D'
(RD' 2
)
− sLD (1 + sRc C )
•
denol ( s ) (4.54)
e − sDTs
−e − sTs
vˆ o ( s ) e − 1 vˆ g ( s) 
− sDTs
 +
 sTs L sTs L 
 ,


V g −1
(
D ' m c D' 1 − e − sTs
(
+e − sTs
) 
 )
 L 

where
176 Chapter 4. A Novel Model

denol ( s ) =
RD' (RD'+ Rc ) (R + Rc ) + s (L + RRc CD ') + s 2 (R + Rc )LC = (4.55)
RD' ((RD '+ Rc ) (R + Rc ) + sRc C ) + sL(1 + s (R + Rc )C ) ,

(4.54) is rewritten:

denol ( s )
Vg
LD '
(m D' (1 − e )+ e )vˆ
c
− sTs − sTs
o (s) =

Vg
LD '
(m D' (1 − e )+ e )RDD' (1 + sR C )vˆ
c
− sTs − sTs
c g (s) +

V g (R + R c ) (4.56)
(RD'+ Rc )D'
(RD' 2
)
− sLD (1 + sRc C ) •

 e − sDTs − e − sTs vˆ o ( s ) e − sDTs − 1 vˆ g ( s ) 

 + .
 sT L sT L 
 s s 

(4.56) is rewritten:

vˆ o ( s )  V g
=
vˆ g ( s )  LD '
( ( ) )
m c D' 1 − e − sTs + e − sTs RDD' (1 + sRc C ) +

V g (R + R c )
(RD'+ Rc )D'
( )
RD' 2 − sLD (1 + sRc C ) •

 e − sDTs − 1  
  •
 sTs L   (4.57)
 
 V
( (
 denol ( s ) g m c D' 1 − e − sTs + e − sTs −
 LD '
) )

−1
V g (R + R c ) e − sDTs − e − sTs 
(RD'+ Rc )D'
(RD' 2
)
− sLD (1 + sRc C )
sTs L


.


The following is obtained if the numerator and denominator in (4.57) are

( (
multiplied with D' Ts V g 1 − e − sTs : ))
Chapter 4. A Novel Model 177

vˆ o ( s )  Ts  − sTs 
=  m c D'+ e  RDD' (1 + sRc C ) +
vˆ g ( s )  L  1 − e − sTs 
 
Ts (R + Rc ) e − sDTs − 1 
RD '+ Rc
( )
RD ' 2 − sLD (1 + sRc C )
1
sTs L 1 − e − sTs
•


(4.58)
  − sTs 
 den ( s ) Ts  m c D '+ e −
 ol
L  1 − e − sTs 
  
−1
Ts (R + Rc ) e − sDTs − e − sTs 
( )
RD ' 2 − sLD (1 + sRc C )
1 

RD '+ Rc sTs L 1 − e − sTs 

The numerator in (4.58) is rewritten:

178 Chapter 4. A Novel Model

Ts  − sTs 
 m c D'+ e  RDD' (1 + sRc C ) +
L  1 − e − sTs 
 
Ts (R + Rc ) e − sDTs − 1
RD '+ Rc
( )
RD ' 2 − sLD (1 + sRc C )
1
sTs L 1 − e − sTs
=

RTs 
D '  m c D '+
e − sTs
+
( )
(R + Rc ) RD' 2 − sLD e − sDTs − 1 1 
•
L 
 1 − e − sTs RDD ' (RD'+ Rc ) sTs 1 − e − sTs 

D(1 + sRc C ) =
RTs 
D '  m c D '+
e − sTs
+
(
(R + Rc ) RD' 2 e − sDTs − 1 1) +
L 
 1 − e − sTs RDD' (RD '+ Rc ) sTs 1 − e − sTs
(R + Rc )(− sLD ) e − sDT − 1 1
s 
 D (1 + sRc C ) =
RDD' (RD '+ Rc ) sTs 1 − e − sT s 

 RTs

 1 D' (R + Rc ) 1 − e − sDTs 1 
D'  mc D '+ sT − +
 L  e −1 RD'+ Rc sDTs 1 − e − sTs 
  
s

(4.59)
RTs (R + Rc )sL 1 − e − sDTs 1 
 D (1 + sRc C ) =
D'
L RD' (RD '+ Rc ) sTs 1 − e − sTs 

 RTs

  D ' (R + Rc )  1
D'  mc D '+1 − 
 sTs −
 L  + −
   RD ' R c  e 1

D' (R + Rc )  1 − e − sDTs 1 1  
 − sT +
RD '+ Rc  sDTs 1 − e s e s − 1  
 − sT

D' (R + Rc ) D 1 − e − sDTs 
 D(1 + sRc C ) =
(
RD '+ Rc D ' D 1 − e − sTs ) 

 RTs

 D ' (R + Rc )  D ' (R + Rc )  1 
D'  mc D '− F f 1 ( s ) + 1 −  H e ( s ) +
 L  RD '+ Rc RD '+ Rc  
   sTs 
D ' (R + Rc ) D 
F f 2 ( s)  D(1 + sRc C ) ,
RD '+ Rc D ' 

where
Chapter 4. A Novel Model 179

1  sTs 1 − e − sDTs sT 
F f 1 ( s) =  − sT s  =
sTs  1 − e s sDT
− sT
e s − 1 
 s
(4.60)
 D  (3 − 2 D )DTs
1 −  − s−
(
1 − 2 D + D 2 DTs2 2 )
s + K,
 2 12 24

F f 2 (s) =
1 1 − e − sDTs
= 1 +
D ' Ts
s +
( )
1 − 3D + 2 D 2 Ts2 2
s + K. (4.61)
D 1 − e − sTs 2 12

and H e (s ) is the same as in (3.10). The Taylor series of F f 1 ( s) and F f 2 ( s )

are also shown in (4.60) and (4.61). Note that F f 1 ( s) in (4.60) is exactly the
same as F f (s ) in (4.27). The denominator in (4.58) is rewritten:
180 Chapter 4. A Novel Model

Ts  − sTs 
denol ( s )  m c D '+ e −
L  1 − e − sTs 
 
T s (R + R c ) e − sDTs − e − sTs
( )
RD ' 2 − sLD (1 + sRc C )
1
=
RD '+ Rc sT s L 1 − e − sTs
T T 1
denol ( s ) s m c D '+ denol ( s ) s sT −
L L e s −1
T s (R + R c ) e sTs e − sDTs − 1 1
RD '+ Rc
( )
RD ' 2 − sLD (1 + sRc C ) =
sTs L e sTs − 1
T
denol ( s ) s m c D '+
L
  RD'+ Rc   Ts 1
 R + R + sRc C  + sL(1 + s (R + Rc )C ) L sTs
 RD '    −
 e −1
  c  
Ts D ' (R + R c )
L(RD '+ Rc )
( )
RD' 2 − sLD (1 + sRc C )
e sD 'Ts − 1 1
sD' Ts e sTs − 1
=

Ts T RD ' RD'+ Rc 1
denol ( s ) m c D '+ s + (4.62)
L L R + Rc e s − 1
sT

RD' sT sT
Rc C sT s + (1 + s (R + Rc )C ) sT s −
L e −1s
e s −1
 T s D ' (R + R c ) e sD 'Ts − 1 1
 RD' 2 +
 L(RD'+ R ) sD' Ts e sTs − 1
 c

D ' (R + Rc ) e sD 'Ts − 1 sTs

RD' 2 Rc C −
L(RD'+ Rc ) sD ' Ts e sTs − 1
D ' (R + Rc ) e sD 'Ts − 1 sTs 
D(1 + sRc C ) =
RD '+ Rc sD' Ts e sTs − 1 
Ts  RD ' 
denol ( s ) m c D '+1 + Rc C + s (R + Rc )C  H e ( s ) −
L  L 
 D ' (R + Rc ) D ' (R + R c )  e sD 'Ts − 1
 RD' 2 Rc C − D(1 + sRc C ) H e (s) +
 L(RD'+ Rc ) RD'+ Rc  sD' Ts
 Ts RD ' RD'+ Rc Ts D ' (R + Rc ) e sD 'Ts − 1  1
 − RD' 2 ,
 L
 R + Rc L(RD'+ Rc ) sD ' Ts  e sTs − 1
Chapter 4. A Novel Model 181

where H e (s ) is the same as in (3.10). The novel expression for the audio
susceptibility of the buck-boost converter with current-mode control is
obtained by using (4.59) and (4.62) to rewrite (4.58):

vˆ o ( s )
=
vˆ g ( s )
 RTs

 D ' (R + Rc )  D ' (R + R c )  1 
D'  mc D '− F f 1 ( s ) + 1 −  H e (s)

+ (4.63)
 L  RD '+ Rc RD'+ Rc  
   sTs 
D ' (R + Rc ) D  D(1 + sRc C )
F f 2 ( s )  ,
RD '+ Rc D '  den( s)

where

den( s ) =
Ts   RD'+ Rc  
m c D'  RD '  + sRc C  + sL(1 + s (R + Rc )C ) −
 
L   R + Rc  
 (R + Rc )RD' 3 Rc C (R + Rc )DD'  e sD 'Ts − 1
 − (1 + sRc C ) H e (s) +

 L(RD'+ Rc ) RD'+ Rc  sD' Ts
 (4.64)
 RD' 
H e ( s )1 + Rc C + s (R + Rc )C  +
 L 
Ts RD ' 3  RD '+ Rc

(R + Rc )D' e sD'Ts − 1  1
− H (s) ,
L  RD'+ R D' RD '+ R sD ' T  e sD ' Ts
 c c s 

F f 1 ( s) is defined in (4.60), F f 2 ( s ) is defined in (4.61), and H e (s ) is

defined in (3.10).
If (4.63) and (4.64) are compared with (3.112) and (3.113), i.e. the
Ridley model, it is seen that there are some differences both in the numerator
and in the denominator.
182 Chapter 4. A Novel Model

-10

-20
Phase (deg); Magnitude (dB)

-30

-40
0 1 2 3 4
10 10 10 10 10

-20

-40

-60

-80

-100
0 1 2 3 4
10 10 10 10 10
Frequency (Hz)

Figure 4.20: The audio susceptibility of a buck-boost converter with a current

controller. X: the simulation results. Dotted line: the novel
expression. Solid line: the Ridley model.

A Comparison of the Novel Expression and the Simulation

Results

Figure 4.20 shows the Bode plot for the audio susceptibility according to
the novel expression in (4.63) together with the results presented in Figure
3.19, i.e. the simulation results and the audio susceptibility predicted by the
Ridley model. From the figure it is seen that the novel expression agrees
closely with the simulation results at high frequencies. However, the novel
expression does not agree with the simulation results at low frequencies. Note
that this is the same conclusion as for the novel expression for the Boost
converter (see Section 4.5).
Chapter 4. A Novel Model 183

4.6 Summary and Concluding Remarks

A novel model for the audio susceptibility of converters with current-
mode control was derived in this chapter. This was made by considering how
the duty cycle depends on changes in the input and output voltages. The
novel model was applied to the buck converter to obtain a novel expression
for the audio susceptibility. The novel expression was compared with
simulation results and the Ridley and Tan models. We concluded that the
novel expression agrees closely with the simulation results also at high
frequencies. The Ridley and Tan models do not agree closely with the
simulation results at high frequencies due to modeling errors. The ramp that
is used for slope compensation in current-mode control can be chosen such
that the audio susceptibility is very small at dc for the buck converter. The
predictions of the simulation results made by the Ridley and Tan models are
worse for this case compared to other cases.
The novel model was also applied to the boost converter. The novel
expression was compared with simulation results and the Ridley model. We
concluded that the novel expression agrees closely with the simulation results
at high frequencies, contrary to the Ridley model. However, the novel
expression does not agree with the simulation results at low frequencies but
the Ridley model does. In the case where the ESR of the capacitor, Rc , is
zero, the novel expression agrees closely with the simulation results also at low
frequencies. An attempt to explain the predictions made by the novel
expression for low frequencies was made. We saw that these predictions are
very sensitive to errors in some parts of the derived model. The good result
for Rc =0 Ω is therefore due to more good luck than good management.
The novel model was also applied to the buck-boost converter. The novel
expression was compared with simulation results and the Ridley model. We
concluded that the novel expression agrees closely with the simulation results
at high frequencies, contrary to the Ridley model. However, the novel
expression does not agree with the simulation results at low frequencies but
the Ridley model does. Thus, the novel expressions for the boost and buck-
boost converters have similar properties.
 Chapter 5 Improved Models

Models for converters with current-mode control were considered in

Chapter 3. We showed that the way the changes in the input and output
voltages are treated in the Ridley and Tan models introduces a modeling error
at high frequencies. We also showed that this modeling error is significant for
the audio susceptibility. A novel model for the audio susceptibility was
derived in Chapter 4. In this chapter, this model is utilized to improve the
Ridley and Tan models. In Chapter 6, the improved Ridley model will be
approximated and then, in Chapter 7, used to analyze some properties that
can be obtained when load current measurements are utilized for control.

5.1 Chapter Survey

The expressions for the audio susceptibility obtained by applying the
Ridley and Tan models to the buck converter are improved in Section 5.2.
The expressions for the audio susceptibility obtained by applying the Ridley
model to the boost and buck-boost converters are improved in Section 5.3
and Section 5.4, respectively. A summary and concluding remarks are
presented in Section 5.5.

5.2 Improved Expressions for the Buck Converter

The expression for the audio susceptibility obtained by applying the
Ridley model to the buck converter is (see (3.58), (3.62), and (3.33))

RTs  L 
D mc D'+ k f  (1 + sRc C )
vˆo ( s) L  DTs Ri  (5.1)
= ,
ˆv g ( s ) den( s )

185
186 Chapter 5. Improved Models

where

RTs
den( s ) = (1 + s(R + Rc )C )Fh−1 ( s ) + (mc D'−0.5)(1 + sRc C ) , (5.2)
L

DTs Ri  D 
kf = − 1 −  , (5.3)
L  2

and Fh (s) is defined in (3.19).

The expression for the audio susceptibility obtained by applying the Tan
model to the buck converter is (see (3.68), (3.72), and (3.23))

RTs  L 1 2  
D mc D '+ k f − 1 − s   (1 + sRc C )
vˆo ( s) L  DTs 2 πω n  (5.4)
= ,
vˆ g ( s ) den( s )

where

den( s ) =
RTs    (5.5)
(1 + s(R + Rc )C )Fh−1 (s) +  mc D '−0.51 − s 2   (1 + sRc C ) ,
  πω n 
L   

DD' Ts
kf =− , (5.6)
2L

and Fh (s) is defined in (3.19).

The novel expression for the audio susceptibility obtained by applying the
novel model derived in Chapter 4 to the buck converter is (see (4.26) and
(4.28))

ˆv o ( s )
RTs
( )
D m c D '− F f ( s ) (1 + sRc C )
(5.7)
= L ,
vˆ g ( s ) den( s )

where
Chapter 5. Improved Models 187

den( s ) = (1 + s(R + Rc )C )(H e ( s ) + sTs mc D ') +

RTs  1 − H e (s)  (5.8)
 m c D'−  (1 + sRc C ) ,
 
L  sTs 

F f (s ) is defined in (4.27), and H e (s ) is defined in (3.10).

In Section 4.3, it was concluded that the three expressions for the audio
susceptibility presented above have approximately the same denominator but
three different numerators. One way to improve the Ridley and Tan models
is to modify the numerators in (5.1) and (5.4) in some way so that they are
equal to the numerator in (5.7). The feedforward gain k f is not present in
the denominators (5.2) and (5.5) and may therefore be used to modify the
numerators in (5.1) and (5.4) without changing the denominators. Note that
predictions of the control-to-output transfer function and the output
impedance made by the Ridley and Tan models are not affected when k f is
changed. This parameter is the gain of the feedforward of the input voltage,
which is constant in the case where the control-to-output transfer function
and the output impedance are considered. Note that modifying only k f may
not be the most suitable solution. This will be discussed in Section 5.5.
The improvement of the Ridley model is first considered. The following
equation is obtained if the numerator in (5.1) is put equal to the numerator
in (5.7):

RTs  L 
D mc D '+ k f ( s )  (1 + sRc C ) =
L  DTs Ri  (5.9)
RTs
L
( )
D mc D '− F f ( s ) (1 + sRc C ) .

The new k f depends on s and this is prepared for in (5.9). The transfer
function k f (s ) is calculated by using (5.9):

DTs Ri
k f (s) = − F f ( s) . (5.10)
L

where F f (s ) is defined in (4.27). Note that the s 0 term in (5.10) is equal to

(5.3).
188 Chapter 5. Improved Models

The improvement of the Tan model is now considered. The following

equation is obtained if the numerator in (5.4) is put equal to the numerator
in (5.7):

RTs  L 1 2 
D  mc D '+ k f ( s ) − 1 − s  (1 + sRc C ) =
 2 πω n 
L  DTs  (5.11)
RTs
L
( )
D m c D '− F f ( s) (1 + sRc C ) .

k f (s ) is found by rearranging (5.11):

DTs   
k f (s) = −  F f ( s ) − 1 1 − s 2   . (5.12)
L 
 2  πω n 


where F f (s ) is defined in (4.27). Note that the s 0 term in (5.12) is equal to

(5.6).

5.3 Improved Expression for the Boost Converter

The expression for the audio susceptibility obtained by applying the
Ridley model to the boost converter is improved in this section. The
methodology is slightly different compared to the one used for the buck
converter since the novel expression exhibits poor low-frequency predictions.
A combined expression is first derived and this expression is then used to
derive a new feedforward gain k f (s ) .
The expression for the audio susceptibility obtained by applying the
Ridley model to the boost converter is (see (3.94) and (3.95))

RTs  RD'+ Rc  R + Rc  2 
D'  mc − 0 .5  + 1 + s 
 
   ωn
2
vˆo ( s) L R RD '  (5.13)
= ,
vˆ g ( s ) (1 + sRc C )−1 den(s)
where
Chapter 5. Improved Models 189

den( s ) =
Ts mc RD'+ Rc   RD '+ Rc  
 R + R + sRc C  + sL(1 + s(R + Rc )C ) −
 RD '   

L R   c  
RD '3 Ts (R + R )L
(1 + sRc C ) 1 − s 2 c2  + (5.14)
2L  R D' 
 2 RD'+ Rc RD'+ Rc 
H e ( s ) + s(R + Rc )C ,
 RD' RD ' 

and H e (s ) is defined in (3.35).

The novel expression for the audio susceptibility obtained by applying the
novel model derived in Chapter 4 to the boost converter is (see (4.38) and
(4.39))

RTs   RD '  1  R+R

+
D'  m c D'− F f ( s) + 1 − 
 sT  RD'+ R
c
vˆ o ( s ) L   RD'+ Rc  s  c (5.15)
= ,
vˆ g ( s ) (1 + sRc C )−1 den( s)
where

den( s ) =
Ts m c   RD '+ Rc  
D '  RD '  + sRc C  + sL(1 + s(R + Rc )C ) −
 
L   R + Rc  
 R D ' Rc C (R + Rc )D '
2 3  e sD 'Ts − 1

 L(RD '+ Rc )
− (1 + sRc C )
 sD' Ts
H e (s) +
 RD '+ Rc  (5.16)
 RD' 
H e ( s )1 + Rc C + s (R + Rc )C  +
 L 
Ts RD'3  RD'+ Rc RD ' e sD 'Ts − 1  1
 − H (s) ,
L  RD'+ R D ' RD'+ R sD' T  e sD ' T
 c c s  s

F f (s ) is defined in (4.36), and H e (s ) is defined in (3.10).

In Section 4.4, it was concluded that the expression (5.13) agrees closely
190 Chapter 5. Improved Models

50

40

30

20
Phase (deg); Magnitude (dB)

10

0
0 1 2 3 4
10 10 10 10 10

150

100

50

-50

-100
0 1 2 3 4
10 10 10 10 10
Frequency (Hz)

Figure 5.1: The denominator in the audio susceptibility of a boost converter

with a current controller ( Rc =14 mΩ). Dotted line: the novel
expression. Solid line: the Ridley model.

with the simulation results at low frequencies but not at high frequencies. It
was also concluded that the novel expression (5.15) agrees closely with the
simulation results at high frequencies but not at low frequencies. These two
expressions will now be combined to an expression that agrees closely with
the simulation results both at low and high frequencies.
The Bode plots for the denominators in (5.13) and (5.15) are shown in
Figure 5.1. The parameter values shown in Table 2.5 are used. The two
denominators agree closely at high frequencies. The maximum difference is
approximately 0.3 dB and 2 degrees in the frequency interval 500 - 25000
Hz. (5.15) is not good at low frequencies in the case where Rc =14 mΩ.
Figure 5.2 shows the same as Figure 5.1 except Rc is changed from 14 mΩ
to 0 Ω. In this case, the two denominators agree closely in the whole
presented frequency interval. The maximum difference is approximately 0.1
dB and 2 degrees.
This discussion shows that it is reasonable to replace the denominator in
Chapter 5. Improved Models 191

50

40

30

20
Phase (deg); Magnitude (dB)

10

0
0 1 2 3 4
10 10 10 10 10

200

150

100

50

0
0 1 2 3 4
10 10 10 10 10
Frequency (Hz)

Figure 5.2: The denominator in the audio susceptibility of a boost converter

with a current controller ( Rc =0 Ω). Dotted line: the novel
expression. Solid line: the Ridley model.

(5.15) with the denominator in (5.13). After this approximation, it is easy to

find a good combined expression. To obtain a good dc gain, the s 0 term in
the numerator in (5.13) is used in the combined expression. The s 0 term in
the numerator in (5.13) is obtained by removing the s n terms where n ≥ 1 .
The s n terms, where n ≥ 1 , in the numerator of the combined expression are
fetched from the numerator in (5.15). These are obtained by removing the
s 0 and s −1 terms from the numerator in (5.15). The s 0 term in the
numerator in (5.13) is

RTs  RD '+ Rc  R + Rc
D '  mc − 0 .5  + . (5.17)
L  R  RD '

The numerator in (5.15) excluding the s −1 term is equal to

192 Chapter 5. Improved Models

R + Rc
RTs
L
(
D ' m c D '− F f ( s ) +)RD '+ Rc
. (5.18)

By using (4.36), the following is obtained if the s 0 term is removed from

(5.18):

RTs   1 
D '  −  F f ( s ) −   . (5.19)
L   2 

The numerator in the combined expression is now obtained by adding (5.17)

and (5.19):

RTs  RD '+ Rc  R + Rc RTs

D '  mc − 0 .5  + − D ' F f ( s ) − 0 .5 = ( )
L  R  RD' L
(5.20)
RTs  RD '+ Rc  R + Rc
D '  mc − F f ( s)  + .
L  R  RD'

By using (5.20), the combined expression for the audio susceptibility is

obtained:

RTs  RD'+ Rc  R + Rc
D'  mc − F f (s)  +
vˆo ( s) L  R  RD' (5.21)
= ,
vˆ g ( s ) (1 + sRc C )−1
den( s)

where den(s) is defined in (5.14) and F f (s ) is defined in (4.36).

Figure 5.3 shows the Bode plot for the audio susceptibility according to
the combined expression in (5.21) together with the results presented in
Figure 4.16, i.e. the simulation results and the audio susceptibilities predicted
by the novel expression, (5.15), and the Ridley model, (5.13). From the
figure it is seen that the combined expression agrees closely with the
simulation results both at low and high frequencies.
(5.13) is the expression for the audio susceptibility obtained by applying
the Ridley model to the boost converter. This expression is written as follows
(see (3.91) and (3.84)):
Chapter 5. Improved Models 193

-20
Phase (deg); Magnitude (dB)

-40

-60
0 1 2 3 4
10 10 10 10 10

-50

-100

-150
0 1 2 3 4
10 10 10 10 10
Frequency (Hz)

Figure 5.3: The audio susceptibility of a boost converter with a current

controller. X: the simulation results. Dashed line: the combined
expression. Dotted line: the novel expression. Solid line: the
Ridley model. Note that the dashed line almost coincides with
the solid line at low frequencies.

vˆo ( s)  RTs  RD'+ Rc L  R + Rc 

1 + s
2 
−
= D '  mc + k f  +
vˆ g ( s )  L  ω2 
  R T s Ri  RD'  n 
(5.22)
T  R  1 L  1
s s 1 + c   + k f   ,
D'  R   2 T s Ri  (1 + sR C )−1 den( s)
 c

where

T s Ri
kf =− , (5.23)
2L

and den(s) is defined in (5.14).

194 Chapter 5. Improved Models

The combined expression will now be used to improve the Ridley model.
Note that the denominator in (5.21) and (5.22) are exactly the same. One
way to improve the Ridley model is to modify the numerator in (5.22) in
some way so that it is equal to the numerator in (5.21). The feedforward gain
k f is not present in the denominator in (5.22) and may therefore be used to
modify the numerator in (5.22) without changing the denominator. The
following equation is obtained if the numerator in (5.22) is put equal to the
numerator in (5.21):

RTs  RD '+ Rc L  R + Rc  2 
D '  mc + k f ( s )  + 1 + s −
 ω2 
L  R T s Ri  RD '  n 
Ts  R  1 L 
s 1 + c   + k f ( s )  = (5.24)
D'  R   2 T s Ri 
RTs  RD '+ Rc  R + Rc
D '  mc − F f (s)  + .
L  R  RD'

The new k f depends on s and this is prepared for in (5.24). k f (s ) is

calculated by using (5.24):

RTs  L  R + Rc  s2 
D'  k f ( s )  +  −
ω 2 
L  T s Ri  RD'  n 
T  R  1 L  (5.25)
s s 1 + c   + k f ( s )  =
D'  R   2 T s Ri 
RTs
L
( )
D' − F f ( s ) ,
Chapter 5. Improved Models 195

Ri  RTs R + Rc s 2 Ts  Rc  1 
− D ' F ( s ) − + s  1 + 
RD'  R  2 
f
L RD ' ω n2 D' 
k f (s) = =
Ri  RTs L Ts  Rc  L 
 D' − s 1 +  
RD '  L Ts Ri D'  R  Ts Ri 
Ts Ri (R + Rc )Ri s 2 T R  R 
− F f (s) − + s s i 2 1 + c 
L 2
R D' 2
ωn2
2 RD '  R 
=
L  Rc 
1− s 1 + 
RD' 2  R 
(5.26)
2 L R + Rc s 2 L  R 
2 F f (s) + −s 1 + c 
T R Ts R D ' ω n
2 2 2
RD ' 
2 R 
− s i =
2L L  Rc 
1− s 1 + 
RD ' 2  R 
L  2 s   Rc 
2F f (s) − s  1 −   1 + 
Ts Ri RD' 2  Ts ω n2   R 
− .
2L L  Rc 
1− s 1 + 
RD' 2  R 

(5.26) is rewritten by using (3.13):

L  2   Rc 
2F f (s) − s  1 − s  1 + 
T s Ri RD' 2  πω n   R 
k f (s) = − , (5.27)
2L L  Rc 
1− s 1 + 
RD ' 2  R 

where F f (s ) is defined in (4.36). Note that the s 0 term in (5.27) is equal to

(5.23). Note also that k f (s ) is an unstable transfer function since its pole is
in the right half plane.
196 Chapter 5. Improved Models

5.4 Improved Expression for the Buck-Boost

Converter
The expression for the audio susceptibility obtained by applying the
Ridley model to the buck-boost converter is improved in this section. The
methodology is analogous to the one used for the boost converter.
The expression for the audio susceptibility obtained by applying the
Ridley model to the buck-boost converter is (see (3.112) and (3.113))

RTs  RD'+ Rc  D  D  s2 
 + s Ts D
D '  mc − 1 −   + 1 + 2 
vˆo ( s) L  R + Rc  2   D'  ω n  2 (5.28)
= ,
vˆ g ( s ) D −1
(1 + sRc C ) −1
den( s )

where

den( s ) =
Ts mc RD'+ Rc   RD '+ Rc  
 RD ' 
 + sRc C  + sL(1 + s(R + Rc )C ) −
L R + Rc   
 R + Rc  
RD '3 Ts
(1 + sRc C ) 1 − s LD2  + (5.29)
2L  RD' 
 RD ' (1 + D ) + Rc RD'+ Rc 
H e ( s ) + sC .

 (R + Rc )D' D' 

and H e (s ) is defined in (3.35).

The novel expression for the audio susceptibility obtained by applying the
novel model derived in Chapter 4 to the buck-boost converter is (see (4.63)
and (4.64))
Chapter 5. Improved Models 197

vˆ o ( s )
=
vˆ g ( s )
 RTs

 D ' (R + Rc )  D ' (R + R c )  1 
D'  mc D '− F f 1 ( s ) + 1 −  H e ( s)

+ (5.30)
 L  + + 
  RD ' R c  RD ' R c  sT s 
D ' (R + Rc ) D  1
F f 2 ( s )  ,
 D (1 + sRc C ) den( s )
RD '+ Rc D ' −1 −1

where

den( s ) =
Ts   RD'+ Rc  
m c D'  RD '  + sRc C  + sL(1 + s (R + Rc )C ) −
 
L   R + Rc  
 (R + Rc )RD' 3 Rc C (R + Rc )DD'  e sD 'Ts − 1
 − (1 + sRc C ) H e (s) +

 L(RD'+ Rc ) RD'+ Rc  sD' Ts
 (5.31)
 RD' 
H e ( s )1 + Rc C + s (R + Rc )C  +
 L 
Ts RD ' 3  RD '+ Rc

(R + Rc )D' e sD'Ts − 1  1
− H e (s) ,
L  RD'+ R D' RD'+ Rc sD' Ts   sD' Ts
 c

F f 1 ( s) is defined in (4.60), F f 2 ( s ) is defined in (4.61), and H e (s ) is

defined in (3.10).
In Section 4.5, it was concluded that the expression (5.28) agrees closely
with the simulation results at low frequencies but not at high frequencies. It
was also concluded that the novel expression (5.30) agrees closely with the
simulation results at high frequencies but not at low frequencies. These two
expressions will now be combined to an expression that agrees closely with
the simulation results both at low and high frequencies.
The Bode plots for the denominators in (5.28) and (5.30) are shown in
Figure 5.4. The parameter values shown in Table 2.6 are used. The two
denominators agree closely at high frequencies. The maximum difference is
approximately 0.4 dB and 4 degrees in the frequency interval 250 - 25000
Hz. (5.30) is not good at low frequencies in the case where Rc =14 mΩ.
Figure 5.5 shows the same as Figure 5.4 except Rc is changed from 14 mΩ
198 Chapter 5. Improved Models

40

30

20
Phase (deg); Magnitude (dB)

10

0
0 1 2 3 4
10 10 10 10 10

150

100

50

-50

-100
0 1 2 3 4
10 10 10 10 10
Frequency (Hz)

Figure 5.4: The denominator in the audio susceptibility of a buck-boost

converter with a current controller ( Rc =14 mΩ). Dotted line:
the novel expression. Solid line: the Ridley model.

to 0 Ω. In this case, the two denominators agree closely in the whole

presented frequency interval. The maximum difference is approximately 0.3
dB and 3 degrees.
This discussion shows that it is reasonable to replace the denominator in
(5.30) with the denominator in (5.28). After this approximation, it is easy to
find a good combined expression. To obtain a dc gain that agrees closely with
the simulation results, the s 0 term in the numerator in (5.28) is used in the
combined expression. The s 0 term in the numerator in (5.28) is obtained by
removing the s n terms where n ≥ 1 . The s n terms, where n ≥ 1 , in the
numerator of the combined expression are fetched from the numerator in
(5.30). These are obtained by removing the s 0 and s −1 terms from the
numerator in (5.30). The s 0 term in the numerator in (5.28) is

RTs  RD '+ Rc  D  D
D '  mc − 1 −   + . (5.32)
L  R + Rc  2   D'
Chapter 5. Improved Models 199

40

30

20
Phase (deg); Magnitude (dB)

10

0
0 1 2 3 4
10 10 10 10 10

200

150

100

50

0
0 1 2 3 4
10 10 10 10 10
Frequency (Hz)

Figure 5.5: The denominator in the audio susceptibility of a buck-boost

converter with a current controller ( Rc =0 Ω). Dotted line: the
novel expression. Solid line: the Ridley model.

By using (4.43), the numerator in (5.30) excluding the s −1 term is obtained:

RTs  D ' (R + Rc )
D'  mc D '− F f 1 (s) +
L  RD '+ Rc
 D ' (R + Rc )  1 
1 − (H e ( s ) − 1) + (5.33)
 RD'+ Rc  
 sTs 
D ' (R + Rc ) D
F f 2 (s) .
RD '+ Rc D '

By using (4.43), (4.60), and (4.61), the following is obtained if the s 0 term is
removed from (5.33):
200 Chapter 5. Improved Models

RTs  D' (R + Rc )   D 
D'  −  F f 1 ( s ) − 1 −   +
L  RD'+ Rc   2 
 D ' (R + Rc )    T   1 
1 −   H e ( s ) − 1 − s s   + (5.34)
 RD'+ Rc   2   sTs 
 
D ' (R + Rc ) D
RD'+ Rc D '
( )
F f 2 ( s) − 1 .

The numerator in the combined expression is now obtained by adding (5.32)

and (5.34):

RTs  RD'+ Rc  D  D
D'  mc − 1 −   + +
L  R + Rc  2   D'
RTs  D' (R + Rc )   D 
D'  −  F f 1 ( s ) − 1 −   +
L  RD ' + R c   2 
(5.35)
 D ' (R + Rc )    T   1 
1 −   H e ( s ) − 1 − s s   +
 RD'+ Rc     2   sTs 
 
D ' (R + Rc ) D
RD'+ Rc D '
( )
F f 2 ( s) − 1 .

One part of (5.35) can be rewritten as follows:

 D  D ' (R + R c )   D 
− 1 −  −  F f 1 ( s ) − 1 −   =
 2 RD'+ Rc   2 
D ' (R + R c )  RD '+ Rc  D 
 F f 1 ( s ) − 1 −  +
D
− 1 −   =
RD'+ Rc   2  D ' ( R + R )
c  2 
(5.36)
D ' (R + R c )   RD '+ Rc  D 
−  F f 1 ( s) +  − 1 1 −   =
RD'+ Rc  
 D ' (R + Rc )   2  
D ' (R + R c )  DRc  D 
−  F f 1 ( s) + 1 −   .
RD'+ Rc   D ' (R + Rc )  2 

Another part of (5.35) can be rewritten as follows:

Chapter 5. Improved Models 201

D D ' (R + R c ) D
D'
+
RD'+ Rc D'
(
F f 2 (s) − 1 = )
D ' (R + Rc ) D  RD'+ Rc 
 F f 2 (s) − 1 + = (5.37)
RD '+ Rc D '   D ' (R + Rc ) 
D ' (R + Rc ) D  DRc 
 F f 2 (s) + .
RD '+ Rc D '   D ' (R + Rc ) 

By using (5.36) and (5.37) to rewrite (5.35), the combined expression for the
audio susceptibility is obtained:

vˆ o ( s )
=
vˆ g ( s )
 RTs

 RD'+ Rc D ' (R + Rc )  DRc  D  
D'  mc −  F f 1 ( s ) + 1 −   +
 L  R + Rc RD'+ Rc  D ' (R + Rc )  2 
 
(5.38)
 D ' (R + Rc )    T  1 
1 −   H e ( s ) − 1 − s s   +
 
RD '+ Rc    
  2   sTs 
D ' (R + Rc ) D  DRc  1
 F f 2 (s) +  ,
RD '+ Rc D '   D' (R + Rc )   D −1 (1 + sRc C )−1 den( s )

where den(s) is defined in (5.29), F f 1 ( s) is defined in (4.60), F f 2 ( s ) is

defined in (4.61), and H e (s ) is defined in (3.10).
Figure 5.6 shows the Bode plot for the audio susceptibility according to
the combined expression in (5.38) together with the results presented in
Figure 4.20, i.e. the simulation results and the audio susceptibilities predicted
by the novel expression, (5.30), and the Ridley model, (5.28). From the
figure it is seen that the combined expression agrees closely with the
simulation results both at low and high frequencies.
(5.28) is the expression for the audio susceptibility obtained by applying
the Ridley model to the buck-boost converter. This expression is now written
as follows (see (3.109) and (3.102)):
202 Chapter 5. Improved Models

-10

-20
Phase (deg); Magnitude (dB)

-30

-40
0 1 2 3 4
10 10 10 10 10

-20

-40

-60

-80

-100
0 1 2 3 4
10 10 10 10 10
Frequency (Hz)

Figure 5.6: The audio susceptibility of a buck-boost converter with a current

controller. X: the simulation results. Dashed line: the combined
expression. Dotted line: the novel expression. Solid line: the
Ridley model. Note that the dashed line almost coincides with
the solid line at low frequencies.

vˆo ( s)  RTs  RD'+ Rc L  D s2 

= D '  mc + k f  + 1 + 2 −
vˆ g ( s )  L R + Rc  
  DTs Ri  D'  ω n 
(5.39)
T D L  1
s s  + k f   ,
D '  2 T s Ri  D −1 (1 + sR C )−1 den( s)
 c

where

DTs Ri  D
kf =− 1 −  , (5.40)
L  2

and den(s) is defined in (5.29).

Chapter 5. Improved Models 203

The combined expression will now be used to improve the Ridley model.
Note that the denominator in (5.38) and (5.39) are exactly the same. One
way to improve the Ridley model is to modify the numerator in (5.39) in
some way so that it is equal to the numerator in (5.38). The feedforward gain
k f is not present in the denominator in (5.39) and may therefore be used to
modify the numerator in (5.39) without changing the denominator. The
following equation is obtained if the numerator in (5.39) is put equal to the
numerator in (5.38):

RTs  RD '+ Rc L  D s2 
D '  m c + k f ( s )  + 1 + 2 −
L  R + Rc DTs Ri  D '  ω n 

Ts  D L 
s  + k f ( s )  =

D '  2 T s Ri 
RTs  RD '+ Rc D' (R + Rc )  DRc  D  
D'  mc −  F f 1 (s) +
 1 −   + (5.41)
L  R + Rc RD '+ Rc  D ' (R + R c )  2 

 D ' (R + Rc )    T  1 
1 −   H e ( s ) − 1 − s s   +
 
RD '+ Rc    
  2   sTs 
D ' (R + Rc ) D  DRc 
 F f 2 (s) + .
RD '+ Rc D'   D' (R + Rc ) 

The new k f depends on s and this is prepared for in (5.41). k f (s ) is

calculated by using (5.41):

RTs  L  D s2  T D L 
D '  k f ( s )  + 1 + 2  − s s  + k f ( s )  =
L  DTs Ri  D'  ω n  D '  2 T s Ri 
RT  D ' (R + R c )  DRc  D  
− s D'   F f 1 (s) +
 1 −   −
 RD'+ R D ' (R + R c )  2 
L  c 
(5.42)
 D ' (R + Rc )    T  1 
1 −   H e ( s ) − 1 − s s   +
 +  
 RD ' R c   2   sTs 
D ' (R + Rc ) D  DRc 
 F f 2 ( s) + ,
RD '+ Rc D '   D' (R + Rc ) 
204 Chapter 5. Improved Models

k f (s) =

DRi  RTs  D' (R + Rc )  DRc  D  

− D'   F (s) + 1 −   −
 L  RD'+ R  f 1 D ' (R + R c )  2 
RD '   c 

 D ' (R + Rc )    T  1 
1 −   H e ( s ) − 1 − s s   +
 
RD '+ Rc    
  2   sTs 
D ' (R + Rc ) D  DRc  D  s2  
 + s Ts D  •
 F f 2 ( s) + − 1 +
RD '+ Rc D '  D' (R + Rc )  D '  ω n2 
 D' 2 

−1 −1
 DRi   RTs L Ts L 
   
 L D ' DT R − s D' T R  =
(5.43)
 RD'   s i s i 
 DTs Ri  D ' (R + Rc )  DRc  D  
−   F f 1 ( s) + 1 −   −
  RD '+ R  D ' (R + R c )  2 
 L  c 

 D ' (R + Rc )    T   1 
1 −   H e ( s ) − 1 − s s   +
 RD '+ Rc   2   sTs 
 
D 2 Ri D ' ( R + R c )  DRc  D 2 Ri 
1 − s T s + s
2 
 •
 F f 2 (s) + −
 D' (R + Rc )  RD ' 2  
RD ' 2 RD '+ Rc   2 ω n2 
−1
 LD 
1 − s  .
 RD ' 2 

(5.43) is rewritten by using (3.12) and (3.13):

 DTs Ri  D ' (R + Rc )  DRc  D  

k f (s) =  −   F (s) + 1 −   −
  RD'+ R  f 1 D ' (R + R c )  2 
 L  c 

 D ' (R + Rc )    T  1  D 2 Ri D ' ( R + R c )
1 −   H e ( s ) − 1 − s s   − •
RD '+ Rc     RD ' 2 RD '+ R
  2   sTs  c
(5.44)
 RD '+ R  s2  
 c 
1+
s
+ 2  − F f 2 (s) −
DRc  •
 D ' (R + Rc )  ω n Q z ω  D ' ( R + R ) 
  n  c 
−1
 LD 
1 − s  .
 RD ' 2 
Chapter 5. Improved Models 205

where F f 1 ( s) is defined in (4.60), F f 2 ( s ) is defined in (4.61), and H e (s )

is defined in (3.10). Note that the s 0 term in (5.44) is equal to (5.40). Note
also that k f (s ) is an unstable transfer function since its pole is in the right
half plane.

5.5 Summary and Concluding Remarks

A novel model for the audio susceptibility was derived in Chapter 4. This
model was in this chapter used to improve the Ridley and Tan models.
The expressions for the audio susceptibility obtained by applying the
Ridley and Tan models to the buck converter were improved by modifying
the gain of the feedforward of the input voltage, k f . The feedforward gain
k f was derived by putting the numerator in the expressions equal to the
numerator in the novel expression obtained by applying the novel model to
the buck converter. We could use this method since the denominators in the
three expressions are approximately the same and k f is not present in the
denominators.
The expression for the audio susceptibility obtained by applying the
Ridley model to the boost converter was also improved by modifying the
feedforward gain k f . However, we could not use the novel expression
obtained by applying the novel model to the boost converter directly since
this expression exhibits poor low-frequency predictions (see Section 4.6).
Instead, we first derived a combined expression that has high-frequency
properties similar to the novel expression and low-frequency properties
similar to the expression obtained by applying the Ridley model to the boost
converter. The combined expression was then used to derive the modified k f
in the same way as for the buck converter.
The expression for the audio susceptibility obtained by applying the
Ridley model to the buck-boost converter was improved by using the same
methodology as for the boost converter.
The modified k f for the boost and buck-boost converter, (5.27) and
(5.44), will provide unstable transfer functions. This does not seem to be so
natural. The reason for obtaining unstable transfer functions may be due to
that no physical interpretation was used when choosing to modify only k f .
Therefore, it is more suitable to also modify the gain of the feedforward of the
output voltage, k r . However, the modification of k r must be made in a way
that causes negligible effects on the control-to-output transfer function and
the output impedance. From Figure 4.12, it is seen that it is possible to write
206 Chapter 5. Improved Models

( )
dˆ ( s ) = f vˆon ( s ), vˆoff ( s ) , (5.45)

where the function f can be obtained by using (4.21). From (4.21) and
Figure 4.12, it is seen that dˆ ( s ) depends linearly on vˆon ( s ) and vˆoff ( s ) .
The function f may therefore be written as

( ) ( )
dˆ ( s ) = f vˆon ( s ), vˆoff ( s) = k f ' ( s )vˆon ( s ) + k r ' ( s )vˆoff ( s ) Fm , (5.46)

for some k f ' ( s ) and k r ' ( s ) . Fm is defined in (3.32). (5.46) is a unified

model for how dˆ ( s ) is affected by vˆon ( s ) and vˆoff ( s ) . It is unified since the
voltage across the inductor is used instead of the input and output voltages. In
(Johansson, 2002b), it is suggested that an improved unified model may be
obtained if the definitions of k f ' and k r ' presented by Ridley (1991) were
replaced by k f ' ( s ) and k r ' ( s ) defined above. This is not a good suggestion,
which is seen from the small-signal model in Figure 3.8. A change in the
input voltage causes a change in the inductor current and since the inductor
current is fed back, dˆ ( s ) depends on three signals in the small-signal model,
iˆL ( s ) , vˆ g ( s ) and vˆo ( s ) . In the unified model, the three signals are instead
iˆL ( s ) , vˆon ( s ) and vˆoff ( s ) . The k f ' ( s ) and k r ' ( s ) in (5.46) must therefore
be modified before they are to be used in an improved unified model. Further
research is needed to find out how these modifications can be made and if the
resulting model is suitable. The problem with poor predictions at low
frequencies must also be considered before an improved unified model is
obtained.
 Chapter 6 Approximations of
Obtained Expressions

Models for converters with current-mode control were considered in

Chapter 3 and they were improved regarding the audio susceptibility in
Chapter 5. The expressions obtained from all these models are rather
complicated. In this chapter, the expressions for the control-to-output
transfer function and the output impedance obtained by using the Ridley
model are approximated. The expressions for the audio susceptibility
according to the improved Ridley model are also approximated. In Chapter 7,
these approximate expressions will be used to analyze some properties that can
be obtained when load current measurements are utilized for control.

6.1 Chapter Survey

The expressions connected with the buck converter are approximated in
Section 6.2. The expressions connected with the boost and buck-boost
converters are approximated in Section 6.3 and Section 6.4, respectively. A
summary and concluding remarks are presented in Section 6.5.

6.2 Approximate Model for the Buck Converter

In this section, the transfer functions obtained by applying the
(improved) Ridley model to the buck converter are approximated. The
common denominator for these transfer functions is first approximated. The
control-to-output transfer function, the output impedance, and the audio
susceptibility are then approximated by using the approximated denominator.
The expression F f (s ) , used in the numerator of the audio susceptibility, is
also approximated.

207
208 Chapter 6. Approximations of Obtained Expressions

The Denominator in the Transfer Functions

The transfer functions obtained by applying the Ridley model to the buck
converter have the same denominator and it is (see (3.62), (3.19), (3.20), and
(3.21))

RTs
den( s ) = (1 + s(R + Rc )C )Fh−1 ( s ) + (mc D'−0.5)(1 + sRc C ) , (6.1)
L

where

1
Fh ( s ) = ,
s s2 (6.2)
1+ +
ω nQ ω n2

1
Q= , (6.3)
π (mc D'−0.5)

Me
mc = 1 + . (6.4)
M1

mc is unity in the case where no external ramp is used (no slope

compensation).
The characteristic value α can be used to check the stability of the
control of the inductor current. The expression for α , (3.1), is rewritten by
using (3.17) and (6.4):

D D Me D
− Me
M1 − − (mc − 1)
M2 − Me D ' D' M 1 D '
α= = = = =
M1 + M e M1 + M e M 1 + (mc − 1)
1+ e
M1 (6.5)
1
− mc
D' 1
= − 1.
mc mc D'
Chapter 6. Approximations of Obtained Expressions 209

Im(s)

jωn

Re(s)

-jωn

Figure 6.1: The root-locus of the three poles for increasing mc .

mc must be chosen such that α < 1 to obtain stability (see Section 3.2), i.e.
m c D' must be greater than 0.5. The denominator in (6.3) must therefore be
greater than zero and Q tends to infinity if the stability limit is approached.
The transfer functions have three poles since the denominator (6.1) is a
third order polynomial. The poles depend on mc according to Figure 6.1
(see Ridley (1990b, Section 5.2.2)). If mc is small, there are two complex-
conjugated high-frequency poles and one real low-frequency pole. As mc
increases, the imaginary part of the two high-frequency poles decreases and
for a certain mc , they coincide at the real axis. If mc is increased further, the
two poles start moving along the real axis, one towards higher frequencies and
the other one towards lower frequencies. The low-frequency pole is at the
same time moving along the real axis towards higher frequencies and for a
certain (rather high) mc , it coincides with the pole that is moving towards
lower frequencies. If mc is increased further, the two poles start moving away
from the real axis and become a low-frequency complex-conjugated pair.
An approximate denominator will now be derived for the case where mc
is so low that the low-frequency pole has a frequency that is much lower than
the frequencies of the two high-frequency poles. This does not mean that the
two high-frequency poles must be a complex-conjugated pair. They can also
be positioned at different places on the real axis.
210 Chapter 6. Approximations of Obtained Expressions

Assume that the ESR of the capacitor is much smaller than the resistance
of the load, i.e.

Rc << R . (6.6)

The denominator (6.1) can then be rewritten as:

 s s 2  RTs
den( s ) ≈ (1 + sRC ) 1 + + 2 + (mc D'−0.5)(1 + sRc C ) =
 ωnQ ω  L
 n 

RTs  RT 
1+ (mc D'−0.5) +  1 + RC + s (mc D'−0.5)Rc C  s +
L  ω nQ L  (6.7)
 1 1  2 1
 + RC s + RC 2 s 3 =
ω2 
ω nQ  ωn
 n
( )
K −1 1 + a1 s + a 2 s 2 + a3 s 3 = K −1 P ( s) ,

where

1
K= ,
RTs (6.8)
1+ (mc D'−0.5)
L

P ( s ) = 1 + a1 s + a 2 s 2 + a 3 s 3 , (6.9)

 1  RT 
a1 = K  + RC 1 + c s (mc D '−0.5)  , (6.10)
 ω nQ  L 

 1 RC 
a2 = K  2 + , (6.11)
ω 
 n ω nQ 

RC
a3 = K . (6.12)
ω n2
Chapter 6. Approximations of Obtained Expressions 211

The following derivation is similar to the one made by Erickson and

Maksimovic (2000, Section 8.1.8). The polynomial P (s ) is factored into

( )( )(
P ( s ) = 1 − p1−1 s 1 − p 2−1 s 1 − p3−1 s = )
(
1 + − p1−1 − p 2−1 − p3−1 s ) + (p −1 −1
1 p2 )
+ p1−1 p3−1 + p 2−1 p3−1 s 2 + (6.13)
(
− p1−1 −)( )( )
p 2−1 − p3−1 s 3 ,

where p1 , p 2 , and p3 are the three poles. The coefficient a1 is identified

by using (6.9) and (6.13):

a1 = − p1−1 − p 2−1 − p3−1 . (6.14)

As mentioned previously, the approximation will be made with the

assumption that the low-frequency pole has a frequency that is much lower
than the frequencies of the two high-frequency poles. This means that
p1 << p 2 and p1 << p 3 if p1 is the low-frequency pole. a1 is then
approximated with

a1 ≈ − p1−1 . (6.15)

An approximate expression for the (real) low-frequency pole is hence known.

The polynomial P (s ) is approximated with

( )( )
P ( s ) ≈ (1 + a1 s ) 1 − p 2−1 s 1 − p 3−1 s =
(1 + a1 s )(1 + b1 s + b2 s 2 ) = (6.16)
1 + (a1 + b1 )s + (a1b1 + b2 )s 2 + a1b2 s 3 .

The coefficients b1 and b2 are real since the poles p1 and p 2 are either
complex conjugated or real. Approximate expressions for b1 and b2 are
obtained by comparing (6.9) and (6.16):

a3
b2 ≈ , (6.17)
a1
212 Chapter 6. Approximations of Obtained Expressions

a3
a2 −
a 2 − b2 a1 a 2 a3 (6.18)
b1 ≈ ≈ = − .
a1 a1 a1 a12

(6.16) is rewritten by using (6.17) and (6.18):

 a a  
P ( s ) ≈ (1 + a1 s ) 1 +  2 − 32  s + a3 s 2  . (6.19)
  a1 a  a1 
  1  

The denominator (6.7) is now approximated by using (6.19):

 a a  a 
den( s ) ≈ K −1 P ( s ) ≈ K −1 (1 + a1 s ) 1 +  2 − 32  s + 3 s 2  . (6.20)
  a1 a  a1 
  1  

The following two assumptions are made before continuing:

Rc Ts
(mc D'−0.5) << 1 , (6.21)
L

1
<< RC . (6.22)
ω nQ

This means that a1 can be approximated with

a1 ≈ RCK . (6.23)

The denominator (6.20) is approximated by using (6.23), (6.11), and (6.12):

den( s ) ≈ K −1 (1 + RCKs ) •
  1 1 1   (6.24)
1 +  + − s + 1 s 2  .
  RCω 2 ω n Q RCKω n2  ω2 
  n  n 

One part of (6.24) is rewritten as follows by using (6.8), (3.13), and (6.3):
Chapter 6. Approximations of Obtained Expressions 213

1 1 1 1 1  1
+ − = + 1 −  =
RCω n2 ω n Q RCKω n ω n Q RCω n  K 
2 2

  
1
+
1
1 − 1 +
RTs
(mc D'−0.5)  =
ω n Q RCω n2   L 
1 1 1 1 1
Ts (m c D'−0.5) =
(6.25)
− − =
ω n Q LCω n
2 ω n Q LCω n π 2 1
Ts π (mc D '−0.5)

1     
2
1−
1  = 1 1 −  ω 0  ,

ω n Q  LCω n2  ω nQ   ω n  
    

where

1
ω0 = (6.26)
LC

is the corner frequency of the output low-pass LC-filter (if the ESR in the
capacitor is negligible). The corner frequency of this filter should be chosen to
be much lower than the switching frequency to obtain small magnitude of the
ripple in the output voltage (see Section 2.2). The following assumption is
therefore also made:

ω 0 << ω n . (6.27)

The denominator (6.24) can now be approximated by using (6.25) and

(6.27):

den( s ) ≈ K −1 (1 + sRCK )Fh−1 ( s ) , (6.28)

where Fh (s) is defined in (6.2).

One of the assumptions used in the derivation of the approximate
denominator (6.28) is that the low-frequency pole has a frequency that is
much lower than the frequencies of the two high-frequency poles. Conditions
that ensure that this assumption is fulfilled are derived in the following by
utilization of the approximate denominator (6.28). It is a little unsatisfactory
214 Chapter 6. Approximations of Obtained Expressions

to utilize the expression for the approximate denominator to derive a

condition that shows if the expression for the approximate denominator is
valid. This is made anyway since the derivation then becomes rather simple.
In the approximate denominator (6.28), the low-frequency pole is equal
to − 1 RCK and the two high-frequency poles are determined by Fh (s) . In
the case where the two high-frequency poles are complex conjugates, both of
them have the frequency ω n . The condition

1
<< ω n (6.29)
RCK

ensures that the low-frequency pole has a frequency that is much lower than
the frequencies of the two high-frequency poles in this case. In the case where
the two high-frequency poles are real, one of them has a frequency that is
lower than (or equal to) ω n . By putting the denominator in Fh (s) equal to
zero, it is concluded that the two high-frequency poles are real if and only if
Q ≤ 0.5 . The frequency of the pole with the lowest frequency can in this case
be approximated with ω n Q according to Erickson and Maksimovic (2000,
Section 8.1.7). This pole is hence moving towards the low-frequency pole as
Q is decreased. This is in accordance with Figure 6.1 since Q is decreasing if
mc is increasing (see (6.3)). It is concluded that the condition

1
<< ω n Q (6.30)
RCK

ensures that the low-frequency pole has a frequency that is much lower than
the frequencies of the two high-frequency poles in the case where the two
high-frequency poles are real.
From (6.8), it is seen that K is a positive number less than unity since
m c D' is greater than 0.5 if the current control is stable (see (6.5)). The
following is obtained if this result is combined with (6.30):

1
<< RCK < RC . (6.31)
ω nQ

Hence, the condition (6.30) ensures that the condition (6.22) is fulfilled.
Chapter 6. Approximations of Obtained Expressions 215

The results are now summarized. The denominator in the transfer

functions obtained by applying the Ridley model to the buck converter, i.e.
(6.1), is approximated with

den( s ) = K −1 (1 + sRCK )Fh−1 ( s ) , (6.32)

where

1
K= ,
RTs (6.33)
1+ (mc D'−0.5)
L

and Fh (s) is defined in (6.2), if

Rc << R , (6.34)

Rc Ts
(mc D'−0.5) << 1 , (6.35)
L

1
<< ω n , (6.36)
LC

1
<< ω n , (6.37)
RCK

and

1
<< ω n Q , (6.38)
RCK

where Q is defined in (6.3). Note that the conditions (6.34)-(6.38) are

sufficient to ensure that (6.32) is a good approximation. This does not mean
that they are necessary, i.e. (6.32) may be a good approximation even if some
of the conditions are not fulfilled.
Current-mode control behaves like voltage-mode control if mc is high
since the feedback of the inductor current then becomes negligible compared
216 Chapter 6. Approximations of Obtained Expressions

50

40

30

20
Phase (deg); Magnitude (dB)

10

0
1 2 3 4
10 10 10 10

200

150

100

50

0
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 6.2: The denominator in the transfer functions for a buck converter
with a current controller. Dashed line: the approximated
denominator. Solid line: the Ridley model. Note that the two
lines almost coincide.

to the external ramp (compare Figure 3.3 and Figure 3.4). In voltage-mode
control, there is a resonance frequency at approximately 1 LC . This is
predicted by the denominator (6.1) since it has two low-frequency complex-
conjugated poles for high mc . However, the approximate denominator
(6.32) does not predict this since it only has real poles for high mc .
The Bode plots for the denominators (6.1) and (6.32) are shown in
Figure 6.2. The parameter values shown in Table 2.1 are used. From the
figure it is seen that the two denominators are almost the same. The
maximum difference is approximately 0.1 dB and 0.3 degrees in the presented
frequency interval.

Control-to-Output Transfer Function

An approximate version of the control-to-output transfer function

obtained by applying the Ridley model to the buck converter is
Chapter 6. Approximations of Obtained Expressions 217

vˆo ( s ) R (1 + sRc C )
= , (6.39)
iˆc ( s ) den( s )

where den(s) is defined in (6.32) if the conditions in (6.34)-(6.38) are

fulfilled. The only difference compared to the non-approximated version
(3.59) is that the approximate denominator (6.32) is used instead of the
denominator (6.1).
(6.39) is rewritten by using (6.32):

vˆo ( s ) R(1 + sRc C )

= −1 = RKFl ( s ) FESR ( s ) Fh ( s ) , (6.40)
iˆc ( s ) K (1 + sRCK )Fh−1 ( s )

where

1
Fl ( s) = , (6.41)
1 + sRCK

FESR ( s) = 1 + sRc C , (6.42)

Fh (s) is defined in (6.2), and K is defined in (6.33). (6.40) is exactly the

same as the approximate control-to-output transfer function proposed by
Ridley (1991).

Output Impedance

An approximate version of the output impedance obtained by applying

the Ridley model to the buck converter is

R(1 + sRc C )Fh−1 ( s)

Z out ( s ) = , (6.43)
den( s )

where den(s) is defined in (6.32) and Fh (s) is defined in (6.2) if the

conditions in (6.34)-(6.38) are fulfilled. The only difference compared to the
non-approximated version (3.60) is that the approximate denominator (6.32)
is used instead of the denominator (6.1).
(6.43) is rewritten by using (6.32):
218 Chapter 6. Approximations of Obtained Expressions

R (1 + sRc C )Fh−1 ( s )
Z out ( s ) = = RKFl ( s ) FESR ( s ) , (6.44)
K −1 (1 + sRCK )Fh−1 ( s )

where K is defined in (6.33), Fl (s) is defined in (6.41), and FESR (s ) is

defined in (6.42). The two high-frequency poles are cancelled by two of the
zeros. (6.44) is exactly the same as the approximate output impedance
proposed by Ridley (1991).

Audio Susceptibility

In this subsection, the audio susceptibilities obtained by applying the

Ridley model and the improved Ridley model to the buck converter are
approximated. The expression F f (s ) , used in the numerator of the improved
expression for the audio susceptibility, is approximated in two different ways.
The first one is by using a Taylor polynomial and the second one is by
minimizing an integral.
An approximate version of the audio susceptibility obtained by applying
the Ridley model to the buck converter is

RTs   D 
D mc D '−1 −  (1 + sRc C )
vˆo ( s) L   2  (6.45)
= ,
vˆ g ( s ) den( s)

where den(s) is defined in (6.32) if the conditions in (6.34)-(6.38) are

fulfilled. The only difference compared to the non-approximated version
(3.61) is that the approximate denominator (6.32) is used instead of the
denominator (6.1).
(6.45) is rewritten by using (6.32):

RTs   D 
D mc D '−1 −  (1 + sRc C )
vˆo ( s) L   2 
= =
vˆ g ( s ) K −1 (1 + sRCK )Fh−1 ( s) (6.46)
RTs D   D 
 mc D '−1 −   KFl ( s ) FESR ( s ) Fh ( s ) ,
L   2 
Chapter 6. Approximations of Obtained Expressions 219

where K is defined in (6.33), Fl (s) is defined in (6.41), FESR (s ) is defined

in (6.42), and Fh (s) is defined in (6.2). (6.46) is exactly the same as the
approximate audio susceptibility proposed by Ridley (1991).
The expression for the audio susceptibility obtained by applying the
Ridley model to the buck converter was improved in Chapter 5 by replacing
the numerator (see (5.9)). An approximate version of this improved
expression is

vˆo ( s)
RTs
( )
D mc D '− F f ( s ) (1 + sRc C )
= L =
vˆ g ( s ) K −1 (1 + sRCK )Fh−1 ( s ) (6.47)
RTs D
L
( )
mc D '− F f ( s ) KFl ( s ) FESR ( s ) Fh ( s ) ,

where

1  sTs 1 − e − sDTs sTs 

F f ( s) =  − =
sTs  1 − e − sTs sDT e sTs − 1 
 s
(6.48)
 D  (3 − 2 D )DTs
1 −  − s−
(
1 − 2 D + D 2 DTs2 2
s +K
)
 2 12 24

(see (4.27)), K is defined in (6.33), Fl (s) is defined in (6.41), FESR (s ) is

defined in (6.42), and Fh (s) is defined in (6.2) if the conditions in (6.34)-
(6.38) are fulfilled.
Since F f (s ) is a rather complicated expression, it is desirable to find an
approximate expression. F f (s ) can be approximated by a Taylor polynomial,
i.e. a truncated version of the Taylor series in (6.48). The higher degree of the
Taylor polynomial that is used, the better approximation is obtained. If a
Taylor polynomial of degree 0 is used, k f (s ) in (5.10) is the same as k f in
(5.3) and the Ridley model is not improved. This is also seen if (6.46) and
(6.47) are compared. If a Taylor polynomial of degree 1 is used, there is an
extra zero in the improved Ridley model. In Section 3.4, it was observed that
the Tan model includes an extra zero compared to the Ridley model
(compare (3.61) and (3.71)). If a Taylor polynomial of degree 1 is used to
improve the Tan model, the extra zero that is already present in the Tan
model, is moved to a more suitable position. The extra zero in the Tan model
220 Chapter 6. Approximations of Obtained Expressions

explains why the Tan model is better than the Ridley model in Figure 4.14.
However, the Ridley model is better than the Tan model in some cases since
the extra zero in the Tan model is not placed at the most suitable position.
This is for instance the case if D is small. If D tend to zero, F f (s ) in (6.48)
tend to 1 and 1 − D 2 in the numerator of (6.46) also tend to 1. The Ridley
model therefore becomes the same as the improved version of the Ridley
model. However, in the Tan model the extra zero remains, which can be
concluded from (3.71). s is equal to the extra zero if

 D s
m c D '− 1 −  + = 0. (6.49)
 2  πω n

(6.49) is rewritten:

  D    1 
s = −πω n  mc D '− 1 −   = −πω n  mc − 1 − D mc −   . (6.50)
  2    2 

It is seen from (6.50) that the extra zero does not disappear (move to infinity)
as D tend to zero.
Figure 6.3 is a modified version of Figure 4.14. It shows the Bode plot for
the audio susceptibility predicted by the Ridley model (3.61) together with
the simulation results. It also shows the audio susceptibilities predicted by the
improved Ridley model ((3.58) with k f replaced by k f (s ) in (5.10)) in the
cases where F f (s ) is not approximated, F f (s ) is approximated by a Taylor
polynomial of degree 1 and 2. From the figure it is seen that the predictions
of the improved Ridley model agree more closely with the simulation results
if the degree of the Taylor polynomial is increased.
A Taylor polynomial approximates a function in a neighborhood of a
point. There are other methods that approximate a function in an interval.
The rest of this section will be devoted to an example of how F f (s ) can be
approximated in the frequency interval dc to half the switching frequency,
ω n . This example can be omitted by the reader without risk of increasing
difficulty to understand subsequent sections of this thesis.
Let the function g (z ) be defined by (see (6.48))

 z  1  z 1 − e − zD z 
g ( z ) = F f   =  − . (6.51)
 −z
e z − 1 
 Ts  z 1 − e zD
Chapter 6. Approximations of Obtained Expressions 221

-30

-40

-50

-60
Phase (deg); Magnitude (dB)

-70

-80
1 2 3 4
10 10 10 10

-50

-100

-150
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 6.3: The audio susceptibility of a buck converter with a current

controller ( mc =1.5). X: the simulation results. Solid line: the
Ridley model. Dashed line: the improved Ridley model with
F f (s ) approximated by a Taylor polynomial of degree 1. Dash-
dotted line: the improved Ridley model with F f (s )
approximated by a Taylor polynomial of degree 2. Dotted line:
the improved Ridley model with F f (s ) not approximated.

Note that the complex variable z is not connected with the Z-transform in
this chapter. g (z ) will be approximated and an approximation of F f (s ) is
then obtained by

F f ( s ) = g ( sTs ) . (6.52)

Let s = jω . If 0 ≤ ω ≤ ω n , then sTs is in the interval [0, jω n Ts ] which is

rewritten as [0, jπ ] by using (3.13). g ( z ) should therefore be approximated
for z in the interval [0, jπ ] .
One way of approximating a function f ( z ) on a curve C is to use the
method of least squares and the result is (Walsh, 1935, Section 6.1):
222 Chapter 6. Approximations of Obtained Expressions

f ( z ) ≈ a 0 p 0 ( z ) + a1 p1 ( z ) + K + a n p n ( z ) . (6.53)

where

a k = ∫ f ( z ) p k ( z ) dz , (6.54)
C

and the functions p 0 ( z ) , p1 ( z ) , K , p n ( z ) are normal and orthogonal on

C . The coefficients a k in (6.54) minimizes the integral

n 2

∫ f ( z) − ∑ ak pk ( z) dz . (6.55)
C k =0

The function g ( z ) is approximated by using (6.53) and (6.54). The

curve C is chosen to be the part of the imaginary axis that starts at the origin
and ends at jπ . A set of functions, p 0 ( z ) , p1 ( z ) , K , p n ( z ) , that are
normal and orthogonal on C must also be found. It is seen from (6.55) that
the approximation error is equally weighed for each z in C . It is possible to
reduce the approximation error for some z by introducing a weight function.
The approximation error will then increase for other z . Another possibility is
to introduce a special condition and then reformulate the approximation
problem so that the condition is fulfilled. If the special condition is that the
approximation error should be zero for z =0 (dc), then the reformulated
approximation problem is to approximate the function f ( z ) defined by

g ( z ) − g ( 0)
f ( z) = . (6.56)
z

g ( z ) is then obtained by

g ( z ) = g (0) + zf ( z ) . (6.57)

It is seen from (6.57) that the approximation error equals zero for z =0.
If f ( z ) in (6.56) is approximated by using (6.53) and (6.54) and the
functions p 0 ( z ) , p1 ( z ) , K , p n ( z ) are polynomials, then the
approximated f ( z ) is also a polynomial. The coefficients in this polynomial
are complex numbers. It will now be shown how f ( z ) can be approximated
Chapter 6. Approximations of Obtained Expressions 223

by a polynomial of degree 1 where the coefficients are restricted to be real

numbers. Note that this approximation will not be as good as the case where
complex coefficients are allowed. f ( z ) is approximated by the polynomial

λ0 + λ1 z . (6.58)

The values of the real coefficients λ0 and λ1 are obtained by minimizing the
integral

jπ π
∫ f ( z ) − (λ 0 + λ1 z ) dz = ∫ f ( jθ ) − (λ 0 + λ1 jθ ) dθ =
2 2

0 0
π
∫ ( f ( jθ ) − λ0 − λ1 jθ )( f ( jθ ) − λ0 − λ1 jθ ) dθ =
0
π
∫ ( f ( jθ ) f ( jθ ) − f ( jθ )λ0 + f ( jθ )λ1 jθ − λ0 f ( jθ ) + (6.59)
0

λ 20 − λ 0 λ1 jθ − λ1 jθ f ( jθ ) + λ1 jθλ0 + λ12θ 2 dθ = )
π π
∫ (
f ( jθ ) f ( jθ )dθ − λ 0 ∫ f ( jθ ) + f ( jθ ) dθ − )
0 0
π
(
λ1 ∫ − jθ f ( jθ ) + jθ f ( jθ ) dθ + λ20π + λ12) π3
3
.
0

Define the following:

π
∫ ( f ( jθ ) + f ( jθ ))dθ ,
1
a0 = (6.60)
2π 0

π
∫ (− jθ )
3
a1 = f ( j θ ) + j θ f ( j θ ) dθ . (6.61)
2π 3 0

(6.59) is rewritten by using (6.60) and (6.61);

224 Chapter 6. Approximations of Obtained Expressions

jπ

∫ f ( z ) − (λ0 + λ1 z )
2
dz =
0
π
2π 3 π3
∫ f ( jθ ) f ( jθ )dθ − λ0 2πa0 − λ1 3
a1 + λ20π + λ12
3
+
0 (6.62)
π3 π3
πa 02 − πa 02 + a12 − a12 =
3 3
π
π3
(λ1 − a1 )2 − πa02 − π
3
∫ f ( jθ ) f ( jθ )dθ + π (λ0 − a 0 )2 + a12 .
0
3 3

It is seen from (6.62) that the minimum is obtained if λ0 is equal to a 0 and

λ1 is equal to a1 . It is also seen that if f ( z ) should be approximated by a
polynomial of degree 0, i.e. λ1 =0, then λ0 should still be set equal to a 0 to
obtain the minimum. (6.60) and (6.61) are rewritten:

π
1
a0 =
π ∫ Re( f ( jθ ) )dθ , (6.63)
0

π π
a1 = −
3
3 ∫
( jθ f ( jθ ) + jθ f ( jθ ))dθ = − 33 ∫ Re( jθ f ( jθ ))dθ =
2π 0 π 0
π
3
−
π3
∫ Re( j (Re(θ f ( jθ ) ) + j Im(θ f ( jθ ) )))dθ = (6.64)
0
π π
3 3
π3
∫ Im(θ f ( jθ ) )dθ =
π3
∫ θ Im( f ( jθ ))dθ .
0 0

The conclusion is that the function f ( z ) in (6.56) can be approximated by

the first order polynomial

f ( z ) ≈ a 0 + a1 z , (6.65)

where a 0 is defined in (6.63) and a1 is defined in (6.64). f ( z ) can also be

approximated by the zeroth order polynomial
Chapter 6. Approximations of Obtained Expressions 225

f ( z) ≈ a0 , (6.66)

where a 0 is defined in (6.63). Note that a 0 and a1 are real numbers.

g (0) is calculated by application of (6.51) and (6.48):

D
g (0) = F f (0 ) = 1 − . (6.67)
2

From (6.52), (6.57), and (6.67), F f ( s ) is written as

 D
F f ( s) = g ( sTs ) = g (0) + sTs f ( sTs ) = 1 −  + sTs f ( sTs ) . (6.68)
 2

F f ( s ) can now be approximated with

 D
F f ( s ) ≈ 1 −  + a 0Ts s + a1Ts2 s 2 (6.69)
 2

(by using (6.68) and (6.65)) or

 D
F f ( s ) ≈ 1 −  + a 0 T s s (6.70)
 2

(by using (6.68) and (6.66)), where

π
1
a0 =
π ∫ Re( f ( jθ ) )dθ , (6.71)
0

π
3
a1 =
π3
∫ θ Im( f ( jθ ) )dθ . (6.72)
0

f ( jθ ) in (6.71) and (6.72) is obtained by using (6.56), (6.67), and (6.51):

226 Chapter 6. Approximations of Obtained Expressions

-0.01

-0.02

-0.03

-0.04

-0.05
a0

-0.06

-0.07

-0.08

-0.09

-0.1

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

D

Figure 6.4: Calculated value of a 0 for different values of D .

 D
g ( jθ ) −  1 − 
g ( jθ ) − g (0)  2
f ( jθ ) = = =
jθ jθ
(6.73)
1  jθ 1 − e − jθD jθ   D
 − − 1 − 
jθ  − jθ jθD θ 
1 − e e − 1 
j 2
.
jθ

The integrals in (6.71) and (6.72) must be solved numerically. The lower
integration limit must be increased to a small positive number since there will
be divisions with zero if θ is set to zero in f ( jθ ) . It is seen from (6.73) that
f ( jθ ) depends on D , i.e. the operating-point value of the duty cycle. a 0
and a1 therefore also depend on D . a 0 and a1 are calculated for 1000
different values of D in the interval ] 0, 1] . The results are plotted in Figure
6.4 and Figure 6.5.
Chapter 6. Approximations of Obtained Expressions 227

-0.001

-0.002

-0.003
a1

-0.004

-0.005

-0.006

-0.007

-0.008
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
D

Figure 6.5: Calculated value of a1 for different values of D .

A polynomial curve is fitted to the 1000 values plotted in Figure 6.4 by

utilization of the method of least squares. An approximate expression for a 0
is obtained:

a 0 ≈ 0.00046 − 0.25892 D + 0.15370 D 2 + 0.01654 D 3 . (6.74)

The same is made for the values plotted in Figure 6.5 and an approximate
expression for a1 is:

a1 ≈ 0.00012 − 0.04933D + 0.09816 D 2 − 0.04892 D 3 . (6.75)

Figure 6.6 is a modified version of Figure 6.3. It shows the Bode plot for
the audio susceptibility predicted by the Ridley model (3.61) together with
the simulation results. It also shows the audio susceptibilities predicted by the
improved Ridley model ((3.58) with k f replaced by k f ( s ) in (5.10)) in the
cases where F f ( s ) is not approximated, F f ( s ) is approximated by (6.70),
228 Chapter 6. Approximations of Obtained Expressions

-30

-40

-50

-60
Phase (deg); Magnitude (dB)

-70

-80
1 2 3 4
10 10 10 10

-50

-100

-150
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 6.6: The audio susceptibility of a buck converter with a current

controller ( mc =1.5). X: the simulation results. Solid line: the
Ridley model. Dashed line: the improved Ridley model with
F f ( s ) approximated by a Taylor polynomial of degree 1. Dash-
dotted line: the improved Ridley model with F f ( s )
approximated by a polynomial of degree 1 obtained by
minimizing the square of the approximation error. Dotted line:
the improved Ridley model with F f ( s ) not approximated.

and a Taylor polynomial of degree 1. From the figure it is seen that the
maximum difference between the predictions of the improved Ridley model
where F f ( s ) is not approximated and where (6.70) is used is a little smaller
than the maximum difference between the predictions of the improved Ridley
model where F f ( s ) is not approximated and where F f ( s ) is approximated
with a Taylor polynomial of degree 1. It is also seen that the difference tend
to zero when the frequency tend to zero in both cases.
Figure 6.7 shows the same as Figure 6.6 except (6.69) is used instead of
(6.70) and that the degree of the Taylor polynomial is increased to 2. The
conclusion made for Figure 6.6 is valid also for Figure 6.7.
Chapter 6. Approximations of Obtained Expressions 229

-30

-40

-50

-60
Phase (deg); Magnitude (dB)

-70

-80
1 2 3 4
10 10 10 10

-50

-100

-150
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 6.7: The audio susceptibility of a buck converter with a current

controller ( mc =1.5). X: the simulation results. Solid line: the
Ridley model. Dashed line: the improved Ridley model with
F f ( s ) approximated by a Taylor polynomial of degree 2. Dash-
dotted line: the improved Ridley model with F f ( s )
approximated by a polynomial of degree 2 obtained by
minimizing the square of the approximation error. Dotted line:
the improved Ridley model with F f ( s ) not approximated.

6.3 Approximate Model for the Boost Converter

In this section, the transfer functions obtained by applying the
(improved) Ridley model to the boost converter are approximated. The
common denominator for these transfer functions is first approximated. The
control-to-output transfer function, the output impedance, and the audio
susceptibility are then approximated.
230 Chapter 6. Approximations of Obtained Expressions

The Denominator in the Transfer Functions

The transfer functions obtained by applying the Ridley model to the

boost converter have the same denominator and it is (see (3.95), (3.35), and
(3.12))

den( s ) =
Ts mc RD'+ Rc   RD '+ Rc  
 RD '  + sR C  + sL(1 + s(R + Rc )C ) −
  R+R c  
L R   c  
RD '3 Ts (R + R )L
(1 + sRc C ) 1 − s 2 c2  + (6.76)
2L  R D' 
 2 RD'+ Rc RD'+ Rc 
H e ( s ) + s(R + Rc )C ,
 RD' RD ' 

where

s s2
H e (s) = 1 + + 2 , (6.77)
ω nQz ω n

−2
Qz = . (6.78)
π

An approximate denominator will now be derived for the case where mc

is so low that the low-frequency pole has a frequency that is much lower than
the frequencies of the two high-frequency poles.
Assume that

Rc << RD' . (6.79)

Since D' is a positive number less than or equal to unity, the following is
obtained by using (6.79):

Rc << RD ' ≤ R . (6.80)

The denominator (6.76) is rewritten by using (6.80) and (3.54):

Chapter 6. Approximations of Obtained Expressions 231

Ts m c
den( s ) ≈ D ' (RD' (D'+ sRc C ) + sL(1 + sRC )) −
L
RD '3 Ts
(1 + sRc C ) 1 − s L 2 
+
2L  RD' 
 2 
1 + s + s − sTs m c D ' (2 + sRC ) =
 ω nQ ω 2 
 n 
Ts m c RD '3 Ts
RD '3 − +2+
L 2L
 Ts m c (6.81)
 RD '3 Ts
RD ' 2 Rc C + Ts mc D '− Rc C +
 L 2 L

D ' Ts 2 
+ − 2Ts mc D '+ RC  s +
2 ω nQ 
 Ts m c D ' Ts 2 1 
 D ' LRC + Rc C + 2 + RC − Ts mc D' RC  s 2 +
 L 2 ω n ω nQ 
 
1
( )
RCs 3 = K −1 1 + a1 s + a 2 s 2 + a3 s 3 = K −1 P ( s ) ,
ωn2

where

1
K= 3
,
RD ' Ts (6.82)
2+ (mc − 0.5)
L

P ( s ) = 1 + a1 s + a 2 s 2 + a 3 s 3 , (6.83)

 2  R D ' 2 Ts  D'   
a1 = K  + RC 1 + c  mc −   − D ' Ts (mc − 0.5) , (6.84)
 ω nQ  L  2   
  
232 Chapter 6. Approximations of Obtained Expressions

 2 RC D ' Ts 
a2 = K  2 + + Rc C  , (6.85)
ω 
 n ω nQ 2 

RC
a3 = K . (6.86)
ω n2

Since the derivation of the approximate denominator is made with the

assumption that the low-frequency pole has a frequency that is much lower
than the frequencies of the two high-frequency poles, (6.19) is used to
approximate the denominator (6.81):

 a a  
den( s ) ≈ K −1 P ( s ) ≈ K −1 (1 + a1 s ) 1 +  2 − 32  s + a3 s 2  . (6.87)
  a  a1 
  1 a1  

The following three assumptions are made before continuing:

Rc D ' 2 Ts  D' 
 mc −  << 1 , (6.88)
L  2

D ' Ts (m c − 0.5) << RC , (6.89)

2
<< RC . (6.90)
ω nQ

This means that a1 can be approximated with

a1 ≈ RCK . (6.91)

The denominator (6.87) is approximated by using (6.91), (6.85), and (6.86):

den( s ) ≈ K −1 (1 + RCKs ) •
  2 1 R D ' Ts 1   (6.92)
1 +  + + c − s + 1 s 2  .
  ω nQ  ω2 
  RCω n RCKω n2
2 2R  n 
Chapter 6. Approximations of Obtained Expressions 233

One part of (6.92) is rewritten as follows by using (6.82) and (3.13):

2 1 R D ' Ts 1
+ + c − =
RCω n2 ω nQ 2R RCKω n2
1 1  1  Rc D ' Ts
+  2 − + =
ω n Q RCω n2  K 2R
  3 
1
+
1  2 −  2 + RD' Ts (m − 0.5)  + Rc D ' Ts =
ω n Q RCω n2   L
c  2R (6.93)
  
D'3 π
(mc − 0.5) + c π =
1 R D'
−
ω n Q LCω n ω n
2 2R ω n

1  1 πD ' Rc 
2
ω 
− πD '3 (mc − 0.5)  0  + ,
ωn Q 2 R 
  ωn  

where

1
ω0 = . (6.94)
LC

Assume that

πD ' Rc 1
<< , (6.95)
2 R Q

2
ω  1
πD ' (mc − 0.5)  0
3
 << . (6.96)
 ωn  Q

By using (6.80) and the fact that D ' is a positive number less than or equal
to unity, it is concluded that the left side of (6.95) is much smaller that unity.
The condition (6.95) is therefore fulfilled unless Q is very high, i.e. the
current control is very near the stability limit. If mc is not to high, the same
234 Chapter 6. Approximations of Obtained Expressions

is true also for condition (6.96) since ω 0 usually is much lower than ω n .
The denominator (6.92) is approximated by using (6.93), (6.95), and (6.96):

den( s ) ≈ K −1 (1 + sRCK )Fh−1 ( s ) , (6.97)

where Fh ( s) is defined in (6.2).

One of the assumptions used in the derivation of the approximate
denominator (6.97) is that the low-frequency pole has a frequency that is
much lower than the frequencies of the two high-frequency poles. Conditions
that ensure that this assumption is fulfilled are (6.29) and (6.30), i.e. the
conditions derived for the buck converter. The same conditions can be used
for the boost converter since the approximate denominators are exactly the
same (see (6.28) and (6.97)) except for the definition of K .
From (6.82), it is seen that K is a positive number less than 0.5. The
following is obtained if this result is combined with (6.30):

2
<< 2 RCK < RC . (6.98)
ω nQ

Hence, the condition (6.30) ensures that the condition (6.90) is fulfilled.
The results are now summarized. The denominator in the transfer
functions obtained by applying the Ridley model to the boost converter, i.e.
(6.76), is approximated with

den( s ) = K −1 (1 + sRCK )Fh−1 ( s ) , (6.99)

where

1
K= 3
,
RD ' Ts (6.100)
2+ (mc − 0.5)
L

and Fh ( s) is defined in (6.2), if

Rc << RD' , (6.101)

Chapter 6. Approximations of Obtained Expressions 235

Rc D ' 2 Ts  D' 
 mc −  << 1 , (6.102)
L  2

D ' Ts (m c − 0.5) << RC , (6.103)

πD ' Rc 1
<< , (6.104)
2 R Q

2
 1 LC 
πD ' (mc − 0.5) 
3  << 1 , (6.105)

 ωn  Q

1
<< ω n , (6.106)
RCK

and

1
<< ω n Q , (6.107)
RCK

where Q is defined in (6.3).

The Bode plots for the denominators (6.76) and (6.99) are shown in
Figure 6.8. The parameter values shown in Table 2.5 are used. From the
figure it is seen that the two denominators are almost the same. The
maximum difference is approximately 0.3 dB and 2 degrees in the presented
frequency interval.

Control-to-Output Transfer Function

An approximate version of the control-to-output transfer function

obtained by applying the Ridley model to the boost converter is

 L 
RD ' (1 + sRc C ) 1 − s 
vˆo ( s )
=  RD ' 2 , (6.108)
iˆc ( s ) den( s )
236 Chapter 6. Approximations of Obtained Expressions

50

40

30

20
Phase (deg); Magnitude (dB)

10

0
1 2 3 4
10 10 10 10

200

150

100

50

0
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 6.8: The denominator in the transfer functions for a boost converter
with a current controller. Dashed line: the approximated
denominator. Solid line: the Ridley model. Note that the two
lines almost coincide.

where den( s) is defined in (6.99) if the conditions in (6.101)-(6.107) are

fulfilled. The only differences compared to the non-approximated version
(3.92) are that the approximate denominator (6.99) is used instead of the
denominator (6.76) and that (6.101) (which gives (6.80)) is used to
approximate the numerator.
(6.108) is rewritten by using (6.99):

 L 
RD ' (1 + sRc C ) 1 − s 
vˆo ( s )  RD ' 2 
= = (6.109)
iˆc ( s ) K −1 (1 + sRCK )Fh−1 ( s )
RD' KFl ( s) FESR ( s) Fh ( s ) FRHP ( s ) ,

where
Chapter 6. Approximations of Obtained Expressions 237

1
Fl ( s) = , (6.110)
1 + sRCK

L
FRHP ( s ) = 1 − s , (6.111)
RD ' 2

FESR ( s) is defined in (6.42), Fh ( s) is defined in (6.2), and K is defined in

(6.100).

Output Impedance

The output impedance obtained by applying the Ridley model to the

boost converter is given by (3.93). The numerator in (3.93) will first be
approximated. By using (6.101) (which gives (6.80)) and the fact that D is a
positive number less than or equal to unity, the following is obtained:

Rc
Rc D << Rc D ≤ Rc << RD' ≤ R . (6.112)
R

The numerator in (3.93) is approximated by using (6.112):

num( s ) = R(1 + sRc C ) •

 Ts m c DD' 2 Rc Rc2 D s s2  (6.113)
1 + + 2 + + 2 + sTs mc D '  .
 L R D' ω n Q z ω n 
 

One part of (6.113) is approximated by using (2.39) and (6.112):

Ts mc DD ' 2 Rc Rc2 D
1+ + 2 =
L R D'
Rc
RD '+ Rc D 2
(6.114)
R + Rc D ' Ts m (1 − D ') ≈
c
RD' L
Rc D ' 2 Ts
1+ (mc − mc D') .
L
238 Chapter 6. Approximations of Obtained Expressions

The second term in (6.114) is approximated by using (6.102) and the fact
that mc is a positive number greater than or equal to unity:

Rc D ' 2 Ts Rc D ' 2 Ts  1 
(mc − mc D') ≤  mc − D '  << 1 . (6.115)
L L  2 

Hence, (6.114) is approximately equal to unity. The numerator (6.113) can

therefore be approximated with:

num( s ) ≈ R(1 + sRc C )Fh−1 ( s ) (6.116)

according to (3.54).
An approximate version of the output impedance obtained by applying
the Ridley model to the boost converter is

R(1 + sRc C )Fh−1 ( s)

Z out ( s ) = , (6.117)
den( s )

where den( s) is defined in (6.99) and Fh ( s) is defined in (6.2) if the

conditions in (6.101)-(6.107) are fulfilled. The only differences compared to
the non-approximated version (3.93) are that the approximate denominator
(6.99) is used instead of the denominator (6.76) and that the approximate
numerator (6.116) is used instead of the numerator in (3.93).
(6.117) is rewritten by using (6.99):

R (1 + sRc C )Fh−1 ( s )
Z out ( s ) = = RKFl ( s ) FESR ( s ) , (6.118)
K −1 (1 + sRCK )Fh−1 ( s )

where K is defined in (6.100), Fl ( s) is defined in (6.110), and FESR ( s ) is

defined in (6.42). The two high-frequency poles are cancelled by two of the
zeros.

Audio Susceptibility

The expression for the audio susceptibility obtained by applying the

Ridley model to the boost converter was improved in Chapter 5 by replacing
Chapter 6. Approximations of Obtained Expressions 239

the numerator (see (5.24)). The improved expression then became the same
as the combined expression (5.21). An approximate version of the improved
expression is

 RTs 1
vˆo ( s)  L
 ( D' 
)
D' mc D '− F f ( s ) + (1 + sRc C )
(6.119)
= ,
vˆ g ( s ) den( s )

where

1 1 1 T T3
F f ( s) = − sT = − s s + s s3 + K (6.120)
sTs e s − 1 2 12 720

(see (4.36)) and den( s) is defined in (6.99) if the conditions in (6.101)-

(6.107) are fulfilled. The only differences compared to the non-approximated
improved version (5.21) are that the approximate denominator (6.99) is used
instead of the denominator (6.76) and that (6.101) (which gives (6.80)) is
used to approximate the numerator.
(6.119) is rewritten by using (6.99):

 RTs 1
 ( )
D ' mc D'− F f ( s) + (1 + sRc C )
vˆo ( s)  L D' 
= =
vˆ g ( s ) K −1 (1 + sRCK )Fh−1 ( s ) (6.121)
 RTs D ' 1
 ( )
m c D'− F f ( s ) +  KFl ( s ) FESR ( s ) Fh ( s ) ,
 L D' 

where F f ( s ) is defined in (6.120), K is defined in (6.100), Fl ( s) is defined

in (6.110), FESR ( s ) is defined in (6.42), and Fh ( s) is defined in (6.2).
k f ( s ) defined in (5.27) will now be approximated. k f ( s ) is first
approximated by using (6.101) (which gives (6.80)):

−1
T R   2   L 
k f (s) ≈ − s i  2F f (s) − s L 1 − s   1 − s  , (6.122)
 RD ' 2 πω n  RD ' 2 
2L   
240 Chapter 6. Approximations of Obtained Expressions

-20
Phase (deg); Magnitude (dB)

-40

-60
1 2 3 4
10 10 10 10

-50

-100

-150
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 6.9: The audio susceptibility of a boost converter with a current

controller. X: the simulation results. Solid line: the Ridley
model. Dashed line: the improved Ridley model with F f ( s )
approximated by a Taylor polynomial of degree 0. Dash-dotted
line: the improved Ridley model with F f ( s ) approximated by a
Taylor polynomial of degree 1. Dotted line: the improved
Ridley model with F f ( s ) not approximated. Note that the
solid and dashed lines coincide and that the dash-dotted and
dotted lines almost coincide in the phase shift plot. Note also
that the dashed, dash-dotted, and dotted lines almost coincide
in the magnitude plot.

where F f ( s ) is defined in (6.120). A second step to approximate k f ( s ) is to

approximate F f ( s ) .
F f ( s ) is approximated by a Taylor polynomial, i.e. a truncated version
of the Taylor series in (6.120). Figure 6.9 is a modified version of Figure 5.3.
It shows the Bode plot for the audio susceptibility predicted by the Ridley
model (3.94) together with the simulation results. It also shows the audio
susceptibilities predicted by the improved Ridley model ((3.91) with k f
replaced by k f ( s ) in (5.27)) in the cases where F f ( s ) is not approximated,
Chapter 6. Approximations of Obtained Expressions 241

F f ( s ) is approximated by a Taylor polynomial of degree 0 and 1. From the

figure it is seen that the predictions of the improved Ridley model agree more
closely with the simulation results if the degree of the Taylor polynomial is
increased.
The phase shift curves for the Ridley model and the improved Ridley
model where F f ( s ) is approximated by a Taylor polynomial of degree 0 are
identical in Figure 6.9, but the magnitude curves are not. The reason for this
will now be explained since it may seem a little strange. The improved Ridley
model where F f ( s ) is not approximated is the same as the combined
expression in (5.21). The combined expression has only one zero (the ESR
zero) if F f ( s ) is approximated by a Taylor polynomial of degree 0. The
Ridley model (3.94) has two extra zeros and these are obtained by putting the
numerator in (3.94) equal to zero and the result is

R 2Ts  RD'+ Rc 
± jω n 1 + D ' 2  mc − 0.5  . (6.123)
(R + Rc )L  R 

Hence, the two extra zeros are complex conjugated and they are located on
the imaginary axis. The distance from the origin to these zeros are larger than
ω n if mc is so high that the current control is stable. Two extra zeros do
therefore not contribute to the phase shift at all for frequencies in the interval
[0, ω n ] . However, the magnitude is affected, especially at the higher
frequencies in the interval.

6.4 Approximate Model for the Buck-Boost

Converter
In this section, the transfer functions obtained by applying the
(improved) Ridley model to the buck-boost converter are approximated. The
methodology is analogous to the one used for the boost converter.

The Denominator in the Transfer Functions

The transfer functions obtained by applying the Ridley model to the

buck-boost converter have the same denominator and it is (see (3.113),
(3.35), and (3.12))
242 Chapter 6. Approximations of Obtained Expressions

den( s ) =
Ts mc RD'+ Rc   RD '+ Rc  
 RD ' 
 + sRc C  + sL(1 + s(R + Rc )C ) −
L R + Rc   
 R + Rc  
RD '3 Ts
(1 + sRc C ) 1 − s LD2  +
(6.124)
2L  RD' 
 RD ' (1 + D ) + Rc RD'+ Rc 
H e ( s ) + sC .

 (R + Rc )D' D' 

where

s s2
H e (s) = 1 + + 2 , (6.125)
ω nQz ω n

−2
Qz = . (6.126)
π

An approximate denominator will now be derived for the case where mc

is so low that the low-frequency pole has a frequency that is much lower than
the frequencies of the two high-frequency poles.
Assume that

Rc << RD' . (6.127)

Since D' is a positive number less than or equal to unity, the following is
obtained by using (6.127):

Rc << RD ' ≤ R . (6.128)

The denominator (6.124) is rewritten by using (6.128) and (3.54):

Chapter 6. Approximations of Obtained Expressions 243

Ts m c
den( s ) ≈ D ' (RD' (D '+ sRc C ) + sL(1 + sRC )) −
L
RD '3 Ts
(1 + sRc C ) 1 − s LD2  +
2L  RD ' 
 2 
1 + s + s − sTs mc D' (1 + D + sRC ) =
 ω nQ ω 2 
 n 
Ts m c RD '3 Ts
RD '3 − +1+ D +
L 2L
 Ts m c RD '3 Ts (6.129)
 RD' 2 Rc C + Ts mc D '− Rc C +
 L 2 L

DD ' Ts 1 + D 
+ − (1 + D )Ts mc D'+ RC  s +
2 ω nQ 
 Ts m c DD' Ts 1 + D RC  2
 D' LRC + Rc C + + − T m D ' RC s +
 L 2 ω 2 ω Q
s c 
 n n 
1
( )
RCs 3 = K −1 1 + a1 s + a 2 s 2 + a 3 s 3 = K −1 P ( s ) ,
ωn 2

where

1
K= 3
,
RD ' Ts (6.130)
1+ D + (mc − 0.5)
L

P ( s ) = 1 + a1 s + a 2 s 2 + a 3 s 3 , (6.131)

a1 =
1+ D  R D ' 2 Ts  D'    (6.132)
K + RC 1 + c  mc −   − DD' Ts (mc − 0.5) ,
 ω nQ  L  2   
  
244 Chapter 6. Approximations of Obtained Expressions

1 + D RC DD ' Ts 
a2 = K  2 + + Rc C  , (6.133)
 ω ω nQ 2 
 n 

RC
a3 = K . (6.134)
ω n2

Since the derivation of the approximate denominator is made with the

assumption that the low-frequency pole has a frequency that is much lower
than the frequencies of the two high-frequency poles, (6.19) is used to
approximate the denominator (6.129):

 a a  
den( s ) ≈ K −1 P ( s ) ≈ K −1 (1 + a1 s ) 1 +  2 − 32  s + a3 s 2  . (6.135)
  a  a1 
  1 a1  

The following three assumptions are made before continuing:

Rc D ' 2 Ts  D' 
 mc −  << 1 , (6.136)
L  2

DD ' Ts (mc − 0.5) << RC , (6.137)

1+ D
<< RC . (6.138)
ωnQ

This means that a1 can be approximated with

a1 ≈ RCK . (6.139)

The denominator (6.135) is approximated by using (6.139), (6.133), and

(6.134):
Chapter 6. Approximations of Obtained Expressions 245

den( s ) ≈ K −1 (1 + RCKs ) •
  1+ D 1 Rc DD ' Ts 1   (6.140)
1 +  + + − s + 1 s 2  .
  RCω 2 ω n Q 2R RCKω n2  ω2 
  n  n 

One part of (6.140) is rewritten as follows by using (6.130) and (3.13):

1+ D 1 R DD ' Ts 1
+ + c − =
RCω n2 ω nQ 2R RCKω n2
1 1  1  Rc DD ' Ts
+ 1 + D −  + =
ω n Q RCω n2  K 2R
  3 
1
+
1 1 + D − 1 + D + RD ' Ts (m − 0.5)  +
ω n Q RCω n2   L
c  (6.141)
  
Rc DD ' Ts 1 D'3 π R DD' π
= − (mc − 0.5) + c =
2R ω n Q LCω n ω n
2 2R ω n

1  1 πDD ' Rc 
2
ω 
− πD '3 (mc − 0.5)  0  + ,
ωn Q 2 R 
  ωn  

where

1
ω0 = . (6.142)
LC

Assume that

πDD ' Rc 1
<< , (6.143)
2 R Q

2
 ω0  1
πD '3 (mc − 0.5)   << . (6.144)
ω
 n Q
246 Chapter 6. Approximations of Obtained Expressions

The denominator (6.140) can now be approximated by using (6.141),

(6.143), and (6.144):

den( s ) ≈ K −1 (1 + sRCK )Fh−1 ( s ) , (6.145)

where Fh ( s) is defined in (6.2).

One of the assumptions used in the derivation of the approximate
denominator (6.145) is that the low-frequency pole has a frequency that is
much lower than the frequencies of the two high-frequency poles. Conditions
that ensure that this assumption is fulfilled are (6.29) and (6.30), i.e. the
conditions derived for the buck converter. The same conditions can be used
for the buck-boost converter since the approximate denominators are exactly
the same (see (6.28) and (6.145)) except for the definition of K .
From (6.140), it is seen that K is a positive number less than 1 (1 + D ) .
The following is obtained if this result is combined with (6.30):

1+ D
<< (1 + D )RCK < RC . (6.146)
ωnQ

Hence, the condition (6.30) ensures that the condition (6.138) is fulfilled.
The results are now summarized. The denominator in the transfer
functions obtained by applying the Ridley model to the buck-boost converter,
i.e. (6.124), is approximated with

den( s ) = K −1 (1 + sRCK )Fh−1 ( s ) , (6.147)

where

1
K= 3
,
RD ' Ts (6.148)
1+ D + (mc − 0.5)
L

and Fh ( s) is defined in (6.2), if

Rc << RD' , (6.149)

Chapter 6. Approximations of Obtained Expressions 247

Rc D ' 2 Ts  D' 
 mc −  << 1 , (6.150)
L  2

DD ' Ts (mc − 0.5) << RC , (6.151)

πDD ' Rc 1
<< , (6.152)
2 R Q

2
 1 LC 
πD ' (mc − 0.5) 
3  << 1 , (6.153)

 ωn  Q

1
<< ω n , (6.154)
RCK

and

1
<< ω n Q , (6.155)
RCK

where Q is defined in (6.3).

The Bode plots for the denominators (6.124) and (6.147) are shown in
Figure 6.10. The parameter values shown in Table 2.6 are used. From the
figure it is seen that the two denominators are almost the same. The
maximum difference is approximately 0.3 dB and 1 degree in the presented
frequency interval.

Control-to-Output Transfer Function

An approximate version of the control-to-output transfer function

obtained by applying the Ridley model to the buck-boost converter is

 LD 
RD ' (1 + sRc C ) 1 − s 
ˆv o ( s )  RD ' 2  (6.156)
= ,
iˆc ( s ) den( s )
248 Chapter 6. Approximations of Obtained Expressions

40

30

20
Phase (deg); Magnitude (dB)

10

0
1 2 3 4
10 10 10 10

200

150

100

50

0
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 6.10: The denominator in the transfer functions for a buck-boost

converter with a current controller. Dashed line: the
approximated denominator. Solid line: the Ridley model. Note
that the two lines almost coincide.

where den( s) is defined in (6.147) if the conditions in (6.149)-(6.155) are

fulfilled. The only difference compared to the non-approximated version
(3.110) is that the approximate denominator (6.147) is used instead of the
denominator (6.124).
(6.156) is rewritten by using (6.147):

 LD 
RD ' (1 + sRc C ) 1 − s 
vˆo ( s )  RD ' 2 
= = (6.157)
iˆc ( s ) K −1 (1 + sRCK )Fh−1 ( s )
RD' KFl ( s) FESR ( s) Fh ( s ) FRHP ( s ) ,

where
Chapter 6. Approximations of Obtained Expressions 249

1
Fl ( s) = , (6.158)
1 + sRCK

LD
FRHP ( s ) = 1 − s , (6.159)
RD ' 2

FESR ( s) is defined in (6.42), Fh ( s) is defined in (6.2), and K is defined in

(6.148).

Output Impedance

The output impedance obtained by applying the Ridley model to the

buck-boost converter is given by (3.111). The numerator in (3.111) will first
be approximated. By using (6.149) and the fact that D and D' are positive
numbers less than or equal to unity, the following is obtained:

Rc
Rc DD ' << Rc DD' ≤ Rc D ' ≤ Rc << RD' ≤ R . (6.160)
R

The numerator in (3.111) is approximated by using (6.160):

num( s ) = R(1 + sRc C ) •

 R 2 − Rc2 D Ts m c DD' 2 Rc Rc s s2  (6.161)
 + + + + + sT m D ' .
 2 L RD ' ω Q ω 2 s c 
 R n z n 

One part of (6.161) is approximated by using (2.39) and (6.160):

R 2 − Rc2 D Ts mc DD ' 2 Rc R
2
+ + c =
R L RD'
R
RD '− Rc DD ' c + Rc
R D ' 2 Ts (6.162)
R + c m c (1 − D ') ≈
RD ' L
Rc D ' 2 Ts
1+ (mc − mc D') .
L
250 Chapter 6. Approximations of Obtained Expressions

The second term in (6.162) is approximated by using (6.150) and the fact
that mc is a positive number greater than or equal to unity:

Rc D ' 2 Ts Rc D ' 2 Ts  1 
(mc − mc D') ≤  mc − D '  << 1 . (6.163)
L L  2 

Hence, (6.162) is approximately equal to unity. The numerator (6.161) can

therefore by approximated with:

num( s ) ≈ R(1 + sRc C )Fh−1 ( s ) (6.164)

according to (3.54).
An approximate version of the output impedance obtained by applying
the Ridley model to the buck-boost converter is

R(1 + sRc C )Fh−1 ( s)

Z out ( s ) = , (6.165)
den( s )

where den( s) is defined in (6.147) and Fh ( s) is defined in (6.2) if the

conditions in (6.149)-(6.155) are fulfilled. The only differences compared to
the non-approximated version (3.111) are that the approximate denominator
(6.147) is used instead of the denominator (6.124) and that the approximate
numerator (6.164) is used instead of the numerator in (3.111).
(6.165) is rewritten by using (6.147):

R (1 + sRc C )Fh−1 ( s )
Z out ( s ) = = RKFl ( s ) FESR ( s ) , (6.166)
K −1 (1 + sRCK )Fh−1 ( s )

where K is defined in (6.148), Fl ( s) is defined in (6.158), and FESR ( s ) is

defined in (6.42). The two high-frequency poles are cancelled by two of the
zeros.

Audio Susceptibility

The expression for the audio susceptibility obtained by applying the

Ridley model to the buck-boost converter was improved in Chapter 5 by
Chapter 6. Approximations of Obtained Expressions 251

replacing the numerator (see (5.41)). The improved expression then became
the same as the combined expression (5.38). To approximate the numerator
in (5.38), an inequality is needed and this will first be derived.
Let the function f ( z ) be defined by

 z  z   12
f ( z ) =  z − 1 −  . (6.167)
 e −1  2   z 2

Figure 6.11 shows the value of the function f ( z ) in the case where the
domain of z is the interval ] 0, jπ ] , i.e. a part of the imaginary axis. From
the figure it is seen that f ( z ) is approximately equal to unity in this domain.
Let s = jω . If 0 < ω ≤ ω n , then sTs is in the interval ] 0, jω n Ts ] which
is rewritten as ] 0, jπ ] by using (3.13). f ( sTs ) is therefore approximately
equal to unity if 0 < ω ≤ ω n and this result is independent of Ts . The result
is rewritten as:

 sTs  sTs   1 1
 −  1 −   ≈ , (6.168)
 e s −1  2   s 2Ts2 12
sT

  T  1 DTs
D H e ( s ) − 1 − s s   ≈ s, (6.169)
  2   sTs 12

where H e ( s ) is defined in (3.10) if 0 < ω ≤ ω n . (6.169) is valid also in the

case where s is equal to zero since the left side (excluding the absolute value
operator) is a polynomial in s with s n terms where n ≥ 1 (see (4.43)). The
Ridley model is designed to be accurate only from dc to half the switching
frequency, ω n , so the condition 0 ≤ ω ≤ ω n should be understood tacitly
from now on. By using (6.169) and the fact that 3 − 2 D is greater than or
equal to unity, the following inequality is obtained:

  T  1 DTs (3 − 2 D )DTs
D H e ( s ) − 1 − s s   ≈ s≤ s. (6.170)
  2   sTs 12 12

The numerator in (5.38) will now be approximated. The s 0 term is given

by (5.32) and it is approximated by using (6.149) (which gives (6.128)):
252 Chapter 6. Approximations of Obtained Expressions

1.5

1
f(z)

0.5

0
0 0.5 1 1.5 2 2.5 3
Im(z)

Figure 6.11: The value of the function f ( z ) (see (6.167)) in the case where
the domain of z is the interval ] 0, jπ ] , i.e. a part of the
imaginary axis.

RTs  RD '+ Rc  D  D
D '  mc − 1 −   + ≈
L  R + Rc  2   D'
(6.171)
RTs   D  D
D '  mc D '−1 −   + .
L   2   D'

The other terms are given by (5.34) and they are approximated by using
(6.149) (which gives (6.128)):
Chapter 6. Approximations of Obtained Expressions 253

RTs  D ' (R + R c )   D 
D '  −  F f 1 ( s ) − 1 −   +
L  RD'+ Rc   2 
 D ' (R + R c )    T   1 
1 −   H e ( s) − 1 − s s   +
 RD '+ R c   2   sTs 
 
D ' (R + R c ) D
RD '+ R c D'
(
F f 2 (s) − 1 ≈)
RTs
D' • (6.172)
L

  D (3 − 2 D )DTs
−
   1 − − s −
(
1 − 2 D + D 2 DTs2 2 )
 D 
s + K − 1 −   +
  2 12 24  2  


  Ts   1  D
Rc

D H e ( s ) − 1 −
RD' 
s  +
2   sTs  D'
(
F f 2 (s) − 1 , )



where F f 1 ( s) is defined in (4.60), F f 2 ( s ) is defined in (4.61), and H e ( s )

is defined in (3.10). (6.172) is approximated by using (6.149) and (6.170)
and the result is:

RTs   D  D

D'  −  F f 1 ( s ) − 1 −    + ( )
F f 2 (s) − 1 . (6.173)
L    2    D'

An approximate version of the numerator in (5.38) is obtained by adding

(6.171) and (6.173):

RTs   D   D 
D'  mc D '−1 −  −  F f 1 ( s ) − 1 −    +
L   2   2 
D D
+
D' D'
( )
F f 2 ( s) − 1 = (6.174)

RTs
L
(
D' mc D '− F f 1 ( s ) + )
D
D'
F f 2 (s) .
254 Chapter 6. Approximations of Obtained Expressions

The expression for the audio susceptibility obtained by applying the

Ridley model to the buck-boost converter was improved in Chapter 5. An
approximate version of the improved expression is

 RTs 
vˆo ( s)  L
 (
D' mc D '− F f 1 ( s ) + )
D
F f 2 ( s )  D (1 + sRc C )
D'  (6.175)
= ,
vˆ g ( s ) den( s)

where

1  sTs 1 − e − sDTs sTs 

F f 1 ( s) =  − =
sTs  1 − e − sTs sDT e sTs − 1 
 s
(6.176)
 D  (3 − 2 D )DTs
1 −  − s−
(
1 − 2 D + D 2 DTs2 2
s +K,
)
 2 12 24

F f 2 (s) =
1 1 − e − sDTs
= 1 +
D ' Ts
s +
(
1 − 3D + 2 D 2 Ts2 2
s +K,
) (6.177)
D 1 − e − sTs 2 12

(see (4.60) and (4.61)), and den( s) is defined in (6.147) if the conditions in
(6.149)-(6.155) are fulfilled. The only differences compared to the non-
approximated improved version (5.38) are that the approximate den( s) in
(6.147) is used instead of den( s) in (6.124) and that the numerator in (5.38)
is replaced by its approximate version (6.174).
(6.175) is rewritten by using (6.147):

 RTs 
 ( )
D ' mc D '− F f 1 ( s ) +
D
F f 2 ( s )  D (1 + sRc C )
vˆo ( s)  L D' 
= =
vˆ g ( s ) K −1 (1 + sRCK )Fh−1 ( s) (6.178)
 RTs D ' 
 (
mc D'− F f 1 ( s ) +
D
)
F f 2 ( s )  DKFl ( s ) FESR ( s ) Fh ( s) ,
 L D' 

where F f 1 ( s) is defined in (6.176), F f 2 ( s ) is defined in (6.177), K is

defined in (6.148), Fl ( s) is defined in (6.158), FESR ( s ) is defined in (6.42),
and Fh ( s) is defined in (6.2).
Chapter 6. Approximations of Obtained Expressions 255

k f ( s ) defined in (5.44) will now be approximated. The s 0 term of the

numerator of k f ( s ) is written as

DTs Ri D' (R + Rc )   D  DRc  D  

−  1 −  + 1 −   −
L RD'+ Rc    2  D ' (R + Rc )  2 
D 2 Ri D' (R + Rc )  RD '+ Rc DRc 
 −1 − =
2 
RD '+ Rc  D ' (R + Rc ) D' (R + Rc ) 
RD '
DTs Ri D' (R + Rc ) (1 − D )(R + Rc ) + DRc  D 
− 1 −  − (6.179)
L RD'+ Rc D ' (R + Rc )  2
D 2 Ri D' (R + Rc )  R (1 − D ) + Rc − (1 − D )(R + Rc ) − DRc

 =
 D ' (R + R c ) 
RD ' 2 RD '+ Rc  
DTs Ri D' (R + Rc ) RD'+ Rc  D  DTs Ri  D 
− 1 −  = − 1 −  .
L RD'+ Rc D ' (R + R c )  2 L  2

The other terms of the numerator of k f ( s ) are approximated by using

(6.149) (which gives (6.128)):
256 Chapter 6. Approximations of Obtained Expressions

 D ' (R + Rc ) 
DTs Ri  D 

−  RD'+ R  F f 1 ( s ) − 1 − 2   −
L  c  
 D ' (R + R c )    T   1 
1 −   H e ( s ) − 1 − s s   −
 RD '+ R c   2   sTs 
 
D 2 Ri D ' (R + Rc )  RD'+ Rc  s s 2  
2
+
RD'+ Rc  D ' (R + Rc )  ω n Q z ω n 
2
(
− F f 2 (s) − 1  ≈

)
RD '  
DTs Ri
− • (6.180)
L
  D (3 − 2 D )DT
 1 − − s
s−
(
1 − 2 D + D 2 DTs2 2)  D 
s + K − 1 −   −
 2 12 24  2  

Rc   T   1 
D H e ( s) − 1 − s s   −
RD'   2   sTs 

D 2 Ri   s s 2  
2  ω Q
+ 2 
(
− F f 2 (s) − 1  ,

)
RD '   n z ω n  

where F f 1 ( s) is defined in (6.176), F f 2 ( s ) is defined in (6.177), and

H e ( s ) is defined in (3.10). (6.180) is approximated by using (6.149) and
(6.170) and the result is:

  D   D Ri
2
−
DTs Ri
 F f 1 ( s ) − 1 −   − (
H e (s) − F f 2 (s) ,) (6.181)
L   2   RD ' 2

where H e ( s ) is defined in (3.14). An approximate version of k f ( s ) is

obtained by replacing the numerator in (5.44) with the sum of (6.179) and
(6.181) and the result is:

−1
 DTs Ri D 2 Ri 
k f (s) =  − F f 1 (s) − (
H e (s) − F f 2 (s)  1 − s LD  ,
)
 L RD' 2  RD' 2  (6.182)
 

where F f 1 ( s) is defined in (6.176), F f 2 ( s ) is defined in (6.177), and

H e ( s ) is defined in (3.14). A second step to approximate k f ( s ) is to
approximate F f ( s ) and F f 2 ( s ) .
Chapter 6. Approximations of Obtained Expressions 257

-10

-20
Phase (deg); Magnitude (dB)

-30

-40
1 2 3 4
10 10 10 10

-50

-100

-150
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 6.12: The audio susceptibility of a buck-boost converter with a current

controller. X: the simulation results. Solid line: the Ridley
model. Dashed line: the improved Ridley model with F f 1 ( s)
and F f 2 ( s ) approximated by a Taylor polynomials of degree 0.
Dash-dotted line: the improved Ridley model with F f 1 ( s) and
F f 2 ( s ) approximated by a Taylor polynomials of degree 1.
Dotted line: the improved Ridley model with F f 1 ( s) and
F f 2 ( s ) not approximated.

F f 1 ( s) and F f 2 ( s ) are approximated by Taylor polynomials, i.e.

truncated versions of the Taylor series in (6.176) and (6.177). Figure 6.12 is a
modified version of Figure 5.6. It shows the Bode plot for the audio
susceptibility predicted by the Ridley model (3.112) together with the
simulation results. It also shows the audio susceptibilities predicted by the
improved Ridley model ((3.109) with k f replaced by k f ( s ) in (5.44)) in
the cases where F f 1 ( s) and F f 2 ( s ) are not approximated, F f 1 ( s) and
F f 2 ( s ) are approximated by a Taylor polynomials of degree 0 and 1. From
the figure it is seen that the predictions of the improved Ridley model agree
more closely with the simulation results if the degree of the Taylor
polynomials is increased.
258 Chapter 6. Approximations of Obtained Expressions

6.5 Summary and Concluding Remarks

In this chapter, the expressions obtained from the (improved) Ridley
model were approximated. To derive these approximations, we assumed that
the distance from the origin to the low-frequency pole is much smaller than
the distance from the origin to the two high-frequency poles. We also made
some other assumptions to obtain simple expressions.
Since the denominators for the control-to-output transfer function,
output impedance, and audio susceptibility are the same (for each specific
converter), we approximated the common denominator only once (for each
specific converter). The most complicated numerators were simplified the
most. Therefore, the approximated numerators are approximately equally
simple. The expressions for the audio susceptibility depend on F f ( s )
( F f 1 ( s) and F f 2 ( s ) ). The expressions for F f ( s ) were therefore also
approximated.
 Chapter 7 Using Load Current
for Control

The output voltage and the inductor current are measured in the case
where current-mode control is utilized. In this chapter, some properties that
can be obtained when the controller also utilizes load current measurements
are analyzed. The results of this analysis are compared with simulation results.
The analysis in this chapter is partly based on the approximate expressions
obtained in Chapter 6. This may result in unreliable analysis (see Section
7.7).

7.1 Chapter Survey

Some of the previous works made in this area are reviewed in Section 7.2.
In Section 7.3, a simple model of the buck converter with current-mode
control is used to give a simple explanation of some of the properties that are
obtained when using load current measurements to control the converter. In
Section 7.4, the model obtained in Chapter 6 for the buck converter with
current-mode control is used to analyze the properties and the results are
compared with simulation results. The same is made for the boost and buck-
boost converters in Section 7.5 and Section 7.6, respectively. A summary and
concluding remarks are presented in Section 7.7.

7.2 A Review
A number of papers suggest that the load current should be measured in
order to improve the control of dc-dc converters. A few of them are
mentioned in this section.
Redl and Sokal (1986) show that the transient in the output voltage due

259
260 Chapter 7. Using Load Current for Control

vo(t)
vg(t)
i (t)
Converter load
vref(t) Voltage ic2(t) Controller ic(t) Current δ (t)
using load iL(t)
controller controller
current

Figure 7.1: The configuration of the controller.

to a step change in the load can be much reduced if the load current is
measured and used to control the converter. They only consider the case
where current-mode control is used. The configuration of the controller is
shown in Figure 7.1. The inductor current, i L (t ) , is fed back in the inner
loop. The inner loop consists of the current controller and the converter. The
load current, iload (t ) , is used in the middle loop. (The input voltage, v g (t ) ,
and output voltage, v o (t ) , are in some cases also used in the middle loop.)
The middle loop consists of a controller and a process where the process is the
closed inner loop. The output voltage, v o (t ) , is fed back in the outer loop.
The outer loop consists of a voltage controller and a process where the process
is the closed middle loop. The control method includes an extra middle
controller compared to current-mode control (compare Figure 7.1 and Figure
3.1). The control signal of the voltage controller is now called ic 2 (t ) and it
affects the middle controller. The reference signal for the current controller is
still called ic (t ) . The middle controller should be as follows according to
Redl and Sokal (1986).

Buck: ic (t ) = ic 2 (t ) + iload (t ) , (7.1)

v o (t )
Boost: ic (t ) = ic 2 (t ) + iload (t ) , (7.2)
v g (t )

v o (t ) + v g (t )
Buck-boost: ic (t ) = ic 2 (t ) + iload (t ) . (7.3)
v g (t )
Chapter 7. Using Load Current for Control 261

The numerator in (7.3), v o (t ) + v g (t ) , is actually not what is proposed by

Redl and Sokal (1986). Instead, v o (t ) − v g (t ) is used but in Kislovski, Redl
and Sokal (1991, Section 11.2) v o (t ) + v g (t ) is used. If v o (t ) − v g (t ) is
used as the numerator in (7.3), then (7.3) does not make sense with the
definitions of signs of voltages and currents made in this thesis.
Schoneman and Mitchell (1989) analyze the proposed use of load current
further in the case of a buck converter, i.e. (7.1).
Redl and Sokal (1986) suggest that the load current is not measured
directly but calculated indirectly. For example, in a buck converter, the
inductor current, i L (t ) , and the current to the output capacitor, icap (t ) , are
measured. The load current can then be calculated as the difference (see
Figure 2.1):

iload (t ) = i L (t ) − icap (t ) . (7.4)

Note that the inductor current is measured in current-mode control so it is

still only necessary to measure one extra current compared to current-mode
control.
Schoneman and Mitchell (1989) propose an alternative approach. From
Figure 3.4, it is seen that the difference ic (t ) − i L (t ) is calculated in the
current controller. This difference is rewritten by using (7.1) and (7.4):

ic (t ) − i L (t ) = ic 2 (t ) + iload (t ) − i L (t ) =
(7.5)
ic 2 (t ) + i L (t ) − icap (t ) − i L (t ) = ic 2 (t ) − icap (t ) .

Hence, it is not necessary to measure the inductor and load currents. Only
the current to the output capacitor must be measured.
Hiti and Borojevic (1993) use the measured load current to modify the
current-mode control for the boost converter. The modification is made in
such a way that the dc gain of the closed middle loop in Figure 7.1 (i.e. the
transfer function that describes how v o (t ) is affected by ic 2 (t ) ) is
independent of the load. The modification turns out to be the same as the
one presented for the boost converter by Redl and Sokal (1986), i.e. (7.2).
262 Chapter 7. Using Load Current for Control

Controller

r u y
C1 Process

C2

Figure 7.2: The basic configuration of the controller.

7.3 Principal Properties

In this section, some properties obtained by using measured load current
are explained by using a simple model of the buck converter with current-
mode control. The properties are:

• Low output impedance.

• An almost invariant control-to-output transfer function for different
loads.
• Risk of instability.

However, two concepts in control theory, feedforward and gain scheduling,

are first reviewed.

Feedforward and Gain Scheduling

Figure 7.2 shows the configuration of the controller used as a base here.
The control signal of the controller is called u and it controls the input signal
of the process that should be controlled. The output signal of the process is
called y and it is fed back to the controller. The reference signal of the
controller is called r . The controller is very general since the compensators
C1 and C 2 can be chosen independently.
The definition of feedforward is that a disturbance signal is measured and
used to make a change in the control signal before the disturbance has caused
Chapter 7. Using Load Current for Control 263

Controller with
feedforward v
y
r u Process
C1

Cff
C2

Figure 7.3: A controller with feedforward.

any change in the output of the process (Åström and Hägglund, 1995,
Section 7.3). Figure 7.3 shows an example where the controller in Figure 7.2
is extended to also include feedforward. The disturbance signal is called v
and it is multiplied with C ff to obtain the contribution to the control signal
from the feedforward part. Feedforward does not cause any stability problems
(Åström and Hägglund, 1995, Section 7.3). A consequence of this is that, for
linear systems, feedforward does not affect the loop gain.
Before continuing with gain scheduling, some more examples of
feedforward will be considered. One input signal to the middle controller in
Figure 7.1 is (in some cases) the input voltage, v g (t ) , and it is a disturbance
signal. If Figure 7.1 is compared with Figure 7.3, it is clear that the use of the
measured input voltage in the controller should be called feedforward.
Another input signal to the middle controller is the load current, iload (t ) .
Whether the use of the measured load current should be called feedforward or
not will be analyzed later in this section. Yet another input signal to the
middle controller is (in some cases) the output voltage, v o (t ) . The output
voltage depends on the state variables in the controller and it should therefore
not be regarded as a disturbance signal. The use of the measured output
voltage should therefore not be called feedforward. Ridley (1990a) uses the
expression “feedforward gains k f and kr ”. From Figure 3.8 and the previous
discussion, it is concluded that it is correct to use the expression “feedforward
gain k f ” but not the expression “feedforward gain kr ”. Both of these
expressions are used in this thesis even though it is not formally correct.
264 Chapter 7. Using Load Current for Control

Gain-scheduling
controller
y
r u
C1 Process y
2

C2

Pcal Gain
schedule

Figure 7.4: A gain-scheduling controller.

A definition of gain scheduling is found in Åström and Wittenmark

(1995, Chapter 9). If the operating conditions of a process vary with time and
the controller that controls the process is time invariant, the dynamics of the
closed loop system also vary with time. Assume that some variables that reflect
the operating conditions of a process are measured. If these variables are used
to change the values of the parameters in the controller, it should be possible
to reduce the influence of changed operating conditions on the dynamics of
the closed loop system. This method is called gain scheduling and an example
is shown in Figure 7.4. Pcal is a time-varying parameter in the controller. It is
calculated by using the gain schedule and the scheduling variables, which are
the input signals to the gain schedule. The scheduling variables should reflect
the operating conditions of the process. The signal y 2 is an extra output of
the process that is measured and used as a scheduling variable. The signal y
that is measured and fed back to the original controller could also be used as a
scheduling variable if it reflects the operating conditions of the process in
some way.

Low Output Impedance

The current through the inductor is controlled in current-mode control.

If this control has high bandwidth, the inductor can be approximated by a
Chapter 7. Using Load Current for Control 265

iload(t)

Vref Voltage ic(t) ic(t) Load vo(t)

C
controller

Figure 7.5: A simple model of the buck converter with current-mode control.

iload(t)

Vref Voltage ic2(t) ic(t)

ic(t) C Load vo(t)
controller
iload(t)

Figure 7.6: The measured load current affects the control signal.

controlled current source. The buck converter with current-mode control can
then be modeled as in Figure 7.5 (compare with Figure 1.2). The current
from the current source is equal to the reference signal for the current
controller, ic (t ) . The ESR of the capacitor is neglected in Figure 7.5.
Figure 7.6 shows how the measured load current should be used to
control the buck converter according to (7.1). If iload (t ) changes, ic (t )
changes by the same amount, provided that ic 2 (t ) is constant, e.g. when the
voltage controller is disabled. Consequently, the capacitor current does not
change. The conclusion is that changes in load current do not affect the
output voltage, that is, the output impedance is zero:

vo ( s )
Z out ( s ) = − = 0. (7.6)
iload ( s )
266 Chapter 7. Using Load Current for Control

The output impedance is zero also in the case where the voltage controller is
enabled since a load change does not cause a change in the output voltage
according to (7.6). Note that the definition of output impedance used in
(7.6) is not the same as the one used elsewhere in this thesis, e.g. (2.91).
According to Erickson and Maksimovic (2000, preamble of Chapter 8), the
load resistance, R , can either be included in the output impedance as in
(2.91) (see Figure 2.6) or excluded as in (7.6).

An Almost Invariant Control-to-Output Transfer Function

for Different Loads

In this subsection, the control-to-output transfer function that describes

how ic 2 (t ) affects v o (t ) is analyzed. Two different types of loads are
considered. The first type is a current source, i.e. the load current is
independent of the output voltage. The second type is a linear resistive load.
First assume that the load is a current source. If the load current is not
used, ic (t ) is equal to ic 2 (t ) and the output voltage is

1
vo (s) = (ic 2 ( s) − iload (s)) . (7.7)
sC

Since iload (t ) is determined by the current source, it can be regarded as an

input signal in addition to ic 2 (t ) . The control-to-output transfer function
that describes how ic 2 (t ) affects v o (t ) is therefore obtained if iload (t ) is set
to zero:

v o ( s) 1
Gvoic 2 ( s ) = = . (7.8)
ic 2 ( s ) iload ( s ) =0
sC

If the load current is used, ic (t ) is according to (7.1) and the output voltage
is

1
vo (s) = (ic (s) − iload ( s) ) =
sC
(7.9)
1
(ic 2 ( s) + iload (s) − iload ( s)) = 1 ic 2 (s)
sC sC
Chapter 7. Using Load Current for Control 267

and Gvoic 2 ( s) is

vo (s) 1
Gvoic 2 ( s ) = = . (7.10)
ic 2 ( s ) sC

It is seen from (7.8) and (7.10) that the control-to-output transfer function
Gvoic 2 ( s) does not change when the measured load current is introduced for
control. Therefore, the loop gain and the stability properties do not change.
Since the load current does not depend on the states of the converter, it can
be seen as a disturbance signal. The conclusion is that the use of measured
load current for control is feedforward in the case where the load is a current
source.
Now assume that the load is a resistor with resistance R . If the load
current is not used, Gvoic 2 ( s) is

1
R
v ( s) sC .
Gvoic 2 ( s ) = o = (7.11)
ic 2 ( s ) 1
R+
sC

If the load current is used, the output voltage is given by (7.9) and Gvoic 2 ( s)
is given by (7.10). It is seen from (7.11) and (7.10) that Gvoic 2 ( s) changes
from the impedance of the parallel of the capacitor and resistor to just the
impedance of the capacitor when the measured load current is introduced for
control. Since R is not present in (7.10), the use of the load current makes
the control-to-output transfer function Gvoic 2 ( s) invariant for different values
of R , i.e. different linear resistive loads.
If the load is a resistor and the measured load current is introduced for
control, the control-to-output transfer function Gvoic 2 ( s) changes, as shown
above, and can also become unstable, as will be shown in the next subsection.
The conclusion is that the use of measured load current for control is not
feedforward in the case where the load is a resistor. It can instead be seen as
gain scheduling, as now will be shown. Figure 7.7 is a modified version of
Figure 7.1. The closed inner loop in Figure 7.1 is regarded as the process to
be controlled in Figure 7.7. The controller in Figure 7.7 controls the input
signal of the process, ic (t ) . The output signals of the process, v o (t ) and
iload (t ) , are measured and used by the controller. The controller consists of
268 Chapter 7. Using Load Current for Control

Controller Process

Gain-scheduling
voltage controller vo

(Outer) i
Vref
voltage
c2 ic Current δ iload
C1 Converter
controller controller
iL

C2

Rcal Gain
schedule

Figure 7.7: A configuration of the controller where the gain-scheduling

property is emphasized.

two voltage controllers in cascade: an outer voltage controller and a gain-

scheduling voltage controller. The outer voltage controller then controls a
“process” which is the closed loop of the gain-scheduling voltage controller
and the process. Assume that the gain-scheduling voltage controller is
designed just to make this “process” invariant for different linear resistive
loads. Assume further that the outer voltage controller is designed thereafter
to obtain the wanted properties of the whole closed loop system. If the
designed gain-scheduling voltage controller is

C1 (t ) = 1 , (7.12)

−1
C 2 (t ) = , (7.13)
Rcal (t )
Chapter 7. Using Load Current for Control 269

v o (t )
Rcal (t ) = , (7.14)
iload (t )

then the control signal is

1
ic (t ) = C1 (t )ic 2 (t ) − C 2 (t )vo (t ) = ic 2 (t ) + v o (t ) =
Rcal (t )
1 (7.15)
ic 2 (t ) + v (t ) = ic 2 (t ) + iload (t ) .
v o (t ) o
iload (t )

Hence, the gain-scheduling voltage controller is exactly the same as the

middle controller in Figure 7.1 for the buck converter, i.e. (7.1). The use of
measured load current for control can therefore be seen as gain scheduling.
It was shown previously that the control-to-output transfer function
Gvoic 2 ( s) is invariant for different linear resistive loads if (7.1) is used. The
designed gain-scheduling voltage controller, (7.12)-(7.14), is therefore
reasonable since it is equivalent to (7.1) and the purpose was to obtain
invariance for different linear resistive loads. The gain-scheduling voltage
controller can be designed by using model matching, but this will be shown
first in the next section.
The simple model in Figure 7.5 gives the following description of the
process in Figure 7.7 in the case where the load is a resistor:

1
R
v ( s) sC ,
Gvoic ( s ) = o = (7.16)
ic ( s ) 1
R+
sC

v o ( s)
iload ( s ) = . (7.17)
R

This is a first order system and the most natural choice for the state variable is
the voltage across the capacitor, which is the same as the output voltage,
v o (t ) . v o (t ) and iload (t ) are both measurements of this state, where the
latter one is scaled by the factor 1 R . If the controller knows the value of R ,
270 Chapter 7. Using Load Current for Control

it will control the process equally well with only iload (t ) measured compared
to if only v o (t ) is measured.
That both signals are measured and used by the controller is in the above
gain-scheduling approach interpreted as follows. v o (t ) is a measurement of
the state and is fed back to the controller. The controller shall try to control
the process such that v o (t ) is equal to reference signal Vref and it is
therefore a voltage controller. The value of the load resistance, R , will be a
parameter in the voltage controller. The operating conditions of the process
vary with time since the load resistance varies. To reduce the influence of the
changed dynamics of the process, the parameter R in the controller is
replaced by the time-varying parameter Rcal (t ) . It should be an estimate of
the load resistance. To be able to calculate this estimate, an extra variable that
reflects the operating conditions of a process must be measured and it is the
load current, iload (t ) . An estimate of the load resistance can now be
calculated with the gain schedule (7.14). iload (t ) and also v o (t ) are used as
scheduling variables. The calculated estimate, Rcal (t ) , is equal to the load
resistance, R , if there are no measurement errors.
An alternative interpretation of that v o (t ) and iload (t ) are both
measured, is that it is the output power, v o (t )iload (t ) , that is measured and
used to control the input power (Hiti and Borojevic, 1993).
Schoneman and Mitchell (1989) use a load consisting of both a resistor
and a current source. The load current is in this case dependent of the output
voltage. The authors say that the load current is fed forward, but this is thus
not correct, strictly speaking. They also claim that the “feedforward” does not
affect the loop gain. The reason for this erroneous conclusion is that the
authors at a point in the derivation neglect the changes of the current through
the resistor.

Risk of Instability

From Figure 7.6, it is seen that there is positive feedback in the load
current loop. This indicates that there can be a problem with the stability in
the case where the load is a resistor. To investigate the stability, Figure 7.6 is
generalized to obtain Figure 7.8, where the gain in the measurement of the
load current is H i . In a real system, H i is not exactly equal to 1, but the
measurement can be made such that H i is close to 1. The output voltage is
Chapter 7. Using Load Current for Control 271

iload(t)

Vref Voltage ic2(t) ic(t)

ic(t) C R vo(t)
controller

iload(t)
Hi

Figure 7.8: A measurement gain, H i , is introduced for the load current.

(ic (s) − iload ( s)) = 1  ic 2 (s) + H i o − o  .

1 v (s) v (s)
vo (s) = (7.18)
sC sC  R R 

Hence, the control-to-output transfer function Gvoic 2 ( s) is

1
vo (s) sC 1
Gvoic 2 ( s ) = = = . (7.19)
ic 2 ( s ) 1
(H i − 1)  1 
1− C  s + (1 − H i ) 
sRC  RC 

Gvoic 2 ( s) has a pole at

1
p1 = −(1 − H i ) . (7.20)
RC

If H i >1, the system is unstable since the pole is in the right half side in the
complex s-plane. If H i =1, the system has a pole in the origin and the system
acts as an integrator. If H i <1, the system is stable and the dc gain is
R (1 − H i ) , i.e. very high if H i is close to 1. The conclusion of all this is
that it is difficult or impossible to obtain a specific output voltage by
manually setting a value for ic 2 (t ) if H i is close to 1. Instead, an outer
voltage controller is used to set ic 2 (t ) and the system can be stabilized.
272 Chapter 7. Using Load Current for Control

Summary

Analysis was made in this section by using a simple model of the buck
converter with current-mode control. We concluded the following.

1. The output impedance can become zero if the measured load current is
used for control.

2. The following properties are obtained in the case where the load is a
current source:
• The use of measured load current for control is feedforward.
• The control-to-output transfer function does not change when this
feedforward is introduced.

3. The following properties are obtained in the case where the load is a
linear resistor:
• The control-to-output transfer function can change when the
measured load current is introduced for control.
• The converter can become unstable when the measured load current
is introduced for control.
• The control-to-output transfer function can be invariant for different
linear resistive loads if the measured load current is used for control.
• The use of measured load current for control is not feedforward. It
can instead be seen as gain scheduling.

Some properties that can be obtained in the case where the load is a linear
resistor and the controller uses load current measurements will be analysed
further in the next sections by using the models obtained in Chapter 6.

7.4 Properties of the Buck Converter

In Section 6.2, approximate expressions for the buck converter with
current-mode control were derived. In this section, these expressions are used
to analyze how the control-to-output transfer function, the output impedance
and the audio susceptibility are affected when using load current
measurements to control the converter. An expression for the output voltage
is first derived and this is then used to derive the three transfer functions. The
results are compared with simulation results and the used simulation models
are presented at the end of this section. In this section, it is assumed that the
Chapter 7. Using Load Current for Control 273

conditions in (6.34)-(6.38) are fulfilled since the approximate expressions

derived in Section 6.2 may not be valid otherwise.

An Expression for the Output Voltage

To make the analysis in this section general, a transfer function, H i (s ) ,

is introduced in (7.1):

iˆc ( s) = iˆc 2 ( s ) + H i ( s )iˆload ( s ) . (7.21)

H i (s ) can represent the dynamics of a filter, which filters the signal from the
load-current sensor, and also the sensor itself. From
Figure 2.6, it is seen that the load current is

v o (t )
iload (t ) = + iinj (t ) . (7.22)
R

Figure 7.9 shows the system obtained by using (7.21) and (7.22). In Chapter
3, the subscript ol was introduced to denote the converter transfer functions,
i.e. the open loop system. This system is controlled by changing the duty
cycle, δ (t ) . If the inductor current is fed back, a new system is obtained
which is controlled by changing ic (t ) . The transfer functions for this new
system will be denoted with the subscript ol 2 . However, it is not used for
the control-to-output transfer function since there is no risk of confusion in
that case. Since linear models are used, the output voltage is obtained by
adding the contribution from each input signal.
The following is obtained from Figure 7.9:

vˆo ( s) ˆ  vˆ ( s)   
vˆo ( s ) = ic ( s ) +  o  iˆ ( s ) +  vˆo ( s )  vˆ ( s ) =
iˆc ( s )  iˆinj ( s )  inj  vˆ g ( s )  g
  ol 2   ol 2
vˆo ( s )  ˆ  vˆ ( s ) ˆ 
 ic 2 ( s ) + H i ( s ) o + iinj ( s )   + (7.23)
iˆc ( s )   R 
 vˆ ( s )   
 o  iˆ ( s ) +  vˆo ( s)  vˆ ( s ) .
 iˆinj ( s )  inj  vˆ g ( s )  g
  ol 2   ol 2
274 Chapter 7. Using Load Current for Control

Converter with a
current controller

v^ g(s) v^ o(s)
v^ g(s) ol2
^i (s)
inj v^ o(s)
^i (s)
inj ol2

^i (t) ^i (s)
c2 c v^ o(s) v^ o(s)
^i (s)
c
^i
1 load(s)
R

Hi(s)

Figure 7.9: The system obtained when using H i (s ) in the control law.

An expression for the output voltage is obtained from (7.23):

ˆ  vˆ ( s)  vˆ ( s )  
 v (s) ˆ
vˆo ( s ) =  o ic 2 ( s ) +  o H i (s) +  o  iˆ ( s ) +
 ic ( s)
ˆ  ˆ
ic ( s )  ˆ
i ( s )   inj
   inj  ol 2 
(7.24)
−1
 vˆo ( s )   ˆ 
  vˆ g ( s)  1 − v o ( s) H i ( s ) 1  .
 vˆ g ( s )   iˆc ( s ) R 
  ol 2 

Control-to-Output Transfer Function

In this subsection, the control-to-output transfer function is derived for

the case where the measured load current is used for control. The result is
analyzed and compared with simulation results. It is shown that the control-
to-output transfer function is almost invariant for different loads in the
frequency interval dc to half the switching frequency if H i (s ) =1. It is also
Chapter 7. Using Load Current for Control 275

shown that invariance (theoretically) is obtained for a specific H i (s ) .

Furthermore, it is shown that this specific H i (s ) can be obtained by
designing a gain-scheduling controller.
The approximate control-to-output transfer function (6.40) is repeated
here for convenience:

vˆo ( s)
Gvoic ( s ) = = RKFl ( s ) FESR ( s ) Fh ( s ) , (7.25)
iˆc ( s )

where

1
K= ,
RTs (7.26)
1+ (mc D'−0.5)
L

1
Fl ( s) = , (7.27)
1 + sRCK

FESR ( s) = 1 + sRc C , (7.28)

1
Fh ( s ) = ,
s s2 (7.29)
1+ +
ω nQ ω n2

ω n is defined in (3.13), Q is defined in (3.20), and mc is defined in (3.21).

To obtain the control-to-output transfer function of the closed loop
system where (7.21) is used, the two input signals vˆ g ( s ) and iˆinj ( s ) are set
to zero in (7.24):

vˆo ( s )
vˆo ( s) iˆc ( s)
= . (7.30)
iˆc 2 ( s ) vˆ ( s ) 1
1− o H ( s)
ˆic ( s ) i R

(7.30) is rewritten by using (7.25):

276 Chapter 7. Using Load Current for Control

vˆo ( s ) RKFl ( s) FESR ( s) Fh ( s )

Gvoic 2 ( s ) = = =
ˆic 2 ( s ) 1 − KFl ( s) FESR ( s) Fh ( s ) H i ( s )
(7.31)
FESR ( s) Fh ( s )
.
R −1 K −1 Fl−1 ( s ) − R −1 FESR ( s ) Fh ( s) H i ( s )

The (last) denominator in (7.31) is rewritten by using (7.27) and (7.26):

R −1 K −1 Fl−1 ( s ) − R −1 FESR ( s ) Fh ( s ) H i ( s) =
R −1 K −1 + sC − R −1 FESR ( s ) Fh ( s ) H i ( s ) =
Ts
R −1 + (mc D'−0.5) + sC − R −1 FESR (s) Fh ( s) H i (s) = (7.32)
L
 1 T 
C  s + (1 − FESR ( s ) Fh ( s ) H i ( s ) ) + s (mc D '−0.5) .
 RC LC 

(7.31) is rewritten by using (7.32):

vˆo ( s )
Gvoic 2 ( s ) = =
ˆic 2 ( s )
FESR ( s) Fh ( s ) (7.33)
.
 1 T 
C  s + (1 − FESR ( s ) Fh ( s ) H i ( s ) ) + s (mc D '−0.5)
 RC LC 

Note that if H i (s ) is set to zero in (7.33), i.e. the measured load current is
not used, Gvoic 2 ( s) is the same as Gvoic (s ) in (7.25).
A new variable, F (s ) , is now introduced:

F ( s ) = FESR ( s) Fh ( s ) . (7.34)

The load resistance, R , occurs only at one place in (7.33) and the more
H i (s ) is in accordance with 1 F ( s ) , the closer invariance for different loads
is the control-to-output transfer function. 1 F ( s ) is approximately equal to 1
at low frequencies since both FESR (s ) and Fh (s ) are approximately equal to
1 at low frequencies. Condition (6.34) sets a lower limit for the corner
frequency of FESR (s ) . Condition (6.38) sets a lower limit for Q and
Chapter 7. Using Load Current for Control 277

therefore also a lower limit for the corner frequency of Fh (s ) (see upper plot
in Figure 7.10). If (7.1) is used, i.e. H i (s ) is equal to 1, the control-to-
output transfer function is almost invariant for different loads at low
frequencies. It will be shown below that the absolute value of the second term
is much smaller than the absolute value of the first term, s , in the (largest)
parenthesis in the denominator of (7.33). This means that the control-to-
output transfer function is almost invariant for different loads for all
frequencies in the interval [0, ω n ] . The conclusion is that there is not so
much to gain by trying to get H i (s ) in accordance with 1 F ( s ) compared to
setting H i (s ) equal to 1.
It will now be shown that

1
(1 − FESR (s) Fh ( s) H i (s))
RC (7.35)
<< 1
s

for all frequencies in the interval [0, ω n ] in the case where H i (s ) is equal to
1. In addition to the conditions in (6.34)-(6.38), the condition

1 ωn
<< (7.36)
RC (1 − ω n QRc C )2 + Q 2
is also assumed to be fulfilled. This extra condition is necessary for showing
that (7.35) is valid also for frequencies near ω n and it sets an upper limit for
Q . The left side of (7.35) is rewritten by using (7.28) and (7.29):

(1 − FESR (s) Fh ( s) ) 1
RC
(
Fh ( s) Fh−1 ( s ) − FESR ( s ) ) RC1
= =
s s
(7.37)
 s2  1
Fh ( s ) 1 + + 2 − (1 + sRc C )
s
 ω nQ ω n  RC
= Fh ( s ) G ( s ) ,
s

where
278 Chapter 7. Using Load Current for Control

20

-20
Magnitude G (dB); Magnitude Fh (dB)

-40
Q decreasing
-60

-80
3 4 5 6
10 10 10 10

20

0
Q decreasing
-20

-40

-60
3 4 5 6
10 10 10 10
Frequency (Hz)

Figure 7.10: The magnitude of Fh (s ) and G (s ) for different values of Q .

{
Q = 0.2, 0.4, 1 2 , 2, 4, 8 . ω n 2π is equal to 25000 Hz. }
 1 s  1
G(s) =  + − Rc C  . (7.38)
 ωnQ ω 2  RC
 n 

Fh ( jω ) and G ( jω ) are shown in Figure 7.10 for different values of Q .

The following parameters are used: R =1 Ω, C =400 µF, Rc =0 Ω,
{
ω n =50000π rad/s, Q = 0.2, 0.4, 1 2 , 2, 4, 8 . It is seen that Fh ( jω ) has }
a peak at approximately ω n for high Q . There is no peak if Q ≤ 1 2 . It is
also seen that the dc gain of G (s ) depends on Q . The high-frequency
asymptote of G ( jω ) is equal to jω ω n2 RC and it is hence independent ( )
of Q and Rc . The asymptote is equal to 1 at the frequency ω a :

ω a = ω n2 RC . (7.39)

Since K is a positive number less than unity (see Section 6.2), the following
are obtained from (6.37) and (6.38):
Chapter 7. Using Load Current for Control 279

1
<< ω n , (7.40)
RC

1
<< ω n Q . (7.41)
RC

The following is obtained by combining (7.39) and (7.40):

ω a >> ω n . (7.42)

The absolute value of the dc gain of G (s ) is equal to

 1  1 1 R
G ( j 0) =  − Rc C  = − c . (7.43)
 ω nQ  RC ω n QRC R

Assume that Rc =0 Ω. By combining (7.43) and (7.41), it is concluded that

G ( j 0) << 1 . If Rc increases, G ( j 0) decreases and it becomes zero for a
certain Rc . For higher Rc ,

1 R R
G ( j 0) = − c < c << 1 . (7.44)
ω n QRC R R

Condition (6.34) is used in (7.44). The conclusion is that G ( j 0) << 1 for all
permitted Rc . Since the slope of the high-frequency asymptote of G ( jω ) is
equal to 1 (20 dB/decade) and ω a is much higher than ω n according to
(7.42), G ( jω ) << 1 for all frequencies in the interval [0, ω n ] .
First consider the case where Q ≤ 1 2 . Since Fh ( jω ) has no peak,
Fh ( jω ) ≤ 1 and hence Fh ( jω ) G ( jω ) << 1 for all frequencies in the
interval [0, ω n ] . Now consider the case where Q > 1 2 . The maximum of
Fh ( jω ) is located approximately at ω n . The maximum of
Fh ( jω ) G ( jω ) in the frequency interval [0, ω n ] is also located
approximately at ω n . The maximum is therefore approximately
280 Chapter 7. Using Load Current for Control

Fh ( jω n ) G ( jω n ) =
−1
1 + jω n + ( jω n )
 2   1 jω  1
  + 2n − Rc C  =
 ωnQ ω n2   ω nQ ω  RC
   n 
2 (7.45)
 1 
2
ω  1  1
2
 Q2 1
Q  − Rc C  +  n2  =  − QRc C  + 2 =
 ω  RC
 ω nQ   n ωn  ω n RC
1
(1 − ω n QRc C )2 + Q 2 1 << 1 .
ωn RC

Condition (7.36) is used in (7.45). The conclusion is that

Fh ( jω ) G ( jω ) << 1 for all permitted Q and all frequencies in the interval
[0, ω n ] . From (7.37), it is now concluded that (7.35) is valid for all
frequencies in the interval [0, ω n ] in the case where H i (s ) is equal to 1.
It is seen from (7.28) and (7.29) that FESR (s ) has one high-frequency
zero and Fh (s ) has two high-frequency poles. If H i (s ) is equal to 0, i.e. the
measured load current is not used in the control law, the low-frequency pole
in (7.33) is

 1 T 
p1 = −  + s (mc D '−0.5) . (7.46)
 RC LC 

If H i (s ) is equal to 1, i.e. the control law (7.1) is used, the low-frequency

pole in (7.33) is approximately

Ts
p1 ≈ − (mc D'−0.5) . (7.47)
LC

There would be no approximation in (7.47) if F (s ) was equal to 1. Since

F (s ) is approximately equal to 1 at low frequencies, the low-frequency pole
can be approximated according to (7.47). The results obtained here for the
position of the low-frequency pole are compared with the results obtained in
Section 7.3 where a simple model was used. According to (7.20), the pole
moves from − 1 (RC ) to the origin if H i is increases from 0 to 1. The pole
moves into the right half side in the complex s-plane if H i increases further
and the system becomes unstable. From (7.46) and (7.47), it is seen that the
Chapter 7. Using Load Current for Control 281

position of the low-frequency pole is shifted approximately

Ts (mc D '−0.5) (LC ) to the left in the s-plane compared to the position of the
pole in the simple model. A greater H i (s ) is therefore needed to obtain
instability according to the model used in this section compared to what is
needed according to the simple model.
It is seen from (7.33) that the dc gain is approximately inversely
proportional to the distance between the low-frequency pole and the origin.
Therefore, the dc gain approximately increases by the same degree as the first
(lowest) corner frequency decreases if H i (s ) is changed from 0 to 1.
Figure 7.11 shows the Bode plot for Gvo ic 2 ( s) in (7.33) when different
H i (s ) and loads, Rmin =1 Ω and Rmax =4 Ω, are used. Except for R , the
parameter values shown in Table 2.1 are used. mc is set to 2. From the figure
it is seen that for H i (s ) =0, the gain and phase shift changes considerably for
different loads. For H i (s ) =1, the gain and phase shift is almost invariant for
different loads. Simulation results are also plotted in Figure 7.11 and they are
in good agreement with (7.33). The used simulation model will be presented
at the end of this section.
Previously in this subsection, we concluded from (7.33) that the more
H i (s ) is in accordance with 1 F ( s ) , the closer invariance for different loads
is the control-to-output transfer function. Simulation results have showed
that when H i (s ) is set to 1 F ( s ) in series with a second order Butterworth
low-pass filter with corner frequency at the switching frequency, the transfer
function is closer to invariance for different loads than when H i (s ) is set
equal to 1. (The Bode plot with the simulation results is not presented here.)
This is in accordance with the conclusion. 1 F ( s ) has two zeros and one pole
and cannot therefore be implemented. To be able to implement a rational
transfer function, it must be proper, i.e. the degree (order) of the numerator
polynomial must be lower than or equal to the degree of the denominator
polynomial. The combination of 1 F ( s ) and a first order Butterworth low-
pass filter is proper since an extra pole is added. To obtain a small
amplification of the load current at very high frequencies, the order of the
Butterworth filter is increased to two in the simulation.
According to (7.33), H i (s ) should be equal to 1 F ( s ) to obtain
invariance for different loads. The rest of this subsection will be devoted to
showing that this result also can be obtained by applying gain scheduling.
It was shown in Section 7.3 that the use of measured load current for
control can be seen as gain scheduling. Assume that the only goal with the
gain scheduling controller is to make the closed loop invariant for different
282 Chapter 7. Using Load Current for Control

20

0
Phase (deg); Magnitude (dB)

-20

-40
1 2 3 4
10 10 10 10

-50

-100

-150
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 7.11: The control-to-output transfer function of a buck converter

controlled by (7.21). Symbol for simulation result is in
parenthesis. Dash-dotted line (): H i (s ) =0 and R = Rmin .
Dotted line (O): H i (s ) =0 and R = Rmax . Solid line (+):
H i (s ) =1 and R = Rmin . Dashed line (x): H i (s ) =1 and
R = Rmax . Note that the two last mentioned lines almost
coincide.

loads and an outer controller is designed later to control the output voltage
(see Figure 7.7). The first step in designing the gain scheduling controller is
to design a controller as if the resistance of the load is constant and known.
This controller will here be designed by using model matching (Chen, 1999,
Section 9.3), which is similar to pole placement but the zeros are also placed.
The design procedure presented by Chen (1999, Section 9.3) will be used and
the notation will be almost the same. One difference is that the polynomial
F (s ) is replaced by B (s ) since F (s ) is already reserved (see (7.34)).
The process to be controlled is Gvoic (s ) in (7.25):
Chapter 7. Using Load Current for Control 283

vˆo ( s) RKF ( s ) N (s)

Gvoic ( s ) = = −1 ESR−1 = , (7.48)
iˆc ( s ) Fl ( s ) Fh ( s ) D ( s)

where

N ( s) = RKFESR ( s ) , (7.49)

D ( s ) = Fl−1 ( s ) Fh−1 ( s ) . (7.50)

It is assumed that no pole in Gvoic (s ) coincides with a zero. This means that
N (s ) and D (s ) are coprime, i.e. they have no common factors. The position
of the zero connected with FESR (s ) does not depend on R . The same is true
for the two high-frequency poles connected with Fh (s ) . However, the
position of the low-frequency pole, p1 , connected with Fl (s) and the dc
gain of Gvoic (s ) depend on R .
The expressions for the dc gain and the position of all the poles and zeros
of the closed loop system Gvoic 2 ( s ) should be independent of R . Since this is
the only goal with the controller, the dc gain and the positions are chosen
such that the expression of the controller is simple. Two poles and a zero are
therefore placed at the same position as the two high-frequency poles and the
zero in Gvoic (s ) . It is not clear at this point how the dc gain and the position
of the last pole in Gvoic 2 ( s ) should be chosen to obtain a simple expression of
the controller. The derivation of the controller will here be made only with
analytic expressions. The derivation will therefore not be much more difficult
if two variables are introduced for the dc gain and the position of the last
pole. The expressions for these two variables are chosen later to obtain a
simple expression of the controller. The dc gain of Gvoic 2 ( s ) is denoted β
and the position of the last pole in Gvoic 2 ( s ) is denoted p1n . Hence, the low-
frequency pole, p1 , in Gvoic (s ) is replaced by a new low-frequency pole,
p1n , in Gvoic 2 ( s ) and the following notation is introduced:

1
Fln ( s ) = . (7.51)
1 − p1−n1 s

Gvoic 2 ( s ) can now be written as

284 Chapter 7. Using Load Current for Control

vˆo ( s ) βF ( s ) E ( s)
Gvoic 2 ( s) = = −1 ESR −1 = , (7.52)
iˆc 2 ( s ) Fln ( s ) Fh ( s ) B ( s )

where

E ( s ) = βFESR ( s ) , (7.53)

B ( s ) = Fln−1 ( s ) Fh−1 ( s ) . (7.54)

It is assumed that E (s ) and B (s ) are coprime.

The fraction Gvoic 2 ( s) N ( s) is now computed:

Gvoic 2 ( s ) E ( s) βFESR ( s )
= = −1 =
N (s) B ( s) N ( s) Fln ( s ) Fh−1 ( s ) RKFESR ( s )
(7.55)
β E ( s)
−1 −1
= ,
Fln ( s) Fh ( s ) RK B ( s )

where

E (s) = β , (7.56)

B ( s ) = Fln−1 ( s ) Fh−1 ( s ) RK . (7.57)

E (s ) and B (s ) are coprime since all common factors ( FESR (s ) ) are

cancelled.
The two compensators in the controller are

L( s)
C1 ( s ) = , (7.58)
A( s )

M (s)
C2 ( s) = . (7.59)
A( s )
Chapter 7. Using Load Current for Control 285

Both C1 ( s ) and C 2 ( s ) should be proper rational transfer functions. Note

that the denominators are chosen to be the same in the design procedure.
From Figure 7.7, it is seen that the closed loop system is

vˆo ( s ) Gvoic ( s )
Gvoic 2 ( s ) = = C1 ( s ) =
iˆc 2 ( s ) 1 + Gv i ( s )C 2 ( s)
o c

N (s) (7.60)
L( s ) D(s) L( s ) N ( s )
= .
A( s) N ( s ) M ( s ) A( s ) D ( s) + M ( s ) N ( s)
1+
D ( s ) A( s )

The following is obtained by combining (7.55) and (7.60):

L( s ) N ( s ) E ( s) N ( s)
Gvoic 2 ( s ) = = . (7.61)
A( s ) D( s) + M ( s ) N ( s ) B (s)

If the two numerators in (7.61) are set equal and consequently, the two
denominators are set equal, there might not exist any solution such that
C 2 ( s ) = M ( s) A( s ) is proper. A new polynomial, Bˆ ( s ) , is therefore
introduced and it must fulfill two conditions. The roots of Bˆ ( s ) must be in
the open left half side in the complex s-plane and the degree of Bˆ ( s ) must
be:

deg Bˆ ( s ) ≥ 2 deg D ( s ) − 1 − deg B ( s ) . (7.62)

For our system, the degree condition is deg Bˆ ( s ) ≥ 2 ⋅ 3 − 1 − 3 = 2 . The

polynomial Bˆ ( s ) can be chosen arbitrarily as long as it fulfills these two
conditions. (7.61) can now be rewritten:

L( s) N ( s) E ( s ) Bˆ ( s ) N ( s )
Gvoic 2 ( s ) = = . (7.63)
A( s ) D ( s ) + M ( s ) N ( s ) B ( s ) Bˆ ( s )

There exists a proper C 2 ( s ) if the numerators in (7.63) are set equal and
L(s ) , M (s) and A(s ) can thus be calculated from

L( s ) = E ( s ) Bˆ ( s ) , (7.64)
286 Chapter 7. Using Load Current for Control

A( s ) D( s ) + M ( s ) N ( s ) = B ( s ) Bˆ ( s ) . (7.65)

The polynomial Bˆ ( s ) is a factor both in the numerator and the

denominator of Gvoic 2 ( s ) and does therefore not affect the dynamics of the
closed loop. The degree of Bˆ ( s ) is here chosen to two and one root of Bˆ ( s )
is chosen to be equal to the ESR zero since this simplifies the analysis of the
controller design. (It is assumed that Rc > 0 .) The second root of Bˆ ( s ) is
denoted p f and Bˆ ( s ) is written as

(
Bˆ ( s ) = FESR ( s ) 1 − p −f 1 s .) (7.66)

As concluded previously, 1 F ( s ) is not proper. Since C 2 ( s ) is proper with

the used design procedure, it will be studied what happens when p f tend to
− ∞ . At the limit, condition (7.62) is not fulfilled and C 2 ( s ) is not proper.
The following notions are introduced:

A( s ) = A0 + A1 s + A2 s 2 , (7.67)

M ( s) = M 0 + M 1 s + M 2 s 2 , (7.68)

B ( s ) Bˆ ( s ) = B0 + B1 s + B2 s 2 + B3 s 3 + B4 s 4 + B5 s 5 , (7.69)

N ( s) = N 0 + N1 s + N 2 s 2 + N 3 s 3 , (7.70)

D ( s ) = D0 + D1 s + D2 s 2 + D3 s 3 . (7.71)

(7.65) is rewritten as

[A0 M0 A1 M1 A2 M 2 ]S m =
(7.72)
[B0 B1 B2 B3 B4 B5 ] ,

where
Chapter 7. Using Load Current for Control 287

 D0 D1 D2 D3 0 0 
N N1 N2 N3 0 0 
 0
 0 D0 D1 D2 D3 0 
Sm =  . (7.73)
 0 N0 N1 N2 N3 0 
 0 0 D0 D1 D2 D3 
 
 0 0 N0 N1 N2 N 3 

M (s) and A(s ) are obtained by rewriting (7.72):

[A0 M0 A1 M1 A2 M 2 ]=
(7.74)
[B0 B1 B2 B3 B4 B5 ]S m−1 .

The polynomials L(s ) , M (s) and A(s ) are calculated by using the
software MAPLE (version 5.00). By executing the sequence of commands
shown in Figure 7.12, the following results are obtained:

 s 
L( s ) = β (1 + sRc C ) 1 − , (7.75)
 p f 


M ( s) =
(RCKp f ) (
+ 1 (RCKp1n + 1) ω n2 Q + ω n s + s 2 Q ), (7.76)
p f ω n2 Qp1n R 2 K 2 C 2

(1 + sRc C )(sRCK − RCKp1n − RCKp f −1 )

A( s ) = . (7.77)
C 2 KRp f p1n

The resulting C1 ( s ) and C 2 ( s ) are (after some rewriting)

288 Chapter 7. Using Load Current for Control

> # Initiation
> Fl:=1/(1+s*R*C*K):
> FESR:=1+s*Rc*C:
> Fh:=1/(1+s/(wn*Q)+s^2/wn^2):
> N:=R*K*FESR:
> D_:=1/(Fl*Fh):
> Fln:=1/(1-s/p1n):
> Ebar:=Beta:
> Bbar:=R*K/(Fln*Fh):
> Bhat:=FESR*(1-s/pf):
>
> # Calculating controller parameters
> with(linalg):
> L:=Ebar*Bhat:
> N0:=coeff(N,s,0):
> N1:=coeff(N,s,1):
> N2:=coeff(N,s,2):
> N3:=coeff(N,s,3):
> D0:=coeff(D_,s,0):
> D1:=coeff(D_,s,1):
> D2:=coeff(D_,s,2):
> D3:=coeff(D_,s,3):
> F0:=coeff(Bbar*Bhat,s,0):
> F1:=coeff(Bbar*Bhat,s,1):
> F2:=coeff(Bbar*Bhat,s,2):
> F3:=coeff(Bbar*Bhat,s,3):
> F4:=coeff(Bbar*Bhat,s,4):
> F5:=coeff(Bbar*Bhat,s,5):
> Sm:=matrix(6,6,
> [D0,D1,D2,D3, 0, 0,
> N0,N1,N2,N3, 0, 0,
> 0,D0,D1,D2,D3, 0,
> 0,N0,N1,N2,N3, 0,
> 0, 0,D0,D1,D2,D3,
> 0, 0,N0,N1,N2,N3]):
> BbarBhat:=matrix(1,6,[F0,F1,F2,F3,F4,F5]):
> A_M:=multiply(BbarBhat,inverse(Sm)):
> A:=factor(A_M[1,1]+A_M[1,3]*s+A_M[1,5]*s^2):
> M:=factor(A_M[1,2]+A_M[1,4]*s+A_M[1,6]*s^2):
> C1:=factor(L/A):
> C2:=factor(M/A):
> L,M,A,C1,C2;

Figure 7.12: The sequence of MAPLE commands used to calculate a

controller for a buck converter
Chapter 7. Using Load Current for Control 289

 
βCp1n 1 − 
s

L(s)  pf 
C1 ( s ) = =− ,
A( s )  
    (7.78)
1 +  p1n 1  1   s 
+  1−
  
RCK  p f   1 
  p f + p1n + 
 RCK 

 
(K+ RCp1n 1 +
−1
)1 

M ( s)  RCKp f 
C 2 ( s) = =− . (7.79)
A( s )   1  1 
FESR ( s ) Fh ( s ) R 1 +  p1n + − s 
  RCK  p 
 f 

With C1 ( s ) and C 2 ( s ) according to (7.78) and (7.79), the expressions for

the dc gain and the position of all the poles and zeros of the closed loop
system Gvoic 2 ( s ) are independent of R if β and p1n are chosen not to
depend on R . Note that both C1 ( s ) and C 2 ( s ) are proper rational transfer
functions.
Assume that p f is a negative real value with great absolute value. C1 ( s )
and C 2 ( s ) can then be approximated in the frequency interval [0, ω n ] by
using (7.78), (7.79) and (7.26):

C1 ( s ) ≈ − β Cp1n , (7.80)

T 
−1 1 + RC  s (mc D '−0.5) + p1n 
K + RCp1n  LC  (7.81)
C 2 ( s) ≈ − =− .
FESR ( s ) Fh ( s ) R FESR ( s ) Fh ( s ) R

It is seen from (7.81) and (7.80) that it is suitable to choose

Ts
p1n = − (mc D'−0.5) , (7.82)
LC
290 Chapter 7. Using Load Current for Control

1 1
β =− = .
Cp1n Ts (7.83)
(mc D'−0.5)
L

None of β and p1n depends on R as required. Note that if H i (s ) is equal

to 1 F ( s ) in (7.33), then the dc gain and the low-frequency pole in (7.33)
are exactly the same as β and p1n in (7.83) and (7.82).
With the choices for β and p1n in (7.83) and (7.82), the following
limits are obtained when p f tend to − ∞ :

C1 ( s ) = 1 , (7.84)

−1
C 2 ( s) = . (7.85)
FESR ( s) Fh ( s ) R

Note that C 2 ( s ) in (7.85) is no longer a proper rational transfer function.

From Figure 7.7, it is seen that the following control law is obtained if (7.84)
and (7.85) are used:

−1
iˆc ( s) = iˆc 2 ( s) − vˆo ( s) , (7.86)
F (s) R

where F (s ) is defined in (7.34).

The second step in designing the gain scheduling controller is to replace
the parameter R in the control law (7.86) with the time-varying parameter
Rcal (t ) defined in (7.14) and the result is

−1 1 ˆ
iˆc ( s) = iˆc 2 ( s) − vˆo ( s) = iˆc 2 ( s) + iload ( s ) . (7.87)
F ( s ) Rcal ( s ) F ( s)

By comparing (7.87) and (7.21), it is concluded that H i (s ) should be equal

to 1 F ( s ) to obtain invariance for different loads. It has now been shown
that this result could be obtained by applying gain scheduling.
Chapter 7. Using Load Current for Control 291

Output Impedance

In this subsection, the output impedance is derived for the case where the
measured load current is used for control. The result is analyzed and
compared with simulation results.
The approximate output impedance (6.44) is repeated here for
convenience:

 vˆ ( s ) 
(Z out (s))ol 2 = −  ˆ o  = RKF ( s ) F ( s ) ,
 l ESR (7.88)
 i inj ( s )  ol 2

where K is defined in (7.26), Fl (s) is defined in (7.27), and FESR (s ) is

defined in (7.28).
To obtain the output impedance of the closed loop system where (7.21) is
used, the two input signals vˆ g ( s ) and iˆc 2 ( s ) are set to zero in (7.24):

vˆo ( s )  vˆ ( s ) 
H i (s) +  o 
ˆc ( s)
i  iˆinj ( s) 
vˆ ( s )   ol 2 (7.89)
Z out ( s ) = − o =− .
ˆiinj ( s ) ˆv o ( s ) 1
1− H i (s)
iˆc ( s ) R

(7.89) is rewritten by using (7.88) and (7.25):

RKFl ( s ) FESR ( s ) Fh ( s ) H i ( s ) − RKFl ( s ) FESR ( s )

Z out ( s ) = − =
1 − KFl ( s) FESR ( s) Fh ( s ) H i ( s )
(7.90)
(1 − Fh (s) H i ( s) )FESR (s)
.
R −1 K −1 Fl−1 ( s) − R −1 FESR ( s ) Fh ( s ) H i ( s )

(7.90) is rewritten by using (7.32):

(1 − Fh ( s) H i (s))FESR ( s)
Z out ( s ) = .
 1 T  (7.91)
C  s + (1 − FESR ( s ) Fh ( s ) H i ( s ) ) + s (mc D '−0.5)
 RC LC 
292 Chapter 7. Using Load Current for Control

-20
Phase (deg); Magnitude (dB)

-40

-60

1 2 3 4
10 10 10 10

100

50

-50

-100
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 7.13: The output impedance of a buck converter controlled by (7.21).

Dash-dotted line (): H i (s ) =0 and R = Rmin . Dotted line (O):
H i (s ) =0 and R = Rmax . Solid line (+): H i (s ) =1 and
R = Rmin . Dashed line (x): H i (s ) =1 and R = Rmax . Note that
the two last mentioned lines almost coincide. Points:
Simulation with R = Rmax and H i (s ) equal to 1 Fh ( s ) in
series with a second order filter.

Note that the denominator is exactly the same as in (7.33) and that the
numerator is independent of R . The conclusions about invariance of R for
the control-to-output transfer function are therefore also valid in this case.
From (7.91), it is seen that the more H i (s ) is in accordance with 1 Fh ( s ) ,
the lower is the output impedance. 1 Fh ( s ) is approximately equal to 1 at
low frequencies. If (7.1) is used, i.e. H i (s ) is equal to 1, the output
impedance will be low at low frequencies.
Figure 7.13 shows the Bode plot for the output impedance in (7.91)
when different H i (s ) and loads are used. The parameter values used for the
control-to-output transfer function are also used here. From the figure it is
seen that for H i (s ) =0, the output impedance is high at low frequencies.
Chapter 7. Using Load Current for Control 293

When H i (s ) is changed to 1, the output impedance is reduced at low

frequencies and it becomes almost invariant for different loads. Simulation
results are also plotted in Figure 7.13 and they are in good agreement with
(7.91). The used simulation model will be presented at the end of this
section. When H i (s ) is changed to 1 Fh ( s ) in series with a second order
Butterworth low-pass filter with corner frequency at five times the switching
frequency, simulation results show (see Figure 7.13) that the output
impedance is further decreased by approximately 20 dB for all the tested
frequencies. Note that (7.91) is not valid for frequencies over half the
switching frequency. It cannot therefore be used to predict what happens
when there is a step change in the load, since the load current in this case
consists of frequency components that are also over half the switching
frequency.

Audio Susceptibility

In this subsection, the audio susceptibility is derived for the case where
the measured load current is used for control. The result is analyzed and
compared with simulation results.
The approximate audio susceptibility (6.47) is repeated here for
convenience:

 vˆo ( s ) 

 vˆ g ( s )  L
( )
 = RTs D mc D '− F f ( s ) KFl ( s ) FESR ( s ) Fh ( s ) , (7.92)
  ol 2

where

D (3 − 2 D )DTs
F f ( s) = 1 − − s−
(
1 − 2 D + D 2 DTs2 2
s ,
) (7.93)
2 12 24

K is defined in (7.26), Fl (s) is defined in (7.27), FESR (s ) is defined in

(7.28), and Fh (s) is defined in (7.29). F f (s ) in (7.93) is used in this section
and it is a Taylor polynomial of degree 2 of F f (s ) in (6.48).
To obtain the audio susceptibility of the closed loop system where (7.21)
is used, the two input signals iˆinj ( s ) and iˆc 2 ( s ) are set to zero in (7.24):
294 Chapter 7. Using Load Current for Control

 vˆ ( s ) 
 o 
 vˆ ( s ) 
ˆv o ( s)  g  ol 2 (7.94)
= .
vˆ g ( s ) vˆo ( s ) 1
1− H i (s)
iˆc ( s ) R

(7.94) is rewritten by using (7.92) and (7.25):

vˆo ( s)
RTs D
L
( )
m c D '− F f ( s) KFl ( s ) FESR ( s ) Fh ( s)
= =
vˆ g ( s ) 1 − KFl ( s ) FESR ( s ) Fh ( s ) H i ( s )
(7.95)
Ts D
L
( )
mc D '− F f ( s ) FESR ( s ) Fh ( s )
.
R −1 K −1 Fl−1 ( s ) − R −1 FESR ( s ) Fh ( s) H i ( s )

(7.95) is rewritten by using (7.32):

vˆo ( s)
Ts D
L
( )
mc D '− F f ( s ) FESR ( s ) Fh ( s )
= . (7.96)
vˆ g ( s)  1 Ts 
C  s + (1 − FESR ( s ) Fh ( s ) H i ( s ) ) + (mc D'−0.5)
 RC LC 

Note that the denominator is exactly the same as in (7.33) and that the
numerator is independent of R . The conclusions about invariance of R for
the control-to-output transfer function are therefore also valid in this case.
According to Section 4.3, it is possible to choose mc such that the audio
susceptibility is very small at dc. This ability still remains in the case where
the control law (7.21) is used since the expression m c D'− F f ( s) in (7.92)
also is present in the numerator of (7.96).
Figure 7.14 shows the Bode plot for the audio susceptibility in (7.96)
when different H i (s ) and loads are used. The parameter values used for the
control-to-output transfer function are also used here. From the figure it is
seen that for H i (s ) =0, the gain changes considerably for different loads. For
H i (s ) =1, the gain is almost invariant for different loads. Simulation results
are also plotted in Figure 7.14 and they are in good agreement with (7.96).
Chapter 7. Using Load Current for Control 295

-10

-20

-30

-40
Phase (deg); Magnitude (dB)

-50

-60
1 2 3 4
10 10 10 10

-20

-40

-60

-80

-100
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 7.14: The audio susceptibility of a buck converter controlled by

(7.21). Dash-dotted line (): H i (s ) =0 and R = Rmin . Dotted
line (O): H i (s ) =0 and R = Rmax . Solid line (+): H i (s ) =1
and R = Rmin . Dashed line (x): H i (s ) =1 and R = Rmax . Note
that the two last mentioned lines almost coincide.

The small differences that are seen in the phase at high frequencies are mainly
due to that the approximate model (7.92) is used instead of the non-
approximated version (5.7). The used simulation model will be presented in
the next subsection.

Simulation Models

In the previous subsections, simulation results were presented. In this

subsection, the utilized simulation models are presented.
Some properties obtained by using measured load current were in Section
7.3 explained by using a simple model of the buck converter with current-
mode control. One conclusion was that an outer voltage controller should be
296 Chapter 7. Using Load Current for Control

magnitude
signal
Vg angle
vg

vghat
vg vo
vo
Iinj iinj iload
iinj iload
delta iL
iL
Buck
iinjhat S Q converter
delta

in out R !Q
ic2 ic
Avoid
algebraic
loop

Ts*Me
ie sawtooth

0.5

Hinum(s)
Hiden(s)
Hi(s)

Hvnum(s)
Hvden(s)
Hv(s)

magnitude
Vo_offset Kp signal
ic2
angle
Voltage
controller

ic2dist

Figure 7.15: The simulation model where an outer voltage controller and two
transfer functions, H i (s ) and H v (s ) , are included.

used to set ic 2 (t ) if H i is close to 1 since Gvoic 2 ( s) then has a pole near the
origin and the system is close to instability. Figure 7.15 shows an extended
version of the simulation model in Figure 3.9. An outer voltage controller
with the proportional gain K p is added. Two transfer functions, H i (s ) and
H v (s ) , are also added. H v (s ) should be used only when the boost and
buck-boost converters are considered and H v (s ) is therefore set to zero.
Gvoic 2 ( s) is a part of the voltage loop and to evaluate the frequency
function of Gvoic 2 ( s) at a specific frequency, a sinusoidal signal must be
injected outside this part. The sinusoidal signal ic2dist is used for this
purpose. The frequency of the signal ic2dist is ω m and it is set in the signal
generator. For each ω m , the magnitude in the frequency function is equal to
Chapter 7. Using Load Current for Control 297

the ratio of the magnitude of the signal vo to the magnitude of the signal ic2.
The phase is equal to the phase of vo minus the phase of ic2. The magnitude
and the phase of ic2 is not known since ic2 is partly set by the outer voltage
controller. Fourier analysis must therefore be made of both ic2 and vo so that
the components with frequency ω m is obtained.
The simulation results for Gvoic 2 ( s) presented in Figure 7.11 are obtained
by using the simulation model in Figure 7.15. The parameters used in the
simulation model presented in Section 3.5 are also used here. K p is set to 2
and the constant Vo_offset is adjusted manually so that the average value of
the output voltage, Vo , is equal to 5 V ( D =0.455).
The linearized system that is analyzed in this section has three input
signals: vˆ g (t ) , iˆinj (t ) , and iˆc 2 (t ) . iˆc 2 (t ) is assumed to be zero in the cases
where the output impedance and the audio susceptibility are considered. If
the signal generator for ic2dist is deactivated and either the signal generator
for vghat or iinjhat is activated in the simulation model in Figure 7.15, the
signal ic2 contains a component with frequency ω m since the output voltage
is fed back. One solution to this problem is to design the voltage controller
such that the gain is zero at the frequency ω m . This controller can be
regarded as one with a notch filter included. One drawback of this solution is
that the controller must be changed when ω m is changed. Another drawback
is that the time for each simulation can be long. We concluded previously in
this section that according to the model derived in this section, the low-
frequency pole is shifted a little to the left in the s-plane compared to what
the simple model in Section 7.3 predicts. The system is therefore stable for
H i (s ) =1 also without the outer voltage controller according to the model
derived in this section. However, the stationary value of the output voltage is
rather sensitive to changes in the stationary value of ic 2 (t ) . Figure 7.16
shows a simulation model where the outer voltage controller is excluded. The
magnitude and phase of the signal ic2 are set directly in the signal generator.
The constant Ic2 is adjusted instead of Vo_offset. The frequency function of
Gvoic 2 ( s) has been obtained with the simulation model in Figure 7.16 but the
result is not presented here since it is almost the same as the result obtained
with the simulation model in Figure 7.15. No difference can be seen without
zooming the two Bode plots.
The simulation results for the output impedance presented in Figure 7.13
and the audio susceptibility presented in Figure 7.14 are obtained by using
the simulation model in Figure 7.16.
298 Chapter 7. Using Load Current for Control

magnitude
signal
Vg angle
vg

vghat
vg vo
vo
Iinj iinj iload
iinj iload
delta iL
iL
Buck
iinjhat S Q converter
delta

Ic2 in out R !Q
ic2 ic
Avoid
algebraic
ic2hat loop

Ts*Me
ie sawtooth

0.5

Hinum(s)
Hiden(s)
Hi(s)

Hvnum(s)
Hvden(s)
Hv(s)

Figure 7.16: The simulation model where the outer voltage controller is
excluded.

7.5 Properties of the Boost Converter

In Section 6.3, approximate expressions for the boost converter with
current-mode control were derived. In this section, these expressions are used
to analyze how the control-to-output transfer function, the output impedance
and the audio susceptibility are affected when using load current
measurements to control the converter. An expression for the output voltage
is first derived and this is then used to derive the three transfer functions. The
results are compared with simulation results and a small difference is
observed. This difference is explained for one case (the control-to-output
transfer function) at the end of this section. In this section, it is assumed that
the conditions in (6.101)-(6.107) are fulfilled since the approximate
expressions derived in Section 6.3 may not be valid otherwise.
Chapter 7. Using Load Current for Control 299

An Expression for the Output Voltage

The control law (7.2) for the boost converter is nonlinear. The linearized
version of this control law will now be considered. To linearize (7.2), the
following partial derivatives is first calculated:

∂ic (t )
=1, (7.97)
∂ic 2 (t )

∂ic (t ) v (t )
= o , (7.98)
∂iload (t ) v g (t )

∂ic (t ) iload (t )
= , (7.99)
∂v o (t ) v g (t )

∂ic (t ) v (t )i (t )
= − o 2load . (7.100)
∂v g (t ) v g (t )

The linearized version of (7.2) is obtained by using (7.97)-(7.100):

V I V I
iˆc (t ) = iˆc 2 (t ) + o iˆload (t ) + load vˆo (t ) − o load vˆ g (t ) , (7.101)
Vg Vg V g2

where I load is the dc value of iload (t ) . From Figure 2.14, it is seen that the
load current is

v o (t )
iload (t ) = + iinj (t ) . (7.102)
R

It is assumed that the dc value of iinj (t ) is equal to zero (see (2.56)). The
following is therefore obtained from (7.102):

Vo
I load = . (7.103)
R
300 Chapter 7. Using Load Current for Control

(7.101) is approximated by using (7.103), (2.122), (2.124), (6.101), and

(2.39):

1 1 1
iˆc (t ) = iˆc 2 (t ) + iˆload (t ) + vˆo (t ) − 2 vˆ g (t ) . (7.104)
D' D' R D' R

The last term in (7.104) is a feedforward of the input voltage and it

reduces the audio susceptibility. The control law (7.1) for the buck converter
does not contain a feedforward of the input voltage. The (outer) voltage
controller (see Figure 7.1) reduces the audio susceptibility at low frequencies.
If this reduction is not enough, a feedforward of the input voltage can be
included in the voltage controller. The voltage controller then uses
measurements of both the output and input voltages and the feedforward
contributes to ic 2 (t ) . The feedforward of the input voltage in the case of the
boost converter is here considered in the same way, i.e. it is a part of ic 2 (t ) .
Hence, the control law (7.104) is replaced by:

1 1
iˆc (t ) = iˆc 2 (t ) + iˆload (t ) + vˆo (t ) . (7.105)
D' D' R

To make the analysis in this section general, two transfer functions,

H i (s ) and H v (s ) , are introduced in (7.105):

iˆc ( s) = iˆc 2 ( s) + H i ( s )iˆload ( s ) + H v ( s )vˆo ( s ) . (7.106)

One possible interpretation of H v (s ) is that it is a part of the outer voltage

controller since it represents a feedback of the output voltage. Figure 7.17
shows the system obtained by using (7.106) and (7.102). If the control law
(7.105) is used, H i (s ) and H v (s ) are identified as

1
H i ( s) = , (7.107)
D'

1
H v (s) = . (7.108)
D' R

(7.105) is rewritten as follows by using (7.102) if iinj (t ) is zero:

Chapter 7. Using Load Current for Control 301

Converter with
current controller

v^ g(s) v^ o(s)
v^ g(s) ol2
^i (s)
inj v^ o(s)
^i (s)
inj ol2

^i (t) ^i (s)
c2 c v^ o(s) v^ o(s)
^i (s)
c
^i
1 load(s)
R

Hi(s)

Hv(s)

Figure 7.17: The system obtained when using H i (s ) and H v (s ) in the

control law.

1 1 2
iˆc (t ) = iˆc 2 (t ) + iˆload (t ) + Riˆload (t ) = iˆc 2 (t ) + iˆload (t ) .
D' D' R D' (7.109)

Hence, if the control law (7.105) is used and iinj (t ) is zero, H i (s ) and
H v (s ) can equivalently be identified as

2
H i ( s) = , (7.110)
D'

H v ( s) = 0 . (7.111)

In the case where the control-to-output transfer function or the audio

susceptibility is analyzed, iinj (t ) is considered to be zero and therefore
302 Chapter 7. Using Load Current for Control

(7.110) and (7.111) can be used. The contribution H v ( s )vˆo ( s ) can always
be moved to H i ( s )iˆload ( s ) in these two cases and H v (s ) can be considered
to be zero. The control law (7.106) can then be replaced by (7.21) (see Figure
7.9).
The following is obtained from Figure 7.17:

vˆo ( s) ˆ  vˆ ( s)   
vˆo ( s ) = ic ( s ) +  o  iˆ ( s ) +  vˆo ( s )  vˆ ( s ) =
iˆc ( s )  iˆinj ( s )  inj  vˆ g ( s )  g
  ol 2   ol 2
vˆo ( s )  ˆ  vˆ ( s ) ˆ  
 i ( s ) + H i ( s ) o + iinj ( s )  + H v ( s )vˆo ( s)  + (7.112)
ˆic ( s )  c 2  R  
 vˆ ( s )   
 o  iˆ ( s ) +  vˆo ( s)  vˆ ( s ) .
 iˆinj ( s )  inj  vˆ g ( s )  g
  ol 2   ol 2

An expression for the output voltage is obtained from (7.112):

ˆ  vˆ ( s)  vˆ ( s )  
 v (s) ˆ
vˆo ( s ) =  o ic 2 ( s ) +  o H i (s) +  o  iˆ ( s ) +
 iˆc ( s)  iˆc ( s )  iˆinj ( s )   inj
    ol 2 
(7.113)
−1
 vˆo ( s )   ˆ 
  vˆ g ( s)  1 − v o ( s) H ( s ) 1  ,
 vˆ g ( s )   iˆc ( s ) R 
  ol 2 

where

H ( s ) = H i ( s ) + RH v ( s ) . (7.114)

Control-to-Output Transfer Function

In this subsection, the control-to-output transfer function is derived for

the case where the measured load current is used for control. The result is
analyzed and compared with simulation results. It is shown that the control-
to-output transfer function cannot be made invariant for different loads in
the frequency interval dc to half the switching frequency due to a right half
plane zero. It is shown how to fix the poles and the rest of the zeros by
designing a gain-scheduling controller.
Chapter 7. Using Load Current for Control 303

The approximate control-to-output transfer function (6.109) is repeated

here for convenience:

vˆo ( s )
Gvoic ( s ) = = RD ' KFl ( s ) FESR ( s ) Fh ( s) FRHP ( s ) , (7.115)
iˆc ( s )

where

1
K= 3
,
RD ' Ts (7.116)
2+ (mc − 0.5)
L

1
Fl ( s) = , (7.117)
1 + sRCK

L
FRHP ( s ) = 1 − s , (7.118)
RD ' 2

FESR (s) is defined in (7.28), Fh (s) is defined in (7.29), and mc is defined

in (3.21).
To obtain the control-to-output transfer function of the closed loop
system where (7.106) is used, the two input signals vˆ g ( s ) and iˆinj ( s ) are set
to zero in (7.113):

vˆo ( s )
vˆo ( s ) iˆc ( s )
= . (7.119)
iˆc 2 ( s ) vˆ ( s ) 1
1− o H (s)
iˆc ( s ) R

(7.119) is rewritten by using (7.115):

vˆo ( s ) RD ' KFl ( s ) FESR ( s ) Fh ( s ) FRHP ( s )

Gvoic 2 ( s ) = = =
ˆic 2 ( s ) 1 − D' KFl ( s ) FESR ( s ) Fh ( s ) FRHP ( s) H ( s )
(7.120)
D ' FESR ( s ) Fh ( s ) FRHP ( s)
.
R −1 K −1 Fl−1 ( s ) − R −1 D ' FESR ( s) Fh ( s ) FRHP ( s ) H ( s)
304 Chapter 7. Using Load Current for Control

The (last) denominator in (7.120) is rewritten by using (7.117) and (7.116):

R −1 K −1 Fl−1 ( s ) − R −1 D ' FESR ( s ) Fh ( s ) FRHP ( s ) H ( s) =

R −1 K −1 + sC − R −1 D ' FESR ( s ) Fh ( s ) FRHP ( s ) H ( s ) =
D '3 Ts
2 R −1 + (mc − 0.5) + sC −
L
R −1 D ' FESR ( s ) Fh ( s ) FRHP ( s ) H ( s) = (7.121)
  D'  2
C  s + 1 − FESR ( s ) Fh ( s ) FRHP ( s) H ( s )  +
  2  RC

(mc − 0.5) .
Ts
D'3
LC 

According to a discussion previously in this section, H v (s ) can be considered

to be zero in the case where the control-to-output transfer function is
analyzed. H (s ) in (7.114) can therefore be replaced by H i (s ) . (7.120) is
rewritten by using this conclusion and (7.121):

vˆo ( s)
Gvoic 2 ( s ) = = D ' FESR ( s ) Fh ( s ) FRHP ( s ) •
ˆic 2 ( s )
 
 C  s + 1 −  2
D'
 FESR ( s ) Fh ( s ) FRHP ( s) H i ( s )  + (7.122)
   2  RC
−1
T 
D ' s (mc − 0.5) 
3
.
LC 

A new variable, F (s ) , is now introduced:

F ( s ) = FESR ( s ) Fh ( s ) FRHP ( s ) . (7.123)

The load resistance, R , occurs explicitly only at one place in (7.122) but it
also occurs implicitly in FRHP (s ) , see (7.118). The more H i (s ) is in
accordance with 2 (D ' F ( s ) ) , the closer invariance for different loads is the
denominator in the control-to-output transfer function. This is not easily
Chapter 7. Using Load Current for Control 305

made since 1 F ( s ) has a right half plane pole and is unstable. If F (s ) is

approximately equal to 1 at low frequencies a good choice is to set H i (s )
equal to 2 D ' which is the same as (7.110). F (s ) is approximately equal to 1
at low frequencies if FESR (s ) , Fh (s ) , and FRHP (s ) are approximately equal
to 1 at low frequencies. Condition (6.101) sets a lower limit for the corner
frequency of FESR (s ) . Condition (6.107) sets a lower limit for Q and
therefore also a lower limit for the corner frequency of Fh (s ) (see upper plot
in Figure 7.10). None of the conditions (6.101)-(6.107) sets an upper limit
for L and therefore no lower limit for the corner frequency of FRHP (s ) .
However, there is a lower limit for the corner frequency of FRHP (s ) in the
design of the converter since a right half plane zero sets an upper limit for the
bandwidth of the closed loop system. However, this bandwidth depends on
the specifications of the converter. Therefore, from now on it will be assumed
that corner frequency of FRHP (s ) is much higher than the frequency of the
low-frequency pole in (7.124).
If H i (s ) is equal to 0, i.e. the measured load current is not used in the
control law, the low-frequency pole in (7.122) is

 2 T 
p1 = −  + D '3 s (mc − 0.5) . (7.124)
 RC LC 

If H i (s ) is equal to 2 D ' as in (7.110), the low-frequency pole in (7.122) is

approximately

Ts
p1 ≈ − D' 3 (mc − 0.5) . (7.125)
LC

There would be no approximation in (7.125) if F (s ) was equal to 1. Since

F (s ) is approximately equal to 1 at low frequencies, the low-frequency pole
can be approximated according to (7.125).
It is seen from (7.122) that the dc gain is approximately inversely
proportional to the distance between the low-frequency pole and the origin.
Therefore, the dc gain approximately increases by the same degree as the first
(lowest) corner frequency decreases if H i (s ) is changed from 0 to 2 D ' .
Figure 7.18 shows the Bode plot for Gvo ic 2 ( s) in (7.122) when different
H i (s ) and loads, Rmin =1 Ω and Rmax =4 Ω, are used. Except for R , the
parameter values shown in Table 2.5 are used. mc is set to 2. From the figure
306 Chapter 7. Using Load Current for Control

20

0
Phase (deg); Magnitude (dB)

-20

-40
1 2 3 4
10 10 10 10

-50

-100

-150

-200

-250
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 7.18: The control-to-output transfer function of a boost converter

controlled by (7.21). Dash-dotted line (): H i (s ) =0 and
R = Rmin . Dotted line (O): H i (s ) =0 and R = Rmax . Solid line
(+): H i (s ) = 2 D ' and R = Rmin . Dashed line (x):
H i (s ) = 2 D ' and R = Rmax .

it is seen that for H i (s ) =0, the gain and phase shift changes considerably for
different loads. For H i (s ) = 2 D ' , the gain is almost invariant for different
loads for low frequencies but not for high frequencies. Simulation results are
also plotted in Figure 7.18 and they are in good agreement with (7.122). The
small difference that is seen will be explained at the end of this section. The
simulation model shown in Figure 7.15 is used, except the buck converter
subsystem is replaced with the boost converter subsystem shown in Figure
2.14. The parameters used in the simulation model presented in Section 3.6
are also used here. K p is set to 2 and the constant Vo_offset is adjusted
manually so that the average value of the duty cycle, D , is equal to 0.382.
Note that the average value of the output voltage, Vo , is a little higher in the
case where R = Rmax compared to the case where R = Rmin according to
(2.124) and (2.122).
Chapter 7. Using Load Current for Control 307

To see if it is possible to make the control-to-output transfer function

closer to invariance for different loads compared to the choice H i (s ) = 2 D ' ,
gain scheduling can be applied. The rest of this subsection will be devoted to
applying gain scheduling and this will be made in the same way as in Section
7.4.
The first step is to design a controller by using model matching as if the
resistance of the load is constant and known. The process to be controlled is
Gvoic (s ) in (7.115):

vˆo ( s ) RD' KFESR ( s ) FRHP ( s ) N ( s )

Gvoic ( s ) = = = , (7.126)
iˆc ( s ) Fl −1 ( s ) Fh−1 ( s ) D( s )

where

N ( s) = RD ' KFESR ( s) FRHP ( s ) , (7.127)

D ( s ) = Fl−1 ( s ) Fh−1 ( s ) . (7.128)

The position of the zero connected with FESR (s ) does not depend on R .
The same is true for the two high-frequency poles connected with Fh (s ) .
However, the position of the low-frequency pole, p1 , connected with Fl (s)
depends on R . The position of the zero connected with FRHP (s ) and the dc
gain of Gvoic (s ) also depend on R .
The expressions for the dc gain and the position of all the poles and zeros
of the closed loop system Gvoic 2 ( s ) should be independent of R . However,
this is not possible since the zero connected with FRHP (s ) is in the right half
side of the complex s-plane and such zeros cannot be moved, i.e. they must be
placed at the same position as they have in the process (Chen, 1999, Section
9.3). To obtain a simple expression of the controller, the other zero and two
poles are placed at the same position as the ESR zero and the two high-
frequency poles in Gvoic (s ) . The dc gain of Gvoic 2 ( s ) is denoted β and the
position of the last pole in Gvoic 2 ( s ) is denoted p1n . Hence, the low-
frequency pole, p1 , in Gvoic (s ) is replaced by a new low-frequency pole,
p1n , in Gvoic 2 ( s ) . The notation Fln (s) defined in (7.51) is used. Gvoic 2 ( s )
can now be written as
308 Chapter 7. Using Load Current for Control

vˆo ( s) βFESR ( s) FRHP ( s ) E ( s )

Gvoic 2 ( s ) = = = , (7.129)
iˆc 2 ( s ) Fln−1 ( s ) Fh−1 ( s ) B(s)

where

E ( s ) = βFESR ( s) FRHP ( s ) , (7.130)

B ( s ) = Fln−1 ( s ) Fh−1 ( s ) . (7.131)

The fraction Gvoic 2 ( s) N ( s) is now computed:

Gvoic 2 ( s ) E ( s) βFESR ( s ) FRHP ( s )

= = −1 −
=
N ( s) B ( s ) N ( s ) Fln ( s ) Fh 1 ( s ) RD' KFESR ( s ) FRHP ( s )
β E (s) (7.132)
= ,
Fln−1 ( s ) Fh−1 ( s ) RD ' K B (s)

where

E (s) = β , (7.133)

B ( s ) = Fln−1 ( s ) Fh−1 ( s ) RD ' K . (7.134)

The degree of the polynomial Bˆ ( s ) is chosen according to (7.62). For

our system, the degree condition is deg Bˆ ( s ) ≥ 2 ⋅ 3 − 1 − 3 = 2 . The degree of
Bˆ ( s ) is here chosen to two and one root of Bˆ ( s ) is chosen to be equal to the
ESR zero since this simplifies the analysis of the controller design. The second
root of Bˆ ( s ) is denoted p f and Bˆ ( s ) is written as

( )
Bˆ ( s ) = FESR ( s ) 1 − p −f 1 s . (7.135)

The polynomials L(s ) , M (s) and A(s ) are calculated by using the
same sequence of MAPLE commands as in Section 7.4 (see Figure 7.12)
except for the initiation part which is replaced by the sequence shown in
Figure 7.19. The following results are obtained:
Chapter 7. Using Load Current for Control 309

> # Initiation
> Fl:=1/(1+s*R*C*K):
> FESR:=1+s*Rc*C:
> FRHP:=1-s*Lind/(R*Dp^2):
> Fh:=1/(1+s/(wn*Q)+s^2/wn^2):
> N:=R*Dp*K*FESR*FRHP:
> D_:=1/(Fl*Fh):
> Fln:=1/(1-s/p1n):
> Ebar:=Beta:
> Bbar:=R*Dp*K/(Fln*Fh):
> Bhat:=FESR*(1-s/pf):

Figure 7.19: The initiation part of the sequence of MAPLE commands used
to calculate a controller for a boost converter.

 s 
L( s ) = β (1 + sRc C ) 1 − , (7.136)
 p f 


M ( s) =
(RKCp f ) ( )
+ 1 (RKCp1n + 1) ω n2 Q + ω n s + Qs 2 D ' 2
(7.137)
(R )
,
2
D' 2 KC + L CKp f ω n2 Qp1n

A( s ) = (1 + sRc C ) •
(R 2
KD ' 2 sC + Ls − R 2 KD ' 2 Cp f − R 2 KD ' 2 Cp1n − RD ' 2 + (7.138)
) ((
RLKCp f p1n D' R 2 D ' 2 KC + L Cp f p1n ) )−1
.

The resulting C1 ( s ) and C 2 ( s ) are (after some rewriting)

Cp1n   s 
K −1  1 −
L
β  RD' +
2

L( s) D'  RC  p f 
C1 ( s) = =− , (7.139)
A( s)  F ( s)  RD' 2
RD' 2 − p1n L +  p1n + RHP − s
 RCK  pf
310 Chapter 7. Using Load Current for Control

M ( s)
C 2 ( s) = =
A( s )
 
(
D' K −1 + RCp1n 1 + )1 
 (7.140)
 RCKp f 
− .
  F (s)  RD ' 2 
FESR ( s) Fh ( s ) RD' 2 − p1n L +  p1n + RHP − s 
  RCK  pf 
 

With C1 ( s ) and C 2 ( s ) according to (7.139) and (7.140), the expressions for

the dc gain and the position of all the poles and one zeros of the closed loop
system Gvoic 2 ( s ) are independent of R if β and p1n are chosen not to
depend on R . Note that both C1 ( s ) and C 2 ( s ) are proper rational transfer
functions.
Assume that p f is a negative real value with great absolute value. C1 ( s )
and C 2 ( s ) can then be approximated in the frequency interval [0, ω n ] by
using (7.139), (7.140) and (7.116):

Cp1n  L 
β  RD' 2 + K −1 
D'  RC 
C1 ( s ) ≈ − =
RD ' 2 − p1n L
(7.141)
Cp  2L T 
β 1n  RD' 2 + + D '3 s (mc − 0.5)
D'  RC C 
− ,
RD ' 2 − p1n L

C 2 ( s) ≈ −
(
D ' K −1 + RCp1n ) =
(
FESR ( s ) Fh ( s ) RD' − p1n L
2
)
 T  (7.142)
2 D'+ RCD '  D '3 s (m c − 0.5) + p1n 
 LC 
−
( )
.
FESR ( s ) Fh ( s ) RD' 2 − p1n L

It is seen from (7.142) and (7.141) that it is suitable to choose

Chapter 7. Using Load Current for Control 311

Ts
p1n = − D '3 (mc − 0.5) , (7.143)
LC

D' 1
β =− = .
T (7.144)
D ' 2 s (mc − 0.5)
Cp1n
L

None of β and p1n depends on R as required. Note that if H i (s ) is equal

to 2 (D ' F ( s ) ) in (7.122), then the dc gain and the low-frequency pole in
(7.122) are exactly the same as β and p1n in (7.144) and (7.143).
With the choices for β and p1n in (7.144) and (7.143), the following
limits are obtained when p f tend to − ∞ :

2L T
RD' 2 + + D '3 s (mc − 0.5)
C1 ( s ) = RC C , (7.145)
RD' − p1n L
2

− 2 D'
C 2 ( s) =
( )
. (7.146)
FESR ( s ) Fh ( s ) RD ' 2 − p1n L

Note that C 2 ( s ) in (7.146) is no longer a proper rational transfer function.

From Figure 7.7, it is seen that the following control law is obtained if
(7.145) and (7.146) are used:

iˆc ( s) =
 2L T  2 D'
 RD' 2 + + D '3 s (mc − 0.5) iˆc 2 ( s ) + vˆo ( s)
 RC C  FESR ( s) Fh ( s )
=
RD' 2 − p1n L
 2L  2 D' (7.147)
 RD' + − p1n L  iˆc 2 ( s ) +
2
vˆo ( s )
 RC  FESR ( s ) Fh ( s )
=
RD' 2 − p1n L
2  L ˆ D' 
iˆc 2 ( s ) +  
 RC ic 2 ( s ) + F ( s ) F ( s ) vˆo ( s )  .
RD' − p1n L 
2
ESR h 
312 Chapter 7. Using Load Current for Control

20

0
Phase (deg); Magnitude (dB)

-20

-40
1 2 3 4
10 10 10 10

-50

-100

-150

-200

-250
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 7.20: The control-to-output transfer function of a boost converter

controlled by using gain scheduling. Solid line (+): R = Rmin .
Dashed line (x): R = Rmax .

The second step in designing the gain scheduling controller is to replace

the parameter R in the control law (7.147) with the time-varying parameter
Rcal (t ) defined in (7.14).
Figure 7.20 shows the Bode plot for Gvo ic 2 ( s) when the gain scheduling
controller ((7.147) with R (t ) = Rcal (t ) ) and different loads, Rmin =1 Ω and
Rmax =4 Ω, are used. Except for R , the parameter values shown in Table 2.5
are used. mc is set to 2. The output voltage should be filtered by
1 (FESR ( s ) Fh ( s ) ) according to (7.147) but a second order Butterworth low-
pass filter with corner frequency at the switching frequency is connected in
series so that C 2 ( s ) is proper. Note that this modifies the closeness to
invariance for different loads a little near the switching frequency. (7.139) and
(7.140) should have been used for C1 ( s ) and C 2 ( s ) instead of (7.145) and
(7.146) to obtain the invariance for different loads that was specified with
proper C 2 ( s ) . If Figure 7.20 is compared with Figure 7.18, where
Chapter 7. Using Load Current for Control 313

magnitude
signal
Vg angle
vg

vghat
vg vo
vo
Iinj iinj iload
iinj iload
delta iL
iL
Boost
iinjhat S Q converter
delta

in out R !Q
ic2 ic
Avoid
algebraic
loop

Ts*Me
2 ie sawtooth

0.5

vofilternum(s)
L/C 1-D
vofilterden(s)
vofilter(s)

p1n*L (1-D)^2

Saturationvo
Rcal

Saturationiload
magnitude
Vo_offset Kp signal
ic2
angle
Voltage
controller

ic2dist

Figure 7.21: The simulation model where a gain scheduling controller is

included.

H i (s ) = 2 D ' , it is seen that the gain and phase shift are closer to invariance
for different loads for the gain scheduling controller. The variation depends
approximately only on the zero connected with FRHP (s ) . Simulation results
are also plotted in Figure 7.20 and they are in good agreement with (7.122).
The simulation model shown in Figure 7.21 is used. The transfer function
vofilter(s) consists of 1 (FESR ( s ) Fh ( s ) ) connected in series with the second
order Butterworth low-pass filter. To avoid problems when calculating the
signal Rcal, the signals vo and iload are limited such that the numerator and
314 Chapter 7. Using Load Current for Control

denominator in the fraction are greater than or equal to a small number. The
parameters used in the simulation model used to obtain the simulation results
in Figure 7.18 are also used here.

Output Impedance

In this subsection, the output impedance is derived for the case where the
measured load current is used for control. The result is analyzed and
compared with simulation results.
The approximate output impedance (6.118) is repeated here for
convenience:

 vˆ ( s ) 
(Z out (s))ol 2 = −  ˆ o  = RKF ( s ) F ( s ) ,
 l ESR (7.148)
i ( s )
 inj  ol 2

where K is defined in (7.116), Fl (s) is defined in (7.117), and FESR (s ) is

defined in (7.28).
To obtain the output impedance of the closed loop system where (7.106)
is used, the two input signals vˆ g ( s ) and iˆc 2 ( s ) are set to zero in (7.113):

vˆo ( s )  vˆ ( s ) 
H i (s) +  o 
iˆc ( s)  iˆinj ( s) 
ˆvo ( s )   ol 2 (7.149)
Z out ( s ) = − =− .
ˆiinj ( s ) vˆo ( s ) 1
1− H ( s)
iˆc ( s ) R

(7.149) is rewritten by using (7.148) and (7.115):

Z out ( s ) =
RD ' KFl ( s) FESR ( s) Fh ( s ) FRHP ( s ) H i ( s ) − RKFl ( s ) FESR ( s )
− =
1 − D ' KFl ( s ) FESR ( s ) Fh ( s) FRHP ( s ) H ( s ) (7.150)
(1 − D' Fh ( s) FRHP (s) H i ( s))FESR (s)
.
R K Fl−1 ( s ) − R −1 D ' FESR ( s) Fh ( s ) FRHP ( s ) H ( s)
−1 −1

(7.150) is rewritten by using (7.121):

Chapter 7. Using Load Current for Control 315

Z out ( s ) = (1 − D ' Fh ( s ) FRHP ( s ) H i ( s ) )FESR ( s) •

 
 C  s + 1 −  2
D'
 FESR ( s ) Fh ( s ) FRHP ( s ) H ( s)  +
   2  RC (7.151)
−1

(mc − 0.5) 
Ts
D'3 ,
LC 

where H (s ) is defined in (7.114). Note that if (7.107) and (7.108) are used
in (7.151) and if (7.110) is used in (7.122), then the denominators in (7.151)
and (7.122) are exactly the same. From (7.151), it is seen that the more
H i (s ) is in accordance with 1 (D' Fh ( s) FRHP ( s ) ) , the lower is the output
impedance. This is not easily made since FRHP (s ) has a right half plane zero.
Fh ( s ) FRHP ( s ) is approximately equal to 1 at low frequencies so a good
choice is to set H i (s ) equal to 1 D ' which is the same as (7.107).
Figure 7.22 shows the Bode plot for the output impedance in (7.151)
when different H i (s ) , H v (s ) , and loads are used. The parameter values
used for the control-to-output transfer function are also used here. From the
figure it is seen that when H i (s ) and H v (s ) are both zero, the output
impedance is high at low frequencies. When H i (s ) is changed to 1 D ' and
H v (s ) is changed to 1 (D' R ) , the output impedance is reduced at low
frequencies but increased at high frequencies. However, the maximum
impedance is decreased. The output impedance is not invariant for different
loads at low frequencies due to FRHP (s ) in the numerator of (7.151).
Simulation results are also plotted in Figure 7.22 and they are in good
agreement with (7.151) except at low frequencies in the case where
H i (s ) = 1 D ' and H v (s ) = 1 (D' R ) . In this case, the phase shift and the slope
of the magnitude curve are almost zero at low frequencies. It has also been
noticed that the simulation results are very sensitive to changes in D in this
case. None of these results are predicted by (7.151). Note that simulation
results for the frequency 20 Hz are included in Figure 7.22. The simulation
model shown in Figure 7.16 is used, except the buck converter subsystem is
replaced with the boost converter subsystem shown in Figure 2.14. The
parameters used in the simulation model presented in Section 3.6 are also
used here. The constant Ic2 is adjusted manually so that the average value of
the duty cycle, D , is equal to 0.382. Note once again that the average value
of the output voltage, Vo , is a little higher in the case where R = Rmax
compared to the case where R = Rmin according to (2.124) and (2.122).
316 Chapter 7. Using Load Current for Control

10

-10

-20
Phase (deg); Magnitude (dB)

-30

-40
1 2 3 4
10 10 10 10

100

50

-50

-100
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 7.22: The output impedance of a boost converter controlled by

(7.106). Dash-dotted line (): H i (s ) =0, H v (s ) =0, and
R = Rmin . Dotted line (O): H i (s ) =0, H v (s ) =0, and
R = Rmax . Solid line (+): H i (s ) = 1 D ' , H v (s ) = 1 (D' R ) , and
R = Rmin . Dashed line (x): H i (s ) = 1 D ' , H v (s ) = 1 (D' R ) ,
and R = Rmax .

Audio Susceptibility

In this subsection, the audio susceptibility is derived for the case where
the measured load current is used for control. The result is analyzed and
compared with simulation results.
The approximate audio susceptibility (6.121) is repeated here for
convenience:

 vˆo ( s ) 
  =  RTs D ' m c D'− F f ( s) + 1  KFl ( s ) FESR ( s ) Fh ( s ) ,
( )
 vˆ g ( s )  (7.152)
  ol 2  L D' 
Chapter 7. Using Load Current for Control 317

where

1 Ts
F f ( s) = − s, (7.153)
2 12

K is defined in (7.116), Fl (s) is defined in (7.117), FESR (s ) is defined in

(7.28), and Fh (s) is defined in (7.29). F f (s ) in (7.153) is used in this
section and it is a Taylor polynomial of degree 1 of F f (s ) in (6.120).
To obtain the audio susceptibility of the closed loop system where
(7.106) is used, the two input signals iˆinj ( s ) and iˆc 2 ( s ) are set to zero in
(7.113):

 vˆo ( s ) 
 
 vˆ g ( s ) 
ˆv o ( s)   ol 2 (7.154)
= .
ˆv g ( s ) vˆo ( s) 1
1− H (s)
iˆc ( s ) R

(7.154) is rewritten by using (7.152) and (7.115):

 RTs D ' 1
vˆo ( s)  L
 ( )
mc D '− F f ( s ) +  KFl ( s ) FESR ( s ) Fh ( s )
D' 
= =
vˆ g ( s ) 1 − D' KFl ( s ) FESR ( s ) Fh ( s ) FRHP ( s ) H ( s)
(7.155)
 Ts D ' 1 
 ( )
mc D '− F f ( s ) +  FESR ( s) Fh ( s )
 L RD ' 
.
R −1 K −1 Fl−1 ( s) − R −1 D' FESR ( s ) Fh ( s) FRHP ( s ) H ( s )

According to a discussion in the beginning of this section, H v (s ) can be

considered to be zero in the case where the audio susceptibility is analyzed.
H (s ) in (7.114) can therefore be replaced by H i (s ) . (7.155) is rewritten by
using this conclusion and (7.121):
318 Chapter 7. Using Load Current for Control

vˆo ( s)  Ts D ' 1 
=
ˆv g ( s )  L
(
mc D '− F f ( s ) +)  FESR ( s ) Fh ( s ) •
RD' 
 
 C  s + 1 −  2
D'
 FESR ( s ) Fh ( s ) FRHP ( s) H i ( s )  + (7.156)
   2  RC
−1
T 
D ' s (mc − 0.5) 
3
.
LC 

Note that the denominator is exactly the same as in (7.122). When H i (s ) =0,
an increase in R makes the numerator and the denominator in (7.156) both
smaller and the audio susceptibility can in some cases be almost invariant for
different loads at low frequencies. When H i (s ) = 2 D ' , an increase in R
makes the numerator smaller. The denominator is on the other hand almost
invariant for different loads at low frequencies. The conclusion is that the
audio susceptibility is closer to invariance for different loads when H i (s ) =0
compared to when H i (s ) = 2 D ' . For the buck converter, the conclusion was
the opposite.
Figure 7.23 shows the Bode plot for the audio susceptibility in (7.156)
when different H i (s ) and loads are used. The parameter values used for the
control-to-output transfer function are also used here. From the figure it is
seen that for H i (s ) = 2 D ' , the gain changes considerably for different loads
at low frequencies. For H i (s ) =0, the gain is almost invariant for different
loads at low frequencies. Simulation results are also plotted in Figure 7.23
and they are in good agreement with (7.156). The simulation model used for
the output impedance and its parameters are also used here.

The Difference Between the Simulation Results and the

Model

From Figure 7.18, it is seen that there is a small difference between the
simulation results and the model Gvo ic 2 ( s) in (7.122) for the case where
H i (s ) = 2 D ' and R = Rmin . The reason for this difference is explained in
this subsection.
The simulation results shown in Figure 7.18 were obtained with the
simulation model in Figure 7.15 (with the buck converter subsystem
replaced). It is interesting to compare these simulation results with the ones
Chapter 7. Using Load Current for Control 319

20

-20
Phase (deg); Magnitude (dB)

-40

1 2 3 4
10 10 10 10

-50

-100

-150
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 7.23: The audio susceptibility of a boost converter controlled by

(7.21). Dash-dotted line (): H i (s ) =0 and R = Rmin . Dotted
line (O): H i (s ) =0 and R = Rmax . Solid line (+):
H i (s ) = 2 D ' and R = Rmin . Dashed line (x): H i (s ) = 2 D '
and R = Rmax .

obtained with the simulation model in Figure 7.16 (with the buck converter
subsystem replaced) where the outer voltage controller is excluded. Figure
7.24 shows the Bode plot for Gvo ic 2 ( s) in (7.122) and the simulation results
obtained with the two simulation models for the case where H i (s ) = 2 D '
and R = Rmin . It is seen that the difference between the simulation results
and the model Gvo ic 2 ( s) is increased when the outer voltage controller is
removed. The model Gvo ic 2 ( s) is independent of any outer voltage controller
and does therefore not predict this change.
The model Gvo ic 2 ( s) is derived from the model Gvoic (s ) in (7.115).
Assume that iˆc (t ) is a sinusoidal signal with the frequency ω m . Gvoic (s )
then predicts the Fourier component in the output voltage with the frequency
ω m . However, there are other frequency components (Perreault and
Verghese, 1997), e.g. components with the frequencies ω s − ω m , ω s + ω m ,
320 Chapter 7. Using Load Current for Control

10

-10
Phase (deg); Magnitude (dB)

-20

-30
1 2 3 4
10 10 10 10

-50

-100

-150

-200

-250
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 7.24: The control-to-output transfer function of a boost converter for

the case where H i (s ) = 2 D ' and R = Rmin . Solid line: model.
+: simulation where the outer voltage controller is included and
K p =2. X: simulation where the outer voltage controller is
excluded.

2ω s − ω m , 2ω s + ω m , K , where ω s is the switching frequency. The load

current is proportional to the output voltage. iˆc (t ) is therefore not sinusoidal
if the load current is used to affect iˆc (t ) . iˆc (t ) is sampled in current-mode
control (see Section 3.3) and the sampling frequency is (in average) ω s . Due
to aliasing (Åström and Wittenmark, 1997, Section 7.4), the components in
iˆc (t ) with the frequencies ω s − ω m , ω s + ω m , 2ω s − ω m , 2ω s + ω m , K ,
affect the frequency component in iˆc ( z ) with the frequency ω m . This means
that the model Gvoic (s ) may not be valid if iˆc (t ) has components with
frequencies higher than ω n . The use of the model Gvoic (s ) in the derivation
of the model Gvo ic 2 ( s) introduces an error since the model Gvoic (s ) is used
in a situation where it is not valid.
The following is obtained from Figure 7.15 if H v (s ) =0:
Chapter 7. Using Load Current for Control 321

10

-10
Phase (deg); Magnitude (dB)

-20

-30
1 2 3 4
10 10 10 10

-50

-100

-150

-200

-250
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 7.25: The control-to-output transfer function of a boost converter for

the case where H i (s ) = 2 D ' and R = Rmin . Solid line: model.
+: simulation where the outer voltage controller is included and
K p = 2 (RD') . X: simulation where the outer voltage controller
is excluded.

iˆc ( s) = ic 2 dist ( s ) − K p vˆo ( s ) + H i ( s )iˆload ( s) =

ic 2 dist ( s ) − K p Riˆload ( s ) + H i ( s)iˆload ( s ) = (7.157)
( )
ic 2 dist ( s ) + H i ( s ) − K p R iˆload ( s ) .

iˆc (t ) is therefore a sinusoidal signal with the frequency ω m in the case where
H i (s ) = 2 D ' and K p = 2 (RD') . All the high-frequency components in
iˆc (t ) are canceled since the load current is proportional to the output voltage
and there is a minus sign in the output voltage loop and a plus sign in the
load current loop. Figure 7.25 shows the same as Figure 7.24 except K p is
equal to 2 (RD') instead of 2 in the case where an outer voltage controller is
used in the simulation model. Previously, it was seen from Figure 7.24 that
the difference between the simulation results and the model Gvo ic 2 ( s) was
322 Chapter 7. Using Load Current for Control

decreased when the outer voltage controller with K p =2 was included. By

comparing Figure 7.24 and Figure 7.25, it is seen that the difference between
the simulation results and the model Gvo ic 2 ( s) decreases further when K p
increases to 2 (RD') . The high-frequency components in iˆc (t ) decreases
continuously when K p increases continuously from 0 (no voltage controller)
to 2 (RD') .
From Figure 7.25, it is seen that there is a difference between the
simulation results and the model Gvo ic 2 ( s) in the case where K p is equal to
2 (RD') even though the model Gvoic (s ) is used in a situation where it
should be valid. The difference is largest at low frequencies. If Figure 7.18 is
zoomed, it is seen that difference in magnitude between the simulation results
and the model Gvo ic 2 ( s) is in order of 0.1 dB for the case where R = Rmin
and H i (s ) =0, i.e. the model Gvoic (s ) is actually considered. Hence, it seems
that Gvo ic 2 ( s) is very sensitive to modeling errors in Gvoic (s ) at low
frequencies. To investigate this, assume that the transfer function that
correctly describes how iˆc (t ) affects (the component ω m in) vˆo (t ) in the
simulation model is denoted by Gvoic (s ) . Assume further that the transfer
function that correctly describes how iˆc 2 (t ) affects (the component ω m in)
vˆo (t ) in the simulation model is denoted by Gvoic 2 ( s ) . The following are
obtained from (7.119) and the fact that H (s ) can be set to H i (s ) :

G voic ( s )
Gvoic 2 ( s ) = , (7.158)
1 − Gvoic ( s ) H i ( s ) R −1

G voic ( s )
Gvoic 2 ( s ) = . (7.159)
1 − Gvoic ( s ) H i ( s ) R −1

Let ∆ voic (s ) and ∆ voic 2 ( s ) denote the relative error in Gvoic (s ) and
Gvo ic 2 ( s) , respectively, such that

( )
Gvoic ( s ) = 1 + ∆ voic ( s ) G voic ( s ) , (7.160)

( )
Gvoic 2 ( s ) = 1 + ∆ voic 2 ( s ) Gvoic 2 ( s ) . (7.161)

The following expression is obtained by using (7.161), (7.159), (7.158), and

(7.160):
Chapter 7. Using Load Current for Control 323

G voic 2 ( s)
∆ voic 2 ( s ) = −1=
G voic 2 ( s)
G voic ( s ) 1 − G voic ( s ) H i ( s) R −1
−1 =
1 − Gvoic ( s) H i ( s ) R −1 Gvoic ( s )
1
•
1 − Gvoic ( s) H i ( s ) R −1

 oc (
 G v i ( s ) 1 − G v i ( s ) H i ( s) R −1
o c
) (
− 1 − Gvoic ( s ) H i ( s ) R −1

)
=
(7.162)

 Gvoic ( s) 
 
1  Gvoic ( s) 
 − 1 =
−1  G 
1 − Gvoic ( s) H i ( s ) R  voic ( s ) 
1
∆ v i ( s ) = S ( s )∆ voic ( s ) ,
1 − Gvoic ( s) H i ( s ) R −1 o c

where

1
S ( s) = . (7.163)
1 − G voic ( s ) H i ( s ) R −1

is the sensitivity function. Since Gvoic ( s ) ≈ Gvoic ( s ) , S (s ) can be

approximated with

1
S ( s) ≈ , (7.164)
1 − G voic ( s ) H i ( s) R −1

where Gvoic (s ) is defined in (7.115). Figure 7.26 shows the Bode plot for the
approximate sensitivity function in (7.164) when different loads, Rmin =1 Ω
and Rmax =4 Ω, are used. The parameter values used to obtain Figure 7.25
are also used here. As expected, the sensitivity is very high at low frequencies
in the case where R = Rmin . The sensitivity is lower at low frequencies for the
case where R = Rmax . To see if the sensitivity function in (7.164) is correct,
324 Chapter 7. Using Load Current for Control

25
20
15
10
Phase (deg); Magnitude (dB)

5
0
-5
1 2 3 4
10 10 10 10

-20

-40

-60

1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 7.26: The approximate sensitivity function. Solid line: R = Rmin .

Dashed line: R = Rmax .

the case where ω m =100π rad/s and R = Rmin is now checked. The
magnitudes will not be expressed in dB. From Figure 7.18,
Gvoic ≈ 0.2820∠ − 5.22o and Gvoic ≈ 0.2784∠ − 5.16o are obtained.
∆ voic (s ) ≈ 0.01281∠175.37 o according to (7.160). From Figure 7.26,
S ≈ 8.114∠ − 42.32o is obtained. ∆ voic 2 ( s ) ≈ 0.1039∠133.05o according to
(7.162). From Figure 7.25, Gvoic 2 ≈ 2.288∠ − 47.54 o is obtained. An estimate
of Gvoic 2 is calculated by using (7.161) and the result is approximately
2.133∠ − 42.87 o . This is compared with Gvoic 2 ≈ 2.128∠ − 43.49 o , obtained
from Figure 7.25. The sensitivity function in (7.164) thus seems to be
correct. Note that the relative error in (7.164) is large where the difference in
the denominator is small, i.e. where the sensitivity is high.
To summarize, there are two main reasons for the difference between the
simulation results and the model Gvo ic 2 ( s) in (7.122) seen in Figure 7.18.
The first main reason is that the use of the load current and the output
voltage when calculating iˆc (t ) (usually) makes iˆc (t ) containing components
with frequencies higher than ω n . The simulation results are affected by these
Chapter 7. Using Load Current for Control 325

components but the model Gvoic (s ) , which Gvo ic 2 ( s) is derived from, is not
designed to handle these components. The second main reason is that
Gvo ic 2 ( s) is very sensitive to model errors in Gvoic (s ) according to Figure
7.26.
Consider again the first main reason. As seen from Figure 7.25, the
change in the simulation results is larger at low and high frequencies
compared to at the frequencies in the middle. It seems to be rather
complicated to analyze how the change varies with frequency since there
seems to be several causal connections involved. However, it seems that the
main reason for the relatively large change at very low frequencies is the high
sensitivity to changes in Gvoic (s) according to Figure 7.26. It also seems that
the main reason for the relatively large change at high frequencies is the
significant increase in the magnitude of the component with frequency
ω s − ω m in iˆc (t ) when ω m increases towards ω n .

7.6 Properties of the Buck-Boost Converter

In Section 6.4, approximate expressions for the buck-boost converter with
current-mode control were derived. In this section, these expressions are used
to analyze how the control-to-output transfer function, the output impedance
and the audio susceptibility are affected when using load current
measurements to control the converter. An expression for the output voltage
is first derived and this is then used to derive the three transfer functions. The
results are compared with simulation results. In this section, it is assumed that
the conditions in (6.149)-(6.155) are fulfilled since the approximate
expressions derived in Section 6.4 may not be valid otherwise. Many of the
expressions in this section are identical to the ones obtained in Section 7.5.

An Expression for the Output Voltage

The control law (7.3) for the buck-boost converter is nonlinear. The
linearized version of this control law will now be considered. To linearize
(7.3), the following partial derivatives is first calculated:

∂ic (t )
=1, (7.165)
∂ic 2 (t )
326 Chapter 7. Using Load Current for Control

∂ic (t ) v o (t ) + v g (t )
= , (7.166)
∂iload (t ) v g (t )

∂ic (t ) iload (t )
= , (7.167)
∂v o (t ) v g (t )

∂ic (t ) v (t )i (t )
= − o 2load . (7.168)
∂v g (t ) v g (t )

The linearized version of (7.3) is obtained by using (7.165)-(7.168):

Vo + V g I V I
iˆc (t ) = iˆc 2 (t ) + iˆload (t ) + load vˆo (t ) − o load vˆ g (t ) . (7.169)
Vg Vg V g2

From Figure 2.20, it is seen that the load current is

v o (t )
iload (t ) = + iinj (t ) . (7.170)
R

It is assumed that the dc value of iinj (t ) is equal to zero (see (2.56)). The
following is therefore obtained from (7.170):

Vo
I load = . (7.171)
R

(7.169) is approximated by using (7.171), (2.166), (2.168), (6.149), and

(2.39):

1 D D2
iˆc (t ) = iˆc 2 (t ) + iˆload (t ) + vˆo (t ) − 2 vˆ g (t ) . (7.172)
D' D' R D' R

The last term in (7.172) is a feedforward of the input voltage and it is

here regarded as a part of ic 2 (t ) (see Section 7.5). Hence, the control law
(7.172) is replaced by:
Chapter 7. Using Load Current for Control 327

1 D
iˆc (t ) = iˆc 2 (t ) + iˆload (t ) + vˆo (t ) . (7.173)
D' D' R

To make the analysis in this section general, two transfer functions,

H i (s ) and H v (s ) , are introduced in (7.173):

iˆc ( s) = iˆc 2 ( s) + H i ( s )iˆload ( s ) + H v ( s )vˆo ( s ) . (7.174)

Figure 7.17 shows the system obtained by using (7.174) and (7.170). If the
control law (7.173) is used, H i (s ) and H v (s ) are identified as

1
H i ( s) = , (7.175)
D'

D
H v (s) = . (7.176)
D' R

(7.173) is rewritten as follows by using (7.170) if iinj (t ) is zero:

1 D ˆ
iˆc (t ) = iˆc 2 (t ) + iˆload (t ) + Riload (t ) =
D' D' R
(7.177)
1+ D ˆ
iˆc 2 (t ) + iload (t ) .
D'

Hence, if the control law (7.173) is used and iinj (t ) is zero, H i (s ) and
H v (s ) can equivalently be identified as

1+ D
H i ( s) = , (7.178)
D'

H v ( s) = 0 . (7.179)

In the case where the control-to-output transfer function or the audio

susceptibility is analyzed, iinj (t ) is considered to be zero and (7.178) and
(7.179) can be used. The contribution H v ( s )vˆo ( s ) can always be moved to
328 Chapter 7. Using Load Current for Control

H i ( s )iˆload ( s ) in these two cases and H v (s ) can be considered to be zero.

The control law (7.174) can then be replaced by (7.21) (see Figure 7.9).
(7.113) is an expression for the output voltage for the buck-boost
converter since (7.174) and (7.170) are the same as (7.106) and (7.102),
respectively.

Control-to-Output Transfer Function

In this subsection, the control-to-output transfer function is derived for

the case where the measured load current is used for control. The result is
analyzed and compared with simulation results.
The approximate control-to-output transfer function (6.157) is repeated
here for convenience:

vˆo ( s )
Gvoic ( s ) = = RD ' KFl ( s ) FESR ( s ) Fh ( s) FRHP ( s ) , (7.180)
iˆc ( s )

where

1
K= 3
,
RD ' Ts (7.181)
1+ D + (mc − 0.5)
L

1
Fl ( s) = , (7.182)
1 + sRCK

LD
FRHP ( s ) = 1 − s , (7.183)
RD ' 2

FESR (s) is defined in (7.28), Fh (s) is defined in (7.29), and mc is defined

in (3.21).
To obtain the control-to-output transfer function of the closed loop
system where (7.174) is used, the two input signals vˆ g ( s ) and iˆinj ( s ) are set
to zero in (7.113):
Chapter 7. Using Load Current for Control 329

vˆo ( s )
vˆo ( s ) iˆc ( s )
= . (7.184)
iˆc 2 ( s ) vˆ ( s ) 1
1− o H (s)
iˆc ( s ) R

(7.184) is rewritten by using (7.180):

vˆo ( s ) RD ' KFl ( s ) FESR ( s ) Fh ( s ) FRHP ( s )

Gvoic 2 ( s ) = = =
iˆc 2 ( s ) 1 − D' KFl ( s ) FESR ( s ) Fh ( s ) FRHP ( s) H ( s )
(7.185)
D ' FESR ( s ) Fh ( s ) FRHP ( s)
.
R −1 K −1 Fl−1 ( s ) − R −1 D ' FESR ( s) Fh ( s ) FRHP ( s ) H ( s)

The (last) denominator in (7.185) is rewritten by using (7.182) and (7.181):

R −1 K −1 Fl−1 ( s ) − R −1 D ' FESR ( s ) Fh ( s ) FRHP ( s ) H ( s) =

R −1 K −1 + sC − R −1 D ' FESR ( s ) Fh ( s ) FRHP ( s ) H ( s ) =
D ' 3 Ts
(1 + D )R −1 + (mc − 0.5) + sC −
L
R −1 D ' FESR ( s ) Fh ( s ) FRHP ( s ) H ( s) = (7.186)
  D' 1+ D
C  s + 1 − FESR ( s ) Fh ( s ) FRHP ( s) H ( s )  +
  1+ D  RC

(mc − 0.5) .
Ts
D'3
LC 

According to a discussion previously in this section, H v (s ) can be considered

to be zero in the case where the control-to-output transfer function is
analyzed. H (s ) in (7.114) can therefore be replaced by H i (s ) . (7.185) is
rewritten by using this conclusion and (7.186):
330 Chapter 7. Using Load Current for Control

vˆo ( s )
Gvoic 2 ( s ) = = D' FESR ( s ) Fh ( s ) FRHP ( s ) •
iˆc 2 ( s )
  1+ D
 C  s + 1 −
D'
 FESR ( s ) Fh ( s) FRHP ( s ) H i ( s)  + (7.187)
   1+ D  RC
−1

(mc − 0.5) 
Ts
D'3 .
LC 

A new variable, F (s ) , is now introduced:

F ( s ) = FESR ( s ) Fh ( s ) FRHP ( s ) . (7.188)

The load resistance, R , occurs explicitly only at one place in (7.122) but it
also occurs implicitly in FRHP (s ) , see (7.183). The more H i (s ) is in
accordance with (1 + D ) (D ' F ( s ) ) , the closer invariance for different loads is
the denominator in the control-to-output transfer function. This is not easily
made since 1 F ( s ) has a right half plane pole and is unstable. If F (s ) is
approximately equal to 1 at low frequencies a good choice is to set H i (s )
equal to (1 + D ) D ' which is the same as (7.178). F (s ) is approximately
equal to 1 at low frequencies if FESR (s ) , Fh (s ) , and FRHP (s ) are
approximately equal to 1 at low frequencies. Condition (6.149) sets a lower
limit for the corner frequency of FESR (s ) . Condition (6.155) sets a lower
limit for Q and therefore also a lower limit for the corner frequency of
Fh (s ) (see upper plot in Figure 7.10). None of the conditions (6.149)-
(6.155) sets an upper limit for L and therefore no lower limit for the corner
frequency of FRHP (s ) .
Figure 7.27 shows the Bode plot for Gvo ic 2 ( s) in (7.187) when different
H i (s ) and loads, Rmin =1 Ω and Rmax =4 Ω, are used. Except for R , the
parameter values shown in Table 2.6 are used. mc is set to 2. From the figure
it is seen that for H i (s ) =0, the gain and phase shift changes considerably for
different loads. For H i (s ) = (1 + D ) D ' , the gain is almost invariant for
different loads for low frequencies but not for high frequencies. Simulation
results are also plotted in Figure 7.27 and they are in good agreement with
(7.187). The simulation model shown in Figure 7.15 is used, except the buck
converter subsystem is replaced with the buck-boost converter subsystem
shown in Figure 2.20. The parameters used in the simulation model
Chapter 7. Using Load Current for Control 331

20

10

-10
Phase (deg); Magnitude (dB)

-20

-30
1 2 3 4
10 10 10 10

-50

-100

-150

-200

-250
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 7.27: The control-to-output transfer function of a buck-boost

converter controlled by (7.21). Dash-dotted line (): H i (s ) =0
and R = Rmin . Dotted line (O): H i (s ) =0 and R = Rmax .
Solid line (+): H i (s ) = (1 + D ) D ' and R = Rmin . Dashed line
(x): H i (s ) = (1 + D ) D ' and R = Rmax .

presented in Section 3.7 are also used here. K p is set to 2 and the constant
Vo_offset is adjusted manually so that the average value of the duty cycle, D ,
is equal to 0.620. Note that the average value of the output voltage, Vo , is a
little higher in the case where R = Rmax compared to the case where R = Rmin
according to (2.168) and (2.166).

Output Impedance

In this subsection, the output impedance is derived for the case where the
measured load current is used for control. The result is analyzed and
compared with simulation results.
The approximate output impedance (6.166) is repeated here for
convenience:
332 Chapter 7. Using Load Current for Control

 vˆ ( s ) 
(Z out (s))ol 2 = −  ˆ o  = RKF ( s ) F ( s ) ,
 l ESR (7.189)
i ( s )
 inj  ol 2

where K is defined in (7.181), Fl (s) is defined in (7.182), and FESR (s ) is

defined in (7.28).
To obtain the output impedance of the closed loop system where (7.174)
is used, the two input signals vˆ g ( s ) and iˆc 2 ( s ) are set to zero in (7.113):

vˆo ( s )  vˆ ( s ) 
H i (s) +  o 
ˆc ( s)
i  ˆinj ( s) 
i
vˆ ( s )   ol 2 (7.190)
Z out ( s ) = − o =− .
iˆinj ( s ) vˆo ( s ) 1
1− H ( s)
iˆc ( s ) R

(7.190) is rewritten by using (7.189) and (7.180):

Z out ( s ) =
RD ' KFl ( s) FESR ( s) Fh ( s ) FRHP ( s ) H i ( s ) − RKFl ( s ) FESR ( s )
− =
1 − D ' KFl ( s ) FESR ( s ) Fh ( s) FRHP ( s ) H ( s ) (7.191)
(1 − D' Fh ( s) FRHP (s) H i ( s))FESR (s)
.
R K Fl−1 ( s ) − R −1 D ' FESR ( s) Fh ( s ) FRHP ( s ) H ( s)
−1 −1

(7.191) is rewritten by using (7.186):

Z out ( s ) = (1 − D' Fh ( s ) FRHP ( s ) H i ( s) )FESR ( s ) •

  1+ D
 C  s + 1 −
D'
 FESR ( s ) Fh ( s ) FRHP ( s ) H ( s )  +
   1+ D  RC (7.192)
−1

(mc − 0.5) 
Ts
D'3 ,
LC 

where H (s ) is defined in (7.114). From (7.192), it is seen that the more

H i (s ) is in accordance with 1 (D' Fh ( s) FRHP ( s ) ) , the lower is the output
impedance. This is not easily made since FRHP (s ) has a right half plane zero.
Chapter 7. Using Load Current for Control 333

10

-10

-20
Phase (deg); Magnitude (dB)

-30

-40
1 2 3 4
10 10 10 10

100

50

-50

-100
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 7.28: The output impedance of a buck-boost converter controlled by

(7.174). Dash-dotted line (): H i (s ) =0, H v (s ) =0, and
R = Rmin . Dotted line (O): H i (s ) =0, H v (s ) =0, and
R = Rmax . Solid line (+): H i (s ) = 1 D ' , H v (s ) = D (D ' R ) , and
R = Rmin . Dashed line (x): H i (s ) = 1 D ' , H v (s ) = D (D ' R ) ,
and R = Rmax .

Fh ( s ) FRHP ( s ) is approximately equal to 1 at low frequencies so a good

choice is to set H i (s ) equal to 1 D ' which is the same as (7.175).
Figure 7.28 shows the Bode plot for the output impedance in (7.192)
when different H i (s ) , H v (s ) , and loads are used. The parameter values
used for the control-to-output transfer function are also used here. From the
figure it is seen that when H i (s ) and H v (s ) are both zero, the output
impedance is high at low frequencies. When H i (s ) is changed to 1 D ' and
H v (s ) is changed to D (D ' R ) , the output impedance is reduced at low
frequencies but increased at high frequencies. However, the maximum
impedance is decreased. The output impedance is not invariant for different
loads at low frequencies due to FRHP (s ) in the numerator of (7.192).
Simulation results are also plotted in Figure 7.28 and they are in good
agreement with (7.192) except at low frequencies in the case where
334 Chapter 7. Using Load Current for Control

H i (s ) = 1 D ' and H v (s ) = D (D ' R ) . The simulation model shown in Figure

7.16 is used, except the buck converter subsystem is replaced with the buck-
boost converter subsystem shown in Figure 2.20. The parameters used in the
simulation model presented in Section 3.7 are also used here. The constant
Ic2 is adjusted manually so that the average value of the duty cycle, D , is
equal to 0.620. To obtain D =0.620 in steady state, not just Ic2 must be
adjusted but also the initial voltage across the output capacitor. The initial
voltage must be close to the output voltage corresponding to D =0.620. The
initial values of the states in the simulator are set by using a powergui block.
Note that this was not necessary for the boost converter where the duty cycle
converged to the correct value even if the initial values were zero. The model
that we derived for the buck-boost converter, which predicted stability, is a
small-signal model and may result in erroneous conclusions if the deviations
from the specified operating point are large.

Audio Susceptibility

In this subsection, the audio susceptibility is derived for the case where
the measured load current is used for control. The result is analyzed and
compared with simulation results.
The approximate audio susceptibility (6.178) is repeated here for
convenience:

 vˆo ( s ) 
  =
 vˆ g ( s ) 
  ol 2 (7.193)
 RTs D ' 
 ( )
mc D'− F f 1 ( s ) +
D
F f 2 ( s )  DKFl ( s ) FESR ( s ) Fh ( s) ,
 L D' 

where

D (3 − 2 D )DTs
F f 1 ( s) = 1 − − s, (7.194)
2 12

D ' Ts
F f 2 (s) = 1 + s, (7.195)
2
Chapter 7. Using Load Current for Control 335

K is defined in (7.181), Fl (s) is defined in (7.182), FESR (s ) is defined in

(7.28), and Fh (s) is defined in (7.29). F f 1 ( s) and F f 2 ( s ) in (7.194) and
(7.195) are used in this section and they are Taylor polynomials of degree 1
of F f 1 ( s) and F f 2 ( s ) in (6.176) and (6.177).
To obtain the audio susceptibility of the closed loop system where
(7.174) is used, the two input signals iˆinj ( s ) and iˆc 2 ( s ) are set to zero in
(7.113):

 vˆo ( s ) 
 
 vˆ g ( s ) 
vˆo ( s)   ol 2 (7.196)
= .
vˆ g ( s ) vˆo ( s) 1
1− H (s)
iˆc ( s ) R

(7.196) is rewritten by using (7.193) and (7.180):

vˆo ( s)
=
vˆ g ( s )
 RTs D ' 
 (
mc D '− F f 1 ( s ) +)D
F f 2 ( s )  DKFl ( s ) FESR ( s ) Fh ( s )
 L D' 
= (7.197)
1 − D ' KFl ( s ) FESR ( s ) Fh ( s ) FRHP ( s ) H ( s )
 Ts D ' 
 ( )
mc D '− F f 1 ( s ) +
D
F f 2 ( s )  DFESR ( s ) Fh ( s )
 L RD' 
.
R −1 K −1 Fl−1 ( s ) − R −1 D' FESR ( s ) Fh ( s) FRHP ( s ) H ( s )

According to a discussion in the beginning of this section, H v (s ) can be

considered to be zero in the case where the audio susceptibility is analyzed.
H (s ) in (7.114) can therefore be replaced by H i (s ) . (7.197) is rewritten by
using this conclusion and (7.186):
336 Chapter 7. Using Load Current for Control

vˆo ( s)  Ts D ' 
=
ˆv g ( s )  L
( )
mc D '− F f 1 ( s ) +
D
F f 2 ( s )  DFESR ( s ) Fh ( s ) •
RD' 
  1+ D
 C  s + 1 −
D'
 FESR ( s ) Fh ( s ) FRHP ( s ) H i ( s )  +
   1+ D  RC (7.198)
−1
T 
D ' s (mc − 0.5) 
3
.
LC 

When H i (s ) =0, an increase in R makes the numerator and the

denominator in (7.198) both smaller and the audio susceptibility can in some
cases be almost invariant for different loads at low frequencies. When
H i (s ) = (1 + D ) D ' , an increase in R makes the numerator smaller. The
denominator is on the other hand almost invariant for different loads at low
frequencies. The conclusion is that the audio susceptibility is closer to
invariance for different loads when H i (s ) =0 compared to when
H i (s ) = (1 + D ) D ' .
Figure 7.29 shows the Bode plot for the audio susceptibility in (7.198)
when different H i (s ) and loads are used. The parameter values used for the
control-to-output transfer function are also used here. From the figure it is
seen that for H i (s ) = (1 + D ) D ' , the gain changes considerably for different
loads at low frequencies. For H i (s ) =0, the gain is almost invariant for
different loads at low frequencies. Simulation results are also plotted in Figure
7.29 and they are in good agreement with (7.198). The simulation model
used for the output impedance and its parameters are also been used here.

7.7 Summary and Concluding Remarks

The output voltage and the inductor current are measured in the case
where current-mode control is used. Some properties that can be obtained
when the controller also uses load current measurements were in this chapter
analyzed and some of the previous works were reviewed.
The main results are summarized here.
Chapter 7. Using Load Current for Control 337

20

0
Phase (deg); Magnitude (dB)

-20

-40

1 2 3 4
10 10 10 10

-20

-40

-60

-80

-100
1 2 3 4
10 10 10 10
Frequency (Hz)

Figure 7.29: The audio susceptibility of a buck-boost converter controlled by

(7.21). Dash-dotted line (): H i (s ) =0 and R = Rmin . Dotted
line (O): H i (s ) =0 and R = Rmax . Solid line (+):
H i (s ) = (1 + D ) D ' and R = Rmin . Dashed line (x):
H i (s ) = (1 + D ) D ' and R = Rmax .

1. The analysis confirms that low output impedance can be obtained.

2. The analysis shows that in the case where the load is a current source the
following properties are obtained:
• The use of measured load current for control is feedforward.
• The control-to-output transfer function does not change when this
feedforward is introduced.

3. The analysis shows that in the case where the load is a linear resistor, the
following properties are obtained:
• The control-to-output transfer function can change when the
measured load current is introduced for control.
• The converter can become unstable when the measured load current
is introduced for control.
338 Chapter 7. Using Load Current for Control

• The control-to-output transfer function can be almost invariant for

different linear resistive loads if the measured load current is used for
control. This is especially the case for the buck converter.
• The use of measured load current for control is not feedforward. It
can instead be seen as gain scheduling.

Expressions for the control-to-output transfer function, the output

impedance, and the audio susceptibility were derived from the expressions
obtained in Chapter 6. There are differences between simulation results and
the predictions of these derived expressions in the case where the boost or the
buck-boost converter is considered. The difference was examined in one
specific case. We concluded that there are two main reasons for the difference.
The first main reason is that components with frequencies higher than half
the switching frequency affects the control signal and the expression is derived
without taking these into account. The second main reason is that the
expression can be very sensitive to errors in the used expressions from Chapter
6. The differences in the other cases are probably also due to these two
reasons.
The approximate expressions obtained in Chapter 6 were used as a
starting point in Section 7.4, 7.5, and 7.6. This is not the best way to derive
reliable expressions for the closed loop system obtained when the load current
is used for control. Instead, the non-approximate expressions should have
been used as a starting point and the derived expressions for the closed loop
system should have been approximated afterwards.
As an example, consider the control-to-output transfer function Gvoic 2 ( s )
for the buck converter, (7.33). The denominator is:

 1 T 
C  s + (1 − FESR ( s ) Fh ( s ) H i ( s ) ) + s (m c D '−0.5) . (7.199)
 RC LC 

If the non-approximate expressions are used and the approximation is made

afterwards, the following denominator is obtained:

 1 T 
C  s + (1 − Fh ( s) H i ( s ) ) + s (mc D '−0.5) . (7.200)
 RC LC 

If H i (s ) is equal to 0, there is no difference between (7.199) and (7.200). It

was shown in Section 7.4 that if H i (s ) is equal to 1, the absolute value of the
Chapter 7. Using Load Current for Control 339

second term is much smaller than the absolute value of the first term, s , in
the (largest) parenthesis in (7.199). The difference between (7.199) and
(7.200) is therefore negligible in the case where H i (s ) is equal to 1 if the
values of the denominators are considered. However, the two denominators
give two different answers to the question of how to set H i (s ) to obtain
(total) invariance for different loads. At high frequencies, it is a large
difference between the two answers H i (s ) = 1 / (FESR ( s ) Fh ( s ) ) and
H i (s ) = 1 / Fh ( s ) .
The results in Section 7.4, 7.5, and 7.6 are planned to be remade in such
a way that the approximations are made after the derivations. The new results
are planned to be published in the Ph.D. dissertation of the author of this
thesis.
 Chapter 8 Summary

The work presented in this thesis is summarized in this chapter.

Suggestions for future work are also summarized.

8.1 Results
This section explains which major models were obtained and how they
were derived. The main conclusions are also presented. However, the method
used to verify the obtained models is first explained.

Verification

Evaluation of a converter by means of a network analyzer is common and

this is one of the reasons for the interest in models that can predict the
frequency functions.
To verify the derived small-signal models, the frequency functions
predicted by them were compared with simulation results. Switched (large-
signal) simulation models were utilized and the output voltage then consists
of several Fourier components. To obtain the frequency function, one
frequency at the time was evaluated. A sinusoidal signal with frequency ω m
was injected and only the Fourier component with frequency ω m in the
output voltage was considered. A network analyzer also just considers this
Fourier component.
The control signal can be considered to be sampled with the switching
frequency. The frequency functions were therefore only evaluated for the
frequency interval dc to half the switching frequency.

341
342 Chapter 8. Summary

State-Space averaging

State-space averaging was used to derive linear continuous-time time-

invariant models for the buck, boost, and buck-boost converters. The control-
to-output transfer function, the output impedance, and the audio
susceptibility were extracted from each one of these models. We concluded
that these transfer functions are in good agreement with the simulation
results.

Current-Mode Control

The Ridley and Tan models were applied to the buck converter with
current-mode control. We concluded that the obtained control-to-output
transfer functions and the output impedances are in good agreement with the
simulation results but the obtained audio susceptibilities are not.
The Ridley model was also applied to the boost and buck-boost
converters with current-mode control. Also in these cases, we concluded that
the obtained audio susceptibilities are not in good agreement with the
simulation results.
The high-frequency extensions in the Ridley and Tan models are based
on an accurate control-to-current transfer function, which is derived with the
assumption that the changes in the input and output voltages are negligible.
The actual changes in the input and output voltages are in the Ridley and
Tan models taken into consideration by including two feedforward gains, k f
and kr . These gains are designed such that the dc gain should be correct.
This design results in modeling errors, especially for the audio susceptibilities
at high frequencies. The reason is that the amplitude of the perturbation in
the input voltage does not decrease at high frequencies since the injection
signal from the network analyzer affects the input voltage directly. However,
when the control-to-output transfer function and the output impedance are
considered, the input voltage is not affected and the changes in the output
voltage is negligible at high frequencies due to the low-pass character of the
output filter of the converters.

A Novel Model

A novel model for the audio susceptibility of converters with current-

mode control was derived by treating the changes in the input and output
voltages in a more refined way. The novel model was applied to the buck,
Chapter 8. Summary 343

boost and buck-boost converters. We concluded that the obtained audio

susceptibilities are in good agreement with simulation results at high
frequencies but in some cases not at low frequencies. For one of these cases,
we concluded that the reason for the poor predictions is that the novel model
is very sensitive at low frequencies to errors in some of models that it is based
on.

Improved Models

The novel model was used to improve the Ridley and Tan models. The
feedforward gains k f were changed such that the two models became equal
to the novel model. However, we used a combined model instead of the novel
model in the cases where the novel model makes poor low-frequency
predictions. To obtain the combined model, the low-frequency properties of
the (original) Ridley model was combined with the high-frequency properties
of the novel model
One disadvantage with the improved models is that k f in some cases was
an unstable transfer function.

Approximations of Obtained Expressions

The control-to-output transfer functions, the output impedances, and the

audio susceptibilities obtained from the (improved) Ridley model were
approximated. To be able to do this, several assumptions were introduced.

Using Load Current for Control

The output voltage and the inductor current are measured in the case
where current-mode control is used. Some properties that can be obtained
when the controller also uses load current measurements were analyzed. The
control-to-output transfer function, the output impedance, and the audio
susceptibility were derived for each treated converter topology: buck, boost,
and buck-boost. The main conclusions are presented here.

1. The analysis confirms that low output impedance can be obtained.

2. The analysis shows that in the case where the load is a current source the
following properties are obtained:
• The use of measured load current for control is feedforward.
344 Chapter 8. Summary

• The control-to-output transfer function does not change when this

feedforward is introduced.

3. The analysis shows that in the case where the load is a linear resistor, the
following properties are obtained:
• The control-to-output transfer function can change when the
measured load current is introduced for control.
• The converter can become unstable when the measured load current
is introduced for control.
• The control-to-output transfer function can be almost invariant for
different linear resistive loads if the measured load current is used for
control. This is especially the case for the buck converter.
• The use of measured load current for control is not feedforward. It
can instead be seen as gain scheduling.

The difference between the simulation results and the derived control-to-
output transfer functions, the output impedances, and the audio
susceptibilities are is some cases significant. For one specific case, we
concluded that the reasons for this are the fact that the measured load current
contains components with frequencies higher than half the switching
frequency and high sensitivity to errors in the models used as a starting point
in the analysis.

8.2 Future Work

Suggestions for future work are summarized here.

• Modify the novel model in such a way that the low-frequency predictions
are not so sensitive to errors.
• Find out how to improve the Ridley model without using unstable
transfer functions. An idea how to do this was presented in Section 5.5.
• Perform experiments to validate the main results in Chapter 7.
• Find out suitable methods to identify the load and use the result in an
adaptive controller in cases where the load is more complex. This was
discussed in Section 1.2.
 Chapter 9 Errata for Three
Papers

Errata for the following three conference papers are presented in this
chapter.

1. Johansson, B. and Lenells, M. (2000), Possibilities of obtaining

small-signal models of DC-to-DC power converters by means of
system identification, IEEE International Telecommunications Energy
Conference, pp. 65-75, Phoenix, Arizona, USA, 2000.
2. Johansson, B. (2002a), Analysis of DC-DC converters with current-
mode control and resistive load when using load current
measurements for control, IEEE Power Electronics Specialists
Conference, vol. 1, pp. 165-172, Cairns, Australia, 2002.
3. Johansson, B. (2002b), A comparison and an improvement of two
continuous-time models for current-mode control, IEEE
International Telecommunications Energy Conference, pp. 552-559,
Montreal, Canada, 2002.

Two page numbers is specified for each error. The second page number is
in parentheses and it is relative to the first page in the paper.

345
346 Chapter 9. Errata for Three Papers

9.1 Paper 1
This section presents errata for the paper “Possibilities of obtaining small-
signal models of DC-to-DC power converters by means of system
identification”, i.e. Johansson and Lenells (2000).

Page 67 (3), left column, line 1-2

The sentence should be:

The noise term v(t ) will not change if h(k ) is doubled and λ halved.

Page 72 (8), figure 12

The figure text should be:

The Model Gain Obtained from the Step Response (dotted

line) and the Parametric (OE) Model (solid line).

Page 72 (8), right column, line 11 from bottom

The line should be:

( ∞
) (
= α Re e iω t ∑k =1 g (k ) e −iω k = α Re e iω t G (e iω ) )

Page 74 (10), left column, line 2 from bottom

The line should be:

ˆ iω
ˆ N (ω )
Φ yu ∫−π Wγ (ξ − ω )YN (ξ )U N (ξ )dξ
G N (e ) = N = π
Φ (ω )
ˆ
∫−π Wγ (ξ − ω ) U N (ξ )
2
u dξ
Chapter 9. Errata for Three Papers 347

9.2 Paper 2
This section presents errata for the paper “Analysis of DC-DC converters
with current-mode control and resistive load when using load current
measurements for control”, i.e. Johansson (2002a).

Page 168 (4), equation 16

The equation should be:

Z out ( s ) = RKFl ( s ) FESR ( s ) .

Page 170 (6), figure 8

The figure should be:

20

0
Phase (deg); Magnitude (dB)

-20

-40
1 2 3 4
10 10 10 10

-50

-100

-150

-200

-250
1 2 3 4
10 10 10 10
Frequency (Hz)
348 Chapter 9. Errata for Three Papers

Page 171 (7), equation 30

The equation should be:

Z out ( s ) = RKFl ( s ) FESR ( s ) .

Page 172 (8), figure 10

The second sentence in the figure text should be:

Dash-dotted line (): H i (s ) =0 and R = Rmin .

Page 172 (8), equation 37

The equation should be:

Z out ( s ) = RKFl ( s ) FESR ( s ) .

Page 172 (8), list of references, reference 6

The year should be 2003.

9.3 Paper 3
This section presents errata for the paper “A comparison and an
improvement of two continuous-time models for current-mode control”, i.e.
Johansson (2002b).

Page 554 (3), left column, line 5

The line should be:

where V g is the DC component of the supply voltage.

Chapter 9. Errata for Three Papers 349

Page 555 (4), figure 4

The figure should be:

-10

-20
Phase (deg); Magnitude (dB)

-30

-40
1 2 3 4
10 10 10 10

-20

-40

-60

-80
1 2 3 4
10 10 10 10
Frequency (Hz)

Page 556 (5), left column, line 2 from bottom

The sentence should be:

The integral in (22) can be rewritten in a similar way.

Page 556 (5), figure 7

In the figure, m 2 (t ) should be replaced with − m 2 (t ) .

350 Chapter 9. Errata for Three Papers

Page 559 (8), left column, line 8-9

The sentence

The Tan model will however keep the extra zero since it
is independent of D .

should be replaced with:

The Tan model will however keep the extra zero, which
can be concluded from (16). s is equal to the extra zero if

 D s
m c D '− 1 −  + = 0.
 2  πω n

This equation is rewritten:

  D 
s = −πω n  mc D '− 1 −   =
  2 
  1 
− πω n  mc − 1 − D mc −   .
  2 

It is seen from this equation that the extra zero does not
disappear (move to infinity) as D tend to zero.

Page 559 (8), right column, line 2

The line should be:

1 1 − e − sTs D Ts D ' (1 − 3D + 2 D 2 ) Ts2 2

F f 2 (s) = = 1 + s + s +K,
D 1 − e − sTs 2 12

Page 559 (8), list of references, reference 7

The year should be 2003.

 Chapter 10 References

Chen, C. –T. (1999), Linear system theory and design, Third edition, Oxford
University Press, ISBN 0-19-511777-8.
Clique, M., Fossard, A. J. (1977), A general model for switching converters,
IEEE Transactions on Aerospace and Electronic Systems, vol. AES-13, no. 4,
pp. 397-400, 1977.
Erickson, R. W., Maksimovic, D. (2000), Fundamentals of power electronics,
Second edition, Kluwer Academic Publishers, ISBN 0-7923-7270-0.
Goodwin, G. C., Graebe, S. F., Salgado, M. E. (2001), Control System Design,
Prentice-Hall, ISBN 0-13-958653-9.
Hiti, S., Borojevic, D. (1993), Robust nonlinear control for boost converter,
IEEE Power Electronics Specialists Conference Record, pp. 191-196, 1993.
Johansson, B., Lenells, M. (2000), Possibilities of obtaining small-signal
models of DC-to-DC power converters by means of system identification,
IEEE International Telecommunications Energy Conference, pp. 65-75,
2000.
Johansson, B. (2002a), Analysis of DC-DC converters with current-mode
control and resistive load when using load current measurements for
control, IEEE Power Electronics Specialists Conference, vol. 1, pp. 165-172,
2002.
Johansson, B. (2002b), A comparison and an improvement of two
continuous-time models for current-mode control, IEEE International
Telecommunications Energy Conference, pp. 552-559, 2002.
Kislovski, A. S., Redl, R., Sokal, N. O. (1991), Dynamic analysis of switching-
mode dc/dc converters, Van Nostrand Reinhold, ISBN 0-442-23916-5.
Ljung, L. (1999), System identification: theory for the user, Second edition,
Prentice-Hall, ISBN 0-13-656695-2.
Lo, Y. –W., King, R. J. (1999), Sampled-data modeling of the average-input
current-mode-controlled buck converter, IEEE Transactions on Power
Electronics, vol. 14, no. 5, pp. 918-927, 1999.

351
352 Chapter 10. References

Middlebrook, R. D., Cuk, S. (1976), A general unified approach to modeling

switching converter power stages, IEEE Power Electronics Specialists
Conference Record, pp. 18-34, 1976.
Mitchell, D. M. (1988), Switching regulator analysis, McGraw-Hill, ISBN 0-
07-042597-3.
Perreault, D. J., Verghese, G. C. (1997), Time-varying effects and averaging
issues in models for current-mode control, IEEE Transactions on Power
Electronics, vol. 12, no. 3, pp. 453-461, 1997.
Poon, F. N. K., Tse, C. K., Liu, J. C. P. (1999), Very fast transient voltage
regulators based on load correction, IEEE Power Electronics Specialists
Conference, vol. 1, pp. 66-71, 1999.
Redl, R., Sokal, N. O. (1986), Near-optimum dynamic regulation of dc-dc
converters using feedforward and current-mode control, IEEE Transactions
on Power Electronics, vol. 1, no. 3, pp. 181-192, 1986.
Ridley, R. B. (1990a), A new, continuous-time model for current-mode
control with constant frequency, constant on-time, and constant off-time,
in CCM and DCM, IEEE Power Electronics Specialists Conference Record,
pp. 382-389, 1990.
Ridley, R. B. (1990b), A new small-signal model for current-mode control,
Ph.D. dissertation, Virginia Polytechnic Institute and State University,
Blacksburg, 1990.
Ridley, R. B. (1991), A new, continuous-time model for current-mode
control, IEEE Transactions on Power Electronics, vol. 6, no. 2, pp. 271-280,
1991.
Schoneman, G. K., Mitchell, D. M. (1989), Output impedance
considerations for switching regulators with current-injected control, IEEE
Transactions on Power Electronics, vol. 4, no. 1, pp. 25-35, 1989.
Tan, F. D., Middlebrook, R. D. (1995), A unified model for current-
programmed converters, IEEE Transactions on Power Electronics, vol. 10,
no. 4, pp. 397-408, 1995.
Tymerski, R., Li, D. (1993), State-space models for current programmed
pulsewidth-modulated converters, IEEE Transactions on Power Electronics,
vol. 8, no. 3, pp. 271-278, 1993.
Tymerski, R. (1994), Application of the time-varying transfer function for
exact small-signal analysis, IEEE Transactions on Power Electronics, vol. 9,
no. 2, pp. 196-205, 1994.
Verghese, G. C., Thottuvelil, V. J. (1999), Aliasing effects in PWM power
converters, IEEE Power Electronics Specialists Conference Record, vol. 2, pp.
1043-1049, 1999.
Chapter 10. References 353

Vorperian, V. (1990), Simplified analysis of PWM converters using model of

PWM switch: Part I and II, IEEE Transactions on Aerospace and Electronic
Systems, vol. 26, no. 3, pp. 490-505, 1990.
Walsh, J. L. (1935), Interpolation and approximation by rational functions in
the complex domain, American mathematical society, Colloquium
publications, vol. XX.
Wester, G. W., Middlebrook, R. D. (1973), Low-frequency characterization
of switched dc-dc converters, IEEE Transactions on Aerospace and Electronic
Systems, vol. AES-9, no. 3, pp. 376-385, 1977.
Åström, K. J., Hägglund, T. (1995), PID controllers; theory, design, and
tuning, Second edition, Instrument Society of America, ISBN 1-55617-
516-7.
Åström, K. J., Wittenmark, B. (1995), Adaptive control, Second edition,
Addison-Wesley, ISBN 0-201-55866-1.
Åström, K. J., Wittenmark, B. (1997), Computer-controlled systems: theory and
design, Third edition, Prentice Hall, ISBN 0-13-314899-8.