SMPS and Pspice
SMPS and Pspice
SMPS and Pspice
POWER SUPPLIES
SPICE Simulations and Practical Designs
Christophe P. Basso
Second Edition
ISBN: 978-0-07-182473-6
MHID: 0-07-182473-1
The material in this eBook also appears in the print version of this title: ISBN: 978-0-07-182346-3,
MHID: 0-07-182346-8.
All trademarks are trademarks of their respective owners. Rather than put a trademark symbol after every occurrence of a
trademarked name, we use names in an editorial fashion only, and to the benefit of the trademark owner, with no intention of
infringement of the trademark. Where such designations appear in this book, they have been printed with initial caps.
McGraw-Hill Education eBooks are available at special quantity discounts to use as premiums and sales promotions or for use in
corporate training programs. To contact a representative, please visit the Contact Us page at www.mhprofessional.com.
Information contained in this work has been obtained by McGraw-Hill Education from sources believed to be reliable. However,
neither McGraw-Hill Education nor its authors guarantee the accuracy or completeness of any information published herein, and
neither McGraw-Hill Education nor its authors shall be responsible for any errors, omissions, or damages arising out of use of this
information. This work is published with the understanding that McGraw-Hill Education and its authors are supplying information but
are not attempting to render engineering or other professional services. If such services are required, the assistance of an appropriate
professional should be sought.
TERMS OF USE
This is a copyrighted work and McGraw-Hill Education and its licensors reserve all rights in and to the work. Use of this work is
subject to these terms. Except as permitted under the Copyright Act of 1976 and the right to store and retrieve one copy of the work,
you may not decompile, disassemble, reverse engineer, reproduce, modify, create derivative works based upon, transmit, distribute,
disseminate, sell, publish or sublicense the work or any part of it without McGraw-Hill Education’s prior consent. You may use the
work for your own noncommercial and personal use; any other use of the work is strictly prohibited. Your right to use the work may
be terminated if you fail to comply with these terms.
THE WORK IS PROVIDED “AS IS.” McGRAW-HILL EDUCATION AND ITS LICENSORS MAKE NO GUARANTEES
OR WARRANTIES AS TO THE ACCURACY, ADEQUACY OR COMPLETENESS OF OR RESULTS TO BE OBTAINED
FROM USING THE WORK, INCLUDING ANY INFORMATION THAT CAN BE ACCESSED THROUGH THE WORK VIA
HYPERLINK OR OTHERWISE, AND EXPRESSLY DISCLAIM ANY WARRANTY, EXPRESS OR IMPLIED, INCLUD-
ING BUT NOT LIMITED TO IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR
PURPOSE. McGraw-Hill Education and its licensors do not warrant or guarantee that the functions contained in the work will meet your
requirements or that its operation will be uninterrupted or error free. Neither McGraw-Hill Education nor its licensors shall be liable
to you or anyone else for any inaccuracy, error or omission, regardless of cause, in the work or for any damages resulting therefrom.
McGraw-Hill Education has no responsibility for the content of any information accessed through the work. Under no circumstances
shall McGraw-Hill Education and/or its licensors be liable for any indirect, incidental, special, punitive, consequential or similar
damages that result from the use of or inability to use the work, even if any of them has been advised of the possibility of such
damages. This limitation of liability shall apply to any claim or cause whatsoever whether such claim or cause arises in contract, tort
or otherwise.
contents
Preface xiii
Acknowledgments xv
Nomenclature xvii
iii
iv CONTENTS
Chapter 6. Simulations and Practical Designs of Off-Line Converters—The Front End 527
Appendix 7D A Small-Signal Model of the Flyback Converter Operated in Quasi-Resonance / 794
7D.1 A BCM Flyback Converter / 796
7D.2 Application Example / 798
7D.3 The Ac Analysis / 799
7D.4 Numerical Application / 801
Reference / 803
Appendix 7E Switching Losses with a nonlinearly Varying Parasitic Capacitor / 803
Reference / 805
Appendix 7F Testing Transformer Core Saturation Level / 807
Reference / 808
Appendix 8D A Small-Signal Model for the Active Clamp Forward Converter
Operated in Voltage Mode Control / 929
8D.1 Revealing PWM Switches / 930
8D.2 Large-Signal Simulations / 933
8D.3 Small-Signal Modeling / 934
8D.4 The Magnetizing Current Resonant Circuit / 938
8D.5 Final Lap: Associating All the Blocks / 945
8D.6 Testing a Prototype Response in the Bench / 948
Reference / 953
Appendix 8E Web Content / 953
Conclusion 955
Index 957
ABOUT THE Author
Christophe P. Basso is currently an engineering director at ON Semiconductor
in France, where he has developed numerous popular switching power supply
controllers, for instance, for the notebook adapter business. He is the author of
several books on power electronics, including McGraw-Hill’s Switch-Mode
Power Supply SPICE Cookbook, and, recently, he released a title 100% dedi-
cated to loop control, Designing Control Loops for Linear and Switching Power
Supplies. He regularly teaches professional seminars at IEEE-sponsored Applied
Power Electronics Conferences and often publishes articles in trade magazines
such as PET and the online newsletter How2Power. Mr. Basso graduated from
the Montpellier University in 1985, and he received his M.S.E.E. in power elec-
tronics from the National Polytechnic Institute of Toulouse in 2008. He holds 29
patents in the field of power electronics and he is an IEEE Senior Member.
Preface
I am glad to introduce the second edition of my 2008 book Switch-Mode Power Supplies: SPICE
Simulations and Practical Designs. I would like to thank all the readers who have contributed to make the
first edition a success. I received numerous warm and supportive messages from all around the world and
it is extremely rewarding. Without you, this new book would not exist. Some of these readers have been
kind enough to report typos and errors they found in the first edition. I compiled them throughout the years
and I used the list to clean equations and figures.
Revising a book is not an easy task, as some readers will object that there is too little renewed content
to make it a new book while others complain that this new edition represents a completely different docu-
ment than the first one they bought! Needless to say, trying to please both parties is a perilous exercise.
Loyal to my original approach, I added topics in which I detailed the mathematical treatment so that you
can follow and learn from the book. In Chap. 1, it is the case for rms constraints concerning the basic
switching cells. In most of the available books, authors give formulas without founding equations and often
limit their analysis to one conduction mode. Here, both conduction modes are explored and detailed, with
clear summary tables at the end. Numerous Mathcad® sheets are provided online at www.mhprofessional
.com/Basso to let you evaluate your own configurations. Small-signal-wise, Chap. 2 has been expanded
with the PWM switch at work in a discontinuous conduction mode boost converter and the derivation of a
feedforward compensator gain. Chapter 3 now includes OTA-based compensators and offers a transistor-
level TL431 model. Chapter 4 includes several revisions on blocks such as the D-flip-flop and the leading
edge blanking timer. Chapter 6 now includes a complete small-signal analysis of the borderline-operated
boost PFC circuit operated in voltage or current mode. Chapter 7 covers in detail all over power phenomena
in fixed-frequency discontinuous or continuous flyback converters, without forgetting quasi-resonance. A
small-signal model of a QR flyback converter is presented in one of the appendices. Finally, Chap. 8
includes a new small-signal model of the active clamp forward converter operated in voltage-mode control.
I hope you will enjoy reading this second edition, in particular the newly added materials. Despite all my
efforts, some typos or mistakes may have escaped my attention and I would be grateful if you would send
your corrections/remarks to [email protected]. As usual, I will keep a record of these findings and compile
them in my webpage http://cbasso.pagesperso-orange.fr/Spice.htm for the benefit of the reading community.
I thank you in advance and wish you the best of luck for your designs!
Christophe P. Basso
xiii
This page intentionally left blank
Acknowledgments
My warmest thanks and love go first of all to my dear family: Anne, my wife, and my two beloved children,
Lucile and Paul. Revising an entire 900-page book cannot be done overnight and I am grateful I could spend
endless hours correcting and writing new paragraphs without affecting family life. Now that it is done, I
will enjoy hiking, cycling, reading, snowshoeing, and spending leisure time with you all again!
The book revision could not have been envisaged without the help and involvement of many people. I
wish to thank Joël Turchi, my friend and colleague at work, who is always available to discuss technical
subjects for hours and review my work. These discussions also took place with the application team with
whom I am lucky to work: Thierry Sutto, Stéphanie Cannenterre, Yann Vaquette, and Dr. José Capilla. They
kindly reviewed this second edition’s materials. Special thanks go to Alain Laprade of ON Semiconductor
in East Greenwich who kindly reviewed several chapters.
I wish to also express my gratitude to my beloved parents, Michele and Paul Basso, who bought me my
first power supply when I was 14 and let me develop my passion for electronic circuits, at the expense of
numerous breaker trips. As we have returned to my youth, “merci” to teachers such as René Vinci and
Bernard Métral from the “Clos-Banet Lycée,” who instilled their passion and knowledge into the restless
student that I was. At the same time, I published my first article in Radio-Plans (1982), thanks to my friends
Claude Ducros and Christian Duchemin, last editors-in-chief of the now-defunct magazine. Finally, Claude
Duchemin from the Montpellier University added the finishing touches and plugged my fingers into the
switching power supply world!
Both the first and second editions of this book incorporate comments and recommendations from pres-
tigious people I have been honored to work with. Their names follow and I wish to thank them warmly for
the amount of time they spent reviewing the first edition’s materials and tracking inaccuracies: Dr. Vatché
Vorpérian (Jet Propulsion Laboratory), Dr. Richard Redl (Elfi), Ed Bloom (e/j BLOOM associates Inc.), Dr.
Raymond Ridley (Ridley Engineering), Dr. Ivo Barbi (Power Electronics Institute of the Federal University
of Santa Catarina), Jeff Hall (ON Semiconductor), Dhaval Dalal (Acptek), and not forgetting Monsieur
Mullett (formerly with ON Semiconductor) for the two appendices kindly contributed on magnetic designs!
Also, Christian Zardini (retired from the ENSEIRB school), Dr. Franki Poon and Dr. S. C. Tan (PowerELab
and the Hong Kong Polytechnic University), Dr. Dylan Lu (Sydney University), Arnaud Obin (formerly
with Lord Engineering), Dr. V. Ramanarayanan (Electrical Engineering Department of the Indian Institute
of Science in Bangalore), Dr. Jean-Paul Ferrieux (Laboratoire d’Electrotechnique de Grenoble), Steve
Sandler (AEi Systems), Dr. Didier Balocco (formerly with Saft Power Systems), and Pierre Aloisi (for-
merly with MOTOROLA).
I would also like to thank the people at Intusoft, Larry and Lise Meares and all their great support team
(George, Farhad, Everett, Tim), who helped me during the testing phase of the numerous book examples. I
want to thank the editors of simulation software who have kindly contributed simulation examples.
Finally, thank you to Mike McCabe, at McGraw-Hill, for giving me the opportunity to publish a new
edition of my original book.
xv
This page intentionally left blank
NOMENCLATURE
User friendliness is a key factor for the commercial success of any simulation program. The growing com-
plexity of integrated circuits and equipment makes this aspect increasingly important. Despite numerous
publications devoted to the Simulation Program with Integrated Circuit Emphasis (SPICE), it still scares
the novice when its name is mentioned.
Developed in the mid-1970s at the University of California, Berkeley, the SPICE program’s primary aim
was to fulfill the needs of the electronics industry––mainly integrated circuit makers. However, with the sup-
port and funds from private editors, the SPICE program has evolved over a number of years into many practi-
cal and affordable packages, with emphasis on providing both low-priced and friendly access to beginners.
SPICE can significantly help you speed up the design phase of the equipment you are currently working
on, even if SPICE is not able to generate an electronic schematic by itself. SPICE is inherently efficient
because if you start working with an unfamiliar concept, it will quickly enable you to grasp the full mean-
ing of any particular architecture by unveiling its peculiar waveforms. You can thus use the simulator to
gain insight into the circuit you have to build and also ensure all parameters are taken into account before
the breadboard phase.
This book is intended for power supply designers, experts in their fields, as well as for beginners who
would like to understand the secrets of switch-mode power conversion. Manipulating virtual components
on a computer screen, without the hazards of high voltage, offers an interesting and safe way to learn the
technique. Furthermore, the “experience” gained in simulation, and it is also true for experts simulating a
novel concept, will let you feel more comfortable when breadboarding on the bench.
How many times have you heard this question when asking for a simulation package or a new computer?
The following statements do not represent an exhaustive list of pros about computer simulation, but they
can certainly be considered a “help list” available during the negotiations:
1. Here is an argument: Simulation can avoid waste of time and money. With its inherent iterative power,
SPICE covers numerous application cases in which you could easily detect any design flaw or product
weakness. The stability of a closed-loop SMPS represents a typical application when some key feedback
elements are moving (i.e., the variable load that affects a pole) or start to degrade with temperature
and aging (as the electrolytic equivalent series resistor). Moreover, design ideas can also be tested or
assessed in a snapshot through a computer and, if they are worth trying, further refined in the lab.
2. You can start to work on a project by downloading components models and becoming familiar with the
key elements, before going to the bench or waiting for the samples to be delivered. Once these samples
arrive, you will have already gained insight by prototyping with the simulator and the practical debug
phase on the bench will clearly take benefit from your first computer-based experiments.
3. Simulate test measurements whenever you do not own the adequate equipment: Bandwidth measure-
ments represent a good example. If you cannot afford a network analyzer, then a proven small-signal
model can help you start to refine your feedback loop. When run on the final prototype, stability assess-
ments will be faster and more efficient.
4. Power libraries are safe: They let you experiment “what if” when amperes and kilovolts are flowing
in the circuit without blowing up in the event of a wrong connection! Also, they let you see how your
1
2 chapter one
design reacts to a short-circuit of the optocoupler or the opening of a resistor. SPICE can begin to give
you the answer.
This book thoroughly details the advantages of SPICE to let you understand, simulate, test, and finally
improve the switch-mode power supply (SMPS) you want to design. By providing you with specific
simulation recipes, this work intends to facilitate as much as possible your SMPS design. Unlike other
books, the author strives to balance the theoretical content, necessary to understand and question simula-
tion results, with practical design examples. This is developed throughout the eight chapters of the book.
Chapter 1 explains switch-mode power supply techniques and types of converters, and it introduces
a few important results to help you better understand averaging techniques. This second edition includes
the comprehensive derivation of root mean square (rms) current constraints of the basic switching cells,
buck, boost, and buck-boost operated in continuous or discontinuous conduction modes. As usual, I
have detailed derivation steps so that you can follow and learn the technique in case you encounter a
different switching cell. A new appendix has been added. Chapter 2 explains how average models were
derived, and different types are described. A good comprehension of this chapter is fundamental: It will
help you question certain weird SPICE data resulting from a bad model implementation. If you do not
understand the way the model has been derived, you will obviously face some difficulties in resolving
these issues. In Chap. 2, you will also learn how to wire an average model and run basic simulations.
This second edition adds the description of the feedforward modulator. I added more details to the small-
signal PWM switch. In particular, I show how keeping the same voltage-mode average model to which
you add a “duty ratio factory” turns the model into a simplified version of current mode. An appendix
has been added focusing on the DCM (discontinuous conduction mode) voltage-mode boost small-signal
transfer function. Closing the loop is obviously an important aspect of converter design that is often
overlooked. This is not the case here, and Chap. 3 will guide you through control loop design, again
using practical examples with a TL431 and not op amps only, as often seen in the literature. OTA (opera-
tional transconductance amplifiers) compensators are now part of this chapter, covering types 1 and 2.
A complete transistor-level model of the TL431 has been added to its dedicated appendix. Because not
every integrated circuit always comes with a SPICE model, Chap. 4 describes how the generic switched
models are derived. This chapter will interest those who want to strengthen their knowledge of SPICE
model writing. A more robust D-flip-flop is described and several new circuits have been added. Chapter
5 describes practical designs of the three basic nonisolated topologies, including the front-end filter.
Before analyzing off-line converters, Chap. 6 shows how to design the rectifying section and explains
the various power factor correction techniques. Small-signal response of the popular borderline-operated
power factor correction (PFC) has been added in a dedicated appendix. Chapter 7 is entirely dedicated to
the flyback converter, with specific design examples at the end. A new section on over power has been
added and it covers all operating modes, including QR converters. An appendix gives the small-signal
response of the QR converter. Nonlinear capacitor switching losses are explored in a dedicated appendix.
Finally, the forward converter appears in Chap. 8, again associated with a design example. The coupled-
inductors section has been updated and a complete small-signal model of the active-clamp converter
operated in voltage-mode is presented with experimental results.
Version syntax is a significant issue with SPICE. Most SPICE editors deal with a proprietary syntax,
sometimes SPICE3 conformant, that makes translation from one platform to another a difficult and painful
exercise. To allow the use of different simulators, the standard models presented throughout the pages are
compatible with Intusoft IsSpice (San Pedro, Calif.) and CADENCE’s PSpice (Irvine, Calif.).
To help you quickly copy and paste examples, simulation files have been made available for download
from a McGraw-Hill website. Please check App. 8E for details. Some selected simulation examples are
offered in IsSpice and PSpice syntax, and you can easily load them on your computer if you have one of these
software programs. For students or newcomers to the SPICE world, some demonstration versions will let
you open files and simulate some of them (those demos are size-limited) to give you a taste of what the full
version can do. The McGraw-Hill download page contains PowerPoint® and Mathcad® files to let you key in
your own design parameters and check small-signal response or rms constraints of the basic switching cells.
Introduction to Power Conversion 3
For professional power supply designers, another library file is separately distributed. This file contains
the design examples presented in the book plus numerous other industrial applications using real control-
lers. Please visit the author’s website for distribution details (http://cbasso.pagesperso-orange.fr/Spice.htm).
This book does not describe the way SPICE operates, nor does it solve typical electric circuits. It assumes
that the reader is already familiar with the basics of SPICE simulations. Numerous books and papers are
available on the subject as the References section details [1, 2]. Whenever possible, the extended bibliog-
raphy will guide your choice if you wish to strengthen your knowledge in a particular domain, such as
some topologies that you are unfamiliar with. If some theoretical results are sometimes delivered just “as
is,” we strongly encourage the reader to dig further into the appropriate literature and acquire the theory
that precedes the result.
The book focuses only on a system approach. No SPICE description of typical discrete power elements
such as diodes, MOSFETs, etc., is proposed.
Finally, here is the important statement, probably the most interesting one! SPICE does not replace the
breadboard phase, nor does it shield you from writing equations or understanding electronics. It looks like
a simple sentence, but the author has often been confronted by designers showing boards in the trash and
claiming, “But SPICE said it would work!!” Yes, all ideas work on paper until they face the soldering iron
condemnation.… Use SPICE as a design companion, a circuit insider that can reveal waveforms difficult
to observe. But always question the delivered data: Is this the real behavior, have I been misled somewhere,
does a simple calculation more or less confirm what I see?
After this brief introduction, it is time to plunge into the intricacy of SMPS design and simulation with
SPICE.
In the electronics world, different types of circuitries must cohabit: logic devices, analog circuits, micro-
processors, and so on. Unfortunately for the designer, these circuits do not cope with a single, fixed, power
supply rail: A microprocessor or a digital signal processor (DSP) will need a stable 3.3-V source or less,
a front-end acquisition board will require ±15 V and perhaps some logic glue around a standard 5 V. For
the final board being supplied from a single power point, for example, the mains outlet or a battery, how is
one to adapt and distribute all these different voltages to the appropriate portions? The solution consists of
inserting a so-called converter to adapt the voltage distribution to the circuit needs.
3.30
1.4.1 Associating Resistors R2
132 Rload2 3V3
Figure 1.1 portrays the simplest option a designer can think of: 50 DSP
resistive dividers. If our DSP consumes 66 mA over 3.3 V, then it
can be replaced by a 50-Ω resistor, the same as for our 50 mA, 5-V R1
logic circuit via the 100-Ω resistor. From a 12-V source, we can 12.0
140
then calculate the dropping resistors: 5.00
+
Vin 5V
Rload1
12V logic
12 − 5 100
R1 = = 140 Ω (1.1)
50 m
Before going further, please note that 0.066 A or 0.05 A was, respectively, noted in the calculation as
66m or 50m. This is done to retain the SPICE notation for units, without any space. It adheres to the fol-
lowing rules and will be extensively used in the remaining portions of the book:
p = pico = 10-12
n = nano = 10-9
m = micro = 10-6
m = milli = 10-3
k = kilo = 103
Meg = mega = 106
• Beware not to mix mega and milli, a very common mistake: 10 mΩ = 10m, 1 MΩ = 1Meg.
Unfortunately, these resistors will be the seat of a permanent voltage drop, and power dissipation (in
heat) will occur. The dissipated power for each resistor is
(12 − 5)2
P1 = = 350 mW (1.3)
140
(12 − 3.3)2
P2 = = 573.4 mW (1.4)
132
From these values, we can now evaluate the system efficiency obtained by dividing the delivered output
power Pout by the power taken away from the source Pin:
52 3.32
Pout = + = 250 m + 218m = 468 mW (1.5)
100 50
122 122
Pin = + = 600 m + 791m = 1.39 W (1.6)
100 + 140 50 + 132
The efficiency, represented by the Greek letter h or “eta,” can be computed by dividing Pout by Pin, or
Pout 0.468
η= = × 100 = 33.6% (1.7)
Pin 1.39
Pout 1
Ploss = Pin − Pout = − Pout = Pout − 1 (1.8)
η η
If the load changes, or if the input voltage drifts, what is going to happen? Well, since our input-to-output
transfer ratio, denoted M, is fixed, the output voltage will also vary. Therefore, we need to think of a kind
of regulated system that permanently observes the output power demand and adjusts the series resistor
Introduction to Power Conversion 5
to maintain a constant output voltage, if the output voltage represents the variable of interest. For a well-
designed system, the converter must also ensure a proper regulation independent of input voltage variations.
To reach this goal, we need to use several specific components such as
• A reference voltage Vref : This voltage is by definition extremely stable in temperature and precise in
value (e.g., ±1%). A programmable shunt regulator, such as a TL431 adjustable Zener diode, could do
the job.
• An operational amplifier (op amp): This device will observe a portion of the output voltage (aVout) and
compare it to the reference Vref . It will actually “amplify” the error, the difference between aVout and
Vref , to drive a series-pass element. The error monitored by the op amp is usually denoted by the Greek
letter e or “epsilon”: ε = α Vout − Vref .
• A series-pass element: It can be a MOSFET or a bipolar transistor but working in a linear mode, playing
the role of the necessary variable resistor. If it is a MOSFET, the static driving power is null. For a bipo-
lar, there is a need to supply a sufficient amount of base current to deliver the right collector or emitter
current. This is called the bias current.
Figure 1.2 finally shows how our resistive converter could be improved, let’s say for the 5-V section.
The error amplifier is made via a voltage-controlled voltage source (E primitive) and features a gain of
10k or (80 dB), 20 log10 (10 k ). One input receives the voltage reference whereas the other one, the inverting
input, is biased by a portion of the output voltage. This is actually a linear regulator, however, limited in
the input voltage range since Vin shall be above Vout by a Vbe, at least, to guarantee a proper drive for Q1. If
Vout is below the target (5 V in our example), E1 output increases and strengthens Q1 bias current: Vout goes
up. On the other hand, suppose the load has suddenly been reduced, therefore Vout exceeds 5 V. Thanks to
E1, Q1 bias current goes down, reducing the output voltage until regulation is met again.
Q1
2N2222 Vout
Ib Iout
Iin E1 Rupper
10k 10k
+
Vin – Rload
12 + 100
–
+
+V Rlower
ref 10k
2.5
The output voltage observation, which delivers aVout (a fraction of the output voltage), is obtained
through a resistive divider made of Rupper and Rlower. Calculating their values is straightforward:
1. Let us fix a current circulating in the divider bridge. Since there is no biasing current for E1 in the
example (this is the case for most MOS-based technologies), we could take Ib = 250 μA, for example.
A lower value is acceptable, but degrades the noise immunity in a noisy environment.
2. Ib equals 250 μA and entirely circulates in Rlower. Thanks to the control loop, 2.5 V is “seen”across
2.5
Rlower . Therefore, Rlower = = 10 kΩ.
250u
Vout − Vref 5 − 2.5
3. The voltage drop across Rupper is Vout - Vref. Thus, Rupper = = = 10 kΩ.
Ib 250u
6 chapter one
90.0
80.0
70.0
60.0
Efficiency (%)
50.0
40.0
30.0
20.0
10.0
0.0
5 7 9 11 13 15 17
Input voltage (V)
Figure 1.3 As soon as M diminishes, the efficiency dramatically drops (Vout = 5 V).
If we neglect the power needed to drive Q1, then all the source current Iin flows into the load as Iout.
Therefore, applying Eq. (1.7), we can derive the efficiency for this linear regulator:
If we now plot the efficiency versus the input voltage, we can see how difficult the situation becomes
in the presence of small M ratios (Fig. 1.3). For these reasons, resistive divider type of converters, that is to
say series-pass regulators, are limited to applications where M does not fall below 0.3. Otherwise the heat
dissipation burden becomes a real handicap. On the other hand, when the user really needs to operate a
regulator to ratios M closer to 1 (Vin very close to Vout), the low-dropout (LDO) regulator made with a PNP
becomes a good choice. The input low limit is now linked to the transistor Vce(sat) (a few hundred millivolts,
or less) rather than its Vbe (around 650 mV at room temperature, 25°C).
To close the study on regulator efficiency, we can take three different output examples with linear regu-
lators, where output and input conditions vary:
5
1. Vin = 14 V Vout = 5 V ∆V = Vin - Vout = 9 V η = 100 = 35.7%
14
12
2. Vin = 14 V Vout = 12 V ∆V = Vin - Vout = 2 V η = 100 = 85.7%
14
3
3. Vin = 5 V Vout = 3 V ∆V = Vin - Vout = 2 V η = 100 = 60%
5
As a result, one can see that a high efficiency can be obtained with a linear regulator if ∆V is small (as
plotted in Fig. 1.3), but also if Vout >> ∆V.
Figure 1.2 is interesting because we can use it to derive general statements, pertinent to the closed-loop
world we are going to enter, linear, or switched. Suppose that we remove the error amplifier and replace it
by a fixed voltage source of 5.77 V, our actual op amp output (look at Fig. 1.2 values), as Fig. 1.4 shows.
The regulator becomes a simple emitter-follower circuit, affected by an output impedance and an output
voltage. As such, it can be described with its equivalent Thévenin generator, what Fig. 1.5 suggests. Rs,OL
represents the open-loop output impedance and Vth, the voltage delivered when biased by a control volt-
age Vc, our fixed 5.77 V in the application. Let us now use this representation and redraw our closed-loop
regulator around it, ignoring, for now, the input voltage contribution.
Introduction to Power Conversion 7
Q1
2N2222 Vout 5.00
12.0
+V
in Rload
12 100 +
+V Vth = f (Vc)
bias
5.77 Vth
In Fig. 1.6,Vout(s) is compared to Vref (s) via a resistive divider affected by a transfer ratio of a. H(0) illus-
trates the static or dc relationship between the output voltage and the control voltage Vc, e.g., Vc = 5.77 V to
obtain Vout = 5 V in this example. The theoretical dc voltage (s = 0, but we purposely avoided this subscript
below for the sake of clarity) you would expect from such a configuration is
Vref
Vout = (1.10)
α
Unfortunately, the whole gain chain and various impedances will affect this value. With a few lines of
algebra, we can write the static output voltage definition (again, s = 0) simply by following the meshes:
Vout
Vout = (Vref − α Vout ) HG − Rs ,OL (1.11)
Rload
Vref HG
Vout = (1.12)
R
1 + α HG + s ,OL
Rload
Vout (s)
H(s)
Rs,OL
Vc(s)
Rupper
a Rload
–
G(s) +
Vref (s)
Rlower
dc bias
0
The static error on the output, actually the deviation between what we really want and what we finally
obtain, is derived by subtracting the Vout expression Eq. (1.12) from Eq. (1.10):
Vref Vref HG
1 1
Verror = − = Vref − (1.13)
α Rs ,OL α 1 Rs ,OL 1
1 + α HG + +α +
Rload HG Rload HG
1 1
Verror = Vref − (1.14)
α α+ 1
HG
which equals zero if
1
α =α + (1.15)
HG
From this equation, we can see that increasing the dc gain, G(0), helps diminish the static error which
finally affects our output voltage precision.
Another important parameter influenced by the loop gain is the closed-loop output impedance. The
output impedance of a system can be derived in different manners. As Fig. 1.5 has shown, our closed-loop
generator can now be reduced to its Thévenin equivalent, that is, a voltage source Vth [Vout measured with-
out any load, or Rload = ∞ in Eq. (1.12)], followed by an output impedance Rs,CL, which we actually look
for. One option consists of calculating a resistor RLX which, once wired between the output and ground, will
V
reduce Vout = Vth to Vout = th . When this occurs, RLX simply equals Rs,CL (we have built a simple resistive
2
divider with equal resistors). We can quickly manipulate Eq. (1.12), assuming Rload = ∞ :
Vth
= Vout ( RLX ) or “What value of RLX will divide the Thévenin voltage by 2?”
2
Vref HG Vref HG
= (1.16)
2(1 + α HG ) R
1 + α HG + s ,OL
RLX
If we call α HG the static loop gain T, then the closed-loop output impedance is
Rs ,OL R
RLX = Rs ,CL = = s ,OL (1.17)
1 + α HG 1 + T
Equation (1.17) teaches us different things:
Why do we talk about a static (dc) and a frequency-dependent gain? Well, this is so simply because Figs. 1.5
and 1.6 do not represent genuine regulators. In reality, G(s) is made via a real operational amplifier, impos-
ing a virtual ground on its inverting pin as soon as local feedback exists. In other words, Rlower simply goes
off the picture in the small-signal model, and a no longer plays a role. This is described in App. 3D.
Introduction to Power Conversion 9
In this example, we purposely did not account for an input voltage perturbation. This assumption is valid
for bipolar transistors as the weak influence of the Early effect makes them good current generators, almost
independent from their Vce variations. However, when Vout and Vin are close to each other, the transistor
becomes a closed switch rather than a current source. Therefore, the input voltage starts to play a role. Let us
redraw the Fig. 1.6 sketch, including the input voltage contribution. As drawn in Fig. 1.7, the term k represents
the open-loop audio susceptibility, denoted As,OL. It represents the input voltage contribution to the output.
+ Vout (s)
H(s)
+ Rs,OL
k
Vc(s) Rupper
Vin
a Rload
–
G(s) +
Vref (s)
dc bias Rlower
0
R
Vout 1 + α HG + s ,OL = Vref HG + kVin (1.19)
Rload
Vref HG kVin
Vout = + (1.20)
Rs ,OL Rs ,OL
1 + α HG + 1 + α HG +
Rload Rload
Vref kVin
Vout = + (1.21)
1 + α HG )
+ α
(1
HG
1. The theoretical output voltage, similar to what was defined in Eq. (1.12), simplified, gave (Rload = ∞)
k
2. The input voltage contribution whose new term is or, sticking to the previous definition,
(1 + α HG )
As ,OL A
As ,CL = = s ,OL (1.22)
1 + α HG 1 + T
10 chapter one
Again, operating with a large dc gain ensures an excellent rejection of the input voltage ripple (100
or 120 Hz for full-wave rectification). When T(s) reduces in the high-frequency domain, the system runs
open-loop. Please note that we purposely selected a positive polarity for k, but a negative value could also
have been chosen. It actually depends on the topology under study.
Thanks to SPICE, we can simulate a completely theoretical regulator by associating blocks. Fig. 1.8 depicts
the circuit where you will recognize the block discussed above. The operating parameters are the following:
Rs,OL = 1 Ω
As,OL = 50 m
Vin = 15 V
Vout = 5 V (target)
Vref = 2.5 V
a = 0.5
Please ignore, for now, the presence of the compensation network Rf - Cf . Applying our above input
numbers reveals these closed-loop values:
Equation (1.17), Rs,CL = 1.996 mΩ, or in dBΩ: 20 log10 ( Rs ,CL ) = 20 log10 (1.996m) = −54 dBΩ
Equation (1.20), Vout = 4.991318 V
Equation (1.22), As,CL = 99.8u, or in decibels: 20 log10 ( As ,CL ) = 20 log10 (99.8u) = −80 dB
X2
GAIN
Vin k = 0.05 X3
SUM 2
15.0 0.750 RSOL
GAIN K1
5.09 1 4.99
SUM 2 Vout
4.34 K2
Alpha = 0.5
Rupper
10k
+ + Rf Cf
–
10
2.50 for Zout
– sweep
0.434 +
– 2.50
+
+ Rlower
Beta V1
E1 2.5 10k
100
Figure 1.8 Our theoretical linear regulator, including the output impedance and the input voltage perturbation.
Introduction to Power Conversion 11
Now, let us compare to what the SPICE simulator will give. We have several options. The first one uses
dV (V )
a .TF statement, which performs a out in calculation as well as an output impedance measurement. The
dVin
SPICE code is the following: .TF V(vout)vin. Once it has run, we obtain the results in the .OUT file:
***** SMALL SIGNAL DC TRANSFER FUNCTION
As one can see, we are very close to our theoretical numbers. The input impedance value does not make
sense here since we do not absorb current from the source. Now, we can transient-step the input and observe
the output. Stepping the input means replacing the fixed Vin source by a pulsewise linear (PWL) statement. This
SPICE function builds the curve you wish by (time, amplitude) couples. It associates linear segments together.
Here, we start for t = 0 at Vin = 15 V and suddenly increase Vin when t = 10 μs with a 1-μs slope to 500 V:
Vin 7 0 PWL 0 15 10u 15 11u 500
Figure 1.9 reveals the output response when the input is suddenly increased from 15 up to 500 V. We
can observe a deviation of 48.4 mV engendered by a 500 - 15 = 485 V input variation. Therefore, the dc
audio susceptibility is 0.0484 / 485 = 99.8u, which matches the closed-loop calculation.
Vout = 5.03V
815 5.03
Vout = 48.4 mV
Vin = 500V
415 5.01
vout in volts
vin in volts
Vout = 4.99V
Plot1
0 4.99
Vin = 15V
–385 4.97
–785 4.95
By stepping the circuit output with a current source, we will be able to extract the static output imped-
ance. To do so, we remove Rload and connect a current source affected by a PWL statement:
ILoad vout 0 PWL 0 0.1 10u 0.1 11u 1end
∆Vout
Here, the output will be pulsed from 100 mA to 1 A in 1 μs. The output impedance will simply be .
∆Vin
Figure 1.10 displays the simulation results. For the 900-mA step, we can see a deviation of 1.79641 mV. The
dc output resistance is therefore 1.79641m / 900 m = 1.996 mΩ, again exactly what our calculations predicted.
Now, we reconnect the compensation network made of Rf and Cf. This network will make T(s), the total
loop gain, depend on the sweep frequency. By connecting an ac source of 1 A in place of our static load
resistor, we can sweep the ac output resistance/impedance, called Zout, of our converter.
• The current source is connected with its positive terminal to ground and negative terminal to the output.
12 chapter one
4.9908
vout in volts
Plot1
4.9904
DVout = 1.79641 mV
4.9900 DIin = 900 mA
Vout = 4.98952V
4.9896
Then, using the graphical interface, plotting Vout alone will reveal Zout since Iout = 1 A, as Fig. 1.11
portrays.
On these graphics, we can observe three areas:
1. The dc region, f < 1 Hz: The loop gain T(0) is extremely high, Cf can be considered as open. Therefore,
Zout is defined by Eq. (1.17) and is extremely small.
2. Above 1 Hz, Cf starts to play a role: The loop gain T(s) starts to diminish and the denominator of
Eq. (1.17) goes down. As a result, Zout increases. An impedance growing with frequency reproduces an
inductive behavior. If we take a point, let us select f = 100 Hz, we can calculate the equivalent induc-
63.24 m
tance Leq. From Fig. 1.11, the equivalent inductance Leq is equal to = 100.6 µ H at f = 100 Hz.
2π 100
3. Cf becomes a complete short, and the loop gain is fixed by Rupper and Rf. Now Zout is close to its open-
loop value, which is 0 dBΩ (1 Ω).
If for filtering reasons we connect a 100-μF capacitor on our regulator output, we create an LC filter that
1
is going to resonate. The resonating frequency is defined by , which is approximately 1.586 kHz.
2π LeqC
–10.0
Zout
Plot1
–30.0 10 dBΩ/div