George C. Chryssis High Frequency Switching Powe WWW - Iranswitching.ir

McGraw-Hit1 Reference Books of Interest Gf ORGE CHRYSSIS /
:
Handbooks
Availone and Baumeisrer * STANDARD I~ANDBOOK FOR M E C H ~ I C hmms
AL
Bemon AUDIOENGINEERING HANDBOOK POWER SUPPLIES:
Benson TELEVISII)EI ENGINEERING ~NDBOOK
Cwrnbs PRMW C I R C UHANDBOOK ~ THEORY AND DESlGN
Coombs BAsrc E E ~ O N I ~CS T K U M E I V THANDBOOK
Di G~acom0 V u l HANDBOOK
Fink and Baav STANDARD HANDBOOK FOR ELECTRICAL ENGINEERS
Fink ~ n Chrisriunsen
d ELEC~RON~C~ ENGINEERS' I~ANDBOOK
FIa~grrn H ~ D B WOF K TRANSFORMER ~LICATION~ SECOND EDITION
Harper HANDBWK
9 OF ELECTRONIC SYSTEMS DES~GN
Harper H A N D B ~OFKTHICK Hve~roM r m o m o ~ ~ c s
Inglis 0 E L E ~ O N COMMUN~CATIONS
~C HA~BUQK
Johnson and Jasik ANTENNA ENGLNEERINF HANDBOOK
juran QUAWTYCONTROL HANDBOOK
Kaufinan and Scidman HANDBOOK OP ELECIRONICS CRLCIJLATIONS
Kaufman and Seidman H A N D BFOR ~ E ~ ~ O N IENGINEERING
CS TECHNICIAVS
' Stoul * HANDBODK OF MICROPROCWSOR DWIGNAND APPLICATIONS
Stout and Kaufman HANDBOOK OF MICROCIRCU~~ DESIGNAND APPLICATION
S!oyt !rand K ~ ~ f m 0 nHAND~WK OF OPERATIONAL AMPLIFIER CERCVIT DESIGN
WiItiams DESIGNER'S HANDBOOK OF INTEGRATEDC I R C U ~ S
Wifliams ~ n Toyfor
d * E ~ m o n r FILTER
c DESIGNHANDBOOK
DICTIONARY OF S c r w m c AND TECHNICAL TERMS

DICTIONARY OF EWEFTR~CAL AND E W ~ R O N I CENGINEERING
D T ~ I O N AOF
R ENGWEERING
Y
Other
E RELEC~RONIC
Grossner T R A N S ~ R M IN S Ct~curn
Mitchell * wioc S W ~ I NREGULATOR
G ANALYSIS
Rombaut POWERELECTRONIC CONVERTERS:
AC/ AC CONVERSION
Seguier POWERELECTRONIC C O N V E R ~A C: ~ D CCONVERSION
For more fformation about otkpr McGmw-Hill materials, MUGRAW-HU MIBUSHINGCOYPANT NEW YORK ST tOUlS SAN F R A W S 0 0
call I4W.2-MCGRAW in the United States. In other & W A N D BOGOTA HAMBURG LONDON MADRID MEXW M W MOMREAL NEW DEWl
countries, call yuwr nearest McGraw-Hill &ce. PAUL0 SINGAPORE SYDNEY WKYO TO R OW
www.IranSwitching.ir
-- - - -----
- -- . --
HIGH-FREQUENCY SWlTCWMG POWER SUPPLIES
TO MY WIFE MARGO AND
IWIZRNATIONAl EDITION TO RANIA, LIUAN,
Copyright 0 1989 AND ALEXANDER
This b k cannot be n-rr-
-
Exclusive rights by McGraw-HI11 Book 0. Singapre for manufactme and -on.
from the country m which ft is consigned by
McGraw-Hill.
Copyright Q 1989, 1984 by McGraw-Hi11. All rights m e d . Excev as permitted
d c r the United Stata Copyright A a of 1976, no pan ofthis publimion may be
reproduced or distributed in any form or by any mans, or stored in a data base or
rnricval system, without the prior written permission of the publisher.
T3w dim$jar this bwk awe Ddnlcl A. Gomrenu and h m 'Nugem, the w
the Rimside Gmphic BudiP, Im, and drc pmdsluion mpwisrn wus S w n ~ eW.
BabPuf. It rwrs m C a a n i a by T e c h Typ, Inc.
Library of C o w u s Cataloging-in-PublimtionData
Chryssip, George
Xih-frequency swivAhg power supplies.
, Includcs indcx
I
-
1. Elmronic apparams and appliancts Power supply.
2. M i c ; o c l m d c s -- Power supply. I. Title
TK7868.P6C47 1989 621.381'044 88-13681
ISBN 947-010951-6
Information mmained in rbis work has been obtained by W r a p

Hill, Inc, from g ~ r c e sklievDd to be reliabla KO-, neither
McGraw-XU nar its authors guaranttcs the accuracy or
complaenas of any information published herein, and neither
McGraw-Hillnor its authors shall bt responsible for any errors,
omissions, or damages arising om of use of this i n f m t i o n . Tbis
work is published with the understanding t h McGraw+HUland
IS authors art supplying information b u ~arc not anemping to
render engineering or oher professional services. If such servlces
arc required, the asslaam of an appropriate professional shwld be
sought.
When ordering this title me ISBB 0.03-1007 13-X
hinted in S i p w e
Preface dff
ONE
WE SWITCHING POWER SUPPLY: AN OVIRVEW i
1-0 Intmhcflon, 1
1-1 The Linear Power Supply, 2
1-2 The Off-the-Une Swltchlng Regulated Power Supply, 3
1-2-1 The Com~leteOf-the-Line Power Supply Bullding Blocks, 4
2-0 The Voltage Dwbl# Tmnlque, 7

2-1 Component Selectlm d Design Criteria. 7
2- 1- 1 lnput Rectifiers, 7
2-1-2 lnput Filter Capacitors, 8
2-2 Input Pmfechve Devices, 9
2-2-1 Iruush Cumnt, 9
2-2-2 lnput Transient Voltoge Protection, I1
MREE
OF POWER CONVERTERS 13
34 DeflnRlons Dlrnenslmlng, 13
3-1 The IsolaM flyback Converter, 14
3-1-1 The Flytmck Cmverter Swifchlw Trmlstor. 16
. 3-1-2 he ~~yback Converter ~rambrmer~hclre, 18
3-13 Variaitms of the mi Fwd Converter, 19
3-2 The lsotcrted Fwward Ccnverfer, 19
3-2-1 The Forward Converter Switching Transistor. 21
3-2-2 The Fwwmd Converter Transform#, 23
3-2-3 Varldlons of the Basic Forword bwerter, 24
3-3 The Push-Pull Converter, 25 4-9 fhe Power MOSFET Used as a Swltch. 93
3-3- 1 The Push-Pull Conveer Tramformer, 27 4-9-1 ImUctIon, 90
- ..
$3-2 fhe Push-Pull Comrerter Tmslstors, 27 4-9-2 Baslc MOSFET Definitiom, 93
3-24 Llmitatiom of the Push-Pull Clrcult, 28 4-9-3 Gate Drlve CcnsEderdrions of W MOSFET,
W if briatlons d the Push-Pull Corwerter, 30 4-9-4 Statjc CperuHng Chorocterlstlcs of tt!8 MOSFET, 93
34-t The Half-Wge Converter. 30 4 - 9 4 MOSFET W e Operating A r m [SOA), 95
34-2 The M e s Ccupliq Capacit~,31 4-94 Design Conslderdionsfor Drtvln~the Power MOSFET, 96
3 4 3 'The Commutalicg Dtcdw, 35 4-9-7 C~rcuitsUsed In Drlving me MOSFET, 98
3 3 The Full-Mdge Clrwit, 36 d-9-8 Power MOSFET SwrtCh P r o M m Cirwits. 102
3.6 A-New Zero Olrtput Ripple Converter, 37 440 The Gute Turn-of (GTO) Switch, 103
3-7 The BWlng Oscllldw or Riqing ChOke Converter, 4t 440-1 G u k Drive Requirements of t k GTO, 1 M
3-7- 1 The Bloclcln~Oscillcrtor Trondormer, 03
3-7-2 A MOSFET Blo&lng Oscillator Converter. 44
3-8 The Sheppord-TaylorCmverter, 45
34-1 Circult Anclvysrs ol the Sheppard-Tayb Cmveiter, 46 FIVE
3-8-2 Fwhrres of the Shwpard-Taylor Cmwrbr, 51 THE HIGH-FREQUENCY POWER TRANSFORMER 109
3-9 HigkFrequmcy Resonant Converters, 3 5-0 Introdudlon, 1W
3-9- 1 The Baslc She Wove Serles Resonant Ccnverter, 52 5-1 Prlnc~plesof E l ~ ~ e i l s r109
n ,
4
3-9-2 TranslstW Selectlm In Series Remont Converter, 53 5-2 The Hysteresis Imp, 111
5-94 Power Tmsfomr Design, 54 5-3 Baic Transformer Theory, 114
3-94 Design of the Serles Rwnont Network LCmM 5U Core Mafehal and G m e t r y Selection, I t 6
' 3-94 Design of the Remmt Incfuctor. 54 5 5 Design of a Power Tmsfomr k r a PulmWidth-Mmulatecl&If-Bddge
s t 0 C U M - W eRegulated Ccmwtws, 57 Convwter, I t 7
3-11 The Ward Gwwerter, 59 5 6 Practical Conslderabwrs, t21
3-1 t-1' a m i t Anolysrs of the Word Ccnverter. 61 5 7 The Flybock Cmvertets lmnsfwmer-ChokeDestgn. t22
3-t t-2 Design Pmedure fw the Ward Converter. 62 5-7- 1 Deslgn Procedure, t23
341-3 Feulures of the Ward Cwwerter, 63 5-8 Some General HlgbFrewency Transfom Considerations, 130
FOUR SIX
THE POWER TRANSISTOR IN COHWRTER WSiGN 65 THE OUTPUT SECTION: RECTIFIERS, INDUCTORS, AND CAPACITORS 133
IntroducSim, 65 k-0 ' Introduction, t33
Tmslstw Selection, 65 6-2 Output Rectificution and Filterlcg Schemes, 133
The Bipolar Power Transistor Used as a Switch 66 6.2 Power Rectifier Charoctehstlcs In Switchlw Power Supply Design, 135
Switch Tim= Mnlttons of Bipdar Transistors [ResistiveCoact), 67 6-2-1 Fast stand Very Fast Reccwery DWes, t35
Inductive loud Swstchlw Reldonships. 68 6.2-2 S c M y Banler Rectiim, 138
Transistor Antiscrhndim Cirwits. M 6.2-3 Tmdent OvewoRage Suppression, t 38
Barn Drive Cirwt Techniques for Bipdar Tmslstors, 71 6 2 4 Calculuting W Rectifier D i e Capdoillty for fhe Flyback, Fwwarcf,
4 4 1 Constant Drlve Current Cirwlts. 71 and Prrsh-Pull Converters, 141
44-2 A Propwhml b e Drlve Clrcutt, 75 6.3 Synt%cnw Rectjflers, 144
4-6-3 An Alterndve P w m a l Base Drtre Cirwit, 78 6.31 Geneiul Ccnidercrtlons. 144
4 6 4 Antl-on Clrwits Used n BasP h e s , 82 6-5-2 The Power MOSFET as a Synchrontxrs Rdlfler, 145
Bipolar Tronslstor Seoonctary Breakdown Cmsi&rotrms, 83 6-3-3 The Bipdar Smchroncus Rectmer, 147
4-7-1 Forword-Bios-cry Breakdown, 83 b3-4 Output VoRage Reguldlm Using Bipolar Synchronous Recilfiers and
4-7-2 RevemBias Swmdary Breakdown. 84 P W Re@utor Techniques, 149
Swftching Tramistor Prcteciive Nehvwks: RC S r u W r s . 86 6-3-5 A Current-Drlven S y n c h m s M f l e r , t50
4
6-rt CMput Power IrdJctor De3gn. 152 NINE
I 64-i General Conslderdons, 152 STABILITY IN SWITCHING POWER SUPPLIES: ANALYSIS AND OESlGN 223-
- - 64-2 Dedv~ngthe Deslgn Equcrtlw, 153
6-5 Magnetlc Amplifiers. 162 9-0 Introduction, 233
-r
I &51 Operohcm of the M a W l c AmplWier, 162 9-1 Thz Laplace Transform, 233
6-52 Design of the Magnetic Amplifier Saturable Reactor, 1bl 9-2 Transfer Functlons, 224
3 653 Contrd Clrcults for MagneticAmplifiers, 170 9-3 Bode Plots, 236
4-54 The UC1838-MngnetlcAmpfif~erController, 171 94 F e d b a a Theory and the Crttetla for Stablliiy. 237
1 4-6 Designlw the O W Fllier CaprrcRor, 178 9-5 CW-theline Switching Fwer Supply Stability Analysis. 242
9-51 Control-to-OvtputTransfer Functbm 242
9-52 Error Arnplif~erCompensation, 243
4 SEVEN 9-6 Stablliiy AnolysiS and Synthesis Using the KFactor, 254
SWIEHIHQ REGULATOR COMTROL ClRCUItS 9-6-1 The K W o r . 254
i
? 7-0 tntrccuctlon. 1&3 9 4 2 Mothemdcal Expressjm of the K Factor, 254
7-1 tsolatkm Technmues of Switching Regulcstar Systems, 183 W 3 Synthesis of Feedback ArmiHiers Utlllzlng the K Factor. 256
j 7-2 PWM Systems. 185 9-7 h p Stability Mewrements. 260
3-2-1 A'Siwl&nded, Discrete Cmponent. PWM Control Clrcuit, 185
i 7-2-2 An I n f e r n PWM Controller, 187
7 TEN
I / 7-3 Swne Commercially AvailaMe Mondlthlc PWM Control Circuits and ELECTEOMAQHnlC AND RADIO FREQUENCY IWERRREHCE [EMI-RFI)
1 t Their APplicat~ms,187
> 7-3-1 The T1494 PWM Control Circuit, 189
COHSIDEPATIONS 265
10-0 Introduction, 265
I ' 7-32 The UC1840 ProgmmmaMe, Off-Llne,PWM Controller, 192
10-1 The FCC and VDE Conducted Noise Specificatlm, 265
1 7-3-3 The UC1524A PWM Contrcl InlegruM ClrcuR, 197
10-2 RFI Swrces In Switching Power Supplies, 267
7-3-4 The UCiW Current-Mde Control tntegrted Clrwlt, 200
I 10-3 AC lrpvt Une Fllters lor RFI Suppression, 268
7-3-5 The UC1860 Resonant-Mode Power Supply Controller. 207
3
EIGHT ELEMN
SWlTCHlNO POWER SUPPLY ANCILLARY, SUPERVISORY, AND PERIPHERAL WWEP SUPPLY ELECTRICAL SAFm STANDARDS 271
CIRCUITS AND COYWHEMS 209 I1-0 Infrodudon, 271
IntraduCtion, 209 11-1 Power Supply Construction mirernenfs fw Safety..271
The Optical Coupler (or OptolsdatorE, 209 I1 1 Spacing Requlremem, 271
A Self-Bias Techn~queUsed in Rlmary Slde-ReferencedP o w 11-1-2 Dtelectnc Test W l a n d , 272
supplies. 211 11-1-3 Ceakaw Current Measurements, 275
Wocwpler Circult Design to b i d e Input-- lsolotlon 11-1 4 lnsuldim Resistqnce. 275
In a Switching Power Supdy, 212 11-14 PC Bwrd Requlremenls. 275
Son St& h Swltchlng Power SuWy Designs, 216 11-2 Power Suppfy Trmsfomr Constnrctlm lor Mew, 275
Current timlt Circurts, 217 -_/ 11-2-1 Tmmfrmer Insuldm. 275
8-51 Cutrent Limit Clrcults for Wmary-ReferencedDirect 11-2-2 Trmsfomer Dlelectrlc Strength. 276
Mve Designs. 218 17-54 Tmnsfwner fnsulcrt~mResistance, 276
8-52 C U MLlmit Clrcults for Desigm Utllblng B m Drlvers. 219
8-53 A Univwwl Current Llmlt Cirwit, 222
/ , 11-+4 Trcmnsfmr Creepoge and Clearance Distances. 277
11-26 Transfmer Mdshrre Reddance, 277
Overvoltage Proteclon Clmlts. 223 11-24 VDE Tmnsfmer Temperahre RoHng, 277
8+- 1 The Zener Sense OVP Circuit, 223 11-2-7 UL and CSA Tmhsfmr Temperature Rding. 280
8+-2 Intqrded OVP Clrcults, 224
AC LIE Loss Dekct~rs,228 c
Index
PREFACE
The rapid advancement of microelectronjcs in xecen t years has demanded

the development of sophisticated, efficient, lightweight power systems whsch
have a high power-to-volume density ratio with no compromise in perform-
ance. The high-frequency switching power supply meets these demands,
and recently it has become the prime powering source-in the majority of
modern electronic designs.
Because the complexity of the power supply i s increasing, dedicated
engineering departments and highly skilled engineers are required to un-
d ~ a k their
e design and development. Unfortunately, very few engineers
are college-trained to become power supply engineers. and the ones who
get involved and make it a career do so either due to .-circumstantial involve-
merit or by demand, -- -
.i- r-
On the other hand, very few colleges and universities offer any power
- '
electronic courses with the emphasis on t h ~ d e s i ~ *~ivitching
of power sup-
plies and magnetics. Consequently. college students or practicing engineers
who want to pursue a career in this truly exciting and fascinating field must
enrich their knowledge by studying material mostly in the form of application
notes, published by different electronic companies in promoting their prod-
ucts, or technology articles published in pdessional magazines. The need
for a comprehensive, coherent, a d easy to understand text, which blends
both theory and practice and also covers most of the latest developments in
the switching power supply field, promoted the writing of this book.
This book is intended to be used by either the engineering student or
,the practicing engineer who wants step-by-step instruction to the theory
and design of switching power supplies, It compiles the knowledge of all
those who.have worked in this fiekl for years in order to help those just
starting. It includes enough theory to make the reader aware of the results,
but long mathematical derivations are limited. The end result of the theory
and its application in the design aspect is emphasized.
This Second Edition of the book follmvs the format set forth in the First
Edition. However, c e M n chapters have been enhanced and expanded with
A1
Idv PREFACE
new material. The net result is a more comprehensive and updated presen- Vl Y e
tation of high-frequency switching power supply design, which takes the
reader bom the 20-kHz region to ioday's megahertz region. THE SWITCHING POWER
The first chapter of the book describes the building blocks of a complete
switcher, and each of the subsequent chapters describes arid analyzes each SUPPLY: AN OVERVIEW
building block in detail. The basic design equations are given, but long
mathematical derivations are omitted. Numerous examples are presented
that enhance and reinforce the theory with practical designs.
The book describes all the basic classical switching power supply topol-
ogies, and new topologies are also presented. The switching power transistor,
both bipolar and MOSFET, is extensively analyzed. Subjects such as s n u b h r
circuit designs and base drive designs are extensiveIy covered. Other semi-
conductor devices, such as the gaie turn& (GTO), synchronous rectifiers,
fast recovery rectifien, and Schottky rectifiers are dso presented in depth.
The analysis and design of magnetic components, such as high-frequency
power transformers, power inductors, and magnetic amplifiers along with
application design examples are covered. 1-0 INTRODUCTION
Power control semiconductor integrated circuits (ICs) are covered, and As the integrated semiconductor technology becomes more advanced, sys-
a number of representative popular commereid 1Cs are descsribed as a tem designers as well as electronic manufacturers are emphasizing size and
reference. weight as important features of their products.
Chapter 9 is devoted to the important subject of power suppIy feedback Traditionally, the bulkiest part of a system is the power supply, with its
loop stability, The analysis and design of feedback amplsers is presented heavy isolation tmnsformer, heat sinks, and cooling fans, as in the case of
in a plactical, easy to understand way, simplifying this traditionally difficult series-pass linear designs. I
to understand subject. The trend therefore in recent years has been toward the development of
Chapter 10 discusses the importance of electromagnetic and radio fre- high-eficiency, lightweight, and compact power supplies to complement the
quency interference (EM1-RFI) and presents circuits for its suppression. new system designs. The high-frequency switching power supply is the
Chapter 11covers national and international safety requirements, and it obvious solution.
presents and explains safety design standards such as UL, CSA, VDE, and However, this new type of power supply is much more sophisticated than
IEC. its linear counterpart. It requires knowledge in analog electronic design and
The First Edition of the book was received with enthusiasm, and I was magnetic component design, aIong with logic and control design.
gratified to receive both compliments and constructive suggestions. These The iask of power supply design is not a side project anymore. The
suggestions prompted the writing of the Second Edition, which includes switching power supply has spawned new exciting interest in the power
material asked for by readers. electronics Geld, and the term "power supply engineer" has been redefined
It is my hope that this edition will become.a valuable reference to all and given new respect. New research has been undertaken by the industry
those seeking howledge and understanding of the ever-expanding field of and academia, to push the frontiers of thts truly fascinating field.
high-frequency switching-power supply design. Theadvaneements in thefield have been rapid-and rewarding. F h e p m r
I would again like to thank my h i l y for their support and understanding supply is finally following the advancement of the rest of the electronics.
during the writing of both editions wfthe book, as well as all those who have Power supplies are becoming smaller, more efficient, exiremeIy compact,
contributed to the enhancement of our profession and the power supply and cost-effective. The trend is for higher switching frequencies, above the
field in general. traditional 20-kHz and up to the megaherh: region. Commercial power sup-
Also I would like to thank Ellen Ddmus and Claudia Mungle for typing plies working at 1MHz are already avaiIablefrom a number of manufacturers,
the manuscript for the First Edition and Barbara Stone and Jami Schmid with more to come.
for their contribution in typing the manuscript for the Second Edition. Truly this is only the beginning,
t
George C.Chyssfs
2 HIGH-FREQUENCY SWITCHING POWER SUPRIES THE SWITCHING POWER UPPLY: A N OVEWEW 3
4-4 THE LINEAR POWER SUPPLY 1-2 THE OFF-THE-LINE SWlTCHENO REGULATED POWER SUPPLY
The linear power supply is already a mature technology, which has been The disadvantages of the linear power supply are greatly reduced or elim-
used since the dawn of electronics. Whether tbb type of bower supply inated by the regulated switching power supply,
incorporates tubes or semiconductors, its consrntion and operation are es- Figure 1-2 shows a simplified block diagram of'a high-frequency off-the-
sentially the same. line switching power supply. In this scheme the ac line is directly rectified
Figure 1-1 shows the simplified block diagram of a series.pass linear and filtered to produce a raw hi&-voltage dc, which in turn is fed into a
regulated power supply, In this type of power supply, a low-frequency, 50- switching element. The switch is operating at the high frequencies of 20
or 60-Hz transformer is used to step down the ac mains to a lower voltage kHz to 1 MHz,chopping the dc voltage into a high-frequency square wave.
ofthe same frequency. This secondary voltage is in turn r e d e d and filtered, This square wave is fed into the power isolation transformer, stepped down
and the resulting dc is fed into a series-pass active element. to a predetermined value and then rectified and filtered to produce the
By sampling a portion of the output voltage and comparing it to a k e d required dc output.
reference voltage, the series-pass element is used as a form of "variable A portion of this output is monitored and compared against a fixed ref-
resistor" to control and regulate the output voltage. However, this mode of erence voltage, and the error signal is used to control the on-oil times ofthe
operation dissipates a large amount of power in the form of heat, conse- switch, thus regulating the output.
quently lowering the efficiency of the power supply to 40 or 50 percent. Since the switch is either on or off, it i s dissipating very little energy,
Although the linear power supply in gneral has a tight regulation band resulting in a very high overall power supply efficiency of about 70 to 80
along with very low output noise and ripple, the disadvantages are obvious. percent. Another advantage is the power transformer size which can be quite
As we mentioned, because of its low eficiency, usually bulky and expen- small due to high operating frequency. .Hence the combination of high
sive heat s i n b and cooling fans are needed, and large isolation power trans- efficiency (i. e., no large heat sinksj and relatively small magnetics, results in
formers are used to step down the ac input voltage. Hence, this type of compact, lightweight power supplies, with power densities up to 30 W l i r ~ . ~
power supply tends to be bulky, heavy, and almost unfit for today's compact versus 0,3 W l i r ~ for
. ~ Iinears. Coupled with very wide input voltage range,
electronic systems. 90 to 260 V ac, and very god hold-up time, typically 25 ins, the switcher
Other disadvantages of the linear power supply are its relatively n m has become the choice for electronic system designers,
input voltage range, normally k10 percent, and its very low output hold-
up time, about 1 ms,
a c
Rectifiers
and
filter
- ---
Switching
element
bolatlon
power
transformer
Output
mtlfiers
and fitter
- =v,
--A Isolaaon
power
1
I
4I filter
1 pa^^
element
vM
Lql'-'
transformer
Feedback
contml
vmf
FIGURE 1-1 Block diagram of a series-pass regutated linear power supply, FIGURE 1-2 The basic oithe-line switching regulatd power supply.
1 HIGH-FREQUENCY SWITCHING WWER SUPPLIES
THE SWrrCHlNG POWER SUPPLY: AN OVERVIEW 5
Of course, there are certain disadvantages associated with the switching

drcuits used to protect the load from power supply failures. It is important
power supply, such as higher output noise and ripple, EMI/RFI generation,
to note that although in a linear power supply, overvoltage conditions were
and higher design complexity. However, with careful design these problems
common during failure of the series-pass element (i.e., shorted), in a switch-
may be greatly reduced or elimina+ed.
I ing power supply, failure of the switching element n m d l y results in a
nonoutput condition. However, the output of the switcher will go high if
1-2-1 The Complete Off-the-Llne the feedback loop is opened.
Power Supply Bufldlng Blocb hput/output isolation is essential to an of-the-line switcher. The isola-
The building blocks of a typical high-frequency of-the-line switching power tion may be o p t i d or magnetic, and it should be designed to comply to
ULlCSA or VDEIIEC safety standards. Thus the UL and CSA require
supply are depicted in Fig. 1-3.
1OOO V ac isolation voltage withstand while VDE and IEC require 3750 V
The purpose of this book is to analpe each building block in a logical
ac. Consequently the power transformer has to be designed to the same
sequence, providing the reader with the understanding and with the tools
needed to undertake the design of switching power supplies. As the name safety isolation requirements also.
implies, the input rectification in a switching power supply is done directly
d-the-line, without the use of the low-fr,equency isolation power trans-
former, as in the case of the linear powef supply.
1. Bid,B. M.. and K. G. King: "An Introduction to Power Electmnim," W e y , New York,
The basic operation of the switchiyg power supply was described in the 1983.
previous section. However, the cpmplete block diagram s h o w in Fig. 1-3
B Chryss~s,G., and 0. Friedman: C l w of Power Supply Terms, EPM, July 1979.
contains other important sections such as the ac line RFI filter block, the
ancillary and supervisory blocks; and the 110 isolation block. 3. Hnatek. E. B.: "Design of Solid State Power Supplies," Zd ed.,Van nostrand kinhold,
The EMIt RFE filter could be either part of the power supply or external New York, 1981.
to it, and it is generally designed to comply to national or international 4. Middlebmok, R. D., and S. Cuk: "Advances in Switched Mode Power Convertem," wls.
specifications, such as the FCC class A or class B and VDE-0871. 1 and 2, ~ e s l a c o ,P a s d e n a Calif., 1981.
3
The ancillary and supervisory circuits are used to protect the power sup- 5. Mitchell, D . M.: "DCIDC Switching Regulator Analysis," McCraw-BIU, New York, 1%.
ply-and the electronic circuits it powers-from fault conditions. Generally 6. Pressman, A. I.: "Switching and Linear Power Supply, Power Converter Design," Bayden, I
each power supply has current-limit protection to prevent its destruction Rochelle Park, N. J., 1977. *
during overload conditions. Overvoltage protection is part of the supervisory 7. Severns. B. F,,and C . E. Blmm: "Modern DC-to-DC Switchmode Power Converter Cir-
cuits," Van Nostrand Reinhold, New York, 1985.
circulls, supervisory
ire.,current circuits,
limit
-
ac
RFI
filter
-C
-
Input
m"els
and ffJter
t
Swftchlng
'Iwnent
,Isolation
power
transform=
Output
rectifiers
and trlter
VOU~
FHTtl 4
Feedback
contrw isciation
FIGURE 1-3 The building blocks of a typical OK-the-line high-frequency switching power
supply.
THE INPUT SECTION
I 2-0 THE VOLTAGE DOUBLER TECHHIWE

i
I
11
As we mentioned previously, an off-the-line switching power supply rectifies
the ac line directly without requiring a low-frequency, line isolation trans-
former between the ac mains and the rectifiers. Since in most of today's
electronic equipment the manufacturer is generally addressing an interna-
1
~1 tional market, the power supply designer must use an input circuit capable
of accepting all world voltages, normally 90 to 130 V ac or 180 to 2613 V ac.
Figure 2-1 shows a realization of such a circuit by using thevoltage doubler
technique. When the switch is closed, the circuit may be operated at a
nominal line of 115 V ac. During the positive hdf cycle of the ac, capacifir
Cl is charged to the peak voltage, 115 V ac x 1.4 = 1W V dc, through
diode Dl.During the negative half cycle, capacitor C2 is charged to 160 V
1 dc through diode D4.Thus, the resulting dc output will be the sum of the
voltages across C, + CZ,or 320 V dc. When the switch is open, Dl to D,
i form a full-bridge rectifier capable of rectifying a nominal 230-V ac line and
~roducingthe same 320-V dc output voltage.
ft 2-1 COMPONENT SELECTION AND DESIGN CRITERIA

I
3 - 2-44 Input-Recfiffets
When choosing either a bridge rectifier assembly or discrete rectifiers, the
designer must look up the following important specifications:
I1
f'
I. Maximum forward rectification current capability. This figure depends
primarily on the power level of the switching p&r supply design, and
the selected diode must have at least twice the steady-state current ca-
pacity of the calculated value.
where C = capacitance, pF
I = load current, A
t = time the caPacit(lr must supply currents, ms
AV allowable peak-to-peak V *
calculate the value of the input filter capacitors of a 50-W switching power
supply working at 115 V ac, 60 8,.
I SOLUTION
i The first step is to calculate thr dc load current. Assume a worst-case efti-
ciency of 70 percent for the Potrvr supply, Then for a 50-\V output the input
power is
plGUm 2.). This drmit may be used for either ).W or 230-V ac L w t , depending on the
Using the voltage doubler techntque (see Fig. 2-11, at 115 Vac, the dc output
paition of the switch. Inrush current limiting, input transient protection, and discharge will be 2015 x 1.4) = 320 V dc. Therefore, the load current will be I =
resistors are also shown. PIE = 71.51320 = 0.22A. Nuw assume the design can toleraze a ripple of
30 V peak-to-peak and that thtl capacitor has to maintain the voltage kvel
for every half c~cle,i.e., one-Iulf the line frequency or for about 8 ms for
2. Peak inverse voltage (PIV) blocking capbility. Since these rectifiers are a 60-Hzac line frequency. Tlrt*ll using ~ q 2-1,
.
used in high-voltage environments, they must have a high PIV rating,
normally 600 V or more.
3. High surge current capabilities to withstand the peak currents associated
with turn-on.
I
1
We choose a standard value or 50 E
Since in the voltage doubIt*r rnnfiguration C = C, + C2, then C1 =
Cg = 100 pF, which is the Qlucity needed for this 50-W design.
L.
Proper calculation and selection of the input rectifier filter capacitors is very
important, since this d l influence the following pedormance parameters:
2-2 INPUT PROTECTIVE P E V ~ E S
the low-frequency ac ripple at the output of the power supply and the
holdover time. Normally high-grade electrolytic capacitors with high ripple 2-24 Inrush Curremi
current capacity and bw ESR are used with a wrkingvoltagr-of-200 V dc An of the-line switching p ~ t 4 - supply
r may dewlop extremely high peak
minimum. Resistors R4 and R5, shown in Fig. 2-1 shunting the capacitors,
provide a discharge path when the supply is switched off. I! inrush currena during turn-ort, the designer incorporates some form
1
of current limiting in the inpyt section. These currents are mused by the
The formula to calculate the filter capacitor is given by charging of the filter capacitor< which at turn-on present a very low imped-
FA ance to the ac hies, generally.tnly their ESR, l f n o protectbn is employed,
I* these surge currents may apprtuch hundreds of amperes.
10 HIGH-FREQUEMCT SWITCHING POWER SUPPUES THE INPUT SECT1M 11
Two methods are widely employed in Introducing an impedance to the is essentially the only impedance across the ac line, thus limiting the inrush
ac line at turn-on and in limiting the inrush current to a safe value. One is current.
using a resistor-triac arrangement, and the other using negative temperature As the capacitors begin to charge, current starts to flow through the
coefficient (NTC) thermistors. Figure 2-1 shows how these elements may be thermistors, heating them up. Because of their negative temperature coef-
employed in a p w e * +upply. ficient, as the thermistors heat up their resistance drops. When the ther-
mistors are properly chosen, their resistance at steady-state load cursentwill
The Resistor-Trloc Technique Using this inrush current limiting technique, be a minimum, thus not affecting the w e r d efficiency of the power supply.
a resistor is placed in senes with the ac line. The resistor is shunted by a
triac which shorts the resistor out when the input filter capacitors have been
fully charged. This arrangement requires a triggkr eircuit which will fire the
triac on when some predetermined conditions have been met. Care must
I 2-2-2 Input Transtent Voltage Protecflon
Although the ac mains are nominally rated at 115 V ac or 230 V ac, it is
be taken in choosing and heat-sinking the triac so that it can handle the full common for high-voltage spikes to be induced, caused by nearby inductive
input current when it is turned on. switching or natural causes such as electrical storms or lightning. During
severe thunderstorm activity, voltage spikes in the order of 5 kV are not
The Thermistor Technique This method uses NTC thermistors placed on uncommon.
either the ac lines or the dc buses &er the bridge rectifiers, as shown in On the other hand, inductive switching voltage spikes may have an energy
Fig. 2-1. content
The resisbnce-temperature characteristics and the relationship of the
W = IW* 12-21
temperature coefficient a o f t h e NTC thermistor are shown in Fig. 2-2.
When the power supply is switched on, the resistance of the thermistor(s) where L is the leakage inductance of the inductor, and 3 is the current flowing
through the winding.
I Therefore, although these voltage spikes may be short in duration, they may
carry enough energy to pmve htal for the input rectifiers and the swiGching
transistors, unless they are successfully suppressed.
The most common suppression device used in this situation is the metal
oxide varistor (MOV) transient voltage suppressor, and it may be used as
shown in Fig. 2-1 across the ac line input. This device acts as a vanable
impedance; that is, when a voltage transient appean across the varistor, its
I
impedance sharply decreases to a low value, clamping the input voltage to
a safe level. The energy in the transient is dissipated in the varistor. The
following procedure gives a guide in selecting these devices:
1. Choose the ac voltage rating of the MOV to be about 10 to 20 percent
greater than the maximum steady-state circuit value.
2, Calculate or estimate the-maximum-transient-energ)* in joules that-the
I- circuit may experience.
3. Make sure that the maximum peak surge current of the device will be
FIGURE 8-2 An NTC thermistor's resistance drops drasti- properly rated.
d y as the temperature increaws;a i s htemperature coef- --
ficient d the thermistor, expressed in percentage per degree After all of the above have been established, the proper MOV may be
matigrade. selected from the manufacturer's data sheets.
. .. www.IranSwitching.ir
.
TYPES OF WWER CONVERTERS 15
14 HGH-FREWENCY SWITCmG POWER SUPPUES
(bl
FIGURE 3-1 Model of tbe flyback o r b u c k - h t converter. (a) S w M
clased, (bl switch open.
FIGURE 3.2 Model of the fomrd w buck converter. (a) Switch clased, [b)
switetr opn.
pulsating, flowing when the switch is closed and abruptly interrupted when
the switch is open.
Figure 3-4depicts an isolated flyback converter with the associated steady-
Last, Fig. 3-3shows the push-pull mverter, which in reality consists of state waveforms, The circuit operates as follows. When transistor Qlis on,
the two forward converters operating in a push-pull (or more correctly push- primary current starts to build up in the primary winding, storing energy,
push) action, with alternate closing and opening of either switch S, or 3,. Due to the opposite polarity arrangement behveen the input and output
This circuit is also known as buck-derived. windings of the transformerchoke, there is no energy h a n s f d to the
load since d i d e D is reverse-biased.
When the transistor is turned off, the polarity of the -windings reverses
3-1 THE ISOLATED FLYBACK CO-R due to the collapsing magnetic field, Now diode D is conducting, charging
The model ofthe flyback converter shown in Fig. 3-1has no safety isolation the output capacitor C and delivering current lLto the load.
between input and output. h off-the-line switching power supply normally Since the isolation element acts as both a transfbnner and a choke, no
requires mains isolation in the form of a transformer, To be more precise, additional inductor is needed on the output section of the flyback converter.
although in the diagram the isolation element appears in the form of a In practice, though, a s m d inductor may be necessary between the rectifier
transformer, its action is that of a choke and therefore it is more correctly and the output capacitor in order to suppress high-frequency switching noise
referred to as transformer-choke. spikes.
Im l i t t
REERENCES TYPES OF POWER
1. Cbiysds, C.: Power Supplies: A Switchhg f i t e r n ~ w D&gtt
. Neurs, M a y 21, 1979.
8, Hnatek, F. R: "Desiga of Solid State Rower Supplies," 2d ed., Van Nostrand Rdnhold,
New York,\!981.
3. Pressman, b 1.: "Switchhga-ld+bear Rower Supply, Power Converter Ddgu," Hayden,
Rahelle Park, NJ,1977,
8 4 DEFINmOMS AND DIMEWKINW

Although there are numerous converter circuits described by a number of
authors and researchers, basically all of them are rehted to three dassid
circuits known as the "0ybwk or buck-hst," the "b~wardw buck," aod
the "push-pull or buckderivedmconverter. k model of the ftyback converter
is illustrated in Fig. 3-1, and operation of the chit is as fuIl&m.
When the switch S is closed (Fig. 3-10}, current flows through inductor
L, storing energy. Because of the vohge poIarity,aiode D is reyerse-
biased, thus no voltage is present across the load R,. When the switch S is
open (Fig. 3-lbj, inductor L reverses polarity because of the cobpsing r
magnetic field, f o d - b i a s i n g dide D, and inducing a current flow Ic iw
the polarity shown. Thus, an output voltage of opposing poIarity to the input
voltage appears across RL.Since the switch cornmutates continuous inductor
current alternately between input and output, b t h currents are pulsating
in form. Therefore, in the buck-boost converter, energy is stored in the
inductor during the switch on m o d ; then this energy is transfkmd to the
load during the Byback or the switch OK period.
Figure 3-2 illusfrates The. operation of a forward converter. Wben the
switch S is closed, k m n t I flows in a h m r d manner through i n d u e l ,
producing an output voltage across the load with the polarity s h tn .Fig.
3-2a. Diode D is also r e v e r s e - b i ~ ~ e . b d ~ e ~ t h e e i B ~ 1 t a g
polarity. When the switch-S opens (Fig. 3-Zb), the magnetic field in L
changes polarity, f d - b i a s i n g diode D and producing a current h u g h
w i t w C, as shown. Therefo~,the output voltage polarity across l&
remains the same. Diode D is d e n called a "free-wheeling" or "flywheel"
diode. - - -- . --"
Because of this switching action, the output current is continuous and
therefore nonpulsating. In contrast the input current is discontinuous, hence
FIGURE 3-3 Model of the push-pull w hckderh.ed con-.
The switching transistor used In the flyback converter must be chosen to

handle peak collector voltage at turn-of and peak coliector currents at
turnon. The peak collector voltage whlch the transistor must sustain at
where Vh is the dc input voltage, and 6,- is the maximum duty cycle.
Equation 3-1 tells us by examination that in order to limit the collector
voltage to a safe value, the duty cycle must be kept relatively low, normally
below 50 percent, i.e., dm, < 0.5. In practice 6- is taken at about 0.4,
which limits the peak collector voltage VCEqrnw 5 2.2V,,, and, therefore,
transistors with working voltages above 800 V are usually used In off-the-
h ef l p b a c k - ~ e s i g n s .
The second design criterion which the transistor must meet is the working
collector current at turnon, given by
where I p is the primary transformerchoke peak current, n is the primary- "c.
to-recondary. turns ratio, and IL is the output load current. FIGURE 311 The fwlated flyback coawtar and L ss-d waveforms.
t8 H I W F M W E W SWITCHING POWER W U E S Wfs OF POWER CONVERTERS 19
To derive an expression of the peak working collector current in teni~n not saturate. An extensive analysis and design is given In Chap. 5. Here it
of converter output power and input voltage, the following equation may bc suffices to say that a core with a relatively large volume and air gap must be
written for the energy transferred in the choke: used.
The effective bmsformer-choke volume is givendby
Volume = Wert-LUt
where q (eta) is the efficiencyof the converter, Bk
The voltage across the h-anshrmer may be expressed as where ILnmr = determined by load current
fi, relative permeability of the chosen mre material
v,, = -
L di
dt
B, = maximum flux density of the core
The relative permeability p, must be chosen to be large enough to avoid
If we assume di = I p and lldt = flaw, then Eq. 3-4 may be rewritten as excessive temperature rise in the core due to restricting m e and wire size
and therefore copper and core losses.
wef
v,, = -
,6
344 Vadutlons of the Bask Rybaelr Converter
As was discussed with the basic flyback circuit, the collector voltage of the
switching transistor must sustain at least twice the input voltage at turn-off.
In cases where the voltage value is too high to use a commercial transistor
$ - 6 Eq. 3-3 we get
~ u b s t l t u t. i~n--~ ' ~ ~ into type, the two-transistor flyback converter may be used, as shown in Fig.
.....* ..-- - >...
,pljpljplj
- . . 35. This circuit uses hw transistors, which are switched on or off simulta-
...,,..-.-.,,. ,,. neously. Diodes D,and Dzact as clamping diodes resMcting the m+mum
.--:M..,"
A%: collector voltage of the transistors to VhnThus lower voltage transistors may
Solving for I;,;.:.
. . be used to realize this design, but at the expense of three extra components, r
+. ,...
..
, . *-. .
,b
i.e., Q,, Dl, and D2 2-
-
->
. --
.
. .,.
.--,+,, I? = zpw An advantage of the flyback circuit is the simplicity by which a multiple
~gV,Srn, output switching power supply may be realized. This is because the isolation
Now substituting Eq. 3-7 into Eq. 3-2 we get the expression for the tmnsistor element acts as a common choke to all outputs, thus only a dbde and a
working current in terms of output power capacitor are needed for an extra output voltage, Finre 3-6 illustrates a
practical circuit.
To simplify Eq. 3-8 wen more, assuming a converter elficiency of 0,8 (80 3-2 THE- m A T E D FORWARD COWYERTER
percent) and a duty cycle,S = 0.4 (40percent), then At first glance, the isolated FoMFard converter circuit .resembles that of the
-
Ryback converter, but there are some definite and iristinct &rences be-
ween the two circuit topologies and their operation. Figure 3-7 shows the
basic f o d converter and its associated waveforms.
Since the isolation element in the forward converter is a pure transformer,
34-2 The Flyback Converter TransformeWhoke a second inductive energy storage element L is requbed at the output for
Since the transforme~hokeof the flyback converter is driven in one direc- proper and egcient enerp) transfer. Notice also that the primary and s e e
tion only of the B-H characteristiccurve, it has to be designed so that it will o n d q windings of the transformer have the same polarity, i.e., the dots
fVPES OF POWER CONMMMS 21
1 Ilt c , ~0;t.l
are at the sarne winding ends. Funct&n of the circuit is as follows. When
FIGURE 3-6 Multiple outplb may be easily derkwl using a f l y b d wnvertcr.
Q, turns on, current builds up in the prfmary winding, storing energy. Botb W H v e and negative voltages are pass* using extra output windimgs, a
Because of the same polarity of the transformer s e m b y winding, this dl&, and a smoothing &tor,
energy is forward-transferredto the output and also stored in inductor L
through diode 4 which is fad-biased. ~ i o d D3 e is back-biased. When 3-14 lhe Fornard Converter Swbhing TranSMor
Q,turns of, the transhmer winding voltage reverses, back-biasing d i d e
D,. Now, the flywheel diode D3 JS f o d - b i a s e d , cmducting current in Because of the tertwy winding and M e Dt (Fig. 3-7j,the voltage across
the output loop and delivering energy to the load through inducior L. transistor Q1at turn-off is limited to
The tertiary winding and diode Dlprovide transformer demagnetization v ~ ~ + =m2vw (3-12)
when Q is switchd~offbyreturning the transformer magnetic energy into --- -
the output dc bus. The dark areas on the waveforms of Fig. 3-7 show the The waveforms nlso show that the peak collector voltage d2Vh is maintahed
magnetizing-dernagnang current, given as for as long as Dlconducts, that is, for a period of .
,dT Also by inspection
of the same m e f o r m s we can see that the transistor collector current at
turn-on will have a value equal to that derived for the flyback converter plus
the net amount of the'magmtizationtion current. Therefore, peak collector cur-
rent in the transistor m a y be written as
where T6, is the m od &hen transistor QIis on, and L is the output
inductance in microhenrys.
'*
n P E S OF POWR CONVERTERS 23
where n primary-bsewndary turns ratio

1' = output inductor current, A
Td,, = period trhsistor is on
L = output inductance, p H
From the expression
Equation 3-13 may be rewritten as
Assuming that the magnetizing current term nTV,,ILis very mdl compared
to the peak collector current, for d practical purposes may be derived
as s h in Sec. 3-1-1 to be I .
Care must be taken when designing the-transformer of the forward converter

to choose the proper core volume and air gap to awid saturation. The
transformer equations given in Chap.5 may be used to design the forward
transformer. The core volume of the transformer is given by
Volume = P&IIB,,L (3-18)

Be,,
where
It should also be noted that the duty cycle of the switch 6- must be kept
below 50 percent, so that when the transformer d t a g e is clamped Zhmugh
. t h e , , t e q winding, the Integral of the volt-seconds between the input
voltage, when QIis On, and the c h ~ ~ p i level,
n g when Q,is off, amounts to
FIGURE 3-7 Realization or the isolated forwurd converter and its assdated zero. Duty cycles above 50 percent, i.e., 6 > 0,5, will upset the volt-seconds
fonm. The crass-hatched artas depict tmnsfonner magnetizing currents.
m
s QF POWER CONVERTEB 2s
balance, driving the tiamformer h t o saturation, which In turn produces high

collector current spikes that may destroy the switching transistor.
~lthoughthe clarnphg action of the tertiary winding and the diode IImlt
1
.. the ti.ansIstor peak collector voltage to twice the dc input, care must be
taken during construction to couple the tertiary winding tightly to the pri-
mary (bifilar wound) to eliminate fatal voltage spikes caused by leakage
inductance.
3-2-3 ~atihonsof the Basle Forward Converter

In a s e s where the input voltage is too high, the two-transistor Forward
converter may be used, as in the flyback case. The two transistors are
switched on or of simultaneously, but the transistor voltage does not rise
above Via' Figure 3-8 shows this circuit variation.
The foward converter may also be used to derive multiple output volt-
ages, although this circuit requires an extra diode and choke in the output
section of each derived voltagZ:Tt should-be' noted"here that the flywheel
diode must have at least the same current rating as the main rectifying diode,
FXGURE 3-9 T k h a r d transformer suits fteslf well to multipk oatputs, altb& the
drcut r e a h t h is more apedw than that of the flyback cmwerter because of the exhe
diodr and choke h each output.
since it provides full output current when the transistor 1s off. Figure 3-9
shows a practical multipk-w2put f o d converter.
-
- 3 3 THE PUSH-W CONVERIER
The push-pull converter is really a n m g e m e n t of two forward converters
working in antiphase. Since lmth halves of the push-pull converter are de-
livering power to the load at each half cycle, a more appropriate name might
have been the push-push converter, but over the years, the term push-pull
hi-prevailed. - -'
FIGURE f -8 The ~ ~ i Pr0-dt omnverter ~ limits the mllfftm vdltage of each FIgure 3-10 shows the basic convenMmd push-pull configuration and its
I
transistor ta V, bemuse of the damphg action of Dr sad Dp associated waveforms, Fmm the waveforms we can see that because of the
WE3 OF POWER CONVERTERS
presence of the two sets of switching transistors and output diodes, the
average rmmnt in each set is reduced to 50 percent wer the equivalent
fornard converter. Note that In the interval between transistor condytion,
diodes Dl and Dp conduct simultaneously, essentially shorting the secrmdary
isolation bmdomers and delivering power to the output, acting as flywaleel
diodes.
The output voltage of this wnverter'may be derived as
The value of ,6 in Eq. 3-20 must remain below 0.5 In order to avoid
simultaneous conduction of the switching transistors and therefore destruc-
tion. Assuming d, = 0.4, then Eq. 3-a0may be rewritten as
where n is the primary-to-secondary turns ratio.
Qi.on 02. m , on
Ql
3-34 l o Push-Pull Converter Transformer .-
v,, Vm - Vin
- - As we noted when discussing flyback and Forward converters, their trans-
formers exploit only one-half of the B-H characteristic curve and therefore
are hub and have an air gap. Assuming In the push-pull converter tHat the
conduction times of the two transistors are equal.(or they are forced to be
equal) the transformer will use both halves of the 8-Hm e and the volumz
'el - - of the core will be halved. An air gap may not be necessaq.
The volume of the transformer is given by
- 2% - - Volume = ~ P o P LJ ~ L 13-22)
"CE1- B:,
k
= nV,,TI4L is the magnetizing current.
where ImI,,
101 - \-
L
4 In Chap. 3 an extensive design analysis is @en for the family of push-pull-
based converters.
'DZ
-=
7 W = P u W P u l l - ~ n v ~ ~ r s
Since each half of the push-pull converter in essence is aforward converter,
IL the voltage across each transistor at t u m d i s limited to
I
-- V C E ,=~ 2Vh
~ (3-23)
I The peak collector current of each transistor Is given by .-,
T
FIGURE 3-10
26
--- --
,
- , -...-.,
--.-. . . . - -
- - . ... .
,
~ 38 HGKFWYINCY WTCHING POWER SUPPLIES 1YPES OF PMYER CONVERTERS 29
Assuming that I, a ILln, then
.. \. The bilowfng expression is derived fbr the transistor coIIector working cur- - 5.
rent in terns of output power, dciency, and duty cycle, as shown in Sec.
3-2-1: v~~ ,
a1
-
Assuming a converter efficiency q = 8Q percent and 5- = 0.8, then the Cumnt l i k e due to
d n g transistor collector current is transformer raturation
L
Q1 -
"2 I
Although the push-pull converter offers some advantages, such as nonisolated FIGURE 3-11 These are prsctid d t q e and mrrent wadormo a-ted with tbe
base drives and simpler driver circuitry, it also presents a number of dis- push-pull mverter s h in Fig. 3- 10. c-
advantages which make its use as as off-thehe switcher aImost not practical.
The first Iimitation is the voltage rating of the transistors, which should all working conditions andtemperatures, If the transistor characteristics are
handle twice the Input voltage to the converter pIus any leakage spike that not identical, "flux walkingn to one direcHon d t h e B-H cum occurs, driving
might result because.of transformer leakage inductance as shown in Fig, the core into the saturating region. Saturation of the core means high col-
3-11. This means that switching transfstors abIe to block over 800 V (for use lector current spike in one of the transistors, as shown in Fig. 3-11. .,
in 230-V ac Iines) must be specified for an of-the-line push-pulI converter. This exMssive current produces large amounts of power loss In the tran-
This may present a problem for high-power converters, since high-current- sistor, heating it up, which in turn further unbalances the transistor char-
high-voltage transistors are not all &&common and also tend to be expen- acteristics, thus saturating the core even more, producing even higher
sive. saturation currents, and so on. This vicious cycle wilI continue until the
F a r e 3-11 aka shows the second and more severe of the problems transistor goes into a thermal maway, which 'leadsto destruction.
associated with push-pull circuits, that is, transformer core saturation. In Two solutions to the problem are possible. First is gapping the core,
today's witching power supplies, Errite core material is widely used because which means an increase in leakage inductance and therefore the need of a
of its low losses at high frequencies of 20 kHz and b e . Unfortunately the dissipative snubber at the expense of lower &ciency. Second is using a
ferrites also have a high susceptibility to core saturation because of their low symrneby correction W i t , which ensures balance operation of the power
- flux density, which is usually around 3000 gauss (G). Therefore, a small - - ~ b . ) L ~ e o n - a F r aof the t driver
l ~generator
amount of dc bias in the core will drive it to saturation. This is exactly what equal. UIlfortunately this method requires extra circuitry, which adds to the
happens with the push-pull circuit. When one transistor switches on, the cost and complexity of the mnverter.
flux swings in one direction of the B-H curve in order to reverse direction The disadvantages of the push-pull circuit may be alleviated by using the
when the first transistor switches d and the. second transistor switches on. W - b l d g e or full-bridge power wnverter. The half-bridge wnverter is very
In order b r the two areas ofthe flux d e n s i ~to be equal, the saturation and - poplib among p=er suppIy designers, and it is extensiveIy discussed In
witching characteristics of the switching transistors must be identical under Sec. 3-4-1.
M HIWREQUWCY SWRCCRNG POWER SUPPUES
3 4 ClRCUlT VARIATIOWS OF THE PUSH-PULL COWERTER sistm to no more than V,. Now lower voltage transistors m y be specified,
34-1 fhe WaCBridge Converter and in practice 400-V transistors are commonly used in this application.
There is one minor price to pay when using the half-bridge drcult, that
h mentioned previously, there are two main reasons h r developing the is, since the transformer voltsge has been reduced to V,,/2, >he transistor
half-bridge drcuit. One is to be able to work the converter from both 115-
and 230-V ac input voltages without worrying about using high-voltage tran-
sistors, and the second is either to incorporate a simple means of balancing
and a duty cycle, ,a -
working current will double. If we assume a converter eBcienclylof80percent
0.8, then the transistor working currtnt is
the vol t-seconds interval of each switching transistor without gapping the
power transformer or to use expensive symmetry correction circuits, Figure
3-12 shows the basic dual input voltage half-bridge converter. The second objective of this topology is to balance the volt-second integral
Notice that in the half-bridge mnfiguration the power transformer has of each switching transistor automatically in order to avoid core saturation.
one side connected to a floating voltage potential created by the series Figure 3-12 shows how this is done by inserting a capacitor in series with
capacitors C1 and Crwhich has a value of Vh/% 160 V dc at nominal input the transformer primary. Referring to Fig, 3-12, assume that the two switch-
voltage. The other end of the transformer is connected at the junction of ing power transistors.have unmatched switching characteristics; that is, tran-
the Ql emitter and Qp mllector through a serfes capacitor C,. When Ql turns sistor Qihas a slow turn-off while transistor QZhas a fast turn-off.
on, this end of the transformer goes to the positive bus, generating a voltage Figure 3-130 shows the e m of the slow turn& of transistor Q t upon
a .. pulse of 160 V When--. Ql---
turns d and Q1t u r n on, the polarity of the the ac voltage waveform at the junction of Qiand Q2.A volt-seconds un-
' transformer primarfreverses, since it is now connected to the negative bus, balance. depicted by the cross-hatched area, is added to one side of the ac
generating a negative pulse of 160 V. The turn-on-turn-08 action of QIand voltage waveform. If this unbalance waveform is allowed to drive the power
QI therefore will generate a 320-Vpeak-to-peak square wave, which in turn transformer, Bulr walking will w u r , resulting in core saturation, producing
is stepped down, re&&, and filtered to produce the output dc voltage.
I transistor collector current spiking, which will lower converter ekiency and
We have therefore succeeded with this converter topology in achieving may also drive the transistor into thermal mnaway to destruction.
the first g d of reduchg the. voltage s&ss imposed on the switching tran- By inserting a coupling capadtor in series with the primary transformer
winding, a dc bias proportional to the volt-seconds unbalance b picked up
by this capacitor, shifting the dc level as s h in Fig. 3*13b,thus balancing
the volt-seconds integral of the two s w i t h g periods. 0
k way of decreasing transistor t u r n 4 t h e is to use W e r clamp diodes

in its base circuit, which in &t do not allow the busistor to saturate fully,
thus decreasing the storage time. We will discuss more about Baker clamps
and their applications in Chap, 4.
%4-2 The Serlso Couplhg Capmcbr

The power transformer coupling capacStor described previously is normally
---
A
a film ~JJXnonpolar capacitor, ,capab1~'ofhandlingthe full primary current.
To minimize heating e h - & w ESR capacitor must be used or a bank of
capacitors may be placed in parallel to lower their ESR and also get the
desired capacitance. The Wowing analysis is a guide in selecting the proper
value for the coupling capacitor.
PlGURE 3-13 Bask half-bdge +don. N o t k &at the same trmistors and ha* - - .
lonner a used for both IW and W V ac inputs. DJand are leakage lnductan~ecam- -- By Wpecting Fig. 3-12, it can be stated that the coupling capacitor and
mutathg diodes. the output filter inductor h m a series resonant circuit. The resonant fie-
I
I-. .- -
. .- .- .- .
-,,.-.-,r.-
-., ::.-I.--->
,
. -. . .-, .
, , A," - -.. .-..
, .-
1 quency, horn simple circuit theory* Is
MPES OF POWER CONYfRTERS 9s
where jn = resonant frequency, kHz t.
-
%
C = coupling capacitance, &F . :I

L, reflected filter inductance, p H
The reflected filter inductance to the transformer primary Is
where N,IN, is the transformer primary-to-secondary turns ratio, and L is

the output inductance (pH).
Substituting Eq. 3-30 into Eq. 3-29 and solving for C we get
In order for the chargmg of the cwpling q a c i t w to be linear, the

resonant frequency must be chosen to be well below the converter switching
frequency. For practical purposes, we will assume that the resonant fre-
quency m t be about o n e - h u h the swiicbg frequency, expressed as
jn = 0.SA ' 63-32)
I
where.& is the converter-&itching frequency (kHz).
EXMdaE 3-1
Find the coupling capacitance of a converter workhgat 20 kHz, which has
an output inductance of 20 pH and a transformer turns ratio of 10.
SOLUTION
Since the switching frequency is U) kHz, the resonant frequency h m Eq.
3-32 wiIl be jn= 5 kHz.The reflected inductance from Eq. 3-30 is L, =
1WU) x 10-7 =- 2.#-Q x = 2 mH. Therefore the coupling a@-
tor is -
FIGURE 3.13 (a) W a w f m shm a k d s d a n c e , depkted by the
eross-bakbed area, ou the ac voltage belore the series capadtor. T h i s unbdalaace
is dw to slow turn-off of tmnslrtor Q,.(b)T k same ac waveform is shorm after
the series capitor has rbifted the h level to balance the v o l t - r e d inte&. -
...
hotbe; important aspect to .the value of the coupling -tor
is its c h r g h g voltage. Since this q a t w charges and discharges every half
cycle and shifts the dc level as shown in Fig. 3-12, Its voltage either adds capacitor. Although this theoretical voltage rating may be quite low, in
or subtracts to the V,12 value impressed on the transformer primary. Of practical designs film capacitors with UK)-V ratings are normally used.
course the most critical design criterion occurs when the voltage d the
charging capacitor bucks the V,/ 2 value, since If this voltage is excessive, 4
EXAMPLE 3-2 .. I,
it interferes with the converter regulation at low line.
There are hHO steps by which to check this voltage and in turn to correct Suppose we were to use the c a m t o r calculated In Example 3-1 in a
the calculated capacitance, The capacitor charging voltage is given by 200-W, 20-kHz half-bridge converter., Verify if the calculated value of
0.50 p F is acceptable. If not, then calculate the appropriate value of the
coupling -tor.
where I average primary current, A

C = coupling capacitance, pF From Eq.3-28we calculate the transistor working current at nominal voltage
dt = time interval capacitor is charging, ps to be
The time interval by which the capacitor is charging is given by
--
hsuming that the converte; has an input voltage tolerance of 2 20 percent,
and then the heaviest current of the transistor will occur at low line. Making the
correction, the worst-case collector current will be ..
where
dm, -
T = switching period, ps
dutrcycle
fs = switching frequency, kHz
Using Eq. 3-33 the coupling capacitor charging voltage is I
For a WkHs converter having a duty cycle of 0.8 (80 percent), the charging A charglng voltage of 90 V is too high and will interfere with converter
interval will be 20 ps. regulation at bw line; therefore a new value of coupling capacitance must
The charging voltage Vcmust have a reasonable valueanywhere between be found. Choosing a charging voltage of 30 V and using Eq. 3-36 we get
10 and 20 percent d V,12; that is, if V,12 = 160 V nominal, then 16 5
Vc 5 32 V for-good converter regulation. If the chwglng voltage exceeds
these limits, a new calculation is needed to get a better capacitance value.
This value then will be given as A standard value of 1.5 p F may be used with a minimum voltage rating of
30 V. Practical capacitor voltage ratings of NOV are more commonly used
for safety purposes.
where I average primary current, A
dt = charging interval. ps
dV, = arbitrary number from 16 to 32 V 3 4 - 3 The Commutcrtlng Dbdos
The arbitrary number for & may be so chosen to give a capacitance value - - In Fig. 3-12, showing the basic half-bridge converter, diodes D5and Dewere
close to a real-world standard value. Substituting the calculated-capacftance used across transistors Q, and Qe,respectively. These diodes are called
given by Eq. 3-36 we can get the voltage rating of the series coupling cwnmutading diodes and they have a dual function.
--.- .......
+-. . .- ..----
. -. ....-
. . . . . . . . www.IranSwitching.ir
36 HGKFMQUENCY SWITCWG POWER WMS
1. When the transistor turns off, the commutating diode steers transformer
leakage inductance energy back to the main dc bus. Thus h~gh-energy
leakage inductance spikes, such as the ones shown in the VCpwaveform
of Fig. 3-11associated with the push-pull circuit, are not present.
2. The commutating diode prevents the collector of the on transistor from
swinging negative in respect to its emitter in the event of a sudden off-
load situation due to an increase in transformer flux. In such an event
the commutating diode will bypatis the transistor untll the collector goes
positive again, preventing the device from an inverse conduction and
possible damage.
The commutating diodes must be fast-recovery types with a blocking voltage

capability of at least twice the transistor collector-to-emitter off voltage. In
practice, diodes with a reverse blocking voltage of 450V are commonly used.
3-5 THE FUU.BRID6t CIRCUIT

During our discussion d the half-bridge k i t we noted that, although we
successfully reduced the voltage stress of the switching transistors at turn-
off to half the input dc voltage, the price to pay was doubling the collector FIGURE 3-14 The full-bridge converter dmit
.'
current at turn-on, as compared to the push-pull clrcuit This constraint is

all-right for low- and medium-power applications, but it becomes more
prohibitive for high-pwer applications. where high-voltage-high-mrrent All other properties of this converter remain the same as For the half~bridge
transistors are less available. converter, anhall the formulas developed for the calculation of other ele-
men& may b used here as well.
,
A way-&getting around this problem is to construct a circuit which retains 0.
the voltage properties of the half-bridge and the current properties of the
push-pull. This circuit is the full-bridge topology and it is shown in Fig:
3-14; In this circuit diagonally opposite transistors Qtand Q, or Q3and Qg 3-6 A HEW ZERO OUtPVC RIPPLE CONWRTER
are simultaneously turned on. All the circuit topologies discussed in the previous sections produce an output
This transistor action causes the voltage imposed on the transformer pri- current which has a certain amount of ripple. In recent years, anew converter
mary to swing between +V,,and - VhnThus the transistors never see a topology has emerged called the Cuk converter, named after its Inventor,
collector OK-voltage of more than V,. Also the current through them is half Dr. S. cuk, which with appropriate transformer design can have an output
of an equivalent half-bridge circuit. r i p p l e d zero.
A drawback of this topology is the necessity of using four transistors, and The basic nonisolated converter is shown in Fig. 3-15. Operation of the
sincediagonally-apposite traasistars-amaut-th- ekwkis-as-MIms. -When-trans&r-Qrid; -diede-Dl-imwitched-m,
drives for each transistor must he used. h g m g capacitor C1 by the input c u m t I ] , Dudng the interval when Q,
Assuming a converter efficiency of 80 percent and a 0.8 duty cycle, then switches on, Dl turns off, grounding thepositive terminal of capacitor C,.
the transistor working current is Thus a current It flows through &, pmducing a negative output voltage at
the load.
S& this converter has the properties of a combined buck-lmst topology,
and since the energy transfer is capacitive, the input and output currents
MPES OF POWER CONvERERS 39
FIGURE 3-15 R d k a H o n of the bask Eulr ooaverter and B mdinput and

t
..-.* . WWt cur_=-y?-p;,-".,
.
-
. --.-.-
1-
a
.
"
-.
.'+.--3.
".
-
+A-....d
are v e v cloi~g&+$ue dc quantities having negligible switching ripple. But

"negligible-$-@$ng ripple" does not actually mean "no ripple," which is
the u l t i m a ~goal$f this topology. Fortunately, the zero output ripple to-
pology may beachieved by making the fbllowing observations. in order for
the average"d<yv~ltage
-+ --
across each inductor to be zero, the two waveforms
must be' qud identical. But in-order to achieve this the two inductors
must share thgsame core and they must have the same number of turns. FIGURE.3-E- The coupled inductor version of the Cuk nmverter and its assodated
I
Figure 3-16. shatGs the implementation of this idea. mmnt waverotms.
C
Since th&.Tvikcoupled inductors now form a transformer, the dective
inductance of each winding is altered by the alternating inductive energy a shown in Fig. 3-176. If we make this Inductance very large, the current
transfer. If we were to set a turns ratio of 1: 1, both inductances would it diverts from the two series capacitors is negligiile and therefore the
double, thus reducing the input and output ripple to half those of the un- operation of the converter remains unaffected.
mupled converter. This is indeed a ver)l important obsemation, since if we Now by changing inductor L into an isolation traasformer, the desired
'
altered the turns ratio to the point where the primary-to-secondary turns dc isolation has been achieved. The basic isolated version of the cuk con-
ratio matches the inductive coupling coelficient of the transformer, the output verter is shown in Fig. 3-17c. ks we mentioned, the isolated version of the
. ..>, current ripple could be completely eliminated, as shown in Fig. 3-16. cuk convertex .retains all the properties of the nonisokted converter, and
Although this is a very useful circuit, its applications are limited to designs therefore, the zero ripple extension may be used successfully in this case
where no inpt=to-output isolation hequkd.-For-use ofthisrmn\wter-in - also.
an off-the-line configuration, isolation of input and output is a must. The Figure 3-18 shows the coupled inductor zero output ripple dc isolated
following discussion shows how isolation may be achieved in the Cuk con- cuk converter and its waveforms. Notice that in this circuit the transfer
verter, and the three basic steps in achieving this are shown in Fig. 3-17. capacitors CAand CBhave been inserted in the other side of their respective
The first step is to split the coupling capacitance CLinto two series ca- bansformer windings. This change does not affect the operation of the con-
pacitors CAand CB, as shown in Fig. 3-17a. Since the average dc voltage al- ' verter.
the connection point of the two capacitors is indeterminate and floating. it Akhough coupling the input and output inductors reduces the output
may be forced to zero by pladng inductor L between this point and ground, ripple, it also introduces an undesirable side dm,that is, an output polarity
PIPES OF POWER CONYEMERS 41
FIGURE 3-18 Coupled inductor zero ripple dc isolated euk converter, Diode 4 ts
the output clamp diode, holding the output to a diode drop due to polarity reversal at
start-up.
I
reversa1 at start-up. Although the reverse poIarity pulse is of short duration:

it may prwe fatal to sensitive electronic loads. The addition of a clamp diode
Dz,as shown in Fig. 3-18, limits the transient pulse to a volt or less, thus
protecting sensitive loads.
At the writing of this book, the c u k converter and its assmiated appli-
cations are protected by a series of patents, and its use in commercial or
other designs may require a Iicense agreement with its inventor.
3-7 THE BLOCKING OSCILLATOR OR RINGING CHOKE CONVERTER

The fixed frequency pulse-width modulated (PWM) power supply design
has been excIusiveIy used in modern designs, mainly due b the avdlability
of P W M mntrol circuits. On the other hand, for many low-power (approx-
imateIy 10 to 50 W), inexpensive power supplies, the variabIe frequency
FIGURE 3-17 The three bash steps of transforming a misolakd Cuk coaverter to blocking oscillator converter circuit has been successhlly implemented in
an isdsted one. b)Splitting the coupilag capacitance C1 into series capacitors C, and many designs. Figure 3-19 depicts a practical realization of a low-power
Cd (b)placlag inductor L between point a d ground; Ic) changing inductor L mto an variable frequency bIockfng oscillator circuit, with its associated waveforms,
IsoIation transfomer.
Operation of the circuit is as follows.
TYPES OF WWER CONVERTERS 4a
At the same time, flux is building up on winding N3 developing a voltage

potential VE. Therefore base current to keep the transistor in saturation is
I
now supplied through resistor Rg. The magnitude of the base current is
Since In a transistor the collector current Is a function of the base current

and the transistor gain, that is, Ic = PIB, then the transformer primary
current will increase up to the point where the a b w e product reaches max-
imum. Beyond this threshold the base circuit cannot sustain any more col-
lector current increase, resulting in transistor Q, cutoff and a transfer of
power from the transformer primaiy to the secondary and consequently after
rectification by D lto the load.
The transistcr used in thb drcuit must be so chosen as to have a minimum
collector-to-base (Vcso) rating given by
,V f (leakage spikes) (3-40)
For off-the-line power supplies, transistors with V,,, 2 800 V are commonly
used. .-
S h e the operatingfrequency of this type of converter is variable, care
must be taken to design the drcuit so that its lowest operating frequency
does not drop below 20 kHz, so that the power supply does not b b m e
audible, Remember that the frequency of operation is higher at light loads
and lower at full load, C
Since this topology is basically a flyback converter, the transformer-choke

has to be designed very carefully. To calculate. the primary turns of the
transformer, a full-load frequency of operation must be chosen by the de-
+~4 signer. Then the transistor on time must -be picked so that it will run at a
p r e h e d 50 percent duty cycle maximum. With these factors established,
FIGURE 3-19 A free nmning bloddng o d b t o r drcuit d t b Its assdated w a ~ ~
the primary t r a n s h e r inductance in microhenries is calculated as follows:
At initial power-up transistor Q1receives a certain amount of base current L -

= VhJm, t, 13-41]
through resistor R,, causing the transistor to sturate. Consequently, a peak IPP
current I p p flows in the primary winding of transformer Tiequal to where,,,V is the minimum Input voltage which the transistor will switch,
t, is the maximum translsbr on time, and I, is the peak primary
current through the transistor.
where (6-T) is the maximum transistor on time t,. Once the transformer primary Inductance is calculated, the primary number
- .. - - - - - - ... ...
. --- . .... . .
44 HIGH.FRWNCY SWRCHlNG POWER SllPPLlES TYPES OF POWER CWV€mERS 45
ofturns is found by
where A, 5 the effective srea of the core and B , Is the maximum allowable
d n g Bux density of the core.
In wder to avoid core saturation, an air gap will be used. The length of the
air gap in millimeters is given by
A practical way to alculate the secondary windings number of turns is

to determine first the tramformer primary turns per volt ratio:
Since we chose t = tg, then the secondary wlnding(s) number of turns

will be FTGUIIE 3-20 h t k d impknmtatfon of a blocklag wciUatw cmvertBr using a power
s' 8a
'(r- + (3-45) MOSFET. (Courtesy Motomla Sedconductor Products, Inc.) .-
where V, is the nominal secondary output voltage desired and V, is the
vdtage drop encountered in the conductors and the output rectifier. of Qsmodulates the gate of Q1,shortening or lengthming its on time, thus
The above calculations will result in a very good first-order approxfmation, keeping the output voltage constant.
and minor adjustments to produce exact results may be made by the designer I
durfng dcircuit implementation and testing. 3-8 THE WEWARD-TAYLOR OONVERTER

C
The following new converter topology, like the c u k converter, has the very
desirable features of-nunpulsating currents in both input and output-prts,
3-7-2 A MOSW Blocklng WZllator Converter thus significantly reducing both conducted and radiated noise, The circuit
A practical implementation of a MOSFET blocking oscillator converter using is best suited for high operating frequency applications and therefm the
a single Motomla (or equivalent) power TMOS transistor is s h o w In Fig. use of MOSFETs is inherent. The new topology is a boost converter cascaded
3-20. This design may well bz used for off-the-line switching power supplies, with a buck converter, and it may work in 08-the-line appiiations from
and its regulation is good enough For many applications. unfiltered dc inputs. The input reservoir capadtor is no longer- needed, at
I
The operation of the circuit is as follws. At power turn-on, in phase I the expense of a total lack of output bold-up time. Its inventors David
I -
transfirmer windings N1and-& d l c a u s e ~ t ~ s d l ~ - O s c i l l a t i o r n Sheppad and Briar~.Taylorclaim no-paten~restdctions-for-he.converter,
1 start when capacitor C1 is charged through a large resistor R l connected to
the supply rail. Resistor Rpis used to limit Q2collector current. The on time
ofthe oscillation cycle fs terminated by transistor Qewhich senses the ramped
i and it is therefore free for unrestricted use. It is the belief of the author of-
this book that the converter has many merits and features, many yet unex-
plored, and it is especdly suited in tday's hl&-frequency applications
source current of QI.Capacitor CI is therefoe charged on alternate- . half ( a h r e 100 kHz). Readers, power supply designers, and students involved
cycles by transistor Q2and folward biased zener Zg. in converter research are therefore encouraged to explore this topology for
Regulation is provided by taking the rectifier output of sense winding N3 b t t e r understanding and optimum usage. The name "Sheppard-Taylor con-
and applying it as a bias to the base of Q2via zener diode 2,.The colector verter" has been arbitrarily coined by &is author.
46 HIGH-FREWENCY SWIICHING POWER SUPPtES WPES OF POWER CONVERTERS 47
3-84 Circuit Analysis of the ~heppard--or Converter output current is now sustained through the free-wheel diode Dg. In order
to derive certain equations to facilitate components rating and selection and
The basic circuit schematic of the Shepp&-Taylor anverter is presented relate inputloutput parameters, the c h i t s of Fig. 3-22 will be used.
in Fig. 3-21, in ia nonisolated version. Operation of the circuit is as follow. From Fig. 3-&4 Q, and Q, are on during p e r i d t,, and off during period
Assume that MOSFETs Q, and Q, are oE, Then an input current li, will td. The duty cycle 6 is
Row through L1,D,, C, and 4 to the negative rail. This current Row will
charge the nonpolarid type capacitor C, so that the plate connected to the 8=f ' (3-46)
cathode of Dl is at an initial potential of approximately V,, while the opposite f, + m
plate is clamped by D, to the negative rail. No current is Raving to the
As previously mentioned, the average value of the voltage across inductor
output of the converter due to D,. Note that as C continuer to charge, the
current in Lldiminishes linearly. When Ql and Q2 are witched on, diodes
4 will be the sum of V, and V,.
Dl and D, d l 1 be reverse-biased, prohibiting any further current Row from
From Fig. 3-22a then, applying Kirchhofs voltage law, during period t,
the input terminals. At that instant, capacitor C is in essence connected v,, = v, - I-Vd
acmss the output terminals of the converter, and becaure of its charged or
. D2- will be f o d - b i a s e d , allowing output current to flow through
mlaritv,
Ls, D3,Q J C, ~ and Qe.
By inspfftion we can see that the voltage acmss L,'is proportional to the and
sum of V,.and V,, the voltage across capadtor C when Q and Q2are off.
The moment Q, and Q, are switched on, the diminishing current in Ll
changes and begins to m p up, storing new energy in the inductor Ll.
During the next cycle when Q,and Q2 are once again turned d capacitor Simrlarly, during the same period the average value of the voltage across Lg
C will start charging, fmm the energy stored in Ll, through diodes Dland will I
X
Dp In attempting to b h c e its volt-second pmduar, 4 will cause C to
change to some new voltage, which is inversely proportional to the mn-
dudion duty cycle of Q and Qz.Meanwhile D3is apainxvene-biased, and and C
From Fig. 3-22c, let I,, and I,,, be the peak-to-peak values of I,, and
I, respectively, during the on period 5. Then by integrating Eqs. 3-48
and 3-49, we get
and
1," -'t (Vc - Vm,)

r, (3-52)
During the of period fa, assuming the peak-to-peak values of I , and IWtto
be and Iwtd,respectively, we can show that
and i
If k a s s u m e that rluuring the on period t,, the input voltage Vh, and
output load R of the converter remain constant, then the charge extracted
h m C during this period will be prqmrtional to the "dmp" in voltage
m s s C. This droop in voltage may be written 9s AV,, and the charge
quation will be
(AVJC = l,tp
where I, = L -t IM.
It can also be shown that
-VL
- - -.1, .- -
RV,,
Therefore
Equally, Ifwe assume durlngthe off period td,that V, andR remain corrstant,
then the voltage ramp across C mn be regarded as having the same hag-
nitude as the drop in voltage and can therefore be regarded also as AVC. I
Hence, 4 .
(AVJC = 1, (3-581
and
Since we assumed that V, and R remain constant during both t, and ta, it
blluws that thevolt-second products across L,are equal. Therefore, equating
Eqs. 3-51.and 3-53, we get
t,(vh-cn=tdK--V;d -v-)
Solving for V,,
e b 4
(F) -- *
FIGURE 3-22 C i d t a d d - t h e Sbppard-Taylor m F t s r . (a) Current
paths wben MOSFETs nrs on, (b)current paths whea MOSFETs are d,and where 6 is the duty cycle defined as .
(c)~~~tp~tcurreatwwefwmsduringtheondipe&.
M S OF POWER CONVERTERS H
Similarly, Eqs. 352 and 3-54 can be equated to give the classic buck regulator and
output voltage relationship:
V,, = V,S (3-621
318-2 Fnatures of the Shmppard-Taylor ConvmMr

As previously mentioned, S&I converter is ideal for high-frequency appli-
The above equation shows an unusual and unique characteristic of the new cations (above 100 HZ). At these frequencies MOSFETs are highly rec-
converter, that is, the output voltage approaches infinity as the duty cycle ommended. Even if 'bipolars were to be considered, examination of Eq.
approaches 50 percent! This characteristic gives the converter some ex- 3-61shows clearly that during an overcurrent condition, the conduction time
tremely interesting and unique features, the most Important being the ability of the power switches is small; therefore, the inherent storage times of
of the converter to hold output regulation with very low input voltage values, bipolars do not provide a control rarige broad enough to be usable. At higher
as shown by Eq. 3-63. frequencies the value of the switching capacitor C is also minimized, a
The discussion so far has centered around the nonisolated version of the desmble effect in redudng both size and weight and also making the value
of C practical. We mentioned before that this converter also holds output
converter. An isolated version of the converter may easily be implemented
as shown in Fig. 3-23. In the isolated version, D3 is replaced by a pair of regulation within a wide input voltage~~ a very desirable feature
particularly for of-the-line switching power supplies enabling the converter
kctihers DM and DS8, and an isolation high-fresuency transformer is In-
serted between this pair of rectifiers. to operate worldwide without modification of the input circuitry. Although
Let the turns ratlo of the transformer be N, where N = N,l N,. Then the converter may, for off-the-line applications, operate from gnfiltered dc
Eqs. 3-52, 3-57, and 3-63 should be modified to read, respectively, inputs, the absence of the input reservoir capadtor may not be desired if a
minimum output hold-up time is required. A reservoir capacitor can be used
to increase hold-up time, and for a given value of capacitance, the optput
hold-up time will increase by a factor of 2 over conventional topologies.
Since the input and output currents of the converter are nonpulsating as
shown in Fig. 3-22c, ac line or dc output filtering twmeet local, nationalr
or international noise requirements, such as FCC or VDE, is greatly sim-
plified, or unnecessary.
Snubbers across the MOSFETs and the power t r a n s h e r are not nec-
e;sary using this topology, greatly reducing component count and improving
efficiency.
Also, perfect synchronization of switching Qland Q2on and of is not
essential. Lack of synchronization does not lead to malfunction or to failure
of any device or component.
High-frequency resonant converters, whether parallel or series, have be-

come increasingly popular among power supply designers, especially at fie-
quendes sbove 100 kHz, because they &r small size, good reliability, and
reduced EMIIRFI. Recent advances and price reductions in both control
ICs and power MOSFETs have made resonant sine wave converters even
more popular. Series resonant topologies are being widely used because of
I OF POWER CONVERTERS M
their tolerance to swftching transistor transition times and reverse recovery

times and their ease of operation. Some of the advantages of the reso~lant
sine wave converter, compred to other mnventfond topologies, follow:
With this converter, there is higher overall efficiencyat a given power level,
mainly due to the absence of switching losses at the power switch and the
rectifiers. Lower losses in turn mean smaller heat sinks, hence reduction in
size and weight of the w e d package. Because the voltage is switched when
the drain current of the MOSFET switching transistors is zero, operation
at higher frequencies is possible, resulting in smaller magnetic elements and
filter components. Since the currents in the series resonant converter are
sinusoida in nature, the .absence of high dild current changes, assodated
with square wave converters, means low levels of E MI) RFI emissions. In
the following paragraphs, a series resonant converter will be discussed, and
the basic design parameters will be given.
3-9-1 The Baslc Slne Wave Wries R m m t Converter

A series resonant power stage and its associated waveforms is shown in Fig.
3-24. Operation of the circuit is as follows.
Assume that MOSFETs Q,and Qe are offand.that capacitor CAis mm-
pletely h h a r g e d . Drive pulse t, is applied at the gate of Q,,turning it on.
Current 6,is thus flowing throughresonant inductor LI,the primary winding
of transformer T,,and the resunant capadtor CI. Since the voltage across
the transformer primary is 6xed by its turns ratio N and the output voltage,
and because the resonant current is amtroUed by the series resonant
network L C B ,it increases fn a sinumidal manner starting h m zero,charg-
ing resonant-capacitor C Rand delivering energy through transformer T2to
the load. when the peak of the sine wave current is reached, capacitor C R
is clamped by diode Dlto the positive rail. At that point the voltage across I FIGCrRE 3.28 A high-frequency sine wave series remaat m v e r k r a d tts
inductor L,,ceases to increase, and its energy is released through T2to the u d a t 6 d wadorms.
load causing the current k to-decrease toward zero. Transistor Q1remains
on until the current in-the resonant network reaches zero, at which point 8-9-2 Transistor Sdecflon InSeries Resonant Converter
Q is turned on and the above cycle repeatr itself but in the opposite di-
rection, since the energy Is now drawn from the previously chargeksoaant- . The maximum current In a series resonant converter occurs at high input
capacitor CR.Thus the composite rcsormt sinusoida current iH is developed, line. Since this current also flows through the switchfng transistors, MOS-
as shown in Fig. 3-24, which pwduces a proportiond sinusoidal primary FETS Q and Q2must be selected to handle a minimum current of
voltage. The output of the center-tapped transformer secondary voltage is
then rectifiedand filtered by capacitor C , to produce the dc output voltage
v,. where I h is the peak currerrt at low line, given by.
Note that the leakage inductance of the transformer is in series with the
resonant inductor, and it should be taken into account when designing the
power stage.
I
TYPES OF POWER CONVERTERS 55
54 HIGH-FREQUENCY SWITCHING POWER SUPPLIES I
d
and ifi is the transformer primary current given by Calculate the energy storage capability of the core given by
The blocking voltage of the switching transistors used must have a value where B is the chosen core operating flux density in gauss. (A good starting
greater than the maximum supply voltage V,,,,. point is B = B,12, and A, is the core effective area in square cen-
timeters.)
Since
3-93 Power Transformer Design
H1, = NI, (3-76)
The design of the power transformer in the series resonant converter follows
classical design procedures, A starting point, however, is to determine the solving for N, we get the required resonant inductor number of turns.
primary to secondary turns ratio given by
In order to prevent inductor core saturation, an air gap must be introduced

in the magnetic path. The value of the air gap i s given by
3-9-4 Design of the Series Resonant Ne-k SC,

The characteristic impedance of the series resonant network as a function where p is the permeability of the chosen core. The gap length is expressed
of input voltage and output power is given by the following equation: in millimeters,
where t~ is the eficiency of the: converter. EXAMPLE 3-3 I
Then the resonant capacitor is Calculate the series resonant network LAC,of a 200-W converter operatin?
at 200 kHz.Assume an eficiency of 80 percent and an input voltage range
of 90 to 130 V ac,
where f is the frequency of operation of the converter.

The resonant inductor is SOLUTION
Solving for minimum and maximum dc input voltage,
V,. = 1.4 X W = 126 V dc
V,,,, = 1.4 X 130 = 182 V dc
3-9-5 Design of me Resonant Inductor

Since the output power is given as Po,, = 200 W and efficiency q = 0.8,
then P,, = 20010.8 = 250 W.
First calculate the maximum circuit energy storage, using Eq. 3-74 Average primary current from Eq. 3-69 is
w,, 1
= -L,Ik
2
I , = - - 2PI. - -2 (-2 W - *
Vin.min 126
50 HIGH-FREQUENCY SWRCHING POWER SUPWE
Therefore
Since the series resonant converter is operational at high frequencies,
extra care must be exercised in selecting components. Litz wire is recom-
mended for the inductor and power transformer to minimize skin effects.
Capacitors must be selected for low ESR and ESL and good ripple current
and
rating. Polypropylene-type capacitors are a good choice at these frequen-
cies.
Or, the other hand, the output rectifier diodes need not be extremely
fast, compared to other converter topologies, due to low dildt during diode
Hence the MOSFETs selected must handle a minimum of 11.86 A of drain turn-off, inherent in the resonant converter topologies,
current.
From Eq. 3.71 we get
3-40 CURREM-MODE REGUIATED CONVERlERS

The current-mode segulated converter differs significantly from conventional
Hence PWM converters. Current-mode converters use an inner loop to directly
control peak inductor current with the error signal, rather than controlling
duty ratio of the pulse-width modulator. Figure 3-25 shows the block diagram
of a fixed frequency current-mode forward converter.
and As shown, the error amplifier compares the output to a fixed reference
voltage, and the resulting e m r signal V,is controlling peak {bitch current,
which is proportional to the average output inductor current.
There are serefal inherent advantages to the current-mode c o n t ~ mn-
l
Since the maximum energy storage in the inductor occurs at high line, the verter when compared to the conventional PWM converter. These advan-
maximum circuit energy storage required is calculated from Eq. 3-74, tages are as follows: I
0
1. Automatic feed-fornard improves line regulation.

2. Automatic symmetry correction, due to peak current sensing, makes this
From Eq. 3-75, and using a ferrite core at 1550 G having an A, = 0.9,
scheme suitable to push-pull converters, without the need for complex
balance schemes.
3. Automatic current-limiting control, due to the sampling technique used.
Using Eq. 3-77, the required resonant inductor turns may be calculated as 4. Simple loop compensation.
follows:
5. lmproved transient response.
; N = -Hi,
=- 67 = 6 turns
6. Outputs of more than one converter may be easily ~aralleledwhile main-
I,, 11.86
taining equal current sharing.
The inductor gap required, using Eq. 3-78, is
-
Of course, there are dso certain limitations to this scheme, the most
important being the following:
1. Loop instability above 50 percent duty cycle.
I 58 HtGH-FMQUEf4CY aWCHING POWER SUPPLIES
frequencies up to 500 kHz. Chapter 7 describes the UC1846 current-mode

control IC.
Two other Unitrode IC controller families-the UC1823 and U C 1 8 2 S
may be configured either in a voltage or current-mode topology, with o p
erating frequencies up to 2 MHz, The reader is referred to Unitrode Cor-
poration's "Linear Integrated Circuits Databook'" for specification details and
applicdion information for these versatile FWM-integrated circuits.
Clock
Error n 3-11 THE WARD WMVERTER
The Ward converter is a new dc-dc converter, and the description which
follows is the first to appear in any literature ever, The converter was in-
vented by Dr. Michael Vlahos Ward, president of Combustion Electromag-
netic~,Inc., Arlington, Massachusetts, and a patent is pending on the circuit.
Although basic research has been completed by the inventor and a practical
working prototype for a specific application has been developed, this con-
verter may have many desirable features still unexplored. It is the belief of
the author that this converter has many merits and that it is well suited for
high-frequency operation. The fundamental reason that it is suitable for high-
frequency operation is that the converter is a member of the family of the
so-called zero-current-switching converters. This means that its single con-
trolled switch is turned on at zero current and is turned of when the current
in it (or in an external antiparallel diode) flows in the reverse direction. In
the Ward converter, as in any other zero-current-switching converter, the
switching losses are very small, and the circuit can probably operate efi-
ciently up to several megahertz. Further understanding and research on thfs
topology are therefore encouraged. Figure 3-26 shows the converter and its
La'ch associated wavefoms.
output The converter operates by storing the input energy sequentially, Erst
FIGURE 3-25 A current-mode fornard converter and its assodated wavefoms. inductively and then capacitively; inductively the time that energy is being
delivered to the output circuit, and capacitively for the remaining period.
With reference to Fig, 3-26, when the MOSFET transistor switch Q is closed,
2. Less than ideal loop response, caused by peak instead of average inductor
energy storage inductor L,begins to build up current II. At the same time
current sensing,
capacitor C1, which was fully charged under steady-state conditions, dis-
3, Tendency toward subharmonic oscillation. charges its energy to poduce the first-half-sinusoidal input current 11,and
4, Noise sensitivity, particularly at very small inductor ripple current, by transformer action this current is transferred to the output, and through
rectifier D2it charges the output capacitor C,,,.When switch Q is opened,
However, with careful design the above problems may be seduced or I inductive current I , is steered to capacitor C1to charge up the capacitor, in
eliminated, making the use of the-current-mode control technique a useful preparation for the next cycle when most of the energy is delivered to the
and attractive alternative for high-frequency switching power supply design. output capacitor. In this second period, capacitor C, completes the second
Unitrode Corporation is marketing the UG1846 and UC 1847 families of half of its discharge cycle with the current I,,flowing through diode Dl across
current-mode PWM-integrated controllers, which offer all the necessary the MOSFET switch, which is then followed by a period when diode DLis
functions to design high-pedonnance, low-cost power supplies, at working back biased and voltage Vk recovers as shown in the last waveform.
TYPES OF WWER CONVERTERS bq
Inspection of the waveforms shows that during the energy storage phase
when Q is on, a negative charge is placed on the output capacitor C2, which
must be dealt with. The proper handling of this "wrong polarity" charge was
the key to the final success of this converter. A prekise phasing relationship
was determined so that during the energy transfer phase, this negative charge
is fed back into the input circuit so as to transfer its charge to capacitor C,
while simultaneously helping define zero (negative) current at the instant
the switch Q is opened, In this way, even under conditions where all the
energy on C, is transferred to output capacitor C, and almost no reverse
current Ig exists to cancel current I , , primary current I ; , which is the trans-
form'ed secondary current I:, will guarantee zero current crossing during
MOSFET transistor tum-off.
3-44-1 CErwN Analysis of the Ward Conve~er

There are some simple relationships which define the circuit parameters
making up the Ward converter,
The relationship between the input capacitor CIand output capacitor C,,
is somewhat arbitrary and is determined by the requirements ofthe system
and is related to the transformer turns ratio N as follows:
The parameter I typically represents the number of cycles of operation of

the converter required to fully charge C,. Clearly, the larger I is, the
higher the frequency of operation of the converter and the Iess sensitive t
h
converter is to a shorted output.
If an output voltage V,,, is required, then the ideal turns ratio of the
transformer will be
where Vl is the voltage to which the energy storage capacitor C, is charged

at the end of the charge cycle (at time tZ), and whose amplitude is
given by
r 1
- -
where IL, is the maximum charging current through inductor Llat the instant
MOSFET Q is turned off, given by
FIGURE 3-26 Tbe Ward Mmerter and its msodated waveforms.
62 HIWFREWENCY SWITCHING POWER SUPPLIES
where lk is the dc current which exists as a consequence of setting the time Another parameter depending on p is defined as v and is given by
period t2short of the value needed to just fully charge capacitor C,.
Thus the time period t1 is set to be slightly longer that the half cycle
energy discharge time t3, given by
which is required for defining the converter cycle period T.This expression
t = n q (3-83) then determines a value for Ll and helps complete the specification of the
where L,is the primary leakage inductance @.en by converter.
The half-cycle period t is given by
The term I, in Eq. 3-81 is the maximum current which would flow in the
series circuit made of V*,,4,and C1 and the primary of the transformer and
when the circuit is connected with zero initial current in LIand zero initial t4 = t 3 4
voltage on C1, that is, which is chosen to be slightly greater than one-half t3, which is the correct
lm vin
=- (3-85)
choice as can be seen by inspection of the waveforms of Fig. 3-26, For
z example, we can take t, = 0.68 v = 0,36, and y = 0.56.
and
7
3111-3 Features of the Ward Converter
The Ward converter was developed to satisfy a certain application need,
which conventional topologies could not satisfy. The comerfer was used in
for &ILL is much less than 1. a boost configurntion, and the theoretically predicted waveforms presented
here were reproduced by actual practical measurements.
Inherently, the converter has the Feature of sustaining a shorked'output
3-14-2 Deslgn Procedure fnr the Ward Converter without additional protection, making the converker useful in situations
In order to design a particular converter, one specifies the desired output where capacitive loads are present. The converter also does its switchinet
power Pd, defined ideally by zero current and therefore prcduces very little EM1 and requires no snubber
networks. The practical operating efficiency of this converker is also high, in
the order of 85 to 90 percent. Another desirable feature of the converter is
its ability to operate over a wide input voltage range.
where q 5 the converter dciency and f is the operating frequency. As mentioned, this converter topology has not been fully analyzed; there-
h e , its potential usefulness has not been fully uncovered. Several practical
Select CIand V1, choose the operating frequency line.,define the switch- aspects of the converter require further investigation, e.g., control-to-output
ing period TI and the t3, which initially Is assumed at one-half of T. A first transfer function, dynamic behavior, device stresses, and start-up transient.
approximation of the transformer parameters may be obtained using Eq. The inventor and this author believe that this is an intriguing topology,
3-80 and referring to Chap. 5. which brings new ideas and opens new horizons for further research by
Typically, V,will be in the range of two to six times V,,, so adjustments individuals who are dedicated to the power-conversion field. Such efforts
in the design may be necessary to bring it to the desired level of operation. are encouraged, and the reader may wish to obtain a copy af the patent or
A critical factor in the successful operation of the Ward converter b the contact the inventor for a more detailed description.-. This author wishes to
placement and precise selection of the capacitor C2across the output of the thank Dr. Michael Ward for his contributfon.
transformer. C2 is given by the equation
REFERENCES
1. Bloom, G,,and B. Severn~:"Unusual DC-DC Power Conversion Systems," MIDCON.
where 0.5 5 p 1 but closer to 0.5. 1980. .. ,
& t'dGH-FREQUEKY SWlTCHlNG POWER SUPPLlES
L Chryssis, G.: Power Supprim A Switching Alternative, Design N m , May 21, 1979.
3. Glen, D.: Self-Oscirlating, Flyback Sdtching Converter, "Motomra TMOS P m r FET
FOUR
Design Ideas.'Xandbmk, 1%. THE POWER TRANSISTOR
4. Hnatek, E. R.: "Design of Solid State Power Supplies," &d.,
New York, 1981.
I Van Nostrand Reinhold,
IN CONVERTER DESIGN
5. Jar], R. B.. and D. R. &TIP: Application Note AN-6743, RCA, 1918.
6. Jones, D. v: "A New Hesonant Converter Topology," BFPC Proceedings, April 1987.
7. Lee, F. C.: "Zero Voltage S w i t d n g Techniques in DC-DC Converter Circuits,'%FPC
Proceedings, April 1887.
8. Middlebrook, R. D., and 5 , cuk: "Advances in Switched Mode Power Converters," vols.
1 and 2. Teslam, Pasadem Calif, 198 I.
9. Motorola, ine.: "LineariSwitchrnode Voltage Regulator Bandbwk," 3d ed, 1987.
10. Pressman, A. I.: "Switching and Linear Power Supply, Power Converter Design," Hayden,
Rochelle Park, N.J., 1977.
11. Shepprd, D. I.. and B. E. Taylor: "A New Converter Topology Imparts Non-Pulsating
Currents w Input and Output Power Lines." Powerconvefiion international, October 1984. 6 0 INTRODUCTION
1R Suva, R.,and R. 1. Bawr: "Engineering Bulletin E B G . EB-86, EBd7, EB-88," Motoroh, In the block diagram of Fig. 1-1, depicting an 08-the-line switching power
1979. supply, we show the block where conversion takes place containing what
13. Unitrode Corp.: ''Linear integrated Circuits Databk,'>1987. we called the switching element. Various types of switching elements, such
as the transistor, the SCR, and the GTO, have been used by power supply
14. Vellhooven, C. Van.: "'Propertk of DC-to-DC C o n w r t m Tor Switched Mode Power
Supplies," A m p e m Electmnic Cwp., 1975. designers over the years. By far the most popular and most often used
element has been the bipolar transistor and in recent years its MOSFET
E. Ward, M.: "DC-~PDC Converter Current Pump,'U.S. Patent Application No. 719790.
counterpart. In this chapter we will discuss both types of transistors and the ,
I
September I%.
GTO, their characteristics, and their use in switch-mode power supplies.Z
16. Won& T.: Compact Power Unit Raism Supplyi Switching Rate by an Order of Magnitude,
EIecsrafc DesIgn, April 1985.
17. Wod, P. N.: "Design Considedons Tor Transistor Convertefi," TRW Power Semicon-
i
ductors, 1977.
44 TRAHSISTOR SELECTION
18. - ."Switching Power Converters," Van N o s h n d Reinhold, New York. 1981. The basic design parameters of a transistor to be used in an off-the-line
converter are first its voltage blocking capability at t u r n 4 and second its
current carrying capacity at turn-on. Both these parameters are determined
by the type of converter in which the transistor will be used. In Chap. 3
we described design equations and criteria for selecting the appropriate
devices.
1 Another important decision which the designer has to face is whether to
use bipolar transistors or MOSFETs in the design. Each of these transistors
ofers distinct advances over the other, the bipolar being less expensive
I at present, while the MOSmT offers emit simplicity because of simpler
I drive circuits.
i It h important to note also thatthe bipolar transistor has a limited working
frequency cutoE which is around 50 kHz, while the MOSFET may be used
in switching frequencies up to 200 kHz. Of course higher frequencies mean
smaller component size and therefore more compact power supplies, a fact Figure 4-L shows the V-1 characteristic of a typical bipolar transistor.
which seems to be the trend in today's power supply designs. Close examination of these curves shows that the saturation region of the
V-I curve is of interest when the transistor is used in a switching mode. At
that region a certain base current can switch the transistor on, allowing a
4-2 TH€ BIPQUR PgWER TRANSISTOR USED AS A SWITCH large amount of collector current to flow, while the collector-to-emitter
voltage remains relatively small.
The bipolar transistor i; essentially a current-driven device; that is, by in- In actual switching application a base drive current is needed to turn the
jecting a current into the base t e r m i d a flow of current is produced in the transistor on, while a base current of reverse polarity is needed to switch
collector. The amount of collector current flow is dependent upon the gain the transistor back off. Since the transistor is a real-world element, it is f
a
(beta) of the transistor, and the following relation holds true: from ideal, and certain delays and storage times are associated with its
operation.
In the following section are some definitions for a discrete bipolar tran-
sistor driven by a step function into a resistive load.
where Ic is the collector current (in amperes) and l Bis the base current (in
amperes).
There are essentially two modes of operation in a bipolar transistor: the
4-3 SWITCH TIMES DEFINITIONS OF B l W l A R
linear and saturating modes, The linear mode is used when amplification TrtLINSlSTORS [RESISTIVE 'LOAD]
is needed, while the saturating mode is used to switch the transistor either
Figure 4-2illustrates the base-to-emitter and collector-to-emitter wave forms
on or off.
of a bipolar NPN transistor driven into a resistive load by a base current
V~~
FIGURE 4-1 Typical output characterfstEcS and symbols of the bipolar
A r L 4 tm I+
Wansfstor. FIGURE 4 4 Bipolar transistor swftehing waveforms.
THE POWER TRANSlSTOR IN CONVERTER DESIGN 7*
TO HIGH-FREQUENCY SWRCHlNG POWER SUPPLIES
Darlfngton circuits may be implemented using discrete components, al-

though monolithic Darlingtons are available from a number of power tran-
sistor manufacturers.
4 4 BASE DRIVE CIRCUIT TLCHNIaWS K)R BlPOLkR TRANSISTORS

4-64 Constant Drive Curren9 Circuits
In the previous paragraphs we examined the turn-on and turn-off limitations
of the bipolar transistor when it is used as a power switch. It is now clear
that in order to minimize the saturation losses, adequate forward base drive
IB1 is required, while to minimize storage time and transistor switching
losses, suRcient reverse base drive In?, is needed.
We should note here that while I,, is increased and the storage and Fall
times decrease, the emitter-to-base reverse bias voltage VEBis also increased.
This means that the reverse-bias secondary breakdown energy, known as
ES8,is also decreased, and if care is not taken in designing the reverse drive
circuit, the switching transistor may be destroyed by going into a s e c o n d q
breakdown mode. In Sec. 47-2we will discuss the significance of Eggand
the phenomenon of secondary breakdown in bipolar transistprs. Here it is
sufficient to say that for practical purposes the reverse base drive circuit
must have a low source impedance; that is, it must provide high ln2and low
FIGURE 4 4 (a) AntisaturMon
dides (Baker damp) preventing VEB - I
transistor Q, from saturaw, (b) The manufacturer"^ data sheets normally provide information on the limits
the use of 1Dadington d A t to of the reverse emitter-to-base bias voltage. Practid design circuits use a
keep Q,out of saturation. V,, bom -2 to - 5 V. Higher reverse base voltage means a reductiondn
storage time delay because, by allowing fewer miers to be neutralized by
recombination, it takes a shorter period of time to remove the stored charge.
A popular base drive circuit used with a floating switching transistor is
shown in Fig. 4-5 along with its waveforms. Operation of the circuit is as
follows. When a positive secondary pulse V, appears across the transformer,
ibrward base current drive Igl flows into the base ofQl, turning the transistor
on. Resistor R1 limits this current within a predetermined value. This vdue
of base current is determined using a forced gain ratio, which in practical
circuits is between 8 and 10. Since the collector current may be easily
Diode D, is a "wrap-around"' type, which is used to pull the reverse base calculated from the output power and type of converter used, the base
current at transistor turn-&, d i s c b n g the base-emitter capacitance and current may therefore be predetermined using Eq. 4-1.
hence decreasing storage time. This positive drive pulse also charges capacitor C very rapidly. The charg-
Figure 4-4b shows a Darlington connection, which works in basically the ing voltage across the capacitor is -. .
same manner as described previously, with transistor Q1preventing QZfrom
fully saturating. An important point here is that Q,must turn of before Q2
clan begin to turn off. Diode Dlprovides a low impedance path for the reverse where Vs = amplitude of transformer secondary voltage
base current of Q, during turn-off. Resistors R1 and R, are low-ohm resistors V,, = saturating baseemitter voltage of Ql
and provide a leakage current path for Q Land Q2. V, = forward-bias voltage of diode D
.-
If we assume VBE= V, = 0.8 V, then Eq. 4-2 becomes
Vc = V3 - 1.6 (4-3)
When the b-ansforrrier primary goes to zero, the transformer secondary also
goes to zero, Now capacitor C, which is fully charged, will fomard-bias the
base of Q,, turning it on and consequently pulling the base of Q1 to the
negative potential.
With the capacitor now connected m s s the base-emitter junction of Q,,
a large reverse base current IBkwill develop. The magnitude of this current
fs determined by the capacitor and circuit resfstanes and the characteristics
of transistors Q , and Q,.
Another base drive circuit technique which has proved very effective in
off-the-line switching power supplies is shown in Fig. 4.6. This circuit has
the distinct advantage of providing adequate I,, turn-off drive while using
a minimum number of components. Circuit operation is as follows, When
transistor Q1is turned on, the base drive transformer primary is connected
to the supply voltage,,V storing energy in the transformer and also inducing
a primary voltage pulse V,,. This voltage pulse is transformer-coupled into
the secondary. Since the transformer winding polarities are the same, a
positive secondary voltage pulse Vs is generated, fuming transistor Q, on.
The resistor A l is a base current limit resistor, designed ti, allow enough
IBl drive to force Qzon, without overdriving and oversaturating it. Resistor
VBE, sat
R, provides a leakage current path for the base-emitter junction d t h e ,switch,
and it normally has a low ohm value, between 50 and 100 R.
YB When transistor Q is switched off, the energy stored in the tmsformer ,
is returned to Vcc by the tertiary winding and diode Dl.Since the polarity
of this winding is opposite to the primary winding, a voltage pulse V,, of
reverse polarity is mdu-d, coupling that negative pulse into the secondary
and thus producing the desired reverse drive IBPcurrent flow.
When designing the base drive transformer, the primary-to-secondary
turns ratio must be chosen not to exceed the transistor Q, published spec-
ifications for Vm and VEB.Normally the primary and tertiary winding turns
are the same.
It should also be noted that the primary and tertiary windings must be
tightly wound (inen,bifilar) to avoid excess voltage spikes due to leakage
inductance. Transistor Q1must be chosen to withstand a collector turn-off
voltage of at least 2Vccn A modlfiation of the above circuit will make the
transformer even simpler, while retaining the advantages described previ-
ously. A practical implementation is shown in Fig. 4 7 ,
If a positive pulse V, is applied to theprimary winding of the base drive
FIGURE 4-5 A base drive Circuit ushg an isohdon trausformer transformer, a secondary voltage puke V, is induced, turning transistor Q1
driving tbe p e r d t c b Q,on and off. Typical switching waveforms on. During the duration of the on pulse the forward drive IB1charges ca-
of vdtages and currents asrociated with the power switch are also pacitor C with the polarity shown in Fig. 4-7. The voltage across the capacitor
710 HIGH-FREQUENCY SWITCHING POWER SUPWE5
FIGURE 4-7 A base drive circuit using a simple isolation transformer to '
generate the transism on pulse. Turn-oddrive 5 generated by the negative
charge of capacitor C.
Because of the simplicity of this scheme it may also be used to drive

direct-coupled power transistors, as shown in Fig. 4-8. An emitter-followes
comprised of transistors Q 1and Q2alternately switches the base of Qjto
Vcc or ground, thus turning it on or off. When Q, is on, transistor switch
Q3turns on, also charging capacitor C as shown, Zener diode 2, limits the
charge on C (in practical circuits VZ = 4.3 V) and also provides a path fbr
the fonvard base drive current I , L, which is set by resistor R1.With transistor
Q 1off and Qzon, the charged capacitor C is effectively connected to the
KICURE 4-6 h transformer-coupled base drive cirmft, which uses the energ). base-emitter of Q,,causing reverse current IBOto flow bemuse of the polarity
stored in the transformer to generate the reverse basedrive through the tertiary
winding. of the capacitor, The value of I I S depends on the gain of Qa, the value and
therefore the charge on capacitor C, and the circuit impedances.
is damped to about 3 V, determined by the forward voltage drops of diodes
DL, .Dp,and D,,Instead of diodes a Zener of appropriate voltage rating could 4-64 A Propottlonal Base Drive Circuit
be used. When the primary voltage goes to zero, the tsansformer secondar, All the base drive circuits described in the previous section provide a constant
voltage will also be zero. At this time the positive terminal of capacitor C drive current to the transistor switch. These-circuits have the drawback that
is at the transistor Q1emitter potentid, Thus the charged capacitor is now at low collector current, the transistor storage time may not be shortened
effectively connected across the base-emitter junction of the switch, pro- suficiently and effectively because of a change in transistor 8.
ducing the necessary reverse base drive Ig2 to turn the transistor off and to Using the proportional base drive technique we can control the value of
reduce its storage time. 8 and in fact we can keep it constant for all collector currents. Therefore,
.. ..
76 HIGH-FREQUENCY S W K H I N G POWER SUPPLIES M POWER TRAFLSlSMR IN CONWMER DESlGN 77
FIGURE 4-8 A capacitor-rwphl direct drive base cir.

cuk Fomard md reverse hase drive currents are gen-
erated by the charged capacitor C. Zener 2, i s used to
clamp the vdtage across the capacitor to a predetermined
Iwea.
in this type of drive, one expects shorter storage times, at low collector
currents, than those achieved by the constant drive current techniques. FIGURE 4-9 A proportional base drive circuit witb base curmt and collector
Figure 4-9 shows a possible way of realizing a proportional base drive current wawfonns sbown.
circuit. The circuit operates as follows. When transistor Q1turns on, trans-
former TI is in negative saturation and transistor Q?is of. In the time interval
where Q1is on, a current flowsthrough windingN1, limited by series resistor offby turning Q1on. The following equations may be used to calculate the
R, storing energy in the winding and also keeping T I in saturation. When transformer turns ratio. Using a forced constant fi value for Q, we have
transistor Q1turns off, the energy stored in N,is transferred to winding N+,
causing base current to flow in Q2,turning it on. With transistor Q2 on,
collector current Ic flows, energizing transformer winding NZ.Therefore all
the dotted ends of transformer Tl become positive, pulling the core from During transformer operation the flux density at tm must be equal to the A
negative to positive saturation. flux density at t d

Since windings N2and N, are now acting as a current transformer, tran-
sistor Q, stays on, forcing a constant at all collector current levels. Q2turns A m m = Am, (4-3
78 HGH-FREQUENCY SWnCHING POWER SUPPLiES
and
A@ 2BmBIAc
Ei
where B , is the maximum operating flux density in gauss, and Ac is the

core a m (in square centimeters).
From the fundainental magnetic equation we get
Combining and equating Eqs. 4-6 and 4-7 we can write for transformer
windings N1 and N, at 50% duty cycle maximum
where VBEis the base emitter voltage of transistor Q2,and Jis the converter
operating frequency (in kilohertz).
An expression of the turns ratio NI/N4may also be derived by dividing Eqs.
4-8 and 4-9.
NI = --
I
Vcc tor
N4 ~ B ton
E
4-63 An Atternatlve Propordfonal Base Drbe Clrcufl

The circuit depicted in Fig. 4-10 is an alternative proportional base drive
circuit, The author has used this circuit in real applications and has kiund
it to perform very well, when arehily designed.
Operation of the circuit is as follows: A positive control pulse at the base
of transistor Q, turns it on. Because of the polarity of the base drive trans-
former windings, transistor Q9is off. At the same time magnetizing current
Id is building up in N d , approaching a steady state. The final value of this
cumnt will be
When Q1 turns off, ldceases, and because of the T,transformer windings FICVRE 4-10 Proportional base drive circuit and its a r
sociated wawfonns.
polarity, energy is transferred to Qp base winding Nb, inducing a current 16
to flow, turning Q2on.
80 HIGH-FREWNCY SWITCHIMG POWER SUPPUES
THE POWER TRAMSlSTOR IN CONMRTER DEStGN 81
Hence, current I, is now flowing through winding N,. This current flow
causes a regenerative increase in base drive to Q2,since windings Nb and
N,act as a current transformer, until Q2is switched fully on. The final value
of I, induces a proportional base current given by the equation
where Vbbis the Q, turn-d base drive source voltage at maximum I,.
While transistor Q2is on and Q,is off, capacitor C charges through resistor
R to the supply voltage Vdd.When Q1turns on again, capacitor C applies
its charge to winding Nd,driving the voltage at the base of Q, sharply
negative, thus providing the reverse base drive turn-off current Ibs, rapidly
turning off transistor QZ. The energy which the capacitor C has to provide
to turn off Q, is given by the equation
A
and
FIGURE 4-11 An i m p m d propodml base drive circuit.
where tf is the published transistor fall time. 4-13. The drive transformer primary winding Nd is given by the equation
Any remaining voltage on capacitor C helps to rebuild the magnetizing
current id, thus repeating the cycle.
Diode D is used to clamp any underdamped ringing that could cause the
upper end of winding Nd to become negative. where A, is the chosen core area In square centimeters.
Figure 4-11 depicts an improved version of the previous circuit, which After the turns ratios of the transformer are calculated, find the magnetic
makes its use practical at high frequencies. In thls circuit transistor, Q3 path length of the desired core using the equation
and its associated components compromise a fast discharge circuit fim capac- L?
itor C.
During the time period that Q 2is on and Q, is off, current flowing through
resistor R is multiplied by the gain P of transistor Q , which significantly where H is the coercive force corresponding to the chosen B,, value of the
reduces the charging time of capacitor C. Diode D, is used to discharge the core to be used. -
capacitor C when transistor Q,turns on. If the actual 1, of the selected core is smaller than the calculated magnetic
In this design it is desimble to set the operating point of the transformer path length, the core will be heavily saturated and will not store enough
near saturation, that is, B,, a B,,, energy to provide the desired base drive currents. Either use a larger core,
Turns ritios for the drive transformer were established by Eqs. 4-12 and or introduce a small gap I, using the equation
www.IranSwitching.ir .. .-
g? HIGH-FMQLIENCV S W H I N G POWER SUPPUES WE POWER TRANSISTOR IN CONVEMER DESGN 83
1, =
le -
- Pa
li
(4-18) I 4-7 BIPOLAR TRANSISTOR SECONDARY
BILEIKDOWH CONSIDIPATKXIS
where p, = BIH the permeability of tbe core, 1, is the length of magnetic 1 4-7-1 Forward-Bhs Secondary Breakdown ,
path of material, and I, is the effective magnetic path length. Thus far we have observed during our discussion of power converters tbat
the switching transistor $ subjected to great stress, during both tr~~n-on and
tum-off. It is imperative that the engineer clearly understands how tl.e power
bipolar transistor behaves during forward- and reverse-bias periods in order
4-6-4 Anflsaturation Circuits Used in Base Drives to design reliable and tmuble-free circuits.
In Sec. 4-5 we discussed two methods used to keep the switcbii~gtransistor The first problem is to avoid secondary breakdown of the switching tran-
of a power conv6rter out of saturation, thus reducing the storage time to sistor at hrn-on, when the transistor is foward-biased, Normally the man-
negligible values. These antisaturation circuits may be incorporated into the ufacturer's specifications will provide a safe-operating area (SOA) curve, such
base drive circuits presented in the preceding discussions with excellent as the typical one shown in Fig. 4-13. In this figure collector current is
results. Figure 4-12 shows a typical application using Baker damps in a basic I plotted against collector-emitter voltage. The curve bcus represents the
base drive circuit. All other base drive circuits described thus far may also
be easily adapted to this scheme. Of course, if tbe switching bansistor is a
Darlington, no antisaturation diodes are needed, since the Darlington in-
~i maximum limits at which the transistor may be operated. Load lines that
fall witbin the pulsed forward-bias SOA curve during tum-on are considered
safe, provided that the device thermal limitations and the SOA turn-on time
are not exceeded.
herently presents antisaturation features,
The phenomenon of forward-biased secondary breakdown is mused by
hot spots which are developed at random points over the workipg area of a
power transistor, caused by unequal current conduction under high-voltage
Secondaw
1 breakdawn
limit
j
i
I!
FIGURE 4-12 Base drive circuit shown in Fig. 4-6, redrawn here to show
I
antisaturath diodes ( W r clamps), which reduce the s t w a g drne of transistor FIGURE 4-13 DC and pulsed S o A c u m lor bipolar power tran-
Q,by keeping it out d saturation. slstors.
84 HGH-FREQUENCY SWCHING POWER SVPPLlES
stress. Since the temperature coefficientof the base-to-emitter junction is

negative, hot spots increase local current flow. More current means more
power generation, which in turn raises the temperature of the hot spot even
more. Since the temperature coefficient of the collector-to-emitter break-
down voltage is also negative, the same rules apply. Thus if the voltage stress
is not removed, ending the current flow, the col!ector-emitter junction
breaks down and the transistor fails because of thermal runaway.
Recently a method for avoiding forward breakdown has been developed
by National Semiconductor using a modified emitter-ballasting technique
during transistor manufacturing. Devices manufactured using this technique
may be operated at maximum rated power levels and collector voltages with
no fear of secondary breakdown. Figure 4-14 shows a schematic of the com-
plete monolithic device.
This technique implants a junction field-dect transistor (JFET) in series
with the power transistor base. The JFET acts as a base ballast resistor whose
resistance changes as a function of collector-to-base voltage. This method
differs From the standard emitter ballasting, where the resistor is placed in
series with the device emitter. Base ballasting also maintains constant power
dissipation regardless of collector voltage. Resistor A (Fig. 4-14) takes over
as the JFET pinches off.
4-7-2 Reverse-Bbs Secondary Breakdamn

It was mentioned in previous paragraphs that when a power transistor is
used in switching applications, the storage time and switching losses are the
two most important parameters with which the designer has to deal exten-
-
sively, If storage time is not minimized, saturation of the transformer takes
place, and also the range of regulation of the converter is limited.
On the other hand the switching losses must also be controlled since they
afect the overall eficiency of the system. Figure 4-15 shows turn-off char-
acteristics d a high-voltage power tmsistor in resistive and inductive loads.
R
I E
FIGURE 4-14 Secmdar). break-
down is prevented In bipolar
pcmw tramistors by integrating a
JFETln series with the base. The
JFET a m as a ballast resistor.
(b)
FIGURE 4-15 Turn-oC characteristics of a high-voltage power transistor
For a resistive [a) and an inductive (b)load. Cross-hatched area in each c u m
represents the switching loss energy.
84 HIGH-FREQUENCY WITCHING POWER SUPPLIES
Inspecting the curves we can see that the inductive load generates a much
higher peak energy at turn-of than its resistive counterpart. It is then pos-
sible, under these conditions, to have a secondary breakdown failure if the
reverse-bias safe operating area (RBSOA) is exceeded,
In early transistor literature the m e t h d of rating reverse-bias secondary
breakdown was to test the tmsistor using an unclamped inductive load.
TRe reverse-bias secondary breakdown energy Ess was calculated as
where
The calculated Es,is given in joules. But, since actual testing of the transistor
may be performed with either open base turnsff or with very hi& base
impedance, the Es8 range may vary from millijoules to joules. Also, taking
into account the Eact that the transistor is operating near the breakdown
voltage VcU, the parameter of ESBis relatively useless to modern transistor
specifications.
An dternative RBSOA rating system has been developed by power tran- v,
VcEO
sistor manufacturers using a damped inductive collector load. The resulting
curve is shown in Fig. 4-16, and in a way it ~esemblesthe fonvard-bias SOA FIGURE 4-16 Rwersebias safe operating area (RBSOA) plot
for hii.vdtage switchingzmdstors as a fuDction of reversebias
curve. The BBSOA curve shows that for voltages below Vcw the safe area voltage Tee.
is independent of reverse-bias voltage VEBand is only limited by the device
collector current Ic. Above Vo the collector current must be derated
depending upon the applied reverse-bias voltage. emitter voltage VCEand (2) reducing collector current with rising collectur
It is then apparent that the reverse-bias voltage VEB is of great importance voltage.
and its effect on BBSOA very interesting. It is also important to remember Of course when the power supply desfgn is an of-the-line type, the second
that avalanching the base-emitter junction at turn-offmust be avoided, since solution seems to be the more realistic one. Figure 4 1 7 shows how this can
t u r n 4 switching times may be decreased under such conditions. In any
be accomplished by using an RC snubber network across the transistor to
case, avalanching the baseemitter junction may not be considered relevant, divert collector current during turnd The circuit worh as follows. When
since normally designers protect the switching transistors with either clamp
diodes or snubber networks to avoid s u h encounters.
4-1 SWlTCHlMG TRAWSISWFL PROTECTIVE WEWORKS: 1K: SNUBBERS

It i s now very clear from the previous discussion that the most critical portion
of the switching cycle occurs during transistor turn-off. During the presen-
tation of the base drive techniques we mentioned that normally reverse-
base current I,, is made very large in order to minimize storage time. FIGWFE 447 A hum-off current
Unfortunately this condition may avalanche the base-emitter junction and snubber a e m k across a swftcb-
h g bhlar tramistor. W e 4 is
destroy the transistor. Alternatively there are two options to prevent this a leakage iaductanoe tommutating
from happening: (1) turning.of the transistor at low values of collector-to- diode.
ME POWER TRAMISTOR IN CONVERTER DESlGN 89
transistor QIis off, capacitor C is charged through diode Dl to a voltage If the resistor is too low and > 0.2SIc, then R may arbitrarily be raised
(V, - V,). When Q 1 turns on, the capacitor discharges through resistor to fulfill the constraint.
R. It is important to note that the snubber actually dissipates a fair amount !
The last step is to dculate the maximum resistor power rating given by
of power, thus relieving the switching transistor, which would haw to dis- !
sipate this pawer in the absence of the snubber. pR = .
!
(4-26)
The following analysis and design procedure is based on quasiempirical where f is the converter working frequency in kilohertz.
results, but the equations may be successfully used to develop snubber I
networks for practical designs. Referring to Fig. 4-15b, the energy area
during turn-off may be written as
I The fdlowlng example gives a numerid verification of the above formulas.
EXAMPLE 4-1
Consider a switching transistor used in a half-bridge converter where
mVCE= 200 V, tf = 2 ps, and t, = 0.5 ps. The canverter is working at 20
w h r e Ic = maximum collector current, A
Vce = maximum collector-emitter voltage, V
t, = maximum collector voltage rise time, ps
9 = maximum collector current fall time. ps
i kHz, and the transistor collector current is Ic = 2 A. Calculate the resistance
R and the mpacitance C of the snubkr network.
SOLUTION
From Eq. 422 we have
Solving Eq. 4-21 for capacitance C, we get
As stipulated earlier, capacitor C charges at turn-& and discharges We will use 22 nF. Assume that t, is 40 percent of the total period (1Jf),
I
tbrough R during the transistor turn-iln time t,. The voltage across the I Then
capacitor may be written as
V, = V,, exp - (t,lRC) (4-Dl I
'Using Eq. 4-24,
In order to guarantee a fully charsed capacitor prior to turn-off near VCE,
we must choose RC such that the expression exp - (t,lRC) approaches
unity. By the same token, we must also choose RC such that the capacitor
will be discharged at the end of the turn-on time ton. We will use 300 a.
From basic circuit theory we know that it takes five time constants (5r; Check for the discharge current
r = RC) for a capacitor to discharge fully through a resistor. Assuming in
this case that the capacitor will be essentially discharged at the end of three
time constants, the following expression may be derived for maimurn d$-
charge xesistor value:
1 This is greater than 25 percent of I,, thus a new R must be calculated.
With value calculated for R in Eq. 4-24, we must check the capacitor dis- Take R = 430 S1.
charge current through the transistor at turn-on and restrict it to about 0.251c Finally the power rating of the resistor is calculated.
using the following formula:
90 HIGH-FREQUENCY SWFCHING WWER SUPPUES
I, WE POWER TRAMISTOR IN OONVERTER DESIGN 91
4-9 THE POWER MOSFET USED AS A SWITCH two symbols of Fig. 4-18, the collector, base, and emitter terminals of h e
bipolar transistor are termed drain, gate, and source, respectively, in the
MOSFET.
Although the field-effect transistor (FET) has been used in circuit designs Although botb devices are called transistors, it is important to understand
for many years, the power metal-oxide-semiconductor field-effect transistor that there are distinct differences in the construction and principles of op-
(klOSFET) has been perfected in recent years tu make it commercially eration between bipolar and MOSFET devices. The first and most important
available for power electronics designs. The MOSFET was developed out difference is the fact that h e MOSFET is a majority carrier semiconductor
of the need for a power device that could work beyond the %kHz frequency device, while the bipolar is a minority carrier semiconductor.
spectrum, anywhere from 100 kHz to above 1 MHz, without experiencing
the limitations of the bipolar power transistor.
Of course there are several advantages to designing converters working 4-9-3 Gate Drfve ConsfderaAom of me MOSFET
at, say, 100 kHz rather than U1kHz,the most important being reduced size
When the bipolar power transistor was examined, we mentioned the fact
and weight. The power MOSFET offers the designer a high-speed, high-
that this device is essentially current-driven, that is, a current must be
power, high-voltage device with high gain, almost no storage time, no ther-
injected at the base in order to produce a current flow in the collector. This
mal runaway, and inhibited breakdown characteristics. Different manufac-
current flow in turn is proportional to the gain of the bipolar transistor.
turers use different techniques for constructing a power FET, and names
The MOSFET, on the other hand, is a voltage-controlled device; that is,
like HEXFET, VMOS, TMOS, etc,, have become trademarks of specific
a voltage of speciiied limits must be applied between gate and source in
companies, The bottom line is that all MOSFETs work on the same principle,
order to produce a current flow in the drain. Since the gate terminal of the
although their variation in construction may enhance certain performance
MOSFET is electrically isolated from the source by a silicon oxide layer,
specifications, a fact which might make a specific type of MOSFET more
only a small leakage current flows from the applied voltage sburce into the
attractive than others for some applications.
gate. Thus, we can say that the MOSFET has an extremely high gain and
high impedance.
4-9-2 S u s ! ~MOSFET PeflnRlons In order to turn a MOSFET on, a gate-to-source voltage pulse is Aeeded
to deliver s d c i e n t current to charge the input capacitor in the desired time.
The circuit symbol for a MOSFET is shown in Fig. 4-18, This is an
The MOSFET input capacitance C , is the sum of the capacitors formed by
N-channel MOSFET, and its counterpart NPN bipolar transistor is also
the med-oxide gate structure, from gate to drain (Codand gate to source
depicted for comparison purposes. Of course there is also a P-channel MOS-
(CGS).Thus, the driving voltage source impedance R, must be very low in
FET in which the arrowhead is pointing the opposite way. Lookiilg at the order to achieve high transistor speeds.
A way of estimating the appmximate driving generator impedance, plus
Drain the required driving current, is given in the following equations:
and
NPN
where R, = generator impedance, 0

Ckr = MOSFET input capacitance, pF
dvldt = generator voltage rate of change, Vlns
Emitter Swrcs
FIGURE 4-18 ElectficaI s)mboIs of an NPN bipolar transistor and its To turn 08the MOSFET, we need none of the elaborate reverse current
equivalent N-channel MOSFET. generating circuits described for bipolar transistors. Since the MOSFET is
91 W R E Q U H U C Y SWRCCBNG POWER SUPPUES THE POWER TRANSISTOR IN CONVERTER DESIGN 93
"3
'31r a majority carrier semicanductor, it begins to turn off immediately upon Since by design the MOSFET responds almost instantaneously to gate volt-
removal of the gate-to-source voltage. Upon removal of the gate voltage the age, it will conduct with a V, somewhere between 2 to 4 V, and it will be
'3 transistor shuts down, presenting a very high impedance between drain and fully on with a gate voltage of 6 to 8 V. Assuming b r this example that the
source, thus inhibiting any current flow, except leakage currents (in MOSFET under cansideration is fully on at 7 V, the current Ic required to
"3 microamperes). Figure 4-19 illustrates the relationship of drain current ver- charge the gate-to-source capaci2anre'CGsis
sus drain-to-source voltage. Note that drain current starts to flow only when
3 the drain-to-source avalanche voltage is exceeded, while the gate-to-source
voltage is kept at 0 V.
+a Therefore, the total generator current I, required to turn the FET on is
EXAMPLE 4-2 I , = I , + I , = 1.775A
A power MOSFET driven from a 12-V dc generator is switching 320 V dc
of supply voltage. Given that the MOSFET has a CkD = 100 pf and Ccs =
500 pf, find the total gate current I, required to switch the transistor on 4-9-4 M e Opeding CharacteMcs of Ae MOSFET
within 20 ns.
Figure 4-20 shows the drain-to-source operating characteristics of the power
MOSFET.The reader may compare these curves to the ones given in Fig.
;I 4-1 for the bipolar transistor. Although the two curves look the same, h e r e
P
7
16
From Eq. 4-28 we deduce that in order to charge the gate-to-drain capac-
itance CCD,a current of magnitude CGD(do/dt) is required to overcome the
Miller effect. Therefore,
are some fundamental differences between them.
The MOSFET output characteristic curves reveal two dic~nctoperating
regions, namely, a "constant resistance"' and a "constant current." Thus as
FIGURE 4-19 Drain-bwurce b l d n g characteristics of tbe MOSFET. Note that

r - ~ when the avalanche voltage is reached, the drain current increases Imearly. FIGURE 4 U l Typical output characteristiccums d a MOSFET.
the drain-to-source voltage is increased, the drain current increases pro-
portionaly, until a certain drain-to-source voltage called "pinchof' is
reached. After pinchoff, an increase in drain-to-source voltage produces a
constant drain current.
When the power MOSFET is used as a switch, the voltage drop between
the drain and source terminals is proportional to the drain current; that is,
the power MOSFET is working in h e constant resistance region, and there-
fore it behaves essentially as a resistive element. Consequently the on-
resistance &, of the power MOSFET is an important figure of merit
because it determines the power loss for a given drain current, just as VcE+d
is of importance for the bipolar power transistor. By examining Fig. 420,
we note that the drain current does not increase appreciably when a gate-
to-source voltage is applied; in fact, drain current starts to flow after a
threshold gate voltage has been applied, in practice somewhere between 2
FIGURE 4-ZZ Curve s m n g the relationship of transcon-
and 4 V. Beyond the threshold voltage, the relationship between drain cur- ductance a to drain current. Note how rapidly the trmwn-
rent and gate voltage is approximately equal. Thus, h e transcondu&nce I ductan~orises to saturation as drain current is increased.
gf$, which is defined as the rate of change of drain current to gate voltage,
is practically constant at higher values of d d n current. Figure 4 2 1 illustrates
the transfer characteristics of I D VS. VCS,while Fig. 4-22 shows the rela- charge the input capacitance in order to enhance the switching speed of the
tionship of transconductance g$ to drain current. MOSFET.
It is now apparent that a rise in transconductance results in a proportional
rise in the transistor gain, i-e., larger drain current Aow, but unfortunately 4-915 MOSFET Safe OperuMng Area (SOA]
this condition swells the MOSFET input capacitance. Therefore, care- In the discussion of the bipolar power transistor, it was mentionedathat in
fully designed gate drivers must be used to deliver the current required to order to avoid secondary breakdown, the power dissipation of the device
must be kept within the operating limits specified by the forward-bias SOA
curve, Thus at high collector voltages the power dissipation of the biporar
transistor is limited by its secondary breakdown to a very small percentage
of full rated power. Even at very short switching periods the SOA capability
is still restricted, and the use of snubber networks is incorporated to relieve
bnsistor switching stress and avoid secondary breakdown.
In contrast, the MOSFET offers an exceptionally stable SOA, since it
does not s&m from the effects of secondary breakdown during forward bias.
Thus both the dc and pulsed SOA are superior to that of the bipolar transistor.
In fact with a power MOSFET it is quite possible to switch rated current
at rated voltage without the need of snubber networks. Of course, during
the design of practical circuits, it is advisable that certain derating must be
observed. Figure 4-23 shows typical MOSFET and equivalent bipolar tran-
sistor curves superimposed in order to compare their SOA capabilities.
Secondary breakdown during reverse bias is also nonexistent in the power
MOSFET, since the harsh reverse-bias schemes used during bipolar tran-
FIGURE 4-21 Transfer characteristics of a pmvw MOSFET sistor turn-& are not applicable to MOSFETs. Here, for the MOSFET to
showing the l i i a r dbldV= relationship. turn off, the only requirement is that the gate is returned to 0 V.
THE POWER T U M O R Dl OONMRTER DESIGN 97
FIGURE 4-a A typlal MOSlFET used as a switch, operating fn cmnmwr

source cdguratfon.
FIGURE 4-23 DC and p u l d SOA c u n e s for a MOSFET power
transistor (solid lines). SOA curves of an equivalent bipolar tran-
sistor are dm s h (dotted lines). Note the superior SOk char- An approxfmation of the rise and fall times is given by' the folIowing
acteristics of the MOSFET device.
equation:
4-94 Deslgn Consldemons for OrMng the Power MOSET
It must be clear by now that using the power MOSFET the designer may
I. t, or tf = 2.2RgC,
where t, = MOSFET rise time, ns
achieve performance much superior to using bipolar power transistors. Since tf = MOSFET f d time, ns
the best performance characteristics oF the MOSFET m e forth when the R, = driving generator impedance, il
device is operated at very high frequencies (normally 100 kHz and above), C, = MOSFET input capacitance, p F
certain design precautions must be taken in order to minimize problems, It is important to note that Eq. 4-23 is vdid if RL % R,. This information,
especially oscillations. Figure 4-24 shows a typical MOSFET driving a re- along with the fact that there are no storage or delay times associated with
sistive load, working in the common-source mode. the MOSFET, allows the rise and fall times to be set by the designer. The
There are basically two very simple design rules associatedwith MOSFET ! resistor REin the circuit of Fig. 4-24 is used to assist transistor turn-&.
application which will prevent the transistor from oscillating when used in I
high frequencies. First, minimize all lead lengths going to the MOSFET
terminals, especially the gate lead. If short leads are not possible, then the
designer may use a ferrite bead or a small resistor R , in series with the
MOSFET as shown in Fig. 4-24. Either one of those elements when placed
I In the circuit &ig. 4-24, a MOSFET is used which has an input capacitanw
Ckr = 500 pF, resistor R1 = 150 R, and RL = UlOO a. What will be the
cbse to the transistor gate will suppress parasitic oscillations. rise time of the driving waveform?
Second, because of the extremely high input impedance ofthe MOSFET,
SOLUTION
the driving source impedance must be low in order to avoid positive feedback
which may lead to oscillations. We must note at this point also that while Using Eq. 429 we have
the dc input impedance of the MOSFET is very high, its dynamic or ac
input impedance varies with frequency. Therefore, the rise and fall times 1
t, = (2.2)(150)(500 X 10"1) - 165 ns
of the M OSFET depend on the driving generator impedance.
98 WH-FREQUENCY SWITCHING POWER SUPPLIES THE POWER ?T?ANXTORIN CONVERTER DESIGN 99
Another Important thing to remember is the Fact that the silicon oxide high gain at high current in order to be able to deliver the current demanded
layer between the gate and source regions can be easily perforated and by the MilIer effect during turn-on and turn-off.
therefore permanently destroyed ff the gate-to-source voltage exceeds manu- The following equations may be used to calculate the current flowing in
facturer's specifications. Practid gate voltages have a maximum value any- each buffer transistor at turn-on (QJ and turn-off (Qz), The charge current
where from 20 to 30 V. Even if the gate roltage is below the maximum is given by '.
lcharge
permissible value, it is advisable to perform a thorough investigation to make
sure that there are not any fast rising spikes, caused by stray inductances,
which may destroy the oxide layer of the MOSFET.
and
4-9-7 Clrcults Used In Dtlvlng the MOSFET Ccs 'Ct,- C,
Driuing the MOSFET fbm TTL Although it is possfble to drive the MOS- where CGs= gate-to-source capacitance, pF
FET directly from the output of some transistor-transistor logic (TTL) fam- C, = input capacitance, pF
ilies, direct driving is not recommended, since the transistor stays in the C, = reverse transfer mpacitance, pF
linear region for a long time before reaching saturation, Thus the perform- V, = gate-to-source voltage, V
ance of the MOSFET may never reach its optimum point with such a gate t r = inputpulserisetime,ns
drive,
In order to improve the switching performance, a buffer circuit must be If we assume that the gate-to-drain capacitance discharges at the same time,
provided, which will present very fast current sourcing and sinking to the then t , = grand the discharge current is given by the following equation:
gate capacitances, Such a simple circuit is a complementary emitter-fo1lower
stage, as shown in Fig. 4-25. Transistors Q1md Q2must be chosen to have br
where V p , !s the drain-to-source voltage (hvohs).

1 In order to calculate the power dissipated In each of the buffer transistors,
the following formula is used: C
P = VmI,trf (433)
where VCE= buffer transistor saturation voltage, V
I, + buffer transistor collector current, A
f = transistor switching frequency, kHz
Another way of driving a MOSFET, instead of with discrete transistors, is
using special integrated buffers such as the one shown in Fig. 4-26, i.e., the
DS0026 high-current driver.
m
Dridng the MOSFETfroPle CMOS Because of the MOSFET high input
impedance, it may be dirtxtly driven by a CMOS gate, as shown in Fig.
*
427a. This configuration will produce rise and fall times of about 60 ns. In
order to achieve faster switching times, an emitter-follower buffer may be
0 used, as shown in Fig. 425, or more than one CMOS gate may be
as shown in Fig. 4-27b,to increase current availability to the MOSFET input
FIGURE 4-25 An emitter-follmver b u b r is used between TIZ and MOSFET tu d m
switching rise and fd times. These transistors must have bigh gain and wide bandwidth. capacitances.
402 HIGH-FREQUEW SWITCHING POWER SUPPLIES ME POWER TRANSISTOR IN CONVERTER DESIGN 103
FIGURE 428 A typical apphcation of a single-supply op-ampemitter-follower drive d d t

wed to drive a MOSFET.
FIGURE 4-29 A btempole circuit used tu drive a load connected tu ground. In this
4-9-8 Power AllOSFET Switch Profedion Circuits case the MOSFET is operated in the common-drain mode. I
!
a
It was mentioned during the presentation of the MOSFET SOA character-
istics that it is conceivable to switch maximum power through a power
MOSFET without the need of snubbers. Although this statement may be
true, it is still a good design consideration to add an RC snubber network The gate turn-off (GTO)semiconductor switch is gaining popularity in switch-
across a MOSFET switch. There are basically two main reasons why this is ing circuits, especially in equipment which operates directly from European
advisable. First, the RC snubber alters the MOSFET load line, increasing mains. The GTO oEers the following advantages over a bipolar transistor:
its reliability to a maximum, and second the snubber dissipates the excess high blocking voltage capabilities, in excess of 1500 V, and also high over-
turn-off energy which othenvise would be dissipated by the transistor switch. current capabilities. It also exhibits low gate currents, fast and efficient tum-
Thus transistor stress is minimized without affecting the overall aciency of o& as well as outstanding static and dynamic duldt capabilities.
the switch. Figure 4-33 depicts the symbol of a GTO. The bvo-transistor circuit is
is verv interesting to note also that by using power MOSFETs, the
It --
-. used to explain how the GTO works. When positive drive is applied to the
leakage- indkbnce mimutating diode, which is used across the switch to 1
G terminal, transistor Qzturns on, driving the base of transistor Q1low.
return inductive energy back to the supply bus, is not necessaq any more. Thus, transistor Q,turns on allowing collector current to flow into the base
This is because all MOSFET structures have a body-drain PN junction in of Qz,setting up regenerative conditions, that is, if the sum of the transistor
shunt with the channel. Figure 432 shows a MOSFET used as a switch, gains exceeds unity, the device will latch in the on state. Unlike a thyristor
depicting the integrated body-drain diode, and also an RC snubber. though, the GTO will turn off only when a negative gate drive is applied.
Although the transition time for MOSFETs is one order of magnitude When the GTO is turned on, its saturation voltage is much lower than bipolar
less than for bipolars, all the equations presented in Sec. 4-8 for calculating i transistors even at high current due to its double-injection manufacturing
the R and C are valid for the power MOSFET also. ; process.
www.IranSwitching.ir 1
100 H W - F W E N C Y SWITCHING POWER SUPPLIES
FIGURE 4-26 A high current integrated bulfer

(Len, DS0026) may be used to interface TTL levels to a
MOSFET, &us ussiderably improving the switdrhg
times.
Vriving the MOSFET from Unear Circuits It is conceivable that the

MOSFET may be driven directly from the output of an operational ampli-
fier (op-amp), such as a power op-amp capable of delivering high output
current. The limiting factor, however, is the slow slew rate ofa power opamp,
which limits the operating bandwidth to less than 25 kHz.
In order to improve both bandwidth and slew rate to make the opamp
usable in driving the MOSFET, an emitter-hllower buffer nay be used. A
typical driving circuit using an op-anip is shown in Fig. 4-28.
S m 0 t h Drioing
~ Comaratim In all the previous driving d m i t s the
FIGURE 4-27 Circuit (a) shows direct drive of a MOS-
MOSFET was used in the common-source configuration. There are times, FET Erom a CMOS gate. I n order to fiaprove the speed
however, where the common-drain configuration may have to be used, for of the MOSFET, more than one CMOS gate may be
example, in a bridge circuit, Such an instance requires a totem-pole circuit paralleied in order to provide larger gate mrrent (b).
and thus driving becomes more difficult. The diEculty arises from the fact
that as the voltage across the load increases, the enhancement voltage ofthe
commondrain MOSFET decreases. high, the charge across C is maintained long enough to turn Q2completel?
This can easily be seen from the circuit of Fig. 4-29. In this configurntion, on. Capacitor C must be made large enough to sustain this charge, and a
when MOSFET Q3turns on, the voltage across 2, rises to the voltage V* good first approximation is to choose C 2 IOC,.
This means that the enhancement vdtage of QJ. decreases, and unless Vl > h o t h e r method used to drive a common-dnin MOSFET is the trans-
VZ,the voltage across ZLnever reaches Vz.It will be necessary, therefore, ; former coupled drive. A typical circuit is shown in Fig. 4-31, and it may b~
to produee a voltage at the gate of Q,which will be greater than the voltage used fbr bridge circuit designs, The input drive pulses V,,are in phase.
across the load, and if such a supply is not available, a bootstmp circuit such Transistor pairs QI-Q2 and Q5-Qsconstitute emitter-follower drivers. Thf
as the one shown in Fig. 4-30 may be used. upper MOSFET Q3 is driven through a transformer, while the I w e
In this circuit, when Q1and Q3are on, capadtor C is charged to a voltage MOSFET Q, is directjy coupled to its drive. Because of the polarit!
-
(V V,) through diode D.When Qland Qjare turned off, the gate voltage in the windings of T,,depicted by the dots, Qjis on when Q,is off, and
of Qzis pulled to the above voltage, and Qgturns on, impressing a voltage
(V - Vo) across ZL. Of course, since the input impedance of Q , is very f versa. Resistors R1and iR3 are used to suppress parasitic oscillations, whilr
resistors R2 and R. are used to assist MOSFFI. turn&.
r
FIGURE 4-32 A power MOS-
used as a switch, showing
the integrated commutatlngdiode.
The RC sndber network b used
to keep the transistor MI- bo
low the breakdorm drsin-twsolm
voltage Warn).
c
FIGURE 4-35 P r d c a l redhtlw of a GTO gate drive circuit.
capacitor CL.!.%en the control voltage at the base of Qz goes positive,

transistor Q, turns on, while transistor Q1turns OR since its base now ii one
diode drop more negative than its emitter. At this stage the positive side of ,
capacitor CJis essentially grounded, and CLwill act as a voltage source of,
approximately 10 V, turning the GTO off.
Isolated gate drive circuits may also be easily implemented to drive the
GTO.
C
FIGURE 8-33 Symbd of the GTO and its two-transistor equivalent circuit.
1. Bailey, B.:"Revme Bies Safe Operating Area," Motorola, AN-7S5, 1978.

Z Clemente, S.: '*GateDrive Chamhistics and Requirements for Power HEXFETs," In-
ternational Rectifier, .W-937, 1980,
3. He~terscheid,W.: "Base Circuit Design For High-Voltage Switching Transistors in Power
Convertem," Amperex Electronic Cwp.,1974.
4. Hnatek, E. R.: "Design of4Solid State Power Supplies," 2d ed.. Van Nostrand Reinbold,
New York, 1981.
5. Hdman, D.: "Dridng MOSPOWER FETs," Silicank, AN-794, 1979.
6. Oxner, E.:'What are MOSPOWER FETsT' Silicanix Design Catalog, 1952.
1 7. Pelly, 8. R: "Using High Voltage Power MOSFETs In Off-Une Converter Applicntions."
I
FIGURE 4-34 Basic drive drcuit of a GTO. ! Powercon 7 P d i n g s , 1979.
106 www.IranSwitching.ir
104 HlGHdREQUEMCY SWFCHING POWER SUPPUES
FIGURE 4-30 A hotstrap circuit is used to improve tbe totem-* cmfigurationwhen

the MOSFET Ss operated in the common drain mode.
410-1 Gde Ddve Requirements of We GTO

Udike the bipolar transistor, the GTO may be turned on, at comparable
applications, with much lower gate drive. Thus, gate drive for a GTO may E -5 s
become very simple, and Pig. 4-34 shows a basic drive circuit. 5-
The GTO is turned on by applying a positive gate current, and it is turned FIGURE 4-31 A transfwmer coupled cmmrm-drain MOSFET configuratfm. Trmsfonner
d by applying negative gate-cathode voltage. A practical implementation of TI is used to mbance the turn-on and turrrd of witching transistor Qo without the need of
htstrapping. Gate resistors it, a d R, are parasitic oscilIatim suppressors, and they must
a GTO gate drive circuit is shown in Fig, 4-35. In this circuit when transistor be physicany hated very close to the MOSFET gates. Resistors R, and R,assist MOSFET
Qzis off, emitter folbwer transistor Q1acts as a current source pumping t u r nff.
current into the gate of the GTO through a 12-V zener Z, and polarized
4 0 8 HIWREQUENCY SWITCHING POWfR SWPLlES
8. Pressman, A. I.: "Switching and Linear Power Supplies, Power Converter Design," Nay-
den, Rochelle Park, NJ, 1977,
FIVE
9, R d , :.)I "Base Drive Condderations In High Power Swfkhlng Tranststoa,'"TRW, AH- THE HIGH-FREQUENCY
120, 1975,
10. Shadore, W. R.:"Methodsb r Vtllizlng High Speed Switch@ in Nigh Energy Switching '
POWER TRANSFORMER
Envimnments, " General Semiconductor Industries, 1978.
11. Walker, R. J.: "Cimft Techniques for Optimlzlng Nigh h r Transistor Switching EP-
kienieng," Powcrcon 5, 1978,
12. Wood, P. N,: "DesignConsiderations for Transistor Converters," TRW, AN-142, 1977.
13. -; "Switcbhg Power Converters," Van Kostrand Reinbold, New York, 1981.
Many engineers consider the design of magnetic components ak a sort of

"black art." Nothing could be more wrong. The design of magnetic wm-
ponents is an exact science, and it follows precisely all the fundirnental elec-
tromagnetic laws developed by the pioneer scientists of the field, such as
Maxwell, Ampere, Oersted, and Gauss. 1
The purpose of this chapter is to introduce the fundamental laws of mag-

netism and to present in a simple, logical, and coherent way the relationships 8
that exist between magnetism and electricity for the design of practicab
electromagnetic components, such as coils and transformers.
511 PRlNClPLES OF EWROMAGNmSM

Consider the simple electrical circuit depicted in Fig. 5-1, consisting of a
voltage source V, a switch 5, and a load L, in the form of an air coil. If at
some instance switch S is closed, a current I will flow through the wires to
the load. As the current passes through the coil, a magnetic field is estab-
lished, as shown, which links the turns of the coil. This is calledftux, and
the lines of field are calledJux linkages.
The flux in this mil, however, is still not very strong. If we were to place
a bar of magnetic material (ferromagnetic) inside the coil, as shown in Fig.
5-2, an additional magnetic field would be induced In the bar, prducing
much more flux. The flux linkages would travel through the bar, and they
would find a return path through the surrounding air. If the ferromagnetic
core were constructed in such a way to present a cons-nuou,s path to the
flux, then the field would be confined within the core, as shown in Fig,
5-3, inducing a strong magnetic fiild.
140 HIWREQUENCY SWRCHING POWER W E S THE HIWREQUENCY POWER TRANSFORMER 111
\cld psth
con
FIGURE 5 3 A umtinuaus ferromagnetic are 4 1 confine all
the flux d t h h , producing a stmng magnetic field.
1 *-- +a&.
FIGURE 5-1 A magnetic dux i s developed in an W r e CQU resulting The magnetic field strength can be written as
from a current flw I .
The degree to which the flux is concentrated is &led the m g M H C j l ~ r

where N = coil number of turns
dewify or m o g ~ H cindudiun, measured at a given point, and designated
I = magnitude of current flow
by the symbol 8. The units of B in the centimeter-gram-second (cgs) system,
which is used throughout this book, are given in gauss (G).On the other lI = magnetic length of core
hand, the magnetizing force &at produces the magnetic flux is known as the Another i m p m t relationship between the magnetic flux and the magnet-#
wnen~field strength H, and the units are given in oersteds (Oe). izing force is their ratio, called the permeability p , given as
p a g n d c flux
Permeability depicts the ease with which a core material may be magnetized
by a certain induction force. The permeability of air is constant and has the
vdue of 1 in the cgs system.
512 THE HVSIERESIS LOOP

Every magnetic material is characterized by an S - s h a d curve, known as
the hysteresis loop. This lwp is a curve plotted on B-H coordinates by
subjecting the magnetic miterial to a complete magnetizing and demsgne-
tizing cycle. Figure 5-4 shows a typical hysteresis curve of a ferromagnetic
core without an air gap in the magnetic flux path, Thus, if we start at point
FIGURE 5-e A bar of Fsrromagnetk materid placed inside the cod
a on the curve, which denotes the maximum positive magnetizing farce, to
will create an additional, rtronpr flux. point b, which denotes zero magnetizing force, then down to point c, the
-cl
F 112 HIGKFREW84CY SWITCHING POWER SUPPLIES
THE HIGH-FREQUENCY POWER TRANFORMEU 143
s 8
where El = length of magnetic path of material
i, = length of magnetic path of air gap
p, = permeability of magnetic material
Applying Ampere's circuital law to the gapped core, one can show that the
core flux density may be written as
Equation 5-4 is a very important relationship, since it states that for a given
ampere-turn product (NI),the flux density of a core with an air gap is smaller
than that of a gapless core. In other words, Bd with an air gap is smaller
than B, without an air gap. Thus the introduction of an air gap in a magnetic
circuit gives a "tilt" to the hysteresis loop, as shown in Fig. 5-5, reducing
the possibility of core saturation at high magnetizing force.
The majority of magnetic core manufacturers describe the B-H property
c
I of their material by the normal magnetizing curve, as depicted in Fig. 5-6.
FIGURE 5 4 Hyskmis loop of a typical magnetic fore &hs). This curve shows that the slope of B vs. H in the region below the "knee'"
may be considered essentially constant. Thus a linear relationship exists at
maximum negative magnetizing force, back through zero point d to maximum this region between the excitation current and the resultant flux, which
positive magnetizing force at point a, a complete magnetic cycle is obtained makes the permeability of the core also constant.
in the form of an S. In the low-level region of the curve, core losses become negligible; thus
Figure 5-4 also depicts certain points on the hysteresis loop that are very the core temperature stays low. Above the knee the core goes in to saturation,
important and are defined as follows: and operation at this region must be avoided for linear applications. '
B, is the maximum flux density (induction) point.

H,, is the maximum magnetizing Force.
B,, is the residual magnetic flux (induction), when the magnetizing force
is zero.
H, is the coercive force or the reverse magnetizing force necessary to
reduce the residual induction to zero.
It is apparent from the B-H curve of Fig. 5 4 that B , is reached at a

certain H,, value and that it cannot be exceeded, even if the magnetizing
force is increased. The value of the magnetic induction at this point is said
to be saturated and is written as B,.
lf we were to introduce an air gap into the core, a hybrid flux path would
be ckated, which would alter the effective length of the magnetic
path. Since the air gap ~ermeabilityis unity, the effective magnetic path
length is FIGURE 5-5 A hpterwb loop of a core with an air gap. Nodm tbe
sharing of tfie loop dw to the addition of the air gap flux and the
r e d o c t h of the B, d u e mmpared to tbe gapless case.
144 HKW-FMQUEKY W C H I N G POWER SUPPLlES ME HIGH+WENCY POWER TRANSFORMER 1111
I Slow is urmwabllity p
1-of core in llnear regwn:
I
I
I
I
I FIGURE 5.7 T y p u twwwindiag transformer.
Low-lwrl region tn whlch I
core lwses are neglbible I
I
1 useful characteristics of a transformer is the electrical isolation wbch it offers
I I + between primary and secondary.
I I F H
0 HL H From the fundamental transformer magnetic relationship
FIGURE 5-6 Typical magneihtfm cum showing linear and saturation reghas.
$3 BASIC TRANSFORMER THEORY we can derive an expression to calculate the flux density B in order to make '
sure that the transformer operates in the linear portion ofthe magnetization
We mentioned in previous discussions that a current flowing through the curve. This expression is given by
windings of a coil wound around a closed core will induce a flux confined
within the core. If this current were periodic and a second coil were wound
around the same core, we would expect that the opposite effect would take
place; that is, the flux would induce a voltage and a current flow in this
secondary winding, Indeed this is the case, and Fig. 5-7 shows the simplest
form of a winding transformer.
! where V p = impressed primary voltage, V
f = frequency, Hz
Normally a transformer operates-with high efficiency in raising or lowering N p = primary number of turns
the output voltage in proportion to the ratio of turns, given by A, = core effective area, cm2
I
I
K = 4,44 for sine waves
Since this book is devoted to switching power supplies, K = 4 will be used
for push-puII and bridge converters, and K = 2 will be used for forward-
Thus, transformers are categorized as step-up or step-down, depending upon converters.
whether the secondary voltage is higher or lower than the input voltage. Of N o d I y , the designer of the power transformer chooses B- arbitrarily,
course more than one secondary winding may be incorporated, which could I so that it lies within the linear region of the B-H cume. A good starting
produce both higher and lower voltages. One of the most important and
www.IranSwitching.ir point is to choose B , = B,12.
116 HIGH-FREQUENCY SWITCHING POWER SUPRlfS THE HIGFMREQIJEWY POWER TRANSFORMER 147
A simple rearrangement of Eq. 5-7 yields the number of primary turns in modern converter designs. Ferrites may not have very high operating
flux densities-most ferrites have a B ,from 3000 to 5000 G--but they offer
low core losses at high frequencies, good winding coupling, and ease of
assembly.
Two more design parameters are very important in the seIwtion ofthe prcper Cores made from ferrites come in many shapes and sizes, and various
core. One is the core's (or bobbin's) winding area, whici' must be selected power ferrite materials specifically aimed at hlgh-frequency transformer de-
large enough to carry the proper wire size in minimizing winding Icsses, sign have been developed by manufacturers. Table 5-1 names some of the
and the second is the core's power handling capability. These parameters most popular ferrite trade materials and their manufacturer.
are related by the following equation: The geometry of the core used for a specific application depends on the
power requirements, E-E, E-I, E-C, and pot cores are some of the most
Po, = (1.16B,,fdA,A,)10-8 (5-91 popular shapes. Because of their construction, pot cores are very well suited
where Pout= power handling capability of core, W for Iow- to medium-power applications, anywhere from 20 to 200 W.They
B, = peak operating flux density, G are particularly attractive for designs requiring low flux leakage, and their
f = frequency, Hz inherent self-shielding design minimizes EMI.
d = current density of wires, Alm2 For higher power levels E-E, E-I, and E-C cores may be used. The
A, = core effective area, cm2 E-C core is a compromise between an E-E and a pot core, combining the
A, = bobbin winding area, cm2 advantages of each shape.
Manufacturers of cores Iist all the important parameters required to design
Some manufacturers use the symbol W nfor window area instead of A,, a power transformer, and if a particular parameter is not listed, it can easily
Normally the current density is expressed in circular mils per ampere, sym- be calculated using the equations in Sec. 5-3. Table 5-2 relates magnet wire
bolized as 0,and related to d by size, AWG (American wire gauge), to current densities,
Substituting Eq. 5-10 into Eq, 5-9 we get 5-5 DESIGN OF A POWER TRANSFORMER K)R A
PW-WIDTH-MODULATED HALF-BRIDGE CONVERTER
The step-by-step design example which follows demonstrates the develop-
ment of a typical high-frequency power transformer. The example is general,
Rearranging Eq. 5-11 we derive a very usefuI equation for calculating and and judicial changes will make it useful for full-bridge or push-pull pulse-
selecting the core size of a transformer: width-modulated (PWM)power converter designs, The materids chosen for
thc design were selected as a representative sample, but any material could
Operating current density D is given by the wire manufacturer based on TABLE %I FERRlTE CORE
MATERIAL FOR HIGH-FREQUENCY
1003 circular mils per ampere {c.m./A). Practical designs use a current POWER TRANSFORMERS
density below that number, and depending upon the appIication and the
M d Manufuchrrer
number of the winding turns, current densities as low as 200 c,m.lA may
be used safelyp 3C8 Ferroxcube
243 Stackpole
n FairHite Products
5-4 CORE MATERIAL AND GEOMETRY SELECTION F, T Magnetics, Inc.
H7C 1 TDK
Although almost any magnetic material may be used in designing high- N27 Siemens
frequency power transformers, ferrites have been almost exclusively used
I." ,,1VI1-,I.LYY.I.-.. .-..-., "".,-.-"
I.."-."...-.
I
1
TABLE 5-2 HEAVY FILM-INSULATED MAGNET WIRE SPECIFICATIONS DESIGN PROCEDURE
c u m
Step 1: Choose the core geometry and ferrite material. For this design
based on
choose a Ferroxcube pot-core ferrite of 3CS material. Use the
AWC Min, M a d arw
cirwk & IWOr.n.lA AWG circuit of Fig. 3-12as a guide for the design.
Step 2: Choose a working Em,. From the Ferroxcube catalog specifications
for the X 8 material we find that the saturation flux density at
100% is B, = 3300 G. Since the converter has to work from
90- to 130- and 160- to 260-V ac, we take B,, at 90 V ac to be
1600 G.This choice guarantees that B,, will stay below 3300 G
at 130 V ac, thus the transformer will not saturate.
Step 3: Find the maximum working primary current. The transformer
primary has to conduct the maximum possible current at the low
input voltage of 90 V ac. The dc voltage after the rectifiers
is V, = 2(90 x 1.4) = 252 V. Using Eq, 3-28, the primary cur-
rent is
Step 4: Determine core and bobbin size. We choose to work with a current
density of 400 c.m. lk Use Eq. 5 1 2 to dculate the A,A, product
Choose a core size close to the calculated A,A, product of 0.&0

cmf e. 2616 core has A, = 0,948, but from the same catalog
the single-section bobbin has a winding area A, = 0.406, giving
an A,A, product d u e of 0,384 cm" which is too lw.
Try the 3019-PL003C8 core and bobbin. From the manufac-
tureis data sheets we get A, = 1.38 cm2 and A, = 0.587 cm2#
Then A,h, = O.Sl0 cmd, which is very cbse to the calculated
! value of 0.850 cmb.
be used as long as the manuhturer's data sheets are consulted for use of Although this core is able to handle the required power, it is
the proper specifications. a good practice to choose an A,A, d u e which is at least 50 percent
higher than the calculated one in order to account for insulation
thickness and air space around round wires. Thus for this design
we choose the 3622-PLA30-3C8 Ferroxcube pot core and the
Design a 100-Wpower transfonner to be used in a 20-kHz half-bridge, 3622FlD bobbin. From the manuf&urerns data we get A, = 2.02
PhVM converter which operates from 90- to 130- and 180- to 260-V ac mains cm2and A, = 0.74s cm2, yielding A,& = 1.5 cm4. This value is
and produces 5 V at 20 A output. more &an adequate for the transformer design.
Tr I20 HGl+%".2+-%'ENCY
SWlTCHlNG POWER SUPPLIES
THE HK;H.F-ENCY POWER TRANSFORMER 1m
.;sr3
>
I
Step 5: Caiculate wire size and primary number of turns, Since we chose
using PWM technique, then Vs = 2V,, at minimum VjOy where
400 c.rn.lA to be the wire current density for this design, the V,,, is the nominal output voitage and 2 is the factor that averages
primary winding requires a wire size of 1.19 x 400 = 476 c.m.. out the approximate 50 percent duty cycle. Since we want to
7 which from the magnet wire specifications of Table 5-2 corre- maintain output voltage regulation at V,,,,,,, the number of sec-
1 sponds to no. 23 AWG. ondary turns required is
3 From the Ferroxcube catalog, t1:e1turns per bobbin graph of
I the 3622FlD single-section bobbin shows that appmximately 180 vs 10 400
turns of no. 23 wire are required to fiil the bobbin. Assuming that
the primary winding fiils 30 percent of the bobbin winding area,
Ns = *,v, = 40-107 = - 107 = 3.74 turns
We wili use 4 turns in the secondary.
if the primary turns are calculated to be 60 turns or less, then the
core and bobbin choice is correct. Step 9: Caiculate the secondary winding wire size and number of layers.
Again taking the worst-case operating condition of 90 V ac, We have already mentioned the fact that the secondary is using
V,,,, = 90 x 1.4 - 20 V dc ripple and rectifier drop = 107 V a full-wave center-ap rectification scheme, therefore each sec-
dc. Using Eq. 5-8 the primary number of turns is cdculated. ondary hdf conducts approximateiy 50 percent of the ioad current,
or 10 A. Taking the current density at 400 c.m.IA, then for each
107 X 10s half of the secondary winding w e need 400 c. rn. / A x 10 A =
Np =
4 x 1600 x 20 x lo3 x 2.02 = 41.3 turns 4000 cm., which corresponds to no. 14 AWG. In order to min-
We round off the number of turns to 40, which is below the imize copper losses due to skin effects, it is advisabie to use lower
theoretical vaiue of 60; therefore the core and bobbin choice is a gauge paired conductors for each winding half? or four wires of
good one. 200 c.m. each, for the entire secondary.
At 2000 c.rn., we can use no. 17 AWG wire, which has a max-
Step 6: Check B,, at V,,. Using the calculated number of turns, we imum diameter of 0.049 in. Then, for the entire secondary, the
are now abie to calculate the maximum working flux density number of turns per iayer is 0.50914(0.049) = 2.69 turns. ?ere-
of the transformer at V,,,,, = 130 x 1.4 + 20 V dc for ripple vol- fore the entire secondary of 4 turns occupies ~ W Olayers.
tage = 202 V dc. Using Eq. 5-8, solving for E,, I
Step 10: Check for fit. From the F e m c u b e catalog, the height of the
3019FlD bobbin window is calcuiated to be approximately 0,260
in. From steps 7 and 9, the two windings stack up to a height of
Z(0.025) + Z(0.049) = 0.148 in. Assuming that an additional 0.010
The value of 3125 G is below the specified saturation flux density
in. of tape is used for insulation and finish, the total height is
of the Ferroxcube 3C8 material, which is specified as B , 2 4400
approximateiy 0.160 in., which is beiow the avaiiabie 0.260 in. ;
G at 2S°C and B,, 2 3300 G at 10°C. If a greater B,, margin is
therefore the bobbin will comfortably accept all the transformer
required, then the value of B , used in step 5 must be chosen
windings.
beiow 1600 G.
Step 7: Calculate the number d Iayers used by the primary winding.
From Table 5-2, we find that no. 23 AWG has a maximum di-
ameter of 0.025 in. for double-insuiation wire. The Ferroxcube 5-6 PRACTICAL CONS1DERAIIONS
catalog gives the bobbin window width as 0.509 in. Therefore the When testing the transformer in the actual application, some fine tuning
maximum number d turns per Iayer, using no. 23 AWG, is 0.5091 may be required to better its overall performance. Although most trans-
0.025 = 20.4 turns. Thus the primary winding occupies two formers are manufactured stacking one winding on top ofthe othe~,as shown
layers, 20 turns per iayer. in Fig. 5-8a, interleaving the windings may be necessary to reduce the effects
of leakage inductance. Interleaving is done by winding half the secondary,
Step 8: Calculate the transformer secondary turns, Because the output
foliowed by the primary winding, followed by the other secondary half, as
voltage is derived From a fuli-wave center-tap rectifier scheme shown in Fig. 5-8bn
before the transistor switch is turned on. Figure 5-9 shows the waveforms
of the two modes of operation.
The complete energy transfer waveforms show a high peak collector cur-
rent during the turn-on period of the switching transistor. This means that
a relatively low primary inductance value is needed to achieve this current
rise, at the expense of increased winding losses and input capacitor ripple
Tape flnlsh current. Also, the switching transistor must have a high current carrying
capacity to sustain the peak current.
The incomplete energy transfer mode, on the other hand, exhibits a
relatively lower peak switching transistor collector current at the expense
of a higher collector current flow as the transistor switches on, a fact which
can lead to high transistor dissipation. However, since a relatively high
transformer-choke primary inductance is needed to achieve this mode of
operation, the residual stored energy in the transformer core assumes that
the volume of an incomplete energy transfer transformer-choke will have a
larger volume than the complete energy transfer one, all other factors being
equal,
FIGURE 5-8 (a) A standard transfomer construction with stacked windings; (b) the
same transfomer in an interleaved canfiguration, where the primary is sandwiched be-
5-74 Peslgn Procedure
hveen a split secondary wimling. In the following steps the necessary equations are given to design a Ryback
converter transformer-choke for a complete energy transfer mode. An in-
i complete energy transfer transformer-choke design follow the basic steps
Some designs may also require a Faraday shield between the primary
and secondary to reduce radio hequency interference (RFI) radiation, al-
though pot cores exhibit excellent shelf-shielding properties, since all the
windings are covered by the core material. Whatever the needs are, a good
transformer design enhances the operation of the power supply in more than
one way, and care must be taken during its design.
5-7 THE FLKBACK CONVERTER'S TRANSFORMER-CHOE DES16M

In Chap, 3 the basic operation of the flyback converter was described, and
Fig. 3-4 depicted some of the hndamental waveforms associated with this
converter. Since the isolation element of this topology has the dual hnctIon
of both transformer and choke, the term transformer-choke has been used
to indicate its usage.
ln the Ryback converter two modes of operation are possible for the
transformer-choke: (1) complete energy transfer, where all the energy stored
in the inductor-transformer is transferred to the secondary before the tran- FIGURE 5-9 (a) The waveforms depict the voltage and current relatioaship of
sistor switch is turned on and (2) incomplete energy transfer, where not all a complete energy transfer tlyback transformer-chok, (b)the waveforms depict
the energy stored in the transformer-inductor is transferred to the secondary those oC an incomplete energy transfer made.
THE HIGH-FREQUENCY POWER TRANSFORMER
given below, with some minor changes m the definition of peak collector and dm,. If the converter input voltage varies from V,,,, to,,V
current (Fig.5-9b), which may be written as (1~1- I p d . then
Step I: Transformer pkak primary current. It is necessary to calculate the
transformer peak primary current first, which is also equal to tran-
sistor peak collegtor current. From the fundamental inductor volt-
age relationshi9,'the rate of rise is determined by where
Since in the complete energy transfer mode, the current ramps up

from zero to peak collector current during the time tc when the
switch is closed, the input voltage may be written as
I
1
Step 3: Calculate transformer primary inductance. Since peak primary cur-
rent is now known, the transformer-choke primary inductance may
be calculated by
Taking ilt, = f15-, then Eq. 5-14 becomes Step 4: Select the minimum size core. From a magnetic core catalog pick
the core material and geometry which best suit your application.
If we were to wind only the primary winding into a bobbin, the
winding area A, and the core effective area A, wwld be related by
where Vh = dc input voltage, V
L, = transformer primary inductance, mH
Zpp = peak transformer currents, A
dm, = maximum duty cycle, ps where D is the diameter of the insulated wire (heavy polynylon
f = switching frequency, kHz wire is recommended), and B,, = B,,12. I
The output power in the complete energy transfer mode is equal Since we are to design a transformer-choke, a secondary windin&)
to the energy stored per cycle times the operation frequency. will also be associated with the design. Assuming that the primary
winding occupies 30 percent of a bobbin available winding area,
70 percent will be reserved for the secondary, for air space around
Dividing Eq. 5-16 into Eqa 5-15 we get round conductors, and for insulating tape. Therefore the right-hand
side of Eq. 5-21 must be multiplied by 3 to take the transformer-
choke secondary into account. Experience dictates that an extra
safety factor must be added; therefore the multiplication factor will
which yields, by rearranging terms, the transformer peak primary be 4, and Eq, 5-21 becomes
current
Of course Eq. 5-22 gives a first-order approximation and the final

Step 2: Relate minimum and maximum duty cycles. In the flyback con- core and bobbin choice may vary,
verter, regulation is aocomplished by varying the duty cycle of the
Step 5:- * Calculate the core air gap length. The flyback converter is operating
switch through predetermined limits, which are designated as d,,
single-endedly; that is, the transformer-chokeuses just half the flux
126 HIW+REQU€NCY S W t T C H l K POWER SUPFUES TtrE HGH-FREQUENCI POWER 7RAMSK)RMER 4 2 7
capacity, since the current and flux never go negative. This fact 7: Calculate number of secondary turns. Tfie secondary voltage V$
may present a potential problem, driving the core into saturation. must be calculated when the input voltage (i.e., primary voltage)
To handle the problem two solutions are possible. First, use a is at its minimum with the duty factor at& maximum. It should
core with a very large roIume, or second, in.troduw an air gap in also be noted that V,,,,, = 1.4Vh., - 20 V dc rippIe voltage and
the flux path to flatten the hysteresis loop, thus lowering the work- diode drop.
ing flux density for the same dc bias. Normally, designers perfer Taking the output rectifier diode drop into account, the output
the second solution, which offers more compact transformers to be voltage of a specific secondary winding is written
manufactured.
The air gap presents the largest magnetic reluctance in the flux
!
path, and most of the stored energy in the transformer-choke is in
the air gap volume v,, whose length is l,. Then Accordingly,
where
ug '.Aslg
and
Design the power transformer to be used in a TOO-W, complete energy
transfer flyback converter {see Fig. 3 4 , delivering 5 V dc at 20 A at the
output, and operating over an input voltage range of 90 to a30 V ac.
po = air permeability = 1. Therefore, the air gap length is given
by DESIGN PROCEDURE
Step 1: Calculate the peak primary current. Assume that the conveher has
a maximum duty factor dm, = 21.45. Since the minimum ac voltage
input is 90 V, then,,,V = 90 x 1.4 - 20 V dc for ripple
If an E-E type core, or similar type, is used to construct the trans- and diode drop = 107 V dc. Using Eq. 5-17, the peak primary cur-
former-choke, the center leg may be gapped to the air gap length rent is
l,, or lg may be equally divided between the outside legs of the
core if a spacer is used,
Step 6: CalcuIate transformer number d primary turns. Knowing the ah A switching transistor which can handle this peak coIlector current
gap lengthz the primary number of turns of the transformer-choke at turn-on must also be used in the design.
may be cdculated as follows:
Step 2: Find the minimum duty factor dm,,. The maximum dc input voltage,
after rectification, is
V,,,, = 130 V ac X 1.4 - 0 V dc ripple = 182 V dc
The same result may also be derived by using the following equiv-
alent equation: Allow a 10 percent margin, then V,., = UH] V dc. Also alIow a
7 percent margin on the V,,,, voltage, then Vha,,, = 100 V dc.
Consequently the input voltage ratio K is
Either Eq. 5-24 or 5-25 will give the same results.
128 HIGH-FREQUENCV SWITCHING POWER %PRIES THE HIGH-FREQUENCY POWER TRANSFORMER 129
Using Eq. 5-18 Since we choose B , = 1650 G at 10% from the 3C8 material
magnetization curves in the Ferroxcube catalog we find H = 0,4
Oe.Therefore
These%re the convelter will operate over the duty ratio range of
0 . 2 9 . ~d' < 0.45 for the input voltage range of 200 V dc > Vin >
lo0 V dc.
Step 3: Calculate transformer primary inductance. Using Eq.5-20 yields This efectlve ungapped core volume requires a very large core
size, since the EC70-3C8 core lists an efective core volume of 18.8
cm3 only. In order to be able to use the EC70-3C8, it must be
gapped to a length I, given by Eq. 5-23:
Therefore
Step 4: Select core and bobbin size, Choose a design current density value The center leg of the E-C core may be gapped to 0.15 cm, or a
of 400 c.m. lA for the winding magnetic wires. Then spacer 0.075 cm thick may be placed between the outer legs of
the core in order tu achieve the same gap eikct.
From Table 5-2, the value of 1660 c.m. corresponds approximately Step 6: Transformer primary number of turns. Now that all parameters are
to AWG no. 18, which has a diameter of 0,044in, known, the primary number of turns to achieve &e desired in-
We will also choose a Ferroxcube 3C8 material, E-C type core. ductance is calculated using Eq. 5-24
The 3C8 ferrite material has'a B,, = 3300 G at 1000C,and for this
design B,, = B,,,12 = 330012 = 1650 G. Accordingly,
We will use 48 turns in the primary. Using Eq. 5-28 to calculate

the primary number of turns
From the Ferroxcube catalog the EC70-3C8 core and 70PTB bob- Thus, both equations give equivalent results.
bin yield
Step 7: Transformer secondary number of turns. Using Eq. 5-26, we get
This A,A, value is of course much higher than the required one,
but it is the only E-C core available in the catalog which satisfies
the requirement A,A, 2 6.7 cmJ. We will use this core-bobbin = 3.52 turns
combination for the design.
Since there will also be some voltage drops in the printed circuit
Step 5: Calculate the air gap length 1,. In order to be able to use an
conductors and the output winding copper conductors which were
ungapped core the effective core volume u, listed in the catalog not taken into account in the above equation, the number of sec-
must be equal to or greater than the theoretical oe value given by ondary turns may be taken to be Ns = 4 turn.
The output voltage on a flyback converter requires a single wind-
ing, one diode, and a capacitor, as shown in Fig. 3-4. In order to
f 30 KGH-FREQUENCY SWlTCHlNG POWER SUPPLIES 7HE HIGH-FRWENCY POWER TRANSFORMER 131
deliver 20 A of output current at 400 c.m. /A, a wire of 20 x 400 Varnish impregnation may not be necessary with a ferrite transformer,
= 8000 c.m. is required. To minimize losses due to skin effect, since oxidization of the core due to moisture, as was the case with iron
four wires in parallel of 2000 c.m. per wire will be used. This laminations, is not a factor. Also the acoustical noise which was asscciated
corresponds to usiqg h u r AWG no. 17 wires in parallel. with low-frequency transformers is not present in the ferrite high-frequency
-, There shauld be no problem in fitting all the windings, plus transformers, which generally operate above the human acoustical range: ,
insulation, in the chosen bobbin because the A,A, product was
I
That is not to say that ferrite transformer assemblies may not generate
much 1argr;r than the calculated one, by a factor of 2. mechanical or acoustical noise. They may, because whatever they are
mounted on acts as a sounding board. A phenomenon associated with ferrites,
called magnetostriction, shortens or lengthens the part due to the applied
magnetic field, which in turn causes mechanical resonance of the core as-
5-8 SOME GENERAL HIOH-FRWENCY sembly, In fact magnetostriction changes polarity, from negative to positive,
WEISFORMER CO#SIDERATlONS as the temperature of the core rises. Therefore, care must be taken to use
e
In the previous paragraphs, some practical and fundamental design equations proper methods when mounting the transformer to the board, in order to
and procedures were given for the design of the isolation transformer of reduce or eliminate any acwstid mechanical noise.
certain types of converters. The majority of the design equations are fun-
damentally valid for any type of magnetic circuit, and they may be adapted
to solve a variety of magnetic-applications, whether transformers, chokes,
or a combination of both. I. Ferroxcube (a) "Linear Ferrite Materials and Components," F e m u b e Corp,, Magnetic
In general magnetic components used in the construction of a switching Catdog; (b) "Linear Ferrite Magnetic Design Manud," Ferroxcube Corp., Bulletin 550,
power supply have to comply with certain national or international safety 1971.
standards. Thus, for the North American countries, standards set by the 8. Manka, W.V.: Design Power lndu~torsStep By Step, Ekchonic De&n, vol. 25, no. 26,
Underwriters Laboratories (UL)for the United States and Canadian Stan- 1980.
dards Association (CSA) for Canada are valid. For European use, the West 3. Mfddlebmk, R. D., and S. cuk: '"Advances in Switched Mcde Power ~onversfan,"~ 0 1 s ~
German Verband Deutscher Elektmnotechniker (VDE) safety standards 1
1 and 2, T e s b , Pade.na, CdfDrnla, 1981.
have become the most popular design guidelines, since they are considered 4. Pressman, A. I.: *Switching and Lfnem P m r Supply, Power Converter Design," Hayden,
to be the most stringent. Rochelle Park, NJ,1977.
There are some fundamental differences between the UL and VDE safety 5. Thibodeau, P. E.: The Switching Transformer:Dmigning It in One Try for Switching Power
standards, the UL concentrating more on preventing fire hazards, while the Supplies, Electwnk b s l g n , d.28, no. 18, 1980.
VDE is more concerned with the safety of the oprator. In constructing an 6. Watson, J. K.: 'Xpplications of Magnetism," Wiley, New York, 1980.
isolation transformer, UL and CSA limit the winding temperature rise to
65°C above ambient for class 105 insulation and to 85OC above ambient for
class 130 insulation.
In any case, it is a good design practice to keep the temperature rise of
a switching power supply transformer to a low value, since most of these
transformers are constructed using ferrites, which have thermal limitations,
Ferrites have a Curie temperature of about 200°C, which limits their op-
erating core temperature to about 300°C. Curie temperature is the tem-
perature at which a material changes its ferromagnetic properties and
becomes paramagnetic.
The VDE safety standards, on the other hand, have strict requirements
for specific winding techniques and input-to-output isolation requirements,
which may require a 37SO-V ac hi-pot test potential. An extensive discussion
of these safety requirements is given in Chap. 11.
SIX
- .- -
THE OUTPUT SECTION:
RECTIFIERS, INDUCTORS,
AND CAPAClTORS
&0 INTRODUCTION
:-
In general the output section of any switching power supply is comprised
of single or multiple dc voltages, which are derived by direct rectification
and filtering of the transformer secondary voltages and in some cases further
filtering by series-pass regulators. These outputs are normdIy Iow-voltage,
direct current, arid cupable of delivering a certain power level to.drive
electronic components and circuits. Most wmmon output voltages are
u
w 5 V dc, 12 V dc, 15 V dc, 24 V dc, or 28 V dc, and their power capability '
C
may vary from a few watts to thousands of watts.
The most common type of secondary voltages that have to be rectified in
a switching power supply are high-frequency square waves, which in turn
require special components, such as Schottky or fast recovery rectifiers, low
I
ESR capacitors, and energy storage inductors, in order to produce low noise
outputs useful to the majority of electronic components.
This chapter describes the characteristics, merits, and Iimitations of the
components used in the output section of the switching power supply. Design
equations and procedures are also developed to aid the reader in the practical
application of these components.
1
6-1 OUTPUT RECnFlCATlON AND FILTERIN6 SCHEMES
The output ratification and filtering scheme used in a power supply depends
on the type of supply topology the designer chooses to use. The conventional
. flyback converter uses the output scheme shown in Fig. 61.Since the
transformer Tl in the flyback converter also acts as a storing energy inductor,
i
i diode Dl and capacitor Cj are the only two elements necessav to produce
a dc output. Some ~ructicaldesigns,
- .. - however, may require the optional
133
134 MGH-FREQUENCY $WITCHING P&ER SUPPLIES
FIGURE 6-1 Output s d o n of r flyback type switching power supply.
insertion of an additional LC filter, shown in Fig. 6-1 within dotted lines,

to suppress high-frequency switching spikes. The physical and electrical
values of both L and C will be small. FIGURE 6 3 Output sectfon of push-pull, bdf.bridge, or full-bridge &tcbing power
.
An important factor in the design of any output power supply section is supply.
the minimum dc blocking voltage requirement of the rectifier diode and the 1
flyheel diode (for designs which require one). For the flyback converter, ! Since each of the lxw diodes Dland D? provides currenf to the output
the rectifier diode D lmust have a reverse voltage rating of [1.2V, (NstN,)], I for approximately half the cycle, they share the load current equally. An
minimum. interesting point is that no flywheel diode is needed, because e i t h e ~diode
The ournut section of a forward converter is shown in Fig. 6-2. Notice acts as a flywheel when the other one is turned of. Either diode must have
the distinct differences in the scheme compared to the flyback. Here, an a reverse blocking capability of [2.4Vw,/V,,-JVh,,)], minimum.
extra diode 4, called the Aywheel is added and also inductor LIprecedes
the smoothing capsitor. Diode D? provides current to the output dudng 6 3 POWER RECTIFIER CMARACreRlSllCS
the off period; therefore, the combination of diodes Dl and D2 must be IN SWltCHlHG POWER SUPPLY 'DESIGN
capable of delivering full output current, while h e i r reverse blocking voltage The switching power supply demands that power rectifier diodes must have
capabilities will be equal to [1.2V,,(N5/NP)]rminimum. The output section low forward voltage drop, fast recovery characteristics, and adequate power
that is shown in Fig. 6 3 is used for push-pull, half-bridge, and full-bridge handling capability, Ordinary PN junctfon diodes are not suited for switching
converters. applications, basically because of their slow recovery and low efficiency.
Three types of rectifier diodes are commonly used in switching power sup-
plies: (1)high-dciency fast recovery, (2) high-efficiency very fast recovery,
and (3)Schottky barrier ce&ers. Figure 6-4 shows the typical farward
characteristics of these diode types. It can be seen from the graphs that
Schottky barrier rectifiers exhibit the smallest forward voltage drop, and
therefore they provide higher efhiencies,
The follow&g discussion explores the daerences and merits of each rec-
tifier type, and it suggests usage in switch-modepower rectification schemes.
6-24 Fasf and Vew Fast Recovery Diodes

Fast and very fast recovery diodes have modelate to high forward voltage
1
FIGURE 6-2 Output section of a frnward rrpe dtrhing power supply. i drop, ranging anywhere from .0.8
.- to 1,2 V. Because of this and because of
F-
1% HGKFREQUENCY SWlTCHlNG POWER SUffllES
0
FIGURE 6-4 Typical f m r d voltage drop cbaractertsth, at variaus f m r d

3 current levels, for ((a) a Schottky barrier d f i e r , (b)a very fast recovery
rectifier, and (c) a conventional fast recover)"d f i e r .
a
their nomaIly high blocking voltage capabilities, these diodes are particu-
larly suited for low-power, auxiliary voltage rectification for outputs above
12 v.
Because most of today's power supplies operate at 20 kHz and above, the
fast and very fast recovery diodes offer reduced reverse recovery time t,,
in the nanosecond region for some types. Usually the rule of thumb is to
select a fast recovery diode which will have a tRItvalue at least three times
lower than the switching transistor rise time.
These types of diodes also reduce the switching spikes that are associated
with the output ripple voltage. Although "soft"' recovery diodes tend to be
less noisy, their longer tRRand their higher reverse current IRy create much
higher switching losses. Figure 6 5 shows the reverse recovery characteristics
of abrupt and soft recovery type diodes.
Fast and very h t switching diodes used in switch-mode power supplies
as output rectifiers may or may not require heat sinking for their operation,
depending upon the maximum working power in the intended application. .-
rJormally these diodes have very high junction temperatures, about 17S0C,
138 HIGH-FREQilENCY SWFlCHlNG POWER SUPPLIES THE WTfW SECTION: RKTiFER.5, INDUCXX1S, AND CAPACITORS 139
and most manufacturers give specification graphs, which will allow the de-
signer to calculate the maximum output working current vs. lead or case No snubber
temperature.
1-24 SchWky Barrier Rectlflers

Thc graph in Fig. 6-4 reveals that the Schottky barrier rectifier has an
extremely low forward volbge drop of about 0.5 V, even at high forward
currents. This fact makes the Schottky rectifier particularly effective in low
voitage outputs, such as 5 V, since in general these outputs deliver high load
currents. Moreover, as junction temperature increases in a Schottky, the
forward voltage drop becomes even lower.
Reverse recovery time in a Schottky rectifier is negligible, because this
device is a majority-carriersemiconductor and therefore there is no minority-
carrier storage charge to be removed during switching.
UnEortunately, there are two major drawbacks associated with Schottky
barrier rectifiers. First, their reverse blocking capability is low, at present
time approximately 100 V. Second, their higher reverse leakage current FIGURE 6.6 Upper wavefonn shows ringing during
Schottky rectilier turn-OK. Lower waveform shows the same
makes them more susceptible to thermal runaway than other rectifier types. rectifier after transient suppression has been added.
These problems can be averted, however, by providing m s l e n t overvoltage
protection and by conservative selection of operating junction temperature.
this method works well, the slow recovery ofa zener may induce noise spikes
at the power supply output, which may not be desirable for low node ap-
6-2-3 transienf Overvoltage Suppression plications.
Consider the full-wave rectifier, shown in Fig. 6-3, using Schottky rectifiers It can be shown that the value of the snubber resistor R, may be found
D Land D, in a PWM regulated half-bridge power supply. The voltage Vs by equation
across each half of the transformer secondary is 2V0,, minimum; therefore
each diode must be capabie of blocking 2Vs at turn-off, or 4V,,.
Unfortunately, the leakage inductance of the high-frequency transformer
and the junction capacitance of the Schottky rectifier form a tuned circuit where LT = transformer leakage inductance, pH
at turn-off, which introduces transient overvoltage ringing, as shown in Fig. C, = Schottky junction capacitance, pF
6-6. The amplitude of this ringing may be high enough to exceed the blocking n = primary-to-secondary turns ratio, NPSNS
capabilities of the Schottky rectifiers, driving them to destruction during
The value of the snubber capacitor C, may be arbitrarily chosen to be
the turn-off period,
anywhere from 0.01 to 0.1 pF.
The addition of RC snubber networks will suppress this ringing to a safe
The power dissipated in the resistor may be found by
amplitude, as shown in the lower waveform of Fig. 6-6, There are two ways
of incorporating RC snubbers at the uutput ofa power supply to protect the
Schottky rectifiers, For high current outputs the snubbers are placed across
each rectifier as shown in Fig. 6-7a, while for low current outputs a single
where f is the converter's operating frequency.
RC snubber across the transformer secondary winding, as shown in Fig. 6-
7b, may be adequate. .inother solution is to place a zener diode, as shown Proper selection of the snubber capacitor C , makes the snubber effective
in Fig. 6-7c, to clamp the excessive voltage overshoot to safe levels. Although and less dissipative.
THE OWPW SECTION: RECTIFIERS. INDUCTORS. AND CAPACITORS 141
6-24 Calculaling tho Recflfier Plodo

Peak Current Cupabillfy for the Flyback,
Fornard, and Push-pull Converters
As mentioned in previous discussions, the output diode in a flyback converter
conducts only during part of the converter switching cycle namely when
the switching transistor is off Accordingly, one has to expect that the output
rectifier must have the current capability of providing full output current
during the conduction time period (Fig. 6-1).
The minimum peak forward current which the output diode must provide
is given by
where,6 is the converter's maximum duty factor.

Assuming dm, = 0.45 for the flyback converter, then
iFM = 3.6 I,, (6-4)
EXAMPLE 6-1
Calculate the output recti5er peak fonvard current rating which may be used
in a 100-WPWM Ryback converter, providing 5 V dc at 20 A output, working
with a duty factor dm, = 0,45, at a frequency of 20 kHz.
I
SOLUTION 0
Fmm Eq. 6-4 we get
Therefore a rectifier which can provide 72 h peak repetitive fonvard current

at an approximate duty cycle of 45 percent must be used in this design.
In the f o d converter, the selection of output diodes becomes more

involved, since one also has to calculate the peak forward current capability
of the flywheel diode (Fig. 6-2). On the other hand, because in the forward
converter energy is continuously flowing to the output load, the peak forward
current will be lower for each diode than in the case of the flyback converter.
The peak Ibnvard current for the output diodes of a fonvard converter is
piven by
RGURE 6.7 Means of protecting output Schottb rectifiers during hrn-off. (a) Snub- IFM = Ioutdo (6-5)
bers placed acron each sectI6er; (b)a single RC mubber placed across the transfonner where S. is the duty factor of either the rectifier or the flywheel diode.
semndary winding (c) using a Zener diode.
1
THE OUTPlJl SECTION: RRXFIERS. INDUCTORS. AND CAPACITORS 443
1
Assuming a dead time of 5 ps between each alternative half cycle, then the
EXAMPLE 6-2 conduction time for each rectifier is 20 ys. AmrdingIy, each diode duty
Cnlculate the rectifier and the flywheel maximum forward current rating for factor is dDR = 20150 = 0.4. From Eq. 6-5we get.
a forward PWM converter with the specifications given in Example 6-1,
working over an input voltage range of 90 to 130 V ac.
SOLUTION Therefore each rectifier may have a minimum rating of 8 A at 40 percent

duty cycle. (In practice 10-A rectifiers should be used.)
The rectifier diode peak forwaml current is, using Eq. 65, As was previously explained, one of the output diode rectifiers acts as a
IFM = IourdDR= 20(0.45)= 9 A flywheel diode when the other rectifier is turned off, In this case, each diode -
conducts for 5 ps in the flywheel mode for dDF = 5150 = 0.1. Therefore
A rectihr diode with a 10-A rating at 45 percent duty cycle is used. The
the current provided to the output during the dead time by each rectifier
maximum duty factor of the flywheel diode is diode is
Take V,,,,, = 9 0 f i - 20 V dc for ripple = 106 V dc. We will use 100 V I
dc. Also, V,,,, = 130lh = 182. Use l90 V dc. Then from Eq. 6-6 It is recommended that a thermal analysis be made for the rectifier's
intended use, and adequate heat sinking must be provided to avoid destruc-
tion due to t h e d mnaway. Normally a manufacturer provides curves of
!
diode current derating vs. case temperature, which shwld be consulted
Consequently, the forward current in the flywheel diode is during the design.
IPM= ZO(0.76) = 15.2 A For higher current outputs, diodes may be paralleled to share the load
i
1 current. Direct paralleling of diodes must be avoided; instead using separate
A diode rated 20 A at 76 percent duty cycle should be used. secondary winding feeding individual diodes, as shown in Fig, 6-8, is rec-@
ommended.
In the push-pull family converters, the output rectifiers provide equal
current to the output load during equal conduction cycles (Fig. 6-31. This
output scheme is also valid for the half-bridge and the full-bridge circuits,
Because the output of a push-pull converter works as two back-to-back
forward converter outputs, maximum f m a r d current for each rectifier fs
given by Eq. 6-5.
Calculate the maximum forward current rating for each output diode rectifier
of a half-bridge PWM converter with the specifications given in Example
6-1.
SOLUTION
The converter's switching period is
FlGURE 6 8 Separate serandary windings feeding indMdual diodes should be used
T = -1= - - 1 - SO ps where increased outpur current capacity is needed. Mrect dhde paraUeling must be
f 20 kHz avoided.
144 ' MGH-FREQUEKY WCHING POWER SJPPLIES
RIE .OLIIPCTI SECTION: RECTIAEm, INWCTORS, AND CAPACrrORS 145
6-3 SYNCHRONOUS REGTIRERS

6-3-1 General Considerations
AS digital integrated circuit manufacturers are working toward implementing Flnd the power losses encountered in the output section of a center-tapped
more electrical functions and circuits on a single silicon chip, the need For full-wave rectifier section of a 3-V,30-A circuit using the FolIwving rectifiers:
lower bias voltages is necessary. Manufacturers of digital ICs have stan- (a) Schottky diodes with a Vf = 0.6 V, "$1 power MOSFET synchronous
dardized the new fogic bias voltages to 3.3 k 0.3 V for 0%-line operation rectifier with RDs., = 0.018 S1 and, (c) 3ISYN synchronous rectifier with
and to 2.8 -i- 0,S V for battery operation. EcEvw " 0.ma.
These new voltage standards have forced design engineers of switching The switching waveform has 6 , = 0 . 4 with a dead-time period of 0.05,
power supplies tn explore the possibility of using devices other than junction
diodes for rectification, in order to reduce the power losses encountered at SOLUTION
the output section of the power supply, For Instance, power losses per diode
1. From Eq. 6-7 using Schottky rectifien: During the positive hdf C Y C ~
at the output circuit utilizing center-tap rectification topology using con-
when the first diode conducts power, dissipation is
ventional rectifiers is given by
PL = (V/1,36,, = (0.6)30(0.45)= 8.1 W
. " (Vflm36,
P (6-7)
where Vf is the diode forward voltage drop, I,,, is the power supply rated During the dead time td9the second diode acts as a flywheel diode
output current, and,,a is the input waveform duty cycle, dissipathg
Since the power dissipation is directly proportional to the diode's forward
voltage dmp, even the best Schottky rectifiers with a 0.4-Vforward voltage
F2 - (0.6)(3)(0.05)= 0.9W
During the negative half cycle of the input waveform, the second diode
drop may account for up to 20 percent of the total input power loss. In has the same power dissipation as ahve. Therefore the total power loss
recent years the power MOSFET with low RDs,, has provided an alternative in the circuit is
in power rectification, and as prices of MOSFETs keep dropping, circuits
using MOSFETs as synchronous rectifiers have started to appear in switching
power supply designs.
Pt - 2(P, + PJ = 2(&,1+ 0.9) = 18 W
*
2. From Eq. 6-8 using the power MOSFET and repeating the above pro-C
Unlike the conventional rectifier the power MOSFET synchronous rec- cedure, total power dissipation is found to be
tifier exhibits practically no switching losses due to its fast switching times,
while its power Ioss when on and working with a center-tap rectification Pt = 2(7.29 + 0.81) = 16.2 W
topology Is given by 3. From Eq. 6-9 the total power loss using a BISYN bipoIar synchronous
rectifier is found to be
P = I:ut(R~~.on)~max
where R,,, is the MOSFET on resistance,
Another practical synchronous rectifier is realized using a specialized
bipolar transistor, such as the BISYN series offered by Unitrode Corporation.
Bipola synchronous rectifiers match or exceed the merits of their MOSFET 6.3.2 me Power MOtFET as a Synchronous Rsdlfier
counterparts with some added Features. The power Ioss of a bipolar syn-
chronous rectifier is directly proportional to its collector-emitter on resist- Figure 6-9 shows a practical implementation of a center-tapped full-wave
ance and for center-tap applications is given approximately by output section of a switching power supply utilizing power MOSFETs as
synchronous rectifiers.
MOSFETs QIand Qeare selected to have minimum possibk RpS,mat
The following'paragraphs describe the characteristics of the MOSFET maximum output current, in order to increase eliciency. 'hnsf'ormer wind-
and bipolar synchmnous rectifiers in order to familiarize the reader with ings N3and N4 are used to turn on the MOSFETs at opposite half cycles of
these devices. the input waveform. It Is beneficial to overdrive the gate-to-source voltage
in order to minimize RpS,m. Careful selection of resistors RI through Re dl
146 H I G H - F W E N C Y SWITCHING WWER SUPWES ME WTFUT SECTION: RECTIFIERS, MUULICJK~,wwub n m v n v v n w
,-,
FIGURE 6-10 C i d t Irngdementatim of a forward converter output section using power

MOSFETs as synchnmous rectaen.
6 3 3 the SIpolar Synchronous Rectifier

Synchronous rectifiers implemented with special bipolar transistors offer all
R3 the advantages of the MOSFET synchronous rectifiers up to frequencies
FIGURE 69 Fu11-war output redhation of a switching p + ~uskg MOS-
suppIy of 200 kHz (above this frequency the MOSFET dominates), mupled with
PETSas synchmous rectifiers. D b k nd DIare parasitic dides dthe MOSFETS- lower prices and improved temperature coeacients compared to power
MOSFETs.
As an example the BISYN series of bipolar synchronous rectifiers offered
also rninimfze the switching times. Diodes Dland are parasitic diodes of by Unitmde Corporation, at their introduction, exhibit very low saturation
the MOSFETs, and they act as free-wheeling diodes during the input wave- resishnce (8 mil for a 4.5-mm chip), with good gain (%25)and symmetrical,
form dead time td to provide current path for inductor t.Since they share blocking capabilities for both positive and negative input voltages. The latter
the load current during this period, their forward voltage is minimized but feature is very important bemuse now the possibility exists of regulating any
in order for the circuit to be practical, the power loss of each MOSFm output voltage of a multiple-output switching power supply using PWM
when on must be much less than the fornard voltage drop of the parasitic techniques to drive the bipolar synchronous rectifier, without the need of
diodes, that is. series-pass linear regulators or magnetic amplifiers, realizing high eficiency
gains.
Practical implementation of bipolar synchronous rectifie~sis shown in
Figure 6-10 shows the Lmplementa€ion of the M O S F m as synchronous Fig. Bll for a center-tapped output circuit. In this circuit synchronous
rectifier in a single-ended fornard converter. In this confipradon transistor rectifier Q1is conducting when the input waveform goes positive providing
Q , turns on by the gate-to-source voltage developed by winding Nl, on the current to the output. Diode D3 IS free-wheeling, which provides a path for
positive cycle of the witching waveform. When the input voltage Ewingr the output inductor current during the input waveform dead-time period
td. At that instance the voltage across the secondary windings has collapsed,
negative, MOSFET Q lturns 08 and output current is supplied to the load
through MOSFET QY which now turns on thmugh coupled winding & but diode D,is forward-biased due to the voltage potential developed amss
using energy stored in the output inductor L. coupled winding N5 thus speeding up t u r n 4 of synchronous rectifier QI.
The criteria for choosing the proper MOSFET parameters lor use i this At the same time diode D2 current flow will turn on synchronous rectifier
circuit are the same as in the bridge configuration output circuit. Flyback QB,and it will remain on for the negative duration of the input waveform,
circuit output sections are implemented as shown in Fig. 6-10, without the untiJ magnetizing current drops below diode Dz current, i.e., during the
use of MOSFET Qz. positive cycle d the input waveform.
14S HIGH-FREQUENCY SWITCHING POWER SUPPLIES THE W SECflON: RECTIFIERS. INDUCTORS. AND CAPACTK)RS 449
FIGURE 6-12 Implementatim of bipdw synchronws d e r s in a single-ended for-
* R2 02
ward converter output section. (Courtesy Untdmde Corporation.)
FIGURE 6-11 Bipdar synchronm rectifiers Q,and Q, used in a center*tapped current overshoot and ringing which, reflected back into the primary, may
circuit. (Courtesy Unttrouk Corpurutlon.) damage the switching transistor.
Figure 6-12 depicts the operation of bipolar synchronous rectifiers in a

singleended forward converter output section, 6-34 O m Voltage Regulation bIng Blpolar
En the following discussion, it is assumed that the single-ended forward Synchronous Redifietsand PWM Rsguldor Techniques
converter has an on-time duty cycle of 550 percent. b k i n g at the circuit As we previously mentioned, because of the symmetrical blocking capabg-
when current flows in the primary winding of transformer TI,voltage is ities of the bipolar synchronous r d e r , any output of a switching power
induced in the secondary windings with the polarity shown by the dotted supply may be rectified and independently regulated using local PWM tech-
ends. At this time synchronous rectifier Q 1turns on, allowing current to niques.
flow to the output. Because of the polarity of the coupled winding N4 to the Figure 6-13 shows a p d c a l implementation of such a circuit, rectifying
output inductor L, synchronous rectifier Q1and diode Dlare off. During and regulating a low-voltage, high-current output in a single-ended fornard
the ofFperiod of the input waveform, rectifier Q 2turns on, sustaining output converter, using a Unitrode Corporation BISYN synchronous rectifier and
current through inductor L. Base drive energy is delivered through winding a UC352SA PWM control circait.
N,,while diode Dlstill remains off. Since the biasing requirements of the BISYN Q , necessitate the use of
When the transformer primary windings are turned on again, the voltage positive and negative power supply voltages, the PWM circuit is biased
across winding N3 is clamped to near zero by diode Dland the f o m r d - using 212 V generated by rectifying and filtering the secondary voltages
biased base-collector junction of Q2. developed by windings N,and N,.
Since winding N3 is essentially shorted, Qi is still off due to lack of base Transistor Qjis used as a current source to level shift the output voltage
drive, while Qe is rapidly shutting off. At that point D lis reverse-biased in order to drive the inverting input of the P W M error amplifier. The output
allowing voltage to build up in the secondary windings, turning synchronous- voltage is developed across resistor R9 and diode L),. This diode is used to
rectifier Q1on through winding N1. temperature compensate for the VBEvoltage drop of transistor Q3
The utilization of windings X3 and N1circumvent the long recovery time The drive current to the BISYN is limited by resistor R, during the FWM
of the synchronous rectifier 92 (300 to 400 ns), thus reducing excess output on time, except at initial turn-on when the bias drive current may be a few
..
-1 n
FIGURE 6-13 A BISYN synchrormus d e r Q , and a PWM c u n d IC are used to ac-
mmplish both r e d m t i o n and regulation of a 3-V output. (Courtesy Unitroh Corporation.)
..
magnitudes higher than steady-state current. Negative turn-off current is
provided by the PWM totem pole output during the off period.
The free running frequency of the PWM is set by resistor R, and capacitor
C, and may be higher than the switching frequency of the power supply
itself. Synchronization is accomplished by transistor Qe diode 9,and R2
and CZ. Diode D5clamps the negative voltage excursion of the differentiator
circuit R2/C2to prevent malfunction of the control circuit.
63-5 A Current-Driven Synchronous R d f i e r

A current-driven synchronous rectifier may be implemented using a power
MOSFET and a handful of external components, as shown in Fig. 6-140.
The power MOSFET utilized in this design must have a low RDs,, at the
required current of operation. During implementation of the circuit shown
in Fig. 6-14a, a Motorola TMOS device, the MTM6ON05, was used, which
has an RDs,o,5 30 mSt at 10 A.
Referring to Fig. 6-14a, transformer T1 is a current transformer with
windings N1:N2:NJIhaving a 1:25 :3 turns ratio. A voltage applied at terminal (C)
A will cause a source current I, to flow through winding N1.En turn, a current
FIGURE 6-14 (a) A current-drhn syixbronws rectifier circuit and its sug-
of 0.041, A is induced in winding Nz,flouring thmugh diode Dl into the gate - gested symbol. ( b )and (c3 Applications of the symhmnous rectifier in a full-wave
of p a ~ e MOSFET
r Q1. This current will charge the FET input capacitance center-tapped rcctikakim scheme and a single-ended r m a r d converter output
C,, turning the transistor on. Transformer T i is designed so that it will scheme. (Courtesy Motorola Semiconductor Products, Inc.)
saturate at some point. Saturation of the core terminates the charging of the
FET input capacitance, since all transformer windings appear shorted. Con-
152 HIGH-FEQUENCY WTCHING POWER SUPPLIES ME WiWT SECTION: RECTIFIERS. INDKTORS, AND CAPACITORS 153
sequently, because of the polarity of winding N,, transistor Qgremains o# Ferrite cores, on the other hand, have to be gapped because of their low
throughout the charging and terminating cycle, diode Dl is reverse-biased, saturation flux density Bw, they are more temperature sensitive, and they
and the input capacitance holds its charge. In this state, Q is fuHy enhanced tend to be bulkier. But if pot cores are used for output chokes, radiated
allowing all the I, current to flow through it. Due to the low Rh,, of the EM1 will be reduced because of the inherent shielding properties of the pot
FET, Q1is acting as a high-efficiencyrectifier diode. The stored magnetizing core. Also, ferrite chokes are easier to wind, especially if heavy-gauge wire
current in the transformer T1,if unopposed, will cause the windings to is involved.
reverse their polarity,
When this happens the base of bipolar transistor Q2becomes positive,
turning it on, thus discharging the FET input capacitance turning Q1off,
which in turn blmks the flow of I,. Resetting of the transformer core takes
place at this time also. The baseemitter drop of Q, clamps the reset and Consider the output section of a PWM halfhidge converter depicted in
guarantees that Q2 is conducting during the critical high d d d t recover)r Fig. 615a.The output waveforms E h and E,, are shown in Fig. 6-15b,as
interval of Q1.This prevents false turn-on of Q1under any circumstance. well as the average load current Imt with ripple AI riding on it.
The absence of any external control or timing signals makes this circuit From basic electrical theory, the voltage across the inductor is given by
a true two-terminal rectifier. Figures 6-14b and c show how this rectifier
circuit may be implemented in a full-wave center-tapped rectification scheme
or in a single-ended scheme.
Since
6 4 OUTPUT POWER IWDUCTOI DESIGN and

4-44 General Considerutlons
Most switching power supply designs use an inductor as part of their output then Eq. 6-7 may be written, solving for L, as follows:
filtering configuration. The presence ofthis inductor is two-fold: fikst it stores
energy during the off or "notch" periods in order to keep the output current
flowing continuousIy to the load, and second it aids to smooth out and average
the output voltage ripple to acceptable levels.
There are a variety of cores that an engineer can use in the design of In the case of the PWM half-bridge or full-bridge converter, the voltage E ,
is roughly twice the value of the output voltage E,, at the maximum primary
inductors. the most popular materials used in present-day hi&-frequency
switching designs are ferrite cores, iron powder cores, and molypermalloy input voltage Vim(see Fig. 6-15),Therefore, El, - E,,, = E,. The time
interval A t is equal to the maximum dead time, or "notch" time, t&, which
(MPP) cores. Ail of these cores are good for power inductor designs, and
basically the criterion of choosing one vs. the other is based on factors such occurs between alternate switching half cycles.
as cost, weight, availability, performance, and ease of manufacture. \laximum td occurs at maximum input line voltage, since the bansistor
Iron powder and MPP cores are generally offered in toroid forms, and conduction time t, is at a minimum. Therefore, the inductor must be de-
they are we11 suited for power chokes because of the following characteristics: signed to store enough energy to provide continuous output current during
the notch periods.
1. High saturation flux density B,, up to 8000 G. Expressing At in terms of secondary voltage E , and E,,, yields
2. High energy storage capability.

3. Inherent air gap eliminates the need of gapping the core.
where f is the converter's frequency in kilohertz, while the factor 1/2 relates
4. Wide choice of sizes. the notch time tor to the entire switching cycle, since the total switching
I TI.IE SECTION: RECTIFIER% INWCTORS. AND CAPAClTORS 155
period encounters two notch time intervals. In order to keep low inductor
r - - - c u r m t and goad output ripple, it is recommended that AIL not be
neak
greater than 0.251,~.
Based on the above, Eq. 6 7 may be rewritten,as follows:
I
Eq. 6-10 gives an inductance value which will be very close to the practical
value, and it may or may not require fine tuning in the actual application.
After the inductance has been &dated, the core size and core materid
has to be chosen in order to complete the design.
The EoHowing examples show step-by-step design procedures using a
ferrite core and an MPP core The first design procedure is analytical, while
the second is graphical. Both methods are equally useful for designing.op-
timum filter chokes.
EXAMPLE 6-5
For a D k H z , 100-WMf-bridge power supply having an ou$put of 5 V dc
at 20 A, calculate the output inductor L using a femite mre.
SOLUTION I
Using Eq. 6-13 to calculate the maximum notch period,
!
The inductor value L capable of delivering output current AIL dudng tor is
\
! Select the minimum size of core using the following equation
1 (5.067) 108 /LI,Dq
A,A, =
m,
where K = 0.4 for toroids and 0.8 for bobbins
D = diameter of wlre to be used
A* = core effective area
A, = bobbin winding area
h
1
A',"t EJ
FIGURE 6.15 (a) The output ~ c t ofh 4 PWM bdf-bridg converter, and (b,C) its
associated voltage and current wavefms.
i! Choose a current density of 400 c.m. /A. Then for 20 A, the wire is 400 x
20 = 8000 c.m., which corresponds to no. 11 AWG wire, with a maximum
diameter of 0,0948 from Table 5-2.
4 5 6 HIWREQUENCT SWITCHING POWER SUPPlIES 1
I
THE SECTION: RECnFIERS, INDUCTORS. AND CAPACITORS 157
Also selecting a B , = 2000 G, the A,A, product is

EXAMPLE 6-6
Design the filter choke of Example 6 5 using an MPP core.
I SOCUTTON
Fmrn ferrite catalogs the 3019 pot core has A, = 1.38 cm2 and A, = 0.587
cm2, which yields &A, = 0.81 cm4.This is good enough for our require- i ~1though
'*\
the'design of a fi1tf:r choke using MPP mres may be analytically

ments, but we may have to use a larger core in order to fit the Iarge-gauge 1 accomplished, this example presents a quick graphical method, developed
wire. In fact, it is advisabIe to use a bundle of smaller wire size to increase
the conductor surface area and reduce skin-effect Iosses. For this example,
eight no, 20 AWG wires wi1I have to be used in place of one no. 11 AWG
1; by Magnetics, Inc., which is fast and accurate. The example is based on
Magnetics, Inc., MPP core data, but many manufacturers make equivalent
i
core sizes with similar magnetic properties; thus the reader may extend the
wire. By using bundled conduetors in parallel, copper losses due to reduced i described method to the manufacturer of choice.
12R losses are minimized; therefore we could reduce the current density 1
requirement, Using only six no. 20 magnet wires, the current density will Step I : Calculate the required inductance. Using Eq. 6-14,
be 300 c. m. /A, still an acceptable vdue. With all of this in mind, the 3622
core and single-section bobbin is chosen for this design.
Because the inductor experiences a large dc bias, it is necessary to gap
the core in order to avoid saturation, The length of the gap is Step 2: Calculate the product of LIZut.Taking L = l 2 pH and I,,, = 20
A, ,then LIZ = (12 pH)(2w = (0.012 mH)(202)= 4.8.
Step 3: Select core size. Fmm the core selector chart of Fig. 6-16, locate
the L12 point of 4.8; following this coordinate the first core sise
encountered that falIs within the solid line permeability family is
size 55548.
Since the air gap interrupts the magnetic circuit twice, if a spacer is I
used to provide the gapping, the spacer thickness will be 1,/2 = 0.0373. i Step 4: Select the permeability. The intersection point of the 4.8 coordinab
On the other hand, the total gap length must be used ifonIy the center leg 1 and the 5554-8 core size coordinate Mls between the 26 p and 60
is gapped, p curves. OnIy those permeabihty lines which intersect the LIZ
Now the number of turns may be calcuIated. mrdinate (4.8 in this case) below the core intersection may be
used. In this case we will use a core with a permeability of 60 as
N = -
Brnal, -
- 2000 X 0.0746 a first try, If a higher permeability core (such as p = 125) is used,
= 5.94 turns
O.4nIwt 0.4 x 3.14 x 20 this choice will yield a lower winding factor, therefore fewer turns.
We will use six turns. Using six no. 20 wire conductors in parallel, the Step 5: Calculate the number of turns to obtain the required inductance.
equivalent of 6 x 6 = 36 turns wodd be required, The number of turns can be calculated as follows:
The turns per bobbin graph (refer to a Ferncube catalog) on a 3622
single-section bobbin data sheet shows that approximately 60 turns of no.
20 wire will fill the lmbbin winding wea, Taking into account air space around
the round wires and tape finish, we conclude that the 3622 bobbin and pot
core is a good choice for this design. Improvements maybe made in the where L is the desired inductance (in millihenries), and Llowis the
actual application, which could include increasing the number of turns for nominal inductance (in milIihenries per 1000 turns).
better filtering or increasing the number of conductors to reduce heating From Table 6-1 or Table 6-2, we find LlW = 61 for the S548 core
effects. with a permeability of 60 p. Therefore the required number of
*
TABLE 6-1 ELECTRICAL MECHANICAL, AND WINDING INFORMAmON
DATA ON MAGNETICS, INC., 5 5 3 8 MPP F A M U CORES
--- DIMENSlONS AFTER FINISH
CnRE
1.332 in. 33.80 mrn
0.760 in. 19.30 mrn
0.457 in. 11.61 m m
WINDING W R N LENGTH
Length1 turn
0.1943 ft 5.93 cm
0.1668 fi 5.09 m
0.1400 fi 4.27 cm
0.1282 fi 3.91 crn
0% 0,1238 ft 3,78 crn
W l d w #a 577 8 0 c m .
C m s t i o n 0.1 ~2 in.' 0.672 cm2 WOUND COIL DIMENSIONS
htthlength 3.21 in. 8-15 cm Unity loindingfobor
Weight 1 . 7 47.W 1.840 in. 46.7 m m
Dl3 (Max.)
HT (Max.) 1.IU3 InI 28.0 mrn
MAGNETIC INFORMATION
Inductance N a m i t ~ l d c fifii9he~ Grading BlNI
part p-, @ 1000 hrme, &am, s t Q ~ ~ - a% G m s .per
p MN*8% aIMH tiom 2%barsds amp. b r n
55551 - 14 14
28
0.335
0.167
A2
A2 ** -2.16(<1500C)
4.00 (C1500 G )
55550- 28
55071- 60 61 0.0768 ALL Yes 9.24 (C15130 C)
553- 125 127 0.0369 ALL Yes 19,3 (<1500G)
55547- 147 150 0.0312 A U Yes 22.6 (41500G)
5j546- 160 163 0.0287 AJdL Yes 24.6 (<1500 G)
55542- 173 176 0.0266 ALL Yes 26.6 (<I500 G)
55545- m 203 0.0230 ALL Yes 30.8 ( ~ 6 0 0 9 )
55543- 300 305 0.0153 A2 and L8 Yes 46.2 (<SO0 G)
555.01- 550 5 59 O.OU83 A2 Yes 84.7 (<?0 C)
WINDLKC; INFORMATION
for unfty winding her
AWG Rdc AWC Rda
rPire&3 Turn a d m size Turns il
8 32 0.00393 23 889 3.50
9 40 O.OD618 24 1100 5.49
10 50 0 . ~ ~ 6 25 1359 6.56
M1for MPF cores. L = inductance with dc bias (in milWlFnriesb I -

FIGURE 6-16 DC bias core selector chart, relating core r b , p m e a b i l i v , HIN1, and
dc current (in
11
12
13
14
63
79
99
123
0.01544
0.0244
0.0384
0.W
26
27
28
29
1699
2139
2625
3209
13.53
21.4
33.3
51,3
15 154 0.0949 30 41111 81.1
16 193 0.1504 31 4937 125.7
17 239 0.234 32 MI17 189.4
turns to obtain an inductance of 12 pH (0.012 mH) is 33 7463 299
16 2% 0.370
19 370 0.579 34 9500 482
N = 1000 -= 14 turns 20
--22
91
432
578
0.909
1.437
35
36
11,788
14,549
758
1173
713 2.24
source: Courtesy of Magnttics, Inc.
Increasing the number of turns by 20 percent we get N = 17 Nms. . . ~ qw
i
160 HIGH-FWENCY SWITCHING POWER SUPPLIES
TABLE 6-2 INDUCTANCE TABLE of 24 percent the number of turns which will fill the core will be
IRductance pw law) turns, nlH
239 x 0.24 = 57.36 turns. Sfnce our design requires 68 turns, a
Part AO.J 55548 with higher permeability should have been chosen.
125p Ilp 16p 60p 125p l47p 160p 173p 2 0 0 ~ 300p 5SDp
Choosing the next higher permeability of 125 p from Table
55140 NA NA NA % 31 33 36 42 62 NA . 6-2,Llooo= 127.Thus, step 5 is corrected to give the necessary
55150 4 7 17 35 41 45 48 56 84 NA ., number of turns to achieve the required inductance.
-
55180
55
O B
55240
55270
5
6
6
12
9
10
11
21
20
24
26
50
42
50
54
103
49
59
H
122
53
64
69
132
57
69
75
144
67
80
86
165
99
120
130
247
NA
220
242
466
N = 1000 Jo* = 9.72 turns
55030 6 11 25 52 62 66 73 83 124 229 Increasing the numbek of turns by 20 percent, N = 12 turns of

55280 6 11 25 53 63 68 74 84 128 232 no, 11 AWG wire is required, or 12 x 4 = 48 turns of four no.
55290 7 14 32 66 78 34 92 105 159 290 17 AWG wires in parallel, which will fit the chosen core.
55040 7 14 32 66 78 84 92 105 159
53130 6 11 26 53 63 68 74 85 127 NA In order to check the results, the following andysis may be used.
55050 6.4 12 27 S6 67 72 79 W 134 Step J : Calculate the dc kagnetidng force. From Table 6-1the 55x8 core
55120 8 15 35 72 88 92 104 115 173 317 shows HIN I = 0.154. Then, the magnetizing force will be
55206 7,8 14 32 68 81 87 96 104 163 320
55310 9.9 19 43 90 106 115 124 144 216 396 H
55350 12 22 51 105 124 135 146 169 253 NA H = - (NI)= 0.154 x 12 x 20 = 36,98Oe
NI
55930 18 32 75 157 185 201 217 251 377 740
55x8 14 28 61 127 150 163 176 203 305 559 Step 2: Check permeability reduction. From the curves of Fig. 6-17, the
55585 9 16 38 79 93 101 109 126 190 348 125-p materid at 3698 Oe experiences a reduction of 30 percent
55324 13 24 56 117 138 150 162 187 281 515 of its initial permeability; therefore, the percent of useful perrne-
%
5523 19 35 81 163 198 215 233 269 403 740 ability is 70 percent.
55438
55089
32
20
59
37
135
86
281
178
338
210
360
228 246
390 450
285
674
427
NA
NA Step 3: Find the inductance of the core at the permeability of step 2. The
'
55715 17 32 73 152 179 195 210 243 365 NA nominal inducatnce of the 55548 core for 125g material is 127 nfEl
55109 18 33 75 156 185 NQ 318 250 374 NA
55866 16 30 63 142 NA NA NA NA NA NA
Note: Relates Magnetics. Inc.. MPPcore number to inductance per 1WO hrrns at various permeabil~tyvalues.
Swm. Courtesy of Magnetks. Inc.
Step 6: Calculate wire size and fit. If we choose a current density of 400
c.m. /A, then a 400 c.m. /A x 20 A = 8000-c.m. wire is needed.
This correqmds. to no. 11 AWG wire fmm Table 5-2.
To reduce skin effect losses, four no. 17 AWG wires in parallel
will be used, the equivalent of 17 x 4 = 68 turns of single no. 17
wire. To check for fit, 68 turns of no. 17 wire (2050c.m.) equals
139,400 c.m. From Table 61, the 53548 family of MPP cores has
a total window area of 577,600 c. m. Therefore, the winding factor
of this core is equal to 139,NO/517,600= 0.24. Also, from Fig. DC Magnetizing f w , oerstads
6-15, the winding infomation data show that a fully wound core FIGURE 6-17 Permeability w. dc bias mnes for Magaetics, Enc., MPP cores. (COURCS~
will accept 239 turns of no. 17 AWG wire. For a winding factof of h f a g n e t b , Inc.)
162 HIGH-FREQUENCY SWITCHING POWER SUPPLIES THE SE&N: RECTIFIERS, INDUCTORS. AND CAPACITORS 163
I
per 1000 turns, At 70 percent permeability the nominal inductance
becomes 127 x 0.70 = 88.5 mH per 1 0 0 turns. Solving for L,
Eq. 6-15 yields
Therefore the minimum inductance of 12 pH has been achieved.
6-5 MAGNETIC AMMIAERS

The magnetic-amplifier technology has been around for a long time, but
renewed interest has emerged lately, especially with the proliferation of the
switching regulator, as an efficient means of regdating the auxiliary outputs
of a multiple-output switching power supply, ,
The term "magnetic amplifier" is really misleading, since the circuit nei-
ther amplifies ;lor does it use any amplifier as its main switching element.
Rather, an inductive element is used as a control switch. Thus a magnetic
amplifier is a reactor, wound on a core with relatively square B-H charac-
teristics, This reactor has two distinct modes of operation: When unsaturated,
it acts as an inductance capable of supporting a large voltage with very little
or no current flow, and when saturated, the reactor impedance drops to
zero, allowing current to flow with zero voltage drop.
Themagnetic amplifier in itself is a pulse-width-modulated buck regulator
and requires an output LC filter to convert its PWM ou~putto a dc voltage.
Since it is a buck regulator, it can only lower the output voltage from what
it would be with the regulator bypassed.
Magnetic-amplifierusage is universal, that is, they work equally well with FIGURE 6-18 A mgneti~ampHfierregulator and its associated waveforms.
any converter topology such as forward, flyback, and push-pull and their
derivatives. At time f = 10+ ps, Vc switches positive again, and the core should be
driven into saturation. But since the core has to be reset by a volt-seconds
product equal to thedamount which took it out of saturation, that is, 60
6-5-1 Operation of the Magnetit Amplifier V-ps, the mag amp delays 'the leading edge of the waveform by 5 ps.
Figure 6-18 shows a simplified magnetic amplifier and its corresponding This condition remains until the voltage across the core drives the reactor
waveforms. Let us examine the operation of the circuit. into saturation, delivering a Sps-long output pulse. In mathematical terms
Assume that N, is the secondary winding of a transformer driven from a we have the following relationships:
square wave which produces a 2 12-V waveform at V,.At time f = 0- , the
reactor & is saturated and V 2 = V3 = i2 V'(diode drops in this discussion
are*ignored for the sake of simplicity, but they should be considered in an
actual design). At time f = O+ to t = 10 ys, the reactor hw a voltage across Therefore the output voltage will be
it of Vk = -12 - (-6) = -6V, since Vc = -6 V. Thus a reset current
flows from Vc through D,to the reactor for a period of 10 ps, driving the
core out of saturation and resetting it by an amount equal to 60 V-its,
164 HIGH-FREQUENCY SWnCHlNG POWER SUPPLIES
1 THE mm O N : r n H E R S , WWCK3RS. AND CAPACIIOrn 165
!
~t is important to note that the reset current of a magnetic amplifier is TABLE 6-3 M A T E W COMPARTSON
determined by the core and thp number of turns and not by the load current.
Hence, a few milliamperes of magnetic-amplifier reset current can and will
control many amperes of load current. In fact, magnetic amplifiers are cost- I
effective and eficient at outputs with bad currents over 2 A. \,
6-5-2 Design of the Magnefic-Amplifier Saturable Reactor

Designing the saturable reactor of a magnetic amplifier requires three steps:
First, determine the withstand A: second, choose the appropriate core; and
third, calculate the number of turns of the reactor. Let us look at each
individual step separateiy.
Step 1: Assuming that the output inductor has been designed for continuous
conduction, the reactor must be designed to delay the kading edge
of the input waveform to the filter inductor long enough to provide
the required output voltage. This excluded poi tion of the pdse is
equivalent to area 8,as shown in Fig. 6-18, and it is defined as the
withstand A in V-p. Hence, !-
cThe values IIstcd are typical oF.002" thick m a t e d ( M d a s ,001")f the types shown. For -teed characterlstb
on all thicknesses ofad alloys available, rekr to Magneticr 1nc.SGuaranteed Tape Wound C? Cham&ristics Bulktin,
A = Vt 1 which is available upon request, fmm the Campanents Saks l k p t m e n t , Wagnetla lnc.. Butler, Pa.
where V = pulse amplitude, V **400~ycbCCFR Coerdwe h r c c k defined as the H1 reset m e r i s t h describad by the Constant Current Flux
Reset Test Method In AlEE Paper #W.
t = leading-edge delay, ps
***Gain Is the eOo cyde are. Gain described by the Constant Current flux Reset Test Method per AIEE P a p a #%932
For practical considerations one should daIb a 20 percent increase For corcs with IDlOD of -75 to .%. 4
over the required withstand to accommodate an equivalent increase @
Swm:Courtesy d Hqnebia, Inc.
&
or decrease in anticipated load current changes.
St@ 3: Determine the appropriate number of turns using the formula
Step 2: Select the wire size based on output current. A practical value of
500 c,m. / A (circular mils per ampere) is an acceptable design rule.
Next choose the core material to determine the saturation flux
density B,,. For best results choose a core material with an outside
diameter of 1 mil or less. Table 6-3 provides several commonly and estimate the control current using the formula
used materials, and it is presented here as a guide to the reader,
Choose a fill factor K using values from 0.1 to 0.3, with the lower
values for large wire sizes Next calculate the core-size based upon
the area product: where H = Magnetizing force, Oe
1, = Magnetic path length, cm
To better understand magnetic-amplifier operation, the following example

where W, = core window area, cmZ gives the reader all the necessary steps to design efficient mag amps.
A, = effective core area, cm2
A, = wire area, cm' EXAMPLE 6-7
A = required withstand, V-s 1-
Consider the forward converter given in Fig. 619 and its associated wave-
.. . Bmaf = core saturation flux density, G $
1 ..
forms. Deslgn the magnetic amplifier to produce a regulated 12-V, auxiliay ,
Kwww.IranSwitching.ir
= fill factor i output voltage at 8 A. Switching frequency is 100 kHz. .
166 HGH-FREQUENCY SWITCHING P O W E R SUPPLIES
SOLUTION
Assume that the pulse height at V, is 40 V, at 100 kHz switching frequency. 1 Feedback f
To produce 12 V at the output, the average vaIue of the voltage at V, must
be 12 V, and the required positive pulse width will be
Assuming a "dead time" at the switching waveform of 2 ps gives a 4 9 s input

pulse width, as shown in Fig, 6-19, for a total period of 10 ps (100 kHz).
With a 4-ps-wide pulse at V,,the saturable core LCmust delay the leading
edge by 1ps, to achieve the required 3-ps pulse width. The core then must
withstand
A = Vt = (40)(1) = 40 V-ps
During reset of the core, the withstand must be equal during the negative
half cycle. The value of the reverse voltage required to achieve equal with-
stand integrals is
Hence, the negative half cycle of the waveform will be clamped by diode
Dlat -40 - ( - 10) = -30 V, as depicted by the V3 waveform of Fig.
6-19.
Step 1: Raving calculated the withstand A to be 40 V-,us, we allow

a 20 percent tolerance for a total maximum withstand of
48 V-ps.
Step 2: To caIculate the required wire size to be used in the saturable
reactor, the root-mean-square f m s ) value of the output current is
caIculated. Thus with a duty cycle of 8 = 12/40 = 0.3, the root-
mean-square current is
Using 500 c. rn. / A, the wire area Is calculated to be 500 c. m ./A x

4.4 X = 2200 c.rn. From Table 5-2, the cIosest wire gauge cor-
responding to this value is 16.
For this example we choose as the core material a Square Permalloy 80,
which has a very Iow coercive force and a very square B-H loop. with a FlGUBE 6-19 4 100-kHz fornard mverter using a magnetic amplifier and its as- ,
B,, = 7000 G {see Table 63). Thus we can now calculate W,A,from Eq. sociated waveforms, for Example 6-7.
THE O W SECTIMV: RECTIFIERS. INWCTORS, AND CAPACITORS 169
Note that since the wire size is 16 gauge, which is relatively large, a fill
factor of 0,1has been chosen as a sensible compromise.
Since the W,Aicalculated above must be at least Om00754X 10s c.m.
cm2, and because the converter frequency is 100 kHz, a core with a tape
thickness of 0.0005 in. (4 mil) will be used.
Table 6-4 lists a choice of high-frequency mag-amp cores. From the table,
the logical candidate for this application is the 50B10-5Dcore. Knowing the
type of core, the number of hrns required may be easily calculated using
Frequency (cyclesper second)

FIGURE 6-20 Average MMF required to saturate Permalloy 80 vs. fre-
quency using s q u a r e - w e current drive. (Courbeoy iCiognetica, Inca)
-.. k
170 HIGH-FREQUEKV SWITCHING POWER SUPPLIES WE SECTION: RECTIFIERS, INWCTORS. AND CAPACITORS 171
Eq. 6-18, Hence

A(l@) x 4 8 ~ 1 0 - ~ x l @
N = = 6.7 turns
2(B,)A, 2(7000)(0.051)
Round of to 7 turns,
Completing the example, the magnetizing current is calculated, in onl&
to estimate the current required to reset the core when designing the control
circuit.
From Fig, 6-20 the magnetizing force at 100 kHz for a $-mil core is H =
0.215 Oe. Thus the control current will be
FIGURE 6-21 A transistor-had mametic-amplikr control circult.
Note that the circuit in this example was designed for regulation only. When that is, it prevents the magnetizing current of the reactor from raising the
independent shutdown is added for short-circuit protection or turn-off from output voltage at no-load 'conditions.
an external logic signal, then the required withstand is the area under the The feedback network Z5and 2,can be designed, using techniques dis-
entire positive input pulse. .In this example, A = (40 V)(4 s) = 160 V-ps, cussed in Chap. 9, to stabilize the loop (see Example 68).
and all the above calculations must be repeated based upon this withstand Full-wave saturable core regulator circuits are also simpk to design,
value. Figure 6-23 shows how a circuit can be used to control a full-wave output,
As mentioned before, the magnetic amplifier is a buck regulator, that is, such as the outputs found in push-pull or bridge-type conveffers. Notice
its output voltage will always be lower than its input voltage. Since the that only one control circuit is used to reset the two separate saturable cores.
power losses of the magnetic amplifiers are smdl, these circuits are proving
to be ideal candidates for deriving low-voltage, high-current outputs, that
is, 3 V or lower, at relatively high dciencies. The applications of magnetic 6-5-4 The UC1838 Magnetic-Amplifier Conlroler
amplifiess to switching power supply design are countless and may be limited
Unitrode Corporation has introduced the UC1838 single-chip integrates
only by the creativity and imagination of the desigh engineer.
circuit, which provides all the necessary parts to implement a high-perform-
ance magnetic amplifier with or without independent current-limit control,
6-5-3 Control Circulis for Magnetlc Amplifiers Figure 6 2 4 shows the block diagram of the UC1838,
Figure 6-91 shows a fundamental magnetic-amplifier control circuit regu-
lating the output of a forward converter. In this circuit, reset of the saturable
core is accomplished by tsansistor Qland its associated components. Thus,
during the negative half cycle of the source voltage, transistor Q1conducts
allwing a current to Row through R1 and D ,resetting the saturable core
LC,The reset RYX level is controlled by the transistor equivalent resistance,
which is a function of the difference behveen the output and reference
voltages.
The drawback of this circuit is its temperature sensitivity and its tendency
to oscillate at extreme load changes. An alternative improved control circuit,
which eliminates the above problems, is shown in Fig. 6-22. Resistor RE
degenerates the transconductance of the transistor, making the transfer hnc-
tion independent. This circuit also has the feature of acting as a "'preload," FIGURE 6 2 2 An & p d magnethmpliller mhd umuit,
ME W l F W SECTk3N: RECTIFIERS. ZNDUCTORS, AND CAPACITORS 173
FIGURE 6-23 A Full-wave rnagnetieamplitier cmtd drcult.

FIGURE 6-25 A maghetic-amplifiercontrol circuit, with current limiting usiag the UC1838
c o n t d , for Exampk 6-8.
Hence, the integrated circuit includes the following basic functions:
1. An independent, precise, 2.5-V reference. This reference is a band gap 2. Two identical, high-gain operational amplifiers. The amplifiers have an
design, accurate to within 1 percent, operating From a supply voltage of input range of - 0.3 V to V,, and a current-sink capability of greater, than
4.5 to 40 V. 1ma, with a slew rate of 0.3 V-ps. The amplifiers have a gain bandwidth
of 800 kHz and may be cascaded if greater loop gain is required, I
3. A high-voltage PNP reset current driver. The driver has the capability of
delivering up to 200 ma of reset current with an 80-V collector voltage
wing. With internal emitter degeneration, the reset drive operates as a
transconductance amplifier providing a reset current as a function of input
voltage. A mag-amp circuit using the UC1838 IC is shown in Fig. 625.
The circuit also provides independent current-limit control.
The following example gives the reader enough information to design the
feedback components of the magneticamplifier controller. In order to un-
derstand the concepts presented here, a knowledge of Chap. 9 material is
necessary, since extensive reference is made to the K-factor techniques for
the amplifier loop analysis.
Consider the forward converter shown in Fig. 6-19. The following specifi-
cations apply:
www.IranSwitching.ir AC input: 90 to 135 V or 180 to 265 V
FIGURE 6-24 The UC1838 mag-amp control IC. (Cotrrte.y Unitsr~Ic
r#,",#,"#,*f#,"
1
174 HIGtlQREQUENCY SWITCHING POWER SUPPLIES THE W SECTION: RECfFlEIZS, INDUCTORS. AND CAPACITQl?S 175
This is a very low ESR; therefore, we have to adjust the value of the capacibr
to achieve the desired results. We will use 10 times the calculated capaci-
Switching frequency: 100 kHz tance, that is, 125 pF, by paralleling tantalum capacitors to achieve the
desired vdue. Consider that a Unitrode UC1524A PWM integrated circuit
Design the control loop op amp, fur the magnetic amplifier of Example is used. In this IC, a control voltage V, is compared with a sawtooth ramp
6-7, at the 12-V output, using a Unitrde UC1838 mnignetic-amplifier con- voltage V, (2.5 V) to establish the PWM drive to transistor Q1.For the
trol IC, foward converter, only one of the two alternating outputs of the IC is used,
in order to limit the duty cycle S to 50 percent maximum and therefore
SOLUTION allow for transformer core reset. Hence,
The schematic of both the current-limit control amplifier and the mag-amp . v,
amplifier control is shown in Fig. 6-25. Using the off-the-line rectification S=--o.5vc - O.SV,
= -
scheme depicted in Fig. 2-1, the dc input voltage applied to the switching v, 2.5 5
transistor of the forward converter at low line will be 252-52 V dc ripple Since the f o m r d convertqr is a member of the buck regulator family,
and the rectifier drops = 200 V dc. the output voltage is related: to both the input voltage and the duty cycle
Calculate transformer primary turns, using the 3622-PL00-3C8 Femx- by the equation
cube pot core, which has an A, = 2.02 cmZ.
(vin,mtn)l@ - 200 x 108
= 28 turns
Np 2fBmAC 2 x 100 x 1V x 1.8 x 16 x 2.02
In order to get a secondary voltage V1 = 40 V, as shown in Fig. 6-25, we To derive an expression for the gain of the 12-V output, the above formuh
need the 12 V, secondary turns, NsP, to be is differentiated with respect to V,:
The filter inductor L1calculation is based on the maximum off time, The
value ofLl is calculated as follows. From Eq. 6-13: Gaind, = 20 log 16.3 = 24.2 dB
The corner frequency of the 12-V output filter is
From Eq. 6-14:
Figure 6 2 6 shows the Bode plots of the output filter transfer function and
its phase shift.
If we choose an output ripple voItage of 0.2 V, the required output capac- Choosing a type 3 amplifier {see Chap,9) and the unity gain crossover
itance may be calculated using Eq. 6-21: frequency to be one-Nth the switching regulator's clock frequency (that is,
20 kHz), we can determine the desired gain and phase boost of the feedback
amplifier. From Fig. 6-26 by inspection, at 20 kHz the gain ofthe modulator
is 0.7 or -3 dB and the phase shift is 158".
The ESR of this capacitor must be As recommended in Chap. 9, at least 6U0 of phase margin is required at
0.2 the crossover frequency. Therefore the phase boost from Eq. 9-43 is
ESR = -
12 = 0.016 n
I WE OUTPUJ SECTION: AECTlHERS, INDUCMRS. AND CAPACITORS 477
The gain at the unity gain crossover frequency of 20 kHz is the recipmcal
of the modulator's gain, that is, + 3 dB or G = 1.4 (see Fig. 6-27),
Using Eqs. 9-49through 9-54 and assuming R1 = 10 kn, the required
feedback-amplifier components are calculated as hllows:
ID
1 - 1
= 1.5 kHz
?-3
FIGURE 6-26 Bode pht of the filter transfer function and phase shift, for Example 68.
= ZnRl(C,+ C,) - (6.28)10 x lCP(0.0106) x lom6
Figure 6-27 shows the plot of the amplifier transfer function. The double-
zero and double-pole frequency location is
1
f , = - =f 20Xlol
b'x m = 4.6 kHz
and
f, = fa= (20 x 1
03)
m= 86.7 kHz
Figure 6-27 shows the Bode plot of the feedback amplifier, The gain
bandwidth product of this amplifier is
GBW = KGf = (18.8)(1.4)(20,000) = 526.4 kHz
Since the GBW of the UC1838 is specified at 800 kHz,the requirements of
the design can be met with this amplifier, The overall system loop gain Bode
plot is shown in Fig. 6-28.
The current-limit section is implemented by using a series-senhg resistor
of 0.01 fl, 1 W, precision wirewound. This circuit will provide adequate
overload protection. In circuits 4 t h high current outputs, the additional
. . efFiciency loss in the sensing resistor may be avoided by providing a current-
FIGURE 6-27 W e plot of the feedkck*arnp2ffier characteristics. sensing transformer to drive the input of the current-limit amplifier. In this
.-
178 HW-FREWENCY SWlTCHlNG POWER SCIPPLIES THE W T P U SCION: RECTIRERS. tNDUCTDRS. AND CAPACITORS 179
film capacitor current capabilities of 2 AlpF at a working frequency of IQO

kHz and above. .
Regardless of the type of capacitor used for output filtering, the following
analysis pertains to the calculation of its value. Referring to Fig. 615c, the
current wavehrm in the output capacitor C, is centered about zero and
has an amplitude of PI. Notice that the current waveform crosses the zero
reference in the positive direction at f,, which i s the middle of the on time,
while at t:, which is the middle of the off time, it crosses the zero reference
in the negative direction. Thus the current will produce a ripple voltage AV
which is given by Eq. 6-20.
But the average current during.the time interval tl and tz is (A1,,12)12 or

AIOut14.Therefore, integrating Eq. 6-20 yields
where T is the total period of the on time t , and the off time t,.
Rearranging terms, the minimum output capacitanw is
FIGURE 6-24 O v e r d system loop gdn Bode plot.
case the current-sensing transformer should be placed between point V, and

where I,, = 0.25 I'; IL = specified load current
the anode of the diode i?;.
AV,, = allowable peak-to-peak output voltage ripple
f = operating frequency
6 4 DESIGNING THE OWPUT FILTER CAPACITOR In order to ensure minimum output voltage ripple, the ESR of the capacitor
The choice of the output filter capacitor depends upon the type of converter may be calculated by the following relationship:
being used as well as maximum operating current and switching frequency. AVO,,
I
Most of today's applications use electrolytic capacitors, preferably of the Iow ESR,, = - (6422)
AIOUt
j ESR type. The ESR of the filter capacitor has a direct effect on the output
It is important to note at this point that proper selection of the LC filter
t ripple and also on the life ofthe capacitor itself. Since the EST1 is a dissipative
element, the power loss in it generates heat, which in turn shortens the is essential, since it influences two important parameters in the perfomance
!
i! of a switching power supply, Fiat, the LC filter combination has a v e v
I+ capacitor's life. strong influence on the overall stabiIity of the switching system, as we will
Modern-day capacitors have temperature ratings of lOSC and very low
't
i ESR at frequencies above 20 kHz. As converter operating frequencies start see in Chap. 9. Second, a small L and large C will result in a low surge
I' - to increase, the majority of capacitor manufacturers offer low ESR electrolytic impedance of the output filter, which means that the power supply e l l have
I
capacitors with guaranteed performance at 100 kHz. With advancement in a good transient response due to load step changes.
i.,
I
passive component technology, the trend at higher than 50-kHzfrequencies In fact, measuring the transient response of a switching power supply,
is the development of film type capacitors which will oEer a high current the important factor may not be how long it takes the output to recover
4. . ' capabiIity, Film type capacitors exhibit extremely low ESR, and they behave during a step load change, but how deeply it deviates from the output voltage
. . much better thanwww.IranSwitching.ir
e1ectroIytics. Already some capacitor manufacturers claim nominal value. For example, a 5-V output may not be suitable for 'ITL if it
THE OUmUT SFCTION: RFCT1FIERS, INWCTORS, AND CAPACITORS 181
480 HIGH-FREQUENCY SWITCHING POWER SUPPLIES
5. Magnerics, Inc.: Output Regulators Uslng Mag Amp Control for Switched-Mode Power
dips more than 250 mV during, say, a 25 percent load change, provided that Supplies, P w r c o n w r s f o n Inkrnatlonal,1985.
this lmd change is antldpated in the actual application.
6. -----"Molypermalloy
: Core Catalog "
7. Middebmk, R. D., and S. Cuk: "Advances in Swftched Mode Power Converston,'\ob.
EXAMPLE 6-9 *I
1 and 2, Teslaco, Pasadena Cdi6, 1#1.
Calculate the capacitance value and the ES1': vdue for an wtput filter ca- 8. Mullett, C. E.: "'Performance of Amorphous Mate&$ in High-Frequency Saturable Re-
actor Output Regulators," HFPC b e e d i n g s , May 1986.
pacitor to be used in the converter specified in Examp12 6-5, allowing a
maximum wtput ripple of 100 mV, 9. -and R, Hirarnatsu: " ' b e n t Advances in High Frequency Mag Amps." HFPC
h e e d i n g , AprIl 1987.
SOLUTION 10. Patel. R.; "Using Bipolar Synchronous Rectifiers Improves Power Efficjency,'"e Power
Sources Conlerence, 1984.
Using Eq. 6-21 we get
11. -
: "Circuit Dacriptlon of a Synchronous PWM ~egulator,'VnitrodeCorp. Appll-
cation Note. 1934.
12, Pressman. A. I.: "Sdtching and Linear Power Supply, Power Converter Design." Hayden,
R d e l l e Park, N.J., 1977,
From Eq. 6-22
13. Unitrode Corp.: "Linear Integrated Circuits Databook," 1987.
14. -. .Application Note U-68A. ,
15. Watson, J, K.: "'Applications of Magnetism," Wiley, New York, 1980.
Although the output capacitance of 312-pF minimum may be adequate

in theory, p m c a l experience shows that it takes a higher value to accom-
plish the required specifications. In fact, a rule of thumb of 300 pFIA
minimum at 20 kHz is n more realistic vdue when electrolytic capacitors
are used.
Using more than two capacitors in p d b l to achieve the required ca-
pacitance and reducing the ESR to extremely low values is also recom-
mended. In any case, careful find circuit measurements and refining of the
design during prototyping will always yield optimum results, and the for-
mulas presented above will give the designer a first-order approximation,
which is good for a start.
1, Archer, W. R.: Current-Driwn Synchronous Rectifier, MotoroIa TMOS P m r FET Design

Ideas, 1096,
2. Hiramatsu, R et al,: Switch M d e Converter Using High Frequency Magnetic Amplifier,
Pouwconwrslon f ntmdaal, March-April 1080.
3, Hnatek, E. R.: "Design of Solid State Power ~upplier,'% ed., Van Nostrand Reinhold,
New York, 1981.
4. Jamerson, C.: "CWaUon of Magnetic hplifikr Post Regulator Voltage Control Loop
Parametea," HFPC Proceedings, Apd 1987.
--
SEVEN
SWITCHING REGULATOR
CONTROL ClRCUlTS
The majority of today's switching power supplies are of the pulse-width-

modulated (PWM) type. This technique varies the conduction Hme of the
switching transistor during the on period to control and regulate the output
voltage to a predetermined value. Although other methods may be used for
control and regulation, the PWM method offers excellent performanct, such
as tight line and load regulation, and stability during temperature variations,
In recent years a number of integrated circuits have been developed,
which include all the necessary functions to design a complete switchi)ng
power supply with the addition of few external components. The purpose
d this chapter is to introduce the reader to some of the techniques and
circuits used to implement the PWM control section of the switching power
supply and to explain how this control is achieved.
7-1 ISOLATION TECWMIWES OF SWITCHING REGULATOR SYSTEMS

The role of an o f the-line regulated switching power supply is twofold. First
it must derive well-regulated low-lwel output voltages, capable of powering
electronic or electromechanical circuits and devices, and second it must
have a high input-to-output isolation in order to protect the user from shock
hazard due to high voltage or leakage currents.
Figure 7-1 depicts two different block diagrams, sh&ng how line isolatior
may be achieved in an o f the-line switching power supply. Blocks sharin~
. common ground are depicted with identical ground symbols, These bloc1
diagrams are universal, and they may be used b r any basic type of switchin!
power supply design, such as half-bridge, full-bridge, flyback, fbrward, etc
184 HIGH-FREQUENCYSWITCHING POWER SUPPLIES
i
isolation version shown in Fig. 7-lb is most commonly used in flyback and
forward type converter designs.
Although many switching techniques can be employed to implement a

switched-mode power supply, the fixed-frequency PWM technique is by far
amplifier, the most popular choice. In a FWM system a square wave pulse is normally
and control
generated to drive the switching transistor on or off. By varying the width
of the pulse, the conduction time of the transistor is accordingly increased
or decreased, thus regulating the output voltage.
The fWM control circuit may be single-ended, capable of driving a dngle
transistor converter, such ys a flyback or forward. If two or more transistors
have to be driven, as with half-bridge or full-bridge circuits, a dual-clannel
Input PWM circuit is necessary. +
+
rectifierti +
and
film
-
-,
1
71211 A Slngle-Ended, Discrete
Component, PWM CeMrol Clrcuft
A very simple closed loop, PWM control circuit may be implemented by
using a small number of discrete components and semiconductor circuits,
as depicted in Fig. 7-2.The function of the circuit is as follows. ,4 clock
pulse generator IC,generates an asymmebic square wave output at a fixed
frequency, i.e, , 20 kHz. This generator may easily be designed using a 555
-
J
- timer, or equivalent clrcuiw. Z
(bl The square wave is differentiated by capacitor C 1and resistor RI to pro-

mGURE 7-1 (a) TransFonner isolation wd (b)optoisoIatim techniques used in off-the- duce a sawtooth waveform which is used to turn d the normally conductive
line switching power supplies. transistor Q ,, The resulting negative going pulse at the collector of transistor
QLis inverted by transistor QO,producing a positive going pulse at the
collector of Q,.
In the blwk diagram of Fig. 7-la, the error-amplified, FWM, and control The low-impedance ougut driver combination of Q3 and Q1is used to
circuit have a common ground with the output rectifiers and filter. Input- switch the main switching transistor Q5on or off, thus transferring energy
to-output isolation is achieved at the power transformer TIand at the driver to the output of the converter through the transformerchoke T1.Regulation
transformer Tz.In general, transformer T2is abase or gate driver. In the is achieved by comparing a portion of the-output voltage, derived by voltage
block diagram of Fig. 7-1S the control circuit and f W M have a common divider R9 and RIP,against a fixed reference voltage VrepAny changes at the
ground with the switching elenlent and the input rectifiers and filter. Input- output due to line or load variations are amplified by op-amp IC, which
to-output isolation is achieved at the power transformer TI and through the drives the photodiode of optocoupler ICB,modulating its light intensity, thus
optoisolator.
-
Both line isolation techniques shown in Fig, 7-1@vevery good perform-
ance characteristics through careful circuit design. Choosin, one circuit vs.
the other is primarily based on economics and-the type of switching power
forcing phototransistor IC,to conduct harder, Consequently the square wave
pulse at the base of transistor QIis differentiated even more strongly, causing
transistors QI, QPIand Q4 to be on for a longer period of time, while
.transistors Q, and QSare switched on for a shorter period. Thus the pulse
converter design. In general, the transformer iiolatior?circuit shown in Fig. width is modulated according to load and line conditlons, stabilizing the
7-la can be used with all different powmcogverter designs, while-the opto- output voltage.
www.IranSwitching.ir . .
The circuit shown in Fig. 7-2 is greatly simplified, and refinements may
be necessary when used in an actual off-the-line swtching power supply
application.
In recent years, a number of integrated circuits have been developed which

include ail the functions necessary to build a P W M switching power supply
in a single package with just the addition of a few external components.
Figure 7-3 depicts the basic building b h k s of a simple PWM controller and
its associated waveforms. The circuit functions ns follows, An op-amp Wm-
pares the feedback signal from the output of the power supply to a fixed
reference voltage VM The'error signal is amplified and fed into the inverting
input of a comparator. The noninverting input of the comparator accepts a
sawtooth waveform with a linear slope, generated by a fixed-frequency 0s-
cillator. The oscillator output is also used to toggle a flip-flop, producing
square wave outputs Q and g.
The comparator square wave output and the Rip-flop outputs are both
used to drive the AND gates, enabling each output when both inputs to the
gate are "high." The result is a variable duty cyde pulse train at channels
A and B. Figure 7-3b shows how the output pulse width is moddated when
the error signal changes its amplitude, as depicted by the dotted lines.
Normally the outputs of the FWM mntroller are externally buffered to drive
the main power switching transistors. This type of circuit may be used to
drive either two tmsistors or a single transistor. In the latter case the outputs
may be externally ORed, or only one channel may be used as a driver.
The merits of such a PWM controller are profound, including the pro-
grammable fixed-frequency linear PWM section with duty cycle
form 0 to 100 percent, adjustable dead time to prevent output transistor
simultaneous conduction, and above all simplicity, reliability, and cost-
effectiveness.
7-3 SOME COMMERC16LIY A Y A I W . MONULrmlC

PWM CONTROL CIRCUITS AN5 WEIR APPLICATIONS
In the early 1970s the switching power supply market started to expand in
the commercial sector, a Eact which brought about the first major attempt
by integrated circuit manufacturers to ofkrPWM m n h l circuits in a single
chip. The first such cimitr to appear in the market were the Motorola
MC3420 Switchmode Regulator Control Circuit and the Silicon General
. SC3524 PUTMControl Circuit, which was destined to h m e the industry
standard. .
. ..
I
C . 1 A . -
SWITCHING REGULATOR COhlTROL CIRCUITS 189
These PWM controllers were and are the heart for a complete switching
power supply design and may be used equally well in single-ended or dual-
channel applications. Shortly thereafter manuhcturers started to introduce
vrefP
more FWM control circuits with improved characteristics and more features.
+ Error The Texas Instruments TL494 PWM Control Circuit is an improved version
of the SG3524, offering features such as adjustable dead-time control, output
transistors with high source or sink capabilities, improved current limiting
~hnal
Fsarlbk control, output steering control, etc,
With the introduction of the power MOSFET, the first PWM control
circuits appeared with totem-pole outputs, capable of driving MOSFETs
directly as well as bipolars, such as the SG1525A and SG1526 series. Besides
offering all the features found in previous control circuits, these new ICs
included additional features such as undervoltage lockout, programmable
soft start, digital current limiting, and operation up to 400 kHz.
Although, as mentioned, all the above circuits may be used in all the
popular switch-mode topologies, recently some companies have introduced
PWM controllers optimized for high efficiency in forward or flyback power
converters. One such circuit is the Motorola itiC34060 PWM controller,
which includes all the features needed to implement a forward or flyback
design using a minimum of external parts.
Another circuit is the Unitrode UC184 series, which promises to have
a major impact on the design of single-ended power converters. The PWM
control1er.includes aH the controi, driving, monitoring, and protectihn func-
tions needed to build a complete switching power supply with just the
addition of a few passive external components, Features of the contro!er
include a low-current, off-line start circuit; built-in protection from over-
voltage, undervoltage, and o v e r c k n t conditions; feed-forward line regu-
lation over a 4: 1 input range; 500-kHz operation, etc,
In the next paragraphs we describe the function of some of the available
PWM control integrated circuits in order to familiarize the reader with the
operation of these circuits. The descriptions presented are purely informa-
tional, and the reader is advised to review carefully all available data sheets
from different manufacturers of PWM control circuits in order to choose the
optimum IC controller for his or her particular application,
The TL494 is a fixed-frequency PWM controt circuit, incorporating all the

(b)
building blocks necessary for the control of a single-ended or dual-channel
FIGURE 3-3 (a) An integrated PWM control ci~cuftand [b)Its asmiakd waveforms. switching power supply. Figure 7-4 shows the internal construction and
block diagram of the TL-494controller. An internal linear sawtooth oscillabr
is frequen~~-~rogrammable by wo external components, R, and C,, con.-
SWITCHING REGUATOR CONTROL CIRCUTS 4#
nected on pins 6 and 5, respectively. The oscillator frequency is deter-

mined by
Output pulse-width modulation is acmmpIished by comparison of the

positive sawtooth waveform across capacitor C T to either of two control
signals. The NOR gates, which drive output transistors Q, and QZ,are
enabled only when the flip-flop clock-input line is in its low state. This
happens only during that portion oftime when the sawtooth voltage is greater
than the control signals. Therefore, an increase,in control-signal amplitude
causes a corresponding linear decrease d output puke width, as depicted
by the timing diagram waveforms d Fig. 7 5 .
The control signals are external inputs that can be fed into the dead-time
control pin 4, the error amplifier inputs at pins 1,2,15, and 16, or the 120-
mV input offset which limits the minimil? output dead time to approximately
the first 4 percent ofthe sawtooth cycle time. This would result in a maximum
duty cyde d 96 percent with the output mode control pin 13 grounded,
and 48 percent with the same pin connected to the reference line. Additional
dead time may be imposed on the output by setting the dead-time control
input pin 4 to a fixed voltage, ranging from 0 to 3.3 V.
The PWM comparator provides a means for the e m r amplifier td adjust
the output pulse width from the maximum percent on time, established by
the dead-time control input, down to zero, as the voltage at the feedbyk
Capacitor Cr
i
Dead-time corn^.^ I
Flip-Rop
clock Input
omut Q2
emitter
Output mode
control
PICURE 7-5 Timing diagram waveforms of tbe T M 4 PWM controller. (Courtesy of.
www.IranSwitching.ir Motorola Semiconductor Produck, lnc.)
:a
r
192 WGKFREQUENCY SWITCHING POWER SUPPUES
pin varies from 0.5 to 3.5 V. Both error amplifiers have a common-mode
input range from -0.3 to (V, - 2) V and may be used to sense power
supply output voltage and current.
The error amplifier outputs are active high and are ORed together at the
noninverting input of the PWM comparator. With this configuration, the
amplifier that demands minimum output on time dominates control of the
loop.
When capacitor CT is discharged, a positive pulse is generated on the
output of the dead-tlme comparator, which clocks the pulse steering Aip-
Aop and inhibits the output transistors Q,and 0,.With the output mode
control pin 13 connected to the reference line, the pulse steering flip-flop
directs the modulated pulses to each of the two output transistors alternately
fbr push-pull operation. Tne output kquency is equal to half that of the
oscillator.
3 Output drive can also be taken from Q , or Q,, when a single-ended
operation with a maximum duty cycle of Iess than 50 percent is required.
3 .-'
When higher output currents are required for single-ended operation, Q,
i
and Q, may be connected in paralIel, and the output mode control pin must
be tied to ground to disable the flip-flop. The output frequency will now be
e q d to that of the osdilator.
Figure 7.6 demonstrates the use of the TL494 in a simple PWM push-
pull converter with current limiting protection.
aI 7-3-2 The UC1840 Rogrammabte, M-Une PWM ControHer
t
Although most of the comlnercially available PWM oontrollers are designed
fbr universal usage, the Unitrode UC1840 family of programmable PWM
controllers was specifidly designed for application as a primary-side, cost-
effective approach to Byback or feed-fonvard designs. Figure 7-7 shows the
o v e d block diagram of the UC1840 PWM controller.
In reference to Fig. 7-7, the UC1840 contains the following distinct fea-
tures:
ij 1. Fixed-frequency operation, user-programmable by a simple RC network

2. A variable slope ramp generator for constaqt vobsecond operation, pro-
viding open loop line regnlation and minimizing, or in some cases elim-
inating, the need for feedback control
41
w
3. A drive switch for low-current start-up, with direct 08-line bias
1 4. A precision reference generator with internal overvoltage protection

5. Complete undervoltage and overcurrent protection including program-
mable shutdown and restart
.~.
- " * .. !
. ...
' '
SWITCHING REGULATOR CONlliu~LIILUII.UIIS 17a
6. A hi&-current, single-ended PWM output optimized for fast turn-of of

an external power switch
7. Logic control for pulse-commandable or dc power sequencing
The following is a discussion of how the UC1840 PWM contml circuit 4'
functions. References to Figs. 7-7 and 7-8 are made throughout the presen-
tation. During the initial power-up, and before the voltage at pin 2 reaches
3 V, the start/undervoltage (UV)comparator pulls a current of 200 FA,
causing an added drop across resistor R,. At the same time the drive switch
holds the driver biar transistor off, ensuring that the only current required
through resistor R , is the start-up current. Also the slow start transistor is
I
on, holding pin 8 of the IC low, thus keeping capacitor CEdischarged.
The start latch flip-flop keeps the undervoltage signal W from being
defined as a fault. The start voltage level is defined as
When the mntrol voltage rises above'zhat level, the startlUV comparator
eliminates the 200pA hysteresis current, sets the start lakh flip-flop to
monitor for an undemltage fault, activates the driver bias output transistor
to supply base current to the power switch, and also turns off the slow start
transistor, providing soft-start start-up which is set by R, and C,.
4
Pin 8 of the UC1840 may be used for both soft-start turn-on and duty
cyde limiting, as well as a PWM shutdown port. The duty cycle may vary
from 0 to 90 percent, and maximum duty cycle limiting is achieved Sy
clamping the voltage on pin 8 with a divider formed by resistors RE and
RDc. When fixed ramp slope operation is employed, resistor Rr is taken to
the 5-V rekrence. Alternatively, for a consQnt volt-second operation, i.e.,
the ramp generator connected as shown in Fig. 7-5, Rs must be mnnected
to the dc input line.
The desired maximum duty cycle is set by the voltage at pin 8, de-
fined by
V (pin 8) = ( R r R+ofim)vmh
This clamping voltage must be equal to the mmp voltage, at the same dc
input voltage level.
The ramp generator 6n the other hand \?rill produce an output ramp
voltage with a slope given by
where V5,,is the voltage to where resistor RR is connected. For a fixed ramp
slope, RR should be connected to the 5-V reference. The peak voltage of
the ramp is damped at 4.2 V, while its valley or. low voltage is typically
0.7 V.
The PWM section of the UC1840 consists of the oscillator, the ramp-,
generator, the error amplifier, the PWM comparator, the PWM latch flip-
flop, and the PWM output transistor, as depicted in the block diagram of
Fig. 7-7. The function of the PWM section is conventional. A constant clock
frequency is established by connecting a simple RC network from pin 9 to
ground and the 5-V reference as shown in Fig. 7-8. The frequency of oscil-
lation is
where RT can range from 1t;C, '100kSt and CT from 300 pF to 0.1 pF.
The basic function of the,ramp generator was described above, The error
amplifier is a voltdge-mode op-amp with a common-mode range of I to
(V, - 2) V. Thus, either input of the op-amp may be directly connected to
the 5-V reference. The other input of the amplifier is sensing an equivalent
portion of the output (or input) voltage which is to be controlled.
The ramp generator output, the error amplifier output, as well as the
slow start input and the current limit output, are inputted to the nFWM
comparator. The comparator generates the output pulse which starts at the
termination of the clock pulse and ends when the ramp waveform crosses
the lowest of the three positive inputs to the comparator. The clock develops
a blanking pulse to keep the duty cycle below 100 percent. The PWM latch
flip-flop ensures one pulse per period rate and eliminates oscillation at com-
parator crossover. The PWM output pulse arrives at pin 12 of the UC1840,
an open collector transistor. The output transistor is capable of supplying
200 mA of output current; thus, it can drive directly bipolars or MOSFETs,
1F higher output current is required, external buffering may easily be im-
plemented, Ancillary circuits such as overvoltage sense, external stop, and
reset are easily implemented.
Current limiting and overcurrent shutdown are implemented with corn-
parators of different thresholds. h the event of an overload, these compar-
ators shorten the PWM output pulse and at the same time turn on the slow
start transistor, discharging the soft-start capacitor, ensuring proper restart
at the end of the fault. -
7 3 3 The UCl524A PWM Control Integrated Clrcult

The UC1524A PWM control .ICi s an advanced version of the first corn-
mercially develci'ped PWM controller, the SG1524. Since many examples
www.IranSwitching.ir contained in."this hook are referenced to the UC1524A, the following de-
t98 HIGH-FRFQUENCY SWITCHING POWER SUPPLIES
scription of its building blocks will adequately introduce the reader to its supply output lines. Its threshold is set at 200 mV, An undervoitage lockout
function. For more in-depth specifications, the individual data sheets should circuit has been added, which disables all the internal circuitry, except the
be consulted. The block djagram of the UC1524A integrated PWM controI reference, until the input voltage is 8 V. This i t i o n holds standby cur-
IC is shown in Fig. 7-9. rent low until turn-on, greatly simplifying the design of low-power, off-line
, An inteinal linear sawtooth oscillator is frequency programmable by a switchers.
resistor RY and a capacitor CT.The oscillator frequency is determined by The power capability of the output transistors is 200 mA, and their.voltage
rating is 60 V. The transistors may be paralleled for increased current ca-
pacity.
As it can be seen, this versatile IC controller can be used in a variety of
and it is usable to frequencies beyond 500 kHz. isolated or nonisolated switching power supplies for a number of applications,
The ramp voltage swings approximately 2.5 V to change the comparator A simple buck reglator application is shown in Fig. 7-10.
output from 0 to I, by comparing it to either one of two control signals, i, e., In this appIication, a wide input range nonisolated buck reguiator is pre-
the error amplifier output or the currenbiimit amplifier output. The error sented, with the addition of a small signal 2N2222 transistor which serves
amplifier input range extends beyond 5 V, eliminating the need for a pair to provide a constant drive current to the output switch, regardless of the
of dividers, for 5-V outputs. Note also that the UC1524A has an on-board input voltage level. Note the 'simplicity of the current-limit function, which
5-V reference which is trimmed to a 21 percent accurzcy. can be implemented using a small sensing resistor in series with the output
Output pulse-width-modulation is accomplished by steering the resuiting bus.
modulated pulse out of the high-gain comparator to the PWM latch along Figure 7-11 shows a low-cost, 50-W of-the-line, fully isolated forward
with the pulse steering flip-flop which is synchronously toggled by the os- converter. At power on, initial start-up is protided by C, yhich charges
cillator output.
The P W M latch insures freedom from multiple pulsing within a period,
even in noisy environments, In addition, the shutdown circuit feeds directly CORE: 2616PA 100337
to this iatch, which will disable the outputs within 200 ns of activation. The
current-limit amplifier is a wide-band, high-gain amplifier, which is useful
for either linear or pulse-by-pulse current limiting in the ground or power
F
regulator
- - 1 FIGURE 7-10 The UC15UA und PIC600hybrid switch provides a wide input range,
FIGURE 7-9 Blcck diagram of thc UC1324A P\1'51control IC. (Corifltbsy Cnitrorlc Cor- 25-W buck regulator, with a minimum number or external components. (Courtesy
porutiot~.1 I
www.IranSwitching.ir i ' ,+
Ur~itrclrleCotp~ration.) . .
200 HIGH-FREQUENCY SWlTCHMG POWER SUPPLIES
frequency switchingpower supplies, with the addition of a handful ofexternal

components.
Figure 7-12 shows the block diagram of the UC1846. The integrated
circuit provides the following nine features:
1. A t 1percent, 5 - 1 4Mmmed band gap reference used both as an external

voitage reference and in ternal regulated power source to drive low-!.eve1
circuitry.
2. A fixed frequency sawtooth oscillator with variable dead-time control and
external synchronization capability. Circuitv features on all NPN design
capable ofproducing low-distortion waveforms in excess dl MHz.Figure
7-13 shows the implementation of the oscillator circuit.
An external resistor RT is used to generate a constant current into a
capacitor CT to produce a linear sawtooth waveform, The oscillator fre-
quency is given by -
where 1kfl a Rr 5 500 162 and CT 2 1000 pF.
FIGURE 7-11 An off-the-line isolated forward converter using the UC1524A. (Courtesy
Unftrode Corpomtion)
through RI. Winding Nz takes over after steady-state conditions have been VI"
reached, providing the drive power. Isolated feedback control is provided SFc
by an innovative circuit comprised of transistor Q3and h-ansformer Te,by RT
sampling the 5-V output level at the switching frequency of 4 l kHz. CT
With every switchingcycle, the output voltage is transferred from winding
N1to winding Niwhere it is peak detected to generate a primary reference (-1 cur
signal to drive the PWM error amplifier. Diode D2is used to temperature sense
compensate for the loss in rectifier D,,and the net result is better than 1 (+) cur
sense
percent regulation.
Diode D3 forms a duty cycle clamp, Forward base drive current to the
switching transistor QIis provided by connecting the internal drive transis-
tors in parallel. Turn-off base current is provided by the combination of Cur llmR
transistor Q2and capacitor Cb. adlust-
auWown
7-34 UC4846 CurrerutlModr Conk01 integrated Circuit

In Chap. 3 the basic current-mode topology was discussed, md the advan-
tages over conventional PWM techniques were presented. C~D- 6'
The Unitrode Corporation UC1846 is an integrated current-mode control FIGURE 7-12 The Hock diagram of the UC1$46 current-mode control integrated circuit.
{Courtesy Unit& Corpomtion.)
circuit, which has all the necessary functions to design state-of-the-art, high-
202 H I W - F W N C Y SWITCHING POWER SlPWES
4. A current sense amplifier which may be used in a variety of ways to sense

peak switch current for comparison with the error voltage. Referring to
Fig. 7-12, maximum swing on the inverting input ofthe PWM comparator
is limited to approximately 3.5 V by the internal regulated supply. Ac-
cordingly, for a fixed gain of 3, maximum differential voltages must be
kept below 1.2 V at the current sense inputs. Figure 7-14 depicts sweral
ways of configuring sense schemes.
Direct resistive sensing is simplest; however, a lowerpeak voltage
may be required to minimize power loss in the sense resistor. An RC
filter is recommended to eliminate any extraneous voltage spikes which
may trip the current sense latch and result in erratic operation of the
PWM circuit, particularly at low duty cycles.
Transformer coupling can isolation and increase' efficiency at
the cost of added complexity. Regardless of the scheme, the largest sense
voltage consistent with lower power losses should be chosen for noise
immunity. Typically, this will range h m several hundred millivolts in
some resistive sense circuits to the maximum of 1.2 V in transhrmer
coupled circuits.
. .
sync
2
(Pin 10)
output dead time (rd)
FIGURE 7.13 The UC1846 oscillator drcuit. (Courtesy U n l f t d e Corpo- o m u t stage
rutton.)
Referring to Fig. 7-13, the oscillator generates an internal clock pulse

used, among other things, to blank both outputs and prevent simulta-
neous cross conduction during switching transitions. The output dead
time is controlled by the oscillator fall time. Fall time is controlled by
CTaccording to the formula Current xlormer
- WPf
For large values of RT
Synchronfzntion of more than one circuit may be accomplished via the

bidirectional SYNC pin.
3. An error amplifier with common mode range from ground to 2 V below FIGURE 7.14 Varims current sense schemes for the UClM6 current sense amplikr: (a)
bias voltage. The output of the error amplifier along with the output of Resistive sensing wltb ground r e f e m , (b)same sensing with R C filter to reduce swith
the current sense amplifier are fed Into the PWM comparator to provide transient spikes, (c) resistive sensing above p d ,and ( d )isolated current sensing. (Courtesy
the pulse steering modulated signal. Unitrode Curpmtton.)
.. - - - .-. www.IranSwitching.ir
204 HIGKFREQUENCY SwlTCHING POWER SUPPUES
I
II
SWITCHING REGULATOR CONTROL CIRCUITS 205
6. A shutdown hnction with built-in 350-mV threshold. May be used in

i
1 either a latching or nonlatching mode, or initiating a ""hiccup" mode of
j operation. The shutdown circuit is shown in Fig. 7-fk. Shutdown is
i
i
accomplished by applying a signal greater than 350 mV to pin 16. In this
u configuration, capwitor C, is used to provide a soft-start or soft-restart
I function, if so desired.
7 . ' ~ n d e rvoltage lockout with hysteresis to guarantee that the outputs will
stay "off' until the reference is in regulation.
8. Double pulse suppression logic to eliminate the possibility of consecu-
tively pulsing either output.
9. Totem pole output stages capable of sinking or sourcing 100 mA contin-
uous, 400-mApeak currents.
i All the above features indicate that the UC1846 current-mode control IC
is a versatile circuit which can be used in many designs.
j Figure 7-17 shows how the UC1846 could be utilized in a conventional
FIGURE 7.15 Peak current-llmft design for the UC1846. (Cwrtesy
Unitrode Corpomtton.)
5. Current limiting through clamping of the error signal at a user-pro-

II
1
push-pull circuit. This is a 20-kHz power supply, and although no e h r t is
given to matching the switching transistors, there is no current imbalance
to cause transformer saturation. For higher operating freqiencies, the
switching transistors may be replaced by MOSFETs. The power supply will
grarnmed level. In fact, one of the most attractive features of a current- 1 also exhibit excellent transient response to output load changes, as c o m p ~ e d
mode converter is its ability to limit peak switch currents on a pulse-hy-
pulse basis by simply limiting the error voltage to a maximum value. 1 to conventional PWM designs.
I
Figure 7-15 shows how this is accomplished.
Resistors R1 and R2 forin a divider network to set a predetermined
voltage at pin 1. This voltage, in conjunction with transistor Q1,acts to
clamp the output of the error amplifier at a maximum value. Since the
base-emitter drop of Q and the forward drop of diode Dl nearly cancel,
the negative input of the comparator will be clamped at the voltage at
pin I minus 0.5 V. Thus the digerenth1 input voltage, V,,,of the current
sense amplifier is
Using this relationship, a value for maximum switch current In terms of

external programming resistors can be derived as follows:
Note also that Ill supplies holding current for the shutdown circuit and 1
I
therefore should be selected prior to selecting RZ. FIGURE 7-16 Shutdown ctrcuft of the UC1846. (Cout-tesg Unitrode Corporacfon.)
i
7-3-5 The UC1860 Resonant-Mode Power Supply Controller
Increased interest in resonant-mode power supplies #hasbeen fueled by the
fact that this class of converters is well-suited for operation in the megahe*
ZOO rn region. Although a number of discrete power components have been de-
veloped to operate in the megahertz region, the lack of similar simple in-
tegrated controllers has been obvious. Unitrode corporation, however, is
in the process of releasing the UC1860 resonant-mode power supply con-
troller family, which will satisfy this need. Information given here is prelim-
UES 1402
inary, since at the writing of this book, only advanced product information
was available, but it should serve very well as an introduction to this versatile
IC. The block diagram of the UC1860 controller is depicted in Fig. 7-19.
The control philosophy employed is fixed on-time, variable frequency.
1 1 1 The fundamental oontrol blocks include a reference and a wide band, high-
gain error amplifier which controls a variable frequency oscillator from 1
FIGURE 7-11 A 20-!db push-pull current-mode controlled converter using the UC1846.
(Courteq Unftrode Corporation.)
Also by disabling the oscillator and error amplifiers (CTgrounded, EIA +

to Vd and - EIA grounded) of one or more slave modules, and connecting
SYNC and COUP pins ofthe slave(s), respectively, similar outputs may lx
p a d d e d together to provide a modular approach to power supply design.
In &is configuration, output current win be equally shared by all units.
Figure 7-18 shows a practical implementation of slaving two units in a pardkel
operation.
Out A
Out B
/
'7 Slave (additlonsl unlts)
FIGURE 7-18 Slaving a l h parallel operation of two or more UC1846 units, with equal FIGURE 7-19 The block diagram of the UC18M remaat mode power supply controller.
current sharing. (Cwtaeq Unftrode Corpwotion.) (Cou~y Unitrode Corpombion.) '
11 3108 HlWREQUENm SWmTCHBrlG POWER WPPUES
kHz to 3 MHz.The error amplifier controls oscillator frequency via a resistor

into the I, pin which is 2 diode drops above ground. The error ampIifier EIGHT
-- - -
is damped so that its output swing is limited from 2 diode drops above
ground to 2 V plus 2 diode drops, thus allowing minimum and maximum SWITCHING POWER
frequencies &be programmed, SUPPLYANCILLARY,
A temperature-stable one-shot timer, triggered by the oscillator, gener-
ates pulses as low as 300 ns defining on-time for the output drivers. Each - SUPERVISORY, AND
output is capable of driving transient currents up to 2 A, makng them ideal PERIPHERAL ClRCUlTS
for power MOSFET gates. The mode of the toggle flip-flop is programmable
for alternate or unison operation of the outputs. AND COMPONENTS
Additional blocks reside in the chip to facilitate more control capability.
An uncommitted open-collector comparator m be used to shorten the on-
time pulse during start-up or under low load conditions. A fast comparator
with a common-mode range from -0.3 to + 3 V is available to sense over
current fault conditions. The chip is versatile in fault disposition with several
soft-start and restart delay options. The restart delay pin can be used to
permanently shut the supply down after a hult, restart after a delay, or
restart immediately after the fault indication has been removed.
A programmable undemltage lockout (UVLO) rounds out the IC. It In general, switching power supplies are closed-loop systems, a necessity
a1ow.s of-line operation with a 16-V start thres hold and 6 V to V, 'hysteresis for good regulation, rippIe reduction, and system stability. Other than the
or operation directly from a dc supply from 5 to 20 V, The UVLO mode pin basic building blocks, which have been described in the previous chapters,
can also be used as a part for gating the entire supply on and off. During there are a number of peripheral and nncilkuy circuits that enhance the
undervolbge lockout, the output stages are actively driven low, and supply performance and reliability of the power supply.
current is kept to a minimum. Components such as optoisolators are extensively used by primary ref-
erenced designs like the flyback or feed-forward converter to offer the net,
essary input-to-output isolation and stfB maintain good signal transfer
information. Other circuits, such as soft-start, overcurrent, and overvoltage
1. Mamano, B.: "Applying the UC1840 to Provide Total Control for Low Cost, Primary- protection circuits, are used to guard the power supply against failures due
Referenced Switching Power Systems," Unitrode, Applkatlon Note U-91, 1982, to external stresses. This chapter describes tbe components used in these
2. -and R, Patcl: "The UC1524A Integrated PWM Control Circuit Rovides New Per- circuits, and presents typical circuit designs showing how these devices may
be utilized to perform their hnction.
romance Levek For Old Standard," Unitmde, AppZhtbn Note U-90, 1982.
3. Motorola, Inc.: "LinmlSwitchmode Voltage Regulator Handbook," 2d ed., 1982.
11 i'! 4. Unitrode C o p . : "Linear Integnted Circuits Databook," 1987.
The optical coupler, or optmupier, is used primarily to provide isolation
between the input and output of the power supply, while at the same time
output noise and low temperature c d c i e n t of 50 ppml°C are desirable
of a typical optocoupler. --
The optocoupler consists mainly of two elements: the light source, which
could be an incandescent lamp or a light-emitting diode (LED); and the
detector, which could be a photovoltaic cel, photodiode, phototransistor,
or light-sensitfve SCR. The most common construction of the optocoupler
. -(GaAs)
consists of a gallium arsenide+ -. . LED
-. - e -. .
and a silicon p*hoto.transistor
I 209
b NO HGKFREQJENCY SWnCWNG POWER SUPPLIES ANCUAW. WPERVISORV. AND PEmPHERAL ClRCUlTS AND COMPONENTS 211
point may be chosen and the current limit resistor can be easily calculated.
The basic parameter of interest in the output section of the coupler is the
collector current lc of the phototransistor.
The amount of current generated in the collector of the phototransistor.
is proportional to the diode forward current IF and the optmoupler's do '
current transfer ratio or coupling efficiency I . If the diode forward current
is known, the collector current of the phototransistor may be calculated by
6 d FIGURE 8.1 A typical o p b
pler drcdt.
The manufacturer's data sheets normally give a dc current transfer ratio
housed in a light-excluding package. In normal operation, a c u m t through curve at a specified collector-emitter voltage VcE.From this information the
the LED produces a light source, the intensity of which depends on the collector current (and emitter current) may be derived, which allows chws-
exciting current, which in turn modulates the phototransistor producing a ing the value of RL in order to get a specified output voltage V,, (see Fig.
collector current proportional to the forward diode current. Figure 8-2 shows 8-21.
an optocoupler connected in a basic linear operational mode.
In order to design the input circuit, the parameters-needed are the diode . - ..
forward current IF, the diode forward voltage VFI and the input voltage V,,. 8-2 A S-BIAS ncnmaur USED IN PRIMARY
SIDE-REFERENCED POWER SUPPtlES
Then the current limit resistor R may be calculated from the following
equation: Although base drive transformers may be used to provide the necessary
input-to-output isolation in a switching power supply, they are most com-
monly used in bridge type designs, mainly because of necessity rather thm
5 In general, diode forward voltage Is plotted vs. diode forward current

and is given in the manufacturer's specification sheets. Thus, an operating
I
choice. The majority of the flyback or forward converter designs use an
optoisolator to achieve the necessary isolation.
The use of an optoisolator greatly simplifies a design, because it eliminates
the need for the-driver transformer and the bias transformer. Sirice, in this "
case, the total c o n b l circuitq may be primary side-refkrenced, start-up
i and self-bias techniques may be used directly from the high-voltage line and
1I the high-frequency transformer to bias the control circuitry.
Figure 8-3 shows the implementation of a self-bias circuit whlch may be
used in this family of power supplies. The function of the circuit is as follows.
When the ac input voltage is applied, the PWM control and drive circuit
receive a bias voltage Vc set by the linear regulator Eormed by R,, Z,, and
Q1connected directly to the dc bigh-voltage bus. After the power supply
starts, an auxiliary winding in the main transbrmer provides a voltage VD,
the value of which is chosen by design higher than the voltage Vc, conse-
quently back-biasing diode D5and turning off the linear regulator. Under
this condition the power supply provides a self-bias voltage VD to keep it
running, while no power is dissipated in the start-up regulator.
Caution must be exercised when designing this circuit to select a high-
voltage transistor capable of sustaining the base-collector high-voltage stress
FIGURE 8-2 Aa optocoupler connected in a basic b a r when it is of. This circuit is a typical application of a self-exciting auxi1iar)r
mode. A diode fomrvd wrrent bgenerates a ligbt source, power supply, and many other c i m i a based on this principle may be de-
.. . ..-
!I
indudng collector current Ic to Bow in he phototramistor. veloped to suit individual needs., -.---= .
www.IranSwitching.ir -. - - .- . .
A
--
212 HIGH-FREQUENCY SWllCHlNG POWER SUPPLES
I 6
PWM control
and drive c ~ r m l t
FIGURE 8-3 Start-up m d self-bias d t y used In pttnery slde-refemnoed, ofLthe.

line switching power supplies.
6
FIGURE 8-4 Regulation control, as well as input-tosutput Isolation, i s provided in tbis
flyback drcuit using an qfOCOUpler.
8-3 WWCOUPUR CIRCUIT DESIa TO PROVIDE INPCII-TO-OM
ISOLATION IN A SWITCHIN0 POWER SWPPLY
in this flyback circuit is derived by the voltage divider comprised of resistors
When t b optocoupler is used in an of-the-line switching power supply for
the purpose of providing input-to-output isolation, the following design cri-
Re and Rg. This voltage is fed into one input of the amplifier A I and is
compared against a fixed rekrenm voltage Vef, which is fed into the other
teria must be kept in mind:
input of amplifier Al.
Any & h e n c e between the two voltages is amplified, and in turn the
I. The optocoupler must sustain an isolation breakdown voltage as dictated
voltage at the output of the amplifier A, produes a current through R3,
by l d andlor international safety standards.
consequently modulating the light intensity of the coupler's LED.The LED
2. The amplffier circuitry driving the optmupler must be well designed to light source induces a proportional emitter current at the phototransistor,
compensate for the coupler's thermal instability and drift. and therefore a voltage drop is developed across R , which mirrors the
3. An optocoupler with good coupling efficienq is preferred.
voltage at the junction of.R, and Rg.
The voltage across RL is fed into the noninverting input of the error
Normally1 the optocoupler in such a~applicatlonis used in a linear rmde; amplifier of the PWM circuit (see Fig. 7-3), while the inverting input of the
that is, a contml voltage at the input of the coupbr results in a proportional error amplifier is sitting at a fixed voltage, derived by dividing down a
output voltage to be used for further control, such as closed-loop regulation. reference voltage source Vmf.Therefore the conduction period of transistor
.-.. . - A typical circuit connected in this mode of operation is depicted in Fig.
Q kis adjusted aoxrdingly, in order to maintain regulation at the output of
the power supply. --.. -. - - .
of this Circuit is as hI1ows. A portion of the output voltage
8-4. ifie functionwww.IranSwitching.ir
214 HIGH-FREQUENCY SWUIING POWER SUPPUES ANCILLARY, SUPERVEOW~AND PEWHERAL Q R C U K AND COWONEMS 215
1
!
Although the circuit shown in Fig. 8-1 is a practical realization, the num- I
ber of components needed to drive tbe LED of the optocwpler may b i

dramatically reduced using the cirmit shown in Fig. 8-5. The function of
this circuit is exactly the same as the one described for Fig, 83. Notice here
the simplicity and the reduction of parts by using the TL431 shunt regulator.
'Fhis shunt regulator is offered by a number of companies, i.e., Texas In-
struments, Motorola, etc. It can be used ns a programmable, low-temper-
ature-coefficient, reference amplifier with current sink capabilities up to 100
mA. 1
8 .
Anode
The internal 2:S-v reference makes it ideal For operation from a 5-V bus, IA1
while ib output voltage may be externally programmed up to 36 V. Its low
output noise and low temperature codcient of 50 ppml°C are desirable
features for such an application. Figure 8-6shows tbe symbol and block
diagram of the TL431. In Fig. 8-5, the CrR, capacitor and resistor network
may be necessary in an actual application for frequency compensation.
Cethde
Reference
IRI
Output =Ion
i+ I
FIGURE 8-6 Symbol and block diagram d tbe TU3l programmable

-----
l nput section
1
precision reference. (Courtesy of Motomla Sem~conductorPmdum,
lnc.)
Another inexpensive, low parts count circuit with good perfOrrnance may
be designed around a single transistor as shown in Fig. 8-7. In this circuit
transistor Q1is biased at a set level using zener diode Z,, The collector
current flow excites the phodiode and sets the control voltage level across
-* I
!
i
RL-
mod-
Output voltage adjust is accomplished by varying potentiometer R,, thus
the light intensity of the photodiode. The.&-C1 I m - h u e n c y
filter is added to improve overall power supply stability.
All the above-described optocoupler driving circuits are typical examples,
and modifications may be necessary to tailor them to fit individual needs.
1 On the other hand these circuits may be used to develop other circuitry,
&- I even to be used in specialized applications. The fact remains that the op-
tocoupler is an important peripheral device in the design of switching power
FIGURE 8-5 Using the TLA31 shunt reguIator to drive tbe optocoupler LED and to supplies, providing input-tosutput isolation and still maintaining all the
provide the necessav amplilication fundon dramatfcally reduces cornponeat count regulation features of the converter.
www.IranSwitching.ir .- - -- - - . - - ..- " - ..
"-
.C
/
211 ' HIGH-FRWENCY W H W G POWER SERPUES ANCILLARY, SUPERV1WW. PERIPHERAL ClRCUlTS AND CCWONENTS a7
Ermr signet -
signal output
FIGURE 8-8 A typical soh-start circuit used in r PWM control circuit aids
FIGURE 8-7 A ~ i n ~ ~ f ~ 8 t I sampIifier-comparator
isfw may be used the gradual increase of the FWM signal to its operdng value.
to drive the uptocoupler in the Feedbck loop of a d t c h h g power .. .
supply.
Diode DZis used to bypass resistor R in order to dischargethe capacitor

C fast enough in case of system shut-down, thus initiating a new soft-start
cycle even during very short interrupt periods. In some second-generation
In most switching power supply designs it is desirable to introduce a certain PWM control integrated circuits, R has been substituted by an oh-chip
delay during start-up, in order to avoid output overshoots and trmsfbrmer current generator, thus the only external element required to implement
saturation problems at turn-on. Circuits which are-employed to perform this the soft-start feature is the addition of capacitor C.
task are called soft-startcircuits, and in general they consist ofan RC network It is obvious that the soft start imposes a certain delay on the rise tim'e
which allows the PWM control circuit output to increase h m zero to its
of the output voltage, thus a reasonable value of R and C must be selected
operating value very "softlys" to keep this delay within practical limits.
Figure 8-8 shows how a soft-start circuit may be implemented in a PWM
i
B
I
control circuit, At time t = 0, when the power supply is just turned on,
qipacitor C is discharged and the error amplifier output is held to ground
through diode DL,thus inhibiting the comparator output.
At time t = 0+,the capacitor starts to charge through resistor R with a
time constant determined by
!
I
!
1
A switching power supply is generally designed to operate safely at a pre-
determined output power level. Operation beyond the nominal output cur-
rent should be avoided, but iacase an overcurrent or short-circuit condition
-! occurs, the power suppIy must have some means of protection against per-
r = RC (8-3) ! manent destruction.
Current limit circuits are basically protective networks which will limit
toward the charging voltage V,+ AS capacitor C attains full charge, diode the output current drain i n a safe level in case of an output short-circuit
D lis back-biased, and therefore the output ofthe ermr amplifier is isolated miidition. There are many ways of implementing a current limit circuit by
from the soft-start network. The slow charge of capacitor C results in the placing it either at the primary (input) side of the power supply or at the
gradual increase of.the PWM waveform at the output of thecomparator, j output section. Of course, the optimum current limit topology greatly de-
and consequently a "soft start" of the switching element is initiated.
1 pends on the specific power supply design whkh it protects. single-oudut
218 H I G K F R E Q U E W SW~CHINGPOWER SUPPLIES W U A R Y . SUPERV1SORV. AND PERlPHERAt ClMXllTS AND COMPOREMS 2l9
designs may be equally protected by placing the current Iimit at either the In Figure 89a, the peak primary current is monitored in terms of a
input or the output section. Thus, for primary-derenced direct drive power proportional voltage drop developed across the current limit resistor Rsc.
supplies, it may be convenient to place the current limit at the input side, The value of resistor RSc is givm by
while for power designs using base drives, it may be beneficial to place the
current limit circuit at the output bus.
Although direct coupling of the current limit circuit with the monitored
bus may be convenient, simple, and may require a low parts count, trans- When the voltage dmp across resistor Rsc exceeds the base-emitter threshold
former-coupled current limit circuits we alsa extensively used, especially voltage, transistor Q2turns on. The coIIectorofQ2is returned to the oscillator
when grounds are not common and voltage level translation is required. output or a shutdown port.
Curreljt Iimit circuits may be implemented by discrete components, or in If an overload or short exists at the output of the power supply, primary '
IC P W M control circuits the integral current limit function may be used. current increases dramatically, turning transistor Q2on, harder and harder.
It should be noted here that the current limit must have a fast response The collector of Qz,in turn, pulls the oscillator output toward ground, or it
in order to protect the power supply before destruction occurs. activates the shutdown circuit, thus limiting the available primary currmt
to safe levels.
A faster and more accurate current limit circuit is shown in Fig. 8-9b.
8-54 Curtent Umit C l w b for Primary- Such a current Iimit is popular among PWM control circuit ICs. Although
Referencd Dl& D r l w Designs
the principle of operation of this circuit is similar to the one described
Primary-referenced direct drive designs, such as flyback or fon*ard con- previously, there are certain distinct advantages in using this circuit vs. the
verters, are easily current Iihited. Figure 8-9 s h m two ways of achieving transistor one. First, the comparator's current limit-activatbg threshold
current limiting in such designs,.
-. . . A
voltage is preset to an accurate and predictable level, as opposed to the wide
Vg, threshold of a bipolar transistor. Second, this voltage threshold is made
small enough, typically 100 to 200 mV, dowing the use of a smaller qrrent
limit sense resistor value, therefore increasing the overall dciency of the
converter. I
#
8-5-2 Current Llmlt Circuits for Dclslgm Wllzlng-laseDrivers

Normally in designs which utilize base drive isolation beween the control
circuitry and the switching transistor, such as half-bridge, full-bridge, andl
or flyback and forward conveprs, the output section shares a common ground
with the control cirhit:In such cases the current limit circuit may be directly
connected to the output bus; Such an implementation is shown in Fig.
8-10.
Under normal operation the load current IL is small enough to develop
a sufficient voltage drop across R x to turn on transistor Q1. Since Q1 is off
and lIcl = 0,capacitor C1 is totally discharged and consequently transistor
Q2is of. If IL is increased to a value such that
then collector current lcl starts to flow, charging capacitor C1with a time
FIGURE 8-9 In &back or f4 converters, a simple resistor in series with the switch constant determined by
will p w d e the mmsary voltage drop c turn om the transistor (a), or d a t e the IC
mmparatw (b),and shorten the drive pulse in case of ofrcu~rentmditfons.
-. .. . .. . ... .
-.
. :
ANCILLARY, SUPERVISORY. 'AND PERPHERAL CIRCWTS. AND C ~ E N I S
Z2l
parators, the circuit in Fig. 8-9b may very well be used by moving the
current limit resistor Rsc at the output positive bus.
Although both methods work well to detect an overcurrent condition, the
presence of the power resistor Rsc may become objectionable, especialIy in
high-current outputs b q u s e of power dissipation and its adverse effect on
system efficiency. If that is the case, the,&cuit in Fig. 8-11 may be used.
T h i s circuit uses a current transbrmer to detect overcurrent conditions, and
sMee no loss elements are involved, the overall efficiency of the power supply
is increased. The circuit operates as folIows, A current transhrmer T1mon-
To
regulator itors load current 1, producing a proportional voltage determined by the
shutdown scaling resistor RI. Diode D j rectifies the pulsed voltage, and resistor R,
and capacitor C1are chosen to srncath the voltage.
When a current overIoad oocurs, the voltage across C1 increases to the
p i n t where Zener diode Z1cpnducts, turning on transistor Q1.The signal
,
at the collector of Q may be 'ked to shut dow11-1the regulator's drive signal.
Current transformer TImay be wound on a ferrite or MPP toroid, but care
. FICUBE $40 Tbir current limit circuit may be wed in almost all must be exercised to keep the core out of saturation. Normally the primary
power-umwrted ckigns, where the control d d t shares the s a m e consists d o n e hrn, while the number of secondary turns needed to establish
ground with the output return bus. the necessary secondary voItage is given by
Tfie charging voltage at capacitor CIwill attain a value

VCI= I 8 8 3 + V ~ ~ , Q ~ @-7) Since Is = VslR1, the number of secondary turns required to prodqce the
desired voltage across the capacitor C,at maximum specified load current
In order to minimize loading on the capacitor voltage, a Darlington transistor I, is given by
with a very high hpEis chosen for Q2.This will limit the base current Ig2 to
microamperes. Choose R4 4 R j in order to allow fast discharge of C1,after
a current overload has been detected.
Rt is selected as foI1aws:
and
ICI = ~ Q ~ 1mPx
IB
T ~ 4 a t w
Therefore, QI shutdown
In a properly designed circuit, Vo reaches its value fast enough to bias

transistor .Q2 on, which in turn shuts down the regulator's drive signal.
Recovery of the circuit is automatic upon removal of the overload. If FIGURE 8-11 6'dIssipalire m a t Ifmlt circuit, using a current transfomer
integrated PWM control circuits are used with built-in current limit com- to detect ovmmment coditiww.
P I
C
222 HIWFMQUENCY WTCClNG POWER SWRlE AhlCILIARY, SUPERVISORY, AND PERIPHERAL CIl?CUltS AND COMPON~MS 223
1
Eq. 8-10 gives accurate information for manukturing the current trans- If an overload occurs, the voltage Vlwill rise above Vd, pulling the output
former. Final adjustments in the number of turns may be made during actual of the comparator low. The high-to-low transition at the input of the 555
circuit tests for optimum performance. p d u c e s a one-shot output, turning on transistor Q1. The collector of tbe
transistor is tied to the shutdown input or across the soft-start capacitor of
the PWM circuit, and by pulling it toward ground it terminates the output
i switching pulse, shutting down the regulator.
8-5-3 A Universal Current LlmH ClrcuH d+ I If the overload persists, the 'puwer supply e!~ters a hiccup mode; that is,
I
A universal current limit circuit may be designed that is equally abk to it turns on and off, with a period d e t e ~ i n e dby the 555 one-shot RC time
perfohn well when placed in either the input or the output section of the constant, until the overload is removed. Recovery then is automatic. Design
power supply. Also, this circuit is well suited for multiple output voltages, of the toroidal current transformer is the same as describd in Sec. 85-2.
where the multiplicity of outputs makes individual current limiting a dficult
task. I
&6 OWRVOLTAOf PROTECTION CIRCUITS
Figure 8-12 shows the basic circuit design. This circuit shows operation
at the input side ofthe power supply. Current transformer T1monitors the Overvoltage protection circuits are networks designed to clamp the output
primary current of power transformer Tp The secondary voltage of trans- voltage to a safe value, shourd this voltage attempt to rise beyond a pre-
former TLis rectified by diodes Dlthrough D4and smoothed by capacitor determined value. Although the threat of an overvoltage condition was quite
C1. Besistor R , is used to set the comparator trip threshold. Under normal likely in linear power supplies, .this is not necessarily the case with switching
opefation the VRrterminal to the comparator.is in a higher potential, and power supplies. In fact, most of the failures in a switching power supply
the comparator output is high. Therefore the 555 one-shot multivibrator has
a low output, holding transistor Q1oE. circuits? -
result at a "no output" condition. Why then use overvoltage protection
The season is twofold. First, in power supplies with adjustble outputs,

wewoltage protection circuits guard against accidental output overadjust-
ment. Second, they reassure the end user about safety against overvoltage,
even during the rare occasions of such a happening.
A better place to use an overvoltage protection (OVP) circuit is right at '
the electronic circuit which the power supply powers. An OVP circuit the&
will definitely protect the circuits from accidental overvoltage application
caused by assembly errors, especially when more than one voltage is to be
wired on the same circuit. The simplest and most effective way of imple-
menting m OVP circuit is to use a "crowbar" SCR across the dc power bus.
When an overvoltage condition is detected, the SCR is turned on by some
means, shorting the output terminals. Since during turn-on the SCB is
subjected to large amounts of current, careful selection of the device must
be made to suit the intended application.
8-44 me Zener Sense OVP Clrcutt

One of the most widely used OVP circuits is the one shown in Fig. 8-13.
Although this circuit provides very poor gate drive for the SCR and also
decreases the dildt handling capability of the SCR,it works fairly well for
a low-cost design. Under normal operation, the gate of the SCB is grounded,
keeping the device off. When an ovemltage is detected, zener Zlconduct.,
k! FIGURE 8-12 A oneshut multMbrator i s used to initiate current Umit "hiccup"when
an werload is detected,
pulling the SCR gate to the zener voltage, thus turning it on and conse-
, - + -- quently shorting the output terminals.
.- .
i
FIGURE 8-11 An OVP circuit
=&g of a Zener diode and an
- SCR.
Once fired, the SCR remains on until its anode voltage is removed. In
power supplies this may be amomplished by removing the input p o ~ for r
a few seconds.
FIGURE 8.15 Typical O W application d tbe MCW23. (Cou*sy of Motoroh Sendcan-
ductor Pmdum, Inc.) ,
-
In recent years a few dedicated OVP integrated cirmfts have been idrodwed The earliest of these ICs was the MC3423, which has becorfle an industry
by a number of msnufadurerr. Most of these circuits are low-cost and offer standard. A basic block diagram of this JC is given in Fig. 8-14, The diagram
the designer a number of design fcahms, such as progwnmable trip voltage shows that the circuit consists of a stable 2.6-V reference, two comparators,
threshold, fast response, low-temperature coefficient trip point, etc. and a high current output, ']The output is activated by a voltage greaer than
2-6 V on pin 2, or by a high logic level on the remote activation, pin 5,
Figure 8-15 shows a typical application of the MC3423 in an OVP appli-
cation. In the circuit, resistors RI and REset the threshold trip voltage. The
relationship between V,, and R1,REis given by
Keeping the value of Ra below 10 &I for minimum drift is recommended,

The value of R1and R2may also be calculated by using the graph of Fig.
8-16. In the graph, RE = 2.7 kSt, while R may be directly calculated from
the intersection of the trip voltage with the desired curve.
The MC3423 OVP circuit also has a programmable delay feature, which
prevents false triggering when used in a noisy environment. in Fig. 815,
capacitor CDis connected From pins 3 and 4 to the negative rail to implement
this function. The circuit operates as follows. When V m rises above the trip
point set by R1 and Rz, the internal current source begins charging the
capacitor CDconnected to pins 3 and 4. If the overvoltage condition remains
present long enough for the capadtor voltage Vm to reach Vd, the output
is activated. If the wemoltage condition disappears before this occurs, the
FIGURE 8-14 The M C W O W drait block diagram. (Cwrteq of Motordo capacitor is discharged at a rate 10 times faster than the charging time,
Saniwnrductor Products, Inc.)
. -. ..
ANCILLARY. S U P E R W R Y . AND PEIXPHRAL
CIRCUITS AND COMWNENTS 227
Vvip volts
FIGURE 8-16 The threshold resistor values mmy be dlrecdy calm
culatd from &s graph, wbich plots R , vs. trip voltage for the h4-
OVP circuit. (Courtmy 01
M o t o r o h Semiconducto~Products, I=)
FIGURE 817 Delay capacitance CDvs. mMmum overvoltage duration to for the MC3432
resetting the ffming feature. The value of the delay capacitor CD may be OVP c i r d . (Courtesy of M o t o r o h Smlcoaductor Products, Inc)
found from Fig. 8-17.
A more elaborate OVP circuit, the MC3425, is in many respects similar
to the MC3423, but the former one may also be programmed for under- I d , changing the external deIay capacitor Ca to 2.5 V. The delay time t d can
1
voltage detection and also line loss monitoring. The block diagram of the be calculated as follows:
I
MC3425 is shown in Fig. 8-18. Notice that this is a dual-channel circuit,
with the overvoItage (OV)and undervoltage (Uw input comparators both
referenced to an internal 2.5-V regulator. The UV input comparator has a
feedback-activated, 12.5-pA current sink In, which is used for programming
the hysteresis voltage VH.The source resistance feeding this input RH de- From Eq. 8-13 the delay capacitor can be easily calculated for a wide range
termines the amount of hysteresis voltage by B of delay times, or it may be directly determined using the graph of Fig,
8-17.
The delay pins are pulled low when the respective input comparator's
Separate delay pins 2 and 5 are provided for each channel to independently noninverting input is less than the inverting input. The sink current Id
delay the drive and indicator output pins 1and 6, respectively, thus providing capability of the delay pins is greater than 1.8 mA and is much greater than
greater input noise immunity. T ~ ~ A V O pins are essentialIy the outputs
delay -
.- the typical 200 FA source current, thus enabling a relatively fast delay
of the respective input comparators and provide a constant current source capacitor discharge time.
Id of typically 200 pA when the noninverting input is greater than the in- I The overvoltage drive output is a current-limitedemitter follower, capable
verting input level. A capacitor connected from these delay pins to ground of sourcing 300 mA at a turn-on slew rate of 2 Alps, ideal for driving
will provide a predictable delay time td for the drive and indicator outputs. "crowbar" SCRs. The undervoltage indicator output is an open wllector,
The delay pins are internally connected to the noninverting inp& of the NPN transistor, capable of sinking 30 m A to provide suficient drive for
OV and UV output comparators, which. are referenced to the internal 2.5- LEDs, small relays, or shut-down circuitry. These current capabilities apply
V regulator. Therefore delay time td is based on the constant current source P
to both channels' operating simultaneously, providing device power dissi-
ANCILLARY. WERVLSORY, AND PERIPHERAL CIRCUITS AND COMPONENTS 229
accomplish two design goat. First, half the MC3425 may be utilized as an
OVP circuit, while the other half may be used to detect ac Iine loss or
brownout conditions. Figure 8-20 shows a typical application of such a circuit.
As a line loss detector, input pin 4 of the MC3425 is connect4 as an
undervoltagt sensing circuit to sense the center tap of a full-wave rectified
signal pmyo -tional to the ac Iine voltage. At each peak of the line the output
of the compamtor dischar5es the delay capacitor Cd.If a halfcycle is missing
from the Iine voltage or if a brownout occurs reducing the peak line volHge,
QV d t ~ the delay capacitor is not discharged and it continues to be charged as shown
in Fig, 8-21. If a sufficientnumber of half cycles are missing or if the brownout
W Ind
continues for a sufficient time, the circuit will detect an ac line fault and will
wtput a Tine fault indication on the indicator output, pulling pin 6 low.
Input~ectlOn W OV I
dfy dlY
FIGURE 8-18 The dual-channel M C M 5 power supply rupelrtso~,wedtage, under-
voltage protection circuit, (Courtesy Motordo Smlconductw Products, Inc.)
pation limits are not exceeded, The MC3425 has an internal 2.5-Vhand gap
reference regulator with an accuracy of 2 4 percent for the basic devices.
Figure 8-19 shows a typical application of the MC3425 in an overvoltage
protection circuit, with undervoltage fault indication. Note that delay m-
pacitors have been added to pins 2 and 5.
i
1
8-7 AC LIME LOSS DETECTORS
In many computer applications, an ac line loss or a brownout condition must
1
!
be detected in timely fashion in order to transfer valuable data to nonvolatile
memory banks or to trigger an uninterruptfble power supply (UPS)to take .
i ?
:I
over before the power failure occurs. This line loss detection must be done
within a cycle or two of the ac frequency, since most of the switching power
supplies have a hold-up time of 16 ms minimum, enough to power the circuits
between line loss detection and take-over time.
FIGURE 8-19 Overvoltage protection and uaderroltagefault fndication with
i. Using the MC3425 integrated circuit, the power supply designer may
1 programmable delay, using the M C M 5 . (Courtesy Motorola Semiconductor
d Products, Inc.)
230 HIGH-FREQUENCY SWITCHING POWER S1IPAUB ANCILLAM: WRVISORY, AND PERlRlERAt ClRCUlTS ANO 233
1 2. Chryssis, G. C.: Avoid Supply Voltage a d Overload Problems by Defining Them Before
1 They R a p p , Electronic Design, August 2, 1979,
!
!I 2. Single Transistor Adds
-----: Latching Function to Overvoltage Protection IC, E h n k
l?ea4gn, August 16, 1979.
t.
I 3. Motorola, Inc. "LincarISwitchrnodeVoltage Regulator Handbook," 2d ed., lsSZ.
I " P m Supply Design Seminar Manual."

4. Unitrode Corp.:
"Linear Integrated Clrcuits Databwk" 1B7.
5. ----:
FIGURE 8-20 The hIC342.5 inkgratd circuit r n be~ as& to detect GC line Faults
and overvoltage conditions, Indeptndendy, (Cov-sy Motorola Semkanductor
Products, IRC.)
The delay capacitor is used to provide some noise immunity and to prevent
the 101s of a single h& cycle from triggering the h e fauk signal. 7 % ~
minimum time the fault condition must occur can be adjusted by changing
the value of tho delay capacitor. The graph of Fig. 6-17 may be used to
spec'& the delay capacitor.
uPin 4v - - , q C j q - -fiZav
UVdlr
Pin 5
--- - ----- 2.5 V
A
UV Ind
Pln 6
.-=
FIGURE 8-21 Waveforms nlustrating brownout conditions and line
loss detection for the circuit in Fig 8-m. (Cou~esyAfotorola Semi-
-.. www.IranSwitching.ir
condrrclor Products, lac.)
1 NINE
STABlLllY IN SWITCHING
POWER SUPPLIES:
ANALYSIS AND DESIGN
1
C..
9-0 IWRODU~TION
The subject of stability, which pertains to the closed-bop frequency response
of switching regulators, has undoubtedly received much amntion, and a
score of papers have been published on and around the subject. To most
practicing engineers as well as students, however, feedback control bop
stability seems to be shrouded by a cloud of mystery. Although most de-
signers do understand what causes a switching power supply to osciIlate,
many of them stabilize the loop using a trial-and-error approach or fancy
i! mathematical models which require extensive use of the computer.
This chapter presents feedback imp stability, blending theory and practice
in a coherent way, in order to give the reader the necessary tools to stabilize
!
switching regulators with little effort, much enjoyment, and great reward.
*I
i
9 4 WE W C E TRANSFROM
In most linear systems, the input-to-output relationship characterizes the
system, and the differential or integrodiferentd equations which mathe-
matically describe the system give art idea of the response to some input
excitation. These equations are generally given in the time domain and are
diLult to manipulate By applying the Laplace mnibrm, the equations
are "transformed" to the frequency domain, thus taking an dgebraic form,
which is easier to handle. After the desired result has been derived, trans-
formation back to the time domain may be achieved by applying the inverse
Laplace transform. '
By definition, iff{*) is any function of time such that'f(t) = 0 for t < 0

I
and the integral Jo'jr(t)e dt has a finite value,. thenf(s.) is the Laplace trans-
-* -
-. - ..--
.- . . . ... * - .- --.-, -
www.IranSwitching.ir * 133
. C
/ #
234 HIGH-FREQUENCY SWlTCHMG POWER SUPPLIES S T W W I N SWrtCHlMG POWm SUP-: ANALYSlS AND D E W -5
form off(t). The Laplace operator s is defined as the complex variable

s = u f j# (9-1)
and the Laplace transform is defined as
E U P L E $1
Find the Laplace transform of the unit step function, defined as f(t) = 1
for t > 0 and f(t) = 0 for t < 0. Wn, dB Phase, d e g m
.O 0
SOLrnON
Using Eq. 9-2 we have
90
therefore
bl
FIGURE sl (a) An RC integrator arcuft and (b)its gam and
ph= plot.
It can be seen from Example 9-1 that any time function may be transformed Using Erchhoffs law the network equations may be written as I
in terms of the complex variable s. In case the result is required in the time
domain, the inverse Laplace transform, given by
I and
may be used to obtain ftt).

There are tables which give bothf(8) and f(t), so that transformation from
time to frequency domain and vice versa may be done quickly and efficiently. I Substituting q = Jf dt, the above equations reduce to
and
9-2 TRANSFER FUNCTIONS
So far so good, but how can the Laplace transform be used in our study of -' '
system stability to derive useful information? The first step is to derive a I - . . Taking the Laplace transform,
relationship between input drivlng signal and output response of a system.
As an example, let us examine the simple RC network given in Fig. 9-1.
and
I STABllTN IN SWffCHNG POWER SUPPUES: ANALYSlS AND D E W 237
To summarize, then, a pole will cause a transition from a + 1to a 0 slope,

or 0 to - 1, or - 1 to - 2, or - 2 to -3, etc. This corresponds to a gain
change per octave of + 6 dB, 0 dB, - 6 dB, - 12 dB,' and - 18dB, associated
with a phase shift of + 90°, 0°, - W ,- 180°, - 27P, respectively. The gain
Dividing Eq. 9-4 into Eq, 9-5 we get change per decade is'\+20 dB, 0 dB, -20 dB, -40 dB, and -60 dB,
associated with a pha$e shift of + 4 8 . O0, - 45', - 9[P, - 135'.
Zeros, on the other hand, are points in frequency where the slope of the
Bode plot breaks upward, causing a transition of the gain plot froma - 1 to
The mtio of Vm(s)SV,,(s)is defined as the transfer function G(s). The im- 0 slope, or - 2 to - 1, or -3 to - 2, etc. Accordingly the phase shift would
portant thlng here is to recognize that this function has a gain and a phase now be leading by 9V. Figure 9-2 shows crkuits which cause poles and
associated with it. Any system, therefore, may 'be described by its tnnskr cfrcuits which cause both poles and zeros,
function i
In order, then, to plot any network in a Bode format, first determine its
. 1 transfer function using Laplace transforms; then arrange the equation in the
i form
In such an equation, the roots of N(8) = 0 are d e d the zeros of the system,
while the roots of D(s)= 0 are called the poles of the system, The most
convenient way of plotting the gain and phase of a transfer function is on a
decibel basis, and such plots are called Bode plots, after the man who Points T,, q, . . . ,r, correspond to the zero break frequencies, while points
,
developed them. T ~ ,~b~ ... q,,correspond to pole break frequencies. Then plot the gain
vs. frequency on logarithmic paper, choosing a gain change slope corre-
sponding to decibels per octave or decibels per decade. a
9-3 BODE PLOTS If phase shift is to be plotted, remember that a pole causes a phase lag
We mentioned before that a transfer Function equation has poles and zeros of 90°, while a zero causes a phase lead of 90' per decade. Since all the
and that these poles and zeros determine the slope of the gain plot, Let us information on gain-phase plots is plotted on logarithmic paper in t e n s bf
now examine Eq. 9-6 and Fig. 9-1. The equation shows a pole in the de- decibels, simple addition of the individual asymptotes is required to derive
nominator, By setting sRC + 1 = 0, we get the final rate of closure.
I
j
and 1 9 4 E D B A C K THEORY AND THE CRITERIA FOR STABILIW
Any switching regulator may be treated as a closed-loop feedback control
system. A block diagram of a closed-loop system is depicted in Fig. 9-3
where the wtput signal is fed back and compared to the input. A referenee
Equation 9-8shows a vew important result-that a pole will cause the
transition of the gain plot from 0 to - 1 at a frequency f, = 112nRCRThis
frequency is called the corner or breuk frequency, so termed because the
/
!
signal A(*) is compared to a feedback signal B(s)at the summing point, and
the error signal E(s) inputted to block G(s), and an output C(s) is obtained.
In order to derive the closed-loop transfer function f(83, we proceed as
asymptote breaks here, i follows: e
If we determine the rate of change of this asyiptote, we see t h t the
slope is - 6 dB per octave, or - 20 dB per decade. An octave is a 2: 1range '~ ( s =) G(s)E(s)
of frequencies, while a decade is a 10: 1 range of frequencies. Likewise the
phase of the network changes at points A110 and 10fc, producing a 900 phase B (s) = H (s)C(8)
1%. www.IranSwitching.ir
-. . .. ..
fa)
Phase.
FIGURE 9-3 Blwk diagram of a closed-loop feedback mntd sys-
deSlr-s tem.
l4 -2 slope
0
Eliminating E(s) from the above equations, we get
. ..... .
-1
t=&e,-& ,
..z=
and the closed-loop transfer function is
Phase,
, (y- Gain,
. '.. - "dB - The term G(s)is the open-loop gain, while the term G(s)H(s)is called the
.++,..,
-.I * - - open-loop transfer function.
... ....
I
*-: G..
,
,.
. , ...
,* .
_ 0
Pok In order to derive a conclusion about the stability of the system, solution
-. ., of the characteristic equation I
.....,, -..,-
-.....-.
"..*.
0
Gain 1 + G(s)H(s)= 0 (9-10)

Vln
I t will give the poles of the closed-bop transfer function, since they characterize
I I
the response of the system. Therefore, the feedback system must be ex-
amined for eacb value of closed-loop gain to determine the rate of closure
between open-loop and closed-Imp gain. The objective of stabilf ty analysis
is to reduce the closed-loop gain rolls6 rate to a - 1slope, i. e., - 6 dB per
octave or -Ul dB per decade, at the region of unity gain mosswer (0 dB).
At that point the phase shift will be less than 360°, a condition for a stable
system. The amount the gain is below unity when the total phase shift is
3600, is d l e d the gain margin, while phase margin is the &rence between
the actual phase shift and 3mm,when the loop gain is unity, as shown in
Fig. 9-4.
A typical switching power supply closed-loop system is depicted in Figure
9-5. The loop consists of two typical blocks: The modulator, where the power
processing t&es place, in series with an amplifier generally called the feed-
back or error amplifier. Although the modulator shown here is a buck reg-
(dl (a
:-- (dand (b)c m w pies, while networks (c) and (4
FIGURE 9-2 Nehwdm www.IranSwitching.ir =use both
pdes and zeros.
/ *
a 0 HIWREQUENCY SWITCHING POWER WWJES
S l ~ ~ l i l TINf SWlTCtPNG POWER SUPPLIES: ANALYSIS AND DESIGN a1
Unity galn
cmamwr polnt
L ----------
Amplifier
1
FIGURE 9-5 A typical feedback controlloop, showing the
modulator and error amplifier.
. ...-~ -
in Figs.9-6 and 9-12.The next step is to determine the unity gain crossover
frequency and the desired phase margin, The uniw gain crossover frequency
Phase I should be chosen based on desired perhrrnance, but a rule d thumb is to
choose a frequency at a b u t one-% the regulator's clock frequency.
The desired phase margin, that is, the difference between the actual phase
shift and 360" when the loop gain is unity (0 dB), must be at least 000 or
better. A good compromise is 600,which also gives a good transient response.
The find step then is to compensate the feedback or error amplifier so that
its gain is equal to the recipmcal of the modulator gain at the desired hO-
quency.
To achieve overall system stability and adequate phase margin, the am-
plifier gain combined with the modulator gain should produce an overall
gain plot which crosses the unity gain (0 dB) line at the desired crossover
frequency, at a -I slope, as shown in Fig. 9-14. The - 1 sbpe introduces
FIGURE 9 4 Pbaoe and gain ma&^ pbts o f a feedback system. Phase 900 of phase lag, which combined with the 180' of phase shift already existing
sblft is plotted here in termsof 1Wbecauseat dc tbe feedback is negative, at dc (inverting amplifier), gives a total phase lag of 270'. Since there is 90'
i.e., there is an additional 18V of pbase shift for a totd of 360" phase left to get to 360°, the phase margin will be 90'. Remember, as mentioned
sbift as defined inthe text.
in Sec, 9 3 , that a - 1 slope introduces 900 of phase lag, a - 2 sbpe, 180"
of phase lag, and a - 3 slope, 270' of phase lag. It is therefore obvious that
a loop system with a - 2 slope at unity gain crossover has no phase margin
ulator type, the discussion to follow holds true for any modulator no matter
because the total phase shift is W, while a loop system with a -3 slope
how complex. -. .
will oscillate because the total phase shift around the loop exceeds 3600.
To optimize the design of the feedback loop in order to achieve stable
The following paragraphs of this chapter will examine analysis and design
overall operation of the system, the first step is to determine and plot the
techniques In tailorfng the gain of the error amplifier, allowing prediction
control-to-output transfer function of the modulator and draw its Bode plot.
Typical Bode plots of a modulator showing both gain and phase are shown and plotting of loop performance of any switching regulator, without tedious
trial-and-error efForts.
241 HIWREQUENCY WTCMVG POWER SUPPUES STABILITY IKSWlTCHlNG POWER SUPPllES: ANALYSIS AND DESIGN 243
*
I
9-5 OFF7WICIINE SWITCHING POWER SUPPLY STABILIIY AMAfYSlS
i A11 off-the-line PWM switching power supplies consist, more or less, of a

% moduIator, an error amplifier, an isolation transformer, and an output LC
*
I
filter. The control-to-output transfer function of a switching power supply
using a PWM controller IC includes the gain of the sawtooth modulator,
the power switching circuit, and the output filter characteristics.
+2I
p
I
In singIe-port direct duty cycle control PWM power supply topologies,
a voltage Vc applied to the control port of the P W M comparator (see Figs.
o
* Phar
degrees
7-3 and 7-5) is compared to a sawtooth voltage of constant amplitude Vs to

change the comparator's output duty cycle from 0 to 1. The resulting duty
cycle 6 of the drive waveform then varies as -20 -90
-40 I 1-180
The gain of the buck family converters, i.e., feed-forward, push-pull, and 1 10 loo
bridge converters, is given by f, kHz
FIGURE 9 6 Control-to-autputtransfer function characteristics of LC
filter and modulator, of a W M switching power supply. The - I slope
at about 20 kHz is caused by a zero due to the ESR of the output filter
.
cqachr. The phase is also shown.
where hrslNp is the transformer secondary-to-primary turns ratio, and
Vh is the transformer primary voltage.
P The gain of the buck-boost converter, i.e., the flyback, is given by In decibels the dc gain is given by I
The output filter on the other hand is normally an LC type, with a slope of
In order to obtain the control-to-output voltage dc gain of a PWM power
supply, Eqs. 9-12 and 9-13 are differentiated with respect to Vc, i,e.,
- 2 ( - 40 dB per decade), as shown in Fig. 9-Zb. f i e overall closed-loop
gain of the power supply is therehre
dV,,JdVc. For the buck famiIy of converters,
avmt -- - [(Gain)H(r)]
avo.,
-=
avc
{dc gain) = - -
V," Ns
v s NF
avc
This means that the Bode curve of Eq. 9-18 has a dc gain which is flat out
.&
- In decibels the dc gain is given by to the resonant frequ'ency of the LC filter and then Mls at a -2 ( -40 dB
per decade) slope, as shown in Fig. 9-6.
{dc gain)dB= U3 loglo
(2
-)-
:
:I.
L 4 For the buck-boost family of converters, -
9-5-2 Error Amplifier Compensation
l n the majority of the PWM kontro1 ICs, the ermr amplifier Is a high-gain
avwt- - (dc gain) = h
'l 3 = + v*.t)' 5
(b
' (9-16) operational amplifier, which generates the ermr signal to the control input
av~ - . v ~ ) eN~
(v5 vfnv~ NF of the modulator. CIosing the loop in a PWM switching power supply in-
I
*.
.'
.
, -- -
:-: A;.-
. . -. .- .
+
. " .
& .. - . . ...
. . .
244 H I G H - F R E W W SWITCHING POWER SUPPLIES STABIUW IN SWnCHNG POWER SU!TUES: ANALYSIS AND 5fSlGN 24S
volves the active role of the error amplifier, and the objective is to design
the feedback network around the amplifier such that the overall loop gain
crosses the 0-dB (unity gain) line at a - 1 (- 20 dB per decade) slope.
In order to plot the amplifier characteristics in a Bode format, its gain
must be written in a Laplace form. Let us examine the properties of a simple -,
operational amplifier first, and see how we will be able to write its transfer "
function in a Laplace format. Figure 9-7shows an operational amplifier and

its feedback impedances. As mentioned previously the transfer function af
this circuit is the ratio of output voltage to input voltage, and for an oper- FIGURF 47 A simple -rattonal amplifier drcuft
ational amplifier with feedback imped-.
The high-frequency amplifier gain is set by Rs and R3. Therefore the gain
at f, is
Since 5 and 2,represent complex impedances, when the Laplace trans-
form is applied to them, Eq. 9-19 may take the form
The operator T represents a time constant RC in Eq. 9-20. The terms in the
numerator represent zeros, while those in the denominator represent poles.
The term ( ~ ~ represents
3) a pole at the origin because it lacks the + 1term.
In order to appreciate the ease by which one can write the transfer function
of any operational amplifier with complex impedances, and also sketch its
transfer function in a Bode plot, the circuit in Fig. 9-8 will be used as an
example. By inspection we write,
&= R3 + lJCss - R1 + lJCss phase,

degrees
Vh Rz + {(RIICl~)J[(RtCls+ l)/Cls]) Rg + [RIJ(RICts+ I)]
-0
- -90
Comparing Eq. 9-22 to Eq.-. 9-20,
.
rl = R3C2
fl 12 f3 Frequency
3 = RlCl
T3= RtCl @1
FIGURE 9.8 (a) Operatima1 a r n a e r with Ledback Impedances and tb)its @
. . 4 = RzCI Bode plot showing zero-pole pfrs. Tbe phase i s also shomr. -- ..
www.IranSwitching.ir ..
2.46 HIGWREWENCT $wlTwwG POWER SUPPUES SldBlUlY IN SWllCHlNG POWER SUPPLIES: ANALYSIS AND DESlGN 247
The gain at fi and fi is
The Bode plot break or corner frequencies are determined by . Z
Galn, Phase,
and dB degrees
0- -0
In an actual design application, the break frequencies are.normally pre- 4

determined by the design objectives. Then the values of the resistors and
I (-270)
I
capacitors may be easily calculated using Eqs. 9-23 through 9-27. The circuit
shown in Fig. 9-8a may be used in any PWM switching power supply for
error ampliiier compensation, with the overall loop gain crossing over be-
tween& andf, in order to achieve a - 1 slope at 0 gain crossover, which is
@I
the final objective of loop stability analysis. 1
FlGURE 4 9 ( a ) A single-pole feedback amplifier and Ib) i t s gain a d . phase
Some other popular error amplifier compensation networks have been plots. Also kwmr as a type 1 amplifier. 4
published, and they are presented here. The simplest form of feedback
#
amplifier with a single-pole rollo$ k-shown In Fig. 9-9. The transfer function
of the amplifier in Fig. 9-9 Is follows
with a break frequency
Another amplifier cclnfiguratton Is shown in Fig. 9-10, where a zero-pole

pair has been introduced to give a region of frequency where the gain is flat The amplifiers of Figs. 9-8 and 9-10 offer improved power supply transient
and no phase shift is introduaed. I'hc region with constant gain occurs be- response when the supply is subjected to output Ioad changes, as opposed
tween the break frequencies fi and fi, This region must be used for loop to the dow response of the amplifier shown in Flg. 9-9.
gain crossover, when the c h i t is used as an error amplifier in PWM power Although it seems to be cornpiicated, the amplifier depicted in Fig. 9-11
supplies. win give a very good transient response. In this chit two zero-pole pairs
The previously discussed method of analysis may be used In, this circuit have been introduced to give a region of frequency where the gain is in-
also to derive the gain and break frequencies, the results of which are as creased at a + 1slope with a 90" p h a s e k d . The performance of this amplifier
. .. ... . . .. -. I
When the circuit of Fig. 9-lla is used as a compensated ermr amplifier in
a PWM switching power supply, loop crossover should occur between f,
and f3 for M e r results.
1I Although many more circuits may be used as error mplifiers with a m -
1 . pensation, the four circuits presented here should be adequate to use in the
!
I
Phase,
degrees
1 1 I 1
fl f2 Frequency
@I
FIGURE 9-10 (a) ~n operational ampli6er with a m p o l e pak and (b)its gain
and phase p h . Also known as a type 2 amplaer.
is very similar to the one shown in Fig. 9.8. The pins and break frequencies
are hund to be as follows:
v
1f1=f,l
fr f2 fg fa Frequency
FIGURE 9.11 (a) An operational mpUfier with iwu zero-& pairs and
(6) its gain and phase plots. Also known as a type 3 ampU6er.
. ....-....-. .-. . . ..
ZW HIWMWENCY SWITCHING POWER SUPPLIES STABUrY LN SWiTWNG POWER SWIPUES: ANALYSIS AND DE%N 2H
majority of PWM switching power supplies for loop stability analysis and +40 +1BO
Gain,dB P h w , degrees
design. The following example shows step-by-step stability analysis fn an
actual switching application, using theory presented in the abwe paragmphs.
Consider a half-bridge power supply designed to accept both 90 to 130 V

ac or 180 to 250 V ac, working at 20 kHz and using the UC3524A PWM
control circuit. The LC wtput filter has been designed to have a resonant
corner frequency at 1 kHz. The power transformer primary-to-secondary
turns ratio N p / N s = 18. Design the error amplifier compensation network
in order to achieve overall power supply stability, and sketch the overall
loop gain Bode plot.
SOLUTION:
First let us choose an amplifier configurationfrom the four circuits presented
in this chapter. Although with careful design all the amplifiers will work, FIGURE 9-Le ~ontrol-t&outputtransfer functioll of Example 9-1. .
the one in Fig. 9-11 is chosen, sink it will give good transient response.
The second consideration is to choose the crossover frequency where the
gain is unity and the Bode plot crosses at a - 1slope (-20 dB per decade). Therefore, for an overall loop gain of zero, the fe;dback amplifier gain must
Theoretical limits set the crosswer frequency at halfthe switching frequency, be made + 12 dB at 4 kHz.
but from practical experience less than one-fifth of the switching frequency The important thing to remember here is that the switching power supply
is used. For this analysis we choose a crossover frequency of 4 kHz, which will be stable if the overall loop gain crosses wer the 0-dB line at a - 1
is one-fifth the switching frequency and one-tenth the modulator frequency. slope. Since the controko-output gain of the converter k Mhng at a -2 1
Since we are using the UC3524A, the control voltage V, swings 2.5 V to slope, as depicted in Fig. 9-12, the feedback ampMer must supply a A 1
change the comparator's drive waveform from 0 to 1. We also take the worst dope at this point for a resulting slope of - 1 (-20dB per decade). Again,
case of input voltage, i.e., 130 V ac. The control-to-output voltage gain using the feedback amplifier gain at 4 kHz must be +12 dB (or 4.0) with a +1
Eq. 9-15 is slope. Since the input line voltage to the switching power supply swings
+
from low to high line, the 1 slope must have some margin to span the
(7;):
(dc gain)dB= 20 loglo - - = 20 loglo --
;8) (z m g e of crossover frequencies as the modulator's gain changes with input
voltage.
Now, let us find the ~nodulatorgain at 1 kHz. It will be
AV, = -
kHz (4.0) = 1.00 or 0 dB
4 kHz
The wtput transfer functiol~characteristic is depicted in Fig. 9-12. Al-
tbwgh in practice the asymptote of Fig. 9-12 has a break frequency due to Let us assume, then, the following characteristics for the feedback amplifier
the ESR of the output capacitor, its effect on the overall loop gain in this and plot i t Bode graph. The gain is + 12 dB at 4 kHz and 0 dB at 1 kHz.
example is of no impofince, and therefore it has been omitted for simplicity. We desire a double zero at 1 kHz, a pole at 10 kHz, and a second pole at
From Fig. 9-12 by inspection, the control-to-output gain is + 12 dB at 30 kHz. The Bode plot is given in Fig. 9-13. From the graph we get
low frequencies, rolling off above 1 kHz at -40 dB per decade, so that at
the chbsen crossover frequency of 4 kHz the control-to-output gain is - 12
dB. The. fact
- that the magnitude of both gains is 112 dB1 is purely coincidental.
- .- "
252 HI&RE~~~ENCY SWITCHING POWER SUPRLES STABlLlfY IN SWITCHING POWER SUPRIES: ANALYSIS ANU UESlGN 253
and From Eq.9-38

f,-f,=lkHz
f, = 10 kHz
f, = 30 kHz * The final amplifier design and the overall loop gain are shown in Fig,
1 1
9-14. Figure 9-14 is plotted by adding the graphs of Figs. 9-12 and 9-13.
Referring to Fig. 911a and Eqs. 9-33 through 9-38, the valueshofresistors Notice &at the w e d gain crosses the 0-dBline (unity gain) at 4 kHz, at a
and capadtors to give the required results depicted in Fig. 9-13 are calculated
as follows. From Eq. 9-33, assuming R 1 = 10 kn,
R2 = AVI(RJ = (1.m10 kR) = 10 kf2
From Eq. 934
From Eq. 9-35
From Eq. 9-36

1
Gain,
lope
0.1 1 4 10 100 Frsqurncv, kHz FIGURE 9-14 The compensatd feedback mpliEer and the werali system l w p
FIGURE 913 Feedback amplfir Bode plet showing deslied fre- gain plot of the swltchiog power supply of Example 91. (Resistw %, has no
quency and gin characterlstlcs. effectand isnot sharm for simplicity.)
.-
s 2Sa HIGH-FREQUENCY SWITCHING POWER SUPPLIES
-$ - 1slope, as desired. Of course the crossover frequency will change as-the

9 input line voltage is varied over the 90- to 130-V ac (or 130- to 260-V ac)
range, but the crossover will still be at a - 1 slope. The reader may veriEy
h this by plotting the loop gain Bode curves at low line, i.e., 90 V ac
(180 V ac). ..
9 4 STABILITY M I U S I S AND SYNTHESIS USlW THE K FACTOR

In the previous paragraphs the subject of stability analysis and synthesis was ,
discussed, and practical mathematical tools were developed to design com-
pensated error amplifiers to be used with any type of+switching regulator.
Three basic amplifier types were presented and analyzed, and the design
equations given with each type have universal usage (see Figs. 9-9 to 9-11).
The following paragraphs will present a new powerful mathematical tool,
known as the K factor, which makes the analysis and synthesis of error
amplifiers even easier, The techniques presented here are vaIid with any
type of modulator used, whether buck; boost, or buckboost.
It is important to note that in all cases a real operational amplifier is used,
either external or part of the IC m t r o l circuit, and that the system is
configured so that negative feedback is required in the amplifier.
9-64 The KFactor

The K factor is a mathematical tool for defining the shape and characteristics
of a transfer function. No matter what type of amplifier is chosen, the K
factor is a measure of the reduction of gain at low frequencies and increase
gain at high frequencies, by controlling the location of the poles and zeros
of the feedback amplifier Bode plot, in relationship to the loop crossover
Frequency f.
Figure 9-15a shows that for type 1 amplifiers K is always 1. This is due
to a total lack of phase boost or corresponding increase or decrease in gain.
For amplifier types 2 and3, as s h m in Figs. 9-15b and 9-15c, the zero
is placed a factor of K below loop crossover a d the pole a factor of K above.
Since f is the geometric mean of the zero and pole locations, peak phase
boost will occur at the crossover frequency. For either case, as K is increas-
ing, so is the phase boost.
FIGURE 9-19 he Bode lot characteristicsof(a1 type

1 ampllfiw, (b)type2 amplier, and (c) type3 amplifier,
It is widely known that phase boost, due to a zero-pole pair, is the inverse in relation to the K factor. ~ l coordinates
l are logarlth-
tangent ratio of the measurement frequency to the zero or pole frequency. mlc. (See also Figs, 9 9 to 9-11.)
The total p h s e shift then is the sum of aII individual zero and pole phase
shifts.
-..*..
.-. -
2W HGH-FFZQUEMCY MCHlWG WWER WPLlES STABILITY H M C H I M G POWER SUPPUES: AMALYSlS AND D E W 257
Step I : Make a Bode plot of the modulator. A t y p i d mdulator B d e plot

showing both gdn and phase characteristics is depicted in Fig.
9-12.
I Step 2: Choose a cmssouerfrequency. The crossover frequency is the point

where you want the overall loop gain to be unity, Remember that
!i the higher the crosscwer frequency, the better the transient re-
sponse of the power supply. However, practical limitations restrict
the range of the crossover frequency. The theoretical limit is half
I the switching frequency, but practical considerations have proven
i that a crossover frequency figure of less than one-fifth of the clock
I
i frequency is a good choice.
r
Step 3: Choose the desired phase margin. This margin is the amount of
phase desired at unity gain, as shown in Fig. 9-4 and 9-14. Phase
margin may have a range of 30 to 980, with 60P being a good
K factor compromise.
FIGURE 9-16 Phsse h s t w. Kfactor for type 2 and type 3 feedback
ampIlers. Step 4: Detemfne required amplger gain. This gain G is the required
amplifier gain at crossover and must equal the modulator loss.
For type 2 amplifiers the boost at frequency f is given by the equation When expressed in decibels, the amplifier gain is simply the neg-
ative of the modulator gain; otherwise, amplifier gain = Itmod-
ulator gain.
Boost = tan" (K)- tan-' (K)tan (9-39) I
Step 5: Calculate the required phase boo&. The amount of phase boost
and from this equation it can be shown that, required from the zero-pole pair in the amplifier is given by the
formula c.
boost
K = tan + a]
For type 3 amplifiers the boost at frequency f is given by the equation where M = desired phase margin, degrees
P = modulator phase shift, degrees
Step 6: Choose the amplger type. Choose amplifier type 1when no boost
is required, amplifier type 2 when the required host is less than
and subsequently, 900, and amplifier type 3 when the required phase boost 5 less
than 180'.
Step 7: Calculate the K factor. The K factor may be calculated using either
The above equations yieId the phase boost vs. K-factor cumes depicted Eq. 9-40 or 9-42 or directly from the curves of Fig, 9-16. For
graphically in Fig. 9-16. These curves are universal and may be used to amplifier type 1, K = 1. Location of the transfer function poles
readily determine the K factor for a given phase boost. and zems will determine the circuit values. The pole at the origin
causes the initial - 1gain slope, and the frequency where this l g e
crosses or would have crossed the 0-dB line (unity gain), is the,
9 4 - 3 Symesls of Feedback Amplifiers WllhIng the K Factor unity gain frequency, U GF.
Using the K f ~ t o to
r synthesize an amplifier, the following steps are rec- The following equations ~rovidethe component values of each
ommended. www.IranSwitching.ir
I amplifier type.
. . .. .
STABlLlTY IN WITCHING POWER SUPPUR ANALYSIS AND DESIGN
fl
\
Using Eq. 9-42, or reading directly from the curves f n Fig. 9-16, to achieve
Type 1 125' of phase boost, we need K = 16. Now the component values of the
";r (Figs. 9-9, 9-15~):
feedback network of the amplifier shown in Fig. 9-18 may be calculated
3 1 using Eqs. 9-50 through 9-54.
Type 2 UGF = (9-45) To facilitate the calculations, resistor R,is arbitrarily chosen to be 10 kfh.
(Figs. 9-10, 9-15b): 2nR3(C3
3 Hence,
3 1
Type 3 = ZnR,(C, + CJ (9-49)
(Figs. 9-11, 9-15c):
-4
Since the crossover frequency f is known to be 4 kHz, by inspection of

the curve shown in Fig, 9-15b we note that a double zero ti located at a
hequency f l f i belaw the crossover frequency, while a double pole is
located at a Frequency of above, I
The double-zero bcation is-

f L =
f
- = = 1 kHz
trK fi
The double-pole bcation is
& = M = 4 f i = 16kHz
For a more accurate lomtfon of zeros and poles, Eqs. 9-35 through 9-38
Consider the power supply requirements given in Example 9-2. Design the could have been used, but for all practical purposes the above results are
error amplifier compensation network to achieve overall loop stability, with acceptable. The amplifier Bode plot is shown in Fig. 9-17.
. a phase margin of 60" at the unity crossover frequency. Unity gain crossover occurs at 1kHz (verrfy.,using Eq. 949). Note that
m m N the results of this example are the same as the ones p~eviouslycalculated
in Example 9-2.
From the modulator Bode plot depicted in Fig. 9-12, at acrossover frequency The only difference is the values of C2and R33which resulted in rnakng
of 4 kHz (one-tenth of the modulator frequency and one-fifth of the clock the poles previously located at 10 and 30 kHz coincide as a double pole
frequency), the amplifier gain must be 4 or - 12 dB, By inspection of the Imted at 16 kHz. This is a direct result of the K-factor assumptions, of a
I
same plot, the phase curve and the crossover frequency line show a m d - type 3 amplifier, the zeros and poles are coincident. Of course the feedback
ulator phase shift of 155'. component values may be adjusted using Eqs. 9-35 through 9-38, so that
Using Eq. 9-43, we calculate the required phase boost: i zeros or poles or both are spread apart. The net effect of this is to broaden
Boost = 60 - (-155) - 90 = 125" 1 and flatten the phase bump, thus reducing the phase margin at crossover.
&$W~?EQJQ~CY SWRCHlNG POWER W S
ac line input frequency. Such a load change forces the feedback amplifier
fmm an open-loop to a closed-loop condition at the end of the recovery time.
Figure 9-19 shows t y p i d transient responre traces for a 225 percent
load change Tne switching waveform of Figure '9-19a causes the output
voltage of the switcher to "dip" or "jump" at the square wave rise and fall
edges, m e magnitude V. of these transients depenh primarily on the ESR
~odbie
@ 1 kHz
' ' Gnlv baln
Imqmncy
WGF)
1 -60
-10 I
I, i 1
j
n .r 1 4 10 16 100 Frwuency.kHz
of Example 9-3.
Optimum performance therefore is achieved when both rem and poles are
coincident, although it is strongly suggested that each switching regulatar
w i t is for its own merltf and acwrdin$ly tailor the amplifier
design to meet optimum p d a n c e .
We have therefore successfully completed the design of the feedback
ampl&er to produce the desired werdl system imp gdn and phase plOt
shown in Fig*9-18. Hence, the overall gain plot crosses the 0-dB line (unity
gain) at 4 kHz, at a - 1 slope up to 16 kHi, at which p i n t it f& of at a
- 3 slope. At the same time the phase plot shows a 60' phase margin at the
unity gain crossover, a3 desired, assuring a stable system operation.
9-7 LQOI STABIUTI ME*SUEMMS

-. Although there are numemus methods of measuring the overall loop gain
of a switching power supply (see references), a very simple and useful way
of deriving information on closed-loop stability is by measuring the power
supply's transient response, Transient response i s measured by switching
the output load from 75 to 100 percent of its full value at a rate of twice the
TEN
ELECTROMAGNETIC AND
RADIO FREQUENCY
INTERFERENCES EMl-RFl
CONSIDE ATloN 6 1
United States and international standards for EMI-RFIhave been estab-

lished whicb require the manufacturers of electronic equipment to minimize
the radiated and conducted interference of their equipmen? to acceptable
leveb. In the United States the guiding document is the FCC Docket 20780,
while internationally the West German Verband Deutscher Elektronotech-
niker (VDE) safety standards have been widely accepted.
It is very important to understand that both the FCC and VDE standards
exclude subassemblies from compliance to the rules; rather, the final equ*
ment, where the switching power supply is used, must comply with the
EMI-RFI specifications, Rightly so, since even if the switching power supply
has an input filter, this filter is matched to the power supply when passive
bads are powered, and its characteristics and suppression capabilities may
drastically change when used to power active electronic circuits.
This chapter attempts to introduce the reader to the conducted RFI
problem and gives some suggestions for minimizing it, whether it is applied
in a power supply or a final system.
10-1 THE FCC AND VDI CONDUCED NOISE SPECIFICATIONS

Both the FCC and VDE are concerned with RFI suppression generated by
equipment connected to the ac mains employng Mgh-frequency digital cir-
cuitry. The VDE has subdivided its RFI regulations into two categories, the
first being unintentional high-frequency generation by equipment with rated
frequencies from 0 to 10 kHz, i.e,, VDE4875and VDE-0879, and the second
dealing with intentional bigh-frequency generation by equipment using fre-
quencies above 10 kHz,i.e., VDE-0871 and VDE-0872.
- E AND RAW FREQUElrlCV INTERFEEICES (EM-RFI) CONSlDERATlONS 267
The FCC on the other hand includes in its SF1 regulations d electronic
devices and systems which generate and use thing signals or pukes at a
rate greater than 10 kHz. F i e r e 10-1 summarizes the FCC and VDE FiFI
requirements.
The FCC E MI-RFI regulations closely follow those of the VDE.The FCC
class A specification covers business, commercial, and industrid environ-
ments, and compliance to the specified EM1 emissions in decibel-pvoIts can
be met by any equipment meeting VDE66875IN or VDE-WTl/A,C.
On the other hand, FCC class $ requirements cover residential environ-
ments and are more stringent than those of class A, Both FCC conducted
E M I - M I specifications, however, cover the frequency range from 450 kHz
I to 30 MHz. Tbe V D E reguhtions extend below the W k H z range; in Eact
the VDE frequency range Eor EMI-WI conducted emissions covers a spec-
trum from 10 kHz to 30 MHz. Figure 10-2 shows the FCC and VDE curves
for conducted MI emissions,
40-2 RR SOURCES IN S m l l S Q POWER SUPPLll

\
Every switching power supply is a source of RFI generation because of the
very fast rise and fall times of the current and voltage wavefarms inherent
I FCC Class A
/ VDE-a7lJA,C
VDE-08751~
Il VDEa71lB
w /VDE~~~WN-~Z
5 I I R
defined luter- i
I I
Fraqueney,MHt
FIGURE 10.2 FCC a d M E + a h cums showing madmum permissl%le I
W
emisdons in d e c i b e l s - m W l t om d t e d h.
4
268 HIM-FMQCIENCY SWITCHING POWER W R I E S ELECTROMAGNETIC AND R A W FREQUENCY INTERFERENCES (EMCRFI) CONSlDERATlONS 269
1
In the converter operation. The main sources of switching noise are the of the power supply. On the other hand, filtering conducted noise becomes
switching tmnsistor(s), the mains rectifier, the output diodes, the protective much easier as the working frequency of the power supply Is increased.
diode for the transistor, and of course the control unit itself. Depending The resistor R across the ac lines of the filter is a discharge resistor for
upon the topology of the converter used, the HFI noise level at the mains the X capacitors, and it is recommended by the safety specifications of the
input may vary from bad to worse. 1
C
VDE-0806 and IEC-380. In hct SEC-380 Sec. 8.8, states that if the RFI X
Flyback converters, which by design have:a triangular input current capacitor is above 0.1 pF, a discharge resistor of the following value is
waveform, generate less conducted HFI noise than converters with rectan- required in the circuit:
gular input current waveforms, such as feed-forward or bridge converters.
Fourier analysis shows that the amplitudes of the high-frequency harmonics
of a triangular current waveform drop at a rate of 40 decibels per decade,
compared to a 20 decibel per decade drop for s coinparable rectangular where t = 1 s, and C is the sum of the X capacitors (in microfarads).
current waveform.
I 10-3 AC INPUT LINE ALTERS FOR RA SUPPRESSION

The most common method of noise suppression at switching power supply
EXAMPLE 10-1
Calculate the discharge resistor R for the filter of Fig. 10-3, given that
C3(X) = C4CX) = 0.1 pF.
ac mains is the utilization of an LC filter for differential- and common-mode
RFI suppression. Normally a coupled inductor is inserted in series with each SOLUTION / .
ac line, white capacitors are placed between lines (called X capacitors) and
between each line and the ground conductor (called Y capacitors). Using Eq. 10-1
I The capacitance and inductance of the components may be within the
1 Following values:
f Further reduction of the symmetrical and asymmetrical interference volt-

age may be accomplished by the insertion of an extra line choke, L2,#as
shown In Fig. 10-4. Insertion of choke & leads to a limitation oftfie charging
current of capacitor C4(X).
Figure 10-3 depicts a standard switching power supply input line filter.
Although the described circuits will suppress the generated RFI to ac-
During filter coinponent selection it is important to make sure that the
ceptable levels, it is important to understand that if the power supply pack-
resonant frequency of the input filter is lower than the working frequency aging or layout changes, a certain filter may or may not work properly. To
elaborate on this claim, if a power transistor or power rectifier which uses
Load 1 -
C 3 h l /
FIGURE 10-3 A mitchlngpower supply input line filter fDr ac

mains RPI noise upp press ion. FIGURE 10-4 I m p 4 ad b e filter incorporatingtwg line chokes.
#
270 HIGH-FR-W SWIK;HlNG POWER SUPPUES
high-frequency waveforms Is directly mounted on the chassis of the power

supply, with only a mica insulator between the ~ w o ,and if the chasrs is
connected to the ac ground conductor, generated RF noise will be coupled
into the ground conductor, thus upsetting the effectiveness of the particular
mains filter. It has been shown that a TO-3switching transistor working at ELECTRICAL SAFETY
20 kHz with a 20-V input, m o u n d on a groun? heat sink through a mica
insulator, will generate an RF current of 1 mA at 1 MHz. A solution is to STANDARDS
sandwich a metal shield between the insulators and to return the shield to
the dc ground. This technique dectively "shorts out" the capacitor created
by the mica insulator, reducing RF noise currents.
Power supply and system layout are very important in reducing or elim-
inating RFI-EM1problems. The designer should take care in analyzing all t
potential proMems before the proper line filter is chosen.
For complete information on EMI-WI regulations, the following original

documents are recommended. National and international safety regulatory agencies have established eke-
1. VDE-W/B.T7 " S p d c u t b a for the Radio Interference Suppression of trical safety standards, formulated and directed toward the manufacturing
Equipment, Machines,and Systems with Opsrational Fre- of equipment andlor electrical components, to provide the end user with a
quencies h 0 to 10 kHz." safe, quality product. These standards aim to prevent injury or damage due
2, VDE437116.78 "Regdationsfor tbe Radio Frequency Intederence Suppres- to eleclrical shock, fire, mechanical, and heat hazards, etc.
sion of High Frequency Appratus and Installatfons," In general, each country may impose local standards for electrical safety,
3, FCCDocket20780 "First Repwt and Order br T e h i c a l Standards for Com-
puting Equipment," part 15, subpart J.
t but most power supply manufacturers use the IEC (International Elwtm-
technical Commission), VDE, UL (Undenvriten' Laboratories) and CSA
,
4. Docket 80-284, FCC 81-63 "FCC Methods of Measurement of Radio Noise Emissions (Canadian Standards .~ssociation)standards as a de facto solution to the
from Computing Devices," majority of the world's safety requirements. The West German safety stan-
dard for business machines, VDE-P806, is h e d on the IEC's recommen-
dation IEC-380,and it is by far the most stringent electrical safety standard
for power supplies, For the United States and Canada, power supplies are
generally designed to meet safety standards for data processing equipment,
i.e., UL-478 and CSA C22.2 no, 154-1975, and ofice equipment safety
standards, i. e., UL-144and CS A-4222.2 no. 143-1975.
In this book VDE, UL, and CSA safety standards refer to the abwe
requirements, unless otherwise specified.
114 POWER -PLY CONSTRUCTION ZHQUllMMS KIR SARCT

94-44 Spacing Requirements -. .
The UL, CSA, and VIDE safety specifications impose specific spacing re-
quirements for between Jlve parts and between live and dead metal parts.
The UL and CSA require that high-voltage conductors of opposite polarity
/ #
272 HIW-FREQVMCY SW~TCWNG

POWER WPW
t
up to 250 V ac, or high-voltage conductors and dead metal parts, other than
field wiring terminals, must have a separation distance of 0.10 in., either
over surface or through air. The VDE requires a 3-mrn creepage or a 2.0-
mm clearance distance between ac lines, and a 4-rnm creepage or a 3-mrn
clearance distance between ac Iines and tbe grounding conductor. The IEC
has even tougher specifications, requidnk a 3-mrn clearance distance be-
tween ac Iines and a 4-mm clearance distance between ac lines and the
grounding conductor. In addition the VDE and IEC require a full 8-mrn
spacing between the input and output sections of the power supply. Notice
that what the UL calls separation over surface, the VDE calls ,creepage,
while the UL.definition of separation through air corresponds to the VDE
clearance distance, 1
Figure 11-1shows the distinction between the measurement of clearance
and creepage distances, h Fig. 11-10 the path under consideration includes
a V-shaped groove with an internal angle of less than 80" and with a width
greater than 1 mm, as well as a parallel or converging-sided groove of any
depth with width less than 1 rnrn. The rule in this case states that clesrance
is "line-of-sight" distance, and it is measured over the grwves. The creepage
distance is measured on the s u b of the grooves, as shown, but short-
circuits the bottom of the V-shaped m e by 1 mm. The contribution to
the creepage of any groove less than 1 mrn wide is limited to its width; that
is, only the clearance distance applies. Figure ll-lb shows a path which
includes a rib. in this case, clearance is the shortest direct air path over the
top of the rib, while the creepage path follows the contour af the rib. 4
Fire 11-2 shows different examples of printed circuit board design to
achieve clearance and creepage distances between primary and secondary
'circuits of a power supply. As shown m Fig. ll-2a, if the primary circuit
track is opposite to the secondary circuit track, the thickness of the printed
circuit board must be 2 mm minimum. When the printed circuit board is
greater than 1 mrn but less than 2 mm, then the primary and secondary
circuit tracks must be separated by at least 3 mm, as shown in Fig. ll-2b.
If the primar)~and secondary circuit tracks face each other, as shown in Fig.
ll-2c, then the full 8-mrn clearance dfstance applies.
11-192 Dieiectrio Test Wibtund

I
For equipment rated 2% V ac or less, both UL and CSA specification;
require an input-to-output and fnput-to-ground hi-pot isolation test of 1000 '
V ac for 1 min, or 1200 V ac for 1s. This ac potential must Ix a sine wave
FIGURE 11-1 Measurement of dear- and weepage distances as &%ed in the
of 50 or 60 As. VIDE safety standards. (a) Path d t h a V-shaped ~IWW, (bl pth Including a rib,
The VDE requires the Mowing dielechic tests: 3750 V ac between each
input ac line and secondary extra low voltage (SELV) output circuits; 2500
V ac between ac lines and the grounding conductor; 500 V ac between the
. .- ., >
,.
POWER m y aEMRlCAL SbffM STANDAROS 27s
t
grounding conductor and secondary SELV output circuits; and 1250 V ac
between ac input lines. All tests are 1 min in duration. The test may be
reduced to 1 s if all test voltages are increased by 10 percent.
4 4 - 4 4 leukage Current Meusuremenis

Secondaw circuit
. The UL and CSA require that all exposed dead metd parts must be earth-
grounded and that the leakage current measured thmugh a 1300-nresistor
connected to earth ground must not exceed 5 mh.
The VDE allows the following leakage values, measured at 1.06 times
rakd voltage through a 1300-Rresistor in parallel with a 130-nF capacitor:
for portable office equipment (<23lcg), 0.5 mA; for nonportable ofice equip-
ment, 3.5 mA; and for data processing equipment, 3.5 mh, maximum.
It is ihteresting to note that Japan a1Iow.s a maximum leakage current of
Primary circuit 1mA, measured through a 1000-il resistor, b r line frequencies up to 1kHz.
For higher leakage currents an isolation transformer at the installation is
I
required. For line Frequencies above 1 kHz the maximum leakage current
I is logarithmically increasing tb a vaIue of 20 inti at 30 kHz.
Secondary circuit
14-4-4 Insulation ResJS?Wlc8
VDE requires 7.0 minimum resistance between input and SELV output
circuits, and 2.0 lvin between input and accessible metd parts, 9 t h an
applied voltage of 3 0 V dc for I min.
#
444-5 PC Board Requlremerb

The UL and CSA hmmability standards apply, i. e., all pc boards must be
ULrecognized 94V-2or better material.The VDE accepts these standards.
Primary circuit I 1b n d a r y circuits
11-2 POWER SUPPLY TRANSWRMER C U N ~ O FOR
N SAFEW
Since the VDE standads for the design, manufacture, and utilization of
transformers impose the most shingent specifications satisfving the majority
a f s a f e p ~ e m m x t s a f o t h e r c ot h~e y d k p r e s d h e r e f n
IcJ
depth. Because the VDE has no flammability requirements for t r a n s h e r
FIGURE 11-2 Proper printed dimit h r d design to meet W E clear- construction, the UL standards may be used, which require that all material
ance and creepage distance rqdrements beh*een primary and second- used in the construction of transformers must have a NV-2 or better sating.
circuits of r pawr supply. ( a ) With primary circuit track oppwite see
ondary CIPCOjt tradl; ( b ) witb Phteti circuit bpard greater than 1 mm hut
less than P mm;(c) with primary and gecDndary cim& facing eacb other, 14-2-1 Transformer lnsulutlon
The windings of n transformer must be separated physically by insulation in
'accordance with tbe requirements shmm in Fig. 11-3 and Table 11-1.
" .
..- -- - .
276 HlGKFEgUENCV WITCHING POWER WmES
TABLE 11.1 DISTASCE THROUGH TRANSFORMER INSULATION

Worb'ng cobage
"0 U<50° 50rUS250 v>m
1 1 ply? - -
2 1P ~ Y 2 plies or 0.5 2 plies .or 0.8
3 3 plies or 0.5 or 3 plies or 0.5 or 3 plies or 0.8
2 piieslscreenl2 plies 2 piieslscreenl2 plies
a - 3 plies or 2.0 or -
2 plieslscreenl2 plies
-
symbol C signifies the swrlbnq dtege W the indimted points.

% minimum thickness lor piics i s A l mm.
and core and frame and screen, with an applied voltage of 500 V dc for, 1
41-24 Tmnsformer Creepage and Clearcince DWernces

The spacings bemeen windings; between windings and terminals, screen,
core, Frame, winding crossover leads; between terminals; and between ter-
minal-core and frame-must be in accordance with the values shown in
Fig. 11-3 and Table 11-2. Creepage and clearance values are based on the
FTGURE 11-3 Distances through transformer insulation m specified I assumption that the winding wire is coated with varnish and the like. I
by the M E .
Enamel, lacquer, or varnish coatings on winding wire or other metallic parts,

asbestos, and moisture-abwrbing m a t e d are not considered to be insulation The transformer must be capable of complying with insulation resistance
within the meanins of this requirement requirements and d i e l e c ~ cstrength requirements immediately after the
tmnsfonner has been subjected to adverse humidity conditions, where the
relative humidity is 92 2 2 percent and the stabilized temperature value
14-2-2 Transformer Dielectric Wen* behveen 20 and 30°C with a stabilization factor of 2l0C. The duration of
F a e n multiple layen of plies of insulation are employed, any two la?-ers conditioning is h, minimum. The transformer may be temperature-sta-
must be capable of withstanding the dielectric strength value shown in Fig. b i k e d not more than C C greater than the humidity conditioning temper-
r lb4,where the layers are in contact and where the test potential is applied ature value prior to conditioning.
to the outer surfaces. The applied ac potential must be a sine wave of 30 or
60 Hz, and the test duntion must be 1 min. S o insulation rupture or
flashover- may
. occur during the dielectric strength test. 14-24 VDE Transfonner Temperahre Rding
The m a h u m stabilized temperature under normal operation for a specific
1112-3 Tramfomer lnull'atlon ResMunce insulation class must not m e e d the temperature value of the insulation
classes as shown in the table below. Consideration must be given to the
The insulation employed in the cqns.tnrctron of a transformer must possess utilization ambient temperature within the product or prnver supply during
'.
a minimum resistance of 10 hIR between windings, and between windings , temperature evaluation. .---. -
PCWER W R Y EK:CTRICAL S
AW STANDAM 279
-
lnput
+
-
-
lnput
cairn poinfe. Tabla 11 -2. fim value appllm acDnd mtue in panrr
w@+@~@whWItha£afmty-is
@ dYew3 h not apply,
8
rmirted, ot the me.
2500: or 2U + 2M)[l when U > 250.
b nongwndsd,
8 1.6whonU<130;2~0wheraIf>130.
Table 11-2. srecrnd value in paren- appllcr.
2500:or3750whrn~safety1ueeniromi~ed
@ mw@+@=3m*me*~crew
iromittEd.~tth#~we.~:..ia~ded
8 5.0: 3.6 for tiensformsrs rwd 80 Hz only.
3.6 where U < 1 3 0 2 0 minimum w Table 31-2
whan U > 250, fimvalue.
I @
@
The irwlrcared value may b M m
U<50.
500 Y when
Whenene~wa safety rn Is mphyed. it b pnferttd

NOW Whare produca; are raod for Hz d y . d a 6 w
ship
sh m .
@becomer the m e ar nfarlomhip @ #
that tha corn, etc. b cwvwctcd to tha gwmding

ODnBCt
FIGURE IIG Transformer creepage and clearanre distamw.
/
280 HIGH-FREQUENCYWTCHING POWER WlllES WWER SUPRY !ZECTKN SAFFlY S A N D A R E z&l
TABLE 11-8 CREEPAGE AND CLEARANCE REQUIREMESTS IN SECOXDMY

CIRCUITS (VALUES IN MXUMETERS) the change of resistance method may still be employed as a find determi-
nation of compliance.
AlInirnum ~tfinirnum 41-2-7 UL and CSA Trunsformer Ternperdure Rmng

clearance cwepage
The UL and CSA rate tnnsformer temperature as a rise above ambient
(E0C).Two methods are employed in temperature measurement, namely
the thermocouple or the resistance method. The following table shows ac-
ceptable temperature rises.
-. . ..
REFERENCES
. .- .
No&: If p d w t s are r n & d ~ ~ > 1 ywtth a &voltage of <I00 IXMS 141VPKldc or If output IS For in-depth and cmplete information on electrid safety standards. the
<2M VA, there nre.nc s
-" L
--
:
@k sprig rquirementr and compliance 3 determid by the dklwtrk
"
'
A
'
reader is referred to the follmving original documents. I
-
: .--- -"
-,
--- 1. UI.47'8: Electronic data pmcffsing units and s?.skrns.
#
2. CLLM: OBce appliance4 and business equipment--el&c.

3. C S b C X Z So. &I9;Z Data processing equipment
4. CS.4-CEZ So.1Gl97Z Office machines.
5. IEC-380: ace machines id* of electridl!. e n e q i d o C ~ ema.
A 1105) ubinesl.
G (For %Hz usage only)
I131 6. LEC-433: Ddta procesdng equipment.
F (155,h60-Hz usage only) 7. lPEM301Part 2P: r p h i o n s for bushas machines.
~ c u l am
H (180, for WHz usage only) 8. \:DE-OS06/6.81: Safety of e l e r t d l y energad o h equipment
Temperature measurements are secured by the change of resistance

method, where the transformer input voltage is supplied at 1.06 times the
nominal rating and with the frequency at 50 Hz for transformers nted at 50
Hz, 50 to 60 Hz,or 50160 Hz and at 60 Hz for transformers so rated.
Thermocouples may be employed for measurement purposes as a means
of acceptance, when the temperature d u e s shown iii;the table are reduced
- by.lS°C. Whenwww.IranSwitching.ir
noncompliance is determined by the t h d n n ~ ~ ~ umethod.
ple
n .
Index
AC brownou( 229 Capadtors:
AC Iine filters, 269 couplh, 31
AC fine lo= detectors, 228-230 ESR, 8, 179
Air gap, 19,44,55,56,113 input rdter, 8
length of magnetic path of, 82,113,126 output fdkr, 178
American wire gauge IAWGI, 118,220, X, 268
121 Y,268
Amplifw Charachxktic equation, 239
compemtea, 244,287,289 circuit($):
error, 187,191,198,202,204,208, antisatumtion, 6%71,82
239,241,243 base &ve, 71,7&82,88 *
high** 173 central:
magnetic bee Magnetic ampMers3 current-mode, MO
type 1,247,254,256,257 *width-modulaWd, 185, I87
type 2,248,251,256,257 ament limit, 217,219,222 I
type 3,249,254,256,251 hiccup mode in, 223

(See dso O p a d o t d amplinm) w w v e l p-4
~ 223 I
An-on Wts,68-7f &2 Zener msct,223 C
Asymmetrfd Interference voI- 269 semndary sx2ra Iow voltage,272

SF-bia8, 211
clamps,6&71,82 &~~resonatlt,39
Base drive &xi&-, 71,M mfk starb, 211,217
propordonal, 1 2 Q
- distanq 272-274,277,280
Bipolar synchronouswMer WSYM, 147 Closed-loop feedback mtrol system, 237,
Bipolar1 " ,66 239
base drive bddques for, 71 C1&-Ioop m f e r function, 237
junction dmching of, 86 Cowdve force, 112
p k c t i v e networks of, 86 cd,109,110
secondary b m k d o m of, 83 Commm mode RFI suppression,.268
d k h idme dermitions of, 67 Cornmutating diodes, 30,35
W e plots, 236 Compete en- transfer flyback
Break frequency, 236,237 converter, 122, 123
Buck-boost{flyback)w e , 13, 14 Compfex variabIe, 234
Buck If&) arnvemr, 13, 15,19 Control4wutput -Wander function, 242
Euck-derid (push-pull) converter,13, Corn]-toattput voltage dc gain, 242
26,25 Conveners:
bl* WIlatOr, 4L 44
buck t f o d , 13, 15. 19
buck-host (flw, 19,14
Hdf-bridge converters, 30 Magnetk force, 110
Convertws (Conk1: Diodes (Cmtl: Magnetic Inductla 110
soft recovery, L36 Hiccup mode in current limit chmiC8,
buck-derived (push*puZU,13,16,25 223 M a p t h t I o n c u m t , 20,21
b k 34 W e n t W n g in, 138,139 Magne&aion, 131
wry f s t remery, 135-138 Holdover or hold-up time, 3, 8
m e n t d e regulate% 57 Hysteresis, 111 Metaldde vexistor W W , 11
whack (buck-bt), 13314 wraparotmd 70 Moisture r & s w , 277
complets and !ncomplete energy IEC safety stan-, 5,272 Molypermalloy W P 1 WE,157
trader, U2, 123 EfFecdve magnetic path length, 82,112 MOSFET:
Efficiency, 1-3 Incomplete e- trader n y b d
f d (buck); 11,19 con-, 122, 123 constant cumnt r s g h 93
full-bridge, 36 of l i n w power suppb, 2 constant resistance d m , 93,94
of BIvitchIng power supply, 3 Inductan=, 153,155
half-bridge, 30 Inductor: driving from CMOS, 99
push-pull (buckderired), 13,16,25 Electromagnetism, principle of, 109 from linar &its, 1m
flyba& la, 122,123
resonant, 51 EMI (electromagnetic interference), 265 magnetic amplifier, 170 from t r d m e r s , 190
ringing choke, 41 Equivalent series resistance (ESR), 179 from TTL, 9.8
output, 152
Sheppard-Taylor, 45 Error amplff~wcompensation, 243,254 m n a n t , 54 gate M v e-hc of, 91
Ward, 59 ESR (equivalent series resistance), 179 Inrush current, 9 on-res-ce 06 94
Core affective area, 116 Insulation, 275 pinchoff of, 94
Core geometries, 119 Faraday shieid, 122 Insulation redstance, 275 d m p m t i n g area (SON of, 95
Core materid, 116, 117 Fast recovsry diodes, 135-138 *tic operating c h ~ s t i c of, s 93
Interleaved windings, 121
Core saturation, 112,119 FCC spedfleatiom, 265,267 as aynchmmou9 d e r , 146
Inverse L a p b transform, 238
Corner frequency,236 Feedback theory, 237 Isolation techniqusr of switching kmmnductance of, 94,95
Creepage distanc~,23,274,277,280 Ferrite core, 116, 117 regulators, 183 MOV (mecaZsto:de varistor), 13
Crowbar, 223 Ferromagnetic m a w , 109 Isohtion v o i w , 5, 130
(CanJim Standslrds AgsOCiari~n), Filters: Negative tempwature d c i e n t MTC)
271 ae input line for RFI sum8iM1, JFET ljunction fleld-effecthnslstnrl, 88 thermistors, 10
Curie temperature, 130 268 Junction avalanching of blpoh transis- Noise,4,268
C u m t limit circuits, 217 input, 8 tor, 86 Noise filters, 268,289
for desigms utilizing bass drivers, 2 19 output, 133,152, 178 Noise suppression, 268
for direct drive designs, 217 Flux, 109 K factor, 254 Notch period, 152 I
hiccup mode in, 223 Flux density, 28,110 m a t h m a t i d expression of, 254
for univerd desiw, 222 Wux linkages, 109 synthesis of feedback mpufiers using, Off-the-line poww supply, 3, 4,7
Current-modecontml circuits, 200 Flux w d b g , 29 255 Omredstance of MOSF3T, 94
I
Current-mde regulated mwrters,67 Flyback (buck-boost) 19,14 Open-loop gain, 239

Current transformer, 221,222 complete and incompkb energy Laplace operator, 234 Operational amplifier% feedback or mar,
transfer, 122, 123 Zaplace transform, 233,234 241,243
Dielectric test withstand, 272 Fiyback transformer design, 18 Leakage current measurements, 275 K f&m, 254
Differentid-mdeR E suppression, 268 Fwheel diode, 13,20 Leakage inductance, 24,28,121 pola of, 244,254
Diode reverse blocking voltage, 134 Forward (buck)m e , 13,16,19 Length of magnetic pa* transfer function, 244
for flyback converter, 134 F o d voltage drop, 135, US of air gap, 82,113 zeros of, 244,254
for f ~ m r a dconverter, 134 Frequency d d e , 236,237 of m a t e d , 81,82, 113 Clptwwpler (optohahtar), 209
for full-bridge wnvetter, 134, 135 Frequency octave, 236,237 Line regulation, 2 .
Wit designs of, 212,214
for half-bridge convener, 134,136 k i l r b d g e con-~,36 LDad regulatioa, 2 coupling efficiency of, 211.
forprrstt.pull convmw 1-34335 h o p gain, 239,241,244 dc current d o of, 211
Diodes: Gain: Loop stabiliw rmmmmeri&, 260 t h m a l atability of, 212
abrupt recovery, 136, 137 of bipolar wanslstor, 66 Output diode3 qmwer recmersl, 135
an:isaturation, 70, 82 of buck-boost Eon-r, 242 Magnet wire specincation%wl%118 fomd voltage drop oE, 135, 9 8
clamping, 19, 24, 41, 52
rcmmutating, 30, 35
loop, 239, 241, 244
of modulator, 241
Magnetic amplifiem 162 - - peak mmnt capahUty of, 141
Schottky, 138
control circuits fm, 170
fast ~ r y 135-138 , of MOSFET, 91 cores for, 168 W e n t 0~e~vg1Cage m ~ n of,
flywheel, 13, 20 open-loop, 239 operation or, 162 141
free-wheeling 13 of operathnd ampliller, 241, 264 saturable raactPr design foF, 164 Output filter capadtor, 178
output, 135, 138, 141 Gain bandwidth product, 177 Magnetic fleld strength, 110 Output power inductor, 152
reverse refoveIy time of, 135, 136 Gain mar@&236,237 Output tFansfent response, 179.261
Magnetic flux densitg; 110
.-
.--. ... Schottky, -. .
138www.IranSwitching.ir GTO [gate mmsff) swlteh, 103 -
.-. . . . -
Index 287 '
W l o a d or w t limit pfltaetlolq 217 Beverse rerwery tlms of diode 135, 136 Transformer Mgn: Unlnterruptable power supply (UPS),
Ooerrtoltagepmtection (OW) R m m ev o w ming of outgut d e r lor flyback, 114-116.l22 228,
223 for bridge canmn, 134,133 for halt bridge, 114-117 Unity gain c r m ~239,241,244
~ ~ ,
for flyhck mmerms, 134 ' l b n d ms a f
estandads: Unity gain factor (UGF),257,258
Pmk in- voltage 0 . 8 . 1 3 4 , WS, for f m - w d om*, 134 creepage and clearance, 279 UPS ludntemghble power mpply), 228
138 for &-pull c ~ n ~ m ?l35 r, dielectric strength. 276 Q
,-&P If 1, 1 3 1l3 RPT (radio frequency i n t e r f m ~265

, imlation, 275- Varnish impregnation of ftansforrners,
Pbae kg, 241 Ring@ choke oonvertw, 41 moisture resistmw 277 131
P m lead, 241 Ripple, 2,159, 180 hperature mtingg, 277, 280 M E 6s-:
P h magin, 239 Transient input voltage protection, 11 EMI-RFI, 265-270
PIV (peak fnwm voltage), 8,136, 135, Safe o m area (SOAI,83,95 Transient ringing in diodes, 138, 139 d s t p , 271,281
138 S&o& did=,138 Transistors: Very fast recovery diodes, 135-238
Pole b d h t ~ w s c g236
, Semndap breakdown: antisaturation circuits, 68, 82 Voltage:
Poles: fmard-bias, 83 base drke circuits, 71, 82
reverse-bias, 84 m e t r i c a l interlemnce, 269
of a functioa,236,237,244 b i p h , 86 revem blocking, diode (see Diode
at the origin, 244 , k m d a q e x m low roltggs (SELV) (Seeaha Biplar t r a n s h r )
Power induaor design 132 chlits, 272 meme bloddng voltage)
Darlington, 70 m-cal interference, 269
Power d e m SelF-bias circuit, 211 delay time,68 Volhge doubler, 7
1as and w w fast -I 135 Separarion: fall time, 68
h g h air,272 V o h refemcss, 214,215
s w , 1% @q 66
syncbnmous, 144 mer- 272 hot 8pbt8,83
215~ ~ Seriw wupling q z i t m , 31
F r f n t c d ~ t i b m a b i U Ward wme&rI 59
junction fieldmeffect, 84
P m p d d base d r i m 7 3 2 Seriw I-EmantM t , 33 linear mode, 66,67 Wiading area of core or bobbin. 119
PubwiZth-modulatedtPFFM) m u d Sbepa~+Taylw converter,45 W h me,118
MOSFET, 90, 102
M t s : S w b w 66,138 W;re9, current denaity of, lllj, 118
rise time,68
m m m d monoli'chi~ IC, 187 SOA W e operating arw. 83,95 saturation mode, 66, a7
dimete qrmponent 185 Soft racwery dide, 136 secondary breakdown, 83
in- 187 ,bR s u m &&is, 216,217 storage itime, 68
Push-pull (6uCkderhdmn-, 13, Stabibitpa d p k thermal runaway, 84
16,25 We:jdors, 236 Wac, 10
ermr ampl5er arrnpsnaioa, 243
OVP drmlt, 223
Radio m e p i~m
WI),265 f d d x themy, 237 UGF gain factor), 257,258 break frequencg, 237
RBSO.4 (-bias d e operaring g Tacm, 234 UL safety standanla, 271,281 Zwcw of a function, 236,237,244
area), 86 Sgmnrdcal interferwce l-ol-, 269
RC wubbers: Spchmnuus mxiileps:
for output mMem 138 bipIar BEYX),I17
for mnsismr, eS mnt&, 150
RmSers: porrer MOXET, 145
m,135
fa5t
input, 7 Twtiap winding, 13,i4
a*, 138 Thermal rmaway, &
144, 1&, 147,150 ThermiFwrJ,10
verp fast reowetg,133,136 Time consranr,68
Mare$mtw Mu- 33 Rarwwmducam, 91
Relaw perm*, 19,113 W e r funcdons. 234
Residual m a g ~ d flu&
c 112 closed-loop, 23:
Rewnant amwtm, 51 mml-twutput, 242
~~t M e n c q , 33 ~ o m e r s :
h n a n t inductor d&n, 54 basic iheorp, 114
Revme bas? drire, 71 cnre marerial, 116
k v e m b ' i safe area inmb5oa 509
(RBSOd, 86 pm&d wnsidemtiorrs, 121,130
www.IranSwitching.ir wmkh i m p m a n of, 131
- -- -
-bh3Sm&qbR&dm&&
About the Author
Georg. C. Chryssia is founder and President of Tntelco Cmporation, a
mufa- of fiberoptic elechonic test equipent.
Previously, he was m-founder and Vice President of Engineering of Power
General Corporation, a successfulpower supply mmpany. There he was
responsible for a l l power supply design and development
The author of numerous technical d d e s , he received B.S.E.E. and
M.S.E.E. degrees from Northeastern Universiq.

George C. Chryssis High Frequency Switching Powe WWW - Iranswitching.ir

Uploaded by

Copyright:

Available Formats

George C. Chryssis High Frequency Switching Powe WWW - Iranswitching.ir

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

George C. Chryssis High Frequency Switching Powe WWW - Iranswitching.ir

Uploaded by

Copyright:

Available Formats

McGraw-Hit1 Reference Books of Interest Gf ORGE CHRYSSIS /

DICTIONARY OF S c r w m c AND TECHNICAL TERMS

Xih-frequency swivAhg power supplies.

Information mmained in rbis work has been obtained by W r a p

When ordering this title me ISBB 0.03-1007 13-X

2-0 The Voltage Dwbl# Tmnlque, 7

The rapid advancement of microelectronjcs in xecen t years has demanded

Of course, there are certain disadvantages associated with the switching

I 2-0 THE VOLTAGE DOUBLER TECHHIWE

ft 2-1 COMPONENT SELECTION AND DESIGN CRITERIA

8 4 DEFINmOMS AND DIMEWKINW

The switching transistor used In the flyback converter must be chosen to

where I p is the primary transformerchoke peak current, n is the primary- "c.

where n primary-bsewndary turns ratio

Equation 3-13 may be rewritten as

Care must be taken when designing the-transformer of the forward converter

Volume = P&IIB,,L (3-18)

balance, driving the tiamformer h t o saturation, which In turn produces high

3-2-3 ~atihonsof the Basle Forward Converter

where n is the primary-to-secondary turns ratio.

Assuming that I, a ILln, then

k way of decreasing transistor t u r n 4 t h e is to use W e r clamp diodes

%4-2 The Serlso Couplhg Capmcbr

where jn = resonant frequency, kHz t.

C = coupling capacitance, &F . :I

where N,IN, is the transformer primary-to-secondary turns ratio, and L is

In order for the chargmg of the cwpling q a c i t w to be linear, the

where I average primary current, A

The commutating diodes must be fast-recovery types with a blocking voltage

3-5 THE FUU.BRID6t CIRCUIT

current at turn-on, as compared to the push-pull clrcuit This constraint is

FIGURE 3-15 R d k a H o n of the bask Eulr ooaverter and B mdinput and

are v e v cloi~g&+$ue dc quantities having negligible switching ripple. But

reversa1 at start-up. Although the reverse poIarity pulse is of short duration:

3-7 THE BLOCKING OSCILLATOR OR RINGING CHOKE CONVERTER

At the same time, flux is building up on winding N3 developing a voltage

Since In a transistor the collector current Is a function of the base current

,V f (leakage spikes) (3-40)

Since this topology is basically a flyback converter, the transformer-choke

At initial power-up transistor Q1receives a certain amount of base current L -

A practical way to alculate the secondary windings number of turns is

Since we chose t = tg, then the secondary wlnding(s) number of turns

durfng dcircuit implementation and testing. 3-8 THE WEWARD-TAYLOR OONVERTER

1," -'t (Vc - Vm,)

318-2 Fnatures of the Shmppard-Taylor ConvmMr

High-frequency resonant converters, whether parallel or series, have be-

their tolerance to swftching transistor transition times and reverse recovery

3-9-1 The Baslc Slne Wave Wries R m m t Converter

In order to prevent inductor core saturation, an air gap must be introduced

3-9-4 Design of the Series Resonant Ne-k SC,

where t~ is the eficiency of the: converter. EXAMPLE 3-3 I

where f is the frequency of operation of the converter.

3-9-5 Design of me Resonant Inductor

3-40 CURREM-MODE REGUIATED CONVERlERS

1. Automatic feed-fornard improves line regulation.

1. Loop instability above 50 percent duty cycle.

frequencies up to 500 kHz. Chapter 7 describes the UC1846 current-mode

3-44-1 CErwN Analysis of the Ward Conve~er

The parameter I typically represents the number of cycles of operation of