This co mprehensive and insig htful book sets ou t in detail how to design gigahertz-speed radio-frequency integrated circuits in C MOS tec hnology.

Slafti ng ,,"'ilh d hi!olury of ,ihliu to c~wblbb OJ fou lk.lOJtiun

THE DESIGN OF CMOSRADIO -FREQUENCY

INTEGRATED CIRCUITS

THOMAS H. LEESial/ford University

U C AMBRIDGEV UNIVERS ITY PR ESS

PU BLIS HED BY TH E PR ESS SYNDICATE OF TilE UNIVERS ITY OF CAMBRIDGEThe Pitt Building, Trump ington Street. Cambridge, United Kingdom

CA MBRIDG E UN IVERSITY PRESSThe Edinbu rgh Building. Cambridge C B2 2RU. UK

40 Wesl 20th Street , Ne w York . N Y 100 11-421 1. USA10 Stamford Road. Oak leig h, VIC 3 166, Australia

Ruiz de Alarcon 13. 28 0 14 Madrid, SpainDock House. Th e Wate rfro nt. Cape Town 800 1, South Afri ca

http://www.cambridge.org

o Cambridge Universi ty Pre ss 1998

This boo k is in copyrig ht. Subjcct lo statutory exception andto the provisions of rele vant co llect ive licen sing agree ments.

no reprcdccncn of any pan may take place wi thou tthe written permission of Cam bridge Un iversity Press.

First pu blished 1998Reprinted with co rrec tion s 1998

Reprin ted 1999,2000,2001

Printed in the: Unued Stares of America

Typeset in Times using AMSTEX

Library of Congress Cataloging-in-Publication DtllaLee, Thomas H. 1959-

The de sign of C MOS radio-freque"')' inleg raled circu its I Thomas U. Lee.p. em.

ISBN 0-521-6306 1-4 ( hb) - ISBN 0-52163922-0 l pb)I. Metal ox ide semiconductors , Complementary - Design and

conwccuon . 2. Integrated circuits - Design and co nstruction .J . Radio - Traesrmuer-recetvers. 1. Tille .

TK7871.99.M44L44 1998621.39'732 - !X21 97-34 158

CIP

A catalog rrcord for this book is m'aiiab/e from the Brit ish Library

ISBN 0 521 6306 14 hardbackISBN 0521639220 paperback

To my parents. who had no idea what theywere starting when they bought me a pairof walkie-talkies for my sixth birthday

CONTENTS

Pref ace pag" XIII

A NONLINEAR HISTORY OF RADIO1. Introduction2. Maxwell and Hertz3. Pre-Vacuum Tube Electronics 24. Hirth of the Vacuum Tube 95. Armstrong and the Regenerative Amplifer/ Detecror/Oscillaror 136. Other Radio Circuits 157. Armstrong and the Supe rregenerator 188. Olcg Losev and the First So lid -Slate Amplifi er 209. Epilog 21

10. A ppendix: A Vacuum Tube Primer 22

2 CHARACTERISTICS OF PASSIVE IC COMPON ENTS ) 4I. Introdu ction J 42. Resistors 343. Capacitors 374. Inductors 475. Summary 576. Appendix: Summary of Ca paci tance Equat ions 57

Problem Set 58

3 A REVIEW OF MOS DEVICE PHYSICSI. lntrodur.-tion2. A lill ie History3. FETs: The Shon Sto ry4. MOSF ET Physics: The Long-Channel Approxi mation5. Operatio n in Weak Inversion (Subthreshold)

vii

6262626)64n

THE LIBRARYUNIVE RSITY OF WEST FLORIDA

\/iii CONTE NTS

6 . MOS Device Physics in the Short -Channe l Regime 737. Other Effects 778. Tran sit Tim e Effects 7~9. Summary XO

10. Appendix: O.5,Hl 1Level-S SP ICE Models XOProblem Sc i XI

A PASSIVE RIC NETWORKS X6I. lr ur oducticn X62. Parallel RLC Tank 863. Se ries RLC Networks ~ I4. Ot her Resonant RLC Networ ks ~ I5. Rt.C Networks as Impedance Transformers ~36. Exam ples 104

Problem Sci 106

5 DISTRIBUTED SYSTEMS 114I. Introd uction 11 42. Link Between Lumped and Distributed Regimes 11 63. Driving-Point Impedance of Iterated St ructures 1174 . Tra nsmission Lines in More Deta il 11 95. Behavior uf Finite-Len gth Tra nsmiss ion Lines 1246. Summary of Transmission-Line Equations 1277. Artificia l Lines 1278. Summary 131

Problem Sci 13 1

6 THE SMITH CHART AND S-PARAMETERS 134I . Introd uction 1342. The S mith C hart 1343. S-Paralllclers I3X4 . Appen dix : A Sho rt Note on Units 1405. Appe ndix: Why 50 (or 75) Q? 141

Problem Set 144

7 BANDWIDTH ESTIMATION TECHN IQUES 146I . Introduction 1462. The Method of Open-Circuit T ime Constants 1473. The Method of Short -Circuit Time Constant s 1614. Further Read ing 1665. Risctime. Delay. and Bandwidth 1676. Summary 174

Prob lem Sc i 174

CO NTENTS

8 HIGH fREQUENCY AMPLIfiER DESIGNI. Intr oductio n2. Zeros as Bandwidth Enha ncers3. The Shunt-Series Amplifie r4. Band width Enha nce ment with Ir Dou blers5. TImcd Am plifiers6. Neutrali zation and Unil areralizarion7. Cascadc-d Amplifiers8. Su mmary

Problem Set

9 VOLTAGE REFERENCES AND BIASINGI . Introd uction2. Review of Diod e Beh avior3. Diodes and Bipol ar Tran sistors in C MOS Technology4. Supply-Independent Bias C ircuits5 . Bandgap Voltage Reference6. Co nstant-g; Bias7. Summary

Problem Sc i

10 NOISEl . Introductio n2. Thermal No ise.1 . Sho t No ise4. Flicker No ise5. Popcorn Noise6. Classica l Two -Port Noise Theory7. Examples of Noise Calculationsl'I . A Hand y Rule of Thumb9. Typical Noise Performance

10. Appendix: Noise ModelsProblem Set

II LNA DESIGNJ. Introd uct ion2. Der ivation of MOSF ET Two -Port Noi se Parameters3. LNA Topo logies: Power Match versus Noi se Milich4. Power-Con strained Noise O ptimization5. Design Exa mples6. Linearity and Large-Signal Performance7. Spurious-f ree Dynamic Range

;,

17M1781791911971992032062162 17

22322322.1225225227235237237

24324324.12502522552562602632M265266

2722722732772842MM295302

CO NTENTS

I~

8. SummaryProblem Sci

12 MI XERSI. Introd uction2. Mixer Fundamentals3. Nonlinear Systems as Linear Mixers4 . Multi plier -Based Mixers5. Subsampling Mixers6. Appendi x: Diode -Ring Mixers

Problem SCi13 RF POW ER AMPLIFIERS

I. Introd uction2. General Conside rations3. Class A. AB. B. and C Power Amplifi ers-I. C1u!Os [) Amplifiers5. Class E Ampli fiers6. Class F Amp lifiers7. Modulation of Power Amplifiers8. Summary of PA Characteristics9. RF I'A Design Examples

10. Additional Design Considerations11. Design SummaryProblem Set

14 FEE DBACK SYSTEM SI. Introd uction2. A Brief History of Modem Feedback3. A Puzzle4 . Desensitivity of Negative Feedback Systems5. Stabili ty of Feedback Systems6. Gain and Phase Margin as Stability Measures7. Root-Locus Techniq ues8. Summary of Stability Criteria9 . Modeling Feedback Systems

10 . Errors in Feedb ack SystemsI I. Frequency- and Time-Domain Cha racteris tics of First- and

Second-Orde r Systems12. Useful Rules of Thumb13. Root-Locus Examples and Compensation14. Summary of Root -Loc us Techniques15. Co mpensation

303304308308309314318335337340

J.I4J.I4344345355357359362364365372379380

385385385390391395396398404404408

4 104 1441 5422423

CO NTENTS

16. Compe nsa tion throu gh Gain Red uction17. Lag Co mpe nsa tion18. Lead Compensation19. Summary of Compe nsalionProbl em Set

15 PHA SE-LOCKED LOOPSI . Introducti on2. A Short History o f I'LL~3. Lineari zed PLL Models-I. Some No ise Propertie s of PLL s5. Phase Detectors6. Sequen tial Phase Detectors7. Loop Filters and Charge PumpsR. I'LL Design Examples9. Summary

Proh lem Ser

16 OSCILLATOR S AND SYNTHES IZE RSI. Introdu ction2. The Problem with Purely Linear Oscillators3. Describing Functions-I. Resonators5. A Catalog of Tuned Oscill ators6. Nega tive Resistance Osci llators7. Frequency Sy nthesis8. Su mmary

Problem Set

17 PHASE NOI SEI . Introd uction2. General Co nsidera tions3. Detai led Considerations: Phase No ise4. The Haj imiri Model : A Time-Varying Phase No ise Theory5. Summary

Pro hlem Set

18 ARCHITE CTURE SI. Introduction2. Dynamic Range3. Subsa mpling-I. Tnansmi tter Archi tectures5. Oscill ator Stabi lity

, i

423426-l29-l32-132

43843H43H44 144745045546 3471147H47H

4844H44844855(X)5().j5""51-1524525

531153115305325365445-15

5511550551565566567

xii CONTE NTS

6. SummaryProblem Sci

19 RF CIRCUITS THROUGH THE AGESI. Intr oduction2. Armstrong3. The "All-American" 5-Tube Superhet4. The Regency TR -I Transistor Radio5. Three -Transistor Toy CB Walkie -Talkie

568568

57\57157\575578580

PREFACE

The field of radio frequency (RF) circuit design is currently enjoying a renaissance.driven in particular by the recent . and largely unanticipated. explosive growth in wire-less telecommunications, Because this resurgence of interest in Rf caught industryand academ ia by surprise. there has been a mad scramble to educate a new genera-tion of RF engineers. However. in trying to synthesize the two traditions of "conven-tional" RF and lower-frequency IC design. one encounters a problem: "Traditional"RF engineers and analog IC designers often find communication with each other dif-ficult because of their diverse backgrounds and the d ifferences in the media in whichthey realize their circuits. Radio-frequency IC design . particularly in C~fOS. is adifferent activity altogether from discrete RF design. This hook is intended as botha link to the past and a pointer to the future.

The contents of this book derive from a set of notes used to teach a one -term nd-vanccd graduate course on RF IC design at Stan ford University. The course wasa follow-up to a low-frequency ana log Ie design class. and this hook therefore as-sumes that the reader is intimately famil iar with that subjec t. described in standardtexts such as Al/aly.\I \' and lk.\';RII /d Analog integrated Circu its by P. R. Gray andR. G. Meyer (Wiley. 19 lJ) ). Some review materia l is provided. so that the practicin gengineer with a few neurons surviving from undergraduate education will be able todive in without too much disorientation.

The amount (If materia l here is significantly beyond what students can comfort-ubly assimilat e in one quarter or semester. and instructors arc invited to pick andchoose topics to suit their tastes. the length of the academic term. and the backgroundlevel of the students, In the cha pter descript ions that follow are included some hintsabout what chapters may he comfortably omit ted or deferred .

Chapter I presents an erratic history of rad io. This material is presented largelyfor cultural reasons. The author recognizes that not everyone finds hislory interest-ing. so the impatie nt reader is invited to skip ahead to the more technical chapters.

Chapter 2 surveys the passive components norm ally available in standard C~10Sprocesses. There is a focus on inductors bec ause of their prominent role in kF

...iii

PREFACE

circuits. and also because material o n this subject i.... scattered in the currentliterature(al though. happily, this situation is rapidly changing).

Chapter 3 provides a quick review of MOS device physics and modeli ng. Sincedeep submicron technology is nowcommonplace, there isa focus on approximate ann-lyrical models that account for short-channel effects. Thischapter is necessarily brief.and is intended only as a supplement 10 more detailed treatments available elsewhere.

Chapter 4 examines the properties of lumped. passive RLC networks. For ad-vanced students. this chapter may he a review ami may be skipped if desired . In theauthor's experience . most undergraduate curricula essentially abandoned the teach-ing (I f inductors long ago. so this chapter spends a fair amount of time examining theissues of resonance. Q. and impedance matching.

Chapter 5 extends into the dist ributed realm many of the concepts introduced inthe context of lumped networks. Transmission lines are introduced in a somewhatunusual way. with the treatment a..oiding altogether the derivat ion of the telegra-pher 's equation with its attendant wave solutions. The charac teristic impedance andpropagation constant of a uniform line are derived entirely from simple extensionsof lumped ideas. Although distributed networks play hut a minor role in the currentgeneration of silicon IC technology. that slate of affairs will be temporary. given thatdevice speeds are doubling about every three years.

Chapter 6 provides an important bridge between the traditional "microwave plumb-er 's" mind-set and the IC designer 's world view by presenting a simple derivationof the Smith chart, explaining what Svparameters are and why they are useful. Eventhough the rypicul IC engineer will almo st certainly not design circuits using thesetools. much instrumentation present s data in Smith-chart and Scparamctcr form. somodern engineers still need to be conversant with them.

Chapter 7 presents numerous simple methods for estimating the bandwidth of high-order systems from a series of first-order calculations or from simple measurements.The former set of techniq ues. called the method of open-circuit (or zero -value) timeconstants. allows one to identify bandwidth-limiting parts of a circuit while provid-ing a typically conservative bandwidth estimate. Relationships among bandwidth .delay. ami risetimc allow us to identify important degrees of freedo m in trading offvarious parameters. In particular. gain and bandwidth are shown not to trade off withone another in any fundamental way. contrary to the beliefs of many (if not most)engineers. Rather. gain ami delay are shown to be more tightly coupled. openingsignificant loopholes that point the way 10 amplifier architectures which effect thaitradeo ff and leave bandwidth largely untouched.

Chapter 8 take!"> a detailed look at the problem of designing extremely high-frequency amplifiers. both broad- and narrowband. with many " tricks" evolving froma purposeful violation of the assumptions underlying the method of open-ci rcuit timeconstants.

Chapter 9 surveys a number of biasing methods. Although intended mainly asa review, the problems of implementing good references in standard C MOS are

., .PR EFACE

large enough to risk so me repetition. In part icu lar. the design of CM OS -co mpatiblebandgap voltage references and consrant-transcon ductunce bias circuits are empha -sized here. perhaps a little more so Ihan in most standard analog texis.

Chapter 10 stud ies the all-important issue of noise . Simply obtaining sufficie ntgain over some acceptable bandwidth is frequently insuff icient . In many wirelessapplications, the rece ived signal amplitude is in the microvolt range . The need 10amplify such minute signals as no iselessly as possible is selt-evident. and this chap-ter provides the nec essary founda tion foridentifyi ng conditions for achieving the bestpossible noise performance from a given technology,

Chapter II follows up on the previou s two or three chapters to identify low-noiseamplifier (LNA) architectu res and the specific cond itions that lead 10 the best possi-ble noise performance . given an explicit constraint on power consumption. Thi spower-con strained approac h differs considerably from standard discrete -orientedmethod s. and explo its the freedom enjoyed by Ie designers to tailor device sizesto achieve a particu lar optimum. The important issue of dy nam ic range is also ex-amined. and a simple analytical method for estimating a large-signal linearity limitis presented.

Chapter 12 introduces the first intentionally nonlinear element. and the heart of allmodem transceivers : the mixer. After identifying key mixer performance parame-ters. numerous mixer topologies are examined. As with the LNA. the issue of dy-namic range is kept in focu s the entire time.

Chapter 13 presents numero us topo logies for bui lding RF power amplifiers. Theserious and often unsatisfactory tradeo ff's among gain. efficicncy. Iincarity, and outputpower lead to a family of topo logies. each with its particular domain of application.The chapter closes with an exa mination of load-pull experimental characterizationsof real power ampli fiers.

Chapter 14 provides a review of cla ssical feedback concepts. mainly in preparationfor the following chapter on phase -locked loops. Readers with a so lid backgroun din feedback may wish to skim it . or even skip it entirely.

Chapter 15 surveys a number of phase-locked loop circuits after presenting basicoperating theory of both first- and second-order loo ps. Loo p stability is examin ed indetail. and ,I simple criterion fo r assessing a PLL's sensitivity to power supply andsubstrate noise is offered.

Chapter 16 exam ines in detail the issue o f oscillators and frequency synthes izers.Both relaxation and tuned o..cillators are considered. with the Jailer category furthersubdivided into LC and crystal-contro lled oscillators. Both fixed and cor urollsble os-cillator s are presented . Predict ion of oscillation amplitude. criteria for start-up. anddevice sizing are all studied .

Chapter 17 extends to oscillators the earlier work on noise. After elucidating somegeneral criteria for optimizing the no ise performance of oscillators. a powerfu l theoryof phase noise based on a linear. time -vary ing mood is presented. The model makessome surprisingly optimistic (and experimentally verified ) predictions abou t what

PR EfACE

one may do to reduce the phase noise o f osc illators buill with such infamously noisydevices as MO SFETs,

Chapter 18 ties all the previous chapters together and surveys arc hitectures of re-ce ivers and transmit ters. Rules are derived for computing the intercept and noisefigure of a ca...cadc of subsys tems. Trad itiunal superheterody ne architect ures are ex-umincd. along with low-IF image-reject and direct-conversion receivers. The relativemerits and disadvantages of each of these is studied in detail.

Finally, Chapter 19 closes the boo k the way it began: with some history. A non-uniform sampling of c lassical (and di stinctly non-CM OS ) RF circuits takes a look atArmstrong's earliest inven tions. the "All-Amcrican Five" vacuum tube table mdlo.the first transistor radio. and the first toy walkie-talkie. As with the first chapter . thisone is presented purely for enjoyment. so those who do not find history lessons enjoy-able or worthwhile arc invited to close the book and revel in having made it throughthe whole thing.

A book of this length could not have bee n comple ted in the given time were it notfor the generous and competent help of colleag ues and students. My wonderful ad-ministrative assistant. Ann Guerra. magically created time by handli ng everythingwith her remarkable good cheer and efficiency, Also . the followi ng Ph .D. studentswent far beyond the call of duty in proofreadi ng the manuscript and suggesting orgenerating examples and many of the problem-set questions: Tamara Ahren s, RafaelHetancourt -Zarnoru. David Collera n, Ramin Farjad-Rad, Mar Hershenson . Joe In-gino. Adrian Ong. Hamid Raregh. Hirad Samavati. Brian Senerbcrg, Arv in Sbahani.and Kevin Yu. Ali Haj imiri , Sunderarajan S. Mohan. and Derek Shaeffer merit spe-cial mention for their co nspicuous contribut ions. Without the ir help. given in theeleven th hour. this txXIK would still be awaiting comple tion,

The author is also ex tremely grateful to the text's reviewers. both known andanonymous, who all had excellent. thoughtful sugges tions. Of the former group, Mr.Howard Swain (formerly of Hewlett -Packard ), Dr. Gilly Nusserbakfu of Texas In-strumc nts. and Professors James Roberge of the Massachusetts Institute of Technol -ogy and Kurtikeya Mayannn of Washington State University deserve speci al thanksfo r spotting typographical and gruphicnl errors, and also for their valuable edito rialsuggestions. Malt and Vickie Darne ll of Four-Hand Book Pnckuging did a fantasticjob ofcopyedui ng and type setting. Their valiant efforts to convert my "sow 's car" ofa manuscript into the proverbial silk purse were noth ing short of superhuman. AndDr. Phi lip Meyler of Cambridge University Press started this whole thing by urgingme to write this book in the first place. so he 's the one to blame .

Despite the delight taken by students in findin g mistakes in the professor 's notes.some errors have managed to slip through the sieve. even after three years of filter-ing. Sadly. this suggests that more await discovery by you. I suppose that is whalsecond edit ions are for ,

i!2

CHAPTER ONE

A NONLINEAR HISTORYOF RADIO

1.1 INT RODUCTION

Integrated circuit engineers have the luxury of lak ing for granted thai the incrementalcost of a tran sistor is essentia lly zero, and this has led ( 0 the high-device-co unt cir-cuits that are common toda y. Of course. this situation is a relatively rece nt develop-ment: during most of the history of electronics. the economics of circuit design werethe inverse of what they are today. II rea lly wasn't all thai long ago when an engineerwas forced by the relatively high cos t of ac tive devices 10 try to get blood (or atlea..1rectification) from a stone . And it is indeed remarkable j ust how much performanceradio pionee rs were ah le to squee ze out of just a handful (If components. For exa m-ple. we' Il SI."e how American radio genius Ed win Armstrong devised circuits in theearly 1920s that trade log of gain for ban dwidth . contrary 10 the conventional wis-dom that gain and bandwidth should trade o ff more or less directl y. And we ' ll SCI.'that at the same time Armstrong was developing those circuits, self-taught Soviet ra-dio engineer Oleg Loscv was experimenting with blue LEDs and constructing com-pletely solid -state radios that functio ned up to 5 MHz, a qua rter ce ntury before thetransistor was invented.

These fascinating stories are rare ly told because they tend to fall into the cracksbetween history and engineering curricula. Somebody ought to tell these stories,though, since in so doing, many commonly asked questions (" why don' t they do itthis way?") arc answered (vthcy used [0 , but it caused key body pans to fall off") .This highly non linear history of radio touches briefly on j ust so me of the main stories,and provides poin ters to the literature for those who want 10 probe further .

1.2 MAXWELL AND HERTZ

Every elec trical engineer knows at least a bit about James Clerk (pronounced "clark")Maxwell; he wrote those equat ions rhar made life extra busy back in sophomore yea r

2 CHAPTER 1 A NONLINEARHISTORY Of RADIO

or thereabouts. Nor only d id he write the electrodynamic equations! that bear hisname. he also published the first mathematical treatm ent of stability in feedback sys-tems ("On Governors: ' which explained why speed controllers for "team enginescould sometimes he unstab tc-) .

Maxwell collec ted all that was then know n about electromagnetic phenomena and .in a mysterious) ami brilli ant stroke. inven ted the displacemen t (capac itive) currentterm that allowed him to derive an equation thai led 10 the prediction of electromag-netic wave propagation .

Then came Heinrich Hertz. who was the first to verify experimentally Maxwell'sprediction thai electromagnetic waves exist and propagate with a finite velocity. His"t ransmitters" worked on this simple idea : discharge a coil across a spark gap andhook up some kind ofan antenna to launch a wavetuninrentionally) rich in harmonics.

His setup naturally provided only the must rudimentary fi ltering of this dirt y sig-nal. so it took extraordinary care and persistence to verify the existence of (and toqua ntify) the interference nulls and peaks that are the earmarks of wave phenomena.He also managed to demon strate such qu intessential wave behavior as refraction andpolarization . And you may be surprised that the fundamental frequencies he workedwith were bctween Sn and S{XJ Mi ll . He was actually forc ed to these frequencie s be-cause his laboratory was simply too small 10 enclose several wavelengths of anythinglower in frequency.

Because Hertz's sensor was another spark gap (integra l with a loop resonator).the received signal had to be large enough to induce a visible spark . Although ade-quate for verifying the validity of Maxwell' s equa tions. you can apprec iate the diffi-culties of trying to use this apparatus for wireless communication. After all. if the reoceived signal has to be strong enough to generate a visible spark. scaling up to globalproportions has rather unpleasant implications for those of us with metal dental work.

And then Hertz died - young, Enter Marconi.

1.3 PRE-VACUUM TUBE ElECTRONICS

For his radio experiments Marconi simply copied Hertz's transmitter and tinkeredlike crazy with the sole intent to usc the system for wireless communication (and not

I Actually. Oliver Hcavisidc was the one who llrst used the nota tional conventions of vec tor calcu lus10 ca'" Maxwell's equeuon.. in me form familia r 10 mo"l cl1~ il1~ totl"y.

2 Proc. Ro)'. Sf~.. 18tl11.J Many electncuy and magnensm (E&MI lexts offer the logical . bul hi"lorically ....ro ng. explana-

tion that Maxwell Inven ted lhe disp lacement current tenn after rea lizing that there was an inconsis-tency between the known taws of E&M and the continuity equation for current . The lruth is thatMaxwell was a genius. and the inspirations of a genius often have elusive or igins. This is one ofthose cases.

1.3 PRE - VACUUM TUBE ELE CTRO NI CS

[A,---O-,j__Lr------' RJFIGURE 1.1. Branly's coherer.

3

0---0Relay/Paper Tape lnker(As\umcd to have highRF impedance]

FIGURE 1. 2 . Typical rece iver with coherer.

incidentally to make a lot of money in the process). Recognizing the inhere ntlir nita-rions of Herta' s spark-gap defector. he instead used a bizarre creat ion that had beendeveloped by Edouard Branl y in 1890 . As seen in Figure 1.1. the device

-, CHAPTER I A N ONLINE AR HISTQRY O F RADIO

- --e:::::::jFIGURE 1.3 . Marconi's coherer .

As can he seen. the co hcre r acti vated a relay (for audible clicks) or paper tapeinker (fo r a perma nent record ) when a received s ignal triggered the transition to alow-resistance sta te. II is ev ide nt thai the cohe rer was ba sicall y a digita l device. andtherefore unsuita ble fo r uses other than rad iote legra phy.

Marcon i spe nt a grea t deal of time improvin g what wus inheren tly u terrible dct cc-lor and finally se ttled on the configuration shown in figu re 1.3. lie great ly reduced thespaci ng be twee n the end plugs (to a minimum of 2 nun). filled the intervening spacewith a particular mixtur e of nic kel and silver filings (i n 19 :1ratio ) ofcarefully selectedsize. and sealed the en tire assembly in a partiall y evacuated tube . As an additional re-finemcm in the receiver. a so lenoid provided an audible indication in the process o f au-tomaticully whacking the detector bac k into its ini tia l state after each recei ved pu l~.'

As yo u can imagin e. many Et\.f even ts other than the desired signa l co uld trigger acoherer. resulting in some difficult-to-read messages. Even so. Marconi was able torefin e his apparatus to the po int of achieving transatlantic wireless communica tio nsby 190 1, with much o f his success attributable to more powerful trunsrmuers andlarge. e levated antennas that used the earth as one terminal (as d id his tran smitter ).as well a~ to his imp roved coherer.

It shouldn' t surp rise you. though . thai the coherer. eve n at its be st. pe rformed quitepoorly. Frustration with the coherer's e rra tic nat ure im pelled an aggressive searchfor bett e r detec tors. Without a suitable theoreti cal framework as a guide , however,this search sometimes took macabre turns. In one case. eve n a human brain from afre sh cadaver was used as a cohercr. with the expe rime nter claiming remarkable sell-sitivity for his apparatus." Let us all be thankful that this particul ar type o f cohcrernever qui te ca ught o n.

Most research was guided by the vagu e intuiti ve not ion thai the coherer's opera-lion depended on some mysterious property of imperfect contacts, and a variety ofexperimenters stumb led. virtually sir nulraneously, on the point-cont act crys tal dcrcc-tor t f'ig urc IA), The first paten t for such a device wa.. awa rded in 190-1 (filed in 19( 1)

~ The cohcrer was most recently used in a radio -contro lled toy truck in the late 1 'J .'i O ~ ,f> A . F, Cotbns, Electr ical World and Enginrrr. \I , :\9 , 1902 : he started out with bruins of othe r spcde~

and worked his way up In humans,

1.3 PRE-VACUUM TUBE ELECTRONI CS

o

5

FIGURE 1.4 . Typical crysk:ll detector.

10 J. C. Bo'-C for a detector thai used galena (lead sulfide j.? Th is appears 10 he thefirst patent awarded for a semiconductor detector . although it was not recognized assuch ( Indeed. the word "semicond uctor" had not yet been co ined ). Work along theselines continued. and Genera l Henry Harrison Chase Dunwoody received a patent inlate 1906 for a detector using ca rborundu m (silicon carbide) . followed in early 1'X)7by a patent to Greenleaf Whini er Pickard (an MIT graduate whose great-uncle wasthe poet John Greenleaf Whinier) for a silicon (!) detector . As shown in the fi gure.one connection to thi... type of detec tor consisted of a small wire (whimsically knownas a cutwhisker} that made a point contact to the crystal surface . The other conncc -tion was a large area contact typica lly formed by a low-melting-point alloy (usuallya mixture of lead. tin, bismuth and cad mium. known ;1.'-' wood's metal . with a melt -ing temperatu re of under 80' C) that surrounded the crystal. One might cal l a devicemade in this way a po int-co ntact Schottky diode. although mea.surerncr us 'Ire not al-ways easily reco nciled with such a description, In any eve nt. we ca ll sec how themodern symbol for the d iode evolved from a depiction of this physical arrangement,with the arrow representing the ca twbiskcr point contact. us seen in the figure,

Figure 1.5 shows a simple crystal" radio made with these devices." An L C cir-cuit tunes the desired signal. which the crys tal then rectifies, leaving the demodul ated

' 1. C, Bose, U,S.I'atellt #755,840, granted 19 March 1904. Actually, Ferdinand Braun had reportedasynuncrrica l conduction ill gale na and copper pyrites (among others ) hack in 11174. in "Ueberdie Suornlcinmg durch Schwefclmctallc ("On Current Fluw through Metallic Sulfides"], /'0I:IW/l od0rff',f Antlllit'/! dc r /'h)'.~i1( und Chemie , v, 153, pp. 556- 63. The large-area contact was madethrough partial immersion in merc ury. and the other with copper, platinum, and silver wires, Noneof the !kImples ~hnwed more than a 2: I forward / reverse current ratio. Braun later shared a NnhdPrize with Marcon i for conmbu nons 10 the radio art .

, In modem electronics. "crystal" usually refers to quartz resonators used. for example, as fre quency-determini ng elements in lI

, CHAPTER. 1 A N ONLI NEAR HISTORY O F RAD IO

L

.. - Long ...in: for A M !>and

C

L: Appru x. 25(~H

C: Af'l'"'~ . JlJ-Jmrf.o IUnc AM ...OO

Headphones~ to be hlJ h.Z (> I fe... kU I

..__ l'Ottd.. good conlle

1.3 PRE-VACUUM rueeElECTRO N ICS 7

Marconi had made head lines in 1899 by contrac ting with the New York Heraldand the Evening Telegram to provide up-to -the-m inute coverage of the America 'sCup yacht race, and was so successful that two additional groups were enco uragedto try the same thing in 190) . One of these was led by Lee de Forest. whom we ' llmeet later. and the other by an unexpected interloper (who turned out to be none otherIbm Pickard ) from American Wireless Telephone and Telegraph. Unfortunately withth rr t: groups s im ult a ne o usly s park ing away lhal ye ar. " 0 o m ' was able 10 receive in-telligible signals. and race results had to be repo ned the o ld way, by semaphore. Athoroughly disgusted de Forest threw his transmitt er overboa rd, and news-starvedrelay stations on shore resorted 10 making up much of what they reponed.

This failure was all the more discouraging because Marconi. Lod ge. and that errat icgenius Nikola Tesla had actually alread y patented circuits for luning. and Marcon i' sapparatus had employed bandpa ss filters to reduce the possibility of interference.12

The prob lem was that . even though adding tuned circuits to spark transmitters andreceivers certainly helped to filter the signal. no practical amount o f filterin g couldever really convert a spark train into a sine wave. Recognizing this fundam entaltruth. a number of engineers sought ways of generating co ntinuous sine waves at ra-dio frequencies. One group. which included Dani sh engineer Valdemar Poul sen' !(who had also invented a crude magnetic recording device called the telegraphone )and Australian-American engineer (and Stanford graduate) Cyri l Elwell. used thenegative resistance associated with a glowing OC arc to keep an LC circuit in con-stant osc illation14 to prov ide a sine -wave radio -frequency ( RF) carrie r. Enginee rsquickly discovered that this approach could be scaled up to impressive power levels:an arc transmitter of ove r I lrIegllwatt was in use shortly after WWI!

Pursuing a somewhat differe nt approach , Ernst F. W. Alcxanderson of Genera lElectric (GE) acted on Reginald Fessenden's request to produce RF sine waves atlarge power le....e ls with huge alternators (realty big. high-speed versions of the thingthat charges your ca r battery as yo u d rive). Th is dead -end tec hno logy culminated inthe construction of an alterna tor that put out 200 kw at 100 kl tz! It was completedjust as WWI ended. and was already obso lete by the time it became ope rational.'?

12 Marcon i was the only one hacked by str ong financial interests (essentiullythe Briti sh governme nt).and his Bnush palenl (#7777. lhe famou s "four sevens" pate nt, granted 26 April I9

a CHAPTER. 1 A N ON LINEAR HISTORY O F RAD IO

nirnc acid -- _

FIGURE 1.6 . Feuendeo's liquid bcr retter.

The superiority of the continuous wave over spark signals was immedia tely evi-dent . and spurred the develop ment of bet ter receiving equipment. Thankfully. thecohercr was grad ually supplanted by a number of improved devices, including thesemicond uctor devices described earlier. and was well on its way to exti nction by19 10 (alt hough as late as the 1950s there was atleast thai one radio -controlled toythat used a cohcrer).

One such improvement . invented by Fessenden. was the " liquid barretter " shownin Figure 1.6. Thi s detector consisted of a th in. silver-coa led pla tinum wire (a "Wol-laston wire" ) encased in a glass rOO . A tiny bit of the wire protruded from the rodand made contact with a small pool of nitric acid. Thi s arrangement had a quasi-quadratic V- I characte ristic near the origin and therefore co uld actually demodulateRF signals. The barreuer was widely used in a number of incarnatio ns since it was a"self-res toring' device (unlike typical coherers ) and requi red no adjustments (unlikecrystal detectors) . Except for the hazards associated with the acid , the barretter wasapparently a satisfactory detector. judging from the many infringements ( includingan infamous one by de Forest) of Fessenden 's pate nt.

Enough rectifying detectors were in usc by late 1906 to allow shipboard ope ra-tors on the east coast of the United States to hear, much to their amazement (despitea forewarning by radiotelegraph three days before). the fi rst AM broadcast by f es-senden himself 011 Christmas Eve. lf> Delight edlisteners were treated to a program ofpoet ry, Fessenden's violin playing of Christmas carols. and some singing. li e used awater-coo led ca rbon microphone i ll series witli the ml1 n /1/tl to modu late a 5-kW (ap-proximately), 50-kHz (also appro ximate) carrier generated by a prototype Alexan-derson alternator loc ated at Brant Rock, Massnchusens. Those unfortun ate enough10 use cohcrers missed out on the historic event, since coherers a." typically used arecompletely unsuited to AM demodulation. Fessenden repeated his feat a week later.on New Year 's Eve. to give more people a chance to gel in on the fun .

" Auken [see Sec non 1.91erroneous ly !lives the dale as Christmas Day.

1.4 BIRTH O F THE VACUUM TUBE 9

The next year. 1907. was a significant one for electronics. Aside front Iollowingon the heels o f Ihe first AM broad cast (w hich ma rked the rrunsition from radioteleg-raphy to radiot elephony ). it saw the emerge nce of important semiconductors. In ad -dition to the patenting of the silicon detector. the LED was also discovered that year!In a brief art icle in Wi l'l'les.f UiJrltl titled "A No te on Carborundum." Il en ry J. Roundof Great Britain reported the puzzling emission of a co ld. blue l7 1ighl from carboru n-dum detectors under certai n cond itions (usually when the ca twtucker porennal wasvery negative relative to thai o f the crys tal). The e ffec t was largely ignored and ulti-mately forgott en , as there were just so man y more pressing problems in rad io at thelime. Today, however. carbo rundum is in fact used in blue LEDs,18 and has beeninvestigated by some 10 make transistors that ca n operate at elevated tem peratures.And as for silicon. wel l. we all know how that turned out .

1.4 BIRTH OF THE VACUUM TUBE

The year 1907 also saw the pa tenting, by Lee de Forest , of the first electronic devicecapable of ampli f ica tion: the triode vacuum lube . Unfortunately, de Forest didn 'tunderstand how his invention actual ly worked. ha ving stumbled upon it by way of acircuitous (and occasionally unethical) route.

The vacuum lube ac tually traces its ancestry to the lowly incandescent lig ht bu lbof Thomas Edison. Edison's bulbs had a problem wit h progress ive dark en ing ca usedby the accumu lation of '100 1 (given off by the carbon filaments) on the inner surfaceof the hulb. In an atrernpt rc cure the problem, he insert ed a metal electrode. hopin gsomehow to attract the '10 0 1 to th is pla te ra ther than to the glass. Ever the ex pcri-mentalist . he applied both positive and negati ve vo ltage s (relative to one of the fila-ment connections ) to this plat e, and noted in 1883 that a current mysteriously Ilowedwhen the plate was po sitive, but none flowed when the pla te was negative, Further-more, the current that flowed depended on how hot he made the filamen t. li e hadno theory to explain these observations (rem ember, the word "electron" wasn 't evencoined unti l IXlJ 1, and the panicle itself wasn' t unambi guously identified until J. J,Thomson' s experimen ts of IXlJ7 ), but Ediso n went ahead and paten ted in IXX4 thefi rst electronic (as opposed to electrical) devic e. one that ex ploi ted the dependence ofplate cUITCnl on filament temperatu re 10 measure line vo ltage indirectly. This RubeGoldberg instrument never mad e it into production since it was inferior to a standardvoltmeter: Edison j ust wanted another patent . that ' s all (tha t's one way he ended upwith over a thousand of the m).

17 1 1e ~ ..... IIrange and yellow, too. He may have been drinking.18 II stMluld he me ntioned lhal GaN-N.

10 CHAPTER 1 A N ONLINE AR HISTORY O F RADIO

cylindrical plate -

~f f-;-, ' \

4+ Plate conrecnon9.- Filament connecnonsFIGURE 1.7. Fleming V'CIIve.

The funny thing about this ep isode is that Edison arg uably had never invented any-thing in the fundamental sense of the term. and here he had stumbled across an elec-tron ic rectifier but nevertheless failed to recogn ize the impl ications of what he hadfound. Part of this blindness was no doubt related to his emotional (and financial)fixation on the DC tran smission of power. where a rectifi er had no role.

At about this time. a consultant to the British Edison Company named John Am-brose Fleming hap pened to attend a conferen ce in Ca nada . He drop ped down to theUnited Slate s to visi t his brother in New Jersey and also stopped by Edison's lab. Hewas greatly intri gued by the " Edison effect" (much more liO than Edison. who fou ndit difficult to underst and Fleming's exciteme nt ove r something that had no obvio uspro mise of practical app lica tion). and eventually published papers on the Edison ef-fect from 1890 to 1896 . Althou gh his expe riments cre ated an init ial stir. Ront gen'sannouncement in January 1896 of the d iscovery o f Xvrays - as well as the discov-ery o f natural radioact ivity later that same yea r - soo n do minated the interest of thephysics co mmunity. and the Edison effec t qu ickly lapsed into obscurity,

Se veral years later . though . Flemi ng bec ame a consultant to British Marconi andjoined in the search for improved detectors. Recalling the Ediso n effect . he testedsome bulbs, found out tha t they worked all righ t as RF rect ifiers. and patented theFleming valve (vacuum tube s are thus still known as valves in the Un ited Kingdom ) inIljOj (Figure 1.7). The nearly deaf Fleming used a mirror galvanometer to prov ide avisual ind ication of the rece ived signal. and included this featu re as part of his patent.

Alt hough not partic ular ly sensi tive. the Fleming valve wax at least continuallyresponsive and req uired no mechanica l adjustm ent s. various Marco ni installationsused them (largely out of contractua l obli gations) . but the Fleming valve was neverpopular (contrary 10 the asse rtio ns of some poorly resea rched histories) - it neededtoo much pow er, filament life was poor. the thing was ex pens ive, and il was a re-markab ly insens itive de tector com pared to. say. Fessenden 's barretter or well-madecry sta l detectors.

De Forest . mean whil e. was busy in America se tting up shady wireless companieswhose so le purpose was to earn money via the sale of stock . "Soo n. we believe.

1.4 BIRTH OF THE VACUUM TUBE \I

the suckers will begin to bite," he wro te in his journal in early 1l)()2. As soo n asthe stock in one wireless installation was sold, he and his cronies picked up Slakes(.... hether or nOI the station was actually completed) and moved un to the next town.Inanother demonstrat ion o f his sterling character. he outright sto le Fessenden's bar-retrer (simply reform ing the Wollaston wire into the shape of a spade ) after visitingFessenden 's laboratory, and even had the audaci ty to claim a prize for its invention.In this case, however. ju stice did prevail. and Fessenden won an infringement suitagainst de Forest.

Fortunately for de Forest. Dunwood y invented the carbo rundum detector just intime to save him fro m bankrupt cy. NOI content 10 develop this legitimate invention.19though. de Forest proceeded to steal F leming's vacuum tube d iode and actual ly re-ceived a patent for it in 1905. He simply rep laced the mirror galvanometer with aheadphone and added a huge forwa rd bias (thus reducing the sensitivity of an alreadyinsensitive detector). De Forest repe atedl y and unconvincingly denied throughouthis life that he was aware of Fleming 's prior work (even though Fleming publi shedin professional jo urnals tha t de forest habi tually and assiduously scanned ): to 001srer his claims. de Forest pointed 10 his usc of bias where Fleming had used none.P'Conclusive evide nce that de Forest had lied o utright finally came to ligh t when his-torian Gerald Tyne obtained the business record" of W. Mccandtess. the man whomade all of de Forest' s first vacuum lubes (de Forest called them "audions"). Therecords clearl y show that de Forest had asked McCandless 10 dupl icate some Flem-ing valves months before he filed his patent. There is th us no mom for a charitableinterpretation tha t de Forest independently invented the vac uum tube diode.

lI is crowning achievement ca me soon after. however. De Forest added a zigzagwire elec tr ode. which he called the grid. between the filament and wing electrode(later known as the plate}, and thus the triode was born (see Figure 1.8). This three -element audion was ca pable of amplifica tion, but de Forest d id nol reali ze this factuntil years later . In fact. hi!'> patent application described the tr iod e audion only as adetector. not as an amplitier.U Motivation for the addit ion o f the grid is thus still curi-ously unclear. He certainly d id not add the grid as the co nsequence (Ifcareful reason-ing. as some histories claim. The filet is that he added electrodes all over the place.Ile even tried "control electrodes" o utside of the plate ! We must therefo re regard hisaddition of the grid as merely the result of haphazard bUI persistent tinkering in his

I ~ Dun.....o ody had performed l h i ~ Iol.ur l as a consulta nt to de F(lrt"~t. lie .....as unsecce ssnr l in hisrtTorts to get de Fo resl to pay him for it .

:to In his efforts 10 establish that he had Iol. ur lcd independe nt ly of Fleming. de Forest repeatedly and!Mide nlly suned Ihat il was hi!> rese arcbes Into the conductivity properties of lIalfl('s that informed

hi ~ .....od in vacu um 1Uhe!>. arg uing thai ionic conduct ion wa~ the l ey 10 the ir operation. As aconsequence. he b oxed him self into a comer thaI he found difficult to escape later . afte r oebersdeve loped the superior hig h-vac uum robes that were e~ntially free of iun~ .

21 Curiou~ly enough . though, his patent fUf the: two -ele me nt audion JoC's Ifl('ntion am phficat ion .

12 CHAPTER 1 A NO NliNEAR HISTORY O F RADIO

Grid connection Plate (Mwing") con nectio n

Grid

Plate

(new)

Wing

G rid- f--/\ (o ld)

Filament

FIGU RE 1. 8 . De Forest triode c udicn and syrnbo15.

search for a detector to call his own. It would not be inaccurate to say that he srum-bled onto the triode. and it is certainly true that others had to explain its operation tohim.n

From the available evidence. neither de Forest nor anyone else thought much ofthe audion for a number of years ( 1906-I9{)t) saw essentially no activi ty on the au-dion ). In fact, when de Forest barely escaped conviction and ajail sentence for stockfraud after the collapse of one of his companies. he had to relinquish interest in allof hi!'> inventions as a condition of the subsequent reorganization of his companies.with just one exception: the lawyers let him keep the patent for the audion. thinkingit worthless.P

He intermittently puttered around with the nudion and eventually discovered itsamplifying potential. as did others almost simultaneously (including rocket pioneerRobert Goddard).24 He managed to sell the device to AT&T in 191 2 as a telephonerepeater amplifier. but initially had a tough time because of the erratic behavior ofthe audicn. Reproducibi lity of device characteristics was rather ptxlr and the tubehad a limited dynamic range. It functioned well for small signals. but behaved badlyupon overload (the residual gas in the tube would ionize. resulting in a blue glow anda frying noise in the output signal). To top things off the audion filaments (made

II Auken (see Section 1.9) argues that de Ft,m,t ha., been unfairly acc used of not unde~tanding h i ~own invennon . However. the bulk of the evidence contradicb Aitken ' ~ ge ocnJus view.

13 The rece ntly unem ployed de Forest the n wem te work fur Elwell at Federal Telephone and Tele -graph in Palo Atto.

N His U.S_Patent III,159,209 . filetl l AugUSl IIJI 2 and granted 2 November 191S.describes an audionuscillaillf and Ihu~ actually predates even Aml~tn,"g 's oocu rrenred won .

1. 5 TH E REG ENERATIV E AMPli fiE R/ DETECTOR/ OSCIL LATOR 13

B+

c

Ond-leak bias resistor Til Fil. Supply

Tbc: ~m"unl of P'" iliw fee'dbocLi, ~olll ll ll kd by 1'-' couplingbct"' ....m U . 00 L.2.

FIGURE 1.9. Armstrong regenerative receiver.

of tantalum) had a life of only about 100 - 200 hours. It would he a while before thevacuum tube could take over the wor ld .

1.5 ARMSTRONG AND THE REGENERATIVEAMP 1IFIE R/ DHE CTOR/ OSC II I ATOR

Fortunately, some gifted people fina lly became interested in the audion. Irving Lang-muir at GE Labs in Sc henectady wor ked to ach ieve high vacua. thus eliminating theerratic behavior ca used by the presence of (easily ioni zed ) res idual gases. De Forestnev er thought to do this (and in fact warned aga inst it. be liev ing that it would reducethe sensitivity ) because he never rea lly be lieved in thermionic emission of elec trons(indeed. it isn't clear he even believed in elec trons at the time ), asserting instead thattheaudion depended fundam entally on ioni zed gas for its operat ion.

Afler Lan gmui r ' s achievement , the way was paved for a bright engi neer to deviseuseful circuits to exploit the uudio n' s potent ial. That enginee r was Edw in HowardArmstrong, who inven ted the regenerative ampli fler/de tector P in 19 12 at the tenderage of 2 1. Th is ci rcuit (a modern version of which is shown in Figu re 1.9) employedp ositive feedback (via a " tickler coil" that coupled some of the output energy hackto the input with the right phase) to boost the gain and Q of the system simultane-ously. Thus high gain (for good sensitivity) and narrow ba nd width (for good selec-ti\'ity) could be obtain,..-d rather simply from one tube. Additionally, the nonlinear-uy of the tube demodulated the signal. Furthermore. ovcrcoupling the ou tput to theinput turned the thing into a won derfully compact RF oscilla tor .

~ Hisno14rized OU(ebouL: e ntry is actually da ted 31 January 191 J.

-"CHAPTE R 1 A NONLIN EA R HISTORY O F RADI O

L

In a 1914 paper entitled "Operating Feat ures of the Aud ion,',26 Arms trong pub-lished the first co rrect explanation for how the triode worked and provided experi-menial ev idence 10 suppo rt his claim s. He followed this paper with another (" So meRecent Develop ments in the Audion Receiver" )27 in which he additionally ex plainedthe operation of the regenerative amplifier/detector . and showed how 10 make an os-ci llator ou t of it . The paper is a model of clarity and quite read ab le even 10 modemaudiences. De Forest. however. was quite upset at Armstrong' s presumptuousness.In a published discussion sec tion fo llowi ng the pape r. de Forest repeatedly attac kedArmstro ng. II is clear from the published exchange that . in sharp contras t with Arrn-stron g, de Forest had difficulty with ce rta in bas ic conce pts (e.g.. that the averagevalue o f a sine wave is zero). and didn 't even understand how the triode, his own in-vent ion (more of a discovery. really) . actually worked. The bitter enmity that arosebetween these two men never waned .

Armstron g went on to deve lop circ uits that continue to dominate communicationssys tems to this day. While a mem ber of the U.S. Army Signa l Corps during WorldWar I. Arm strong became involved with the problem of detecting enemy planes froma distance. and pursued the idea of trying to home in on the signals naturally gener-ated by their ignit ion systems (spark transmitters ag ain). Un fortunate ly, little usefulradiation was found below about 1 MH z, and it was exceedingly difficult with thetubes ava ilab le at that tim e to get much amplifica tion above that freq uency. In fact.it was o nly with ex traordi nary care that II . J. Round (of blue LED fame) ach ieveduseful gain at2 Mll z in 19 17, so Armstron g had his work cut OUI for him.

li e solved the pro blem by emp loyin g a principle ori ginally used by Poul sen andlater eluc idated by Fessenden. When demodulating a co ntinuous wave

1.6 OTHER RADIO CIRCU ITS- - -----

15

.......... Demod j-. AF Amp~_"Audio0",

IF Ampx~ RF Am. ~Tuning contml--.

FIGURE 1.10 . Superheterodyne receiver block diagram.

sensitivity so that the limiting factor was actually atmospheric noise (wh ich is qui relarge in the AM broadcast band) . Furthermore. a single tuning control was madepossible. since the IF amplifier works at a fixed frequency.

He called this system the "superheterody ne" and patented it in 19 17 (see Figure1.10). Although the war ended before Armstrong could use the supe rhet to detec tGerman planes, he continued to develop it with the aid of se..'eral talented engineers,fi nally reducing the number of lubes to five from an ori ginal complement of ten (goodthing. too: the prototype had a total filament current requ irement of ten amps). DavidSarnoff of RCA eventually negotiated the purchase of the superhet rights. and RCAthereby came to dominate the radio market by 1930.

The great sensitivity enabled by the invention of the vacuum tube allowed trans-miner power reductions of orders of magni tude while simultaneously increasing use-ful communica tions distances. Today. 50 kW is considered a large amount of power.yetten times this amo unt was the norm r ight after WWI.

The 1920s saw greatly accelera ted deve lopment of radio e lectronics. The warhad spurred the refinement of vacuum tubes to an astonis hing degree. with theappearance of improved filaments (longer life, higher emis sivity, lower power re-quirements), lower ir uerclectrod e capacitances, higher transcond uctance, and greaterpower-handling capability, These develop ments set the stage fo r the invention ofmany clever circuits, some designed to challenge the dominan ce of Armstrong's re-generative receiver.

1.6 OTHER RAD IO CIRCUITS

1.6 .1 THE TRF AN D THE NEUTRODYN E

One wildly popular type (If rad io in the early days was the tuned radio -frequency(TRF) rece iver. The basic TR F circuit typica lly had three RF bandpass stages. eac htuned separately, and then a stage or two of audio after demodulation (the latter some-limes accomplished with a cry stal diode). The user thus had to adj ust three or more

-16 CHAPTER 1 A NONLIN EAR HISTO RY O F RA DIO

eN

Adjus t eq ual 10 en>

B+ I~ /t~To next stageO utput bandpass filler

From previous stage

Input bandpass filter R I.sur r 1ylbia., control

FIGURE 1.11. Basic nevtrodyne amplifier-.

knobs to tun e in each station, Whil e this array of controls may have appealed to thetinkering-di sposed tecbnophile. it was rather unsuit ed to the average consumer.

Oscillation of the T RF stages was also a big problem. ca used by the pa rasitic feed-back path provided by the grid- pla te capaci tance C"p.2r. Althou gh limiting the gainpe r stage wa.. one way to red uce the tend ency to oscillate. the attendant degrad ationin sensit ivity was usuall y unaccept able.

T he problem ca used by eftr was largely eliminated by Harold Wheeler' s inven-tionl'J of the ncutrod yne circuit (see Figu re 1. 11 ) .311 Recognizing the cause of theprob lem. he inserted a compe nsating ca pacitance (C N ), termed the neu tralizing ca -pacitor (actua lly, "con denser" was the term back then ). When properly adjusted . thecondenser fed back a current exactly equal in magnitude but oppo site in phase withthat of the plate -to -grid capacitance, so that no inpu t current wa s required to chargethe capacitances. The net resu lt was the suppression of C HI , ' S effec ts, permit ting alarge increase in gain per stage without oscillat ion." Afte r the war. Westin ghou seacquired the righ ts to Armstro ng 's rege neration patent , negotiated licensing agree-ment s with a limited number of radio manufacturers. and then aggressively prose-cuted those who infringed (wh ich was just abo ut eve rybody) . To pro tect them selves,

2MIt is len as "an exercise for the reader" to show that the real part o f the input impedance nf an in-o.IUl: livc1y loadco.l l:u Ill11l01J-l:a lhode a mpli fie r can I;...., le ~~ tha ll / l:rt. hc~' au,>

1.6 O THER RADIO CIRCUITS

Ar- .. AuJio OUI .

fiGURE 1.12 . Reflex receiver block diagram.

17

those "o n the outsid e" organi zed into the Independ ent RadioManufacturers Associ -ation and bou ght the rights to Hazelt ine's circuit . Tens of tho usands of ncut rodynekits and assembled conso les were sold in the 1920 s by members of IRMA, all in ananempr to compete with Armstrong's regenerat ive circuit .

Meanwh ile, de Forest was up to his old tricks . He bou ght a compa ny that had a li-cense to make Armstro ng 's regenenuivc circui ts . Although he knew that the licen sewas nontransferab le, he nonetheless started to se ll regenerative radios until he wa..caught and threatened with lawsuits. He eve ntually skirted the law by selling a ra-dio that m ulti be hooked up as a rege nerator by the cu stom er sim ply by reconnect -ing a few wi res between binding posts that had been co nveniently prov ided for thispurpose.l2

1.6.2 THE REFLEX CIRCUIT

The reflex ci rcuit ( Figure 1.1 2)enjoyed some prominence in the early 1920s, but wasmore popular with hobbyists and ex perime nters tha n with commercia l indu stry . Theidea behind the reflex is wonderfu l and subtle, and perhaps even the inventor of thecircuit himsel f (believed to he French engineer Marius Latour .H ) did not full y ap-prcciatc just how marvelous it was. The bas ic idea was this: pass the RF throu ghsome numbe r (say. one) of amplifier stages, demodu late. and then pass the audio backthrough tho se same amp lifiers. A given tube thus simul taneou sly amplified both RFand AF signals.

)2 Oneanecdotal report ha~ it that de Fores t su ld receivers with a wire thaI protruded from !:he hac kpanel. maned ..... ilh a lallclthat ~id so melhing like t'Do 001cut this wire'; it converts this receiverinto a regeneranve one." I have nul found a primary source for this information. bUl it is e ntirelylon~i!ll.enl with all we know ahuut de Poresi's charac ter.

l) It should be noeed thai AmlStnmg 's SCX'uod paper on the s u~rhelerodyne (pu hlished i ll 1924)contains examples of rd leJ; circuit s.

18 CHAPTER 1 A N ONLINE AR HISTORY O F RADIO

The reason that this arrangement made sense beco mes convincingly clear onlywhen you consider how this connection allowed the overall system to p O S5eSS a gain-bandwidt h product thai exceeded that of the active device itself. Suppose that thevacuum lube in question had a certa in constant gain- bandwidth prod uct limit. Fur-ther assume thai the incoming RF signal was amp lified by a factor G RF over a brick-wall passband of bandwidth B. and Ihat the audio signal was also ampli fied by afac tor G Al' over the same brickwall bandw idth B. The ove rall gain-bandwidth prod-uct was therefore (GRFGAF) B, while the gain- band width product of the combinedRF/AF signal processed by the amp lifier was ju st (G RF + GAF) B . For the reflex cir-cuit to have an adva ntage we need only the product o f the gains to exceed the sum ofthe gains, a criterion that is easily satisfied .

The reflex circ uit demonstrates that there is nothing fundame ntal about gain-bandwidth , and that we are effective ly fooled into bel ieving that gain and bandwidthmust trade off linearly just because they com monly do . The reflex circuit showsus the error in our th inking . For this reason alone, the reflex circuit deserves moredetailed trea tment than it commonly receives.

1.7 ARMSTRONG AND THE SUPERREGENERATOR

Armstrong wasn' t content to rest , although after having invented both the regenera-tive and superheterodyne rece ivers he would seem to have had the right.

While experimenting with the rege nerator, he noticed that under certain condi-tions he could, for a fleeting moment , get much greater amplifica tion than normal.fie investigated further and developed by 1922 a circ uit he called the superregenera-tor, a circuit that provides so much gain in a single woe that it can amplify therma land shot noise to audible levels!

Perhaps you found the reflex princip le a bit abstruse ; you ain 't seen nothin ' yet.In a superrcgenerator the system is pu rposely made unstable, hut is periodica lly shutdown (quenched) to prevent getting stuck in SOITlC limit cycle .

How can such a bizarre arrangement provide gain (lots of gain)? Take a look atFigure 1.13. which strips the supcrrege nerator 10 its basic etemc nts." Now. duringthe time that it is active (i .e., the negative resistor is connected to the circuit), thissecond-order bandpass system has a respo nse that grows exponentially with time.Respon se to what'! Why, the initial conditions, of course! A tiny initial voltage will,given sufficie nt time, grow 10 detectable levels in such a system. The initial voltagecould conceivably even come from thermal or shot-noise processes.

.14 Theclassicvacuum tube superregenerator looks a lu( like a normal regencrauve amplifier. excepttha t tbe grid-leak bias nerwork ume cons tant is made very large and the feedhad . (via (he lick-ler coil) is large enough 10 guarantee instahihty . As (he amplitude grows, the grid- leak bias alsogrows until n curs off Ihe tube. The rube remains l;UIoff until (he hias decays 10 a value that returnsthe tulle to Ihe active region. Th us, no separate quenc h oscillator is n-es.\ ary.

j

1. 7 ARM STRON G AND THE SUPERREGEN ERATO R 19

.. AuJ io Oul .Demod.

Quenched when ope n;Regenerative when closed

a

iC,,

I.e is Iliited to de~i red inplII signal

Quenc hO",illalor

FIGURE 1.13. Soperregenerative receiver ba~iC5 .

The problem with all rea l systems is that saturation eventually occurs. and no fur-ther amplificat ion is possible in such a state. The supe rrcgenerator evades this prob-lem by period ica lly shutting the system down. Th is period ic "q uenching" ca n bemade inaudible if a sufficiently high quench frequency is chosen.

Because of the exponential gro wth of the signal with time. the superregencratortrades off log of gai n for bandw idth . As a bonus. the unavo ida ble no nlinearity ofthe vacuum tube can be ex ploited to provide demodulation of the amplified signal!As you might suspect , the superregenemror ' s action is so subtle and co mplex that ithas never been understood by more than a handful of peop le at a given time. It's aquasipenodicall y time-varyi ng. nonlinear sys tem that is allowed 10 go inrermiuentlyunstable, and Armstrong invented it in 1922.

Armst rong sold the paten t rights 10 RCA (who shared Armstrong's view that thesuperregenera tor wax the circuit to end all ci rcuits), and became its largest share-holder as a resutt." Alas, the supcrregcnerator never ass umed the dom inant positionthat he and RCA's David Sarnoff had envis ioned. The reason is simple for us to seenow: every superrcge ncmtivc am plifier is fund ament all y also an oscillator . Th ere-fore, every superregcncrativc receiver is also a transmi tter thai is capable of caus-ing interference to nearby receivers. In addi tion. the superrege nenu or produce s anannoyingly loud hiss (the amplified thermal and shot noise) in the absence o f a sig-nal, rather than the relative qui et of oth er types of receivers. For these reasons, thesuperregenerator never took the radio world by storm.

The ci rcuit has found wide applica tion in toys. however. When you've got toget the most sensitivity with abso lutely the minimum number of ucuve devices, you

)j In a hil of fort u jtous liming. Armstrong so il! his stock just before the grea t stoc k-market cra sh of1929.

20 CHAPTE R I A NONLIN EAR HISTORY OF RADIO

ca nnot do better than the superrege nerative receiver. Radio -controlled ca rs, au-tomanc garage -doo r ope ners, and toy walkie-talkies almost invariabl y usc a circuitthat co nsists of j usI one transistor ope rating as a superregenemtive amplifier/detec tor,and perhaps two ur three more as amplifiers of the demodulated audio signal (as in awalk ie -talkie) . The ove rall sensitivity is often of the same order as thut provided bya typical superhet. On top of those attributes, it can also demodulate FM through aprocess known as slope demod ulation: if o ne tunes the rece iver a bit off frequencyso thai the receiver gain versus frequency is not flat {i.e. has so me slope. hence thename ). then an incoming FM signal produces a signal in the rece iver whose ampli-tude varies as the frequency varies: the signal is converted into an AM signal that isdemod ulated as usual (""it's both a floor wax (mil a dessert toppin g") . So , if most ofthe system cos t is associated with the number of active devices, the superregenerativereceiver provides a remarkably economical solution,

1.8 OLEG LOSEYAND THE fiRSTSOLlDSTATE AMPlifiER

Surely one of the most amaz ing (and little- known ) stories from this era is that ofself-taught Soviet engineer Oleg Losev and his solid-stale receivers of 1922. Vac-uum tubes were expensive then, particu larly in the Soviet Union so soo n after therevolution. so there was naturally a great desire to make rad ios on the cheap .

Losev's approach was to investigate the mysteries of crystals, which by this timewe re all but forgotten in the WeM. l ie independently rediscovered Round 's carborun-dum LEDs. and actually published about a half doze n papers on the phenomenon.Be co rrectly deduced that it was a quantum effect. descr ibing it as the inverse of Ein-stein's photoe lectric effect. and correlated the short waveleng th cutoff energy with theapplied voltage , He even noted that the light was emitted from a particular cry stallineboundary (which we would ca ll a ju nction), and cas t doubt on a prevailing theory ofa thermal origin by showing that the emission could he electronically mod ulated upto

211.9 EPil OGT - - -_+----.c"- I

v, - I hI -2kU t}p.

./

Diode ("har,M;leri sl il;~

FIGU Il: E 1.14 . lmev ', crystadyne receiver {$ingle -slagel .

The reaso n almost no o ne in the United Sta les has ever heard of Losev is simple.First , 31010:-.1no one has eve n heard of Armstrong - it seems there isn' t much interestin preserving the names o f these pio neers . Plus. most of Lose... 's papers arc in Ger-man and Russian. limiting readers hip. Add the generally poor relations betwee n theUnited Stales and the Soviet Union over most of this ce ntury, and it's actually a won-der that unyotlt' knows who Losev was. Lose v himself isn't around beca use he wasone of many who slanted 10 death duri ng the terrible siege of Le ningrad. brea thinghis last in January of 1942. His colleag ues had advised him 10 leave. hUI he was j usttoo interested in finishin g up what he termed " pro mising experiments with silicon:'Sadly, all record s of those experiments have apparently been lost.

1.9 EPILOG

By the ea rly 1930s, the supe rhet had been refined to the poin t that :1 single tuningcontrol was all that was required. Th e superior performance and case of use o f thesuperhet guaranteed its dom inance (as well as that o f RCA ), and virtually every mod-em receiver, ranging from portable radio", to radar sets, employs the superheterod yneprinciple; it seems unlikely tha t this situation will change in the ncar future, It is atribute to Armstrong's genius tha t a sys tem he conceived durin g World war I stilldominates on the C\lC otthe 2 1st ce ntury.

Armstrong, annoyed by the static that plagues AM radio, went on to develop (wide-hand) frequency mod ula tion . in defiance of theoret icians who declared FM usele ss."Unfortunate ly. Armstrong's life did not end happi ly. In a sad example o f how our le-gal system is often ill-equ ipped to deal intelligent ly with tech nical matters, de Forest

.l6 Bell Laboratories mal~mal ician John R. Carso n (no known rcunon to the entertainer) had cor-reedy ~htJwn lhal FM a lway!'> ret.ju ires more bandwid th than A~1. disproving a prevailing belief Inthe contrary . Rut he went loti far in declanng FM worthlevs.

I

22 CHAPTER I A NON LINEAR HISTO RY OF RADIO

cha llenged Armstro ng's regeneration paten t and ultimately prevailed in some of thelongest patent litigation i ll history (i t lasted twe nty years). Not long afte r the courtsha nded dow n the final adverse decision in this case. Armstrong began locking horn swith his forme r friend Sarnoff and RCA in a biller batt le over FM that raged for wellover another decude . His energy and mon ey all bur go ne. Arm strong comm itted sui-cide in 1954 at the age of63 on the fort ieth anniversary of his demonstration of regen -oration 10 Sarnoff. Armstrong's widow. Marian. pic ked up the tigh t and eve ntuallywen! on to wi n every legal battle; it took fifteen years.

De Forest eventually wentlegit . He moved to Hollywood and worked o n devel-oping sound and co lor for motion pictu res. A few years before he died at the ripe oldage of 87, he penned a characte ristically self-aggrandizing autobiography tit led TheFather of Radio that sold fewer than a thousand copies. l ie also tried to get his wifeto write a book ca lled J Marri ed a Genifu, bu t she so mehow never got around to it.

FURTHER READING

Th e stories of de Forest . Armstrong , and Sarnoff are wonderfully recounted by TomLewis in The Empire of the Air, a book thai was turned into a film by Ken Bums forPBS. Although it occasionally gets into trouble when it ven tures a technical explana -tion . the human foc us and rich biographical material tha t Lewi s has unearthed muchmore tha n compe nsa tes. (Prof. Lewi s says that many corrections wi ll be incorporatedin a la ter paperback edition of his book .)

For tho se interested in more technical details. there are two excellent books byHugh Aitken. SyntollY ami Spark recounts the earliest days of radiotelegraphy. be-ginning with pre -Hert zian ex periments ami end ing with Marco ni. Tile ContinuouslVm'e takes the story up 10 the 1930 s, covering an; and alternator technology in addi-tion to vacuum tubes. Curio usly, thou gh . Armstro ng is hut a minor figure in Aitken'sportrayals.

The story of early crystal de tectors is well told by A . Dou glas in " The Cry stal De-tector" (IEEE Spectrum, Apri l 1981 , pp. 64-7) and by D. Th ackeray in " When Tube sBeat Cry sta ls: Early Radio Detectors" (IEEESpectrum, March 1983, pp. 64-9). Ma-terial on oth er carly detectors is found in a delightful volume by V. Phillips, EarlyRadio Uin'f' Detectors (Peregrinus. Stevenage. UK. 1980 ). Finally, the story of Lo-sev is recounted by E. Loe bner in "Subhistories of the Light-Emitting Diode" (IEEETram . Electron Devices, July 1976, pp. 675-99).

1.10 APPENDI X: A VACUUM TUBE PRIMER

1.10 .1 IN TRODUCTION

Sad ly, few eng inee ring students are ever exposed to the vacuum tube . Indeed, mos teng inee ring faculty regard the vacuum tube a qu aint relic . We ll. maybe they're right .

J

1.10 A PP EN DIX: A VACUUM TUBE PRIMER

J\ = J

I I[}=~

e .....~..........

23

Cathode Plale (t'wing")

FIGURE I,1.5 . Idealized diode ~tructure .

but there are still certain engineering provinces (such as high-power RF ) where thevacuum lube reigns supreme . Thi s appe ndix is intended 10 provide the necessarybackground so that an engineer educated in solid-state circuit design can develop atleast a superficial familiarity with this historically important dev ice .

TIle operation of virtua lly all vacuum tubes can be understood rather easily onceyou study the physics o f the vacuum diode. To simplify the develop ment. we' llfollow a historical path and consider a para llel plate structure rather than the morecommon coax ial structures. The results are easier to derive bUI still hold generally,

1.10.2 CATHODES

Consider the diode structure shown in Figure 1.15. The left-most elec trode is thecathode, whose job is to emit electrons. The plate' s job is to co llect them.

Allthe early tubes (Ed iso n's and Fleming's diode, and de Forest' s triode audio n)used directly heated ca thodes, meaning that the light-bulb fi lament did the work ofemitting electrons. Physically all that happens is that , at high enough temperatures.the electrons in the filament material are given enough kinet ic energy that they canleave the surface; they literally boi l off.

Clearly, materials that emit well at tem peratures below the meltin g point make thebest cathodes. De Forest 's first filament s were made of the same carbon variety usedin Edison's light bulbs, although tantalum. which has a high melting point (abou t3100 K), quickly replaced carbon. Usefu l emission from tantalum occ urs only if thematerial is heated to bright incandescence, though , so the early uudion s were prettypower-hungry. Additionally, tantalum lends to crystallize at high temperature. andfil ament life is unsatisfactory as a consequence of the attendant increasing brittleness.A typical audion filament had a lifetime as short as 100- 200 hours. Some uudionswere made with a spare filament that could be switched in when the first filamentburned out.

24 CHAPTER 1 A N ONLINE AR HISTORY O F RADIO

Resea rch by W. D. Coolidge (same guy who developed the high-power X-ray tube )at GE allowed the use of tungsten (melt ing point: 3600 K ) as a filame nt materia l.He found a way to make filament s out of the unwieldy stuff (tungsten is not ductile.and hen ce it ord inari ly cannot be drawn into wires) and opened the path to great im-provernenrs in vacuum tube (and light bulb) lon gevity because of the high mel tingpoint of Ih OlI material. "

Unfortunately , luis ofheal ing power is required to maintain the operating tem pera-ture of about 24(X) K . and portable (or even luggable ) equipment ju st could not evolveunti l these heating req uiremen ts were reduced. One path to impro vement (d i5COV-cred accidenta lly) is to add a little tho rium to the tungsten . If the temperature is heldwithin rather narrow limit s (around 19

1.10 APPENDIX: A VA( UUM TUBE PRIM ER

cathode. without worrying (much) abo ut the injection of hum that would occur if ACwere used in tubes with di rectly heated cathode s.

The draw back to oxide-coated ca thode s is that they are ex traordinarily sensitiveto bombardment by positive ions. And to make thin gs worse. the cathodes them -selves tend to give off gOis over time. especially if overheated. Th us. rather e luboruteprocedures must be used to maintain a hard vacuum in tubes using such cathodes .Aside from pumping ou r the tulle at te mperatures high enough to ca use all the ele-ments to incandesce .

26 CHAPTER I A NO N LIN EA R HISTORY OF RADIO

Now, the current density J (i n A/ml ) is j ust the product of the volume chargedensity p and velocity. and must be independent of x . Hence we have

so thai

t q,,(X)J = p(x)v(x) = p (x )

m,

R'p (x) = J 2 .q ,,( x )(3)

(4)

Thi.. last equation gives us one relationsh ip betwee n the charge density and the p0-tcrniul for a given current density. To so lve for the potential (or charge den sity) wetum 10 Poisson's equation. which in one-dimensiona l fonn is j ust

1,10 APPE ND IX: A VA(UUM TUBE PR IMER

I

-+-~=---------- V

FIGURE 1,16. V-I charoderis& s of diode (space cha~ imitedl .

27

The 3j2-power relationship betwee n voltage V and current I (sex Figure 1.16 ) isbasic to vacuum tube operation (even for the more common coax ial structure ) andrecurs frequently. as we shall soo n see.

As stated previously. the V- I characteristic j ust derived assumes thai the currentflow is space chargc- Iimited." Tha t is, we assume that the ca thode's ability to sup-ply electrons is not a limiting factor. In reality, the rate at which a cathode ca nsupply electron s is not infinite and depen ds on the cathode temperature. In all realdiodes, there exists a certain plate voltage above which the curre nt ceases to followlhe 3/2-power law becau se of the unavailabi lity of a sufficient supply of electrons.This regime. know n as the emission-lim ited region ofoperation. is usually assoc iatedwith power dissipa tion sufficient to cause destruction of the device. We will gener-ally ignore operation in the emission-limited regime. although it may be of interestin the analysis o f vacuum tubes ncar the end of their useful life, or in tubes ope ratedat lower-than-normal cathode temperature.

The diode structure we have j ust analyzed is normally incapab le of umpliticarion.However, if we insert a poro us contro l electrode (know n as the grid) between cath-ode and plate, we can modulate the now of current. If ce rtain elementary co nditionsarc met , power gain may be readily obtaine d. Let's sec how this work s.

Figure 1.17 shows a triode tha t is quite similar to the structures in de Forest' s firsttriode audions. and its operation can he understood as a relatively straightforward ex-tension of the diode. The field that controls the current flow will now depend o n boththe plate-to-ca thode vo ltage and the grid-to- cathode voltage . lei us assume that wemay replace the vo ltage in the diode law with a simple we ighted sum of these twovohages. We then write, using notational conventions of the era:

}II And. a\ staled earlier. it also as..\ umes zero initial velocity of elec trons e mitted from the cath -ode and ~glec ts contact potential diffe re nces between plare and curhode. Th is correct ion usuallyamounts 10 less than a volt and therefore is important on ly Ior low pkue -ro-carhod e voltages.

-28 CHAPTER. 1 A N ON LIN EAR HISTORY O F RA DIO

Cathode Grid Plate ("wing")

FIGURE 1.17. Idealized planor lriode structure .

"c,

_I-_~I-_~I-_~L .. More nega tive Ec

+~:::.'""':::::..._~=-~:::._------ .. Ell

FIGURE 1.18 . Triode characteristics.

(E ) ' /211'1 " ~ = K Ec + ILH , (1 2)

where K is the triode perveance. Ec is the grid-to -cathod e voltage. En is the plate-to -cathode voltage, and In is a roughly constant (though geometry-dependent) pa-rumcter known as the amplifica tion factor . Figure 1.18 shows a family of triodecharacteristics conforming to this ideal relationship.

Physically what goes on is this: elec trons leaving the ca thode feci the infl uence ofan electric fi eld that is a function of two vo ltages. Vol! for volt. the more proximategrid exerts a larger influence than the relatively distant plate. If the grid potential isnegative then few electrons will be attracted to it , so the vast majority will now ontothe plate. Hence, little grid current flows. and there can he a very large power gainus a consequence.

The negative grid-to -cathode voltage and tiny grid current that characterize nor-mal vacuum tube operation is similar to the negative gate-to -source voltage and tinygate current o f depletion-mode n-channcl FETs (field -effect transistors). althoughthis comparison seems a bit heretical to o ld-timers.

j

T 1.1 0 APPEND IX: AVACUU M TUBE PRIMER

cathode

FIGURE 1.19. Incremental model For triode vacuum lube.

29

The analogy betwee n FETs and vacuum tubes is close enough thai even their in-cremcntal models arc essentially the same (see Figure 1.19). Approximate eq uationsfor the transcondu ctance g", (sometimes ca lled the "mutual conductance") and in-cremental plate resistance rp are readily obtained from the V- I rela tionship alreadyderived:

and

( 13 )

( 14 )

Note thai the prod uct of 8m and Tp is simply JI , M ) that If represents the open-circuit amplifica tion factor. Add itionally. note thai the transconductance and plateresistance are only weak functions (cube roots) of operating poin t. For this reason,vacuum tubes generate less harmonic distortion than other devices working over acomparable fractio nal range about a given ope rating point. Rec ull rhat the exponen-tial V- I relationship ofbipolur transistors leads to a Hncur dcpcndencc of 1:m on I . andthat the square-law depe ndence ofdra in current on gate voltage lends to a square -roo tdependence of 1:", on I in PETs. The relatively weak dependence on plate' cur rentin vacuum tubes is apparently at the core of arguments that vacuum tube amplifi ersare "cleaner" than those made with other types of active devices. It is certainly truethat if amplifiers arc driven beyond their linear range that a transistor version is likelyto produce more (perhaps much more) distortion than its vacuum tube counterpart.However, there is co nsiderably less mer it to the argument that audible differencesstill exist when linear ope ration is maintained.

The triode ushered in the e lectronic age, making possible transcontinental telc-phone and radiotelephone communications. As the rad io art advanced, it soo n be-came clear thai the triode has severe high-frequency limitations. The main prob lemis the plate-to-grid feedback capacitance. since it gets amplified as in the Miller c f-feet. In transistors. we can get around the prob lem using cascoding, a tech nique Ihalisolates the output node from the input node so that the input doesn' t have 10charge amagnifi ed capaci tance. Although this technique could also he used in vacuum tubes.

30 CHAPTE R I A NONLINE A R HISTORYOF RADIO

" Ec IncreasingAppnJll. imale screen grid voltage,

j!

r/

FIGU RE l.20 . Tetrodechorocteristics.

(here is a simpler way: add anothe r grid (cal led the scree n grid) be tween the old grid(called the control grid ) and the plate . If the scree n grid is held at a fixed po tential.it acts as a Faraday shield between ou tput and input . and shunts the capacitive feed -hack to an incremental ground. In effect. the ca...coding device is integral with therest of the vacuum tube .

The scree n grid is traditionall y held at a high DC potential to prevent inhibitionof current flow. Besides gelling rid of the Miller effec t prob lem. the addit ion of thescreen grid makes the current flow even less depe ndent on the pla te voltage than be-fore. since the control grid "see s" what' s happening

1.10 APPEN DIX: A VACUUM TUBE PR IM ER 31

.. Increasing ~:

r:

/ (Scree n and SUP~'iOf grid voltages held I;nn, la nl)

FIGURE 1.21. Pentacle characteristics,

Well, one grid is good, and IWO arc bette r, so guess what? One way to so lve theproblem of secondary emission is 10 add a third grid (called the suppressor grid ),and place it nearest the plate. The suppressor is normall y held at cathode potentialand works as follows. Elec trons leaving the region past the screen grid have a highenough velocity that they aren' t go ing to be turn ed around by the suppressor grid'slow potential. So they happily make their way to the pla te, and some of them ge ner-ate secondary electro ns, as before. But now, with the .~ ll ppressor grid in place, thesesecondaryelectrons are attracted back to the mo re positive plate, and the negat ive re-sistance region of operation is avoided . With the add itional shielding pro vided bythe suppressor grid, the output curren t depe nds less on the pla te -to -cathode voltage.Hence, the output resistance increase s and pe ntodes thu s provide large amplificationfactors (tho usands, compared with a typica l triode's value of about ten or twenty)and low feedback capacitance ( like 0 .0 1 pF, excl ud ing ex ternal wiring ca pacitance) .Large voltage swings at the plate are therefore allowed, since there is no longer aconcern abo ut negative resistance (see Figu re 1.2 1). For these rea sons, pentodes aremore effic ient as pow er output devices than tctrod cs.

Later, some very clever people at RCA figured OUI a way to get the equivalentof pentode act ion without add ing an ex plicit suppressor grid. Since the idea is j ustto devise conditions rbat repel seco ndary electrons hack to the pla te, you might heable to exploit the natural repulsio n between electrons to do the same job. Suppose,for example, we consider a stream o f electrons flowing be tween two locati ons. Atsome intermediate po int, there can be a region of ze ro (or eve n negative) field if thedistance is sufficiently great.

The effect of mu tua l repul sion ca n be enhanced if we bunch the electrons to-gether. Beam-forming electrodes (see Figure 1.22), working in concert with contro land screen grids wou nd with equ al pitch and aligned so that the grid wires overlap,

32 CHA PTE R I A NONLI NEAR HISTORY O F RAD IO

f IGURE 1.22. BeomptJ"Wef $trvcture (top view).

Calhude -

Plate ---

IElectron Beam

\Contro l Grid Screen Grid

II

force the electrons to flow in shee ts . The conccmrutcd electron beam then generatesa nega tive field region (a virtual suppressor grid) withou t requ iring large electrodespacings . And , as an unexpected bonus. it turns OUI thai the characteristics at lowvol tages are actually superior in some respec ts (the plate CUITCn l and output resis-ranee are higher) 10 those of true pentodes and are thus actually more desirable than" real" pcn todes for powe r applica tions.

Well. this grid mania didn't stop at the pe ntode , or even the hexode. Vacuum tubeswith up 10 seven grids have bee n made. In fact . for decades the basic superhet AMradio (the "All- American Five-Tuber ") had a heptode. whose five grids allowed onetube (usually a 12BE6) to function as both the local oscillato r and mixer, thus re-ducing tube cou nt. For trivia's sake. the All-American Five abo used a 35W4 rec-rifler for the po wer supply, a 12BA6 IF amplifi er, a 12AV6 triode /duo-diode as ademodul ator and audio amplifier, and a 50e5 beam-power audio output tube .

Here's some other vacuum tube trivia: for tubes made utter the ea rly 1930s, thefi rst numerals in a U.S. receivin g vacuum tube's type number indicate the nominalfilamen t voltage (with o ne exce ption: the " loktals" 411 have numbers beginning with7, but they are actually 6 -volt tubes 1110st of the time). In the typical superhet men-tioned previously, the lube filament voltages sum to about 120 volts, so that no fil-ament transformer was required . The last num bers arc supposed to give the totalnumber of clements. but there was widespread disagree ment on what constituted anclement te.g., whether one should count the filamenn. so it is nnly a rough guide atbest . The letters in between simply tell us something about when that lube type was

4" L...klals had a spec ial base that locked the tulles mechanicall y into lhe soc ket 10 preve nt theirworking IUfl:'oe in mob ile operenon s.

1.10 APPE N DIX: A VACUUM TUBE PRIMER OJ

registered with RETM A (which later beca me the EIA). No! all registered rube typeswere manufactured, so there arc many gaps in the sequence.

In CRTs. the first numbers indicate the size of the screen' s diagonal ( in inches inu.s.eRTs and in millimeters e lsewhere). The last segment has the leiter P fo llowedbynumbers. The P stands for " phosphor" and the num bers following it tell you whatthe phosphor characteristics are. For example. P-I is the standard phosphor type forblack-and-white T V CKTs, while P22 is tne co mmon type tor color TV rubes.

The apex of vacuum lube evo lution was reached with the development (If the tinynuvisrorby RCA. The nuvistor used adva nced metal-end-cera mic construction. andoccupied a volume about double that of a TO-5 transistor . A number of RCA co lortelevisions used them as VHF RF amplifiers in the ea rly 1970 s before transistors fi-nally look over completely, RCA's last vacuum lube rolled o lT the assembly line inHamso n, New Jersey soon after. markin g the end of about sixty years of vacuumtube manufacturin g and. indeed, the end of an era .

CHAPTER TWO

CHARACTERISTICS OFPASSIVE IC COMPONENTS

2.1 IN TRODUCTION

We 've see n that RF circuits generally have many passive compo nents. Successfuldesign therefore depends cri tica lly on a detailed understanding of their character-istics. Since mainstream integra ted circ uit (Ie ) processes have evolved largely tosatisfy the demands of d igital electronics. the RF Ie designer has been left with alimited palette of passive devices. For example, inducto rs larger than abo ut lO nHconsume significant die area and have relatively poor Q (typically below 10) and lowself-resonant frequency. Capaci tors with high Q and low temperature coe fficient areavailable. but tolerances are relatively loose (e.g. order of 20% or worse). Addi tion-ally. the most area -efficient ca paci tors also lend to have high loss and poor voltagecoe fficients. Resistors with low self-capacitance and temperature coefficient are hardto come by. ami one must also occasional ly contend with high voltage coe fficients,loose to lerances. and a limited range of values.

In this chapter. we examine IC resistors. ca paci tors. and inductors ( including bond-wires. since they are often the best inductor s avai lable), Also, given the ubiqui ty ofinterconnect , we study its prope rties in de tai l s ince its parasitic at high frequenciescan be quite import ant.

2. 2 RESISTORS

There are relatively few good resi stor options in standard CMOS (compleme ntarymetal-oxide silicon) processe s. One possibili ty is to use pnlysilicon ("poly" ) inter-connect material. since it is more resistive than metal. Ho