Harold Black and the

Negative-Feedback Amplifier
Ronald Mine
n August 2, 1927. Harold Black, a young Bell Labs engineer just six years out of college, invented the negative feedback amplifier in a "flash ofinsight''whileridingthe Lackawanna Ferry across the HudsonRiveron his way to work. (Bell Labs was then located on West Street in Manhattan.) Black recalled, "I felt an urge to write but had nothing to write on so picked upmy moming paper. By sheer coincidenceonepage was blank. Here was a perfect set-up, lots of room and fully dated. With I this to implementmyjob, started the first written record pertaining to the stabilized negative feedback amplifier. Years of studyandmany failures preceded this sudden conception of stabilized feedback. Despite immediate recognition of its importance.yearsofadditional work were required before it foundsubstantialcommercial use" [ 1, p. 7231. Negative feedback became widespread [2]. It allowed the One of the newspaperp u y e s used by B/ac,kto jot down early ideas on feedbar:k. /Photo: AT&T Alrhj1.eJ.i Bellsystemtoreduceovercrowding of lines and extend itslong-distancenetworkby means o f carrier telephony.It enabled the design of accurate fire-control systems in World War II. and i t fonned the basis of earlyoperationalamplifiers, as well as precise, variable-frequent) audio oscillators.Withthetransitionfromvacuum tubes to microelectronics after World War 11. negative feedback retained its status as an integral part of communicationsand control system3 because it was associated

~ith netRork theory from the beginning. Like Maxwell's circuit equations, which were de1860s t o veloped in the understand the induction coil. negative feedback was not tied to the technology of its origins (vacuum tubes) and became a fundamental principle of electrical engineering with innuapplications merable independent of theoriginal hardware used for its invention.

Invention
The journey to theLackawanna Ferry began in 192 1. whenHaroldBlack (18981 9 8 3 ) g r a d u a tf erd om WorcesterPolytechnicInstitute udh a Bachelor's degree in electricalengineeringand took a job with the EngineeringDepartmentatWestern Electric. (The research branch of this department formed the nucleus of Bell Labs when it was established in 1925.) A pressingproblem in the Bell system at thetime was that the his of distortionandinstability vacuum-tubeamplifierswere compounded when they were connected in tandemovera long-distance system based on carrier telephony. (Carrier telephony became feasible in1917 u iththeinventionofa practical wave filter by George Campbell, an ATkTresearcher who had invented the of Michael loadingcoilindependent Pupin around 1900. The gain of the amplifiers varied with plate voltage, temperature. aging of the tubes, etc.. while the nonlinearity of the

82

IEEE Control Systems

tubes created intermodulation distortion in the multichannel carrier system. In 1921, engineers were facing these difficulties on the transcontinental telephoneline.builtin1915 butrecentlyupgradedtoa three-channelsystemwith twelve amplifiers. Bell built a second transcontinental line in 1923 withfourchannelsand twentyamplifiers,andwas planningmorecomplexsystems for the future[3]. Blackbecameintrigued with the problem that year and asked his supervisor for permission to work on improved amplifiers for a system numbering thousands of channels. The supervisor was apparently bemused by the scale of his thinking [4:p. 631 and agreed it did to let him pursue the topic. provided not interfere with his other work. Black approached the problem by trying to make vacuum tubes operate in a more linear fashion. Le., he wanted to make the active elements of theamplifierproduceless distortion. Mervin Kelly, well-known later as the head of Bell Labs when his researchers invented the transistor in 1947. was then in charge of electron tube research and cooperated with Black on the endeavor, but to no avail [SI. Blackthenhadanimportantinsight that tielped him reframe the problem. He recalled that in 1923, .I attendeda lecture by C. P. Steinmetz (chiefengineer at General Electric] at an ALEE meeting [the American Institute of Electrical Engineers, a forerunner of the IEEE] and was impressed by the Steinmetz way of getting down to the fundamentals of a problem. As a result I restated my assignment as beins that of removing distortion products from the amplifier output. I immediately observed that by reducing the output to the same amplitude as the input and subtracting one from the other, the distortion products only would remain which could then be anplified in a separate amplifier and used to cancel out the distortion products in theoriginal amplifier output ... Thus, the Feedforuard Amplifier came into being [4. pp. 64-65]. Although not based on negative feedback, this amplifier was an essential step to its invention because Black had reformulated the problem. He was no longer trying to prevent vacuum tubes from causing di3tortion.Heaccepted that distortionand
August 1993

high-gain device. Correctly designed, negative feedback reduceddistortionandnoise. while stabilizing gain by making it dependent on the passive feedbacknetwork,instead of the troublesomeactiveelements (vacuumtubes).Twoother AT&T researchers, H.T. Friis and A.G. Jensen, had investigated the feedbackcaused by the plate-to-grid capacitance in a vacuum tube and its influence on amplification in 1924. They noted this feedback could increaSe or decrease the amplification(i.e.,bepositiveornegative). But they regarded plate-to-grid feedback as an unwanted phenomenon in amplifiers, to be neutralized by adding a capacitor. and did not see the benefits of nepative feedback [91. ~. sought a way to reduce it at the output of Black probably knew about their work the amplifier. This was a critical step and because it was published in the Bell Sysput him on the track to the eventual solu- tem Tcr,hr7iccr(J m m a l , but it is unclear if tion. it influenced his research. Black recalled that he was working on the feedforward Although resurrected in the 1970s for and other related topics when he had the single-sidebandmicrouaveradio.the feedforward amplifier did not work well insight about negative feedback. Mewin for carrier telephony in the 1920s. Black Kelly. however, said, Finally, a mathethat by appliedfor a patent ontheinvention in maticalanalysisconvincedhim February 1925. which was issuedin Octo- merely inserting apart ofthe outputpower ber 1928 [6], and then, as a member of a into the input in negative phase, he could development group at Bell Labs. tried to obtain an) desired reduction in distortion products at the expense of a sacrifice in make it norh in a systems environment. He built an experimental amplifier to re- amplification. His f i m 1 marhematical was duce the invention to practice, hut precise ar~cr/ysi;\wasconceivedwhilehe balances and subtractions o f signals were crossing the Hudson River on the Lackawanna Ferryboat en route from home to hard to achieve and maintam in practice. This \\as the state of Blacks work on the laboratories [5. p. 722. my emphasis]. Black had the page of The New York telephone amplifiers in 1927. the year of Time5 witnessed by a co-worker the momthe famous flash of insight on the Lackm a n n a Ferry. What occurred betueen the ing of his insight, then set out to build his feedforward experiments and that fateful amplifier and prepare a patent application. morning onthe Ferry is not clear. Negative He submitted an extremely long application feedback was a concept diametrically op- (5.2 pagec. 126 claims) in 1928. but the posed to feedforward and was not an ob- patent offke objected to many ofthe claims, vious directionto explore. Black probably apparently because his concept of negative knee about the two main types of feed- feedback flew in the faceof accepted theory. back at the time: 1 ) the comparison of The examiners finally awarded the patent nine years later, in December1937 [IO], outputandinputsignalstogeneratean error signal to control the output, which after Black and others at AT&T developed had been M idely used in mechanical and both a practical amplifier and a theory of 7 1 :and 2) posi- ne&\ e feedback. electrical control cystems1 tive feedback. which had been used for oscillationandincreasedamplification Development (regeneration) in radioequipment since As Black recalled. the road from the about1913 [8]. But Black took a much Lackau, anna Ferry to a practical amplifier different approach. He used negative feed- wac long and rocky. He had adifficult time back to reduce the amplification of a Pery withtheamplifiersinging[breaking x3

into oscillations) and devised a designrule to guardagainst this instability. In May1928. Hany Nyquist (1889-1976) and other communication engineers at AT&T conferred with Black about using his amplifier for a newcablecarriersystem. Nyquist, who received his Ph.D. in physics from Yale in 19 17, thought Black's design rule was too suingent anddid an analysis of negative feedback. This work led to what later came to be known asthe "Nyquistcriterion" for determining when an amplifier with negative feedback was stable. He published the paper containing the criterion in 1932 during the patent office deliberationsonBlack'spatentand n 1934 [ I 11. joined Bell Labs i Black recalled. "Although this criterion is simple in expression and application. Nyquist's derivation of it required a mathematical-physical intuition given to few men" [ 1, p. 7231. Black's classicpaper on the negative feedback amplifier, published in 1934, referred to Nyquist's paper and his stability criterion 1121. In that same year, during the developmentofacoaxial-cable camer system with a passband of 1 MHz and the possibility of several hundred amplifiers. anotherBellLabstheorist.HendrikBode (19051982). led a group of mathematicians in the development of design techniques that took full advantageof Black's invention [13]. Bode, an applied mathematician who received M.A. an from Ohio State in 1926 and a Ph.D. in physics from ColumbiaUniversity in 1935 1141. published his paper. "Relations Between Att e n u a t i o na n dP h a s ei nF e e d b a c k Amplifier Design." in I940 [ 151. The paper, and a resulting book in 1945 (which included what engineers now call "Bode plots"), used thepowerfultools of network theory to show how to design feedback amplifiers with the desired gain and frequency response in a precise manner [16]. Thus, although Blackis usually recognized as the inventor of thenegative feedback amplifier, its deyelopment and the recognition of its possibilities in communications, measurements, and control systems were the result of a group effort between engineers and theorists with ad~~

.Jt

! & a+"%&J-lLf-.'

vanced training in mathematics and physics - a common a q e c t of the histor! of electronics in this period.

Hewlett-Packard and the Negative Feedback Amplifier

The application of the negative feedback principle was taken u p fairly rapidly at Stanford University by electrical engineering professor Frederick Terman and hisstudents. I n 1939. Terman.William Hewlett, Robert Buss, and Francis Cahill wrote a paper that described many uses for negative feedback: in a laboratory audiofrequency amplifier, an audio-frequency voltmeter, a tuned-radio receiver, high-Q circuits. and laboratory oscillators. Negative feedback resulted in low distortion, stable gain. and a small phase angle,while protecting the circuitsfromtheharmful effects of vacuum-tubeaging,variable supply voltages. and so forth. These characteristics were of high value in precision voltmeters. osc~lloscope\, and oscillators. Terman. who often is called the founder of what came to be known as "Silicon Valley" hecause of his promotion of commercial ties between Stanford and local clectronicsfirms 1171, recalled in the 1970s that the "resis-

tance-capacitance tuned variable frequency audio oscillator described in this paper was WilliamHewlett'scontribution to the paper, and was furthermorethefoundationon whichtheHewlett-Packard Company was built" [18]. T h eh i s t o r yo fH - P ' s founding has elements of the stereotypical story of independent inventors working alone in agarage.DavidPackard andHewlettdidstarttheir company in Packard's garage (Hewlett lived in a cottage on thegrounds). But theinfluence of Stanford and Terman was considerable. Hewlett and Packardreturned to thearea after graduating from Stanford in 1934. Packard took a leave of absence from GE to accept a fellowship in 1938. while Hewlett returned to work underTermanafterfinishinga Master's degree at M.I.T. in 1936. Terman encouraged the to marpair to form acompany ketHewlett'svariable-trequencyoscillator, lent them $538. helped them get a bank loan of $1000, and helped them work out a deal with IT&T who bought their international patent rights in exchange for underwritingtheir U S . patentapplication. An earlycustomerfortheaudio-frequencyoscillators was DisneyStudios, who ordered eight for the film "Fantasia." By 1940, the fledgling company had nine employees and had moved out of the garage to develop a full line of products based initially on thenegative-feedback amplifier 1191.

Common Themes
The story of thi\ amplifier illustrates many themes in the history of technology. Although the in\.ention can be traced to a "flash of insight" by a single person. the inventor was well trained in mathematics, engineering. and science. and had been working with others in this area. The development of the amplifier was truly a team effort at Bell Labs: theorists with excellent mathenutical <killsdeveloped a theory that helped engineers understand the original invention and developit further. The application of the in\ ention by Hewlett and Packard not only started a new business. but reinforced the connections between the univer-

84

IEEE Control Systems

.4cknowledgment
The author thanks Richard Compton andSheldonHochheiserforreading an earlier draft of this paper.

ley." Anre,, Heritqyr 1 f 7 1 ~ n / , T c ~ h n o i io1.6, , pp. 18-24, Spr./Sun?. 1990: and Stuart W . Leslie and Bruce He\ I!. "Steeple buildmg at Stanford: Electrical englneennp. physic<.and micro\mw research." Prnc. IEEE. vol. 73. pp. 1169.1 180, July 1985.
[ 181 F.E. Terman. W.R. He\+letr e / a/,"Some application> of negatlve feedback wlth panicular reference to laborator) equlprnent." Pro<. IRE. vol. 27. pp, 619.655. Oct. 1939; reprinted ~n J.E. Brittain. Ttvniny Point.\ in Anwicuti Eiec.rric.al H i m r , ~pp, , 35 1 - 3 7 : quotation on p. 350.

References
[I] Harold Black."lnvention in engineering. Eng., vol. 77, pp. 722.723. Aug. 1958.
Elel.

[2] SruanBennett,A Hi.iror? ~~C~onl~.o/E,fcifreer-rng, 1800-1930. London: Peter Peregnnus. 1979.

[3] Harold S. Black, "Inbenting the negatne feedback ampllfier." / E Spe(rrunr. vol. 14. pp, 5360, Dec. 1977.
[4] E.F. O'Neill, Ed.. A H D f o l : i ofE,r,e~~?eer~qq and Srienw i n the B e / / S?sien~ Trmrsn?f.s.\fofi TecArrolop) //9X-/Y751 Bell L h . 108.5.

[I91 TeklaPerr)."When the car bas OUI. the busmess H B S 'III'.'' / E Specrrufn. vol. 2 5 , pp. 44-45. Apr. 1988.

[20]Ronald Kline. "An overview of t\wnly-fiW bears of electrical and electronics e n p e e r i n g In the Pro( eedifi,y\of rhr / E . 1963-1987," P1.o~ / E . vol. 78. pp. 469-485, Mar. 1990.

BuildSim"'1 .O
The new approach to sirnulatfon o n the M a c ~ n t o c h & PC
Point & Click

CORRECTIOIV
The footnote to the article "Sliding Mode and Classical Controllers in Magnetic Levitation Systems." b) Dan Cho, YoshifumiKato.andDarinSpilrnan (IEEE Control Systems Maqa:ir?e. Feb. 1993, pp. 42) incorrectly identified palt of Y . Kato's company affliation. Kato wasonleave from "Nippondenso Co.. Ltd." (not "NipponDenso Co.." a \ stated in the footnote).

Modeling

Automated Proiect Management

,
1
I

6 lnte ration
I

MePhods

Reader Service Sumber 4