Università degli Studi di Padova

Padua Research Archive - Institutional Repository

Solidifying [Historical]

Original Citation:

Availability: This version is available at: 11577/3271203 since: 2018-06-13T09:33:26Z

Publisher: Institute of Electrical and Electronics Engineers Inc.

Published version: DOI: 10.1109/MIE.2018.2791062

Terms of use: Open Access This article is made available under terms and conditions applicable to Open Access Guidelines, as described at http://www.unipd.it/download/file/fid/55401 (Italian only)

(Article begins on next page) Historical

by Massimo Guarnieri

Solidifying Power Electronics Massimo Guarnieri

ore than a century ago, in electrochemical plants. The idea of us- 1902, American Engineer ing grid control in combination with M Hewitt (1861– phase retard to modulate ac power had 1921) derived the -arc recti- emerged in the early 1920s, and, by fier, enclosed in a glass bulb, from his 1925, the name inverter was introduced. mercury-vapor lamp of the previous Fundamental converter topologies ap- year. He devised its use for feeding dc peared during those two decades. New motors from alternating currents. As York City’s underground was dc fed the first for power uses (two through 3-MW mercury-vapor recti- years before John Ambrose Fleming’s fiers in 1930, and German railways ad- diode and four years before Lee de opted mercury-arc cycloconverters for Forest’s audion [1]), the mercury-arc a universal motor drive in 1931. Such rectifier marked the birth of power power grid applications called for de- electronics. The device was used in vices capable of higher voltages. 1905 in Schenectady, , for By 1932, AEG was producing high- powering a dc line for incandescent power (750 kW at 3 kV) mercury-vapor lamps and, soon after, in for rectifiers provided with a glass casing battery chargers and electrochemical (Figure 1). GE used similar devices in processes, including aluminum reduc- the 12-kV dc transmission line con- tion and electroplating. Aiming for necting a 40-Hz power station to 60-Hz higher powers, Cooper Hewitt intro- loads at Mechanicville, New York, in duced the metal casing with water 1932 [4]. The device, known as an ig- cooling in 1909. nitron, is a high-current, high-voltage In the same year, the Hungarian en- FIGURE 1 – A large GE thyratron (used in (HV) mercury-vapor controlled rec- gineer Béla B. Schäfer (1879–?), of the pulsed radars) next to a miniature 2D21 tifier developed by Joseph Slepian German company Hartmann & Braun thyratron (used to control relays). (Courtesy of (1891–1969) at Westinghouse in 1931. Wikimedia Commons.) (H&B), applied for his first patent on It has a water-cooled steel casing for mercury-arc rectifiers, which were first withstanding high currents between delivered to a German foundry in 1911. also entered the market in the 1920s, a cathode and an anode, started by a After an agreement between H&B and while the devices’ ratings gradually triggering pulse in the third electrode, the Swiss-owned Brown Boveri Compa- increased [3]. and was used in topologies such as an ny, high-power models (300 kW at 600 Early studies on controlled rectifi- inverse parallel for ac control in welding V) with a metal casing were produced cation in gas tubes were developed and heating and in bridge rectifiers for in 1913 [2]. Steel-tank rectifiers could at GE by Irving Langmuir (1881–1957) supplying railways and mills. Recently, carry 750-A currents by 1915, and and his colleagues in 1914, when they it has been replaced by semiconduc- General Electric (GE) started produc- discovered that a negative potential tor switches, but it is still used in some ing similar devices in 1919; Siemens– introduced between a cathode and an high-power pulsed applications because Schuckert introduced a water-cooled anode by a third electrode prevented of its robustness. model in 1920. Westinghouse and Allge- the anode current. High-power mercu- The device, known as a thyratron, meine Elektricitäts–Gesellschaft (AEG) ry-arc controlled valves, i.e., switches, was first used close to 1928 [5]. It was were developed in the 1920s and 1930s, a hot-cathode mercury-arc controlled

Digital Object Identifier 10.1109/MIE.2018.2791062 providing significant improvements to valve that was developed by building Date of publication: 21 March 2018 electric railway converter stations and on the argon-filled vacuum tubes used

36 IEEE INDUSTRIAL ELECTRONICS MAGAZINE ■ MARCH 2018 in radio detectors. It featured an anode, galena cat’s-whisker detectors [8]) with one frequency to the other in operating a heated cathode, and a grid acting as current densities up to 310 mA/cm2 ac motors at variable speed and with the control electrode that could start without a heat sink and limited direct/ transforming high ac voltage to low ac an anode-to-cathode positive current inverse voltages (i.e., lower than 6 V). voltage [10]. and was enclosed in a glass casing However, by 1948, copper-oxide rectifi- The invention of the point-contact filled with mercury vapor to provide ers had grown considerably. Assembled transistor at Bell Labs by John Bard- low direct voltage. It became popular in series and in parallel, they could cov- een and Walter H. Brattain in late 1947, for a number of low-power applica- er a wide range of different applications and the junction transistor by William tions. Uncontrolled diode versions, with currents up to 25–100 kA (recti- B. Shockley in early 1948 [11], opened known as phanotrons, were also pro- fiers for plating) and voltages up to the door to a completely new genera- duced. The first variable-frequency ac 1,500 kV (atomic physics experiments) tion of power devices, starting with drive of a synchronous motor was de- [9]. Building on early observations a controlled semiconductor rectifier. veloped by Ernst Alexanderson (1878– by Braun, William G. Adams, and others After joining Bell Labs in 1951, Jewell 1975) at GE Corporate Research and in 1874–1876, Ernst Presser at TeKaDe, James Ebers (1921–1959) developed Development (GE-CRD) in 1934 using Germany, developed the selenium rec- a model of a p-n-p-n device as the thyratrons [6]. These devices are still tifier in 1928. It consisted of a layer of combination of a p-n-p and an n-p-n used for some high-power (tens of ki- selenium applied to an aluminum plate. transistor. This had to be developed loamps and tens of kilovolts) pulsed Despite operating at a current density as a three-electrode solid-state device operation systems, e.g., lasers, radio- lower than the copper-oxide rectifier, made with four layers of alternating p- therapy devices, and crowbar circuits it was the first solid-state metal diode and n-type semiconductors. Analyses of TV broadcasting stations. suitable for more general power uses, showed that it could act as a bistable After Schäfer sold his know-how to by virtue of rated voltages of up to 30 V. switch conducting between an anode Allmanna Svenka Elektriska Aktiebo- In the 1940s, power electronics dealt and cathode when the gate received a laget [(ASEA), now ABB after merging with the conversion of electric power trigger current and continued to con- with the Brown Boveri Company in from ac to dc (or vice versa) in dc power duct while the voltage across the device 1998] in 1927, an advanced HV mer- transmission, with the conversion from was not reversed (forward biased) [12]. cury-arc valve was made pos- Shockley later claimed he sible by the grading electrode originated the device by reason design invented by Uno Lamm of the hook collector, i.e., the (1904–1989) of ASEA in 1929. It negative resistance effect that allowed HVdc lines operating he interpreted for the transistor, at hundreds of kilovolts to be which also appeared in the p- put into service in subsequent n-p-n switch. The construction decades. Multianode devices of the p-n-p-n switch was under- for multiphase operations, taken by a Bell Labs group led with both steel and glass cas- by John L. Moll (1921–2011), who ings, were available in the late argued in favor of using silicon 1930s (Figure 2). technology to make the device Early power solid-state instead of germanium, as was rectifiers appeared during this customary in the transistors of era. Grondahl [7] and his co- the day. The first silicon transis- workers at Union Switch and Sig- tor was made by Morris Tanen- nal Company had studied the baum (b. 1928) at Bell Labs in conduction between copper January 1954. After Moll’s group and copper oxide since 1920, es- had been reinforced by James tablishing the basis of metallic M. Goldey (b. 1926) and Nick rectifiers. They obtained early Holonyak (b. 1928), prototypes practical devices four years of the p-n-p-n switch were built later and patented the solid- by Tanenbaum, Goldey, and Ho- oxide rectifier in 1927. Thanks lonyak in 1954–1955, all based to planar geometry, it could on silicon. withstand relatively high cur- Bell Labs did not arrive at rents (i.e., 7 A, much higher than an industrialized design; rath- the solid-state galena contact er, the GE rectifier department point rectifier introduced by under the supervision of R.A. FIGURE 2 – A multianode mercury-arc rectifier still in operation Karl F. Braun in 1874 and devel- in October 2010 at Manx Electric Railway substation at the Laxey York started a program in 1957 oped some years later in the Isle of Man. (Image courtesy of Tony S. at Talk Electrician Forum.) to build a silicon-controlled

MARCH 2018 ■ IEEE INDUSTRIAL ELECTRONICS MAGAZINE 37 rectifier (SCR) that was actually a Bell Labs Engineer Sidney Darlington three-terminal p-n-p-n switch [13]. (1906–1997) in 1953 by combining two Starting with a crude technology, (or three) bipolar junction transistors York’s group was able to produce a (BJTs) that share a common collec- kilowatt-level (i.e., hundreds of volts tor [15]. Westinghouse announced a and tens of amperes) prototype by solid-state controlled rectifier called a July 1957. Later the same year, Ho- trinistor in July 1959. By the 1960s, lonyak joined the GE group and the improved switching speed of BJTs worked on improving the device. ushered in BPTs suitable for high- Thanks to the efforts of Frank W. “Bill” frequency dc/dc converters, a pivotal Gutzwiller (c.1926–2011), it was suc- innovation in power electronics that cessfully marketed in early 1958, thus was marketed in the late 1970s. starting a revolution in power elec- In 1958, GE Engineers York and Hol- tronics (Figure 3). Thyristor, the name onyak developed the SCR concept into for the p-n-p-n switch, was coined in a symmetrical switch, building on the 1966 by merging thyratron and tran- shorted-emitter idea. The device they sistor, since it was intended to be a created operated properly and reli- competitor of the former. ably, and it became the prototype of Silicon technology, specifically con- the triode for (TRI- ceived for thyristors, later played a AC), a device capable of conducting fundamental role in the development currents of either sign when triggered. of electronics. Its use became common However, it suffered from turn-off fail- in other solid-state devices and chips, FIGURE 3 – An early thryristor. (Courtesy of ures when used with reactive loads ABB Company.) starting with silicon-wafer process- and from turn-on failures after HV ing to make p–n-junction devices by derivatives, which called for snub- impurity diffusion. Shockley, despite and Gordon Moore), who could take ber circuits. In the case of demand- resigning from Bell Labs in 1953, had the silicon technology in a different ing circuits, two inversely paralleled access to this new silicon technology direction [14]. Other power devices, SCRs were preferred. Holonyak was and took it with him when he moved including power diodes based on ei- the only researcher who participated to California, in an area now known as ther silicon or germanium, were pro- in the first Bell Labs studies as well as Silicon Valley. He made the informa- duced in the mid-1950s and started the early GE work on p-n-p-n switches tion on the silicon p-n-p-n available to replacing mercury-arc valves. Another (SCR and TRIAC), and he did not limit his recruits at Shockley Semiconduc- early device for power uses was the his contributions to power devices. He tor Laboratory, established in 1956 in planar-type Darlington bipolar power built the first real light-emitting diode Mountain View (with Robert Noyce transistor (BPT), a device invented by capable of emitting visible light and the red-light (visible) semiconductor laser, both in 1962 [16]. In the 1970s, the evolution of elec- tronic converters called for devices capable of self-turn-off, fast-switch- ing, and high-withstand capability; they were achieved by building upon SCR technology [17]. A solution was found in the gate turn-off (GTO) thy- ristor. Invented in GE’s laboratories in 1958 and made available in 1962, it was a switch similar to the TRIAC but was fully controllable since it could be turned off at will by ceasing the gate signal. However, GTOs required powerful driving circuits capable of quickly extracting high-reverse cur- rents form the device gate and ex- ternal snubber circuits to shape the turn-on and turn-off currents to pro- FIGURE 4 – A family of GTOs of different sizes from ABB, rated up to 4.5 kV and 4 kA. (Courtesy tect the device against destructive of ABB Company.) commutations. Both devices presented

38 IEEE INDUSTRIAL ELECTRONICS MAGAZINE ■ MARCH 2018 a beneficial, low forward voltage. HVdc line, superseding ASEA’s domi- Beginning in the 1980s, ABB and GTOs, produced in high-rating mod- nant mercury-valve technology [4]. Mitsubishi collaborated to develop the els by Japanese companies, and BPTs, During this period, PWM was intro- integrated gate-commutated thyris- both marketed in the late 1970s, were duced in ac motor control. Though tor (IGCT) to provide more superior per- the main devices for medium- and switching devices were not yet com- formance than that of the BPT, which high-power electronics applications mercially available, in suffered from low current in the early 1980s (Figure 4). 1976, Gyugyi and Pelly gain when designed for HVs Another switch that appeared during presented a comprehen- THYRISTOR- AND and required snubber cir- this period was the power metal–oxide– sive analysis of switching cuits. The IGCT, introduced TRANSISTOR- semiconductor field-effect transistor converters in various ap- in 1997 as a competitor of (MOSFET), which became commercially plications [18] that en- BASED DEVICES the IGBT, combines a MOS- available in 1976 and found preferential visioned the future of DOMINATED POWER FET operation and a bipo- use in the lower voltage applications PWM-driven converters. ELECTRONICS lar transistor that allows it (e.g., switching-mode power supplies) To impact higher volt- FROM THE 1960 s to be switched on and off because of the negligible control power age systems, a new class TO THE 1980 s. by the control electrode required by the MOS gate structure and of power devices based and results in a compact, of the fast-switching characteristics. on a combination of bipolar and MOS low-cost gate drive circuit and low- The latter feature was crucial in coping physics was needed. The concept of forward voltage. The first generation with increased carrier frequency neces- a composite device was introduced of the emitter turn-off (ETO) thyris- sitated by the development of pulse- for the first time by Yamagami et al. tor was developed by the Center for width modulation (PWM) control, but it in a Japanese patent regarding a wide Power Electronics, Virginia Tech Uni- also exhibited poor power handling ca- base p-n-p transistor driven by an versity, in 1999. This composite device pabilities at higher breakdown voltage, n-MOSFET in 1968 [19]. Subsequent consists of a thyristor that uses a MOS- compared with the BPT. Power MOS- discoveries led to the development of FET to turn on and turn off and has two FETs are still popular for low-voltage, the insulated-gate bipolar transistor gates: one normal gate for turn on and high-frequency applications. (IGBT), consisting of four alternating one with a series MOSFET for turn Thyristor- and transistor-based p-n-p-n layers controlled by a MOS off. In this way, it merges the benefits devices dominated power electronics gate structure that resulted in a high- of both the GTO and MOSFET with clear from the 1960s to the 1980s. They current power transistor with null advantages with respect to the IGCT. evolved into a strong alternative to the consumption at the control electrode. As of 2016, ratings as high as 8.5 kV/5 kA thyratron, drove the development of Early successful experiments were for SCRs, 6.5 kV/3.8 kA for IGCTs, electric power converters, and became reported by J.D. Plummer and B.W. 6.5 kV/0.75 kA for IGBTs, and 4.5 kV/4 kA the workhorse of the power industry. Schaft in 1978 and, independently, for ETOs have been achieved [22]. Their superior switching and control by Indian-born, American engineer If, on one hand, each generation of allowed advanced topologies to be Baliga [20] (b. 1948) in 1979. devices addressed the development introduced that were capable of ad- Improvements came in the follow- of power electronic controllers and dressing poor power factor and high ing years. Nakagawa et al. proposed their applications, the history of power harmonics issues. William McMur- the non-latch-up structure, marking electronics, on the other hand, has not ray (1926–2006) of GE-CRD proposed the birth of the practical IGBT in 1984 been as straightforward. Several power forced-commutation converter topolo- [21]. Its design and fabrication process devices, typically derived from tran- gies in 1961. After SCR-based, variable- technologies were derived from the sistors and thyristors, were proposed voltage constant-frequency drives for power MOSFET, with much effort made in the 1970s and 1980s, but most were induction motors were proposed in to implement trench technology. With not successful. This was the case with 1964, the McMurray inverter and the important contributions from numeri- the static induction thyristor, the static McMurray–Bedford inverter were es- cal simulations, more improvements to induction transistor, and the MOS- sential to the introduction of variable- IGBTs came in the following years. This controlled thyristor, despite intense re- frequency drives, which expanded the work led to an IGBT that, unlike GTOs, search efforts and costs. Nevertheless, operability of the induction motor and does not require snubber circuits or the general advancement has been of provided rated torque at any speed. complex driving circuits. Although paramount importance. Self-commu- Used in suitable converter topolo- early devices were marketed in 1982, tated devices (e.g., power MOSFETs, gies, e.g., a six-pulse Graetz bridge, successful commercial models didn’t BPTs, GTOs, IGBTs, and IGCTs) have SCRs were at the basis of second-gen- appear until 1984. IGBTs are now the largely replaced SCRs and allowed the eration HVdc technology. By 1979, top- most important devices for medium- introduction of advanced topologies, performing 3.2-kV SCRs were developed to-high power applications and are which, together with advanced control by AEG in Germany under GE’s license widely used in motor drives, uninter- techniques, resulted in power convert- for the 1.9-GW ±533-kV conversion sta- ruptible power supply, and renewable ers with superior performance. Notable tions of the Cabora Bassa–Johannesburg energy sources. examples are the popular voltage-fed

MARCH 2018 ■ IEEE INDUSTRIAL ELECTRONICS MAGAZINE 39 converter, the hysteresis-band sinu- The success of solid-state power [3] M. C. Duffy, Electric Railways 1880–1990. Insti- devices allowed power electronics to tution of Engineering and Technology, Lon- soidal current control (1979), and the don, United Kingdom, June 2003. matrix converter, invented by Marco steadily progress over the years, with [4] M. Guarnieri, “The alternating evolution of dc more workers working on power elec- power transmission,” IEEE Ind. Electron. Mag., Venturini in 1980 and based on inverse- vol. 7, no. 3, pp. 60–63, Sept. 2013. parallel ac switches. The neutral- tronics projects. By the early 1980s, [5] A. W. Hull, “Gas-filled thermionic valves,” power electronics achieved recogni- Trans. Amer. Inst. Elect. Eng., vol. 47, pp. 753– point clamper multilevel converter, 763, 1928. also introduced in 1980, is now very tion as a technology in its own right. [6] B. Bose, “Power electronics—Historical per- The Council of Power Electronics was spective and my experience,” IEEE Ind. App. popular. Modular multilevel convert- Mag., vol. 20, pp. 7–13, 81, Mar.–Apr. 2014. ers and step-wave approximations established in 1983 to advance and co- [7] L. O. Grondahl, “Theories of a new solid junc- ordinate work in the field carried out tion rectifier,” Science, vol. 64, no. 1656, pp. have mainly been used from the 1990s 306–308, Sept. 1926. to 2010. through the IEEE. Today, a substantial [8] M. Guarnieri, “Trailblazers in solid-state elec- Notable developments were the number of devices employing power tronics,” IEEE Ind. Electron. Mag., vol. 5, no. 4, pp. 46–47, Dec. 2011. PWM-selected harmonic elimination electronics are present in our facto- [9] L. O. Grondahl, “Twenty-five years of copper- invented by Fred Turnbull at GE-CRD in ries, offices, and homes. In the future, copper oxide rectifiers,” Trans. Amer. Inst. Elect. Eng., vol. 67, no. 1, pp. 403–410, 1948. 1963 and perfected by Patel and Hoft power electronics devices are expected [10] E. T. W. Alexanderson and E. L. Phillipi, “His- (1973), the sinusoidal PWM technique to change power grids and influence tory and development of the electronic power converter,” Trans. Elect. Eng., vol. 63, pp. 654– of Schonung and Stemmler (1964), and technologies, e.g., ultra-HVdc (UHVdc). 657, Sept. 1944. the space vector PWM technique (1982) Extended dc grids, flexible ac transmis- [11] M. Guarnieri, “70 years of getting transistor- ized,” IEEE Ind. Electron. Mag., vol. 11, no. 4, pp. perfected by Lipo (1995). With regard sion systems, and flexible substations 33–37, Dec. 2017. to motor drives, vector controls were will become a significant technologi- [12] N. Holonyak, “The silicon p-n-p-n switch and controlled rectifier (thyristor),” IEEE Trans. introduced in the late 1960s (Hasse, cal field of development. UHVdc is ex- Power Electron., vol. 16, no. 1, pp. 8–16, Jan. 1969) and early 1970s (Blaschke, 1972), pected to reach ratings of 1,100 kV and 2001. [13] E. L. Owen, “SCR is 50 years old,” IEEE Ind. and direct torque control, an advanced 20 GW on extensions of 3,000 km. For Appl. Mag., vol. 13, pp. 6 –10, Nov.– Dec. scalar control technique, was present- decarbonization programs, power elec- 2007. [14] M. Guarnieri, “The unreasonable accuracy of ed by Manfred Depenbrock in Germany tronics is closely involved with renew- Moore’s law,” IEEE Ind. Electron. Mag., vol. 10, (1984) and by Isao Takahashi in Japan able energy systems, photovoltaic and no. 1, pp. 40–43, Mar. 2016. [15] D. A. Hodges, “Darlington’s contributions to (1985). In smart and flexible ac grids, fuel cell power conversion, energy transistor circuit design,” IEEE Trans. Circuits static synchronous series compensa- storage, battery management, elec- Syst. I, Fundam. Theory Appl. vol. 46, no. 1, pp. 102–104, Jan. 1999. tors allow for fast control of equivalent tric and hybrid vehicles, smart grids, [16] M. Guarnieri, “More light on information,” IEEE capacitance or inductance, and thyris- and microgrids. Power electronics Ind. Electron. Mag., vol. 9, no. 4, pp. 58–61, Dec. 2015. tor-controlled series capacitors provide has emerged as a high-tech frontier in [17] N. Iawamuro and T. Laska, “IGBT history, fast capacitive reactance adjustment power engineering, promising to evolve state-of-the-art, and future prospects,” IEEE Trans. Electron. Devices, vol. 64, no. 3, pp. 741– in volt–ampere reactive and harmonic into future devices with superior per- 752, Mar. 2017. formance based on the use of large- [18] L. Gyugyi and B. R. Pelly, Static Power Frequen- compensators as well as improved cy Changers: Theory, Performance and Applica- load performance. Power electronics bandgap materials. tion. New York: Wiley, 1976. has found other major applications [19] K. Yamagami and Y. Akakiri, “Transistors,” Japanese Patent S4 721 739, June 19, 1968. in electrochemical processes, heat- Acknowledgment [20] B. J. Baliga, “Enhancement- and depletion- ing and lighting control, electronic mode vertical-channel M.O.S. gated thyris- I wish to thank Paolo Tenti of the Uni- tors,” Electron. Lett., vol. 15, no. 20, pp. 645– welding, industrial automation, and versity of Padova, Italy, for his valuable 647, Sept. 1979. high-frequency heating. Power elec- [21] A. Nakagawa, H. Ohashi, M. Kurata, H. Yama- assistance with writing this column. guchi, and K. Watanabe, “Non-latch-up 1200V tronic systems have largely benefitted 75A bipolar-mode MOSFET with large ASO,” in from the use of digital microelectron- Proc. IEEE Int. Electron Devices Meeting—Tech- References nical Digest, Dec. 1984, pp. 860–861. ics in their control circuits, including [1] M. Guarnieri, “The age of vacuum tubes: Early [22] Y. Qiu, C.Dai, R. Jin, “Impact of power elec- microcomputers/digital signal proces- devices and the rise of radio communica- tronic device development on power grids,” in tions,” IEEE Ind. Electron. Mag., vol. 6, no. 1, pp. Proc. 28th Int. Symp. Power Semiconductor De- sors and application-specified inte- 41–43, Mar. 2012. vices and ICs (ISPSD), Prague, Czech Republic, grated circuit/field-programmable gate [2] A. Moglestue, “From mercury arc to hybrid June 2016, pp. 9–14. breaker—100 years in power electronics,” ABB array chips. Review, no. 2, pp. 70–78, Nov. 2013.

40 IEEE INDUSTRIAL ELECTRONICS MAGAZINE ■ MARCH 2018