The Transistor Revolution (Part 2)
Dr Bruce Taylor HB9ANY concludes his look at the history of the transistor
Dr Bruce Taylor HB9ANY concludes his look at the history of the transistor, relating how the invention impacted amateur radio and moving on to modern developments in integrated circuitry and microprocessors.
Wireless World first reported the invention of the transistor in October 1948. Using little more than a microscope, an Avometer and a pulse source for point-contact forming, enterprising amateurs were soon making their own transistors by replacing the single cat’s whisker of selected germanium diodes by two. By 1950 enthusiasts in the UK were building simple receivers using commercially-available Raytheon point-contact transistors that were primarily made for hearing aids.
Radio amateurs also started experimenting with transmitters as soon as suitable transistors became available at accessible prices. The February 1953 issue of QST described how RCA manager K2AH even made a 146MHz CW QSO with W2UK over 25 miles with a power input of 24mW to a single selected experimental point-contact germanium T165/6 transistor. In the UK, a miniature 3.5MHz transistor transmitter by G5CV aroused great interest when operated at the 1953 Amateur Radio Exhibition. A simple receiver using two GEC point-contact transistors was described in the January 1954 issue of Wireless World, while a topband (160m) transmitter by G3IEE using a Mullard OC50 featured in the RSGB Bulletin for March. At this time manufacturers such as Philips made ‘experimental transistors’ available to amateurs at low prices. These were devices that failed to meet the full professional specifications and would otherwise have been scrap.
In mid-1956 IDEA launched the Regency ATC-1, a simple two-transistor mobile converter that tuned the five HF amateur bands. One germanium npn transistor acted as oscillator/mixer, with output at 1230kHz for a broadcast receiver, and a second pnp one as Q-multiplier/BFO. It was sold for $79.50.
In September of that year W1OGU, a technician with Raytheon, achieved the first transatlantic QSO with a transistor transmitter, working OZ7BO in Copenhagen and G3AAM in Birmingham on 14MHz. His crystal-controlled rig used one 2N113 germanium alloy junction transistor as 7MHz oscillator, driving a second as doubler/output stage with 78mW input.
One of the first shortwave transistor receivers for radio amateurs to be made in the UK was the Heathkit GC-1U Mohican, which was sold in 1961 for £38 15/- (around £800 in today’s money). The 10-transistor kit set used three Mullard germanium AF115s as RF amplifier, local oscillator and mixer, with four OC45s for the BFO and three IF amplifiers coupled by 455kHz piezoelectric ‘transfilters’. After replacing the audio output stage by a higher power amplifier, I used the compact set as a mobile receiver for several years but its performance didn’t match that of good valve receivers of that period. Starting with the Model IM-30 in 1961, Heathkit also produced a series of rudimentary transistor testers and curve tracers that evolved with the device technology until the final Model IT-2232 was phased out in 1990.
It took several years for many of the traditional shortwave receiver manufacturers to change from valve to transistor designs and not all did so successfully. Eddystone’s first solid-state communications receiver was the S960; virtually a model S940 with 12 transistors in place of valves. Its performance was inferior to its parent and it was dropped two years later. The more compact EC10 that was introduced in 1963 was more successful and over 6,000 were made, followed by around 10,000 of the Mk2 version with an S-meter and fine-tuning control. In 1971 you could choose between the valved Model 830 receiver and the transistorised 1830 version with slightly better performance. The 830 was Eddystone’s last valve receiver and continued in production until January 1973.
National HRO receivers had been a long-time favourite with radio amateurs, and around 10,000 sets were used by the Y-Station service that supplied intercepted enemy wireless messages to Bletchley Park during WW2. In 1964 National introduced a transistor version of the receiver that retained the classic ‘PW’ gear drive but no longer had plug-in coils. This transitional set had 37 germanium transistors in sockets connected by conventional point-to-point wiring, just like its valve predecessors.
The first successful method of making a junction transistor involved creating the base layer by dropping a tiny p-type pellet into the n-type melt while drawing the crystal, and then converting it back to n-type. With this double-doping process, and also the improved rate-grown variant, it was difficult to accurately produce and connect to the very narrow base region required for high frequency performance. Problems also arose with the alternative alloy junction process developed by GE and RCA, since precise control of the alloying temperature and the thickness of the base layer was difficult. In a batch of 100 transistors, the gain could vary from 20 to 50dB.
Once again, chance played an essential role in the development of a hugely important new technology. While doping by gas diffusion had been used to introduce the donor impurities into germanium crystals, attempts to use the process at the very high temperatures required for processing silicon were initially unsuccessful because of damage to the wafer surface. But while Bell Labs chemist Carl Frosch was experimenting with diffusion, the hydrogen carrying the dopant impurities accidentally caught fire, causing water to be produced in his diffusion chamber. Frosch discovered that the fine green silicon dioxide layer that this formed on the surface of the wafer sealed it and protected it from damage. Unlike germanium oxides, silica is strong, inert and an excellent insulator. Initially considered a problem, the oxide turned out to be a key element of reliable solid-state electronics.
Photolithography techniques from the printing industry had already been used for the production of printed circuit boards, using a photosensitive resist that was exposed through an optical mask. The technology was readily adapted to chemically etch precisely-dimensioned windows in the oxide layer covering the silicon wafer, through which the n and p-type impurities could be diffused to make a double-diffused transistor and this important invention was announced by Bell Labs in June 1955. With the Cold War in full swing, ample military funds were available to develop these new transistors that could operate at high temperature and high frequency.
It had been expected for several years that the invention of the transistor merited a Nobel Prize, so it was no real surprise when Bardeen and Brattain were chosen to receive the 1956 award for Physics. It was less certain that Shockley would also share the coveted prize but in the end his name was included although the vote was not unanimous. Shockley tried to find out from the Swedish Royal Academy of Sciences who had opposed him but was told that he had no right to know.
All three men attended a celebration dinner in New York but the Bardeen and Brattain families travelled to Stockholm separately from the Shockleys and only shared the formal ceremonies there. For all three Laureates, it was the pinnacle of their careers. For Shockley, it was the prelude to his demise.
In view of his abrasive management style, Shockley was repeatedly passed over for promotion at Bell Labs. So, during 1955 he left to launch his own enterprise and, with the financial backing of Beckman Industries, Shockley Semiconductor Laboratory started up in April 1956. The facility was located in a new industrial park that was being set up in the Santa Clara Valley on land owned by Stanford University. It was the start of silicon in Silicon Valley, the extraordinary zone of innovation and technology that now accounts for one-third of all the venture capital investment in the US. Shockley planned that his company would “set the world on fire”. Indeed, it did, but not at all as he envisaged.
None of Shockley’s Bell Labs colleagues wanted to join him. But because of his reputation as a scientific genius, he had no difficulty recruiting very talented staff, including the chemist Gordon Moore (of Moore’s Law), Swiss physicist Jean Hoerni and Robert Noyce, who was employed at that time by the radio and TV manufacturer Philco but was seeking research-oriented work. In the UK, Ferranti began producing bipolar silicon diffused transistors in 1954. Shockley could also have directed his new team to ramp up production of silicon transistors to profit from a rapidly expanding US market. But instead, he insisted that they devote their efforts to developing a special four-layer pnpn diode that was difficult to fabricate reliably and of limited application. As a result, Shockley Semiconductor Lab never turned a profit.
Shockley proved an even worse manager in his own company than at Bell Labs. He constantly feuded with subordinates, insulted his staff and refused to listen to their advice. He became paranoid about trivial incidents, recorded all phone calls, and at one point even ordered all his employees to take a lie-detector test. Genius without ethics wasn’t a winning formula and by September 1958, after a mutiny of eight dissidents lead by Noyce, the company had fallen apart. Shockley did almost no further work of scientific value and went on to pursue a racist dysgenics agenda, lecturing about the inferiority of African-Americans and advocating the sterilisation of people with an IQ under 100. He died in 1989 without fortune but knowing that he had been the intellectual driving force behind a technological revolution and the catalyst of an industry that turned a quiet Californian valley into the most concentrated source of wealth on the planet.
The ‘traitorous eight’ that defected en masse from Shockley Lab went on to found Fairchild Semiconductor with a $1.38 million loan from Sherman Fairchild, a New York playboy-inventor who was the largest shareholder in IBM. Each of the eight had to make an initial investment of $500 to purchase 100 of the company’s shares, a sum that Noyce had to borrow from his grandmother since neither he nor his parents had any savings. He had no difficulty repaying her with interest, for less than two years later his stock was worth $300,000 and his personal fortune was eventually to grow to over $3 billion.
The timing was fortuitous. Alarmed by the surprise launch of Sputnik 1 in October 1957, the US Military suddenly had major requirements for lightweight miniature solid-state components for its catch-up programme and total annual purchases soon exceeded $100 million. Fairchild’s first contract was to manufacture silicon mesa transistors for IBM’s navigational computer for the XB-70 Valkyrie bomber. To make arrays of transistors, they cobbled together a step-and-repeat camera using ordinary 16mm movie camera lenses. The yield was reasonable but reliability proved an issue and technicians had to ‘tap test’ the transistors with pencil erasers to check whether tiny loose particles trapped in the hermetic package could contaminate the exposed p-n junctions.
To solve this problem, Hoerni built on the earlier work of Frosch to invent the ground-breaking planar technology that protected the sensitive junctions with oxide and allowed all the contacts to be made on one side of the silicon wafer. This passivation process reduced leakage currents and the configuration automatically created graded bases that reduce the charge carrier transit time, as in drift transistors. It also supported the aluminium-over-oxide interconnection scheme that was used by Noyce to make the first commercial monolithic silicon integrated circuits in 1960, after the problem of aligning successive photolithographic masks had been solved. Patent litigation between Fairchild and TI over the invention of the IC lasted many years but out of court the two companies entered a cross-licensing agreement with a net payment to Fairchild.
The individual components in the production versions of these ICs were isolated by reverse-biased p-n junctions, a technique that had been patented by Kurt Lehovec while at Sprague Electric, who paid him only $1 for the rights to this key invention. (A common practice at the time). Until it was overtaken by TI in 1967, Fairchild became the undisputed leader of the semiconductor industry and in due course, spin-off ‘Fairchildren’ spawned the creation of dozens of successful companies in Silicon Valley, such as AMD, Intersil, National Semiconductor, Signetics and Teledyne.
The silicon dioxide layer also proved the key to neutralising the troublesome surface states that had thwarted attempts to make an insulated-gate field-effect transistor. John Atalla finally succeeded in making a working FET at Bell Labs in 1959, paving the way for modern MOSFET integrated electronics. In the UK, Plessey was involved in integrated circuit development at an early date but the first ICs to be produced commercially in Europe were the Micronor devices introduced by Ferranti from 1962. These ICs had a wired OR function and were used mainly in computers developed by Marconi for the Admiralty. The second generation of this family used an enhanced DTL technology that was faster but less dense than TI 74 series TTL, which emerged as the most popular logic configuration after it appeared in 1966. Micronor II ICs powered the Argus 400 computers that were used in many industrial control applications, such as nuclear power stations and the Jodrell Bank radio telescope. In 1966, Plessey started the first European production of MOS ICs at Swindon.
In 1968 Moore and Noyce left Fairchild to found Integrated Electronics Corp, a name they abridged to the snappier Intel. Inspired by HP, they practised a non-hierarchical management style that encouraged the exchange of ideas and innovation and in 1971 their four-bit 10-micron 4004 CPU chip with 2300 MOS transistors launched the microcomputer age. At that time few people envisaged the wide-ranging consequence of the synergy of microprocessors and wireless, and in the same year a marketing study commissioned by AT&T reported that “there is no market for mobile phones at any price”!
The first European 16-bit microprocessor was the Ferranti 3.5-micron F100L, introduced in 1976. This mil-spec rad-hard chip, designed in Bracknell and fabricated in Manchester, had about 7000 components and 2m of aluminium track interconnects. It was orbited in OSCAR-9, although it wasn’t frequently enabled in that satellite because of its high power consumption. Two years later Intel introduced the commercial 16-bit 3-micron 8086 microprocessor and the x86 family architecture rapidly became ubiquitous in desktop PCs and laptops. In due course this led to the next iteration in wireless technology. Although the concept was much older, in 1995 Stephen Blust coined the term Software Defined Radio and soon SDR hardware and software allowed any PC to be turned into a high-performance transceiver capable of handling a wide range of communications modes.
Today there are 19 silicon manufacturing facilities in the UK. A single modern FPGA can have as many as 50 billion microscopic transistors and (including memory) a high-end smartphone contains a total of around one trillion, or ten times more than the number of neurons in the human brain. To visualise the total number of transistors that have been manufactured, we can take as a metric the number of cells in the human body – about 100 trillion. Then a Fermi estimate of the total number of transistors made since 1947 exceeds the number of human cells in the entire population of the UK and within a few years the number of transistors produced may equal the number of human cells on the planet. Even visionaries like Bill Hewlett and Dave Packard would have been astounded.
In the years immediately following its invention, the transistor was conceived as a replacement for the thermionic valve. Indeed, it was, and a very good one but it also triggered revolutionary growth in communications technology and created a world that has become interconnected as never before. Apart from wireless itself, perhaps no other modern invention has had a greater influence on almost every aspect of our daily lives.
This article was featured in the December 2018 issue of Practical Wireless