Patents and innovation in the early semiconductor industry.
In the 1940s, Bell Labs did a lot of the pioneering work in the semiconductor world. Just before WWII, they discovered the p-n junction and understood the mechanism behind semiconductor doping. After the war, they got back to work, invented the transistor, then the junction transistor, figured out how to grow large semiconductor crystal, how to control doping, and how to purify semiconductor materials.
At that time, they held essentially all the patents around semiconductors.
Then they did something that now seems unthinkable: they realized that letting other companies build on top of those ideas would benefit their own company back. In 1952, they had licensed the relevant patent portfolio to 40 different companies for $25,000 per company (about $200,000 in today’s dollars), and they even organized a two-week training session for those companies to be able to get started quickly.
The rest is history, literally.
In 1952, Germanium Products Corporation provided the transistor that powered the first hybrid transistor/tube hearing aids. In 1953, Raytheon provided the transistors that powered the first fully-transistorized hearing aids. In only 2 years, vacuum tubes essentially disappeared from hearing aids. In 1954, Texas Instruments manufactured the transistors for the first all-transistor radio receiver, and a year later Sony was making the transistors for their own such receiver. In 1953, Standard Telephones and Cables provided the transistors that powered the first transistor-based computer. Bell Labs themselves got a 1MHz transistor computer running in 1954, and Philco’s transistors were used in a 5MHz transistor computer in 1955. In 1954 Dupont was in the business of refining silicon to the highest purity, and Texas Instruments used that material to commercialize the first silicon transistors.
Not resting on their laurels, Bell Labs kept their own research work. They also made progress with silicon in 1952, and invented the diffusion process with which modern chips are made. They use their knowledge to manufacture the first solar cells, and the first high-speed transistors.
Then, Bell Labs did it again: in 1956, just like they had done in 1952, they shared their discoveries with other companies, to help speed up innovation.
Bell labs (still them!) had discovered the possibility of making silicon oxide masks with tiny holes to actually constrain the diffusion effects to small areas on the silicon wafers, and they published their results. They then figured out how to create those holes with a combination of optics and chemistry instead of mechanical action, paving the way for the photolithography method with which silicon chips are still made today.
In all that openness, the US Army figured out how to deposit metal traces on the silicon with a similar photolithography technique. Fairchild then put all those techniques together: in 1958 they invented the step-and-repeat process that allows to make many identical chips from a single wafer, which was automated further when the GCA Corporation created machines that did that in 1961. In 1959 Fairchild invented the planar process, and licensed the relevant patents to other companies. All of those are still used today.
Still in the late 50s, many projects attempted to put multiple elements side-by-side on a single semiconductor device. RCA did some work in that domain. So did IBM, the MIT, and the Ministry of International Trade and Industry in Japan. Bell Labs looking in that area as well, and eventually Texas Instruments (them again) managed to create resistors and capacitors on a semiconductor wafer and to commercialize the result.
Fairchild (them again) put it all together and created in 1960 the first actual integrated circuit with different types of electronic components created on the same silicon wafer and linked together by metal traces deposited on the silicon, i.e. what we are still using today. To reach that result, they used a method to turn semiconducting silicon into an electrical insulator that had been invented by the Sprague Electric Company which had licensed its patent to other companies.
In just 8 years, thanks to all those licensing agreements, the industry had moved from being able to manufacture individual transistors to fully integrated circuits.
In 1959, Bell Labs figured out how to create a MOSFET transistor which had been theorized more than 30 years earlier but never realized. They didn’t quite know what to do with it, but Fairchild and RCA did. Those are the transistors used in our modern CMOS chips. In 1960, Bell Labs figured out epitaxial deposition which made silicon transistors much faster, and they let lots of other companies used the results of that research: Fairchild, Texas Instruments, but also Rheem, Sylvania, Motorola.
After using transistors from the General Transistor Corporation while working at Univac, Seymour Cray’s Control Data Corporation had them created faster germanium transistors, then financed Fairchild research to create even faster silicon transistors.
In the early 60s, Texas Instruments, Fairchild and Signetics started to sell to their competitors the test equipment that they were using themselves to test their semiconductor chips.
Starting in 1961, integrated circuits started to be used by NASA to build computers for space applications where size mattered enough to justify the high price, including the Apollo computers designed in 1962.
In 1963, Fairchild presented their CMOS invention in a conference paper. Around the same time, standard logic families started to appear, and in just 3 years the TTL gate design invented by Pacific Semiconductor was used in Texas Instruments’ 5400, which later evolved into the ubiquitous 7400 family that set the standard for TTL logic levels that many simple chips use today, using a chip package that had been designed by Fairchild.
After the pioneering work by the US Army and RCA, IBM spread in 1964 the usage of hybrid microcircuits that use multiple semiconductor chips in a single package, finding a middle ground between the level of complexity that can be reached with a printed circuit board and the small size of semiconductor chips. Such devices are still widely used to miniaturize systems that can’t be integrated onto a single chip.
In 1964 as well, General Microelectronics sold the first MOS Integrated circuits, and that work was continued by AMI Semiconductors and General Instrument, while Rockwell, Mostek and TI continued the work toward higher integration of MOS circuits. From 120 transistors in 1964, we’re now producing chips with more than 6 billion transistors. Gordon Moore wrote a first version of his eponymous law in 1965.
In 1965, the semiconductor ROMs were invented, by Sylvania (TTL) and General Microelectronics (MOS), and lots of companies ended up manufacturing variants, including Fairchild, TI, Motorola, Signetics, AMI, Rockwell, but also Intel, AMD, National Semiconductors and Electronic Arrays
RAM followed shortly afterward, e.g. as a Fairchild design used by IBM, or as a cooperation between Scientific Data Systems and Signetics. Components Division also designed a RAM that was used by IBM. Transitron designed a RAM chip that was later manufactured by Fairchild, TI and Intel.
Computer-aided design of chips originated at IBM in 1964, got quickly used by Texas Instruments and Fairchild. In turn, that allowed to create blank gate arrays that could be connected into custom ICs in a matter of days rather than months, which was then commercialized by Ferranti, Interdesign, Fairchild, International Microcircuits, Motorola, VLSI Technology and LSI Logic.
Performance improved in the TTL world since the very beginning since Texas Instruments has the idea to use a Schottky-Barrier Diode in 1964 to improve speed. Fairchild refined the design in 1967, and Intel used it in their very first product in 1969. This was also used in the 74S and later 74LS series from Texas Instruments, with similar products later manufactured by AMD, Fairchild, Motorola, National Semiconductors and Signetics.
Bell Labs investigated in 1967 using polycrystalline silicon to replace metal gates in MOS transistors, making them faster and smaller, in a way that had first been described by General Microelectronics. Fairchild then used that idea in commercial chips in 1968. A lot of additional work in that area was done by Intel, from which they created high-density RAMs, dynamic RAMs, EPROMs, which wouldn’t have been practical with metal gates. This allowed MOS chips to compete with bipolar ones in terms of speed.
RAM sizes improved a lot. Fairchild did some of the early work, which General Microelectronics improved on, and Fairchild in turn improved on General Microelectronics’ designs, which were then re-used by Four Phase Systems and Advanced Memory Systems. Honeywell got in the loop with Intel, Cogar got involved, and research by IBM and Siemens allowed Mostek to make high-density RAMs.
ROM also improved. Radiation Inc invented PROM in 1970, and Data I/O, Spectrum Dynamics, Harris, Monolithic Memories, Motorola and Signetics made such devices. On top of some 1967 research at Bell Labs and Sperry Rand, Intel invented the EPROM in 1971, and later then EEPROM in 1978. In 1981, Seeq would then invent an EEPROM that could be reprogrammed in-place, and Toshiba invented flash memory in 1984 which got first manufactured by Intel in 1988. In parallel, the work on PROMs paved the way for Signetics for create the first programmable logic array, which was built upon by Monolithic Memories, AMD, National Semiconductors and Texas Instruments. Cypress, Lattice and Altera then did similar things with EPROMs. As a next step in that evolution Xilinx, Actel and Quicklogic introduced FPGAs in 1984, 1985 and 1988 respectively, which essentially eliminated the need for custom circuits.
Eventually, all those chips became more and more advanced. Fairchild invented partial ALUs, then Four Phase Systems invented CPU slices in 1969, then Garrett AiResearch had a whole processor in a handful of chips in 1970, and finally Intel’s 4004 in 1971 was the first single-chip microprocessor. Intel’s breakthrough led to a lot of other microprocessors, including Motorola’s 6800 in 1974, MOS Technology’s 6502 in 1975, and Zilog’s Z80 in 1976. TI started work in 1971 to add RAM, ROM and control logic on the same die to create the microcontroller, which shipped in 1974, and which is one ancestor of our current single-chip systems along with the digital watch, to which had contributed RCA, Microma, Seiko, Intel, Texas Intruments, Timex. Motorola and Hitachi later re-used the microcontroller idea for high-performance microcontrollers while General Instrument took the idea toward the cheaper end of the range. In 1979, Bell Labs added complex math capabilities to the usual microcontrollers, creating the first DSPs, and that idea was later re-used by NEC, Texas Instruments, Motorola and Analog Devices.
Thanks to Bell Labs’ visionary licensing, the whole semiconductor industry moved incredibly fast. In just 20 years, the normal duration of a patent, the industry went from being just able to produce individual junction transistors all the way to mass-producing MOS microprocessors. If Bell Labs had instead decided to keep those ideas to themselves, creating a culture where the norm would involve secrecy and legal action, the entire world of computers and consumer electronics would have evolved much more slowly, in a way that might have wasted 20 to 30 years of technology evolution.
Looking back at that period of intense innovation that happened between 40 and 60 years ago, I think, maybe naively, that those were simpler times. Back then, when a company built upon your ideas to create a better product, your response was additional motivation to build upon their ideas to create an even better product. Right now, it seems that it’s not possible to do a single thing any more without involving a dozen lawyers. Between copyright issues, patent issues, trademark issues, laws, regulations, creators are paralyzed at every step, innovation slows down, and users are unhappy because of it. We’ve reached the point where some entire products become obsolete before they can ship at all because of all the legal issues that surround and stifle technology.
A Pun Involving Travel 