In part 1 of this three-part series, we considered the birth of the transistor, how William Shockley Jr. ended up in Silicon Valley, the origins of Fairchild Semiconductor, how the pioneering startup was funded and what eventually happened to Shockley. In part 2, we look at the evolution of planar technology, the “family tree” of semiconductor startups that evolved from Fairchild (the “Fairchildren”), including Intel, and the competition with Texas Instruments.
Setting up shop on 844 East Charleston Road, on the border between Mountain View and Palo Alto, founded in intrigue, Fairchild recorded a long history of innovation, having produced some of the most significant technologies of the second half of the twentieth century. It quickly grew to be among the top semiconductor industry leaders, spurred on by the successful development of the silicon planar transistor.
Transistors, however, were already presenting a new challenge, dubbed the “tyranny of numbers”. If you wanted to make a simple flip-flop, it needed four transistors. About 10 wires were needed to connect them. Interconnecting two flip-flops required not only twice the number of transistors and wires but also four or five additional wires to connect the two devices. So, four transistors needed 10 wires, eight needed 25, 16 needed 60 to 70 wires. In other words, as the transistor count increased linearly, the number of connections grew exponentially, where the exponential was greater than one but less than two.
While transistors were relatively easy to mass produce, connections were much more difficult since wires had to be soldered together by hand and took up a lot of space. The industry’s desire to build bigger and more complex systems was stymied by the difficulty in wiring everything together. To this point, few had paid much attention to wiring, but connections would soon become a potential show-stopper, driving the need for the integrated circuit.
In 1958, Jack Kilby of rival semiconductor company Texas Instruments demonstrated the ability integrate a pair of transistors on a semiconductor substrate. Kilby’s transistors were wire-bonded, however, leaving unresolved the connection problem. That problem was solved by Bob Noyce with the help of Jean Hoerni (who provided the technique) and Jay Last (who eventually made it work).
Part 1: Founders, Legends, Legacy
Part 3: Startup Fever and Venture Capital
Hoerni had been working on a fix for reducing transistor defects. Defects were traced to unprotected transistor surfaces inside a package, allowing particles to contaminate and degrade the device over time. Hoerni’s solution was to protect the transistor surface with a passivation, or protection, layer of silicon dioxide (SiO2), grown or deposited on top of the structure. Rather than depositing the emitter and base regions on top of the substrate, as with the current Mesa process, Hoerni saw another way: If the surface was completely covered with SiO2, the emitter and base areas could then be selectively diffused. The net result was a much flatter surface, allowing for greater automation during production.
Planar technology, announced in January 1959, would become the second most important invention in the history of microelectronics — after the invention of the transistor — laying the foundation for future integrated circuits. At the time, the advance went virtually unnoticed, with the key exception of Noyce, who recognized that a glass layer was an insulator, providing a means for connecting wires laid on top and patterned like a printed-circuit board.
Noyce filed his patent in April 1959, triggering a legal battle between Texas Instruments and Fairchild (Kilby and Noyce remained friends, with high regard and respect for each other). Texas Instruments claimed that Kilby’s patent claim– “electrically conducting material such as gold laid down on the insulating material to make the necessary connections”–was a pre-existing description of Noyce’s patent claims, and that Kilby had only used wire bonds as the quickest way to a prototype. Had this assertion been upheld, Noyce’s later-dated patent would have been declared invalid.
Texas Instruments lost the argument and both patents were declared valid — and a cross-licensing agreement was reached between the two firms.
Kilby, by nature, was a very humble person and, even though his patent pre-dated Noyce’s, he generously announced that both he and Noyce jointly invented the integrated circuit, contrary to the position of Texas Instruments’ management.
In 1959, Sherman Fairchild exercised his right to purchase the founding members’ shares, an event that turned former entrepreneurs and partners into ordinary employees, thereby undermining the company’s team spirit and sowing the seeds of future friction.
Isolation was another big problem yet to be solved before integrated circuits could become a commercial reality. The problem was how to stop adjacent transistors interference. Noyce delegated this thorny problem to Jay Last, who was running the R&D group. It was no easy task, taking some 18 months before the first working device was produced on September 27, 1960.
Trouble at Fairchild
Development also met with strong internal resistance. Tom Bay, Fairchild’s vice president of marketing, accused Last of squandering resources. In November 1960, Bay demanded termination of the project, with the money saved to instead be allocated to transistor development. Moore refused to help, and Noyce declined to discuss the matter, leaving Last to fight the battle on his own. The conflict flared up barely a month after Fairchild announced the transition of its transistor production from mesa to planar technologies. Moore refused to credit this achievement to Hoerni, fanning the flames of the already developing tensions between the eight founding partners.
Jay Last continued to develop six more parts, but ongoing conflicts were the last straw. Flush with their planar and isolation process success, Last and Hoerni left Fairchild on January 31, 1961, to launch Amelco in Mountain View, with financing from Teledyne Corp. arranged by Arthur Rock. Their plan was to develop ICs to support Teledyne’s military business. Eugene Kleiner and Sheldon Roberts joined the pair a few weeks later. With this high-level defection, the eight founding members had been split into two groups.
Fairchild announced the world’s first standard logic family of ICs, direct-coupled transistor logic, in March 1961. The device was based on Hoerni and Last’s resistor-transistor logic (RTL) planar process under the µLogic trademark. Among these, the µL903 3-input NOR gate, became the basic building block of the Project Apollo guidance computer. Designed by MIT and built by Raytheon, the lunar navigation computer required 5,000 devices and was the first major IC application. Miniaturization for space applications was driving early scaling.
Fairchild’s lead, however, was short-lived. David Allison, Lionel Kattner and others also left at around the same time as Hoerni and Last to launch Signetics (Signal Network Electronics). One year later, in 1962, the firm announced a much-improved, second-generation logic family, the SE100 Series diode-transistor logic (DTL). Fairchild quickly responded with its own DTL family, the 930 series, undercutting Signetics and rendering them unable to compete against Fairchild’s marketing juggernaut.
NE555 Timer: Most Popular IC Ever?
Signetics’ most famous legacy part was the NE555 timer. Designed in 1971, the 555, along with the ubiquitous TTL 7400 Quad 2-input NAND Gate, was probably the most popular IC ever sold. Signetics was acquired by Philips in 1975.
Early ICs were housed mainly in either TO-5 or TO-18 adapted metal can transistor packages. These worked fine for three-lead devices, but scaling them to provide more connections proved to be limiting, given they can could only be made so large and the radial leads could only be packed so tight. Ten leads were about the practical limit, and would not support the more complicated ICs in the pipeline. It fell to Fairchild’s Don Forbes, Rex Rice and Bryant “Buck” Rogers to provide a fix in 1964, via the invention of the now-familiar dual in-line package, the tiny oblong “millipedes” that would crawl across circuit boards for the next 40 years.
The packaging innovation stemmed from a ceramic flatpack design devised in 1962 by Yung Tao, a Texas Instruments engineer, as an industry standard for surface-mount ICs for the U.S. military. The concept was adapted for through-hole, rather than surface mounting, with an eye toward ease of handling for electronics manufacturers and easier PCB layout design for delivering power to the ever-increasing number of ICs, routing their signals around the board. Another consideration was cost, given the growing consumer IC market. The 0.1″ (2.54 mm) package pin spacing left plenty of room for PCB tracks to be routed between pins, and the 0.3″ (7.62 mm) spacing between rows of pins left room for other tracks.
Fairchild launched its dual in-line package in 1965, originally in ceramic, but it took off with a vengeance when Texas Instruments introduced a plastic resin version, driving the unit cost down dramatically. As a result of great design, low cost, and support for increasingly complex ICs, the plastic dual in-line package became the industry standard, with its basic 14-pin design extended to support more leads, up to 64 pins in a 0.6”-wide form factor, and more complex ICs. It was eventually surpassed by second-generation surface mount devices in the late 2000s as chip complexity and pin count requirements surpassed the capability of dual-in-line packages.
With as many as 15,000 die now on a single wafer, assembly and test now outweighed wafer fab costs. Hence, the need to reduce labor costs as a matter of survival. After some early failed ventures, for example in Shiprock, N.M., at a Navajo reservation, along with early attempts at automation, offshoring test and assembly to Asia ultimately proved successful, at least in the short term. Bob Noyce, an investor in a small radio company in Hong Kong, suggested to Charlie Sporck that he and Jerry Levine scout the region.
They were attracted by the low labor cost, non-unionized facilities, western-educated technicians, good engineering schools, and tax incentives and other government subsidies. In 1963, Fairchild set up the industry’s first Far East assembly and test operation in a former shoe factory on the Kowloon side of Hong Kong. Other semiconductor manufacturers subsequently followed Fairchild to the Far East, primarily Malaysia.