This article reviews the origin and costs of integrated circuits in the semiconductor industry. Sometimes, a significant integrated circuit achievement might not receive history’s recognition because it simply did not work well. The market might politely applaud a new development, but if another came along that was considerably cheaper, it would be seized hungrily. It would become the market leader and later in history's eyes, might appear to have been first. Jean Hoerni from Fairchild Semiconductor, in Mountain View, CA, developed a new transistor with all parts on one plane. He made connections through an insulating and protecting layer of oxide. Fairchild formed two design teams, one under co-founder Cordon Moore, to develop NPN, and one under co-founder Hoerni, who was working for Moore, to develop PNPs. Both projects were successful—Moore’s NPN, the 2N967, the first high-performance silicon transistor, and Hoerni’s PNP, the 2Nl132. But the NPN was easier to make at first, so Moore chose that one for IBM. Today, ICs house millions of transistors at costs of a fraction of a mil per transistor.
They didn't know they were making history. Many were young men in their 20s with enormous responsibility and power. They frequently didn't have enough knowledge to know what couldn't be done, so they did it.
As they look back on those days, many of these men find it difficult to believe that something so intensive on a day-to-day basis could be so important historically.
They made costly mistakes. The group in Northern California became the fathers and grandfathers of hundreds of companies, mostly semiconductor companies. They converted plum, pear, apricot, and cherry orchards into Silicon Valley.
Today, they look back in awe at what they wreaked. They know now that they created one of the greatest forces of the century—the integrated circuit. But they didn't keep careful records, so the history of their contributions lies largely in memories—memories often warped by time.
So we have a history of memories, not documents. There are small inconsistencies that stem, mainly, from human frailties. Because modesty is not distributed in equal measure to all people, some magnify their own roles or those of their companies. But most in this wondrous industry minimize their own early contributions. They award credit to others and assert that they were merely lucky to be in the right place at the right time.
Who, then, was responsible for the great achievements? Even then, it was hard to tell because most important developments were team efforts and weren’t carefully programmed and planned. Someone would get an idea for an integrated circuit and talk about it with friends or colleagues and, pretty soon, there would be a team. There might be a circuits man, an architect, a logic man, a process man, and a technician. Very few did anything alone.
Many ideas emerged from scientists and researchers, who would rush to present papers before others might. And yet, while many important developments came from the laboratories of semiconductor 9 manufacturers, many ideas also came from customers who disclosed their needs to salesmen. These salesmen would rush home to demand action. Several companies might develop similar products at the same time, but instead of rushing to publish papers about it, they would rush out to make sales. Again, it was difficult to know who did what, and when.
Who’s on First
George Rostky is Editor Emeritus of Electronic Engineering Times. Previously, he was editor-in-chief of EE Times, Electronic Design, and EEE magazine. As an electronics engineer, he worked on the last vacuum-tube computer at Underwood Computer.
So when we ask, for example, “Will the real developer of transistor-transistor logic stand up?” several individuals—honest individuals—are likely to rise. And when we ask who was first, we must define first. Do we mean who first developed the circuit on paper? Who first made a discrete-component model? Who first managed to get one good chip out of a wafer? Who was first into production with some “reasonable” yield and volume? Who was first to make a sale? Who was first to deliver in quantity? Who first presented a paper? Who first issued a press release or an advertisement? Who made the first “successful” product? Who made a chip with “reasonable” noise margin or output level? Who made the first 5-volt part? “First” has flexible meaning.
Sometimes, a significant integrated circuit achievement might not receive history’s recognition because it simply didn’t work well. Or the company never licked its yield problems so customers couldn’t get what they needed. Part A may have been first, but Part B may have become the historic device because people could use it and they could buy it.
The great technical achievements weren’t always awarded approval by the marketplace. The market might politely applaud a new development, but if another came along that was considerably cheaper, it would be seized hungrily. It would become the market leader and later, in history’s eyes, might appear to have been first.
Cheaper was frequently better. Cheaper was often an engineering achievement, but not always. Some companies—outstanding is Texas Instruments—might price circuits below cost because they believed in the experience-curve thesis put forth by the Boston Consulting Group.
The BCG concept was that costs dropped by a certain percentage every time accumulated volume doubled. So one could base present prices on anticipated costs to gain market dominance and future profitability.
Because those times were fiercely competitive and enormously invigorating, legends grew. Executives, scientists, engineers, salesmen, marketers, technicians, and secretaries would gather at watering holes like the Wagon Wheel in Silicon Valley. While quenching his thirst, one man would tell a story of some great success or disaster he had heard about. After a few transmissions, that story would become established fact. Its veracity would be certified when the tale emerged a week later at the Velvet Turtle. And it would pass down through history. People worked long hours, making wonderful history but lousy marriages.
There’s no real beginning to any history of technology because every great development grew from an earlier one. Most engineers and scientists readily grant that they stood on the shoulders of others. But we can start with Jack Kilby, an engineer at Texas Instruments who came up with something interesting in 1958.
Kilby started his engineering studies at the University of Illinois four months before the Japanese attacked Pearl Harbor and not long before he acquired the title of private. After the war, he returned to the University of Illinois and received his bachelor’s degree in electrical engineering in 1947, then joined Centralab in Milwaukee and went to the University of Wisconsin, where he earned a master’s degree in electrical engineering in 1950.
Developments in Dallas
At Centralab from 1947 to 1958, he worked on screened-and-fired circuits and, in May 1958, he joined Texas Instruments in Dallas. At the time, the Signal Corps Micro-Module program was pursuing a concept that individual circuit components—like resistors, capacitors, and inductors—could be mounted on small square ceramic wafers and interconnected by three riser wires along each edge. RCA was the prime contractor.
Kilby didn’t like the idea. The project eventually died, and his work with the integrated circuit had a great deal to do with its demise.
Kilby thought it would be nice, instead, to manufacture all circuit components in one operation with one material. Kilby joined TI just before vacation time— everybody else’s vacation time. So he was left to his own devices and started pursuing his idea that resistors, capacitors, and transistors could be combined in one block of semiconductor material. Resistors were to be diffused in silicon or to use the bulk effect in silicon, and capacitors were to use pn junctions.
He showed a sketch of the idea to his boss, Willis Adcock, on Adcock’s return from vacation, then assembled a unit of discrete silicon elements and showed it to Adcock on Aug. 28, 1958. On Sept. 12, he showed three silicon-integrated phase-shift oscillators to Mark Shepherd, TI’s president; Cecil Dotson, operations manager; and Adcock. A week later, he built an integrated flip-flop, a type of logic circuit.
Kilby’s success stemmed, in part, from the work of Adcock, who had hired him. Adcock got his B.S. in chemistry and math from Hobart College which, he likes to point out, was created in 1822 to bring education to American Indians. It was one of the first colleges to drop a knowledge of Latin and Greek as an entrance requirement, as most native Americans lacked that classical background. Hobart awarded a degree to the first woman doctor, Elizabeth Blackwell.
Adcock worked on the Manhattan Project at Oak Ridge, Tenn., developing the first atom bomb and acquiring a fascination with electronics. He earned his doctorate in chemistry at Brown University in 1948 and, he says, was almost awarded a master’s in the soldering iron.
Old Brown Ties
Adcock joined the research lab of Standard Oil of Indiana as a research chemist and worked on the conversion of natural gas to gasoline. When a school chum from Brown, Gordon Teal, invited him to work for Texas Instruments, one of the early companies to pay AT&T’s Western Electric $25,000 for a license to manufacture transistors, he was ready.
Gordon Teal had come from AT&T’s Bell Labs in late 1952 to head TI’s Central Research Laboratories.
When Adcock announced that he was leaving for Texas Instruments, his boss tried to dissuade him. “You’re out of your mind,” he said. “That company is smaller than our research lab. And there’s no future in that stuff. Just remember that people will always buy gas.”
Adcock joined TI in 1953, becoming one of the first to work in Gordon Teal’s lab. When Teal told him to grow some silicon, Adcock didn’t know how difficult it was, so he grew what may have been the first silicon crystal. The following year, TI introduced the world’s first silicon transistor, the ancestor of the transistors that Kilby used in his first integrated circuits.
Mark Shepherd, then vice president and general manager of what came to be known as the semiconductor components division, was charged with getting silicon transistors into full-scale production, which he did in 1954, four years before any competitor.
Within a few years, silicon became almost the exclusive semiconductor material for transistors and integrated circuits. But it was not an immediate success. It had to battle germanium in a marketplace where it was five times as costly and where competitors, who had not yet mastered silicon, fought against it bitterly in the field, while they wrestled to make it in the lab.
Adcock moved up the ladder. He was technical director of the consumer electronics division at one time and served on the corporate staff, retiring in 1986.
On March 24, 1959, TI showed off the fruits of Kilby’s work and announced what it called solid circuits (a name that didn’t stick) at a press conference at the Waldorf-Astoria Hotel in New York during the annual convention of the Institute of Radio Engineers. The IRE was the forerunner, through merger with the American Institute of Electrical Engineers in 1962, of the Institute of Electrical and Electronics Engineers—the IEEE.
TI demonstrated an integrated multivibrator with the discrete equivalent of two capacitors, eight resistors, and two diffused-base transistors; and a phase-shift oscillator, with the equivalent of three capacitors, five resistors, and one transistor.
Texas Instruments recognized the importance of what ultimately came to be known as integrated circuits. In a press release, the company wrote that “they are considered to approach the ultimate in miniaturizing complex electronic circuitry and components.” In years to come, it became apparent that they weren’t quite ultimate, as million-transistor ICs became commonplace.
That demonstration was the beginning of a dramatic history. In October 1961, the company announced its Series 51, with five different digital-circuit logic modules—flip-flops, counters, NOR gates, NAND gates, and exclusive ORs. They cost $95 in sample quantities and $65 each in quantities of 100.
The Series 52 linear amplifiers made their debut at the end of 1962. And, in 1964, TI introduced the transistor-transistor logic line that swept the marketplace, the Series 54.
Kilby’s invention had a serious drawback. It used mesa transistors, the prevalent construction of the day. The transistors protruded above the plane of the silicon. While Kilby used no discrete passive components, he did have to make a few interconnections above the surface of the silicon with lengths of gold wire. That technique would hardly have been suitable for hundreds or thousands or millions of components. He recognized that metallization over an oxide layer could be used for interconnections and said so in his patent, but no photo in his patent showed such an implementation.
The California Connection
Meanwhile, about 1,500 miles northwest of Dallas, Jean Hoerni of Fairchild Semiconductor, in Mountain View, Calif., developed a new transistor with all parts on one plane. He made connections through an insulating and protecting layer of oxide.
Not surprisingly, Hoerni’s device came to be known as a planar transistor, and it’s easy to argue that the integrated circuit of later days would not have been possible without Hoerni’s planar transistors. Kilby paid great tribute to Hoerni’s development, but he argued that ICs just had to move forward. Without Hoerni’s planar transistors, and Bob Noyce’s work that followed, another way would have been found.
Hoerni, one of eight founders of Fairchild Semiconductor Corp. in 1957, showed his planar transistors to some of his co-founders, including Bob Noyce.
Tom Bay, who was marketing manager at the time, suspects that Hoerni’s planar transistor was designed almost in retaliation for losing a contest involving transistors developed for IBM. As Bay remembers it, in early 1958, IBM gave Fairchild an order for 100 transistors, at $150 apiece, to be used as ferrite-core drivers in computer memories.
Fairchild formed two design teams, one under co-founder Gordon Moore, to develop npn’s, and one under co-founder Hoerni, who was working for Moore, to develop pnp’s. Both projects were successful—Moore’s npn, the 2N967, the first high-performance silicon transistor, and Hoerni’s pnp, the 2N1132. But the npn was easier to make at first, so Moore chose that one for IBM.
This was a huge order, and Fairchild hadn’t developed suitable shipping procedures—or containers. It charged Jay Last, another of the founders, with shipping IBM’s 100 $150 2N967 transistors. He shipped them in a Brillo carton.
Tom Bay feels that Hoerni, unhappy that his pnp didn’t prevail, decided to make the best transistor anybody had ever seen. He created the first planar transistor, an npn. (But Hoerni may have had the idea for a planar transistor in his notebook already.)
When Noyce saw Hoerni’s planar transistor, he got a bright idea. If we could make one planar transistor on a bar of silicon, we could make several, and if several transistors could be fabricated on one plane, he reasoned, we could interconnect a large number of transistors and other circuit elements through holes etched in an insulating layer with metal over the insulation. (But Noyce may have had some earlier ideas like this anyway.)
That thought and Noyce’s work that followed led to his and Kilby’s being accepted as the co-inventors of the integrated circuit. But that followed 10 years of legal disputes. In the summer of 1966, before a final legal opinion was issued by the Supreme Court, TI, Fairchild, and others agreed that Kilby and Noyce were co-inventors of the IC.
Noyce and his partner, Gordon Moore, earned the further distinction of founding two major semiconductor companies. Not only were they among the eight founders of Fairchild, but they also were among the founders of Intel, which today is the largest semiconductor company in the world.
Today, ICs house millions of transistors at costs of a fraction of a mil per transistor. Without the pioneering work of Kilby and Noyce, the devices would have been impossible.