Electronics Guide

The Dawn of the Integrated Circuit Age

The period from 1960 to 1975 marked one of the most transformative eras in electronics history. It built on two inventions that arrived just before it: the transistor, demonstrated at Bell Telephone Laboratories in 1947, and the integrated circuit, which Jack Kilby demonstrated at Texas Instruments in September 1958 and Robert Noyce reconceived in planar form at Fairchild Semiconductor a few months later. What the following fifteen years supplied was everything those inventions still lacked, namely a manufacturable process, a catalog of standard parts, a cost curve steep enough to reach ordinary products, and an industry capable of producing chips by the million.

At the beginning of this era, even sophisticated electronic systems relied on individual transistors, resistors, and capacitors wired together on circuit boards, and the first integrated circuits were laboratory curiosities priced beyond any commercial use. By 1975, single silicon chips carried several thousand transistors, semiconductor memory had begun to displace magnetic cores, a complete central processing unit fit on one die, and the semiconductor industry had grown from a handful of research-oriented firms into a global manufacturing business with its own supply chain, equipment vendors, and offshore assembly plants.

The technical achievements of this period were remarkable in both scope and pace. Engineers learned to fabricate increasingly complex circuits on silicon wafers, developed processing techniques for creating ever-smaller features, and established the infrastructure needed to produce semiconductor devices in enormous quantities at falling prices. In 1965, Gordon Moore of Fairchild Semiconductor observed in Electronics magazine that the number of components on an integrated circuit had been roughly doubling each year and predicted that the trend would continue for another decade, an observation later named Moore's Law. In 1975, drawing on ten more years of data, he revised the projected pace to a doubling roughly every two years, the figure most often cited today.

Articles in This Category

The Process Innovations Behind the Chip

The semiconductor revolution rested on a chain of interconnected process innovations, and the first of them was the planar process, developed by Jean Hoerni at Fairchild Semiconductor in 1959. Earlier mesa transistors left their junctions exposed at the edge of an etched plateau, where contamination and moisture degraded them. Hoerni left a thermally grown layer of silicon dioxide in place over the surface, passivating the junctions and using photolithographically opened windows in that same oxide to control where dopants entered the silicon. The result was a device that was dramatically more reliable and, equally important, flat. Robert Noyce recognized that a flat, insulated surface could carry evaporated aluminum wiring, and his monolithic integrated circuit replaced hand-soldered connections between components with a metal pattern deposited over the oxide in the same photolithographic step sequence that formed the devices themselves.

Kilby's germanium demonstration of September 1958 proved that an entire circuit could be built in one piece of semiconductor material; Noyce's planar version, sketched in his laboratory notebook in January 1959, proved that such circuits could be manufactured in quantity. The resulting patent conflict between Texas Instruments and Fairchild ran for years before the two companies cross-licensed in 1966, and the industry has since credited both men. Kilby shared the Nobel Prize in Physics in 2000; Noyce had died in 1990, and the prize is not awarded posthumously.

Patterning improved alongside the devices. Layouts were cut by hand at large scale in Rubylith film, photographically reduced, and then repeated across a mask plate by step-and-repeat cameras. For most of the era the mask was pressed directly against the wafer in contact printing, which resolved fine features but damaged both surfaces and scattered defects across the wafer. Perkin-Elmer's Micralign, introduced in 1973, projected the mask image onto the wafer through an optical system instead of touching it, which extended mask life from a handful of exposures to effectively unlimited use and raised yields substantially. Projection lithography became the foundation of every subsequent generation of chipmaking.

Doping methods advanced as well. Thermal diffusion in furnaces, adequate for the first bipolar circuits, gave way in the early 1970s to ion implantation, which accelerated dopant ions into the wafer and allowed engineers to set dose and depth precisely and reproducibly. That control mattered most for MOS circuits, whose threshold voltage depends directly on the channel doping. Wafers grew during the period from roughly one inch in diameter to two and three inches, spreading the fixed cost of each processing step across more chips. The trade-offs of these steps are treated in detail in Semiconductor Wafer Fabrication.

Yield governed the economics of all of it. A single particle or mask defect anywhere on a die ruins that die, so the fraction of good chips per wafer falls steeply as die area grows. Cleanrooms, filtered air, deionized water, particle discipline, and statistical process control were not refinements but requirements, and most of the industry's cost reduction during this era came from raising yields rather than from any change visible in a data sheet.

Standard Parts: Logic Families and Linear Circuits

Integrated circuits became useful to ordinary engineers when they arrived as catalog products rather than custom designs. Fairchild introduced the first commercial family, Micrologic, in 1961 using resistor-transistor logic; its gates later flew in the Apollo Guidance Computer. Diode-transistor logic followed with better noise margins, and then transistor-transistor logic settled the question. Texas Instruments introduced the 5400 series in a military-grade metal flat package in October 1964 and the commercial plastic dual-in-line 7400 series in the third quarter of 1966. The 7400 family became the standard building block of digital design for the next two decades. Motorola's emitter-coupled logic, introduced in 1962, offered the highest speed available at the cost of high power consumption and became the choice for mainframes and instrumentation.

A fourth family pointed toward the future. RCA introduced the CD4000 series in 1968 under the name COS/MOS, using complementary MOS transistors that draw almost no current except when switching. The parts were slower than TTL and easily damaged by static discharge, but they operated from a supply anywhere between roughly 3 and 18 volts and consumed so little power that they made battery-operated digital equipment practical. Electronic watches, portable instruments, and later almost all digital logic followed that path.

Analog design changed in the same way. Bob Widlar at Fairchild designed the µA702, the first monolithic operational amplifier, in 1963, and the far more successful µA709 in 1965. The µA709 was fast and inexpensive but required external frequency compensation and had no output protection, so careless use destroyed it. The µA741, introduced by Fairchild in 1968, integrated the compensation capacitor on the chip and added short-circuit protection, making it nearly impossible to misuse; it became the archetypal general-purpose op-amp and remains in production. Monolithic voltage regulators, comparators, data converters, and the Signetics NE555 timer of 1972 followed the same pattern. Circuits that had previously demanded matched discrete transistors, careful thermal layout, and individual trimming turned into part numbers ordered from a distributor. The device categories that emerged from this period are surveyed in Integrated Circuits.

The catalog itself reshaped engineering practice. Second sourcing, in which competing manufacturers produced pin-compatible and function-compatible versions of a popular part, protected buyers against supply failure and turned successful part numbers into de facto industry standards. Complexity climbed through recognized tiers: small-scale integration offered a few gates per package, medium-scale integration delivered counters, registers, decoders, and multiplexers as single functions, and large-scale integration brought whole subsystems onto one die by the end of the era.

MOS Technology, Memory, and the Microprocessor

Metal-oxide-semiconductor technology emerged during the 1960s as an alternative to the bipolar transistors used in the first integrated circuits. Mohamed Atalla and Dawon Kahng demonstrated a working MOS field-effect transistor at Bell Telephone Laboratories in 1960, but contamination of the gate oxide by mobile sodium ions made early devices unstable, and manufacturable MOS processes did not arrive until the middle of the decade. Once they did, the advantages were decisive for certain applications. MOS circuits were slower than bipolar circuits, but they required fewer masking steps, drew far less power, and occupied much less area, in part because a MOS transistor can serve as its own load device. Higher density and lower power made MOS the natural technology for memory arrays and for putting a processor on a single chip.

The decisive process improvement came in 1968, when Federico Faggin and Tom Klein at Fairchild developed the self-aligned silicon-gate structure. Replacing the aluminum gate with polycrystalline silicon deposited before the source and drain regions were formed allowed the gate to mask its own doping step, so the gate aligned itself to the channel automatically. That eliminated the overlap capacitance caused by mask misalignment, improved speed and reliability, and provided a second interconnection layer. Silicon-gate MOS became the dominant technique for building dense logic and memory, and every major MOS milestone that followed depended on it.

Semiconductor memory then attacked the last stronghold of an older technology. Magnetic core memory was nonvolatile and well understood, but it was bulky and largely hand-assembled. Intel's 1101, introduced in 1969, was an early commercial MOS static RAM holding 256 bits. The 1103 dynamic RAM, introduced in October 1970 with 1,024 bits, stored each bit as charge on a capacitor refreshed periodically, which cut the cell to a fraction of the area a static cell required. It was the first semiconductor memory to reach cost parity with core, and by 1972 it was the best-selling semiconductor device in the world. Dov Frohman's erasable programmable read-only memory, the Intel 1702 of 1971, completed the set by letting engineers erase firmware with ultraviolet light and reprogram it, instead of committing a program to an expensive mask. Contemporary memory technology is examined in Memory Devices.

The microprocessor was the culminating achievement of the period. The Intel 4004, released in November 1971, originated in a contract with the Japanese calculator maker Busicom and placed a complete 4-bit central processing unit of roughly 2,300 transistors on one silicon-gate MOS die. The 8-bit 8008 followed in 1972, and the substantially more capable 8080 arrived in 1974, the same year that Motorola introduced the 6800 and Texas Instruments introduced the TMS1000, which combined processor, memory, and input-output on a single inexpensive chip as a microcontroller. Priority claims remain contested, in part because comparable work stayed hidden: the Garrett AiResearch MP944 chipset built for the central air data computer of the F-14 fighter was classified for decades. The larger point is not who was first but what became possible once a processor was a component that any engineer could order by part number.

Moore's Law and the Economics of Scaling

Moore's 1965 article was as much an argument about cost as about complexity. He noted that for any given process there is a die complexity that minimizes the cost per component: below it, packaging and testing dominate; above it, falling yield dominates. His central claim was that this optimum point moved steadily upward, so that the cheapest way to build a given function kept getting more integrated. Carver Mead later gave the observation its familiar name, and Moore's 1975 revision to a doubling every two years reflected the shift from the early period, when much of the gain came from simply learning to make bigger and better dice, to a slower and more sustainable combination of finer lithography, larger chips, and cleverer circuit design.

What kept the projection true was not physics but a reinforcing economic loop. Each process generation lowered the cost per function, which opened markets that could not previously afford electronics, and the resulting volume paid for the equipment and process development behind the next generation. Cumulative production experience drove costs down along a predictable learning curve, and manufacturers routinely priced new parts below their current cost in the expectation of reaching that cost later, a practice that accelerated adoption and punished competitors who could not match the volume.

The consequence for design was permanent. As the cost of a transistor fell relative to the cost of a package pin, a printed circuit board trace, or an hour of assembly labor, engineers learned to spend transistors freely in order to save everything else. That inversion, more than any single device, is what separates electronic design after this era from electronic design before it.

Industry Transformation

This era saw the region south of San Francisco emerge as the global center of semiconductor innovation. Fairchild Semiconductor, founded in 1957 by eight engineers who had left William Shockley's laboratory, served as the proving ground for the planar process and the monolithic integrated circuit. It also became a wellspring of new companies as its people left to found ventures of their own: Signetics in 1961, Intersil under Jean Hoerni in 1967, Intel under Robert Noyce and Gordon Moore in 1968, and Advanced Micro Devices under Jerry Sanders in 1969, while Charles Sporck left Fairchild to run National Semiconductor in 1967. Journalist Don Hoefler gave the district its name in a series for Electronic News in 1971, and the label Silicon Valley stuck.

A financing model developed alongside the technology. Arthur Rock, who had helped arrange Fairchild's original backing, raised the capital for Intel in 1968 within days on the strength of a very short business plan. Equity funding for unproven companies, broad distribution of stock options to employees, and rapid movement of engineers between firms became normal practice in a way they were not elsewhere in American industry, and Stanford University supplied both graduates and, through its industrial park, physical proximity. Established firms including Texas Instruments, Motorola, and RCA competed hard against the newcomers and set much of the industry's technical agenda. The period also marked the beginning of sustained international competition, as Japanese manufacturers moved from transistor radios into calculators and then memory; the Busicom contract that produced the Intel 4004 illustrates how closely the two industries were already coupled. These commercial dynamics are pursued further in Economic History of Electronics.

The workforce and methods of semiconductor manufacturing changed just as dramatically. Early production had involved painstaking hand assembly of discrete components; integrated-circuit fabrication demanded photolithography, controlled diffusion and oxidation, cleanroom discipline, precision equipment, and tight statistical process control. New engineering specialties appeared, among them process engineering, device physics, mask design, and test engineering, while traditional wiring and assembly skills lost their value. Labor-intensive back-end work moved offshore early: Fairchild opened an assembly plant in Hong Kong in 1963, beginning the global division of labor between capital-intensive wafer fabrication in one country and assembly and test in another that still characterizes the industry. The human side of that shift is covered in Labor and Workforce Evolution.

From Military Priority to Consumer Product

The first integrated circuits cost far too much for commercial use, so they went to programs that valued size, weight, and reliability above price. The Minuteman II missile guidance computer became the first large-volume customer for integrated circuits in the early 1960s, and the Apollo Guidance Computer was built almost entirely from a single part type, using roughly 2,800 dual three-input NOR gates in its Block II version. These programs bought chips in quantity when no one else would, and the yield learning they financed is what later made commercial pricing possible. By the early 1970s the commercial market had far outgrown military demand, and the direction of technology transfer reversed permanently, with defense systems increasingly adapting commercial parts rather than driving their development. That relationship is explored in Military-Industrial Electronics Complex.

Computing absorbed integrated circuits in stages. IBM's System/360, announced in 1964, used hybrid Solid Logic Technology modules rather than monolithic chips, a conservative choice that reflected how new the technology still was. The System/370 generation adopted monolithic circuits, and the Model 145 became the first IBM computer whose main memory was entirely semiconductor rather than magnetic core. Digital Equipment Corporation's PDP-8 of 1965 created the minicomputer category with discrete transistors; the PDP-8/I of 1968 rebuilt the design around integrated circuits, and the PDP-11 of 1970 became one of the most influential computer architectures ever produced. Falling logic and memory prices put real computing power into laboratories, factories, and hospitals that could never have justified a mainframe.

Consumer products followed as prices collapsed. The electronic calculator was the first mass market to be built on integrated circuits: Sharp's QT-8D of 1969 was among the first calculators to use large-scale integration, and the Busicom LE-120A of January 1971 paired a single Mostek calculator chip with a light-emitting-diode display in a case small enough for a pocket. Within four years, ferocious competition drove four-function calculators from hundreds of dollars to a few tens of dollars and effectively ended the slide rule. The first commercial digital wristwatch, the Pulsar introduced by the Hamilton Watch Company in 1972, used an integrated circuit driving an LED display, and low-power CMOS soon made quartz timekeeping cheap enough for any wrist.

Video games arrived in the same year, though the technology deserves precision: the Magnavox Odyssey, the first home console, used discrete transistor and diode circuitry rather than integrated circuits, and Atari's Pong arcade game was built from standard TTL logic packages with no microprocessor at all. Their significance lies in what cheap logic made economically sensible, not in the sophistication of the parts. Elsewhere, electronic telephone switching, industrial process control, medical instrumentation, laboratory test equipment, and the first automotive electronic ignition and engine control systems all advanced on the strength of integrated-circuit technology during these years.

The Foundations of the Digital Age

By 1975 every essential element of the modern digital age was in place: the planar integrated circuit, a mature catalog of standard logic and linear parts, silicon-gate MOS, semiconductor memory in both volatile and reprogrammable forms, the microprocessor, and a manufacturing base that improved on a schedule reliable enough to plan products around. The miniaturization and steep cost reduction achieved in these fifteen years laid the groundwork for the personal computer, for global digital communication networks, and for the embedded electronics now found in nearly every manufactured product.

This era takes up the story where Transistor Revolution Beginning leaves off, and hands its results to Microprocessor Age (1975-1985), when the components perfected here reached ordinary consumers as products they could buy and program themselves. Readers tracing the underlying device lineage may also follow Vacuum Tube to Transistor to IC, and those interested in the people involved will find them in Transistor and IC Inventors.