From This Story
The Tomcat’s Brain
As a 1968 graduate in electronic engineering, Raymond Holt was in a sandwich generation, caught between vacuum tubes and transistors. When he showed up for his first job, at Garrett AiResearch, the personnel manager told him he was the only one in his department who’d had a class in computer design. AiResearch had just been hired by the Grumman Corporation to design the F-14 Tomcat’s central air data computer, a microprocessor that would, with data from pressure sensors, calculate and report (on cockpit displays) Mach number, altitude, and airspeed, and, accordingly, move the aircraft’s control surfaces and adjust the sweep of its variable geometry wing. AiResearch had designed the McDonnell F-4 Phantom’s central air data computer, but that had been electromechanical. “It had gears and cams,” says Holt. “It’s like a transmission on a car, except that it’s chrome and gold. It was about two feet long and three feet high. When they showed it to me and said I had to make it electronic [to fit on a circuit board 40 inches square] I about freaked out.”
Northrop Grumman engineer Dave Wolfe worked at Garrett AiResearch at the time, where his job was teaching Grumman technicians how the F-14’s central air data computer worked. (“I saw the airplane and I was so impressed that I left [AiResearch] and joined Grumman,” he says.)
“Electromechanical systems were a maintenance nightmare,” says Wolfe. “You needed a very special type of artisan to repair those things — almost a watchmaker. People were looking for something to replace that system.” Garrett AiResearch was the only company to submit a bid for a digital, instead of electromechanical, design, but the small, powerful processors required didn’t exist.
In 1968, the integrated circuit was 10 years old. Jack Kilby of Texas Instruments and Robert Noyce of Intel Corporation had independently developed methods for assembling transistors — switches that in the binary language of computers say “on” or “off” —on a single slice, or chip, of a semi-conducting metalloid, like silicon. (Kilby’s work won him the 2000 Nobel Prize in Physics.)
As a 10-year-old, the integrated circuit had come a long way. It could integrate hundreds of transistors on a single chip, as opposed to the first integrated circuits, used on the Apollo guidance computer, which made do with tens. But it hadn’t come far enough for the F-14’s computer, which demanded as many as 3,500 transistors on a single chip.
In 1968, the circuits were mapped on huge, multi-color Mylar sheets “the size of an entire wall of a commercial building,” recalls Holt. The degree to which that layout could be shrunk — and its detail preserved — to form the pattern for etching on a chip was one limit to a chip’s capacity.
Holt’s team and chip maker American Microsystems created a six-chip processor containing 65,000 bits of data, an achievement Holt believes has been unrecognized because of semantics. Intel Corporation gets credit for producing the first single-chip microprocessor in 1971, but Holt contends that with the F-14A’s first flight in 1970, his team beat Intel to the first microprocessor by a year. Unfortunately, a hydraulic failure on the first flight caused the F-14 to crash (the pilot safely ejected), but the central air data computer performed flawlessly.