But Eldon Hall, who had been at the Instrumentation Lab since 1952 and led the design of the flight computer, had been intrigued for years by the prospect of integrated circuits. So he initiated two parallel design programs: one to build a computer using core transistors, and one using integrated circuits. By the fall of 1962, Hall says, “it was clear to both sides that it was easier to build a machine with Micrologic.” The integrated circuits could perform calculations more than twice as quickly as the core transistors, and their space savings meant that the computer would have much more room for memory circuits. Moreover, wiring them together was much simpler and posed fewer opportunities for something to go wrong. That winter, Hall persuaded NASA to redraw its contract with Raytheon, the company that was to manufacture the computer, and take a gamble on the new technology. “How did I con these managers into letting me use integrated circuits?” Hall says. “I didn’t have to con them. They didn’t care. I could do what I wanted to do.” NASA managers, concerned with more imminent missions such as the 1965 and 1966 Gemini flights, simply weren’t paying much attention yet (see “Getting the Job Done,” p. M16).
By today’s standards, the Apollo computer had a peculiar architecture: it used only one type of logic circuit, the NOR gate–so called because it outputs an electrical signal only when it receives a signal from none of its inputs. A computer built from NOR gates is less efficient than one that also uses other types of gates–the AND gate, for instance, which outputs a signal when it receives signals from all of its inputs. When asked why the Apollo computer relied so heavily on the NOR gate, however, Hall laughs and says, “Because that’s what Fairchild was able to build.” Once Hall and his team had a hardware design that worked, they weren’t about to take chances on even newer technologies. Instead, they worked closely with several potential manufacturers to ensure that the NOR gates could be built reliably.
By the time the first Apollo mission flew, Fairchild had abandoned its NOR-gate chips for more sophisticated architectures, so Philco supplied the computer’s logic circuits. Reliability, which had once been the integrated circuit’s chief drawback, was now its chief advantage. “Computers in those days wouldn’t work for more than a few days without repair,” Hall says. In 15 Apollo flights, however, the guidance computer never suffered a hardware failure–even when lightning struck during the Apollo 12 takeoff.
The design of the guidance computer, like those of the optics and the IMU, was largely complete by 1966. From then until the Apollo 8 mission put astronauts in orbit around the moon in December 1968, the lab focused on software development.
In the beginning, nobody would have predicted that. Early in the Apollo effort, “the lab sort of bifurcated,” says Fred Martin, SM ‘59, ScD ‘65, who would become project manager for the command module’s software. One group designed hardware. The other group, Martin says–the analysis group–“dealt with how you were going to get to the moon, and what kind of measurements you were going to make, and when you were going to fire this big engine, and what direction you were going to point it, and how to figure out the trajectories to get to the moon, and how to worry about the errors you were going to have.” Of course, the analysis group’s calculations would eventually have to be embodied in software; but devising equations was seen as the heavy lifting.