1965 · Dual NOR Gate Flat Pack · Work 853 of 1,871
An assembly bench in a semiconductor works in Pennsylvania, under the lens of a grey stereo microscope a narrow strip of finely grained gold-coloured sheet metal three centimetres long with a round hole at one end, along both long sides rows of parallel gold lead strips still joined to it, in its middle a tiny flat square package with a gold rim around an open window, inside a dark silicon chip with ten fine gold wires, beside it a shallow clear tray holding more strips in rows, two hands in light blue smock sleeves reaching in from the lower edge holding fine tweezers, behind a row of microscopes on empty benches, fluorescent light.
What it was used for
The device was the only logic type in the Block II Apollo Guidance Computer. The list of MIT specification control drawings printed in the Computer History Museum's article carries it as a flight-qualified Dual NOR Gate in a flat pack under Philco number PL9939; Block I, by contrast, had a single Micro NOR Gate per package. Fairchild designed it, Philco in Lansdale, Pennsylvania built most of them. According to Paul Ceruzzi of the National Air and Space Museum, MIT deliberately chose only one type so that it could be tested more thoroughly; if a chip failed the tests, the entire lot it came from was discarded. The processor was built entirely from NOR gates: registers as latches of 2 gates each, an adder, and hardwired control.
What was new
The device carried 2 NOR gates of 3 inputs each in one package, built from 6 transistors and 8 resistors, with 10 leads - the Block I predecessor had one gate per package. The AGC was not the first computer to use integrated circuits; Ken Shirriff calls it one of the very first. Minuteman II ran in parallel: in the autumn of 1962 Texas Instruments won the contract for 13 custom circuits for Autonetics, and according to the Computer History Museum's object record Minuteman overtook Apollo in 1965 as the largest single consumer of integrated circuits. The barb is in Ceruzzi: by the time Apollo 7 flew in October 1968 the six-device circuit was long obsolete, and there was no way to test the newer chips and incorporate them - the very qualification method that made the part reliable also locked it in.
What the sources leave open: The unit counts count different sets: the Computer History Museum gives 2,756 flat packs with 5,530 gates for Block II, of which the CPU used only about half, and, for Block I, 4,100 IC packages for the whole electronics. That is not a contradiction, but a text must say what is counted. No failure rate is documented; only that no AGC suffered a hardware failure in flight. No unit price for the Block II part was found.
Connections
Every connection names its kind. “Forerunner” on its own would be a claim.