An electronics workshop in California on a foggy morning, on the bench an open black hard-shell case with an aluminium rim, its lid lined with white foam blocks, in its black lower panel two green sockets with small levers and a cream-white key pad with plain smooth keys, at the left a long cream-white box with vents, its small round viewing hole giving a cold blue glow, on black foam in front three white ceramic chips three centimetres long with a small square quartz window framed in gold, the tiny crystal shimmering beneath, plain smooth and unmarked, a hand in a pale blue shirt sleeve laying a fourth chip beside them with tweezers.
What it was used for
An ultraviolet lamp for erasing EPROMs. Tom McHugh describes the routine at Dave Nutting Associates in an interview: Dave Nutting drew out what should appear on the screen, McHugh turned it into hexadecimal and then into a table of bytes that went into the EPROM; Jeff Fredricksen came up with the blacklight used to erase the EPROMs. Per All About Circuits, erasure uses ultraviolet light at 253.7 nm through the quartz window onto the die, typically for 10 to 30 minutes, and always the whole chip: single bits or bytes cannot be erased. The part should then rest before being reprogrammed.
What was new
What the part could do: hold data without power and still be cleared again. The Smithsonian calls the 1702A the world's first non-volatile, electrically programmable, UV-erasable ROM. By Intel's own account the invention came out of a fault: from autumn 1969 Dov Frohman investigated reliability problems in the 1101 memory and found that the silicon dioxide absorbed electrons and held charge - what spoiled the device could be used to store. Intel first saw it as a tool for developers, cutting prototype design time from days or weeks to hours; it became a business that grew alongside the microprocessor business into the 1980s.
What the sources leave open: The work's year, 1974, matches no part in my sources: Intel gives 1971 for the EPROM's introduction and February 1971 for the prototype talk at the Solid-State Circuits Conference, the Smithsonian dates its 1702A to 1972, and All About Circuits dates the 2708 to "around 1975". Which device the work shows therefore remains open. McHugh describes the work on Gun Fight, which the interview's introduction dates to November 1975.