Gregory's ServerThere is no intelligence in the punch: 80 wires from the computer drive 80 electromagnets in the punch, punching a row of the card at a time. The punch supplies -20 volts to the electromagnets and the computer pulls the appropriate wires to ground at the right time. 3/9
|
7 comments
 The punch mostly worked, but why didn't it punch the first 16 columns of a card? The IBM 1401 computer is a decimal machine, so 10 failing columns would make sense, not 16. And the cables in the punch were in groups of 20. So 16 failures was a mystery. 5/9 I checked the computer's driver boards; each board has flip-flops to drive two coils. Like all the 1401's circuitry, the boards used germanium transistors. Silicon hadn't caught on yet. The boards were pulling the lines to ground okay, but I didn't see -20V from the punch. 6/9 We checked the punch and it was generating -20V pulses correctly. So where were they getting lost? By measuring continuity at various points, we determined that the problem was on the 1401 computer side. 7/9 A thick cable connects the 1402 punch to the 1401 computer with 80 wires for the electromagnets. Inside the computer, "paddle cards" connect these wires to the computer's circuits, in groups of 16. I spotted a paddle card that had come loose. This was our problem! 8/9 Pushing the paddle card back into place fixed the punch. An easier fix than most! Now we can use the punch to create new card decks for demonstrations. You can see the system in action on Wednesdays and Saturdays https://twitter.com/ComputerHistory in Mountain View, CA, so stop by. 9/9 @kenshirriff I was there on Thursday! Hope to be able to stop by & see it it in action before I leave SF! |
The electromagnets aren't powerful enough to punch the hole. Instead, a powerful electric motor drives a metal "bail", stopping just before the sharp punch. The electromagnet holds an "interposer" tab in between, causing the bail to force the punch through the card. 4/9