Over a weekend in 2012, PMF and I cleaned an IBM 129 Card Data Recorder and were able to fairly reliably punch cards once we were done. When we started it would frequently jam during feeding, mis-feed during the punch, and not cleanly stack the cards in the output bin.
Most of the problem was thirty years of dust, card fiber and grime built up in the mechanisms. The output hopper was full of it and needed a good cleaning to reliably pick up cards into the output stack:
The card punch and keyboard only appear to be seated on the table. They are just the tip of the iceberg -- much of the power and computerized bits fill the space underneath the desk. The bottom of the keyboard is in a cut-out in the desk surface and has a traditional typewriter mechanism that closes reed switches when each key is pressed, except for the special keys like "FEED", "REL" and "REG" that directly actuate lever switches.
Luckily the "monolithic memory" was still fully functional. Debugging the logic part of the system would be an immense task -- it is implemented with a wire-wrapped backplane and fills the bottom portion of the desk. Hidden behind it was a bookshelf with full schematics and a "Maintenance Theory" manual that described the interaction of the various components.
The individual boards in the backplane appear to be hand soldered and mostly consist of the what appears to be the same component, likely Solid Logic Technology blocks:
Once it was cleaned up, PMF tried puching some cards with great success:
We were able to verify that the registration of the cards and the punch were fairly accurate with this card guage:
The ink tape of the print head that marks a human readable version of the EBCDICHollerith Code was totally dry. Perhaps we can adapt a typewriter ribbon to work:
Also hidden in the bottom of the card punch was an incident report log, with the last entry from 1980. For a device manufactured starting in 1973 that is a very long life. The logbook is sitting on an IBM card sorter, which will be cleaned up later.
We ended up not using the IBM-6 Oil Spray:
Here's a 1971 advertisement for the IBM 129 that notes its new features, including the monolithic memory, the ability to tabulate columns of the cards, and the automated card feeding, punching and stacking mechanisms:
And here are more photos and video of the card punch. Once we have the manuals scanned and cleaned up, I'll post links, too.
Retrocomputing Punchcards 2012 IBM