Friday, September 5, 2014

Progress made on core memory issues, more to do, plus keypunch restoration well along


I completed assembly and wiring of the keypunch with the aid of Doug Martin who stopped by today. The keypunch fed cards well and stacked them quite nicely too, but several things were not right.

Card registration and punching were not yet working. The program card unit was frozen in the up position, with no movement at all on the lever that lowers the star pins down onto the card drum to read the program card. The keyboard wouldn't rotate freely when inserted, staying rotated about 30 degrees to the right. The mechanism was dusty and some parts could use new lubrication. The aluminum parts had some surface corrosion but the stainless steel sections were all good except for one rail that sits in front at the bottom of the stacker, which had rust spots.

Looking beyond the temporary problems and cosmetics, however, I could see that this was lightly used, with little visible wear mechanically. The rubber rollers were in great condition, no flat spots or slippery surface sheen. For a unit that hadn't been turned on in decades, it surprisingly stayed well adjusted and was remarkably quiet in operation.

I worked away on cleaning it up, applied the grease and oil everywhere it belonged, even freeing up the frozen axle for the program card unit. The mechanism sounds perfect now that it is properly lubricated and the star pins lower and raise with ease. I repaired the keyboard rotation guides and have it oriented as I like.

The contacts on all the CBs - cam operated contacts that make and break circuits at specific rotational points in the machine operating cycle - were black with oxidation. That meant that most of them are not conducting electricity, an excellent explanation for why the machine doesn't recognize that a card is registered for punching. I did some work on them with IPA and with an electronics contact cleaner, but more is needed before it will work properly.

The relay contacts were less oxidized. I give them a precautionary treatment with spray on contact cleaner, but they all seemed to be doing their job well. On a side note, not all contact cleaners are the same - some include mineral oil or other lubricants, which just accelerate the accumulation of dust in the future.

I believe that once I properly clean the CB contacts - three for the feed cycle and 5 or 6 for a punch cycle - I should have a solidly working keypunch machine. I expect to complete this Saturday (tomorrow), before my trip to the east coast which will take me away from this rescue/restoration for most of a week.


Something was injecting a 1 bit into certain positions of the B register anytime it was transferring IO fields - which it does as part of a CE storage load or when Load mode is set by the operator on the rotary mode switch and Program Start is pushed.  I had to find and fix whatever was doing this, which takes a bit of work since the inverted inputs to the B register are all wired together - what IBM calls a 'dot OR' - thus if any gate is pulling it down to 0 the result is a 1 bit.

Each section of the machine is expected to only pull down when that device has exclusive use of the IO bus, but something is jamming the bus all the time. I can check the state of any output that feeds this 'dot OR' and they will all be at zero if some gate is actively low. No separation that would allow me to look independently at each source. No means of disconnecting some of the signals, other than yanking out the logic cards that contain the output gate wired into the 'dot OR'.

I picked a bit position - bit 1 - and methodically yanked every card that drives the IO bus for position 1. With seven SLT cards sitting on the table, I brought up the machine and saw that it was not jamming in the 1 any more. I could see the B register flicker the 1 bit when the program had been occuring, now it was off unless I set bit 1 to on with the console data switches. In that case, it was solidly on.

I seemed to me that one set of cards I was removing were already a bit loose, which might have been the cause, but I was in a hurry to try out my test with the seven removed, so I didn't investigate that any further. I began adding in cards, one peripheral device controller at a time, looking to see when it went back to the flickering bit 1 on the panel. It never did - causing me to naively believe that my problem was fully fixed.

However, when I went to normal mode and displayed a location, I found that now I was getting bits 1 and 5, like before, but also the entire lower half of the word - bits 8 to 15 - were all on!. The problem progressed to a wider set of symptoms, although I had at least fixed the flickering bit 1 and 5 during storage load.

It is my guess that there are loose cards in the compartments, one or more of which I made looser while pulling and replacing the seven specific cards. It was getting late, so I shut down for the night but the plan of action is to pull and re-seat all cards in all compartments of the A and B gates, then see what shape the machine is in.

No comments:

Post a Comment