Saturday, September 6, 2014

Core memory situation worsens, but 029 keypunch restoration fundamentally complete


I opened the connector on the printer cable, which I believed had a number of signal cables severed by the connector entry hole rim due to the sheath not being tightly cinched in the cable clamp. There were certainly quite a few wires cut off right at the point where the cable entered the connector housing.

Here is the removed connector housing, put aside for safekeeping since I anticipated a long task marking all the signal lines, cutting them even, cutting back the sheath, resoldering all the lines and then properly cinching the sheath into the cable clamp as I reassembled everything.

Connector housing, removed and set aside
When I inspected the connector without the housing around it, it became obvious that all the signal lines are properly attached - the severed wires are extra lines that come in the standard cable IBM uses, not signal lines to or from the printer.I don't have this onerous task ahead, which is good news considering what is happening with the core memory problems of the 1131 itself. All I need to do is reassemble the connector housing and this cable is ready to reattach to the 1130.

Severed wire ends out of cable but no severed wires on the connector!

All signal lines intact, ready to work

Another view showing intact signal lines to all connector pins 
My conclusion is that IBM manufacturing used a standard cable with more wires than needed for the 1132

I opened the gates and reseated every single SLT card in the machine, just in case I am dealing with loose cards. It was a long, painstaking process to verify that there are no bent pins, that every card is seated properly and to check all the cable connections at the same time. By 2PM I had completed gate A, with gates B and D still to do.

By 3:00PM I had completed B gate as well, leaving just the SLT cards in the memory (D) gate to reseat. There were only two compartments in that gate and they were partially filled with the core stack, so it went faster. By 3:30PM I was ready to power up and test again.

To my surprise and disappointment, the machine continues to get sicker and sicker. The same bad bits are showing up in memory reads, but the machine now thinks it only has 256 words of memory, judging by the addresses through which it is looping with the CE Storage Load function.

I single stepped through a couple of 'instructions' from memory, but even here I see very bad behavior. The memory reads out with bits 1, 5 and 8-15 on, but when that is transferred to the D register it appears as 0xFFFF.

I have to wonder whether there is oxidation on the pins and SLT card contacts, so that when a working card is pulled out a bit and reseated, it may sit atop oxidation thus having no continuity. Or, are the wire-wrap connections or signal cables bad enough that the least disturbance causes them to go open? That doesn't make sense if the machine was so healthy after thumping its way across the country in a truck.

There has to be some systemic reason for the continued decline of my 'patient'. Needless to say, this is a low point for me.

I may have to start with the very basics - logic analyzers and scopes from the system oscillator outward, verifying the basic states and operations at a low level then spiral outward to find all the incorrect results.

If I find failed cards, I have some test facilities to troubleshoot down to specific bad components. Whether I can repair those or will need an entire replacement card, depends on the card and the bad component.

Test jig to fit up to two double size SLT cards

Example SLT card (not from the 1130)


I picked up Deoxit, an oxide remover, which I will carefully use only on the contact points and clean it off quickly with isopropyl alcohol to ensure no long term interaction with the plating on the contacts. I don't want to use sandpaper or other abrasives because the contacts may have a thin plating which would be punctured by scratching.

I was able to get most of the contacts working, although I ran out of the $20 can of Deoxit before I was done. I can feel that it has quite a bit of the solution inside, but the propellant is all gone and I have no way to get my $15 worth of solution out of the can. Grrr.

I tested most contacts to verify they were working but when the relay to register that a card is in the punch station was still not closing reliably. When I checked the relays, it was clear that their contacts were also oxidized to open circuits. I used electronic contact cleaner to clean them off, but on some it took three or four courses of spraying, working the contacts by closing and wiggling the armature, and testing continuity until they were all closing and opening when they should.

At this point, registering a card resulted in an immediate release of the card out to the next station. I hunted down the path that causes this to happen and found that an Escape Interlock relay is what was not closing. I found an Escape Latch Contact in the path to energize the relay, located it underneath the mechanism, and verified that it was a perfect insulator.

I hooked up an ohmmeter to the contacts and tried to clean them, using the contact cleaner spray and IPA, but even with lots of wiping, rubbing, and other actions, I never got a glimmer of a connection. Finally, I had to resort to a fine emery cloth, something I would have liked to avoid, but was able to gently stroke the contact faces until I got good connectivity.

Success! The machine registered cards, punched perfectly and printed the characters on the top strip of the card. It is solid as can be, which is what I expected for a unit that Doug Martin had stored, which is why I waited to get this one. He figured out that he acquired it back in 1978 and, other than a few brief times when he powered it up to punch a card, it has sat without use for 36 years.

Machine working properly, still needing a few cosmetic touches
Quick video of the basics - the 029 feeding, punching on a card and releasing. See it at IBM 029 restored, shown handling the basics

My last task today is to apply my new sound deadening foam, after removing the tarry, crumbling IBM foam from inside the cover. Once I get the covers that go over the ribbon mechanism and in front of the blank card hopper, plus clean up some oxidation of aluminum surfaces and other cosmetics, it will be checked off as a completely restored machine.

New sound deadening foam installed - see bottom and behind program drum for a few pieces
It was certainly helpful to have a big success with the keypunch on the day I watched the 1131 slip downhill further in spite of my efforts and testing to date.


I have a business trip to the NY area from Sunday through Wednesday, arriving back by late morning on Thursday. I have some free hours and have arranged to visit with some fellow hobbyists who have been in contact via email because of the blog.

As well, I will get a chance to visit the Mid-Atlantic Retro Computer Hobbyist group's museum exhibits at the Infoage Museum, sited in New Jersey at the former Camp Evans site. I have been a long time member of the MARCH forum but was too far away to visit or help with any of the restorations.

While there, I will contribute a bit of time helping with the breakdown and boxing of several 029 Keypunch machines, as they decided to get rid of the frame and tabletop to gain back useful space, but are holding onto all the mechanisms and covers in case they are needed in the future. I can help them disassemble this machines rapidly, so they can be boxed, labeled and stored.

As they are discarding the legs, frames and tabletops, I can pick up a part of the leg assembly and bring it home, to replace an area that is corroded on my 029. It will fit easily in my luggage for my trip home.

Once I arrive back home, I will resume part-time restoration of the 1130 system. You should expect blogs to resume late in the week.

No comments:

Post a Comment