Ran core memory diagnostics and adjusted various items on the IBM 1130

I USED THE CONSOLE LOADER TO LOAD VARIOUS DIAGNOSTICS TO MEMORY

Because this loader is operating switches normally intended for the human operator, it isn't very fast. We load a bit more than four core words per 3 seconds elapsed time, so that a full 8K machine load would require one hour and forty minutes. Fortunately the diagnostics are all developed to fit in the smallest 1130 configuration, 4K words, thus the load time is always less than 50 minutes and sometimes substantially less for simple tests. 

If my unreliable Lenovo laptop would avoid blue-screening for that length of time, I would load memory. It was quite hard today and I finally dug out an old Win 8.1 tablet and set it up to load core. 

CORE MEMORY TESTS

I ran a set of core memory tests that hammer away at all the core locations to find any issues. It runs as both a high and a low test. Since the code for the test has to sit in memory, it means those relatively few addresses are not checked with one test. The two different tests use different regions to place the test code, thus running the pair ensures that all 8,192 words are checked out. \

ADJUSTED ALL THE FEEDBACK SWITCHES IN THE 1053

I found that several of the feedback switches were badly out of adjustment. I am not surprised since so many of the machine's adjustments were also very bad. I patiently worked through all seven until I was certain the switch made and broke at the correct times. 

Three of the switches cover the one-cycle long functions, ensuring that no new commands are sent during portions of the cycle when nothing else should be moving. The print cycle is one such function, where the switch informs the controller electronics that the print is in progress and later lets the electronics know when the cycle can be considered over. 

Even if the there is a bit more time spent with the type ball returning to rest, for example, it is safe and efficient to trigger the character selection solenoids for the next character to be printed. The machine does not need to stop at the end of a cycle and start up again, in this case, so that it can achieve its maximum print rate of 15.5 characters per second. 

A shift cycle to reposition the typeball between the upper and lower case sides of the element are times when the control electronics should not trigger a character print cycle. This switch reflects the shift activity and informs the electronics when it is safe to start the following print cycle. 

Finally, a space cycle involves one spin of the operational cam and movement of the carrier one column to the right, but should not occur with other functions like print, shift or tab. A switch reflects the state of the space operation back to the controller electronics.

Adjusting these involved a combination of actions. Some of the switches had to be disassembled and portions bent to achieve specific conditions, then all of them needed the position of their bracket adjusted. A couple also had cams whose relative position on the shaft needed a change in order to achieve the various make/break times within the tolerances specified.

To measure these, I used a continuity meter and the hand cycle wheel. It has degree markings to assess contact timings for shift, space and parts of tab and carrier return. A different wheel is attached to the print shaft on the other side of the typewriter to measure the degree of rotation for the switch during a print cycle. 

There are two indefinite duration functions on the 1053 - carrier return and tab. The time it takes is multiples of cycles and depends on the start and stop columns as well. While these are underway the controller electronics should not be printing or actuating other functions, so these are interlocks to block other actions. 

For both of them, the are triggered by an operation clutch which rotates to engage the mechanisms that move the carrier forward to a tab stop or backwards in a return operation. It takes finite time for the clutch cycle to latch in and begin the return or tab, but after that the function continues until it hits a tab stop or the margins of the machine. 

To cover these, a pair of switches in parallel are implemented for each of those indefinite duration functions - thus four switches in total. One switch is activated as the operational clutch begins to activate the return or tab. Once the mechanism has been latched into place for carrier motion, another switch detects that the pawls are forced out of the rack so that the carrier can zoom left or right. 

While the operational clutch cycle completes and its switch opens up, the switch on the return or tab latches takes over and continues the interlock interval until the ending event stops those functions. That is, either the carrier hits a tab stop or the carrier hits a margin. Having both switches for a function provides an interlock from when CR is first initiated until it is complete, or similarly from when a tab is initiated until the carrier stops at a margin or tab stop. 

ONE BIT NOT LIGHTING ON DISPLAY, WAS ONLY A LOOSE WIRE

As I was running and studying various diagnostics, I had noticed that bit 11 of the Storage Buffer Register (memory contents) was not lighting. I had recorded this, assuming that a bulb had burned out which needed replacement. However, when I did the Lamp Test function on the machine, all bulbs were illuminated.

When I opened the back of the display pedestal I could see that the wire bringing the signal for SBR bit 11 was not on the pin of the display board. I pushed it back in place and could check off this task as well. 

MINOR CONCERN OVER UNDOCUMENTED FEATURE INSTALLED ON THIS MACHINE

This machine has a featured called IBM Channel RPQ installed on it, which installs 37 SLT cards into compartment A-A1 and adds six unique connectors that don't exist on standard 1130 systems. These feed a connector that I believe was connected to a third party line printer. This feature provides access to fetch and store data in memory, to raise interrupts (generally on IL3), and to respond to XIO commands on area code 24. 

A failure in some of those functions could trigger spurious interrupts, modify memory without warning or corrupt data being transferred around in the system. This system did NOT come with its ALDs. Since I have never seen an ALD for a system with this RPQ installed, I have no documentation for what it does in case I need to debug it. 

Because IBM would customize ALDs for its 1130, 1800 and 360 systems, displaying only the parts that were configured for the specific serial numbered machine, it is very important that systems which need restoration have their specific ALDs on hand. 

This is only a minor worry because there is some documentation about the RPQ in this manual on bitsavers - Bitsavers manual - which has just enough that I could trace out the logic to reverse engineer this if needed. 

TOMORROWS ACTIVITY PLAN

I expect to perform final runs of the CPU diagnostics and the Keyboard/Printer diagnostic as a checkout prior to preparation of the machine for shipment.