BTW, what does MCC and C2/C4/C5 mean in these pinouts? I was able to associate those with something that I recognized.
I didn't know, but I had a guess. With several things that have the same name that starts with C and has no other hint, it must be something fundamental, something so common that they'd expect a technician to guess right away. I traced it back in the 5100 logic diagrams to its source:
Sure enough: C is for
clock. Specifically, we have five separate clock phases. IBM explains the symbology on PDF page 264 of the
5100 MIM:
The probing you've done is pretty interesting stuff, and I think it's bound to reveal a lot! One of the things that occurs to me is that some of the intermediate voltages you're seeing could be because the signal on a pin you're measuring is doing a lot of switching between 0V and +5V. A scope of course will tell the tale.
Some remarks based on what you've said:
although I thought 0.2V was the threshold for 0, the DisplayOff bits go to 0.38V
The 5100 MIM says you're logic low if you're below 0.8V (PDF page 256):
I'm not sure if the Display can turn itself off, or if it has to be Commanded so from the Executive ROS?
I don't know either, but I do know that the display
can be commanded to go off by software. By default, the BASIC and APL interpreters disable the display while they're running a program, since the Display Card will suspend the processor whenever it needs to access RAM to redraw the screen. When the display is off, the processor does not get suspended. The 5100 will only turn the display back on while a program is running if the program prints something; otherwise, the 5100 assumes that you want your result as fast as it can compute it, and it also assumes that you don't want to see it think.
There may be other things besides the processor that can disable the display, potentially including logic on the Display Card itself. Maybe the mechanism can be found in the logic diagrams somewhere.
The BusOutError (M5) and DeviceAddressError (M9) basically have the same value between these two trials (below 0.2V).
That's a good puzzle. However, it looks like several cards in the 5110 can pull down -Machine Check as a result of a +BusOutError or a +DeviceAddressError. Maybe it would make sense to examine the outputs of other cards, identify the pins that correspond to those other error types, and spot the card that's actually triggering the halt?
The ALARM pin (P5) appears to be blaring using 8.32V.
I think you can rest assured that this behaviour is correct. Look at diagram 420 in the
5100 System Logic Manual. +Alarm On is connected to 8.5V through a 240 ohm current-limiting resistor.
But for sure, EvenBit0 shouldn't be reading over 5V.
Extremely weird. You may have noted that this line is an
input to the Display Card, not an output.
As a pin on the storage read/write bus, this pin can be driven from several places. The RWS cards can drive the bus; so can the Controller and the Executive ROS card and many others. One of the few cards that can't drive the pin is the Display Card.
Shenanigans like multiple cards trying to drive the bus at once could cause this voltage to go too high like this.
We've already established that it's the Display Card as a primary suspect in the fault --- we know that if you replace the broken card with another one, then the whole machine works.
I think there are a couple of categories of problem.
Category 1: the bad Display Card is messing with its inputs somehow.
Maybe a buffer chip got fried and something that should be an input is now driving the line. Maybe consider taking ALL the cards out except for the Display Card. Then check voltages on the pins for the storage read/write bus lines. Don't be surprised if they are high at first: floating inputs do strange things. But use a reasonably chunky resistor (say 200K?) and try pulling the pin down to ground through the resistor. If the pin stays high or won't be pulled down, then I'd say that this is suspicious.
You may want to carry out this experiment with your homebrew power supply. It's possible that without the load of the usual complement of logic cards in the 5110, the PSU will generate abnormally high voltages; your custom job probably won't.
Category 2: the bad Display Card is sending outputs that cause the problem.
The Display Card doesn't have too many outputs, and the ones that go to the display we can almost surely ignore, since the display can't cause Machine Checks. You've basically got the bits it sends out on the storage address bus, +Printer Clock Pwrd, +Display Req Pwrd, and -I/O Display Off. How could those things conspire to upset the entire machine?