I swapped out all but one of the 2.7uf tantalum capacitors on the servo board. One of them is just under a thin wire jumper on the back side, looks like an engineering change, which I don't want to disturb. There are still several 10uf and 22uf tantalums of the same type that I should replace.
I put the removed components back - two transistors, a diode, and another 56 ohm resistor. I want to isolate as much as possible so that I can figure out the source of my problems without the confusion of multiple power supply levels and interconnected boards.
I set up my own -20V power supply and tried to isolate all but the negative power supplies in order to track this problem down and obliterate it. So far, still chasing voltages and validating the behavior. First thing I realized is that I didn't have the power transistors from the heatsinks plugged into the servo board, which is necessary for the board to develop the full negative voltage.
I removed both heatsinks so they could be hooked up as needed on the bench, to continue my isolated testing. With the heatsink for the -10V hooked up, I got an exact -10V as previously determined, but my -5V was down at about a third of a volt (negative). The reference voltage to the op amp is -5.01V, an excellent value, but the other input is our defective -0.373V yet the output is not swinging to a negative value near the low rail (-20V).
My current guess is that the op amp that should steer the overall regulated voltage to -5 is not working properly, based on the lack of a big negative output. However, there are some components that might be causing the symptoms, so I will methodically eliminate them before replacing the op amp. Nothing looks likely, other than the op amp, but checking will proceed carefully since there is a several-day lag for some parts orders to arrive.
CHECKOUT OF A P390 SERVER
I am checking out a donated P390 server - an x86 server (IBM x235) with the P390 processor card that hosts it a S/390 5 MIPS processor that will run MVS or Z/OS on the server. The x86 server has a bad CMOS battery, which is not surprising, but OS/2 came up fine and some initial x-server diagnostics gave the machine a thumbs up. I started the P390 and it did a bunch of 390 processing before it halted with a disabled wait state.
I would need to look further to see why MVS stopped. It could be a startup issue I could solve with a different kind of start, it could be lack of a virtual peripheral or console session . . . I don't know that much about the P390 system yet. However, all signs are good that this is a complete and working historical artifact.