Keithley 2612 SMU Help with repair

[euph0ria]

Greetings people. I am having a problem with a dual channel Keithley 2612 SourceMeter. Moments after I power the unit on, I am greeted with an error : "5046 - SMU B: SMU too hot". Now, I can see that this error is considered SERIOUS (https://www.manualslib.com/manual/1228418/Keithley-2600a-Series.html?page=645). I have disassembled the device and I have swapped the SMU cards, I am then greeted with "5046 - SMU A: SMU too hot". At least I know that the problem is somewhere on that particular SMU board and not on the motherboard.
Has anyone any similar experience with recovering from such error code? I foolishly thought that I would locate some sort of thermistor or temperature sensor on-board but NAH, that thing is pretty crowded and complex for my experience.
I am willing to post a step-by-step tear down and repair if I can have some feedback on where the F*** to look for the temperature detection/sensing. I appreciate any input.

*EDIT* I am also linking the awesome post that helped me with the teardown: https://www.eevblog.com/forum/metrology/keythley-2612-teardown-and-repair/ which unfortunately does not have information for my problem. I am sure that someone more experienced might just spot the appropriate subcircuit from the high-res images.
Additionally, I attach an IR image which shows (under no load) that the thermal profiles are quite similar for both PCBs, but the error occurs.

[MegaVolt]

Image from the thermal imager shows two channels? A and B?

The images of both channels are the same, which indicates that the boards seem to work the same way. The error may not be caused by overheating, but by a message about it. It makes sense to take waveforms at the input connector before the error occurs and find the difference between them. Perhaps some kind of optocoupler ...

[euph0ria]

Thanks for the feedback, I appreciate it. I will report back if I find something interesting. Why optocoupler though? Are signals coupled like that for robustness?
I was wondering, how is this measuring temperature on-board? If a faulty sensor exists and the error propagates from the on-board FPGA to my connector I won't be able to find it even if I capture all the waveforms on boot.
Thanks alot

[Kleinstein]

Optocouplers are parts that are somewhat prone to failure - not as bad as the drop shaped tantalums, but still parts with an above average failure rate, at least in past with old parts.
So it could be a communication problem. I don't think this is very likely as one would than more get a communicaion error as the same channel would likely be used for more than just the temperature.

There are different possibly temperature sensors - in the simple form a diode, but in new model it could also be digital chip (e.g. I2C). With SMD parts a temperature sensor is no longer that obvious as an old times LM35.

Is it possible to read the temperature with the working PCB ? In this case one could use the working one on local heating to locate the sensor by moving a heat source over the board.

Chances are the sensor (could be more than 1) would be near a part that is somewhat prone to overheating, so like the power transistors or at the heat sink, possibly part of a µC.

[euph0ria]

Thank you for the response, I might just go and locally heat the working PCB until I get the same over-temperature error and work my way from there. Sounds very difficult but I am out of other options.
Thanks again for any additional input - Cheers

[MegaVolt]

Why optocoupler though? Are signals coupled like that for robustness?

All signals from the connector pass through optocouplers. Therefore, first of all, you need to check that they are working properly transmitting something. It is also worth checking the cables to the board. They can be interchanged. Perhaps the temperature sensor is connected directly to the outlet for guaranteed temperature checks.

The idea of locally heating the board should work.

Spartan if I'm not mistaken does not have a built-in temperature sensor. He must be somewhere outside. Considering that the exchange goes through optocouplers, the sensor must produce a digital signal.