High voltage troubleshooting of Keithley 237

[nikonoid]

I have a peculiar problem with K237 that I am not sure how to troubleshoot.

I had been working on repairing Keithley 237 for a little while. Initially it would pump -162V to the output even if I try to output something very small. First indications pointed to one of 15V rails being low, but eventually I traced it to Q19 in output module being shorted and dragging 15V down. After Q19 was replaced I got problems with resonant converter not generating +/-1200V, so I had to replace 5 electrolytic caps.

After I fixed resonant converter the K237 worked at +/-1000V for 10 minutes or so. The first thing I tried was to measure leakage on a cheap cable. When doing this measurement (source 1000V, measure current) K237 flashed something crazy on display and rebooted itself.

At this point the K237 would happily output -1000V, but output of +1000V would cause the screen flash and reboot. It would generate +850V without an issue. I tested it at +500V under 2ma load without a problem for few minutes.

Then even +850V would cause the reboot, while now +700V is still working ok.

I suspect the output board being at fault. Visual inspection did not turn out anything obviously burned on components and wires.

My best guess would be that one of high voltage capacitors is at fault, shorting at high enough voltage. This can also be a diode or a transistor.  The failure is only happening at high voltage and every time I go there I risk burning something else.

Schematic for K237 is available. Is there a good troubleshooting technique for something like that?


Here are some photos. Output module top and bottom:


Digital board and closeup of the ADC on digital board:


Analog board and overall view:


Resonant converter. You can find a great post on making your own and converting Keithley 236 to Keithley 237 here https://www.eevblog.com/forum/testgear/playing-around-with-an-old-keithley-236-source-measure-unit/msg1149883/#msg1149883


For now everything upto 100 volts works pretty well. I just need to get +1000V under control.

[TiN]

Perhaps worthy to leave funny but useful video from Marco..

He designed own HVDC-DC to pimp his 236
I'd test high-voltage module separately w/o the unit, to make sure it's stable and can handle rated load without excessive noise or instability.

[floobydust]

If the main PSU is ok, I would check the series-array of pass transistors, to ensure they are equally dividing up the voltage. The caps are all CDE mica and very reliable.
Set it to say +500V output, and carefully measure the (drain-source) voltage on each transistor. Repeat for the -ve bank. Sometimes one part shorts and breakdown stresses the remaining ones.

[nikonoid]

If the main PSU is ok, I would check the series-array of pass transistors, to ensure they are equally dividing up the voltage. The caps are all CDE mica and very reliable.
Set it to say +500V output, and carefully measure the (drain-source) voltage on each transistor. Repeat for the -ve bank. Sometimes one part shorts and breakdown stresses the remaining ones.

Thank you very much for a suggestion. I found one transistor behaving a bit strange. Parts are ordered.

I found a number of substitution transistors for 2N50 and 2P50. They have different max currents and also R(on) resistances (between 5ohms and milliohms). I picked the transistor with low on resistance. Is there much difference for this application or can I use pretty much any?
 


Sent from my iPhone using Tapatalk

[floobydust]

The majority of power MOSFETS are designed/rated for switching applications.
Here we are in linear-mode, so replacements best have a safe operating area curve that includes DC; not just pulse ratings.

Choosing much higher current/very low on-resistance rated parts means they have higher capacitance and are slower in this circuit. I would try to keep Ciss ballpark as the originals- but I don't know their make/part numbers. If you mix and match old/new parts in a series-string, the new parts should not be vastly different.

HV P-channel MOSFETS, not as many are offered so this limits things. Possibly:
FQP3P50 P-ch 500V 2.7A, 510pF, SOA 0.2A@400V
MTP2P50EG P-ch 500V 2A, 845pF, SOA 0.2A@400V but last shipments, almost obsolete.

N-channel MOSFETS with similar ratings:
FDP5N50NZ N-ch 500V 4.5A, 330pF, SOA 0.2A@400V
FDP8N50NZ N-ch  500V 8A, 565pF, SOA 0.4A@400V

[nikonoid]

Thank you, Floobydust. The actual transistor that I am suspecting is Q30. The service manual calls for MTP2N50. Actual transistor installed is BUZ 215 (I think from from Siemens). I am not sure if it is an original part or someone already tried to repair this. I have ordered FDP22N50N from DigiKey, but now second-guessing my choice. Is there a better transistor to try?


PS. In case of Q4 (the opposite P-channel fet), the service manual lists MTP2P50 and on the board I indeed have MTP2P50 branded as Motorola. This makes me suspect that BUZ 215 was not the original part.

[Le_Bassiste]

i concur w/ floobydust in terms of what to look for in the datasheet when trying to find a replacement. that said, the BUZ215 is indeed quite similar to the MTP2N50, so it seems to be a legit substitute "as  per datasheet values". however, the FDP22N50N seems to be far superior in terms of DC FBSOA, so, in my book, would be way to go.

[nikonoid]

Thank you, guys. It was in fact Q30 that was the problem. Since datasheet specs for new transistors seems superior, decided to replace both transistors Q30 and Q32, so they can share the load equally.

The 237 beast is back in action. I had to replace 5 capacitors and 3 transistors in total to get it going. Not too bad. I really appreciate the help.

The triax connectors look ok, but I would consider cleaning them anyways. I have some 99% IPA and also Methanol. Any recommendations on cleaning triax connectors?

[TiN]

No easy ones, I had to disassemble triaxes on Keithley boards to gain access to inner guard shell for cleaning. (Involved desoldering wires if required to get nut off).
Wear gloves to avoid contamination of cleaned surfaces.

Here's photo of my bulkhead triaxes from 7172 matrix module:

.