Digital noise from Keithley 238

[zrq]

I acquired a Keithley 238 SMU a while ago, which appeared to work well before I connect it to a scope. It seems it's somehow generating way more noise at the output then the specification (15 mVpp DC-20M), regardless with whether it's in standby or CV operation. And interestingly, a majority of the noise are spikes that have a regular pattern, appearing "semi-randomly" on a grid of integer microseconds. The FFT in attached screenshots confirms this. The appearance of the spikes is also correlated with the CPU activity, when it enters the memory test, the noise shows reproducible patterns.

Does anyone have seen a similar problem, have any clue on the cause or can give some generic suggestions on troubleshooting? Failing decoupling capacitors somewhere?

[zrq]

It appears the feedback common (F) is also noisy against the output common (connected to chassis common), and the noise have a strong correlation with the noise at the output. It also correlates with the clock transmitted through the opto coupler.
Now I'm stuck again, how can I probe the noise against the floating feedback common, do I need a fancy low noise differential probe? Another basic question is what's the relationship of the DAC AGND (S), Vss (B) and the feedback common (F)?
I can share my copy of K238 schematics if you would like to have a look.

[Kanbus]

Have you tryed disconnecting ALL other equipment and lights in your lab from the mains. Some switching psu's have horrible noise.

[zrq]

Have you tryed disconnecting ALL other equipment and lights in your lab from the mains. Some switching psu's have horrible noise.

I don't think that's likely the problem, as I have seen the noise have a strong phase correlation with the digital switching at the optocouplers.
I tried to replace some (easier to reach) capacitors and adding some MLCCs to improve the decoupling, but didn't see any improvement.
Curiously, I noticed that the noise get around 2x higher in the 1A and 100 mA range compared to lower ranges. There is some gain....

[Hydron]

I'll have a look at my 237 when I get a chance. From memory the noise performance wasn't amazing when I looked at it last, but that was a while back so I can't recall any numbers (and I was concentrating on fixing output amplifier stability with my 236->237 conversion, so I didn't record anything).

[alm]

I just did a quick and dirty test with not particularly great measurement technique: Tek P6101B 15 MHz 1x probe (channel 2) connected directly to the center contact of the triax connector on the back and to the lo binding post, while disconnecting the shorting link between lo and earth. A better setup would be with a differential probe, but I don't have time to set that up right now. Still, performance looks within the ballpark in my opinion. The scope was set to AC coupling, 20 MHz bandwidth limit, and the 238 was set to source 0V. Both pictures are at 5 mV/div, the first at 0.5 ms/div and the second at 1 us/div to mirror yours.

Ignore the traces other than channel 2. I have it setup for a different measurement.

[zrq]

So from this it seems that your 238's noise is around 1/5 of mine, and there is much less such weird time correlation. I'm testing with the output common shorted with the link to the chassis ground, but floating that doesn't change the noise by much. There is something wrong with mine SMU.

I probed around in the SMU with a DIY H-field probe similar to Dave's. And somehow on the digital corner of the analog board where there are a few 74HC74 and MC14094B, I find a very good correlation of the noise at output (ch1) and what the H-field probe picks up (ch2). Elsewhere I see usually less correlation. I'll then try to replace the electrolytics on the analog board, anyway they are old and deserves replacement.