Keithley 182-M NVM - offset with excessive noise

[MiDi]

By accident the front panel connector was hit and unit suddenly showed large offset in conjuction with excessive noise.
Nothing can return instrument to normal behaviour (range change, turn off, long time waiting, w/ or w/o battery, etc.), except tapping the connector multiple times until offset is gone.
I optically checked the guts for loose connectors, bad solder joints, suspicious parts, but could not find anything obvious (unit is already recapped).

Any ideas what could cause the issue or where to look at?

To give a better impression of the issue video and pictures attached:

[Dr. Frank]

As the connector (with plug attached?) got a hit, the most probable defect would be a break inside this connector.. undetected so far, may be hard to detect at all.

Check both input leads for continuity

Frank

[dietert1]

20 uV could be thermal EMF. I would unplug and replug the input cable 10 or 20 times in an attempt to remove oxide from the contacts. Or use a chemical like vinagre to clean the copper surfaces. You can test your method on another copper part, to see whether it gets shiny.
By the way these connectors exist in many variants (with different contact materials). Not all of them are low thermal EMF. Also there should be a rubber piece with two holes for the wires to fill air space inside the connector.

Regards, Dieter

[HighVoltage]

As the connector (with plug attached?) got a hit, the most probable defect would be a break inside this connector.. undetected so far, may be hard to detect at all.

Check both input leads for continuity

Frank

Very likely so,
I would short the input connectors at the PCB input to see if the offset and noise is gone, to quickly confirm that it is a problem with the connections in the connector.

[MiDi]

Guess what, both crimps on the input resistor are bad 
Resistance between both resistor legs and lugs varied a lot if touched and the lugs can be twisted slightly.
Now specimen has to live without crimps 🦐



After repair everything seems fine, no sensitivity to vibration spotted anymore, noise and values are as expected.
21nVacrms/240nVpp @1PLC all filters off over 29h (1.4 Mio. datapoints)
2.6nVacrms/20nVpp @100PLC equivalent (means of 100rdgs each w/o overlapping @1PLC - same dataset)

[HighVoltage]

Nice repair, what an interesting failure.
One solution would have been to re-crimp the lugs.
But you were really lucky that the legs were long enough to be mounted like this.