Keithley 2100 drift and calibration?

[dmderev]

Hello, is there anybody familiar with normal behavior and calibration of Keithley 2100 (or its brothers Picotest M3500A or Array M3500A, Atten ATM3511A, BNC 1201, Chroma 12061, DL-2060, DME 1600, Kusam-Meco M3500A, Time Electronic 5065, BNC 1201)? I have a DMM which claims Tektronix calibration, but it is ~80uV off on range 10V. I tried to calibrate it following calibration manual, and it appears to adjust readings and change calibration date. And it measures OK before next power cycle.
It is unclear if the calibration did not make it into EEPROM, or something else is wrong, but the readings now are different from before calibration (~40 uV), but still not the same as the measurements after calibration and below power cycle.
There is also strange behavior of readings at cold startup - they are appearing closer to used voltage standard, but then drift to incorrect values within next ~20 mins. The drift is slow, followed by a steep change of readings after ~150 readings, then slow again, then step etc... I am attaching the plot of drift vs time since cold power up . It settles after 18-20 min, and at that time I am doing calibration. But on the next power cycle the behavior repeats...

The manual says that the DMM does autozero on every measurement, and there is no explanation to periodic step-wise moves.
Did anyone experience similar behavior or prooblem? Any suggestions what can be wrong? There is no schematics to look at, unfortunately.

[Smokey]

That is a crazy number of rebadges.. Who is the OEM?
WTF is Keithley doing rebadging a 6-1/2 digit multimeter?  They already make a killer 6-1/2 digit multimeter... I mean the only thing that would come of that is weakening the brand.

[Kleinstein]

AFAIK the 2100 is still very similar to the Keithley 2000. There are least pans available for the K2000.
For the rebadges with is really confusing. It is not so clear who actually produces them - there is a chance that they sold the design so a different source like picotest or array builds them and they just now buy them back as there low cost K2100 model.
After all it is a proven design with not too many weak points.

The details on how the K2000 and likely the new meter work internally are not given, but there is a chance that the meter does a correction for the ADC gain from time to time. Also the AZ cycle may not be as simple as alternating between zero and the input. At least with other Keithley models there are indication of some odd averaging or similar going on that efffect the noise and reading rate.

Normally adjustment should be done with a much longer warm up time, more like 3-6 hours for meter without a fan.

Chances are the steps seen in the warm up can be from one such drift corrention methods.  There is a chance to have some defect that makes the drift larger than normal and the steps thus more obvious.  Chances are one would need more than just a new adjustment, but more a repair - at least a more detaild test of the meter. Adjusting a defect meter is a bad idea - it may read OK at the test points, but not at other points.

The easy tests that should be done are:
is the zero stable
input bias current  (some defects cause extra bias)
check zero in 1000/100 V ranges with open input - gives a first hint on bias
turn over test (so compare the +10 V and -10 V reading, maybe also +-1.5 V )
A crude linearity test from recording a slow smooth drifting voltage in the 10 V range (e.g. some 5000 µF with some 10 M) to check for DNL type errors.
Check the noise at zero

[Mickle T.]

1) When performing calibration from the front panel, the DMM must NOT be connected to the PC, otherwise the correction factors will not be automatically saved in the EEPROM.
2) If your DMM has a recent firmware version (>1.05), its calibration procedure will differ from the one described in the service manual. Tek sends the current procedure upon request. As far as I remember, the differences are in the additional steps of calibrating the negative scale and 10 Mohm/>10G input impedance.