Calibrating a Keithley 2000 without any calibrator?

[petemate]

I am (eventually) gonna have to calibrate a Keithley 2000 multimeter that I am trying to repair at the moment. I don't have access to a calibrator, as the calibration manual suggests using. However, I do have access to another calibrated Keithley 2000. Would I be able to basically just take a power supply and generate a voltage of +10V, -10V and +100V, verify that with the calibrated multimeter and then basically enter the values of the calibrated multimeter into the multimeter that I am calibrating? That is, I measure a voltage of 10.01000V with the calibrated multimeter, then plug in the uncalibrated multimeter and enter that value using the arrow keys?

Would I be able to do the same with resistance(1k,10k,100k,1M), by basically going to my parts drawers and find one of these resistors, measure it with the calibrated multimeter and then enter those numbers in the uncalibrated multimeter? What about DC current?

If the above solution work, I guess that I could use the same principle to generate the AC voltages and currents from a power amplifier being fed by a signal generator? Provided that I can find something that can output 50kHz, which I think I can procure. The only issue I have would be to generate the 700V voltage. Any idea?

Also, finally, you are supposed to use a "Keithley 8610 Low-thermal shorting plug", which basically shorts the four inputs(low, high, sense high, sense low) during calibration. I see that someone has given it a go at making one themselves, but what would be the consequences of just soldering four banana plugs together, perhaps via a blank pcb?

Thank you for reading!

[zhtoor]

this could work, provided that the measuring instrument (the reference meter) has a measurement
accuracy which is 3-10 times better than the one being calibrated.

regards.

-zia

[petemate]

this could work, provided that the measuring instrument (the reference meter) has a measurement
accuracy which is 3-10 times better than the one being calibrated.

regards.

-zia

Thanks for replying. Unfortunately, the calibrated meter is also a Keithley 2000. So they have the same accuracy and digits(6.5) available. Why would the accuracy have to be better than what i am trying to achieve? I suspect its some math thing related to multiplication of tolerances, but what exactly?

[TiN]

Just think of it.
You take reference meter (R) and unknown meter (X), connect to the value source (Z).
Done the measurements, using same settings, for example.
R = 10.005643 V, which means true Z signal is +/-noise of the source and +/- uncertainty of the K2000. Assuming that K2000 reference has recent and valid calibration.

For clarity pick worst case and assume, you hit the largest devication of the spec + and add to that largest deviation in source noise/stability in + too. Now add uncertainty of the calibration to that too.
Now take X and get = 10.003152 V.

As result you can tell if meter X is within K2000 spec + source stability (which is for even good PSU many times worse than calibrator) + calibration uncertainty. But if you add all these errors, you will be outside of the K2000 spec. That's why you need use higher level (best to be 10 times more accurate, but not always possible/viable) standard/reference meter to make sure total errors are still inside the spec of DUT.

 

[hwj-d]

Look at this thread:
https://www.eevblog.com/forum/metrology/keithley-2000-adjusting-single-ranges-possible/?all

Conclusion: you can't do it the easy way.
TiN does it, as i understand, correct me, by externally describing the eeprom with particularly single calibration data, but after saving the original.

[jfphp]

And the power supply spécifications ?  Stability and noise of an ordinary power supply cannot match with a 61/2 voltmeter !

[petemate]

Just think of it.
You take reference meter (R) and unknown meter (X), connect to the value source (Z).
Done the measurements, using same settings, for example.
R = 10.005643 V, which means true Z signal is +/-noise of the source and +/- uncertainty of the K2000. Assuming that K2000 reference has recent and valid calibration.

For clarity pick worst case and assume, you hit the largest devication of the spec + and add to that largest deviation in source noise/stability in + too. Now add uncertainty of the calibration to that too.
Now take X and get = 10.003152 V.

As result you can tell if meter X is within K2000 spec + source stability (which is for even good PSU many times worse than calibrator) + calibration uncertainty. But if you add all these errors, you will be outside of the K2000 spec. That's why you need use higher level (best to be 10 times more accurate, but not always possible/viable) standard/reference meter to make sure total errors are still inside the spec of DUT.

 

Thank you for replying. I don't get why the stability factor of the PU isn't eliminated when measuring it with the calibrated meter?

Look at this thread:
https://www.eevblog.com/forum/metrology/keithley-2000-adjusting-single-ranges-possible/?all

Conclusion: you can't do it the easy way.
TiN does it, as i understand, correct me, by externally describing the eeprom with particularly single calibration data, but after saving the original.

I don't have the equipment to do this..


Edit: Just to be clear, consensus is that I could perform a calibration like described above if the reference multimeter was at least ten times better than the multimeter to be calibrated, right? E.g. a Keithley 2001.

[Kleinstein]

Besides a better meter, one also needs a reasonable stable /low noise voltage source. With a normal, usually rather noisy PSU both meters would show rather scattering numbers. So it would take quite some time for averaging and ideally some care to have the two meters really measure at the same time (in sync), which is not that easy with AZ modes.  One could however use a simple set of batteries as a short time low noise source.

The DC Volts part is the relatively easy one, but there are also current ranges and AC mode. Especially AC might need a suitable generator.

If one uses just a second K2000 for calibration the uncertainties would be higher due to the added uncertainties.  It could still be done for an initial step / test if the old calibration data is lost.  A poor accuracy cal might be better than just a random new reference. However the AC part might still be OK - so partial cal would be nice if possible.