Keithley 2000 repair - fails in DCV/ACV/Ohms

[kloetpatra]

Hi. I have two Keithley 2000. One is working and I am using that as a reference for repairing the other one, which is showing the following behaviour:

• Built-In Test completes without any errors
• Amps & Continuity fully working
• Power rails looking good
• DCV values are multiplied by 20 at ranges: 10V, 1V, 100mV for positive voltages
• DCV values are multiplied by 10 at ranges: 10V, 1V, 100mV for negative voltages
• DCV values are multiplied by -145000 at ranges: 100V, 1000V for positive voltages
• DCV values are multiplied by -78000 at ranges: 100V, 1000V for negative voltages
• Ohms is also wrong by a factor. I assume fixing DCV measurement will fix ohms too.

For example:

• when applying 5mV at 10V/1V/100mV range it will display 0.10010V
• when applying 50mV at 10V/1V range it will display 1.01348V
• when applying 500mV at 10V range it will display 10.13488V

Autorange doesn't work for Voltage/Ohms and OVERFLOW is displayed due to those factors (except if you are using very small voltages, so they are within range after applying the factor).

Since I have a fully working Keithley 2000 (it has minor changes in the PCB layout) I tried comparing some measurements. Without any luck. I found no difference so far.
The following signals seem to be equal/correct:

• TP104 (DCVO)
• TP105 (A/D_IN)
• ISIG
• U142.6
• SIGN
• R131.1 - 16 (replaces R271?)
• R128.1 - 16 (replaces R271?)

It looks like the ADC is receiving the correct signal. Is it possible that the factor is applied by software due to bad calibration?
How can I check calibration without losing it? Is it possible to show the raw values?
How does the overflow detection work? Is it software too? I didn't find it in the schematics...

[ManateeMafia]

I haven't worked on one of these but I am interested in how you get the problem resolved.

When you apply the test voltages, does it appear to change ranges? For example, the 50mV  will give you the same value on the 1V as well as the 10V range? I would expect the decimal point to shift but it could be I am not understanding what you wrote.

Also, was the firmware upgraded? There is a note on xDevs that a calibration is needed if updating to A20.

https://xdevs.com/fix/kei2000/

[kloetpatra]

Yes, the decimal point will shift (when switching ranges manually). There is no significant error between the ranges, if the value can be displayed.
It has the latest firmware (A20) installed. I don't know if it has been upgraded earlier since I recently scored it on ebay with that overflow error described.

Unfortunately my two units differ slightly from the schematic available at xdevs. For example I do not have the resistor network R271 (TF-245). Instead of it, there are 4 other resistor networks:
http://www.mouser.com/ds/2/414/tt%20electronics_660-334130.pdf
R128: 668 A 5002B (8x 50 kOhms)
R131: 668 A 5001B (8x 5 kOhms)
R119
R126

I tried to do a new calibration, to see if it has a bad calibration stored.  But right now I do not have a +100V supply which is needed for one step. And without completing all calibration steps and saving to EEPROM,  it won't use the new calibration.

[ManateeMafia]

I am sure there are better ways but my first thought is a stack of 9V batteries.
It looks like this has already been covered. https://www.eevblog.com/forum/testgear/using-9v-batteries-in-series-to-calibrate-a-6-5-digit-multimeter/

[plesa]

Can you make a calibration data dump from both units and compare them?
 Try following comand :CAL:PROT:DATA?
or :CAL:PROT:DC:STEP3? for 10V

[saturnin]

Cal data obtained by :CAL:PROT:DATA? command could definitely reveal whether there is something wrong with the calibration or not.

Btw. :CAL:PROT:DC:STEP3? seems to be unsupported command. Or at least my unit does not like it 

It should be also possible to perform a partial calibration (over GPIB), as Alex Nikitin reported: https://www.eevblog.com/forum/repair/keithley-2015-repair-and-the-input-buffer-replacement/msg890947/#msg890947, so I would try to perform only STEP0-STEP4 cal steps of DC cal and then save cal constants...

[kloetpatra]

Thank you very much! I didn't know that partial calibration would be possible. 

After DC calibration Step 1, 2, 3, 4, 5  DC mesaurement is working again perfectly. 
Either the previous owner did change components without recalibration, or there is something I'm not seeing right now.

Major differences: cal value 39, 40, 41 (see image)
I appended the files if someone wants to have a look.

Now I will try to calibrate Ohms measurement, which is still not working.

[plesa]

It can be firmware upgrade or unfinished previous calibration.
Sometimes it is enough to restore previously backed up  cal data.
Strange is that instrument does not complain on such low values. Maybe it is not so sophisticated
If you have programmer you can upload data from your second DMM to see if ohms are working.

[saturnin]

@kloetpatra

So, it was wrong calibration, not bad unit - that's good news It also explains the Self-test was ok. Lines 26-28 in your table are values entered (by user) at +10V, -10V and +100V cal steps, while lines 39-41 are actually measured voltages (at A/D_IN probably) during calibration. They are expressed as multiples of Vref, which is internal reference voltage ~7V (i.e. 1.43*7V ~ 10V).

From "bad" cal data, one can derive real voltages connected to the unit during calibration:

+10V cal voltage ~ 0.0705*7V = 0.4935V
-10V cal voltage ~-0.1399*7V = -0.9793V
etc.

I am surprised it was allowed to save obviously bad values to cal ram. The multimeter can report “10 vdc full scale error” (+402) and “-10 vdc full scale error” (+403), so it should know something is wrong and abort cal process...strange