Keithley 2000 repair - new reference?

[petemate]

I got my hands on a Keithley 2000, which is great.. Except that it doesn't work. Menus and buttons work quite fine, but the measurements are way off. There is no relation to any input(neither voltage or current). Running the self-test, I get an error already at 100.1(the first test). It is related to a problem with the AD conversion. The service manual suggested to first check power supplies and references, followed by U165(The ADC itself). I started by checking the power rails and they were all fine. Then I measured the reference, and it doesn't put out the 6.95V that the LM399 should. In fact, its something fluctuating around 0.5V. I'd say that the reference is somehow damaged. I don't know why that would be the case, but the measurement doesn't lie. Power to the reference is fine.

My question is related to changing the reference. Is it straight forward? Of course all calibration factors would become void, but apart from that is there anything I should be concerned about? I read somewhere that the reference is "especially selected" or something. What does that mean? Could I just buy a LM399 from digikey and replace the unit?

[Kleinstein]

The original refs are likely burned in, by running them in a test circuit for some time and checked for properties like noise and maybe drift in the initial phase. So the worst of the LM399 are likely sorted out.  An of the shelf LM399 might work, but with a slightly higher chance of not so good performance. One could do the noise test in circuit - so if the DMM is too noisy with the new reference, one might consider getting a second try.

Changing the reference is kind of straight forward. However the LM399 might show some drift for the first few months of operation. So it would be a good idea to plan for calibration after those first few moths. The initial adjustment would be less accurate and likely not worth an official cal.

There is also a slight chance that the circuit around the LM399 is not working. With just 0.5 V this might be just from the startup path. So maybe check the ref amplifier around U139, if the +14 V level is really at twice the +7 V level.

[saturnin]

I have encountered failed LM399 twice. One had defective the reference part (zener voltage around 8V or so) and the other one had shorted heater (which even resulted in partial melting of the heat shield). So LM399 does fail. Still it is rather unusual, so I would check TF-245 resistor network and U139 opamp (LT1124) at first.

LM399's reference is fed through R216, so if TF-245 and U139 are ok you can try to remove it and add a temporal resistor (~ 2 kOhm) parallel to R315. This way the reference will be powered from +15V rail. If there is not ~7V across the reference, it is most likely dead.

I can confirm what Kleinstein wrote about the selection of references - based on my experience they are certainly selected to have lower noise than average part you would buy from a public supply chain.

[alm]

Pretty sure they burn in those references for a while to make sure the instruments will meet their long-term stability specs, to remove some of the initial drift, and as has already been mentioned to remove the noisy ones.

I would definitely second the suggestion of not shelling out for an expensive calibration without burning in the new reference for at least a couple of months.

[petemate]

Thank you for your replies. I have not had a look at the associated circuitry yet, so I'll definitely have a go at that before replacing the reference. Will report back some time after New us with updates. Merry Christmas!

[Jens01]

I stumbled upon this topic and i recently bought a unit with exactly the same issues as yours. My 2000 also outputs error 100.1 (and many more ), which indicates the ADC ASIC (U165) (or the communication path between asic - mcu) is broken. In my case the ASIC is broken and the reference was outputing 0.65V, just like yours. It seems like the ASIC is needed to start up the reference: when you place a resistor of 3 - 6K parallel to R315, the reference voltage (and +14 etc.) should come up. I think there is some sort of start-up bias circuit controlled by the ASIC, but it is a bit unclear to me how it exactly works.

[petemate]

I stumbled upon this topic and i recently bought a unit with exactly the same issues as yours. My 2000 also outputs error 100.1 (and many more ), which indicates the ADC ASIC (U165) (or the communication path between asic - mcu) is broken. In my case the ASIC is broken and the reference was outputing 0.65V, just like yours. It seems like the ASIC is needed to start up the reference: when you place a resistor of 3 - 6K parallel to R315, the reference voltage (and +14 etc.) should come up. I think there is some sort of start-up bias circuit controlled by the ASIC, but it is a bit unclear to me how it exactly works.

Hi Jens, sorry about the late reply. I haven't gotten around to work at it anymore, but I will definitely keep you updated. Last statues is that I can get it to start-up with a correct reference voltage. However, after some time the voltage does indeed drop to something like 0.5V. I can reproduce this fairly consistently, but like I said, I haven't gotten around to look at it any further. I am fixing both a 2000 and a 2700 at the moment, so I took some spares from the 2700 to use in the 2000, as it was more likely to get up and running. However, I finally got the parts i ordered(i was cheap and ordered from china, so shipping took a loooong time). Hopefully I'll have the time to look at it next weekend.

[petemate]

Just a little update here..


As some may know, I have both a 2700 and a 2000 that are in need of repairs. I worked on the 2700 for some time, as it was somewhat functional. It had trouble on voltage ranges, but current works just fine. I managed to get it working. It was a mix of Q105,105,108(the JFETs that switch different signals around), U115(The LM339 that controls the JFETs) and U121(The shift registers that control, among other things, what ranges to enable). I manged to get it working, except the lower ohms range, since I had no LM393(The 2700 uses an LM393 instead of an LM399 as U131).

I then started working on the 2000. It shows a problem with the reference that have been described previously. However, I could not recreate the reference voltage issue. It is now perfectly stable at about 7V. There is however another problem now. U165, the ASIC, switches around signals to the ADC and controls the integrator system of said ADC. It is not switching correctly. The data to the shift registers is not consistent, but rather in some sort of initial boot mode all the time. I deduced this by comparing with the 2700. I was concerned about U165 being defective, so I took the ASIC from the 2700 to test with. That turned out to be a lot of trouble. In short, I managed to somehow mess up both DMMs. Reversing the swap of U165 now causes the 2000 to not produce any clock or data at all. I can see that it is trying, but it is held high by something, almost like it has a defective ground connection. The waveforms are there, but instead of swinging between 5V and 0V, they now go between 5V and something like 4.9V. Like it can't pull to ground correctly anymore.

On the 2700, the data/clock still works, but now I ran into problems with an unstable 2.5V reference, which is used throughout the system by the comparators. Why that happened I don't know. I managed to locate the issue to the U131, which was already defective before(as mentioned above), but not in that way. Also, something else is wrong, cause the voltage readings, which before were perfectly correct, are now again erroneous.

Well, thats the status of the system. I suspect I'll work on it again whenever my frustrations have subsided and I have the new parts I need.

[Macbeth]

...since I had no LM393(The 2700 uses an LM393 instead of an LM399 as U131).

Got me scratching my head there - I'm sure you meant LM339 as the K2000 U131. Of course the LM399 is the VREF in both models. (ok, maybe the higher spec defunct 199 or 299, but you catch my drift)

[petemate]

...since I had no LM393(The 2700 uses an LM393 instead of an LM399 as U131).

Got me scratching my head there - I'm sure you meant LM339 as the K2000 U131. Of course the LM399 is the VREF in both models. (ok, maybe the higher spec defunct 199 or 299, but you catch my drift)

Yes, you are right! The 2000 uses LM339s, while the 2700 uses both LM339's and LM393s. Sorry about that.