Keithley 2100 Resistance Error - Current Source

[namwos]

Hi all,

Picked up a Keithley 2100 from eBay which fails resistance calibration. DCV, DCI, ACV, ACI, Freq, Cont, all work perfectly.

The unit displays OVLD in every range with no load. With a short circuit, it measures about 2R in 100R range, 0.2R in 1K range, etc.

It seems that the current source is not outputting the correct current: but seems to be outputting 14mA in all ranges. Therefore only by applying a tiny resistance between HI-LO terminals can the voltage be brought down to below the supply rails of the op-amps and thus doesn't display OVLD.

I've been looking at the board and trying to figure out how the current source is meant to work. Attached a photo of the general area. Has anyone has one of these units with a similar fault? Particularly, I am unsure about the SOT-23 BJT ladder (these are marked 2Z).

Additionally, I have found an SOIC-8 chip which I cannot identify, marked M3500-TF812. From resistance measurements across its pins, it appears to be some kind of resistor network. Can anyone identify it?

Jon

[Mickle T.]

M3500-TF812 is custom TF resistors network.

[namwos]

Thank you. Do we know anything about it, resistances etc?

Also, why would such a thing be used rather than discrete passives?

[Mickle T.]

We are know nothing about it And about TF811 too.

[Kleinstein]

Unless you have documentation (schematics) or measure it, it would be difficult to get the exact resistor values.

They use these resistors as the matching in TC of the resistors can be quite good. So when it come to a stable resistor ratio such a custom resistor array can be cheaper or better than single resistors.

Would not expect the resistor array to be broken. I would more like expect a broken OP or FET, maybe a cold solder junction or tin whisker.

[namwos]

Thank you. I agree the network is unlikely to be faulty - I am just trying to put together a schematic of the current source and this part is a bit of a mystery.

Does anyone have any clues as to that network of 8 BJTs (MMBT6520)? I will draw out a schematic shortly.

[xani]

Any progress in your repair? I have unit having exactly same problem

[Oldtestgear]

Mine has the same fault. Any updates would be really appreciated as I am lost on this one.

TIA

Phil

[giosif]

If it helps, the Keithley 2100 is based on the Array M3500A.
When I had a 2100 with resistance measurement problems, I contacted Array and asked nicely for the service manual of the M3500A and they kindly provided it.
I suggest you do the same (I wouldn't mind sharing the manual I got, but I don't know if I am supposed to, so better go to the source).

For my meter, one or two op-amps were dead, but I am not sure if that was related to the resistance measurements issue.
In any case, after some troubleshooting, I ended up with the metter sourcing the correct currents for the corresponding resistance ranges, but the resistance measurements were still off.
After checking and re-checking all I could think of, I decided I would do a calibration for resistance side and, what do you know, that fixed the resistance measurement problem.

[Oldtestgear]

Email sent as you suggested. Thanks for the tip. Greatly appreciated.

Phil

[xani]

Apparently it appears unlisted (not in the download section, but in page index) on Picotest page:

https://www.picotest.com/downloads/M3500/

[Mickle T.]

This service manual of the M3500A is almost useless without a schematic diagrams 

[Oldtestgear]

But it does have a parts list..........

Phil

[amps]

Thanks all forthe comments in this thread, particularly the M3500A service manual. I've just managed to fix my 2100.

Mine was outputting just over 18mA on all current ranges, had an appreciable offset on current and the 100V range well out of calibration. I probed the current source components carefully and eventually determined there were two leaky JFETs, Q401 and Q306. I also replaced U401 before replacing these. I now get the expected current source values and the resistance readings look reasonable.

Probing backwards from the input to the ADC for the 100V range the signal seemed to be the correct level, not near the displayed value, througout - so I figured it must just be calibration. I therefore did the zero calibration and just this voltage range and all seems good now.

I checked versus a Datron 4705 (which admittedly isn't recently calibrated either) and the readings look very plausible. I suspect the resistance ranges are out of calibration now. The Datron is a very nice toy, though not mine and was last calibrated in '09 so I remain to be convinced that I'd lower the uncertainty much if I decided to further calibrate my meter with it.

[essele]

Hi,

Sorry to bump an old thread, but I have a similar problem ... I'm seeing 16-17mA output on all of the resistance ranges, and all the resistance ranges and DCV show OVLD (unless very low resistance.)

In my probing so far I'm seeing a problem with U507 (LTC1050), this seems to be setup as a voltage follower and yet I'm seeing +0.6v on the non-inverting input (relative to low rail) and just a few mV on the output ... the output looks like it just goes to R524 which is 475R so it doesn't seem like it's being pulled down, although I could be missing something. U508 seems to be fine, as does U201.

Therefore my assumption is U507 is dead.

I just wanted to see if my logic seems sound before I fork out on replacements parts and try to get the thing out without screwing anything else up??

Thanks for any input.

Cheers,

Lee.

[Kleinstein]

I would be a little surprised to see an LTC1050 as a voltage follower in the Ohms current source. It may still be used, but more like controlling a fet.

From looking at the photos my guess would be more that U507 would be an bootstrapped follower for the input, maybe similar to the input of the K2000.

I looks very much like the string of SOT23 are the protection for the Ohms source and Q401,Q402 and U402 would be the main part of the current source. R407 / R408 could be relevant resistors.

[essele]

Thanks for the input. With some more digging I think there are likely a number of issues...

Sticking to DCV for the time being ... on the <=10V ranges I'm seeing the correct input at U508, yet on the other ranges I'm getting -16v. With a bit more probing I can see a problem similar to https://www.eevblog.com/forum/repair/keithley-2100-repair-jfet-issue/msg1359980/#msg1359980

Q203/204 seem to be suspect, I'm seeing a constant -16v on the drain of these which means the voltage divider is then completely thrown off.

Given other people have seen the same issue it seems like a good place to start -- although I still have my doubts about U507 I can definitely see situations where it's not behaving, so that will be item 2 on the list.

I'm also slightly worried about U504, this is setup as a dual voltage follower, the first one seems to be fine, but I've got a concern on the second one where it also looks like it's driving to the negative rail even with a 0v input. More investigation needed on this one while I order some bits.

[essele]

More progress...

I've replaced the two JFETs (Q203/204) and that has solved the constant -16v on the drain problem. After this it went from 10 errors down to 8, so that's good.

DCV is now working on <= 10V, and I get 100x too high a reading on >10V ... so still problems with the divider.

I then replaced U507 (LTC1050) and the errors went up to 9, sigh. I am seeing slightly different behaviour, so I'm still unsure whether this was really faulty or not ... I may solder it into a breakout board and experiment.

BUT ... the gate of Q203/204 was constantly at -18v, so the divider was never actually being properly connected to GND ... after a bit of probing I realised that the gates were actually shorted to -18v, so I began a slow and painful look for something shorted. Eventually I found U503 (LM339) seemed to have pin 2 (OUT1) shorted which didn't seem right ... it seems logical that a comparator is used to drive the gates (amongst other things).

I've removed it and the short has gone, so I'm now just waiting for some of them to arrive.

I suspect there are other faults, because I don't see this addressing the resistance issue ... one step at a time.

[essele]

The good news ... DCV is all fixed and working fine, and looks to be pretty well calibrated. Replacing U503 did the trick.

The bad news ... resistance is still not working properly, I've replaced Q401 and Q402 ... Q401 definitely helped, 402 was desperation on my part, I think it was ok. I've also eliminated U507 (LTC1050) by removing it and the meter actually runs perfectly fine, not sure if it's used outside of the test mode. So I've probably replaced a perfectly good part here.

Anyway, I now have the 100R, 1K and 10K ranges seeming to be reasonable, the 100K and up ranges read progressively lower ... if I'm being picky then I think the 10K range is slightly low also, so I have a suspicion about the resistor network.

I think I've figured out most of the constant current source ... U402 with the 10v reference creates a 500uA or 50uA constant source which it feeds through a 2K resistor (in U403), this creates a 1V or 100mV drop which the other half of U402 re-establishes across the other resistors in U403 (switched in by U401.) In the pack, along with the 2K, there are 1K, 10K, 100K and strangely 160K (which I'm confused by, it should be 200K, but this may be the issue.)

So that should give you 1mA, 100uA, 10uA, 5uA, and 500nA sources. I can measure the drops across each of the resistors so can see that the switching is all working and I get a pretty constant 100mV or 1V depending on the range.

The problem really seems to be the 160K ... if I measure a 100K resistor on the 100K range it reads 68K, so that would point to that resistor being wrong ... however on the 1M range it drops to 50K, and a 1M resistor measures as 280K.

The 160K resistor does actually drop fairly dramatically when I measure it, especially when not on my meter's low power mode, whereas the 1K, 2K and 10K are rock solid, and I'm wondering whether it suffered as a result of the previous failures.

Does this sound feasible? It seems like a custom part, so I may have to bodge something together to test my theory, and help appreciated.

For info, pin 1 is common, with 100k to pin 2, 200k to pin 3, 1k to pin 4, 10k to pin 5, and 2k to pin 6. Pins 7 and 8 appear not connected.

Lee.

[Kleinstein]

To little current in the high ohms ranges could be some leakage current as well. So some current may be lost somewhere. This could be a leaky part or possibly just flux residue.

100 mV across 160 K would be a higher current, not a smaller current -  so having 160 K instead of 200 K would more like lead to reading too much current. Even it is would be 160 K in series with 100K the current would be not that small.

So what are the actual measured currents ?

Another point to measure maybe input current in the volts mode - some extra bias could also confuse the high ohms.

[essele]

Thanks Kleinstein, of course you are correct, it's been a long day ;-)

I had already removed the resistor network and checked it out of circuit and it's completely fine, so I've just been thinking about whether the Maxim switch, or the U402 could be the problem, but I think you are right about leakage. I will remeasure and post the currents tomorrow.

What do you mean about extra bias?

[essele]

Ok ...  problem identified!

Thanks Kleinstein, you got me thinking again about leakage and I started thinking about the protection string ... then I found the 34401A schematic which is basically identical for this part of the meter!

The only candidate was Q311 and as it seems to be used as a protection diode, I've removed it and all ohms readings are spot on and all the self tests pass!

It's an SST4117-T1 according to the manual, and these are discontinued. The 34401A appears to use MMBF4117A ... which interestingly are marked "61A" which is what was in there, so all good!

Parts on order!

Thanks again for all the help, this has been great fun. I just need to figure out how many parts I replaced unnecessarily ... and thank god I didn't start trying to bodge in a resistor network!

[Kleinstein]

MMBF4117 and SST4117  are essentially the same - SOT23 versions of the 2N4117, just from different manufacturers.

If the diode connected fet in the protection circuit is broken, it may be a good idea to do a quick check on the transistors in the string too. A simple in circuit test should be good enough.

After a repair in the input stage it would be a good idea to check the input current in DCV mode. The simple version is from the voltage read with a 10 M or 1 M resistor. To also check other voltages one could use a good quality capacitor (some 10 nF PP or PS type) and check the drift rate seen if only that capacitor is at the input.

[essele]

So if I understand correctly you are suggesting measuring DCV input current by applying a voltage to the meter, but with a series resistor so we can measure the voltage drop across it and hence measure current?

If so, then using a 1V supply, with a 1M resistor, I get about 83mV drop, thats 83nA, which is seems comparable to a quick test on another couple of meters, although doesn't seem to marry with the >10G input impedance from the datasheet, unless I'm measuring or calculating this wrongly.

Is that what you'd expect?

I've checked the transistors and they seem ok.

[Kleinstein]

For the quick input current test I would not use the extra voltage, so just the 1 M resistor across the input. Some 80 nA of input current would be way to high (if measured correctly). At such a high current the difference with 1 V on top should still work. Usually it should be less than some 50-100 pA. The >10 GOhms input impedance is usually meant as differential resistance - so change in current with change in voltage.

80 nA at 1 V make me think that the 10 M input resistance may be still enabled (input not set to high impedance) or the replaced FETs at the divider may not not turn off all the way.