Keithley 2010 Repair

[BillyO]

This meter is still in production.  Parts should be available for at least the warranty period past the discontinuation date.  Probably longer as companies will have signed service agreements.

[KungFuJosh]

I updated the measurements with an out of circuit list, that's much closer to the others. Some are still a little off though. If it's supposed to be .1% or something like that, then it looks like a fail.

[BillyO]

Yes, looks much better.

On a meter like this with it's calibration capabilities I think stability would be more important than absolute accuracy.  Indeed I could be wrong, but for any meter that can build a calibration map for every range greater stability would surpass accuracy.  No?

[KungFuJosh]

Well, no, if that issue is what's causing the fail. I dunno WTF is causing the fails though, so... 🤷

Reading the other threads, and even checking the numbers I referenced, it didn't take much for the tests to fail. For that member, replacing the TF-245 resolved his meter's fails.

Besides that, accuracy should be able to be adjusted during a calibration. At that point, stability would maintain the accuracy, no?

[Kleinstein]

Not so nice, U138 (lack of) voltages:

U138 OPA177
1: +14.98
2: 0 (- in)
3: 0 (+ in)
4: -15.78
6: 0 (V Out)
7: +15.1489
8: +14.98

The voltages look very resonable. The supply looks OK and the inputs and output are supposed to be close to 0 V - there may be a few µV at the input and maybe a few mV at the output (to counteract the AD744 offset).

The resistor array is all about the resistor ratio stability and in some case the ratio, not so much the absolute values. To some degree R287 for the run-down fine slope can depend on the absolute value of the 30 K range resistors, but it may not be super critical. AFAIR there is a normal slope ratio of 128, but I don't know if the meter uses the nominal ratio for the result or the measured actual ratio of the individual meter. Some ratios can have quite some tolerance and the gain factor are anyway adjusted in calibration. An error in the resistor / slope ratio would cause some DNL errors, that would be especially visible with short integration. I don't think the self tests would catch this.

There could be different versions for the K2000 that uses directly a LM399 ref. and the K2010 and K2182 that use an extra low voltage zener.
If measured in circuit, there are also other parts that can effect the resistance readings.

[KungFuJosh]

Not so nice, U138 (lack of) voltages:

U138 OPA177
1: +14.98
2: 0 (- in)
3: 0 (+ in)
4: -15.78
6: 0 (V Out)
7: +15.1489
8: +14.98

The voltages look very resonable. The supply looks OK and the inputs and output are supposed to be close to 0 V - there may be a few µV at the input and maybe a few mV at the output (to counteract the AD744 offset).

Thanks! I thought it was weird they were all at near zero. I should be more patient or something.

[Kleinstein]

Without a describtion the self tests are a bit hard to interprete. The numbers give a crude idea, but hard to tell what is actually bad.
It may help to do some external tests on the meter - one would need thes later anyway.
A relatively easy one is checking for input bias current with a low loss capacitor and looking at the drift rate. This could detect extra input leakage - a not so uncommon kind of damage.

Another easy point is to look at the noise / drift with a shorted input.

An ADC issue with the small ref. step / slow slope would result in DNL errors that would repeat and would be more visible with relatively short integration time. A possible simple DNL test is looking at a very slowly drifting voltage (e.g. slow dischage or a large electrolytic, or drifting batttery), using the 1 PLC AZ mode. One could than subtract a smooth fitted background (e.g. 2nd order polynominal, or offset+ exponential) to make DNL and hard INL errors visible for small zoomed in ranges.

[KungFuJosh]

These are the descriptions of the two errors...

[Kleinstein]

With the ADC part at least somewhat working. The issue is likely more with signal routing than the actual ADC.
I could help to looks a the signal (test point) at the ADC input.

[KungFuJosh]

With the ADC part at least somewhat working. The issue is likely more with signal routing than the actual ADC.
I could help to looks a the signal (test point) at the ADC input.

I think you're referring to TP105. It's a little weird, but I dunno what it's supposed to look like.

[BillyO]

New scope?

[KungFuJosh]

New scope?

No, I got the Magnova last year. It's awesome.

[KungFuJosh]

My TF-245 is perfect. I just compared my values to the one in the schematic, and they're spot on.

There might have been a connection issue to it though, I'll have to take a look. A couple of the pads were messed up.

Thanks,
Josh

[Kleinstein]

TP105 looks like it is the ADC input. The waveforms look surprisingly complex. What is see there will depend on the DMM settings (AZ mode active or not, integration time) and of cause the input signal.Without that information it is hard to judge the shown curves.
There is quite some hum (likely mains) visible in one of the phases - that part should not be there, except from maybe an open input.

[KungFuJosh]

There is quite some hum (likely mains) visible in one of the phases - that part should not be there, except from maybe an open input.

That was scoped before I replaced the filter caps. I will check it again after I reinstall the TF-245. We'll see if it's any better.

Thanks,
Josh

[KungFuJosh]

I reinstalled the TF-245, and I had to bodge a couple of the connections, and now fail 400.2 is gone. I could be wrong, but something gave me the impression that somebody had tried working on this before.

For anybody wondering, the TF-245 is still available from Tek, and the current price is $37. I'll probably buy at least one to keep as a spare, even though I don't appear to need one.

Anyway, down to 101.2 ("NO 7V AT A/D").

I haven't measure the reference directly, but any other point I tried that should have the 6.9V present reads 6.2V. Is that enough of a deviation to cause the failure? It seems to me like it would be enough to fail, but I'm well aware that I don't know shite. 

Thanks,
Josh

[Kleinstein]

AFAIK the 2010 has the extra lower voltage zener and 6.2 V for this is at the lower end, but still OK. There is also the 7 V refrence (LM399) - so depending on the test step the HW could provide both.
The self test description could still be for a different meter or not fully adapted to the changes. So what was 7 V with the K2000 or 2002 could be 6.2 V with the 2010. The internal limits should take care of this, but the documentation may not.

Where is the information on the self tests from ?

[KungFuJosh]

Straight from the horse's mouth. I took the screenshots I posted previously from the K2000 service manual, but they agree with the meter.

You can press the Shift key on the meter at any fail and it will show the description.

Thanks,
Josh

[KungFuJosh]

To test the LM399, I tacked a lead on to pin 1, and measured it referenced to AGND. 6.9258V. So, the reference itself isn't the issue.

[KungFuJosh]

U139 LT1124CS8:

1: 0
2: -15.8V
3: +6.2V
4: +6.2V
5: 0 (unstable)
6: +15.146V
7: -5.936V
8: 0

Anybody with a working 2010 mind checking their voltages?

Thanks,
Josh

[Kleinstein]

The voltages at pins 5 and 7 depend on the zener diodes used for level shifting. These zeners are not critical in there voltage and may vary quite a bit between meters. Near 0 V at pin 5 is odd - it should be more like some 4-7 V. It could be contract issue with the probe.
Getting 0 V at pin 8 is a good sign. It means the OP-amp is in regulation and not saturating.
Are the +-12.x reference voltages as output of the ref. amplifier part OK and stable ?

[picburner]

Measurements on U139 (on for 10 minutes, input shorted, DC Volt)

1: 0.000
2: -15.799
3: +6.437
4: +6.437
5: +7.374
6: +15.174
7: -7.439
8: 0.000

[KungFuJosh]

Measurements on U139 (on for 10 minutes, input shorted, DC Volt)

1: 0.000
2: -15.799
3: +6.437
4: +6.437
5: +7.374
6: +15.174
7: -7.439
8: 0.000

Thanks!

U139 LT1124CS8 (on for 10 minutes, input shorted, DC Volt):

1: -0.3mV
2: -15.8V
3: +6.2V
4: +6.2V
5: +6.86869V
6: +15.16V
7: -5.93V
8: -0.4mV

I thought it was weird that shorting the input made such a difference, so I pulled the short and nothing changed. Maybe I didn't let it warm up enough (or maybe it was on too long?), or maybe it was a probing issue yesterday. 🤷

Source voltages look good, but outputs are weird.

Now, call me crazy, but the part is rated to 70C, and was around 100C. Mayyybe I should replace it with the nicer LT1124AMPS8#PBF that's rated to 125C?

Thanks,
Josh

[KungFuJosh]

Are the +-12.x reference voltages as output of the ref. amplifier part OK and stable ?

Seems that way:

Voltages:
R271 TF-245
1: +12.4207
2: +6.21061 (too low, right?)
3: -12.4196
4: +12.4208
5: -12.4193
10: +/- 3V (unstable)
13: -24.85mV
14: -4.1384
16: +11.7987

I did not retest those voltages yet after reinstalling/repairing the TF-245 connections.

Thanks,
Josh

[Kleinstein]

The LT1124 has quite higher power consumption, as it has a rather high supply current. It is anyway a slightly odd choice, as there is no real need for such a fast amplifier at that position. However it is one of the few parts with a rotated pinout in SO8 - so no so easy replacement. AFAIK the LT1013 would fit, but it could be slightly high noise.
One could consider adding some heat sink for that part.
The temperature rating should be for environmental temperature and allow for a little self heating on top.