Keithley 2002 repair help

[Kleinstein]

The AC signals don't look very good, but also not that bad. So it is hard to tell if this could be the cause of the problem or just normal.

Due to the charge pumps (LTC1043) on the K2002 board (1 should be used to generate the 14 V ref level for the ohms circuit, not sure what the second one is good for), there is a chance that some spikes are normal for the K2002. However to much could cause trouble in precision DC circuits. One should be able to see which spikes originate from there, by checking the frequency / timing.

At that level, some could be a problem of where the ground is connected - there might be comparable background between the com terminal and corresponding ground on the board. It somewhat depends on the environment and scope, how much "noise" is picked up. Finding out if such small spikes are real of an artifact from the measurement is a little difficult and where experience makes a difference.
However it looks like the noise spikes at the amps part are larger and thus very likely real - possibly the source of the spikes. I see no legitimate reason to have them there, except if there is a chopper amplifier used for the amps circuit.

It might be worth comparing at least the measurement at the amps terminals to the working K2002. At least the large spikes look suspicious. The large spikes could cause some current in protection diodes as a bootstrapping/guard amplifier might not follow that fast. A possible cause could also be the guard/bootrapping amplifier in the amps section oscillating. Also the DC offset looks a little high, especially in the 200 µA range. 62 µV at an 1 K Shunt would be 62 nA (about what the DC reading is).

Have you checked the AC reading of the "bad" meter ? No need for a throughout test, but is it at least reading near zero on something like AC amps or AC volts with a shorted input. The AC circuit could be another possible source for higher frequency spikes (e.g. the rectifier circuit switching hard between positive and negative).

[nikonoid]

Here is an updated OHMS schematics, version 3. Changes to V2 are fairly minor and include Q226 transistor. By the way I still cannot figure out what that transistor is. It seems to be a matched pair with Q227. If someone knows, please post that info.

In trying to troubleshoot AMPS functionality, I also traced most of AMPS circuitry. It is very similar to K2001 schematics with addition of an opamp. So I used 2001 schematics as a basis and just changed what needed to be changed.

Let me know if these look ok.

TiN, if you have a chance please incorporate changes to the nice drawings you did. Thanks.

[nikonoid]

Back to troubleshooting. It appears that problems with AMPS and OHMS are NOT related to each other, but both seemed to be related to temperature.

When starting a cold meter it would start reading -6.3uA with open inputs. It quickly comes down to -2ua, -1ua and so fourth as meter is warming up. If I blow some cold air in the general direction of the APMS components the reading goes back up. To me that sounds like a clear proof of temperature dependency.
I measured voltages relative to Input LO while meter is still cold. Please see attached file.

 
For OHMS, I used hot air soldering station at its lowest setting of 100C and medium airflow. Blowing hot air focused around U232 for about 5 seconds fixed 2MOhm range readings (+/-40ppm of perfect). Blowing for 5 seconds more did not change reading from being good. I am still not sure if one of components is temperature sensitive or contamination on the board is responsible for this behavior. While hot air fixes OHM issue, AMPS reading problem appears unaffected, so they seemed like independent issues.

Klienstein, seeing this temperature dependency should I still further pursue AC analysis?

[Kleinstein]

The Ohms section working only when warm is kind of odd. Usually leakage gets worse with higher temperature.
Also the leakage current is really high (like up to 500 nA) this is a lot for leakage due to contamination.
Just looking at the DC readings would strongly point to the max326 already changed. So the 2 main suspects are leakage due to contamination/burned board and some AC effects. Checking for AC trouble might be easier, at least a quick check. There should be very little AC visible around the Ohms circuit, especially at Q213 (with a 100 K or 1 M resistor at the output), output of U233, emitter of Q251, U223 pins 9 and 11. No need to measure around Q214 this is close to having the probe in the air.
Instead of directly the output, one might use the low value 4 wire shunt as ground. How are the old measurements done (e.g. Probe). the old measured 25 mV_pp at Q213 might indicate a problem, but could be just an artifact from poor ground / probe. For a precision instrument I tend to prefer a quieter signal.

As the leakage problem got slightly better with cleaning, there is still a chance it could be just contamination. Residue from the caps might leave behind rather thin layers that might not be so visible. For cleaning DI water (usually plenty in several steps) might be more effective than IPA or similar. The main suspect would be the area around U223.

The extra transistor (Q226) seems to be part of the high ohms part. I don't think it is really related with the problem, though it might be nice to have the Ohms schematics complete. It is kind of separate from the normal ohms source, more like having the 14 V reference (or ground) connected through a large series resistor.

The amps section look rather conventional, though leakage in the µA range would be a lot even for the DG411. One suspect would be U234, especially the 2 gates connected to the DCA line. There might also be leakage from the DCA side. It is likely going towards some precision OP or maybe the main mux before the gain stage. Again a lot of leakage from the board might be the culprit too.  For surface leakage it would be more like relative humidity than temperature that is the culprit. So cold and really dry should be similar to warm.

[nikonoid]

I already replaced U234, U229, U225, and other and cleaned everything under them, so I had to look elsewhere.

I was just checking the board again and Q217 caught my attention. The source to drain resistance was about 4Ohm in both directions and gate to source and gait to drain measured between 52k and 54k. Not sure is this is OK. Does it make sense to you guys?

The solder on leads of Q217 looked dull and pitted. Since one of the legs is sitting at -15V, I decided to clean between the legs of this transistor.

Immediately after cleaning the AMPS performance improved to almost perfect state:
On cold start:
200uA    +000.0087uA
2mA       +0.000121mA
20mA     +00.00116mA
200mA    +000.0183mA
2A           +0.000103A

After 5 minutes, readings stabilized around:
200uA    +000.0037uA
2mA       +0.000043mA
20mA     +00.00044mA
200mA    +000.0050mA   (this reading was a bit strange, it would go as high as +000.0095 when turning this range on and then go back to 000.0050 fairly quickly)
2A           +0.000042A

These reading did not change after additional 30 minutes of warm up, so calibration might be able to take care of this offset.

This seems like a proof that the contamination was the reason for trouble with AMPS. I will try to do more cleaning with DI water and IPA.

Kleinstein, I will try your suggestions tomorrow. Thank you.

[TheSteve]

Sounds like you're making great progress.
FYI I've had a power supply from an HP network analyzer before with caps that leaked. No amount of alcohol would remove the contamination film on the PCB. I couldn't see the film, but an ohm meter could measure it. I had to scrub with water etc and then finish with lots of alcohol to bring it back to life. And the power supply was likely nowhere near as sensitive to contamination etc compared to your K2002.

[Kleinstein]

The critical of the max326 used in the ohms section is U223 according to the plans (TiNs and the last one below) shown.

If even the rather large amount of leakage in the amps range seems to have been due to leakage, chances are the the ohms part leakage is also due to contamination.  So my suggestion would be a little more cleaning, especially around U223 and the 75 K ref resistors. Even if the first try on cleaning would not solve the problem, any significant change would give an indication that contamination is the culprit.

[nikonoid]

Thank you very much, Guys.

Today I did a bit of DI cleaning followed by IPA and now it is time for some baking    at 130F   (see attachment)

Should 2 hours be enough?

[floobydust]

That's great news. It was a lot of work.
After an IPA wash, a hour in heat should be fine. It's just the water (if not 100% IPA) needs time to evaporate underneath IC's.

I've never seen leakage on a PCB like you were observing.
Unless they used the incorrect flux/wash, or creep corrosion is occurring. Not to be confused with tin whiskers.
Creep corrosion is usually blackened copper in dendrite-like patterns.


I have Keithley 2001,2001 service manuals but they contain no schematics, just long flow chart of parts to replace, kind of gross.

[Zucca]

I have Keithley 2001,2001 service manuals but they contain no schematics, just long flow chart of parts to replace, kind of gross.

Here the schematics K2001 and tons of info about the K2001

https://xdevs.com/fix/kei2001/#fix_2001

Back to the K2002 repair, nikonoid go on like this... you will win for sure!

[nikonoid]

Yes, Guys, I used TiN articles a lot.

This saga is slowly coming to the end. After cleaning with DI water, IPA and doing some fine baking, I gave meter a day to rest and cool.

Testing today showed that almost surprisingly OHMS function is working now.   It reads about 100pm high on a cold start, but then in just 5 minutes gets to good readings. At the same time AMPS results got a bit worse after this round of cleaning, but not much. After 30 minutes warm up the meter is pretty usable, considering it is running without top shield, metal case, etc.

So now I am pretty confident that everything that had to be fixed is fixed. CONTAMINATION is the major problem that is left!

This meter was missing case and top shield when I got it and was literally full and dust, insects and mice litter. The fact I could get it to this point is somewhat unbelievable. What else can I do to clean contamination, short of removing all the components?

Thanks.

[Kleinstein]

Congratulations 

If cleaning worked so far, it might be worth a second round. Dissolving contaminations is sometimes slow. So it might take quite some time and DI to get rid of some stuff. One could try a rather dilute acid - it can help with some salts / deposits. With now not much lower leakage, there should not be very much dirt left - so maybe just a few more steps of dilution.

[Zucca]

My .02$: Soak it in a bath of 50% distilled H2O and 50% IPA... leave it there for a day or two... then rinse it with fresh distilled water and finally in oven again. repeat if necessary.
Maybe an antistatic brush could help.

[nikonoid]

I have done special oil free air dusting and anti-static brush before using liquids. I wanted to minimize dirt flowing under microchips.

As the bath is concerned, could some elements be damaged by it? Like mechanical relays, voltage reference, large capacitors and maybe others.

Is shallow bath, right to the top surface of pcb, a good option?


Sent from my iPhone using Tapatalk

[Kleinstein]

Relays could have a problem when the full board is in a bath.  It depends on the types - some are washable, some are not. Especially reed type relays could be a problem due to increased leakage if the glass gets contaminated. The caps should not be a problem. The tricky part could be leakage under chips or relays. There can also be residue moved from the dirty power supply part towards the more sensitive parts (Ohms and volts input). So I would prefer more local cleaning.

One could use the liquid locally, like a few drops or small spill, so it would only effect the area of interest. Local washing would be adding a few ml of water and remove it with a syringe or similar. So more like a shallow local bath, that is repeated several times. For the parts between chips the brush could also be good option for an initial cleaning.  One can hardly avoid the part below the chips to get wet. Compressed air to blow the liquid out can be tricky, as some will end up as small droplets in other areas. It might still be the best option. The idea is usually to use enough water to dilute the dirt so much that there will be very little left. IPA or similar is than to a large part to dilute the residual water and allow better drying.

[nikonoid]

With input from the forum I did a "sponge bath" for a meter. I would saturate small section with liquid for a minute or two, then wipe it out with dry q-tips. I did it over and over again working one section at a time.

Two days later the meter is now in close to PERFECT condition. It passes all self tests even on cold start.
Current still have a bit of a drift on start and stabilizes after 10 minutes. Ohms seems to read right. Again about 20ppm drift from cold to warmed up. I also noticed noise during OHM measurements of ranging from +/- 1ppm to about +/-5ppm. This might be as good as it gets considering I am running meter without top shield and case.

The plan now is to finish recaping, buy shield and case from Keithley and do full evaluation of the meter. I can take it through few more rounds of cleaning, if becomes necessary.

Thank you everyone for staying with me on this and specifically thank you Kleinstein for your continued help and comments.

[saturnin]

Although it is definitely great progress (considering the initial state) and passing self test may be good indication there is no critical failure, I would still hesitate to declare the unit anywhere near to "perfect" condition. In such old units there still might be hidden serious issues (intermittent glitches in readings, slow long-term drifts...). I saw multimeters which were passing built-in self test with flying colors, but after detailed verification of their performance it was obvious there was an issue. It could take weeks or even months to find and solve these ones...

[TiN]

Well, that to be determined by full low-level calibration and verification (which in turn require access to 5700A + booster) .

[Le_Bassiste]

Although it is definitely great progress (considering the initial state) and passing self test may be good indication there is no critical failure, I would still hesitate to declare the unit anywhere near to "perfect" condition. In such old units there still might be hidden serious issues (intermittent glitches in readings, slow long-term drifts...). I saw multimeters which were passing built-in self test with flying colors, but after detailed verification of their performance it was obvious there was an issue. It could take weeks or even months to find and solve these ones...

so, how would you actually declare this unit to "anywhere near to "perfect" condition"? what are your criteria to do so?

[Zucca]

With input from the forum I did a "sponge bath" for a meter. I would saturate small section with liquid for a minute or two, then wipe it out with dry q-tips. I did it over and over again working one section at a time.

Which liquid? 50% distilleld H2O and 50% IPA?

Congrats Sir nice job! 

Although it is definitely great progress (considering the initial state) and passing self test may be good indication there is no critical failure, I would still hesitate to declare the unit anywhere near to "perfect" condition. In such old units there still might be hidden serious issues (intermittent glitches in readings, slow long-term drifts...). I saw multimeters which were passing built-in self test with flying colors, but after detailed verification of their performance it was obvious there was an issue. It could take weeks or even months to find and solve these ones...

Yes you are right, but I believe nikonoid will not give up  easly.  Moreover the probability a SELF PASSED unit has more problem inside is cam'on small. It happen to me too, but it was a single event.

[saturnin]

I am sorry, but "self test passed" is necessary but not sufficient condition to say that K20xx units work perfectly. As I wrote performance/stability verification needs to be done. This is especially true about nikonoid's unit which was affected by severe leakage due to surface contamination (0.5 uA!). Ideally, F5700 or similar calibrator should be used, but even hobbyist can reveal a lot issues by using a decent set of DYI standards.

If I were at nikonoid's position I would check long-term stability of all DC and ohm ranges (because they have tightest specs) before declaring the unit to be close to perfect state...

I would like to add it is not so rare a DMM has an issue although it passes its self test. Let me mention two examples I solved some time ago:

1/ K2000 - very clean unit, never repaired before, it passed all built-in tests, at first glance everything was ok. During stability tests, I found out there were glitches in readings (see K2000_noisy_test3.png). Reason? A zener in an auxiliary power supply was dying...

2/ K2010 - very clean unit, never repaired before, it passed all built-in tests, very stable and quiet on DCV ranges. After some time, I realized there was a drift on all <10 MOhm ranges. (see K2010_ohms_drift_issue.JPG). Reason? Slowly drifting TF-245 resistor network...

That's why I don't exaggerate relevance of built-in tests...

[Zucca]

Thanks saturnin, very interesting, yes you are right.

Now it looks like you have more experience than me, quick questions: how many devices you saw passing self test and have no problem then? More/less then 90%?

Thanks!

[nikonoid]

I might have overstated the "perfect" state.


Saturnin, K2002 self test conditions are very wide. I think some of them are as wide as 3% accuracy, so I am completely aware that passing self test is not guaranteeing anything. My assessment of the condition was based on measurements outside of the self test. I got it to provide believable, relatively stable readings, considering absence of shield and case.

Questions of calibration, time stability and especially temperature coefficients still need to be addressed. My other K2002 is @CalMachine, for low level calibration, so it will be fun to compare them against each other when it is back.

TiN, I wish I had 5700....     However I can check stability and consistency even before going for low level calibration. I am also quietly experimenting with a DIY temperature controlled chamber.

Zucca, I did use 50/50 mix. My previous attempts with 99% IPA were not as successful. By the way, my original K2002 was sometimes failing self test. In the end it was MAX326 chip that sometimes worked properly and sometimes not. So I actually share the saturnin's concern. Self test, especially if it ran just once, does not guarantee anything.

[saturnin]

Saturnin, K2002 self test conditions are very wide. I think some of them are as wide as 3% accuracy, so I am completely aware that passing self test is not guaranteeing anything. My assessment of the condition was based on measurements outside of the self test. I got it to provide believable, relatively stable readings, considering absence of shield and case.

That's my point exactly. Sorry, it seemed to me you relied on results of built-in tests only. Thanks for posting updates on the actual status of your repair project! Hopefully, you will get fully functional and reliable unit in the end.

[nikonoid]

After fifth, yes fifth, round of cleaning and baking I did measurements for two modes that gave me problems before:
1) AMPS with open inputs and
2) 100k Ohms in 2M and 200k ranges.

Both measurements were done from cold meter (off for 24 hours). I did measurements without top shield but I borrowed the metal casing from a donor K2001.

Amps took 45 minutes to an hour to completely settle. After that it would fluctuate from time to time by about 3ppm of the range.

Ohms took a bit longer - 60 to 120 minutes to settle.

Can someone try something similar with fully functioning K2002 from cold, so we can compare? Thank you very much.


Today I am getting a parts 2001 unit that will donate the case and top and bottom shields. I got it for $150 - cheaper than replacement parts.

PS. The score so far on this K2002: Two blown capacitors, one shorted zenner diode, 3 traces killed by a hole burned in the board, two traces eaten away to electrolyte, two bad vias, and LOTS AND LOTS OF CONTAMINATION!