Keithley 2002 repair help

[Kleinstein]

I think TIN got a mistake around Q213 / Q210. The hand drawn circuit looks much more plausible. I also doubt that Q213 would be a VN0605 - it should be more a high voltage (e.g. 1200 V) MOSFET, as it is for protection against high voltage in ohms mode.

VR216 should be either the other way around or a zener diode.

[TiN]

Fixed a bit. I think "wrong" direction for opamps cause the perception rejection  . That thought crossed my mind...
Usually it takes 2-3 redraws for clarity once more blocks filled in. Anyhow still hopefully better than K2001's schematics, which are pure joy.

[nikonoid]

TiN,

Thank you for doing drawings. It is a lot of work.



Q213 is a "2SK1413" not "25K1413". That is another mistype in the repair guide.
I am attaching a list of errors I found so far.

[nikonoid]

The 100 K resistors don't need to be accurate, they are just there to cope with any leakage from CMOS switches above. One should measure the voltage on both sides of CR208 - ideally it should be very close to zero voltage over that diode, unless the resistor is chosen as the active one. The voltage should also very close to the point where the 4 precision resistors join together. As there is some extra current through the resistor, the polarity of the voltage over CR208 should tell it the current flowing this way (e.g. leakage in the diode) or the other way (through the other CMOS switches). In the 2nd case is would be de-soldering the switches U223: as it would be either that chip, or maybe residue - likely under it. Likely a new chip for U223 is than needed - as there is quite some risk it is damage (either from the beginning or when removing it).

If it is current through the diode, one could for a test remove CR208 - trouble is than more like with the other CMOS switches (U221-Q6).

Kleinstein,

I measured the voltage across diodes (100k resistor on 2MOhm range)
CR206:   .41V
CR207:    50mV
CR208:    .1mV
CR209:    -4.6mV

On a good meter I got:
CR206:   .41V
CR207:    -0.26mV
CR208:    -0.20mV
CR209:    -0.13mV

Broken meter seems inconsistent in terms of both magnitude and sign.


I am not sure though what these numbers are telling me, though.

PS. I have a spare DG411. I am ordering MAX326 now.

[Kleinstein]

The positive (though small) voltage over CR208 tells us, the 0.5 µA leakage current is not flowing through this diode (would be the other direction). The 50 mV for CR207 indicates there might be a problem it that path too (or maybe the diodes / readings are interchanged: 50 mV across the diode at the 75 K resistor would make more sense). Anyway this would point towards a problem with U223(max326), or maybe some leakage on the board around that part.

A slightly different voltage over the diodes for different meters is not a problem. A main contribution would be from the offsets of U228 and U233. Ideally the drop should be the same one all 3 non active paths. Differences would be due to leakage of U223 and less critical U221/U220.

[nikonoid]

This sounds right. Thank you. I mixed up the sequence of diodes. It should be the opposite sequence. The max326 chip is in the mail. The surgery is scheduled for Wednesday. 


Sent from my iPhone using Tapatalk

[nikonoid]

Today I replaced the Max326  (U223) with a new one. The situation did not get worse.... and also did not get better. The board underneath U223 looked clean. I cleaned it more anyways.
I am still getting same errors: 306.2, 500.6, 500.7, 600.2.

I guess next step is to desolder diodes. Any other ideas?


Sent from my iPhone using Tapatalk

[Kleinstein]

The direction of the leakage current is in a way that it can't flow through the diodes (higher voltage at the other side), unless there is massive photocurrent. So unless there are other ports connected, what I somewhat doubt, the current must go through the max326 or maybe leakage on the board.

As the current is through the turned off 75 K resistor, it can't be the OP following the max326. It is slightly strange that the 20 K range (using the 75 K resistor active) also has a current more on the low side - so leakage is not between the two sides of the switch in the max326, but more like towards the negative supply.

One might want to check the supply of the max326 - a wrong supply might cause trouble. One might need to use a scope here to check for AC trouble. A wrong DC supply level should give different effects.
Also a borderline / oscillating / out of bound (e.g. to negative) signal at the control input could be a cause. Oscillation in the current regulation part is another possibility.

It would make sense that the 2 ref voltages for the ohms circuit are generated from the charge pumps (LTC1043) - so AC ripple there might be a possible source of errors too. To understand which voltages are needed for the supply the value of the two ref level would be interesting too (could be 14 V and 7 V or 21 V and 14 V).

[nikonoid]

Thank you, Kleinstein. I will recheck power on the chips this weekend.

I am now looking to get an insulation transformer, so I can safely do oscilloscope measurements.

[Kleinstein]

Measuring on the analog side of the DMM does not need an isolation transformer. The analog part is isolated from ground anyway. Even with the digital part one can get around the transformer, accepting slight possible noise from a ground loop.

[nikonoid]

I did a number of measurements comparing good 2002 that I have and a Bad 2002.

I used 100k resistor on a 2MOhm range. Most measurements were done relative to Input LO.
On Bad instrument ohm reading was at least 10% low, sometimes more depending on how warm the instrument was.
Please keep in mind that all values from good 2002 were stable, while many values on bad 2002 were drifting with temperature and since readings were taken at different time they may not correspond to each other precisely.

I compared values and highlighted any place where I found significant difference. Please tale a look at scans attached and let me know what you think. Pardon my horrible handwriting .

Some most notable differences include Q251, Q226, VR204, Q210 - Q214. Thanks a lot.

PS. I have not had a chance to get to measurements with oscilloscope. It will have to be sometime this week.

[Zucca]

 

PS: Excel next time?

[Kleinstein]

It would be much easier to have to voltages from both meters on the same sheet.

The voltages around Q251 look a little odd. One thing is the rather high supply (collector or Q251) with the bad meter. So there might be a slight problem with VR216. However this should not cause this problem it would be more a problem causing higher temperature at Q251 in the low resistance ranges. Still worth checking VR216.

Ar the pins in the same sequence ? It is a little odd to have the base slightly below the emitter. This might an indication that there is some AC super-imposed. If swapped the two values would make more sense. Similar a base voltage around 15.3 V instead of 14.3 V (this would also match U231 pin 1)

The difference in the voltage around Q213 and so one could be a minor difference in CR217 ( different type of diode, with higher voltage drop on the bad unit. This normally should be of no consequence - just in case one could temporarily have a second diode (like 1N4007) in parallel.

So far the only part really indicating a fault is the voltage over CR208 and thus R272. The polarity of the voltage indicated the false current can not flow through the diode. If at all the diode would add a few pA to the output current. So far I only see U223 (pin 10) for the current to flow through - somehow leaking toward ground.

The DC voltages around U223 (max326 - the one that got replaced ?) look good. Still a little odd they supply +-15 V to the chip, as there is no need for a high negative supply and this would only add to leakage. But it seems to be the same way on the good one.

I would really look for AC trouble (like a control loop that oscillates or maybe spikes on the supply).
One might learn a little more about the missing current, with testing in the 200 K range. This would change the voltage at the critical node.

[nikonoid]

Thank you. I will remeasure everything on transistors and make note of pins in case I made an error before. I also prepared an oscilloscope to do AC measurements.
I did notice before when working on power supply that bootstrap voltage was a bit high. As per manual it was supposed to be +34V to +38V and I was getting something like +38.6.

This comes to be the most difficult repair I ever attempted.

[TiN]

This comes to be the most difficult repair I ever attempted.

I bet you learned a lot already as result. 

If you have more schematics works, I'd be happy to digitize as well., 

+38.6 sounds fine to me, I saw up to +42/-42 few times. It depends on line frequency/voltage a little.

[Kleinstein]

The slightly annoying part is not having a schematics. But this is now there for the Ohms part  .

The only thing that might want a recheck would be the voltages around Q251. The protection part so far looked good. And just for the DC part, it can't be the reason for the current through the 75 K resistor.

Worst case Q251 would see some 7 mA (don'T know the extact max current) at 24 V , thus about 150 mW. So it could get a little warm, but no too bad and it won't help that much moving some of the heat to VR216.

[nikonoid]

This comes to be the most difficult repair I ever attempted.

I bet you learned a lot already as result. 

If you have more schematics works, I'd be happy to digitize as well., 

+38.6 sounds fine to me, I saw up to +42/-42 few times. It depends on line frequency/voltage a little.

I added few small details to schematic. I am trying to collect more of them before passing them to you. Meanwhile, if you or anyone else has any pictures of 2002 analog board with components off it, please send me these pictures. Even if it is not OhmS section, if would still be useful. I am putting together a drawing for analog PCB, and sometimes what is happening under microchip or even resistor is hard to decipher.

If I find time, I may also draw something for amps section, as it has problems on my meter too.

PS. The design of Keithley 2002 reminds me a way how some software developers write code. They managed to squeeze performance that is close to 3458 into a space that is less than a half of it, if you take space for multiplexing cards into consideration. It is Smart, highly optimized and agile, yet very difficult to understand and troubleshoot.


Sent from my iPhone using Tapatalk

[nikonoid]

Another small report. I remeasured voltage at all the pertinent transistors. I did it twice, once on a cold meter and second time after a bit of a warm up. It seems consistent with numbers I posted previously: Transistors.JPG is attached.

It seemed that behavior of cold unit vs hot is very consistent, so I measured the progression of Ohms measurements as meter warmed up. It was on 2MOhm range measuring 100kOhms. See KOhms_Drift.jpg attached. Horizontal axis is seconds.

Then I measured amps on 2A range with open inputs, again starting from completely cold meter. See mA_Drift.jpg.
It looks like these two problems could related. After all, the electrolyte spill from a location point of view affected both ohms and amps circuitry.

Seeing that performance improved with a warm up, I let it run for 1.5 hours and then I ran self test and it passed completely this time, where before I got some ohms and amps related issues.

I do not see much circuitry in common between amps and ohms. Do you think it is leakage on the board itself rather than bad components? I am considering replacing U234  (DG411) to thoroughly clean everything around and under it with IPA. Under similar considerations are diodes and 750k resistor.

What do you think? Any other suggestions?

[Kleinstein]

Leakage due to the spill is still a possibility. However this would be relatively localized around R272 and the U223 (the max326). I doubt that critical node would extend very far. It is more like there should be a guard trace around it - so maybe check that guard. Guard line are sometimes rather thin and might got damaged despite of the gold coating.

The other point to look for would be AC noise - this could be a shield missing / disconnected and thus just 60 / 120 Hz hum or maybe an OP oscillating or to much noise on the supply. Reading the 20 mV of AC at Q214 - which is essentially a high impedance node, suggests that there is some AC floating around. The reading of cause depends a lot on the probes - so hard to tell if 20 mV AC is a lot or little.

Except for the close proximity to the spill I see not much in common with the amps part. Especially the 2 mA range should not be that easy to disturb by leakage. So it might be worth looking at the amps part first - it might be easier to find the trouble there. An oscillating amplifier in the amps part might cause trouble with the ohms.

[nikonoid]

Kleinstein, I did a quick check of guard traces. They look Ok. I will spend a bit more time on it tomorrow. By the way, my traces are led covered - no sign of gold flash. Maybe too old of a unit.

Today I did a pretty thorough IPA cleaning with ESD safe Q-tip equivalent. I covered both amps and ohms section. Interestingly ohms did not improve at all, while amps improved significantly. See attached before/after charts.

As the AC noise is concerned, it went down a bit on Q214 to 11mV AC. I am measuring between input LO and collector.

I also measured AC noise on U225 (OpAmp in amps section) where I have a pretty ugly bodge (attached) and found them to be under 0.7mV AC.

By the way what would be a proper point to attach ground of oscilloscope to?

[Kleinstein]

The logical ground point for the scope is the negative (com) input. This could be either the plug or maybe a connected point on the board (it should not make a big difference at lower frequencies).

Unless very dirty a leakage in the 0.5 µA is quite a lot for just dirt. So it would need way more than just a missing guard. It would more like a dead short towards a guard that could cause that much leakage.
Even this would not match the measured voltages.

So I would either check the amps circuit first of look for AC (with the scope) on the supply and the critical points (e.g. outputs of U228 and U238) of the Ohms circuit. The control lines (e.g. U223 Pin 9 ) might be interesting too.

The strong drift in the amps part looks odd and could be easier to find. The circuit for the amps part should be relatively simple compared to the ohms part.

[nikonoid]

After the last post, I tried quite a few things. Let me list them somewhat chronologically:

I did detailed review under microscope of guard traces and also checked continuity of them and possible shorts between them and microchips - nothing suspicious.

Then I did measure a Drain on Q214 for AC component with oscilloscope. I placed scope ground on Common test post that sticks up next to voltage reference under ADC. I see some low frequency and high frequency components, but I am not sure if this is significant, or I am just picking up some unrelated spikes. See attached.

I let meter to warm up for 3.5 hours and remeasured voltages at att transistors involved. See the Transistors1.jpg attached. Interestingly the meter drifted closer to be in spec, yet some voltages are almost double of previous measurements of good 2002 meter. Is it possible that from 1994 to 1997 they changed not only some microchips to their equivalents but also changed operating modes?

Today, being out better options, I replaced U234 (DG411) and two diodes next two it. When I removed U234 I noticed that R277 was pushed to a side and down (not by me) and basically laying on some traces and vias under it. I measured about 10 to 20MOhms between either lead of the resistor and its metal body. I did not like that for a 750k resistor. I used a small screwdriver to push the resistor higher. Afterwards the resistance between leads and the body became greater than 100M. I thoroughly cleaned the area around and reinstalled fresh U234 and diodes.

After U234 change and this last round of cleaning around U234, performance of the meter improved quite a bit in both Ohms and Amps. Another positive is that it now passes all self tests, even when cold!

If comparing 1 hour warm up readings, it used to read 98.6k and now is reading 99.95. The phantom current on 2A range after 1hr used to be -950uA. After initial cleaning it went down to -337uA and now I get -23uA.

After 2 hours the meter is very close to be fully functioning. I get 99.985k that is no longer improving, +6uA reading on 2A range.

I am still not sure if I am dealing with bad component/components or dirty board. When I got the meter the board was dirty to the point that I was initially removing mice excrement and dead insects from it.

The temperature dependence of performance is quite puzzling. I am considering putting a gloved finger on microchips one by one, trying to find one that reacts to temperature the most. Also I can use cold canned air to cool different areas of the board.

What else can I try? How can I distinguish components trouble from board trouble?

[Kleinstein]

The drain of Q214 is not a good test point: it is an essentially floating point in normal operation. So the readings are essentially noise picked up by capacitive coupling and DC is determined from non critical leakage currents. I don't really understand why they have Q214 at all - it should have very little influence. I would not be surprised if the meter would even pass the self test if it is missing or broken (shorted).

A suitable point to check for a superimposed AC signal would be more like drain (or source - should be essentially the same) of Q213. Another interesting test-point would be the emitter of Q251. Checking the supplies (e.g. +-15 V for the DG411 / max326) for spikes would be a good idea too.

For the current part it might be interesting to see if there is some AC residual at the terminals when in the lowest amps range. Also the measured DC voltage in this range might give an indication. There might still be some cleaning needed in the path (e.g. upper end of highest value shunt) of reading the voltage at the shunts.

I still can't see a path from leakage around R277 towards the R272, were the odd leakage toward ground was initially measured. So far I don't see a path coupling the the ohms and amps part for something like an automatic adjustment or just an extended self test. So it is odd to see an error in both parts that is so close coupled. For me this somewhat point's to an AC problem, of some noise / oscillation that couples to both areas.

Much of the different voltage readings for the good and bad meter around Q213 can be due to a different type of diode for CR217 (higher forward voltage for the bad meter). Such a change would not have a significant effect on the meter operation. Also slightly different levels of the reference voltage can change the readings (e.g. around Q251).

In TiN's circuit drawing, I just noticed the guard connected through R370 to U238. This would be rather odd. The more logical guard potential for the resistor part would be the output of U228. I would more guess this is an error in the plan: the output of U238 would be good for the guard around Q213, CR217 and connected, but not for the range resistors. There might be a second guard for the resistor part too.

[nikonoid]

Kleinstein, I did a number of AC measurements today, as you suggested. Please see attached.

I do see spikes and sometimes even combination of several frequencies but I am not sure what is important and what is not. I noticed one strange thing, though: the power on U220 is +20 (pin 13) / -15 (pin 4), while U223 has +15/-15, as expected. Not sure if this is important. I checked and working 2002 has same power on these +20/-15 on U220 and +15/-15 on U223.

I also used K2000 to measure voltage between Amps and Lo inputs of broken K2002. I did this as both AC and DC voltage. They are in the last attachment here. Please note that K2000 has a 40uV AC offset with inputs shorted, so supposedly 40uV AC should be subtracted from AC measurements in attached table.

I do see at least two guards around ohms circuitry. I will look into tracing them in a bit.

Thanks a lot for taking a look.

[nikonoid]

Kleinstein, you were spot on about the guards. There are at least 4 different guards in the ohms circuitry. I see one guard going from output of U238 through 10k resistor, I also found another guard from U238 though 100k resistor. That one covers Q213, Q210 and other nearby elements.

I also found that output of U228, as you predicted, through R371 (10k) resistor drives guard covering majority of OHMS switching including U220, U223, U232, U233 and precision resistors. and some legs of Q211, Q212. I am making some changes to my drawings and I am planning to pass them to TiN.

I also found a bit of circuitry that is conected to Q226, including precision divider R335, R339 and also involving compactor U239 and register U203. I am not sure though if they could be considered part of OHMS circuitry.