Keithley 2002 repair help

[TiN]

I'd remove DG411 near diodes, remove two diodes (big ones at right), remove 1Meg resistor in metal case, clean everything thorougly with IPA.
Once you done cleaning, dry it. Clean again. Then carefully put new DG411 in, clean 1Meg resistor, clean diodes, put them in.
Test again If same, buy JFETs and transistors to replace Q210-Q214, and iterate again.

[saturnin]

Before starting replacing parts using "try and error" approach (which you can use as a last resort anytime), I would suggest to analyze how ohms circuitry works actually...

In K2002 Repair manual, you can find quite detailed ohms circuitry description (see info on tests 304.1 and 306.2). You have already tracked multiple connections on the PCB, so put these information together and try to draw a partial schematic of ohm circuitry. Then, it should be much easier to understand how it works and find error...

[Kleinstein]

Even if the voltage reading seems to work, it might be worth checking for input bias, as too much input bias could be an issue that can also effect the high ohms readings.

Having it working better when hot, somewhat points to a problem with dirt / electrolyte / flux residue. This would be one of the very few cases where current leakage would decrease when hot (as humidity will go down).

The suggestion from TIN is thus not so much about replacing parts, but cleaning in a area that might be influenced be leaking caps. A new DG411 is just more convenient / easier than a really good cleaning of a part that might already be damages from unsoldering.

A little more understanding of the Ohms circuit would really help. Having an idea from the manual could help a lot here. One would at least know if the control of the FETs should be OK.

As the 2 M range is off so much, a leakage current of that size should be visible in the 200 K range too. Not that obvious, but still well higher than uncertainty specs. One could at least check if the missing current flows through the resistor for the 200 K range.

[nikonoid]

Guys, thank you for the feedback.

I think D411 next to metal 750K resistor is actually part of current measurement, not resistance. I will try to verify it. The actual board is much cleaner now than what you see in the picture. I even have a replacement part for it. However it was so close to precision network, I was not sure if I can remove it without accidentaly touching it with soldering iron. I may have to try if everything else fails.


I just ran few more tests. My resistance standards are not the greatest, some are +/-100ppm, some are +/-20ppm.
2k range off by 30ppm
20k range off by 150ppm
200k range off by .25%
2M range off by 2.5% this is after 1 hour warm up. It is much worse before warm up.
20M range off by 30ppm
200M range off by 30ppm
1G range off by 40ppm

Seeing that performance improves at 20M, I doubt it is leakage that is responsible for the problem.

I also checked bias voltage with 10M resistor and got -0.3mV

I see the pattern with two parallel diodes pointing in opposite direction is used in at least 3 places:
CR210, CR211   (by that DG411)
CR212, CR213   (by large rad COTO relay)
CR214 - actually 2 diodes in one case (by large rad COTO relay)

I think it is used for some sort of voltage clamping. Would someone know what exact use is? Thanks.

[saturnin]

Both 200k an 2M ranges use the same setup of the ohm circuit. The only difference is 7.2V appears across R277 (=750 kOhm) for 200k range, which gives 9.6 uA and 1.44V appears across R277 for 2M range, which gives 1.92 uA (see description of 306.1 and 306.2 tests). According specs these currents have absolute tolerance of 5%. R277 has 1% tolerance, so voltages across should be 4% close to nominal values (this is a rough estimate).

I think the cause both 200k and 2M ranges are way off is the same. Do 200k measurements drift upon warm up as 2M measurements do? Since high ohm ranges (20M and above) are basically spot on, I would not suspect surface contamination by electrolyte (it would have to be very localized to affect only these two ranges).

At first, I would check stability of voltage across R277 for 2M range  (has to be very stable). Is it stable or does it change? If it is changing, does the change correlate with change in 2M readings? From R277 I would follow path of the constant current and draw a partial schematic to get ideal where the current could be leaking...

[macboy]

Just thinking... Monitor the control register outputs (which drive the various relays, analog switches, and FETs) while manually switching between 20k, 200k, 2M, 20M ranges. You will be able to determine which switches or FETs are used differently for the different ranges, and hopefully narrow in to specific FETs or switch input-output combinations only used for 200k and 2M ranges. Then look for leakage in or around those components.

[nikonoid]

Today I measured 200k and 2M ranges for cold meter.
For 100k resistor on 200K range I got stable 7.11670V across R277 (750MOhm) or 9.49uA - within 1.1% off the target - good.
I got 295uV across R322 (33Ohm) or 8.9uA - 6.2% off target not so good. Resistor read at 96.8k

For 100k resistor on 2M range I got stable 1.42343V across R277 (750MOhm) or 1.9uA - within 1.1% off the target - good.
I got 35uV across R322 (33Ohm) or 1.06uA - almost 45% off target horrible. Resistor read at 55k instead of 100k.

So far it looks like we have a proper current on R277 and loose it by R322. There are only few transistors inbetween.


I also liked the idea with similarity of 200K and 2M ranges. Today I retested the meter cold instead of hot, here is what I got:
2K range off by 160ppm
20K range off by 2200ppm
200K range off by 3%
2M range is off by 40%
20M range is off by 30ppm

So really the problem is not on the 2M range, but on every range below it too. It just gets smaller on lower ranges.
It is just small enough on 200k that self test error is not catching it.


Now this also reminds me of the current problem I have. When measuring 0 current I get:
200uA range: -000.0461 uA
2mA range:   -0.013751 mA
20mA range:  -00.03802 mA
200mA range:  -001.7962 mA
2A range:   -0.018315 A

So far I can see one similarity all failing modes including current should have U221 pin 4 set as 1, and 20M, 200M and 1G have it set as 0. I previously could not trace what that pin controls, so there is more work to be done.

As always suggestions are very helpful and welcome.

[saturnin]

Today I measured 200k and 2M ranges for cold meter.
For 100k resistor on 200K range I got stable 7.11670V across R277 (750MOhm) or 9.49uA - within 1.1% off the target - good.
I got 295uV across R322 (33Ohm) or 8.9uA - 6.2% off target not so good. Resistor read at 96.8k

For 100k resistor on 2M range I got stable 1.42343V across R277 (750MOhm) or 1.9uA - within 1.1% off the target - good.
I got 35uV across R322 (33Ohm) or 1.06uA - almost 45% off target horrible. Resistor read at 55k instead of 100k.

So far it looks like we have a proper current on R277 and loose it by R322. There are only few transistors inbetween.

It seems at least the core of the constant current source is working properly.

So far I can see one similarity all failing modes including current should have U221 pin 4 set as 1, and 20M, 200M and 1G have it set as 0. I previously could not trace what that pin controls, so there is more work to be done.

If U221 pin 4 is not set according the manual, I would definitely check what input data go to U221. U221 might be faulty. Btw. Don't you mean rather pin7 (Q4 signal)?

[TiN]

Is there any reason why you not willing to reverse engineer related circuitry schematics? I'd see that much more beneficial, than guessing correct/incorrect voltages/currents around parts.
There are not lot of parts involved, so I'd expect that to be just few evenings effort. You don't need to trace all back to 4094's, just up to LM339's would do just fine.

[nikonoid]

Is there any reason why you not willing to reverse engineer related circuitry schematics? I'd see that much more beneficial, than guessing correct/incorrect voltages/currents around parts.
There are not lot of parts involved, so I'd expect that to be just few evenings effort. You don't need to trace all back to 4094's, just up to LM339's would do just fine.

TiN, the reason is simple.... I simply do not know how to draw circuits. I have never done that. It is on my TODO list, but so is many other things.

I know how to use Photoshop, so I used that to reverse engeneer connections to the best of my abilities.

If someone skilled at drawing circuits can work together with me on that that would be great. Anyone interested?


Sent from my iPhone using Tapatalk

[TiN]

Piece of paper with squares and pencil will work just fine.
Then scan it and I'll be happy to digitize it for community happiness.

[Kleinstein]

For reverse engineering the first drawings are often on paper, especially if the circuit is rather simple. There are a few programs to draw schematics, like LTspice, Kicad or Eagle (e.g. free limited version).

 As the Ohms circuit uses extra diodes, chances are the circuit is similar to that of the 3458. Also the Keithley 2001 might be similar from the general layout, though it is different from the usual path. The current sources are usually not that different. There are a few choices, like using CMOS MUX or a set of JFETs. A second point is the question of switching the resistors at one side only or at both side - for a high quality meter I would expect switching both sides, especially because they use the extra diodes.

[nikonoid]

Here is what I have so far as Ohms schematics is concerned.

I have traced it all the way from R277 resistor to Input Lo and path to Input Hi (R322 and on).
I found what is controlled by U221: Q1, Q2, Q3 and Q4.
You will see that some connections are going into nowhere, that either means that I have not traced them yet, or they don't seem to be relevant for Ohms understanding.

It is collector of Q212 that is at .4V on working meter and .75V on broken one. Let me know what you think.


TiN, let me know if this is enough for you to start working with. The layout can definitely be improved from what I did by hand. Thanks.


PS. I have also been compiling a file with all mistypes/mistakes I found in the repair manual so far.

[Kleinstein]

There are still a few open points in the partial schematics. So it is a good start, but it needs a little more.

One interesting area would be the upper end of R277 (750 K). Another part would be how the other current setting resistors (one more would be enough) are connected.

There also needs to be a kind of path from the lower side of R277 to an OP ( could be U233 Pin 2)

There seems to be a connection missing from the emitter of Q210 to something like the source of Q213.
I am a little confused there is no diode in series with Q213, though is might be somewhere following R322.

The voltage at the collector of Q212 depends on a few non critical parts (e.g. the diode mentioned above) and also on the input connection. So I am not sure this would be a significant difference.

[nikonoid]

There are still a few open points in the partial schematics. So it is a good start, but it needs a little more.

One interesting area would be the upper end of R277 (750 K). Another part would be how the other current setting resistors (one more would be enough) are connected.

There also needs to be a kind of path from the lower side of R277 to an OP ( could be U233 Pin 2)

There seems to be a connection missing from the emitter of Q210 to something like the source of Q213.
I am a little confused there is no diode in series with Q213, though is might be somewhere following R322.

Thank you Kleinstein. I am working on version 2.0, so feedback like this is very helpful.



Sent from my iPhone using Tapatalk

[TiN]

What is Q211 device? You sure it's 4-terminal FET?
Also mark all precision resistors and route them both ways, that would be great.

We would need to know how resistors are selected/switched on various ranges, to make it all meaningful.
All that happening thru pair of MAX326 low leakage switches U223 and U232. I'd expect arrangement is similar to one in 2001, just with better parts.

[nikonoid]

Yes, Q211 is a FET with 4-terminals. Two of them are shorted together. So I guess it is happening on outside of the can rather than on inside.
I started tracing precision resistors. It is a little tricky. They connect to all 4 channels of max326 and also to four diodes above them. Then these diodes connect to all four channels of dg411, next to max326.
I cannot visually see these connections in my "X-ray" photographs. So tracing them one by one is very slow.

I also so far cannot find where the output of u228 goes. It is likely finding its way to these resistors. I just cannot find it yet.

TiN, what software are you considering using to draw this?


Sent from my iPhone using Tapatalk

[Kleinstein]

connecting the resistors to diodes + DG411 and max326 makes kind of sense. This is one option to do the switching: The DG411 and the diodes to carry the current and the max326 to connect a sensing amplifier.

So the overall type would be rather similar to the circuit in the HP3458 and not very similar to the Keithley 2001. Only the protection part would be more similar to the K2001 (using high voltage FET and PV opto-coupler instead of cascaded PNPs).

p.s.:
Possible leakage paths to cause the current to be to small could be:
1) the diodes Cr212/SR213 , it might be worth checking the guard potential for this.
2) the guard amplifier - easy to check via voltage across R294.
3) The P-channel 4 pin MOSFET - might be visible at R290
4) U233 (OP) - that is likely connected to the 750 K lower side.
5) backwards through other precision resistors (for some on could check at the upper end).

[nikonoid]

Kleinstein, I did find the connection from U233 pin 2, missing link from Q210 and also diode after resistor R322. All of your comments are right on the money. New version of schematics are coming in a day or so. Thanks.


Sent from my iPhone using Tapatalk

[nikonoid]

It took quite some effort and now I have a fairly comprehensive schematics (attached).

I did not bother to draw power to any ICs here. Power is fairly obvious.
I traced Q1-Q8 bits of U221 and also found some things controlled by U203. This is far from perfect, as I still have few points that I cannot trace, like pin 1 of U231 of few resistors next to it. There is also some connection from input divider to Q226.

Still, I think I got most of the functionality. I am asking for help to see if I made any mistakes. Maybe even compare schematics to some clues from repair manual, to see if I got them correctly.


On the topic of fixing my meter, I also measured voltages across many resistors, like Kleinstein suggested. Current is correct on R277 and wrong on R386. It appears that some current is leaking through precision resistors, especially R272. It is loosing .5uA there and that is close to the total amount lost. I am still not sure what specific component would be responsible for this leak.

See the second attachment for voltages and associated currents. My measurements were done on a cold instrument with 100K resistor in the 2M Ohm range.

Thanks a lot.

[saturnin]

It is much better now! Good job Although there are still some missing connections to important nets/pins (U231-3, U231-5, U228-3...), it is not necessary to track them completely - it is enough to measure voltages at these points. It will tell you a lot about circuit function...

As concerns the leakage, check U228 works properly, i.e. there are same voltages at its pins 3 and 6. If so, CR208 or the switch connected to R272 may leak some current. (When 2M range selected, there should not be any significant current flow though R272, definitely not 0.5 uA...)

[Kleinstein]

The circuit plan so far looks reasonable - At least I don't see an obvious mistake. There is some connection missing around R276 / U232 pin 2. probably a connection to a fixed reference voltage (around 7 V).
The still open connection of  3 of U231 mit be towards pin 5 or 7 of U231. At least this would be a logical target.

Leakage through R272 has only 3 ways to go: dirt on the board (e.g. below U223), the diode or the CMOS switch (U223). Checking the diode path it likely easier, up to the point of de-soldering the diode. U228 is likely OK since otherwise there should have been even more leakage trough the other resistors. It would be more like the 100 K resistor open or U221 Q6.

[nikonoid]

Kleinstein,

You are right on the money again. I found Pin 3 to Pin 5 connection for U231.

I did not have much time today to troubleshoot the circuit. I did check all 4 100k resistors in the vicinity of U220. I got readings of about 98k for all of them. For 100k resistor connected to Q251 I got 83k. I guess there is some parallel circuit somewhere.

I also checked how much current is flowing through R270,R271,R272 and R277 on working meter and it is quite a different picture. I only got between 0 and 0.003uA there.

I measured all four diodes (in circuit) and got pretty analogous readings between meters: 0.66V in one direction and 2.5V or 2.6V in the opposite direction.

Not sure what to check next...

[Kleinstein]

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).

[TiN]

Prettified nikonoid's handiworks just a bit. Time to get back with pensil again

PDF-version.

EDIT: Bugfix ver3.1