Keithley 2002 repair help

[nikonoid]

Thank you. I will give it a try.


Sent from my iPhone using Tapatalk

[ManateeMafia]

How about some soldering tip tinner? That may help remove any oxide on the via.

[TiN]

Why even bother with transformer wire? Just get piece of regular multi-strand UTP cable (LAN RJ45 cable) and take a strand from there. It will be bare copper metal, which solders like a champ.
When I was fixing my first 2001, I just heated via from the bottom side with the iron set to +240c, and poked a copper strand from top side till it goes thru and sticks from the bottom.
Then you just cut the top remaining wire and solder it down on top as well. Takes just a minute once you practice a bit. if via on either side peels off, you just use same strand to extend to the PCB trace track and connect there.
Don't use solder only bridges, as solder != copper tempco (%0.4/C!) which might be important for analog traces!

[martinr33]

Another good alternative is wire wrap wire - solid core, 30 AWG, silver plated, Kynar insulated.

If your solder is not taking well, you probably have contamination on the exposed traces. The capacitor electrolyte has combined with the copper to create a compound that is hard to solder. Your solder bridge is probably making contact in tiny areas, which is no good. I would clean the solder off, and look to see how much of the underlying trace is tinned.

This is a precision instrument and the board is sensitive to contamination from cleaning products. Or finger oils. Or just being in open air. So whatever you use to clean it, pick something that is easy to wash off. I'd start with white vinegar, being very careful not to get it on the rest of the board - use a damp rag. On really bad corrosion, I use MG Chemicals tin plating solution. As well as dissolving junk, it plates a thin layer of tin onto the copper - but you have to clean it off carefully. The plated tin makes the trace easier to solder.


On your question about standards and the 732a - 8.5 digit meters are quite noisy in the last digit or two. A 732a gives you a stable, quiet voltage so you can see how much noise your meter has.

My K2002 has an SD of anywhere from 300nV to 1.2uV on the 10V range with shorted inputs, depending on NPLC and filter settings, and other variables like time of day. My Advantest units do 200nV - 700nV.I'm bringing up a 732a now - it is nice to see similar noise from the 732a 10V output as I do on a shorted input. And the two Advantest units I have hooked to it will differ by anywhere from 1 - 5uV. Now, I am using rubbish cables but the temperatures should all be stable by now.

Good luck with sorting out that board - it's worth the effort, the K2002 is a good meter. You also have one of the first run of units there.  Post if you need help with the EPROMS.

[macboy]

Another good alternative is wire wrap wire - solid core, 30 AWG, silver plated, Kynar insulated.

+1 on this. It should be the first choice, not an alternative. The strand of cat5e cable would be an alternative to this. You can straighten it after untwisting by grabbing a length of it and pulling it until it just gives a little. The copper will then be perfectly straight.

[nikonoid]

Guys, thank you very much for suggestions. It took me few days to find time to try them.

First of all I got a cheap iron tip thermometer from ebay, calibrated my iron and used it at 250C.

You might try some no-joke RA flux, I use MG chems 835, pretty aggressive stuff.

I had MG chemicals 835, both older bottle (stuff is more like a paste by now) and fresh bottle. It helped in few places, however did nothing for vias affected with electrolyte. Tip cleaner worked better there. Used it very sparingly with a tip of a toothpick. It looks cool under microscope.

I bought the silver plated Kynar insulated 30awg wire and I am very impressed by it. I removed solder bridges and made all the trace repairs with this wire. My multiple attempts to get the wire to go through the hole did not work. Maybe a wire is too large for microvias. Nonetheless I was able so  to solder it partially going in. Pictures are coming.

I will have to try TiNs "going through the via" suggestion another time.

In the end I had to do two trace repairs and 3 bodge wires. Luckily they all are power related, not measured signal.

It took 8 hours to repair the board and solder all ICs and capacitors back. Electrolyte compromised areas are not fun to work with.     

After this I partially reassembled the meter for a quick power up test. It came up without a hitch. As I was shooting a thermal image of it, I noticed a diode CR119 getting really hot. It was shorted and somehow I missed it on initial check.
 
I did not have a replacement handy for this BAV103 250V switching diode. I had a 600V zener diode in a similar package (BYD57J), so I used that for a quick test. Proper part will be in next week.

With this zener unit powered up with thermals looking much more reasonable. Small U108 regulator (78L05A) was getting to 180F, but apparently it it is specked to at least 250F.

This time I did the open inputs test and got errors:
306.1, 306.2 ohms current source
308.3    2M Ohms overload
400.5    +4.48V DAC output
401.1    Regulator (not sure what this is yet).

My next action item is to check all voltage rails before tracking these individual errors.
Thank you everybody for your your help this far.

[TiN]

That's it, no more errors? Looks too good to be true

[Kleinstein]

The DAC  error and the 401.1 regulator error might be hiding more to come. AFAIK the DAC is used extensively in self-tests (that't the main purpose of it) so some self test might not really work right now.

The list is rather short and the first two errors might even be related. Still I would start with the other two errors.

[martinr33]

For corrosion repair you want this stuff:

http://www.mgchemicals.com/products/prototyping-and-circuit-repair/prototyping/liquid-tin-421

I use it for battery contacts. First application dissolves off the crud,  rinse, apply again for tin plating. The plating is rough  ( not shiny) but it works well, and readily takes solder.

Well worth having a bottle to hand.

[nikonoid]

Shorted diode replaced. New round of testing shower problems with power rails.

Repair manuals does not clearly define some testing points. They refer to some diagrams like 2002-100 but I cannot find any of these diagrams available. If someone has it, please let me know, otherwise I might have to produce my own. There are lots of unmarked test pads in the [power supply section, but no description. Attached is a very rough draft. I am not very sure about difference between COM, Isolated Common and Digital common.

As per power supply testing procedure (also attached) step 3 (U108) gave me 15 volts instead of 5. I think I installed a right part (LM78L05ACZ) and in proper orientation. By the way this is also a part that is getting really hot. What could it be?

Step 4: U110. I am getting 1.4V instead of +5V

Then +15 and -15 are just fine and bootstrap is a bit high 38.9 and -38.8.

U105 is producing something completely wacky with with 1.4V and -7V

And last two steps (11&12) again completely fine.

U108 was replaced by a new part, the rest of regulators are originals. What would be my next thing to check? I am especially puzzled by getting 15V instead of 5.

Thanks a lot.

[alm]

Maybe the silkscreen is wrong? Check the voltages on the pins of the 7805 and compare with the datasheet.

[nikonoid]

Another update.

For power test 9 and 10, Voltage on the U105 is actually fine. Before I used a wrong point as an Isolated Common (it is not in the manual). This time using a negative leg of a bridge ratifier CR104 did the trick.

Power test 3 is actually a mistype in the repair manual (yes, another one). Sink screen is fine, but the manual was supposed to refer to Pin 1 not Pin 3 for 5V measurement on U108.

After these measurements I reran the self-test and only saw errors 306.1 and 306.2. Somehow other errors disappeared after diode change!

However this was only an open inputs test. I reconnected all the inputs and ran the full test. This time I got more variety of errors.
306.1/306.2
500.1/500.2/500.3/500.4/500.6/500.7     basically current measurements are shot
501.2
600.2

Regular spot checking of measurement functions showed that voltage works reasonably well, 2W/4W ohms work on everything but 2MOhm range, Amps do not show anything resembling input signal. It is almost like the amps path got blown somewhere. 


PS. I just realized that I forgot to recheck Power Test 4 results. This will be a task for another day.


 

[TiN]

Stupid questions: Is the current fuse ok, and did you short HI input to AMPs jack before going with current test?
Btw, what's the FW rev?

[nikonoid]

Yes, I shorted Input Hi to Amps and I did check fuses. 
I also checked both front and rear connections in case of a switch problem. For the front I was shorting directly into solder pad on the board as my white wire for current is still de-soldered.

I got exactly same list of errors for front and rear connections. It sounds like some sort of protection on amps input is blown. There is also another possibility: when meter is going through tests I can hear that one of relays is making a strange sound rather than regular click. Maybe a have a bad relay that is responsible for switching in current?

Also I tried to measure resistance between input low and amp as I am going through different current ranges. It did change I think from about 2 Ohms to 90 Ohms, but was fluctuating. Tomorrow I am going to compare it to resistances I get from my other 2002. There is definitely a benefit from having multiples.

The firmware in the meter is A5. If you like a copy, I have a programmer and can load it in before upgrading the firmware.

[TiN]

My older 2002 has A05, thanks. I'm still after miracleous B-version firmware for 2002 (saw once at ebay unit with B02).
Resistance from DCI jacks on my 2002 is:
200uA = 1.0194E3
2mA = 119.636
20mA = 14.5467
200mA = 1.2568
2A = 0.3491

[nikonoid]

Thanks, TiN.

I actually got somewhat very similar numbers on both working and broken 2002. The only exception is that on the broken one they seem to fluctuate rapidly by 1-2%. It was a very quick test and I might have to redo it.

Overall resistance switching seems to work ok. Then in a pinch I used a different meter in a resistance mode to serve as a current source. When I was pocking around all of the sudden current started to work then stopped.

So far it appears that proper voltage is not reaching Pin 3 of U225 (unity opamp LT1097S8, incorrectly marked in the manual as V425 instead of U225).

My next stop is to check for compromised vias or solder pads on the way to U225, especially considering that it was heavily compromised by electrolyt and replaced by me.

As promised, here are photos of the unit after repairs. I am considering shortening bodge wires later, however I will have to get a shield first before I decide on the final route for them. I am considering hot melt glue or E6000 glue to semi-permanently hold them to the board.

[Le_Bassiste]

gents,
i need some clarification on your current (Amps) input findings.  were all those measurements taken from outside, i.e., using AMPS  and LO input jacks?
if so, then it's little wonder that measurements are fluctuating. the internal spring loaded white cable "pigtail" connector on the fuse holder is just f* crap.
some intensive IPA cleaning of the fuse holder interior and fuse itself helps alot.

[TiN]

In my case, yes, outside AMP + LO jack. I'm not taking apart 24/7 online 2002's anytime soon, unless they go kaboom  .
I didn't see much fluctuation (random 3458 in reach was used to measure resistance), but again I don't ever remember using 2002 to measure current, other than during calibration  .

[nikonoid]

Le_Bassiste, for rear testing I was using rear terminals (from outside). For front I was using common (Input Low) from the outside and the hole on the PCB where the white AMPS cable goes to. I do not have it soldered yet. So I was bypassing the fuse and fuse holder.

This "fluctuating reading" behavior comes and goes, so I cannot yet pinpoint what is going on. One possibility is something going wrong with control signals that switch in different resistances. Maybe I get a behavior when specific resistor is not completely turned on.

I found what looks like a bad via where the signal after chain of resistors is traveling to pin 3 of U225. I may have to learn how to copper stitch vias. After a did a quick fix to bridge that via, the situation improved just a bit, but not much.

As I am pocking around the miter it behavior seems to change. It could be a mechanical effect or thermal, as I had with a previous K2002. There I had a bad IC switch that depending on a temperature would sometimes be partially closed.

With me fixing that one via the test 500.1 is now passing most of the time, while others still fail. After the fix, the current function seem to work to a degree, it is couple of %s off on 200uA range, little bit more off on 2mA range, and so on with 1A range being about 40% off.

I was tracing the U234 chip that controls 2mA and 200uA switching and found that while 3 channels would get 0V or 5V on their control inputs, one channel got -3V on it, so now I will have to check where control signals are coming from.   

So we are basically down the rabbit hole....


 

[nikonoid]

Today I found another bad via. After fixing it, I am down to only TWO test errors 306.2 and 600.2   

Current passes the self test, but it is working strangely. With shorted inputs 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

If you take this offset into consideration, then DCI reading works pretty well.

I will trace the problem with Ohms next and get back to current later.

[Kleinstein]

Usually the test for zero current would be with open inputs. For a classic solution with shunts this should not make a significant difference though - it can be a big factor for a configuration with TIA (for low currents).

The data look like quite some offset - it might be just a question of adjustment / calibration, but could also be a problem with amplifier offset or similar.  Before doing any adjustment / recalibration, the rest should be working of cause.

For the Ohms source, one could externally measure the test current.

[TiN]

Any update? Can't want for next episode of 2002 repair. 

[nikonoid]

TiN, thanks for checking on me. Yes. it had been a long saga.

I had been stumbling for a week, trying to figure out a problem with 2M Ohm range (Test 306.2). This is also the same range that I had a problem repairing first 2002.   

What I know so far is that on 2M Ohm range I get between 0.5uA and 1.8uA of current between Input High and Low, instead of 1.92uA. In many cases it starts at 0.5uA on warm up and within minutes would get to 1.5uA, then to 1.7uA and fluctuate from there.

When I check voltage drop across R271, I get 1.42V instead of 1.44V in manual - close enough and R271 being 750k it corresponds to correct current.
When I do the same thing across R322 - 33 Ohms I calculate the current that corresponds to what I have on the "Input High".

Somewhere in between we seem to lose some current. From reading the manual in between these resistors we have Q211, Q212 and Q213. My best guess is that either one of them is at fault or a circuit that controls one of them. I am having hard time tracing the circuit in that area.

Any suggestions are welcome. Thanks.

[TiN]

I'd suggest to reverse ohm schematics, for our all enjoyment. Then it will be obvious what currents are expected.

[nikonoid]

The process is slow. Attached images are as much traces as I was able to redraw so far.

Meanwhile I have some additional information. 306.2 error seems to be highly temperature dependent.
I plugged 100K Ohm standard resistor and used 2M Ohm range. When unit is cold it would start reading at 20k, then in minutes gradualy rise to 65k, in an hour the reading went up as much as 95k. At that point when I reran self test every OHM test passed successfully, including 306.2.
If I let the meter cool off a little bit is goes back to reading 80k, if it cools a lot is goes back to 20k.

So the issue appears to be stable, repeatable and temperature related.

I compared voltages to working Keithley 2002. The performance is identical for all pins on:
U220  DG411
U221 MC14094B
U223  MAX326
U228  AD711
U231  AD706
U232  MAX326
U233  AD705

Where I get the difference is:
U238 pins 2,3,5 and 6, respectively
0.787  0.787  0.788  0.859  on the bad meter
0.476  0.476  0.476  0.539  on the good meter

Q210
10.159  0.785  0.785  on the bad meter
10.005  0.475  0.474  on the good meter


Q211
12.792  10.198  0.787  on the bad meter
12.694  10.174  0.475  on the good meter

Q212
12.809  12.792  0.785  on the bad meter 
12.707  12.698  0.475  on the good meter

Q213
10.155 0.786  0.784  on the bad meter
10.011 0.475  0.469  on the good meter

Q214
0.000010  between (63mV and 125mV) 0.004  on the bad meter
0.000009   0.001  0.004 on the good meter   

The major difference is the Q214 transistor and 0.786 voltage instead of 0.475 voltage.

It is almost like performance on one of these transistors (Q210-Q214) is changing with temperature. I do not yet completely understand the role these transistors are playing.

Can someone please suggest a next thing to try?

Thanks a lot.