Keithley 2001 repair log and Heat Sink Clip request

[vsmith]

Hi all,

I see that there is quite a bit of info regarding repair of Keithley 2001, especially on TiN's epic thread "Restoration Glory of Keithley 2001"
I have read it through several times, and still have a few questions/comments, and hopefully can contribute knowledge as I go along.

First off, the 2001 I purchased (eBay, naturally) came missing the heat sink clip for the power FET and Resistor on the left rear of the chassis/analog board. It is also missing the rear bezel, handle, and the long screw which holds the analog board shields to the pcb. I can probably scrounge up the last 3 items, but need your help with that Heat Sink Clip: do any of you have a spare parts unit and could spare that clip for me? Willing to pay reasonable price and shipping of course.

Now for the details. The unit is the M model, serial number 1316678, and the FW rev is B17 A02. I think the unit may date from 2010 or so. But maybe I am reading the date codes on components wrong?

The unit was described as working but failing numerous self tests. And I received it in that condition. Display is good, digital/processor function seems good. But the errors, which began with 200.1 and went all the way up through the 300s (I stopped recording at that point) indicated a problem with the A/D and Reference sections. I checked the power supply voltages and saw that +15 was +18, and the regulator appeared to be shorted. All other voltages checked out OK.

I took the shields off and immediately realized that while the electrolytic caps had been replaced (with good 105 degree caps), whoever was doing the repair stopped and closed the unit back up LEAVING THE UNCUT LEADS OF C101 and C104 TO SHORT OUT WHATEVER THEY COULD TOUCH ON THE BOARD!

Unbelievably, the only actual short circuit that resulted appeared to be the +15v regulator, meaning that the +15 rail was sitting at about 18 volts. Once I cut the leads from those caps, the 15 volt regulator came to life and all the PS voltages were correct.

(By the way, aside from the uncut leads, the board rework was done well, and the board itself showed no signs of corrosion or damage or even leakage)

Re-ran self tests, and this time the errors began at 201.2, (7 volt reference) continued thru 201.3 (1.75 volt reference) and then 300.2, 302.2, and 303.2, plus all the resistance range checks (304.2 thru 304.7). As before I stopped there, knowing there was a fundamental problem that needed attention before any higher level stuff got done.

At this point I discovered that several of the op-amps in the reference and a/d mux/buffer circuitry appeared to be fried. Evidently that +15 rail going high wreaked havoc. I think the lm399 is OK. It's companion op-amp u328 is bad tho, so can't be sure. I have ordered a bunch of replacement parts (there are at least four different op-amps that took a hit) and now have to wait for those to come in.

When I get those circuits working again, I plan to post dc readings at the key points for each of the 201 series of test steps, which I hope will help with future repair attempts.

I'm a little sad that this unit's recent calibration is now useless, I was hoping that a simple power supply repair would allow the unit to come back to life cal intact. But I guess I lost that gamble. I am still very happy to have a 7-1/2 digit DMM, if I can get it repaired!

Question: I quickly realized that without the A/D daughter board plugged in to the analog board, the reference switching/routing circuitry on the analog board did not work. So reading signatures/bit patterns on the various ICs can't really be done, since the A/D card obscures that part of the analog board circuitry when it's mounted, and if it's taken off, the bit patterns are not valid. What have other repairers done to probe that area of the board while troubleshooting? Has anyone made an extender for the daughter board? It's a really annoying mechanical design!

Comments, please!

Vic

[vsmith]

Update: I finally got the parts I ordered and had time to do the rework on the Analog board that I hoped would resolve the problem. I replaced U327, U328, U324, U331, U333. All of these were opamps, and all had inputs with large differences in their voltages, which I interpreted as them having been damaged by the +15v regulator being short circuited.

Unfortunately, nope. U328 (the +7v reference circuit) still has inputs 7 volts apart (7v and 0v) and it still gets hot when the meter is powered on. I probed around a little more, this time looking at u330 (the precision switched capacitor reference divider) and found that the drive voltage at the Cosc input was only 150mv p-p. Meaning that the output of the Mosfet driver U535 was only 150mv, rather than the 5 volt square wave I would expect to see. I dont fully understand how the reference circuit works, but without u330 switching properly, I think that the voltages in the circuit would not be correct.

Without breaking the connection between the driver IC and the switched capacitor IC, I cant be sure which device is causing the problem. They are both supplied by that same 15V rail, though. I have ordered both and will try replacing them next.

Have any of you with experience repairing K2001's ever found and troubleshot problems with the reference circuitry? Any thoughts on theory of operation of this circuit?

[openloop]

What have other repairers done to probe that area of the board while troubleshooting? Has anyone made an extender for the daughter board? It's a really annoying mechanical design!

My speculation is that the ADC board can be plugged upside down and to the left (it has holes for that), but only when the analog board is out of the case.

Have any of you with experience repairing K2001's ever found and troubleshot problems with the reference circuitry? Any thoughts on theory of operation of this circuit?

It is very simple.
-Vref is coming from the ADC board and sets the operating current for the LM399. Check that -Vref is there (about -10V). ADC board has its own zener - that's how Vref is made.

U330 should not load U328 pin6 (op amp output). Moreover, K2001 probably will work without U330, so you can remove it for testing - just some ranges will be dead.

Also, bad COSC on U330 is a possible indication that something's wrong on the ADC board where the clock is made. Pray it's not U808
Or simply U330 is so dead that it loads everything...

[Zucca]

If you need some IC I have them.

[vsmith]

Openloop, Zucca, thank you both for your replies! I have read your previous K2001 related comments, and I value your input!

As far as the bad (low amplitude) COSC waveform, the input to U535 (Mosfet driver) labelled "600Hz" is a full 0 to 5v square wave. So I think that things up to that point are OK.

Thanks for the tip on -Vref...I will check on that. And removing U330 will definitely help pinpoint the problem IC, will try that.

[Kleinstein]

The reference configuration with the K2001 is a little odd. The ADC uses its own Zener reference (suppossedly low noise, but not very  low TC). The LM399 reference is measured in the 20 V (or 10V ?) range and the 2 V scale factor is from 1/4 of the 7 V ref. They measure the reference voltage in parallel to the normal AZ mode (not sure about the details).
The old K19x did it similar with a simple 3 conversion cycle (0 , ref, signal ).
Using the extra gain measurements give a very stable reading (low TC) but additional noise. At least in newer meters like K2002 they seem to use some kind of averaging for the gain and also the zero - to much averaging with the zero leads to additional  low frequency noise, as the 1/f noise supression no longer works as well.

[TiN]

What have other repairers done to probe that area of the board while troubleshooting? Has anyone made an extender for the daughter board? It's a really annoying mechanical design!

I just cooked a wide ribbon cable with 2.54mm pitch headers/connectors that plugs into ATOD PCBA and mainboard to probe. Or if you just want probe few points, solder a piggy wire to access from outside and test that way.

K2001 is one of the most noisy 7.5d meters out there, so don't expect very low noise readout even in fully working unit. Noise of K2010 or HPAK meters is much lower, some good K2000's even have lower noise than K2001.

I might have a clip somewhere, but finding it in my TE black hole is a challenge. 

[vsmith]

Hello all

It's been a while since I had time (and the needed parts) to continue trying to fix this blown up meter! And I wanted to thank TiN for that suggestion regarding an "extender" cable. I haven't done that yet, but at least feel like I have a strategy if I need to do a lot of point to point troubleshooting underneath the AtoD Daughter board.

I was able to obtain replacements for U330 and also U535. I did the replacement and checks one at a time. First U330, and that definitely changed things. After replacing it, the first test that failed was 201.3, rather than 201.2 as before. A quick check showed the 7 Vref was now present (7.053). Checking the output of U535 showed that the switching waveform for U330 was still only 150mV p-p. So changed U535, and voila: now 5v p-p.

Great, I thought, and ran the self test, only to find that now the errors start with 200.1. and run all the way thru 200.7, and continue thru 201.2 and 201.3. So the picture actually looks worse. (201.2 formerly passed as did all the 200 tests)

However, progress was obviously made, since the reference circuit now is outputting voltages that seem correct, 7.053, and 1.771. (-Vref btw is -10.374). I have no idea why that suddenly coincided with finding all the AtoD errors (200.x). Back to troubleshooting, this time the AtoD card.

TiN, if you ever do find that clamp, I could surely use it! Actually, at this point I wish I had a "known good" AtoD card to swap in. I'm unsure where to start troubleshooting that. Suggestions?

Thanks, all.

[vsmith]

Update

Ad In is sitting at -12V regardless of what test mode the unit is placed in. So there is more trouble (caused by the supply shenanigans before I got it?) on the Analog Board. Starting with U322 and working back.

[vsmith]

Hi All, Happy New Year!

I finally returned to this project after a 6 month hiatus. The first update I have is to say that I built an extender cable for the AtoD daughter card, so that I can probe the components and signal paths in the circuitry hidden underneath that card. It turns out that readily available 0.1" headers used widely in Arduino projects are just right for creating a 25 position female to male extender cable. I chose to make mine about 4" in length, whic his enough to sit the AtoD card on the top edge of the front panel. Pictures attached. I used an already assembled female header and cable (stripped down to 25 positions) and soldered onto a male header at the other end. Subsequently I discovered 4" male to female jumpers for sale on Amazon, which I have ordered.

The second thing that happened was that I realized that the cause of the mystery of why after finding and replacing blown components in the reference circuitry, the Self Test results actually got worse (even though the reference voltages were now correct). The problem turned out to be an intermittant bad connection to P/J 1027 which communicates digital controls from the Digital board to the Analog board. Once that was corrected, my unit began measuring DC volts with impressive accuracy! Running the self test now proceeded without errors up to 301.2, and errors continued all the way thru the 300 block of tests.

I discovered what seems to be a nonfunctioning NJFET (Q328) in the signal path, and here I am, once again, stuck. I cannot find any info on this part. It is a plastic to-92 case NJFET, and it is marked 1046, LS, 439F (on three lines). Googling does not bring any useful info.

Can anyone on this thread (If any of you see my post) help with finding a part number for this JFET, and better yet, a useful parts list for the meter?

Thanks!

[Kleinstein]

Q328 is used to switch the lower leg of the input divider. So it should be a low resistance JFET for switching. The usual candidate would be a PN4393 or J113 , possibly using 2 of those in parallel to get lower resistance.  These types should at least be good enough to see of things are working, maybe with slightly higher error for the higher for the 100  and 1000 V ranges due to more resistance.

Edit: With the last line 439F , there is a chance the part is actually a selecte PN4393

[vsmith]

Thank you!

Zucca also PM'd me with similar thoughts, he found that the part is listed as Keithley 31841A. I googled that and discovered that the 196 meter uses several JFETS with a Keithley part number of 31841 (no "A"), which is crossed to a PN4392.

Any thoughts on PN4392 vs your suggestion of PN4393? I haven't yet checked the data sheets to see how different these two are.

[openloop]

I too believe it's PN4392

I also think that they were preselected for low Idoff.

[vsmith]

Thanks openloop!

Will post my results after I get some pn4392 in.

[Kleinstein]

The PN4392 and PN4393 are very similar, same mask and just different in the threshold voltage. The ranges even overlap and some parts can qualify for both - those are probably the ones you like best. To low a threshold and the R_on is relatively high, to high a threshold and at a rather negative voltage it may not turn off all the way and the leakage current can be a bit high.

[vsmith]

I have ordered 10, and will select the best ones for this project. Thanks for the info!

[openloop]

Just realized.
The same part is used in Q312, Q320, Q324, Q328 and Q329.
Presumably, at least some of them were not fried. 
So you can pull one or two and measure their Vgs(off) and Rds(on) - to see what were Keithley's binning criteria.

[vsmith]

Ah, good idea! I am working on it today, so am happy I saw this!

[TiN]

I did some tests of 2N4392 from Keithley 2182A while back. Maybe this would be helpful.

[vsmith]

Well, things are starting to look up for this meter. On the other hand, I would have to rank my analytical and repairing skills as somewhat below par.

First off, the theory that I had a failed JFET in the Ohms measuring circuitry turned out to be false. When I removed Q328 and tested it out of circuit, it tested good. So basing my troubleshooting on measurements made with an AT Mega 328 component tester, with a JFET in circuit, in this case, was not a valid test. Once I replaced the JFET and started probing the voltages found in the current source circuit, I realized that opamp U324 (which I replaced earlier in this saga) was not connected to the positive rail (bad solder joint). And resoldering it only revealed that running it without a positive supply had killed it.

So I replaced it once again, and VOILA. The meter now passes the complete Self Test routine. And is doing a pretty good job of DC and Ohms measurements, although I have only spot checked various ranges.

There is still an intermittent in the Digital to Analog interconnect ribbon cable assembly, so I need to fix that before I can put everything back in place and close it up for performance testing.

So I apologize for prompting all of this great investigation into selection criteria for PN 4392 NJFETs. Not that it's wasted, all information is useful! And stumbling into it, or not, I now have a salvageable Keithley 2001. Much of the credit goes to the contributors here, and especially TiN and Zucca, with valuable assists from Kleinstein and Openloop. Thanks, guys.

[vsmith]

I would also like to mention some discrepancies I have found in the documentation I am using to troubleshoot the meter.

Hopefully this might help others trying to solve problems in the Ohms measurement circuitry.

The description of signal flows and component designations in the Repair Manual (2001-902-10 Rev C / May 2011) seem not to agree with the component designations seen on the schematic (Repair Manual Addendum Schematics Part 1 (2001-902-10 Rev C / May 2011) (page 6 of 9).

For example, The source current setting resistors have different designations. The schematic does agree with the actual board in my unit. The current setting resistors are r356, r358, r365 and r366, whereas the description for test 302.2 describes r355, r356 and r357 being placed in parallel to set the test current. I don't see either r355 or r357 on the schematic, but I do see an r367 (22Meg) which is part of the circuit. I wonder if the self test step descriptions were pulled in from earlier revisions of the 2001, and the designations were changed in the schematics and board screening, but not brought up to date for the descriptions. Another case is the custom resistance divider designated on the schematic and board as r607 (schematic page 5 of 9), which is described as r394 (9.9Meg leg) in the test step description.

If anyone is interested, and sees an error in my statement above, please do post your correction! I often overlook things in such densely populated circuits!

[openloop]

While setting up my "tracer" for unrelated activities, I remembered about this issue of undetermined Keithley's JFET transistors. I also remembered that I have the same Keithley part 31841A transistors removed from the ADC during my restomoding activities.

So I scanned them. Device marking is TG128 (parts list TG-128).

Code: [Select]

Vds=5V    Idss, mA    VgsOff @1nA, V    RdsON, Ohm
Q807      60           -3.9               30
Q809      46           -2.9               30
Q811      41           -2.8               30

I think I see a pattern here. 
So these indeed appear to be PN4392 switching JFETs preselected for low Ron and, maybe, for VgsOff > -4V.