Restoration glory of Keithley 2001 DMM

[bitomaxsp]

I think the main issue is the hum, not so much leakage.

One can do a crude test for leakage by slightly (e.g. +10 K range) heating suspects parts. Semiconductor leakage goes up quite a bit with temperature and this most of the time also applies to parts with excessive leakage.

P.S.:
The 4052 switches are likely selected for low leakage. So unless there is more pointing to a defect I would not change them.

I'm will check 4052 soon. Thanks a lot.
How can I reliably check leaking FETs? I mean i have DCA75 which shows that fets are ok and they match between each other of the same kind quite well.
But things like gate->source o gate->drain resistance, what are the suitable values there?
I have checked Q330, and it has 6.8Mohm between gate and source on one side, but on the other its open gate to other terminals.

[Kleinstein]

Testing the FETs is tricky. 
One can test the overall input current of the meter, if that is OK (< 20 .. 50  pA), there is likely no leakage issue in the relevant part. Switching JFETs that are on may still no leakage as the gate source voltage would be very small.
A good test for the input current is connecting a low leakage capacitor (e.g. 10 nF PP or PS) to the input and watch the drift rate. One can also test with non zero starting points (e.g. more to the ends of the range).

Another test is the reaction to temperature rise: a leaky fet is usually also temperature sensitive. If it is about input leakage, maybe use the 10 M input impedance mode and other wise open input.

[bitomaxsp]

What I did so far.

Swapped Q320 and Q324. They are both seem to be good. But that didn't help.

I also measured some outputs HIOHM and /HIOHM. It seem very weird to me for HIOHM signal.

HIOHM controls Q324
/HIOHM (inversed of HIOHM) controls Q320/Q312 (both bood)

For 304.2-304.5 tests:
U309.1: -9.1V HIOHM. OK
U304.2: 1.25V /HIOHM


But in the following test:

304.6: ADin: 0.6840 (jumping)
   - R354: 6.99V
   U305.7(Q4): 4.9v ok
   U309.1: 0.2 HIOHM Q324.D: 0.2 okish I'd expect it ~ 1-2V And this is fluctuating, but less!
   U304.2: -9.3 /HIOHM Q320.D: -9.1V ok
   
304.7: ADin: 1.5866 - 1.7123 AVG: 1.6455
   - R354: 0.449-0.47 ok
   U305.7(Q4): 4.9v ok
   U309.1: ~-0.1-0.2 -0.35 HIOHM Q324.D: -0.3 NOK And this is fluctuating!
   U304.2: -9.3 /HIOHM Q320.D: -9.1V ok

Bot sure what to check next. I already cleaned all the board several time in isopropyl bath.
I think i'm going to swap 4052 U332 and U325

[Kleinstein]

Th eobvious problem with the test 304.7 is the large rippel. The ripple goes all the way to the limits. With that much rippple it is normal to get odd DC readings. This both from ripple overloading the meter used for the tests and extra current from clamping diodes.
One would really need some shielding to replace the likely open case. The may be a shield that is not connected.

[bitomaxsp]

Th eobvious problem with the test 304.7 is the large rippel. The ripple goes all the way to the limits. With that much rippple it is normal to get odd DC readings. This both from ripple overloading the meter used for the tests and extra current from clamping diodes.
One would really need some shielding to replace the likely open case. The may be a shield that is not connected.

When I fixed the device and it passed all test for 2 hours in a row i assembled it all back as it should be. And then I started testing all the modes.
The reason why I started digging in 304.6. and 304.7 in the first place was that i was not able to measure Ohm on high ranges. So problem was already there with all the shields closed. I can say that just by how reading behave. Then I removed the case and plastic cover and it didn't change anything wrt high Ohm reading.
Then I started poking into 304 series of tests to check all the conditions described and only then I noticed 304.6 and 304.7 are misbehaving.

So it is not shielding. It is something that is either leaking or noisy i suppose. I'll do more tests.

What I don't currently get at all, is how signal in 304.6 and 304.7 gets into U322(A/D buffer). Because in these 2 tests Q525 is off and Q328 is on according to truth tables. The only option is that it gets there through Q333 and normal input path. Which in this case i haven't checked manually, but test 303.2 passed.

301.1: 12.22mV  OK
301.2: 4.20mV  OK
302.1: 4.21mV OK

302.2: 5.897V OK
   R355: 0V
   R356: 7V
   CR335: 5.9V OK

303.1: 4.26mV OK
303.2: 5.897V OK
303.3: 3.199V OK

I'm kind of puzzelled now

[bitomaxsp]

I did a bit of clean up with alco again around High ohm section.
I've checked all the component in the input protection part. All good.
I swapped switched U332 and U325.

I did the 304.x test with plastic covers closed (but the board is out of the case with both lids on).
Things got much better in the sense that 304.6 has more stable AD/in measurement right on the money.
304.7 still jumping and about 1.65V, which I thin is low. 1.5866 - 1.7123 AVG: 1.6455

304.6: 0.7013V Vpp:60mV (AC coupled)
   U305.7: 4.9v ok
   U309.1: 0.2 HIOHM Q324.D: 0.2 Vpp:60mV AC <---- Why 0.2? Yes, Q324 is open. But I check in other modes. FETS are pulled to ~1.95V
   U304.1: 0.017 /HIV Q328.D: 0.017V ok
   U304.2: -9.3 /HIOHM Q320.D: -9.1V ok
   U308.4: -6.2V /DEVIDER Q525.D -6.1V closed ok

304.7: 1.5866 - 1.7123 AVG: 1.6455
   U305.7: 4.9v ok
   U309.1: -0.3 HIOHM Q324.D: -0.34V Vpp:60mV AC <---- Even less 0. But still should be open.
   U304.1: 0.017 /HIV Q328.D: 0.017V ok
   U304.2: -9.3 /HIOHM Q320.D: -9.1V ok
   U308.4: -6.9V /DEVIDER Q525.D -6.7V closed ok

I don't understand why noise goes to OpAMP output. Inputs seems to be clean. I can suspect that U309 is dead, but I replaced it as I already mentioned.

But most importantly, I managed to measure 10M and 200M resistors well. I tried to repeat it many times and it seemed fine.

I'll do more tests in current configuration.
Then I'll switch chip back again and redo the tests again. New chips are on the way.

I also have a suspicion that U341 (AD548) might be cooked. So I found that ADA4610-1ARZ should be good replacement for that. Ordered the part just in case.
Will keep updating.

[Kleinstein]

As a replacement for U341, the AD4610 should work, but an OPA140 / OPA141 might the better choice (lower bias spes and less LF noise - may help a little with the 20 V range noise). I somewhat doubt that U341 is really bad, as this would effect also the 20 V range and possibly others. It could still have higher than normal input bias.

U309.1 should follow the output of U331 and this way the reference voltage from U330.
For the 2 tests 304.6 and 304.7  Q324 should be on and the comparator output U309.1 thus not pulling the voltage down.
The seam rather low, to me, but I don't know what U330 gives out. One could check that point (e.g. U331 out, U330 pins 7 and 13).  The somewhat strange point is why there are 60 mV AC. This should be a relatively stable reference voltage.
It may be worth measuring with open input and shorted inputs.

[bitomaxsp]

As a replacement for U341, the AD4610 should work, but an OPA140 / OPA141 might the better choice (lower bias spes and less LF noise - may help a little with the 20 V range noise). I somewhat doubt that U341 is really bad, as this would effect also the 20 V range and possibly others. It could still have higher than normal input bias.

U309.1 should follow the output of U331 and this way the reference voltage from U330.
For the 2 tests 304.6 and 304.7  Q324 should be on and the comparator output U309.1 thus not pulling the voltage down.
The seam rather low, to me, but I don't know what U330 gives out. One could check that point (e.g. U331 out, U330 pins 7 and 13).  The somewhat strange point is why there are 60 mV AC. This should be a relatively stable reference voltage.
It may be worth measuring with open input and shorted inputs.

I did the measurements. (Relative to SCOM. J1026.2)
"U309.1 should follow the output of U331" - I don't get why?
They are not related to each other if I read the schematic correctly.
In my case they are not the same.

304.5
U331.6: 2.1183V
U330.7: 2.1180V
U330.13: 9.1096V

304.6
U331.6: 0.6966V
U330.7: 0.6964V
U330.13: 7.6882V

304.7
U331.6: 42.68-42.70mV unstable
U330.7: 42.49-42.50mV unstable
U330.13: 7.0340V
 
In addition to above I also did the following (Just in case, my unit passes all the test w/o errors):

301.1: 0V ±20mV : -9.89mV
301.2: 0V ±10mV : -9.89mV
302.1: 0V ±20mV : -9.89mV

303.1: 0V ±20mV :  -9.89mV
303.2: 5.9V ±0.59V : 5.8811V
303.3: 2.5V ±2.5V : 3.1636V

304.1: 0V ±20mV : -9.904mV
304.2: 0.78V ±8mV : 0.7592V
304.3: 0.65V ±65mV : 0.6317V
304.4: 0.56V ±56mV : 0.5462V
304.5: 0.7V ±70mV : 0.6974V
304.6: 0.7V ±70mV : 0.6865V
304.7: 2.2V ±0.6V : 1.6351V

305.1: 0V ±20mV : -9.9mV
305.2: 2.95V ±0.295V : 2.45V
306.1:  0V ±1mV @U322.6 :

307.1: 0V ±20mV : -9.9mV
307.2: 3.8V ±0.38V : 3.9474V
307.3: -2.9V ±0.4V : -2.9461V

308.1: 0.78V ±78mV : 0.7995V
308.2: 0V ±10mV @U322.6: 1.27mV

309.1: 89mV ±8.9mV : 88.88mV
309.2: 98mV ±9.8mV : 97.79mV
309.3: 92mV ±9.2mV : 93.57mV
309.4: ~10mV : 10.36mV
309.5: (0.475V + Test 309.4) ±47.5mV : 475+10.36=485.36 /481.93mV
309.6: (46.2mV + Test 309.4) ±4.62mV : 46.2+10.36=56.56 /56.90mV

310.1: 1.7V ±0.3V : 1.7330V

401.1: 0V ±1mV : 42uV
402.1: 32mV ±5mV : 33.03mV
403.1:  1mV ±5mV : -15uV

404.1: 4.21V ±0.4V : 4.1475V
404.2: 2.08V ±0.34V : 2.0543V
404.3: 1mV ±0.28V : 2.77mV
404.4: 2.25V ±0.34V : 2.2086V
404.5: 4.33V ±0.4V : 4.2622V

410.1: 7V ±0.18V : 6.958V
411.1: 7V ±0.18V : 6.958V
411.2: 7V ±0.16V : 6.958V
412.1: 1.7V ±0.3V : 1.7353V

305.2 seem suspicious. Rest is just fine.

[Kleinstein]

U331 is buffering U330.6.  So that part is working at least.
The main path to drive (turn on) the gate of Q324 and Q312+Q320 is via U331  -> R381 (100 ohms) -> R380/R379 (1M)
The comparator (U309.1) is there to turn off the JFET switches by pulling the voltage negative. If the JFET switch is on the gate voltage should be essentially at the level of U331 out.

A strange part is that U330.13 is droping for the tests 305.6 and 305.7.  This should normally be a relatively stable ~ 9 V and if at all go up (via R271).
This makes the transistor Q540 that drives U330.13 suspect (too little gain ?).

For the test 305.2 one could check the hi output during the test. This should be some 5.9 V according to the test. I would expect it to be a little lower.
The description of the test is a bit odd and I cant find the 9.2 mA source and R397 (may be different schematics version ?).
A lower voltage could be from some parts in the ohms protection

[bitomaxsp]

First of all that you for helping me.

Second. I swapped U332/325 back as it was in the original. The unit continued to work in the sense that it measures high ohm.
So I'd assume I didn't clean it properly before.

But test 304.7 and 305.2 still behaves same weird way. So I'll debug those further.
I did full set of measurements as I posted above, values didn't change, only 2-3 digit after dot.

U331 is buffering U330.6.  So that part is working at least.
The main path to drive (turn on) the gate of Q324 and Q312+Q320 is via U331  -> R381 (100 ohms) -> R380/R379 (1M)
The comparator (U309.1) is there to turn off the JFET switches by pulling the voltage negative if the JFET switch is on the gate voltage should be essentially at the level of U331 out.

A strange part is that U330.13 is droping for the tests 305.6 and 305.7.  This should normally be a relatively stable ~ 9 V and if at all go up (via R271).
This makes the transistor Q540 that drives U330.13 suspect (too little gain ?).

Thanks a lot for the pointers. I'll check Q540.

For the test 305.2 one could check the hi output during the test. This should be some 5.9 V according to the test. I would expect it to be a little lower.
The description of the test is a bit odd and I cant find the 9.2 mA source and R397 (may be different schematics version ?).
A lower voltage could be from some parts in the ohms protection

For 305.2 I did the debugging, But I can't understand the outcome.
So R394.1 (which R607 on the schematic) measures 5.88V rock solid. Which is 5.9V from test 302.2 as it should be.
R394.4, Q525.D = 58.31mV - FINE, Division by 100 is ok.
Q330.2: 58.31mV
Q330.6: 48.99mV

Rest of the voltages I put on the picture.
So if I do 2.45V/50=49mv This is rock solid. Which means x50 is working fine.

How is that possible that Q330.6 is 48 and not 58?

Could that be that Q330 is bad in low signals?
Could that explain low-current/high ohm instability in 304.7?

Again, thanks for helping with this

[Kleinstein]

The JFET pair J330 has some offset. 10 mV offset is not great, but also not unusual for a dual JFET. So this can still be OK and meter should still work with this. It can still cause some of the self test readings that don't use a difference / extra zero measurement to be slightly off the nominal value.
So the 305.2 step looks OK.
The actual test may include the 305.1 step before and look at the difference.
The 305.1 step also measures the offset of some -10 mV with a tolerance of +-20 mV. So the 10 mV offset is likely acceptable.
It is a bit odd that the step 305.1 is using gain 1 and not a gain of 50 as the next step.

[bitomaxsp]

The JFET pair J330 has some offset. 10 mV offset is not great, but also not unusual for a dual JFET. So this can still be OK and meter should still work with this. It can still cause some of the self test readings that don't use a difference / extra zero measurement to be slightly off the nominal value.
So the 305.2 step looks OK.
The actual test may include the 305.1 step before and look at the difference.
The 305.1 step also measures the offset of some -10 mV with a tolerance of +-20 mV. So the 10 mV offset is likely acceptable.
It is a bit odd that the step 305.1 is using gain 1 and not a gain of 50 as the next step.

I'm still digging.
I completely forgot to check power lines to COM resistance.

Here are my measurements (rel to COM):
+BS: 30k
-BS: 25k
+8V: 4.5M
-8V: 30k
5V: 490Ohm
+15V:105 Ohm that seem to low to me.
-15V: 2k

Does anyone have measurements? Or opinion should I dig dead part on +15 rail?

On another note, I check Q540 by desoldering. Working piece with hfe=200 @5mA.
But what I found interesting is R270. On the schematic it is 2k7 5%.
On the board it is blue (most likely 1%) 2k21.
If I get things right, then lower R270 value would result in not enough opened Q540. Which might explain not enough current to Q330.
Or am I raving here?

UPD: did the math correctly. 7V+8V*2.21/49.21=7.35V
Kleinstein was right, of course

[Kleinstein]

Measuring the resistance for the voltage ranges is tricky. Resistance tends to be nonlinear than thus the meter used can matter.

The change in R270 is only changing the voltage for the ohms mode a little. The base of Q540 should see some 7.3 V.
So U330.13 is driven via U324 - R271 and not Q540. That explains why the voltag can drop. Q540 just sets a lower limit, that is not yet reached. The limit would make sure the ouput would not go negative very much.

If not done, it may be worth checking the input bias current of the meter in the 2 V or 200 mV range in non AZ mode ? This would test a similar path the the voltage sense part for the high resistance modes. Extra leakage there could make the testcurrent unstable.

[bitomaxsp]

Measuring the resistance for the voltage ranges is tricky. Resistance tends to be nonlinear than thus the meter used can matter.

The change in R270 is only changing the voltage for the ohms mode a little. The base of Q540 should see some 7.3 V.
So U330.13 is driven via U324 - R271 and not Q540. That explains why the voltag can drop. Q540 just sets a lower limit, that is not yet reached. The limit would make sure the ouput would not go negative very much.

If not done, it may be worth checking the input bias current of the meter in the 2 V or 200 mV range in non AZ mode ? This would test a similar path the the voltage sense part for the high resistance modes. Extra leakage there could make the testcurrent unstable.

I did last series of tests, measuring voltages on different rails rel to SCOM.
Measurements tripple checked.
Since all test pass any way, i'll try to run the unit and test it in different modes.
If someone will endup with the same issue, we can debug further.
Or until I get the second unit

304.5: 0.697V
   - R366: 6.992V
   U305.7(Q4): 0.02
   U309.1: -9.3 HIOHM Q324.G: -9.1V
   U304.1: 0.017 /HIV Q328.D: 0.017V
   U304.2: 1.95 /HIOHM Q320.D: 1.95V
   U308.4: 0.7V /DEVIDER Q525.D 0.709V
   U330.7: 2.118V
   U330.13: 9.1103V
   U331.6: 2.118V
   U324.6: 9.109V
   Q540.B: 7.346V
   Q540.E: 9.109V
   J-R381: 2.1185V
   U316.14:  2.117V
   
304.6: 0.6863V
   - R354: 6.99V
   U305.7(Q4): 4.9v
   U309.1: 0.2 HIOHM Q324.D: 0.2 Vpp:60mV AC
   U304.1: 0.017 /HIV Q328.D: 0.017V
   U304.2: -9.3 /HIOHM Q320.D: -9.1V
   U308.4: -6.2V /DEVIDER Q525.D -6.1V closed
   U330.7: 0.6962V
   U330.13: 7.688V
   U331.6: 0.6962V
   U324.6: 7.687V
   Q540.B: 7.346V
   Q540.E: 7.688V
   J-R381: 0.234V
   U316.14: -15V
   
304.7: 1.5866 - 1.7123 AVG: 1.6455
   - R354: 0.4338V
   U305.7(Q4): 4.9v
   U309.1: -0.3 HIOHM Q324.D: -0.34 Vpp:60mV AC
   U304.1: 0.017 /HIV Q328.D: 0.017V
   U304.2: -9.3 /HIOHM Q320.D: -9.1V
   U308.4: -6.9V /DEVIDER Q525.D -6.7V closed
   U330.7: 0.0425V
   U330.13: 7.034V
   U331.6: 0.0427V
   U324.6: 7.559V
   Q540.B: 7.346V
   Q540.E: 7.034V
   J-R381: -0.4V
   U316.14: -15V

[bitomaxsp]

I have been playing around with second K2001M#2 that I have and it behaves much better than the one I described above. Even if powered with the cable that has grounding.

I accidentally found a thing with First device:

Iff I power FIRST device with the cable where Ground is not connected to socket Ground, then measurements in 304.6 and especially 304.7 are stop on!
I got 34465A in my possession (new one) and I did all the measurements again on first device

Rel to COM
5V: 4.895V 1.365mVAC
+15V: 14.914V 0.360mVAC
-15V: -15.079V 0.390mVAC
+8: 6.2V/ 2.7Vpp
-8: -7.0V/ 2.5Vpp

Rel To LO:
304.5: 0.697V
304.6: 0.6863V
304.7: 2.1V (this much much better that with grounding on) 

Any pointers where to look to debug it? or is it normal ?

[Kleinstein]

The AC amplitude and change with ground connected of not is still a bit suspect.  Maybe check if the shields are really connected / isolated where they should be. This includes shield windings at the transformer.

AFAIk the +-8 V are relative to an input boostrap and this can easy pic up hum with an open input.

[bitomaxsp]

The AC amplitude and change with ground connected of not is still a bit suspect.  Maybe check if the shields are really connected / isolated where they should be. This includes shield windings at the transformer.

AFAIk the +-8 V are relative to an input boostrap and this can easy pic up hum with an open input.

which shields do you mean?

green wire from socket is on the chassis. green wire from the transformer is on the chassis as well.

[Kleinstein]

The ground to chassis and transformer shield looks OK for the digial / output side. Still maybe check if the resistance is really low

There should also be some shielding for the analog side. This should be isolated from the mains ground and connected to the LO terminal. I don't know the details on how the K2001 is build and how much of the analog part is actually shielded. There should be some shielding near the input and likely also the AC input part.
The 2nd meter would be a part to compare.

[bitomaxsp]

I have done thorough comparison check between 2 devices I have, FIRST one that has fishy high ohm path and. the second one. I figured why measurements are on the edge for 304.7 (4.4 nA ohm range). Same reason as for test 305.2. High ∆Vgs ~10mV. It causes voltage at Q330.5 to drop by ~10mV which is that amplified by x50 and hence ADin sees ~1.65V.

I did a test, took Q330 from second device which has ∆Vgs ~3.3mV and 305.2 and 304.7 measurements are way better. I ordered new transistor. There is no guarantee it will have less ∆, but it worth trying.

[Kleinstein]

If the problem is only the offset from the dual JFET, one could have "trimmed" R402 or R404, by soldering a resistor (100 K range ?) in parallel to one of them.