Keithley 2010 With Errors 300.1, 300.3, 301.1, 301.2

[J-R]

Since there is no full repair manual for the Keithley 2010 available, which error codes are most likely correct, the ones listed in the K2001 or the K2000 service manual???

In addition to the error codes listed, the symptoms are that the 10V DC range goes to below -10V and sometimes overload with nothing connected, and with a 10V reference connected it is about 2mV low on the 10V range, but fine on the 100V and 1000V ranges.  It's also spot on if given -10V DC.

So far I've checked all the power supply values, the DCV analog signal switching table 4-5 and cleaned the front/rear switch.

[Kleinstein]

The Keithley 2001 is quite a bit different from the 2010.  The 2010 is more like an improved version of the 2000, with a few aspects (2nd reference) taken from the 2182.
Especially the resistance part is quite strange with the 2001. So expect the error codes to be closer to the 2000.
So chances are the problem is with the resistance measurement part, like the 2 sense inputs.

With the input in high impedance mode it is normal that the input voltage is kind of undefined when open and it usually drifts when removing a short. So reading overload when open is not a problem, that is not unusual.  One can use the speed of the drift to get an idea on the input bias current.

A problem with the sense H input could lead to extra current flow also in normal voltage mode. This could be causing the problem with the +10 V. A test could be neasuring the input current during this test, e.g. with a handheld DMM in voltage range in series. The 200 mV range with 10 M input impedance acts like a 20 nA rang. Similar one could measure the current to the sense H and sense L terminals in 4 wire ohm mode (e.g. with some 10 K or 100 K DUT).

[J-R]

It drifts to about -10V over a few seconds.  Oddly, waving my hand over the open case sometimes causes it to crash to around zero, then it zooms back down to around -10V again.

The resistance measurements are off a bit, I would say maybe 0.5%.  2W and 4W are the same (2W was nulled).

I grabbed a couple DMMs to test the input H/L and the current flow is absolutely zero for all DCV ranges EXCEPT 10V, which is approximately -330nA.  No current flow detected via any other combinations of input/sense hi/lo.

[Kleinstein]

The current with 10 V would explain the voltage error with the 10 V reading.
It may be interesting to see if the ohm sense H input also shown a similar current when the voltage reaches 10 V. This would likely need the external ref used when in 4 wire ohm mode.

0.5% off is an only small error for a broken meter. So it could be related to the problem of the 10 V range with some extra leakage.

AFAIK the input section is still relatively similar to the K2000, but with a lower noise AZ amplfier for the buffer.

[J-R]

I'm not following everything you're saying about the "ohm sense H input".


I did run some more tests:

When in 10V DC range there is no current detected on the sense terminals or voltage using a >1G Ohm DMM, just the Input Hi/Lo.

With Ratio enabled but Sense Input disabled, the leakage current flips between two values, approximately -300nA and -125nA.  I assume it is switching between the inputs to make the measurements.

With Ratio enabled and Sense Input enabled, results are similar.

With Ratio disabled but Sense Input enabled, there is no leakage current from the main Input Hi/Lo.


Has anyone done more work on the reverse engineering of the 2010 schematic?  I found the "Keithley2010_Schematics_and_Firmware.zip" with file dates from 2017, but am not sure which schematic is relevant to this issue, as they are not complete and some are not labelled.  Also, xDevs seems to have taken down a lot of this info.

[Kleinstein]

The ratio mode switches between reading the normal voltage and the voltage between sense_H and sense_L.
I don't see how one can enable / disable the sense inputs in ratio mode - usually the use of the sense inputs (for the "reference" voltage is implied by the ratio mode.

Form the block diagram switching between the Sense H and the normal input is via JFETs before the buffer. So some current spikes are expected from switching between the 2.
Chances are this part is relatively close to the K2000 circuit.
For the board photo I have, the input buffer looks a bit like the AC buffer in the K2001 - so a JFET pair stabilized with an AZ OP-amp, and not the AZ-op-amp alone.

I have the likely same old 2017 partially reverse eng.  crude plans. Parts have quite some errors in the amplifier/buffer part.  Another point to note is that there are different versions. Especially there are versions that use the old TF245 resistor array and some with more standard arrays (5 K +50K) and also better OP-amps at the reference (no longer the poor choice LT1124 for the reference amplification).
The input section part with the JFET switching may still be relevant and better. AFAIR the plan came with repair of a damages input section - it could be worth looking for the old thread as the current damage could be similar.

My main suspects would be the JFETs (Q104,Q105,Q108,Q113) for input switching before the buffer.  Suspect are especially the FETs that are switched off when the extra leakage happens as these see a large negative gate voltage when the input is positive. So the fets towards the input and sense H (if I interprete the tests with sens H right) are less suspect.
A possible test would be to measure the leakage current with 10 V at the input and than selectively warm up (e.g. some 5-20 K) the FETs one by one (e.g. place a hot resistor close to it). Chances are that the bad fet would react most to a temperature rise.

[J-R]

I looked into the Sense feature.  Page 2-20 of the user manual covers the behavior of the various combinations of Ratio and Sense In.  Whether you have SENSE IN enabled or disabled with RATIO being enabled is basically used to control how the Range buttons work.  Do they control the main Input range or the Sense input range. (RA vs. RS on the display.)

Of course enabling SENSE IN by itself simply allows you to use the 100mV, 1V, and 10V ranges for DCV.  I did try this out and was surprised to find that I can use this as a crude work-around for my issue.  The meter is perfection on the 10V range using this method, AND also even if the Input Hi and Lo are directly bridged to the Sense Hi/Lo.  So whatever is impacting the 10V range using Input Hi/Lo is not in-circuit when using the Sense terminals (and to re-iterate, even with the Input Hi/Lo bridged to Sense Hi/Lo).  However, when Ratio is enabled and the 10V range is selected for the Input, the ratio reading is still way off.

Next I looked closely at the "Analog signal switching states" and I do not see any combination of on/off that would ONLY impact the 10V range, even for a single JFET.

I did run a heat test by placing the tip of a soldering iron next to each JFET and got really nothing out of any of the JFETs, except Q105, which doubled the leakage to about -700nA over a few seconds.  The state table shows Q105 on for 100mV and 1V and off for 10/100/1000V.  The big difference I see between 10V and 100/1000V is >1G Ohm input vs. 10M Ohm.  But the math doesn't quite line up with what I would expect to see by simply changing the input impedance.  Also, Q105 is OFF when using the Sense inputs too and they are not affected.

For the next step I might go ahead and try swapping Q105 with another JFET to see what that does unless there are some better ideas...

[Kleinstein]

Q105 should be a relatively normal switching JFET like MMBF4393. So one could try a new FET or for a first try even just remove the FET and keep the path open. If the fault persists Q105 is likely OK. So no need to remove another FET to swap.

There could also be an effect of the JFETs used to switch the lower end of the divider  (Q114 and Q136 in the K200n plan, which seems to be quite similar, including the part numbers). They also get the full input voltage and if they don't turn full off they could be an issue.  These could be a bit tricky parts if they choose  MMBF4392 or similar to get a lower R_on, there is a chance for them to not off fully with +10 V.

There are a few reports of LM339 chips failing in DMMs that are used to control the gates. A gate not going all the way to -15 V (or the like) could be such an issue.
So maybe check the LM339 (maybe also a LM393 as dual) outputs if they really reach the neg supply. A visulal check at the chip that controls the HS signal may also be good - maybe a weak link to a neighboring pin ?

[J-R]

Yes, I do believe the currently available part number is MMBF4393LT3G.  Initially I was going to nuke them all from orbit as they are only $3 for 10.  But I suppose the logical approach is better.

I was concerned the DMM would fault if I removed the JFET completely, but hey, I just yanked it and the leakage current is gone.  So it's probably safe to say either Q105 is the culprit or something directly associated with it.  I might take the easy path first and just replace it with new.

All tests pass except a couple (600.1/600.2) which hopefully are due to the missing JFET as it didn't throw those errors previously.

The 2010 service manual mentions LM339D and LM393D, my board has LM339M and LM393M.

[J-R]

I see some Keithley 2010 units have C288 ("CAP, .1UF, 10%, 1000V, CERAMIC") installed and others do not.  Does anyone know what the story is on that?

(To clarify, it's the very large, orange ceramic capacitor next to the resistor divider.)

[J-R]

Received the replacement JFET (MMBF4393LT3G) and installed.  Zero detectable leakage (or below 1nA) and all tests pass.

Certain resistance measurements are still a bit off, although after looking at the specifications it seems I might be expecting too much.

Also, the "trick" of dragging the backs of the buttons on a piece of paper for a few inches worked perfectly.

Last calibration was 4 years ago, tempted to ship it off...

[Kleinstein]

Before sending it to calibration it would be a good idea to do a bit more testing, e.g. looking at the noise, drift, turn over error and input bias currents. There are quite a few tests that can be done at home and the tests also help to get familiar with the "new" meter.  It would be bad to find out that meter still has a problem after calibration. Especially for hobby use one usually has the time for a good initial test before sending out the meter.  After a repair there is always a slightly higher chance for failure, even if very careful.

The input leakage should normally be below some 50 pA for the voltage input, maybe more is acceptable for sense low.

[J-R]

We'll see, I'm not sure how far I want to go down the rabbit hole.  I've had this K2010 for a bit over a year and it's been working quite well until the sudden 10V issue.

I did run some more tests and couldn't detect any input leakage down to 100pA.  Also used the ~10M Ohm input of another DMM but things were a bit noisy, perhaps even from leakage from the other DMM.  Also, we're firmly in the volt-nut area now, having to pine over temps, leads, drafts, external interference, PCB contamination, etc.

[Kleinstein]

A nice way for testing the input current is using a low loss (e.g. PS or PP or good C0G, ~ 1-10 nF) capacitor and whatch the charging discharging. This way one can get to the single digit pA resolution and also test at different voltages.

I agree that testing the full specs of the DMM gets you in the voltnut area. It still helps to anderstand the effects of drafts and possible interferences to make full use of such a meter.

[J-R]

A 10M Ohm load was too noisy to get meaningful measurements.  Assuming proper maths, paralleled a 1.111...M Ohm resistor with 10M, got reproducible results of less than -5pA of leakage on the 100V and 1000V ranges, but about -30pA from Sense In and about -40pA from Input.

Replaced Q108, Q114 and Q136 (last, of course!).  Q136 appeared to be the offender and I am now down to about -15pA from Sense and about -25pA from Input.  So only a 15pA reduction.  Seems like this should be adequate?

[Kleinstein]

The 100/1000 V ranges are usually not critical with input current, as there is the divider in between and thus internal bias causing some offset, that can be compensated to a large part.

AFAIK the specs should be a something like < 50 pA. So even the 40 pA were OK, though not great. 25 pA of input current looks OK / good enough.

Leakage for Sense_low is far less critical - a few 10 nA would still be good enough for normal use, though ideally it should still be in the < 100 pA range or like the others for this meter.
With less curcuit connected sense H has it a bit easier to get low input bias.

The test with 10 M ohm could be a little noisy, but for a current measurement a lower resistance would not help. Part of the noise can also be actual variations in the input current, e.g. with fluctuations of the internal temperature.

[J-R]

Turns out my math was super-shady, so went back to the 10M via another DMM method and got -40pA of leakage on the 10V/1V/100mV ranges on the main inputs and -25pA when using the Sense Inputs (common jacks connected).  Verified with multiple DMMs.  Decided to call it good for now.

Tested everything out using the calibrated references I had and used the typical tricks of checking readings on multiple ranges and flipping the voltage reference polarities to check negative values.  I found quite a few things I wasn't happy with at all, so went ahead and did the DCV/DCI/Resistance calibration.  Now everything is much tighter, especially resistance.  I think it's safe to say there was some pretty serious leakage going on when the last calibration was made, so correcting that threw things off a bit.

I'll check out a few more things over time.  I have a bunch of references that are about due for calibration, so maybe when they get back I'll see what's what.  Probably will be worth shipping this out for a professional cal/adjust+data if it behaves itself.