Keithley 195 Multimeter repair -

[TheUnnamedNewbie]

I bought a broken Keithley 195 on ebay last week hoping to keep myself busy over the weekend. So far, it's been a succes (the being busy part, the repair part not so much). I won't be posting much pictures of the actual unit, since as far as I can tell, there is no visual damage. Manual

Just to be clear: I've already spent most of my weekend working on this!

What does the meter do?

Digital/interface wise, the meter seems to operate fine. It does not complain about anything, really, and all self-tests pass. It can still store settings I change, so that also seems to work.

When I apply no voltage to the input, the following values are displayed:

Range	Value
1 kV	87.7 V
200 V	1.50 V
20 V	-OFLO
2 V	- 1.19 V
200 mV	-OFLO
20 mV	-OFLO
20 MOhm	18.47 MOhm
2 MOhm	OFLO
200 kOhm	91.654 kOhm
20 kOhm	OFLO
2 kOhm	0.916 kOhm
200 Ohm	OFLO
20 Ohm	OFLO

When I zero the meters on the ranges where they display something that isn't OFLO, I get half-usefull measurements. (like, to a few percent - also very poor linearity).


Troubleshooting

I checked all the voltage rails - they are fine. Reference voltage is also fine.

From the fact that some ranges show some sign of life suggests to me that this problem must happen before the ADC.
The digital section also seems to work. All the control signals go through one single pair of optocouplers, so those must be fine too.

So that makes me consider two main areas: The input multiplexer and the VGA buffer. When considering the switching of the input mux and buffer stages in the different ranges we find the following (Every range also tends to use Q103 and Q114 in each measurement stage)

Range	OFLO?	Input MUX switches	Gain switches
1 kV	No	Q111	Q117
200 V	No	Q113	Q117
20 V	Yes	Q113	Q118
2 V	No	Q109	Q117
200 mV	Yes	Q109	Q118
20 mV	Yes	Q109	Q119
20 MOhm	No	Q111	Q117 and Q118
2 MOhm	Yes	Q111	Q118
200 kOhm	No	Q111	Q117 and Q118
20 kOhm	Yes	Q111	Q118
2 kOhm	No	Q111	Q117 and Q118
200 Ohm	Yes	Q111	Q118
20 Ohm	Yes	Q111	Q119

So this shows that only when Q117 is on something usefull comes out (Or is it perhaps that when Q118 or Q119 are on, something goes wrong?).



In order to make my life easier, I tried to find the locations of most components on the board (as there is no silkscreen):

[coromonadalix]

You may have some resistors network bad, or some of the input mosfets switching  busted ??

Check all the voltages +15 -15  +5 -5 vdc  and the unreg voltage too


Figure 8-7 is the input schematic,  check everything related to the manually choosen range  up to the input buffer ic U131 and the -2vdc reference

Table 7-17, 7-18, 7-19 explains lots of checks

[TheUnnamedNewbie]

You may have some resistors network bad, or some of the input mosfets switching  busted ??

Check all the voltages +15 -15  +5 -5 vdc  and the unreg voltage too


Figure 8-7 is the input schematic,  check everything related to the manually choosen range  up to the input buffer ic U131 and the -2vdc reference

Table 7-17, 7-18, 7-19 explains lots of checks

Thanks for the reply, but I was still adding my progress I have found all the scehmatics and the manual, and checked all the rails. The steps for troubleshooting were of little use to me, as I don't always understand what is going on. I will re-do these checks later now that I better understand the meters inner workings.

I'm gonna get my matlab started so I can plot som graphs of the input voltages VS time to show some interesting things going on.

[coromonadalix]

the section table  7-12 explain all the paths taken on any range ...  this is your holy grail   loll

[TheUnnamedNewbie]

the section table  7-12 explain all the paths taken on any range ...  this is your holy grail   loll

Yes, I am aware... How do you think I got to the table where I look at the relation between the ranges where there is OFLO and the ranges where certain switches are active?

[coromonadalix]

No problemo

[TheUnnamedNewbie]

No problemo

Sorry, my reply was a bit harsh, I was frustrated (about something unrelated) and it got to me. Thanks for your input and for trying to help!

[Kleinstein]

A not working JFET is very real possibility. Quite often a broken JFET causes problems when off, by having extra gate leakage or not really turning off all the way. The on case is sometimes still working.

Just to avoid a common mistake: in the high input impedance modes (e.g. 0.2, 2 and maybe 20 V) the voltage with an open input is not well defined. So the point to test would be with shorted inputs.

With the odd R120 as a bridge, a good test might be to measure the voltage at R120 with a scope. In AZ mode for DC volts there should be a sequence with 0 V,  a 7 V reference and the selected input voltage.  The instructions should tell the sequence.

[TheUnnamedNewbie]

Yes, I should have added the fact that those voltage and current measurements were all with input shorted.

I am working on getting some pictures of the waveforms with my scope but wanted to try using MATLAB to do so and then that didn't work so I had to troubleshoot that and lost track of time.... Will post some screenshots of the scope in an hour or so. I found some interesting results (buffer output shoots to -10V in the broken modes, when q118 switches).

So a damaged and leaky q118 might explain some stuff, though I'm still confused (but will explain when I get home)

[TheUnnamedNewbie]

So here is an update, with plots!

First off: I removed R120 and input some DC voltages externally, and measured the output voltage. (I used the test programs built in to the meter to stop it from switching gain every x ms). First glance some wierdness was going on - the x100 range started at -0.5 V when DC input. I realized I'm not sure of my measurements now, so I will repeat these later today or tomorrow.

When I place R120 back, and measure the output at TP5 (the output of the input buffer), I get the results below. I think what is most interesting about this is the fact that there is a clear relationship between the ranges that output nonsense and the ranges that output some data. What I think is even more interesting, is that the ranges that do spit out data, still have a reference voltage of just -1.5 V and not, as one would expect, -2 V.

Note that I made an error on the time-axis of these graphs, and they should go from 0-100 ms instead. So multiply everything on the x-axis by 2

30 V DC in, 1 kV range



30 V DC in, 200 V range:



15 V DC in, 20 V range:


1 V DC in, 2 V range:


0.1 V DC in, 200 mV range:


0.01 V DC in, 20 mV range:

[Kleinstein]

There seem to be several things wrong. One is that the zero level seems to be more near -1 V. So there might be something really bad with the amplifier !

The -10 V level seen in some cases could be because of Q118 not working, but it could also be a problem with the driver. So it may be worth to check the gate driver level first. If the amplifier has so much offset (e.g. - 1 V) this alone could cause trouble for the modes with gain.

The offset of the amplifier could be due to U138 (excessive bias current) or U131.

[TheUnnamedNewbie]

I'm going to have a look at the amplifier tomorrow. I imagine what I could do is simply remove R120 again, and then really characterize the amplifier for the various gain settings. During my break at work today I'm also going to write down what I expect in terms of levels for a given DC input around the feedback resistors - hopefully by comparing this to what I measure I can then try and figure out what is going on.
The -10 V on the output of the amplifier also corresponds perfectly to the signal applied to Q118 I think, but more proper measurements should be done to make sure this is always the case.

Also, is there a way to scale my images on here?

[Kleinstein]

With the amplifier output at below -10 V the JFETs may no longer turn off reliably. This could cause some confusion.

It might help to also look at the voltage at R120 - the wrong voltage could also happen if the -2 V and GND channel are on together. One could also force the R120 node to ground, as the sources are high impedance.

The x axis scale is still odd: I would expect the times for each voltage step to be at least 1 power line cycle. So the total time should be a little longer.

For the images, one could scale the graphics before hand. For such graphs the .PNG format would be better than JPEG. 

[TheUnnamedNewbie]

For the images, one could scale the graphics before hand. For such graphs the .PNG format would be better than JPEG. 

Yes, I think I will have to scale them more in the future.  I'm used to keepng everything in SVG/PDF formats for all that vector-graphics goodness, hence I had no idea of scale.



That aside: the more I think about it, the more I suspect it is not just a single JFET not working. I will try and do input/output sweeps of the amplifier on all the range settings later this week. Hopefully I can figure out what is wrong before Thursday so I can get components before the weekend - otherwise it will have to wait till 2019.

[MadTux]

Have you tried to "recalibrate" (like apply 10V from PSU, if you have nothing better) a range and see if that makes the readings more plausible?

I had a similar problem with weird readings on a K196 once and all that was wrong with it, was that the NVRAM with calibration data was erased/corrupted and thereby the data on the display was calculated with rubbish in NVRAM. After a basic recalibration, all readings were much more close to the real input.

[wn1fju]

I concur with the previous post.  I had a K195 with similar symptoms - the cal memory was full of nonsense, hence the readings on the
display were being scaled and offset with nonsense too.

Also, check to make sure if there is something like a relay to switch in the ohms function, that it is not faulty.  My meter's relay was closed
permanently and the previous owner, not realizing anything was wrong, calibrated the unit.  That's how the cal memory got corrupted.

A simple check is to measure the input resistance of the meter in the DCV ranges.  If something in the ohms line is affecting it, the resistance
will be lower than it should be.

[Kleinstein]

The measured curve do no look like a calibration problem, as the output voltage of the amplifier is already way off in essentially all cases.

It is more like a problem switching the right input (e.g. one or more if the JFETs or maybe controlling chips - likely LM339). A bad amplifier (e.g. U131 or U138 ) could be another possible cause. Gate leakage from Q117-Q119 might also cause a similar problem.
The most likely cause is trouble switching before R120, as the voltage change is about 1/4 of what it should be. So a little like there are 3 bad channels always on, with a negative offset, e.g. from the ohms source. As in in the cases without gain there still is a reasonable sequence, it is likely that  the JFETs for Zero, the -2 V reference and the input for 2 V and 200 V and 1 kV  (Q114,Q103,Q109,Q113,Q111) are still working. So not that many left.

[rastro]

I suspect that Q117 is bad.  If it is shorted or partially shorted it will stay "on" when it should turn "off" on other settings and causing them to fail (overload).  It may also impact the settings that it should be "on" if there is leakage through the gate also causing non-overload but erroneous readings.

rastro

[Ordinaryman1971]

On a different note, did you get it with the AC measurement board or it's just DC and Ohms? Option 1950 I believe.
I've had one of those not taking measurements at all and it was problem with the j-fet transistors. Changing them solved the problem.

[TheUnnamedNewbie]

I concur with the previous post.  I had a K195 with similar symptoms - the cal memory was full of nonsense, hence the readings on the
display were being scaled and offset with nonsense too.

Also, check to make sure if there is something like a relay to switch in the ohms function, that it is not faulty.  My meter's relay was closed
permanently and the previous owner, not realizing anything was wrong, calibrated the unit.  That's how the cal memory got corrupted.

A simple check is to measure the input resistance of the meter in the DCV ranges.  If something in the ohms line is affecting it, the resistance
will be lower than it should be.

I do agree with Kleinstein that the calibration seems unlikely to be the culprit. Later today or tomorrow, I'm going to run a program that will sweep all the ranges of the input buffer and plot gain. This will hopefully give a better indication if the amplifier is to blame or not. If these results come back good, I think it is safe to assume Q117, Q118 and Q119 are good - after all if they were not, their errors should show up in gain issues.

I'm also thinking of just ordering a bunch of LM399's ahead of time, just so I can replace them right away if they are indeed the broken parts here (my local farnell dealer closes down for the holiday period so..)

I will have to also locate a handfull of replacement JFETs, but am unsure which one I should get. The one used is a keithley part. Ideally, just one JFET will be broken and I can salvage the one of the option input?

On a different note, did you get it with the AC measurement board or it's just DC and Ohms? Option 1950 I believe.
I've had one of those not taking measurements at all and it was problem with the j-fet transistors. Changing them solved the problem.

Nope, just got the basic one. Building my own option 1950 might be a fun project though, so I will see what I do once the meter is fixed.

[rastro]

You could also just un-solder Q117 and see if the range/functions that don't use this transistor (20 V, 200 mV, 20 mV, 2 MOhm, 20 kOhm, 200 Ohm, and 20 Ohm) go back to non-overflow condition.

rastro

[Kleinstein]

The usual switching JFET would be 2N4393/4392 or J112.  J202 could also work. For high ohms (and low leakage) parts 2N4118 would be a good choice. However it gets increasingly difficult to get JFETs in TO92 or TO18 case.

[rastro]

The usual switching JFET would be 2N4393/4392 or J112.  J202 could also work. For high ohms (and low leakage) parts 2N4118 would be a good choice. However it gets increasingly difficult to get JFETs in TO92 or TO18 case.

It looks like Q117 has only a keithley designation.  Unless there is a marking on one of the 'TG-128' JFETs you'll have to make a best guess.  Obviously an N-Channel JFET but you also want to check location of Gate,Source, and Drain pins on this unknown part.  On the K197's some of the switching JFET's have oddball pin placements.  In those cases I used a readily available  N-Channel JFET but had to cross 2 of the legs.  It's been a few years and I can't remember specific part numbers for a replacement - but it was easy sourced.

To determine pin outs you need to look at the PCB and schematics to find other components that connect to the gate for example.  Just don't assume pinout compatibility with a replacement JFET. 

I recently repaired a K485 that also had a bad range JFET (switch).  It actually used a J201 J210 N-Channel JFET.  I used a J201 in it's place without a problem.  So you may want to consider a J201 but you may need to account for different pin outs if you use it in place of Q117/TG-128.  J201 should be easy sourced.
https://www.mouser.com/ProductDetail/ON-Semiconductor-Fairchild/KSD5041QTA?qs=ljbEvF4DwOOPfenfyneIug%3D%3D&gclid=CjwKCAiA9efgBRAYEiwAUT-jtCP2RC4zxu9kIwT1TR2kEsuem2GYHaeJeAigEqs4EP31QAA_MXoqcBoCfN0QAvD_BwE
rastro

[Kleinstein]

Those switching JFETs are usually symmetric, so drain ans source are interchangeable. This simplifies the pinout question to having the gate right.

At least for the first test one could use a more standard JFET and if need look for a low leakage FET later. The low leakage type may be only more stringently tested to ensure better worst case ratings.

[rastro]

If they have dots of paint on them they are probably hand selected out of a batch of JFETs.  I suspect they are also looking for good frequency response for any AC signals.