FIXED : Keithley 199 ohms readings never settle

[Kleinstein]

Things look much better now.  Chances are no more current going the wrong way. The 2N3904 is not such a bad replacement, unless high voltage is applied. So currently no need for a exact replacement if the meter is not use on mains or similar.

There is still the problem with the reading, which is likely from the reference voltage reading finding it's way to the ADC.
However the path was test up to Q32, and a defect here is very unlikely as this would also effect most other readings.

AFAIK the DMM is using the same ADC to take 4 readings:   both sides of the DUT (sense Hi and Sense Lo)  and the "reference" resistor inside the meter on both sides. For some reason it looks like reading of the reference resistor does not work well.

One could try to use the scope to measure the voltages going to the ADC, e.g. behind the JFET mux (where Q30,Q32,Q34.. meet).
The G1 signal used to drive the FETs (where all the 100 K resistors for the JFET gate drive comes from) could also be interesting  - should be the same level but buffered.

Is the meter still reading overflow with a 0 Ohms resistor ?

[coromonadalix]

@Diplover   if needed i could poke in my 196 to help ??



I did the full reset command,  meter seems better, had to read the manual  how's to  loll,  and my probes are not good,  tossed them for new ones.

I'm on firmware b14  outch,   will backup them and the calibration chip too, and do a b17 upgrade, and recheck the readings.

This puppy will cost more than my recent 34401a calibration ??

[macboy]

@Diplover   if needed i could poke in my 196 to help ??



I did the full reset command,  meter seems better, had to read the manual  how's to  loll,  and my probes are not good,  tossed them for new ones.

I'm on firmware b14  outch,   will backup them and the calibration chip too, and do a b17 upgrade, and recheck the readings.

This puppy will cost more than my recent 34401a calibration ??

I also received a very high quote for calibration of a K199. I don't know why. Fortunately, adjustment is simple to do manually, and this can be done for any desired function/range without needing to do all of them. Your recently calibrated 34401A may be just good enough to act as a reference for calibrating a K199, but maybe not for a K196  (if following metrology guidelines of >3x accuracy). Even so, adjusting the the Keithley to agree as closely as possible with the 34401A may be desirable.  If the K196 isn't trusted and definitely requires adjustment, then that's what I'd recommend if you don't want to pay for a professional calibration. In my lab, I have adjusted my two K199's to match a K2001, and that's good enough for me.

[DIPLover]

Kleinstein :
 
Definitely not showing overflow measuring 0 ohm

Measuring a 60cm loop of banana cable
300Mohm : 000.000 M
 30Mohm : 00.0000 M
  3Mohm : 0.00005 M
300Kohm : 000.009 k
 30Kohm : 00.0012 k
  3Kohm : 0.00069 k
300Ohm  : 000.376 ohm

Readings are even quite stable, except in 30K and 300K ranges, where they remain jumpy.

Tried next with a big old 5W 51ohm resistor (meauring 51.69 according to my fluke 8050A)
300Mohm : 000.000 M
 30Mohm : 00.0000 M
  3Mohm : 0.036327 M
300Kohm : 004.010 k
 30Kohm : 00.4298 k
  3Kohm : 0.21392 k
300Ohm  : 116.589 ohm

[Kleinstein]

With a current through the DUT that looks OK (at least with the 5 K). The off readings must be a problem with reading one of the 4  ADC conversions:  Sense H +L ,  5 V and ref. resistor low side.  As the error is quite large it should be visible even with the scope. 

[coromonadalix]

@Diplover   if needed i could poke in my 196 to help ??



I did the full reset command,  meter seems better, had to read the manual  how's to  loll,  and my probes are not good,  tossed them for new ones.

I'm on firmware b14  outch,   will backup them and the calibration chip too, and do a b17 upgrade, and recheck the readings.

This puppy will cost more than my recent 34401a calibration ??

I also received a very high quote for calibration of a K199. I don't know why. Fortunately, adjustment is simple to do manually, and this can be done for any desired function/range without needing to do all of them. Your recently calibrated 34401A may be just good enough to act as a reference for calibrating a K199, but maybe not for a K196  (if following metrology guidelines of >3x accuracy). Even so, adjusting the the Keithley to agree as closely as possible with the 34401A may be desirable.  If the K196 isn't trusted and definitely requires adjustment, then that's what I'd recommend if you don't want to pay for a professional calibration. In my lab, I have adjusted my two K199's to match a K2001, and that's good enough for me.

Just read the manual, can be calibrated as you say with the front panel,    for the ohms calibration, it ask for  190ohm, 1.9k, 19k  190k, 100meg ohms values,  damn  i dont have thoses values.

[Kleinstein]

A new calibration would only fix a possible problem with the scale factor. So resistors near the end of the scale reading wrong.  Without a known good resistor as a reference one would not notice this.

A wrong reading for a short is a different problem. For 2 wire Ohms there usually is a separate point for zero adjustment, that is simpler than a real calibration. This adjustment does not need a special reference - maybe a 4 wire short for a very accurate 0 adjustment for the 4 wire ohms ranges. If one knows how, a 100 Ohms resistor can make a good 4 way short.

Unless one has really good reference values and knows well what to do, one should not do a full calibration with adjustment at home.
Even than the normal way is to first just check of performance / accuracy. Only if really needed and the reference values are known good an adjustment is done. Just a check can get away with a less accurate reference, as the meter is not altered - so no harm done. 

[macboy]

Just read the manual, can be calibrated as you say with the front panel,    for the ohms calibration, it ask for  190ohm, 1.9k, 19k  190k, 100meg ohms values,  damn  i dont have thoses values.

You can enter the actual value using the front panel numeric keys, to override the default calibration point. For either Keithley 196 or 199 look at section 6.4.5 step 4, 8 and 10 in the respective manual from tek.com .  In this way, you only need a stable reference, not an exact value.  Connect the voltage to both the K196 and the reference meter (your calibrated 34401A). Enter the reading from the reference meter as the calibration point on the meter being adjusted. Similarly for resistance, first measure the resistor on the reference meter, then connect it to the meter being adjusted and use that value (don't touch the resistor as this will cause thermal shift in value). It is best to keep the reference at close to full scale on both meters, which isn't possible in this case since the 34401A is 1000000 count and K199 is 300000 count. So for the 3 VDC range on the Keithley, you could choose around 3 V (100% full scale) and measure with the HP in 10 V range (30% FS), or choose 1 V for 100% FS at 1 V range on the HP but only 33% FS on the Keithley. You could try both ways and see where you get a better result.

[coromonadalix]

I dont want to hijack the  OP thread
thks for the infos  Kleinstein and Macboy

Checked for precision foil resistances at 0.02%   they all come around 12$ usd each minimum, 17 weeks lead time ?,  they go as high as 39$ usd ??

I'll do that when i'll be on version b17, just need to erase the eeproms and recap this puppy.   I'm impressed with the low pin counts for the data transfers between the meter section and digital section. Love to see the isolation coils for signal coupling

The need is : >10megs impedance at 300mv and 3 volts, its a huge help for one sensitive system at my job, i dont need absolute precision.

I could choose any  resistors values near the end of the scale 
Say  : 300ohm, 3k, 30k, 300k, 3 meg and 30 megs (If i find them)  since it overrange at 3.03 ??    its a full 3,030.000 counts display for many of the old 6 1/2 digits meters series like this.

[DIPLover]

With a current through the DUT that looks OK (at least with the 5 K). The off readings must be a problem with reading one of the 4  ADC conversions:  Sense H +L ,  5 V and ref. resistor low side.  As the error is quite large it should be visible even with the scope.

Back on target, some more measurements.
Testing previous path on 30K setting
5VR        : 5.09727 V
U22 pin 10 : 0.87887 V
U22 pin 11 : 0.89050 V
Q11 pin  1 : 0.87878 V
External 5K: 0.70187 V

U20A pin 2: 0.87423
U20A pin 3: 0.87384

Next looked at Q32 (see attachements)
Note actual frequency on Q32_pin3 is 5.7Hz like all others. (176ms period)

and A/D Input buffer amplifier U46 (see attachements)
The 20% gain on what I understand to be a unity gain buffer, is worrying.

[Kleinstein]

The signals so far look good, though a little odd with the long Zero phase at Q32 pin3.  However this is more like a software oddity and should not be something that breaks. It looks like all the 4 signal levels (5 V, 0.9 V, 0.7 V and 0 ) get through to the ADC.

The gain of the ADC input guard buffer is indeed odd. I first though it could be due to loading by the scope, but the voltage looks like 6 V_pp, so more than going from 0 to 5 V reference. It not just a simple buffer, but coupled to the bootstrapped supply for the amplifier.
So the next points to check might be the amplifiers for the bootstrapped supply of the input amplifier. In the K196 schematics this is U37A+B.

This could also effect other ranges, if measuring from a positive higher impedance source - but this may not be such a common case in normal operation it is only -3 V to +3 V, while in Ohms mode it is up to 5 V and possibly with high high impedance. The effect might be visible as extra input current in the 3 V range with 2-3 V at the input. 

[DIPLover]

This whole region is much simplified on the K199.
Please see attached page 3 of the schematic.

[Kleinstein]

The K199 circuit is indeed much simpler. The extra gain is also trivial to see: there is the 1.1 K resistor in series to the OPs (U46) output and the feedback from there. So the gain is limited to directly at the output of the chip.  So nothing wrong here.

So far it looks all normal from the analog side. This makes me think that the wrong readings could be due to a corrupted calibration for the ohms ranges.  So maybe someone tried an ohms calibration with the defect hardware - this would have corrupted the calibration and would now give wrong readings.

Ideally the calibration would be just a single constant per range (the reference resistor value). So even if the calibration is wrong it should still be linear. So one could do a short linearity test on one of the ranges (e.g. the 30 K range).  Something like values of around
0 / 100 / 500 / 1 K /  5 K / 10 K / 20 K .
If the readings are still linear, I would consider a wrong calibration.

[DIPLover]

I think you are on to something...
Tried with some 1% resistors and identical leads on both meters.
Remember previous tests with precision 5K showed Overflow on K199.

Linearity test on 30K ohms range
ohms      K199             K196         Factor
0       : 00.0013* K      00.00027 K    4.815
100     : 00.8310* K      00.10035 K    8.281
510     : 04.1876* K      00.50569 K    8.281
1 K     : 08.2864* K      01.00045 K    8.283
3 K     : 24.8115* K      02.99537 K    8.283

Linearity is very good.
By this factor, the 5K would measure over 40K which explains the overflow.

The last 2-3 digits are still very jumpy on the K199 whereas they are stable after a few seconds on the K196. This was my original concern (the readings were much closer to reality in spring 2018 than when I reactivated this thread in winter 2019).

Could that be just an averaging setting on the meters? I will check.

For calibration, I understand I could measure some resistors in the R30 series with the K196 and just enter the exact values in the K199 right? I don't need super precise magical resistors or calibration standard?

[Kleinstein]

The resistor used for calibration would set the scale factor. So it would effect the accuracy. The reading from the K196 could be a good value.

Doing calibration would delete the old values.  I don't know for sure if one can do a calibration of the ohms ranges only. The instructions should tell.

The scattering could be due to the speed settings. The extra factor from the scaling would also apply to the noise - so a higher noise is also expected.

[DIPLover]

The K196 is in cal so it's values should be good enough for the 199. I bought the K199 mainly for monitoring with the great display and GPIB, no critical measurements will be made with it, but still, I want it as accurate as possible.

I am buttoning-up the K199 (what splendid construction, really a joy to work in, kudos to whoever did this work at Keithley in the late 1980s).

I will investigate the Calibration process. If I can do just the ohms, I will do that as the DCV and ACV readings were already spot-on (need to re-check now).

Did try using the front panel filter and it cleans up nicely.

Will report back later.

[coromonadalix]

Me too     i'm impressed for this kind of meters  "all the brown boxes of this era"

[DIPLover]

Alright, we are getting there.

Fortunately, the meter allows calibration of single ranges of single functions. 
Manual asks for R19 series resistors for calibration (probably because suggested calibrator gives these values)
But you can enter your own calibration point easily, so after letting both meters warm up for 3 hours, I carefully measured a range of R20 series 1% resistors using the K196 and proper 4-wire hookup in the low ranges.
Did the 300 ohm, 3K, 30K, 300K and 3M ranges using 200 ohm, 2K, 20K, 200K and 2M respectively.

Now for the good news :
1. My original problem, jumpiness of readings is SOLVED.
The meter now converges reasonably quickly to a stable reading on all ranges.

2. The 300, 3K, 30K ranges are working fine across their full scale. 30M and 300M were already good, I did not touch them.

The bad news is the 300K and 3M are only good at the calibration point (200K and 2M).
Their error factor is not linear, but makes a nice curve whose amplitude is higher for the 3M range.
Some raw data :

K199 measurements on 300K range :
ohms    K196            K199          factor
200     200.1315        000.357 K     1.784
2 K       1.995296 K    003.538 K     1.773
20 K     20.10081  K    033.272 K     1.655
51 K     50.5125 K      075.213 K     1.489
200 K   198.7545 K      198.756 K     1
300 K   299.9202 K      244.916 K     0.8166

K199 measurements on 3M range :
ohms    K196            K199        factor
200     200.1315        0.00179 M   8.944
2K        1.995296 K    0.01777 M   8.906
20 K     20.10081 K     0.16720 M   8.318
51 K     50.5125 K      0.37790 M   7.481
200 K   198.7545 K      0.99867 M   5.025
300 K   299.9202 K      1.23055 M   4.103
510 K     0.509145 M    1.51577 M   2.977
1 M       0.994468 M    1.80842 M   1.808
2 M       2.014896 M    2.01504 M   1
3 M       2.959862 M    2.08937 M   0.706
OPEN      Overflow      2.26766 M

K199 measurements on 30K range :
200 ohm :  00.2001 K   
2 K     :  01.9954 K
20 K    :  20.1015 K

So there is something common to these 2 ranges but unused by the lower and upper ranges that is still broken.
I will get to enjoy another disassembly it seems

What those 2 ranges have in common is the use of R17B, which is controlled by Q13. I don't really know what to look for however.

[Kleinstein]

The 300 K and 3 M range look like there is still some resistance in parallel to the DUT. So a little like the original broken Q16, but to a lesser extent.

Is the 30 M really working ?   (e.g. checked at at least 2 points, like 100 K and 10 M).


This could be the replcemant for q16, if there is one. As this is current toward the negative side we can exclude Q9.
One of the FETs that are used as switches could also leak towards the gate. This would only effect those switches that are off. So Q13 abd Q3 are not candidates, but  more like Q11. Q34 should be on when in 2 wire ohms mode.

One might be able to see the current at the 1 M resistor at the gate (e.g. R4b).
 

Chances are the broken part is also at the same input / line as the broken Q16 was.

[DIPLover]

30 M range is working quite well with the old calibration :

K199 measurements on 30M range :
200     200.1315        00.0001 M   
2K        1.995296 K    00.0019 M   
20 K     20.10081 K     00.0200 M
51K      50.5125 K      00.0505 M
200 K   198.7545 K      00.1987 M
300 K   299.9202 K      00.2998 M
510 K     0.509145 M    00.5092 M   
1 M       0.994468 M    00.9945 M
2 M       2.014896 M    02.0149 M   
18.2 M   18.21961 M     18.2193 M (2x 9.1M in series)

[DIPLover]

Here is page 1 of K199 schematic so we can have it embedded.
Q9 is not populated on K199.
Q16 is the one we replaced with a 3904.
300M range is as good as 30M.

[DIPLover]

Finally made a man out of me and attacked this systematically. 

I printed 5 copies of the schematic page and highlighted the active signal path for each range.
Then I started looking at each transistor and its control while measuring a 1M resistor with the 30M and 3M ranges (which had the most similar signal paths, differing only in Q13 being ON for the 3M range and OFF for 30M).

I started from the right side of the schematic.
Q35 was OK.
Q32 was OK.
Q33 was OK.

It appears Q11 has failed short. All three pins read +4.68 V (30M) or +4.56 V (3M) even though it should be off in both cases. Control pin U2p13 properly reads -14.8 V in both cases.

Q12 is OK.
Q3 is OK.
Q34 is OK.
Q13 is OK.
Q13 gate is -13.4 V and control U2p2 is -14.8 V for 30M range. (OK, should be off)
Q13 gate is 4.3 V and control U2p2 is 4.7 V for 3M range. (OK, should be on)

Now to find a proper replacement for Q11, which is Keithley part number TG-128, which I have seen in other threads is a pn4392. Unfortunately now discontinued. Pinout is Drain(1), Source(2), Gate(3). TO-92 case.
Vgd 40 V
Vgs 40 V
Ig    50 mA
Pd   625 mW

I don't have ANY jfet on hand.

Closest thing I have is 2N7000 MOSFETs. Pinout is Source(1), Gate(2), Drain(3). Could be twisted. TO-92
Vdss 60 V
Vdgr 60V
Vgs 20 V (plus or minus), 40 V peak.
Id    200 mA
Pd   350 mW

Could that work in a pinch? What do you suggest as a pn4392 replacement?

[Kleinstein]

Q11 failing makes absolute sense. A MOSFET replacement would not work, as the control signal is wrong and the mosfets have an internal parasitic diode in parallel.  If only used for the ohms part and thus one polarity, there would be a small chance a depletion mode MOSFET could work, but these are quite special parts - not sure if TO92 parts exist.

So it should be a JFET. The exact type is not that important and for the pinout drain and source are usually interchangeable (most JFETs are symmetric and no difference between drain and source).  Using wires and a SOT23 JFET could be a last resort. Otherwise check the TO92 JFETs available at your source.  One wants low gate leakage and at least a 30 V (better > 35)  voltage rating.

[DIPLover]

PN4392 are discontinued and hard to find, but Digikey had PN4391s in stock for a dollar apiece.

Specs are quite similar, except for IDss and Vgs(off) which are doubled.
I figured the higher Vgs(off) would be OK since the LM339 (U2 pin13) sends -14.8V when the transistor should be off, well above worst case Vgs(off) of -10 V.

So I ordered some yesterday, got them today and replaced Q11.

Some measurements looking for linearity of 300 K and 3 M ohm ranges.
Cold start, open case without shields, 2 wire measurements


K199 measurements on 30K range :
200 ohm :  00.2003 K
2 K     :  01.9958 K
20 K    :  20.1028 K


K199 measurements on 300K range :
ohms    K196            K199          factor
200     200.1315        000.083 K     0.415
2 K       1.995296 K    000.828 K     0.415
20 K     20.10081  K    008.348 K     0.414
200 K   198.7545 K      082.525 K     0.415
300 K   299.9202 K      124.409 K     0.415
510 K     0.509145 M    211.459 K     0.415
OPEN      Overflow      Overflow K

K199 measurements on 3M range :
ohms    K196            K199        factor
200     200.1315        0.00005 M   0.25
2K        1.995296 K    0.00057 M   0.286
20 K     20.10081 K     0.00579 M   0.288
200 K   198.7545 K      0.05730 M   0.288
300 K   299.9202 K      0.08640 M   0.288
510 K     0.509145 M    0.14684 M   0.288
1 M       0.994468 M    0.28686 M   0.288
2 M       2.014896 M    0.58108 M   0.288
3 M       2.959862 M    0.85381 M   0.288
OPEN      Overflow      Overflow M


K199 measurements on 30M range :
200     200.1315        00.0002 M
2K        1.995296 K    00.0020 M
20 K     20.10081 K     00.0201 M
200 K   198.7545 K      00.1987 M
300 K   299.9202 K      00.2998 M
510 K     0.509145 M    00.5092 M
1 M       0.994468 M    00.9947 M
2 M       2.014896 M    02.0150 M
3 M       2.959862 M    02.9618 M   

Quite promising, I think we nailed it 

Now stitching up the patient and letting it warm up before doing a new calibration of the 2 ranges.
K196 is already warm and ready to go.

[DIPLover]

SUCCESS!   

The Keithley 199 is now good on all ohms ranges, across their whole span.
And no more jumpiness, all readings stabilize very quickly.

HUGE THANKS to Kleinstein for his guidance!!

Thanks also to everybody else who took the time to read this topic and/or offer advice.
I have (re)learned a lot doing this repair, it was well worth it and good fun.

I copy this here from edited first post, for future reader jumping directly to the end :

****** EDIT on 2019-03-20 ******
In the end my choice of 18 series resistor for the original test was very bad, since the calibration point for this unit are at 190, 1.9K, 19K etc. Thus being very close the errors didn't look too bad and I was more concerned with the jumpiness. It turned out all ranges were WAY out of wack except 30M and 300M.

In the end I replaced transistors Q16 BJT NPN with a 3904 (it was leaky). This fixed the 300, 3K and 30K ohms ranges.
I also ended up replacing Q11 N-Channel JFET with a pn4901 (failed short, all three pins reading the same 4.5+ Volts when it should have been closed and gate was receiving -14.8 V control signal). This fixed the 300K and 3M ranges.
I also had to recalibrate all ohms ranges except 30M and 300M since they had been calibrated (some by me) with the faults.
*************************************