Author Topic: Another Keithley 2000 repair  (Read 931 times)

0 Members and 1 Guest are viewing this topic.

Offline jchw4

  • Newbie
  • Posts: 4
  • Country: ca
Another Keithley 2000 repair
« on: November 18, 2018, 10:53:00 pm »
I bought faulty Keithley 2000 on E-bay. After power supply repair it's now working fine (checked with DMMCheck Plus). All tests pass.

And everything seems to be working correctly, except for the strange problem with DCV.

When nothing is connected to the meter, it's display value is fluctuating.
In autorange mode, it starts with ~0, but then it goes to -1V in approximately 3 seconds. Then it goes back to 0, and cycle repeats.

In manual ranges below 100V, it decreases, and quickly goes to the overflow (less than a second in 1V range, and approximately 30 seconds in 10V range).
In 100V and 1000V display is stable.

A very nice resource https://xdevs.com/fix/kei2000/ (many thanks to the author!) ,  gave me schematics and a lot of information.

Connecting my U1252 to the inputs drops displayed value to ~.4mV, and matches what I get on the handheld DMM.
(Measuring current on the shorted 2000 inputs gives picoamps on my U1252 and drops displayed voltage to ~zero.)

Measuring TP104 always matches what I see on the display.

So the problem is somewhere around Q104,Q105,Q108,Q113, U113, U114 (with C132, R134, C251, C252).
I've replaced U113, U114, but it did not help.
Removing Q104, Q105, Q108 does not change behavior. (So everything before Q104 probably does not affect the behavior.)
C132, C251, C252 seem to be fine (I don't have LCR meter, but U251 measured 220pF as expected).
C252 was probably broken (measured 50pF), so I replaced it temporary with through-hole.

The only part that I did not try to remove was Q113, but it is covered by R117 on a separate PCB, so I decided to leave it as it is for now.
(I also did not touch R134, as it seem to be on the bottom side of the PCB. But it should not be important, as I have replaced U114, and checking C251 with my DMM also brings displayed voltage to zero.)

Well, now I am somewhat stuck.
I can try to remove Q113 (actually R117, then Q113), but this is difficult (I am worried about R117), so I need some help.

Did I miss something obvious?
« Last Edit: November 19, 2018, 12:51:48 am by jchw4 »
 

Offline RobertPS

  • Contributor
  • Posts: 19
  • Country: nl
Re: Another Keithley 2000 repair
« Reply #1 on: November 19, 2018, 01:19:17 am »
I bought faulty Keithley 2000 on E-bay. After power supply repair it's now working fine (checked with DMMCheck Plus). All tests pass.

And everything seems to be working correctly, except for the strange problem with DCV.

When nothing is connected to the meter, it's display value is fluctuating.
In autorange mode, it starts with ~0, but then it goes to -1V in approximately 3 seconds. Then it goes back to 0, and cycle repeats.

Did I miss something obvious?

That's normal behavior.
 
The following users thanked this post: TiN

Offline LapTop006

  • Supporter
  • ****
  • Posts: 441
  • Country: au
Re: Another Keithley 2000 repair
« Reply #2 on: November 19, 2018, 01:55:44 am »
That's normal behavior.

Yep, it's odd if you're used to other meters, but these Keithley's do it. If you look up youtube videos of them you'll usually see this.

You can buy "ghost voltage eliminators" (ex from Fluke) if you really hate it, but it's almost never a problem in practice once you're used to the oddity.
 

Offline alm

  • Super Contributor
  • ***
  • Posts: 1240
  • Country: 00
Re: Another Keithley 2000 repair
« Reply #3 on: November 19, 2018, 02:49:17 am »
Pretty much all meters with a > 1 GOhm input impedance do this. Regardless if they're made by Keithley, Keysight/Agilent/HP, Fluke or Datron. It is the input impedance charging due to induced current. Try to calculate how much current it takes into > 10 GOhm for the meter to read 10 V.

Once you switch the meter to a range with only 10 MOhm input impedance (100V or higher), it will go away. Connecting a 10 MOhm resistor across the input achieves the same. The relay clicking can be annoying and eventually wear out the relays, so I usually make sure to set it to the 100 V range or some function other than DCV if I have the leads disconnected. Because of the high resolution, 100 V is usually fine when I'm browsing with handheld probes (e.g. troubleshooting a circuit).
 

Offline macboy

  • Super Contributor
  • ***
  • Posts: 1918
  • Country: ca
Re: Another Keithley 2000 repair
« Reply #4 on: November 19, 2018, 03:24:59 pm »
I disagree with the others' assessment that this is normal. 3 seconds is way way too fast. If the drift took 3 minutes to overflow the range then it may be normal for at 6.5 digit class meter.

You may just need to clean the PCB really really well. Leakage on a dirty board could easily cause this, even though the contamination is invisible.
 

Online TiN

  • Super Contributor
  • ***
  • Posts: 3925
  • Country: tw
  • xDevs.com/live - 24/7 lab feed
    • xDevs.com
Re: Another Keithley 2000 repair
« Reply #5 on: November 19, 2018, 03:45:25 pm »
macboy

Maybe he have leads connected and floating and lot of stray fields :)
YouTube | Chat room | Live-cam | Have documentation to share? Upload here! No size limit, firmware dumps, photos.
 

Offline jchw4

  • Newbie
  • Posts: 4
  • Country: ca
Re: Another Keithley 2000 repair
« Reply #6 on: November 19, 2018, 08:57:06 pm »
That's normal behavior.

Yep, it's odd if you're used to other meters, but these Keithley's do it. If you look up youtube videos of them you'll usually see this.

You can buy "ghost voltage eliminators" (ex from Fluke) if you really hate it, but it's almost never a problem in practice once you're used to the oddity.

Wow! Probably I could leave it as it was. But after replacing C252 fluctuation became a little slower.

I disagree with the others' assessment that this is normal. 3 seconds is way way too fast. If the drift took 3 minutes to overflow the range then it may be normal for at 6.5 digit class meter.

You may just need to clean the PCB really really well. Leakage on a dirty board could easily cause this, even though the contamination is invisible.

I did that initially, before I started changing components. The drift was there.

I have not disassembled and cleaned the Front-Rear switch (as it is also difficult), but I don't expect any ghost voltage near it.

macboy

Maybe he have leads connected and floating and lot of stray fields :)



Hmmm..

Right now leads are not connected. Rate = Med. Filter is enabled (default).

After switching from 100V range to 1V  (by quickly pressing "Down" twice)  display value starts with -.5V and goes to overflow in 3-3.5 seconds.

After switching from 100V to 100mV I never see any output. It always say "Overload".


For tests below I used coax ("BNC to alligators" and "BNC to banana adapter"). Hoping this reduced some environment.

After I connect shorted leads , 1V range displays -0.000001 .. -0.000003. Sometimes  4. Slow rate gives -0.000002. .. 0.000017

100mV displays -000.0025 .. -000.0110. (Med). Slow rate gives -000.0022mV . .. 000.0069mV


100Ohm resistor results in the same display as shorted leads.

100k resistor (on my DMMCheckPlus), 1V  range gives -0.000004 .. -0.000015. Slow rate: -0.000009 .. -0.000020 . (it was mostly 15, but while I was writing this, it go to 20).

100k resistor (on my DMMCheckPlus), 100mV range and slow rate gives -000.0111mV .. -000.0250mV. (it started with 111, and then went down to 250 in two minutes, I believe).


What else could be wrong? Should I go on and desolder R117 and Q113?

I can probably connect 200 to my PC using RS-232 (what is the popular software for simple logging?).

If I start recording log, what drift should I expect?
 

Online TiN

  • Super Contributor
  • ***
  • Posts: 3925
  • Country: tw
  • xDevs.com/live - 24/7 lab feed
    • xDevs.com
Re: Another Keithley 2000 repair
« Reply #7 on: November 19, 2018, 09:19:34 pm »
If it's not clear from previous posts, you trying to fix the "problem" that is not a problem at all, but lack of experience with high impedance meter. All meters with GOhm DCV frontend pickup stray charge, no matter what brand.

This is by design, so meter input does not load your sensitive circuit when you try to probe stuff. So your soldering stuff in meter likely to cause more harm than good. Test input leakage by connecting 10Meg resistor at DCV or configure meter in ohm mode on highest GOhm range (should read overflow).
YouTube | Chat room | Live-cam | Have documentation to share? Upload here! No size limit, firmware dumps, photos.
 

Offline jchw4

  • Newbie
  • Posts: 4
  • Country: ca
Re: Another Keithley 2000 repair
« Reply #8 on: November 20, 2018, 01:30:12 am »
If it's not clear from previous posts, you trying to fix the "problem" that is not a problem at all, but lack of experience with high impedance meter. All meters with GOhm DCV frontend pickup stray charge, no matter what brand.

This is by design, so meter input does not load your sensitive circuit when you try to probe stuff. So your soldering stuff in meter likely to cause more harm than good. Test input leakage by connecting 10Meg resistor at DCV or configure meter in ohm mode on highest GOhm range (should read overflow).

My 2000 does not have GOhm range. On 120MOhm range it reads "Overflow".

100mV DC range, 1M (5%) resistor connected. Slow + default filter. BNC-Alligator cable.
Store -> 100 .

1st attempt:
Min: -0.3471 mV DC
Max: -0.3384 mV DC
Avg: -343.8332 uV DC
Std Dev: 1.495335 uV DC

2nd Attempt
Min: -0.3474 mV DC
Max: -0.2691 mV DC
Avg: -342.5604 uV DC
Std Dev: 9.378771 uV DC

3rd Attempt
Min: -0.3607 mV DC
Max: -0.3418 mV DC
Avg: -347.7259 uV DC
Std Dev: 3.895346 uV DC

4. Let's try 500 readings:
Min: -0.3608 mV DC
Max: -0.3418 mV DC
Avg: -345.0136 uV VC
Std Dev: 2.347543 uV DC

24h accuracy in 100mV range is specified as 30ppm of reading + 30ppm of range.  0.003% * (0.34mV + 100mV) = 0.0030102 mV = 3 uV

1V range, the same setup. 500 readings:

Min: -0.000355 V DC
Max: -0.000331 V DC
Avg: -340.996 uV DC
Std Dev: 3.369198 uV DC


24h accuracy in 1mV range is specified as 15ppm of reading + 6ppm of range.  0.00034 * 0.0015% + 1V * 0.0006%  +  = 6E-6 V = 0.000006 = 6 uV

So standard deviation is within tolerance, but I was hoping that displayed value should be within tolerance.

Right now board has a lot of flux on it, I'll clean it up and improve soldering a bit in a few days.

Do you still think that meter is working fine?
 

Offline jchw4

  • Newbie
  • Posts: 4
  • Country: ca
Re: Another Keithley 2000 repair
« Reply #9 on: November 20, 2018, 01:53:16 am »
10M (5%) resistor, 100mV range,  500 readings:

Min: -3.6063 mV DC
Max: -3.4504 mV DC
Avg: -3.532425 mV DC
Std Dev: 18.20320 uV DC

10M (5%) resistor, 1V range,  500 readings:

Min: -0.003549 V DC
Max: -0.003475 V DC
Avg: -3.509665 mV DC
Std Dev: 6.907359 uV DC
 

Offline macboy

  • Super Contributor
  • ***
  • Posts: 1918
  • Country: ca
Re: Another Keithley 2000 repair
« Reply #10 on: November 20, 2018, 02:12:46 am »
With only 1 MOhm on the input, I would expect a reading relatively close to zero, after having nulled the meter using the REL button with inputs shorted.  Although TiN didn't mention that step, it is quite important to establish this zero first. Only then can you know that the reading is due to current through the resistor rather than thermal EMFs, instrument drift over time, or some other source of offset voltage.  So to null the meter for this test, clip both alligator clips to one lead of the resistor (rather than to each other), then press REL to establish a true zero reading. This will null out all sources of voltage error up to and including the thermal EMFs in the alligator clips and even between the clips and the resistor leads. Then move one clip to the other lead of the resistor to measure the voltage across it.

Nevertheless, with readings that high and consistent, it looks like you have a leakage current issue, approximately 340 pA.

Another way to test is to observe the charge in a capacitor. Unless I calculated wrong, a 10 nF capacitor should charge up to 100 mV in about 3 seconds, or to 1 V in about 30 seconds. Connect the capacitor across the input, short it, remove the short, then observe the change in voltage. A steady climb in voltage will confirm that it is a stray current at work, rather than stray voltages.  Use a film capacitor like polystyrene or mylar, not ceramic.

I am not an expert on repairing these things, but I would look at the FETs in the input section. These are the essential components that provide the high impedance, low leakage input. TiN may have more repair advice as he is far more experienced with DMM repair.
 

Online martinr33

  • Regular Contributor
  • *
  • Posts: 178
  • Country: us
Re: Another Keithley 2000 repair
« Reply #11 on: November 25, 2018, 01:49:54 pm »
The 100mV range on these long scale DMMs always seems a bit suspect to me. Plus, a K2001 is kind of noisy. In my case, I have to max out every filter and averaging option to get something reasonable. The 3458a sidesteps the issue by droppinto 7.5 digits on the 100mV range (presumably, the fast reading susbsystem takes advantage of this).

For now, I would check out the readings on the 10V scale.

Something I have always wanted to do (and wll get around  to someday) is to look at the zero noise level by using another meter to measure at the ourput of the analog subsystem, before it goes into the ADC.

Doing this could tell you if you have a noisy ADC, or if the input amp and signal conditioning is noisy.
 

Offline 3roomlab

  • Frequent Contributor
  • **
  • Posts: 958
  • Country: 00
  • can you tell me how to cal sesame street meter?
Re: Another Keithley 2000 repair
« Reply #12 on: November 25, 2018, 03:32:50 pm »
@ jchw4

i was just like you when i got it, "why does the numbers jump around"

to be sure of what condition is the front end, you should take a comparison look at a few things
1) shorted noise on 10V 10NPLC, compare to specs = 1.5uV rms or less (less than 4.5uV P-P?). my used unit was dirty and noisy initially, it sometimes go 6uV.
https://www.eevblog.com/forum/metrology/dmm-adc-noise-comparison-testing-project/
modded units have even lesser noise
also, if the front end is "clean", the noise should appear as a nice well balanced noise bell curve, mine was not initially.

2) input impedance test compare (some tips here https://www.eevblog.com/forum/beginners/o-qn-on-funny-nplc-multislopenoise/msg830688/#msg830688)
this tip is learnt from alex nikitin lurking in the forum
charge a really good film cap (eg kemet 560V PP/PET goes at around 440G ohm, R75LF22204000K is rated 1.3T ohm) to around 11volts, measure the drop vs time, then calculate ohms.
from K2000 specs is > 10G. my used set when working well was at >50G 100G. when it was dirty, it only made 8G ohms.

i dont understand what you mean when you said your set is 120Mohm input? your set is special impedance?
so with the DMM alone, you can try to validate its own noise, and with a good capacitor its own input impedance. with the DMM logging itself over a long time like hours, it could also capture intermittent problems not noticed in short spans, like crazy noise spikes (from phones?)

regarding the schematics, which originally come from 38bbshot (china). as far as i know there are no known recent full revised schematics. but there are some mistakes, of which i highlighted here https://www.eevblog.com/forum/repair/repaired-keithley-2000/

there are also some who modified it for lower noise, TC, etc
https://www.eevblog.com/forum/metrology/project-pimp-a-keithley-2000/
https://www.eevblog.com/forum/beginners/o-qn-on-funny-nplc-multislopenoise/

since you are new, you might also want to know RS232 to PC logging if you are going cheap w/o GPIB
https://www.eevblog.com/forum/projects/is-this-a-real-ns-lm399-(from-polida-ebay)/msg769388/#msg769388

keep calm and log some DMM noise !
« Last Edit: November 26, 2018, 01:52:39 am by 3roomlab »
there should be a new global law, every new mobile phone model released should consume at least 1% less power than its previous highest model @ full load W/Gflops.
 

Offline feedback.loop

  • Regular Contributor
  • *
  • Posts: 128
  • Country: us
Re: Another Keithley 2000 repair
« Reply #13 on: November 25, 2018, 07:12:54 pm »
Why all this guesswork? Why not just follow a performance verification procedure from the service manual? At least approximately. I understand you may not have precise voltage and current sources, but most probably you can come up with something close to that just to see if the overall behavior makes sense.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf