Keithley 2002 8.5 digit DMM review and teardown

[VintageNut]

I don't have anything stable enough for high voltages just yet, if not counting 2400 SMU for that.

The 2400 is a good idea. It is not necessarily accurate enough, but it may be stable enough.

Let both 2400 and 2002 warm up for 3 hours. Local sense on the 2400 is probably good enough for this test. Set the 2002 for 20VDC, 10NPLC, 10 Digital Filter, Window:NONE, Filter Type:REPEAT. Resolution 8.5 digits.

Set the 2400 to output 10V. Turn output on.  On the 2002 turn the filter off. Filter on. Record the reading on the 2002 after a minute or two for settling.

Do not change anything on the 2400.

Change range on the 2002 to 200V DC. Filter off. Filter on. (saves time) Record the reading on the 2002 after a minute or two for settling.

Change range on the 2002 to 20V DC.

Set the 2400 to output 20V. Filter off. Filter on. Record the reading on the 2002 after a minute or two for settling.

Do not change anything on the 2400.

Change range on the 2002 to 200V DC. Filter off. Filter on.  Record the reading on the 2002 after a minute or two for settling.

Set the 2400 to output 100V. Filter off. Filter on. Record the reading on the 2002 after a minute or two for settling.

Set the 2400 to output 200V. Filter off. Filter on. Record the reading on the 2002 after a minute or two for settling.

The last two steps may not indicate anything but they should be included for completeness.

Post a table of results and we can see if there are any obvious differences between the 20V and 200V DC measure ranges on the 2002. The 2002 I used had a noticeable repeatable offset between the ranges. 

[Dr. Frank]

You need a precise 10:1 and a 100:1 transfer to calibrate 100V and 1kV.
Additionally, the 100:1 transfer on 1kV level must be compensated for self heating effects, which is quite difficult to realize.


Either a KV divider like the 720A or the Keithley 2002 might provide a 10:1 transfer from 10V to 100V, with about 1ppm uncertainty.

(Both are linear to 0.1ppm of input, that is 1ppm of output).

So your proposed transfer from 20V range to 200V range might  work, but I recommend to do the transfer from 10 to 100V, instead of 20 to 200V.
The 200V self heating might already affect the 2002 accuracy.

The self heating effect inside the 2002 HV divider will kill your uncertainty if you try to transfer from 100V to 1kV, as it is having about 20ppm heating effect, see footnote (9) in the specification.

Therefore, you'll need a Hammon type transfer standard, like the Fluke 752A or similar.

This has 0.2ppm of output for 10:1 and 0.5ppm of output for 100:1 transfer, latter including self heating effects.

As I also wanted to calibrate my former 332B/AF, I've designed my own ratio transfer standard, estimated (by design and by measurements) to have 0.5 and 1ppm transfer capability.
It's been described somewhere else on EEVBlog.

Frank

[VintageNut]

Hi Dr. Frank

You make very good points and I will have to do some reading to understand some of what you discuss.

The only reason for my suggestion TiN is that the 2002 I used (it was within the 1 year calibration)  seemed to be out of spec measuring 100V on the 200V range. This was demonstrable by measuring a stable voltage on the 20V range and immediately measuring that voltage on the 200V range.

[plesa]

Bingo, TL866 got new software 6.13 which finally supports DS1245 NVRAM.
Steps, NVRAM dump and details for those who interested here, as usual.

Just upgrade my 2002 from STD to MEM2 (128k). With Dallas DS1248Y-70+ without programmer, only through secret menu and works OK on FW A09.
Thanks for sharing info.

[egonotto]

Hello,


in the video https://youtu.be/v887C9I9hCE one sees the noise of a shortend
Keithley 2001.
Is these in a Keithley 2002 similar?
When yes, what is the last digit worth?

Best Regards
egonotto

[ruffy91]

I think you're measuring the thermoelectric effect, not noise. Try to touch each side of the shorting jumper. The voltage should go up when touching near the positive input and down when touching near the negative input (or inverse).

[egonotto]

Hello,

thanks for the answer. It can be, but for the Keithley 2010 you can find DC NOISE PERFORMANCE.
For instance for 5PLC 7,5 Digits and 100mV Range they have 110 nV RMS (2 min).

In Video https://youtu.be/J1tCSehiNyM they show a example at times 6:00.
They got a STD DEV from 71 nV. So the last digit seams meaningless.

Best regards
egonotto

[TiN]

2002 have less bouncing for last digits.

plesa
Did you managed to fix your 2002?
Thanks should also go to ManateeMafia, as he was the one to confirm that 2002 using DS1248.

[plesa]

Repair is in progress, last time I suspected Q211 and Q212 (3N163 and 2N3906) both were OK.
I can confirm that with the 5V version (DS1248Y-70+) it works like charm.

[TiN]

One of our members uploaded Keithley 2002 A10 firmware, thanks! I did not see it anywhere before, so far.

I need to get new UV EPROMs to try it, as mine meter came with OTP ROMs without UV window .
ROM binaries available for grabs, as always

P.S. Also added internal photos of 2440 SMU.

[TiN]

I did few tests before, with results here and here. CSV files with short on terminals (no leads) are there as well.
10 or 1NPLC 20VDC range results noise ~0.4-0.8uV.

Perhaps you can post your data?

[Macbeth]

I need to get new UV EPROMs to try it, as mine meter came with OTP ROMs without UV window .

Is there not an EEPROM Flash available? I would imagine much cheaper and more convenient than old UV EPROMS.

For my K2000 and K2015 I've found a pair of AM29F040B work perfectly as replacements for the OTP despite being twice the capacity and a fraction of the price. Easy reprogramming on the TL866 is a bonus!

[TiN]

I don't mind oldschool UV erasure, more fun to it. Was looking on 29F040 but those have A18, VPP, WE~ pins different from OTP ROM in my meter, so I did not want to do bodgery and jumpwiring in case it would not work drop in.

[Macbeth]

so I did not want to do bodgery and jumpwiring in case it would not work drop in.

Hahaha! When have you ever NOT done bodgery and jump wires 

But yeah, I haven't had to do that at all in my case. Of course if yours are wired up different then I agree.

I'm also annoyed at finding an old school 27C256 UV EPROM that despite an hour total in my PCB UV exposure box it didn't erase anything! Wrong wavelength it seems.

[SeanB]

125W mercury vapour arc tube at 2cm from the Eprom for 5 minutes pretty much guarantees erasure, or a melted EPROM. My first eraser was one with a 20W choke ballast, inside a box with a "safety" microswitch. Pretty much cooked the plastic box after a year or so, being used for 5 minute increments.

[TiN]

Remove any leads and just toss piece of fresh-cut copper wire into terminals. Secure them with ear cotton swabs (tin's patented technology), so you get nice contact there.
Keep any LED/CCFL or other equipment away from meter. Let it sit for 1 hour at least to warm up. Repeat measurements.

[Macbeth]

Remove any leads and just toss piece of fresh-cut copper wire into terminals. Secure them with ear cotton swabs (tin's patented technology), so you get nice contact there.
Keep any LED/CCFL or other equipment away from meter. Let it sit for 1 hour at least to warm up. Repeat measurements.

Hahaha. I am not the only one to do this it seems! I even bought a length of copper bar to turn on my little lathe and then worried about oxidation. Looked into gold plating thinking a simple electrolysis would do the job and soon realised that aqua regia, etc, is not so easy.

Just stripping some mains flex and jamming in with ear buds seems to do ok!

[plesa]

I just built a very short banana jack short, and the result is the same.

Another strange thing is in ACmV mode the noise is lower than DCmV, only less than 2uV pp, most of the time it is just 1uV pp.

The DCmV noise is now 12uV pp, at 8.5 digit, 1PLC, no filter, line sync ADC, line sync zero.

I think I will return this thing before its 30 day warranty expires.

As TiN mentioned use clean copper wire and also do not forget to clean input terminals (oxide cleaner like Deoxit or Kontakt Chemie Kontakt 61).
Before returning unit I will rather ask for discount instead of returning such a good meter

[plesa]

I just built a very short banana jack short, and the result is the same.

Another strange thing is in ACmV mode the noise is lower than DCmV, only less than 2uV pp, most of the time it is just 1uV pp.

The DCmV noise is now 12uV pp, at 8.5 digit, 1PLC, no filter, line sync ADC, line sync zero.

I think I will return this thing before its 30 day warranty expires.

As TiN mentioned use clean copper wire and also do not forget to clean input terminals (oxide cleaner like Deoxit or Kontakt Chemie Kontakt 61).
Before returning unit I will rather ask for discount instead of returning such a good meter

Just did exactly what TiN suggested. No improvements at all.

Same results form front and rear terminals? What about cleaning input jacks?

[plesa]

I don't mind oldschool UV erasure, more fun to it. Was looking on 29F040 but those have A18, VPP, WE~ pins different from OTP ROM in my meter, so I did not want to do bodgery and jumpwiring in case it would not work drop in.

Which type you are using? I would like to upgrade C05 to 6517A electrometer with upgrade 2002 to A10.
I remember that for some Keithley firmware there was mentioned in attached document what are the compatible chips.

[plesa]

Same results form front and rear terminals? What about cleaning input jacks?

The rear jacks are WTF. It reads ~50mV when shorted, and drifts heavily that can not be effectively RELed. Definitely needs a return.

By rear jacks I mean banana socket on rear side, this can help to isolate issue with front/rear switch. It depends on price you paid for it ($2k for working and $500 for not working unit). Keithley gear is good candidate for repairs ( K2002 is based on of the shelf components).
Did you pass self test after warm up?

[plesa]

All tests passed, this unit comes with a fresh cal cert, I paid 3k for it. I already contacted the seller, since I do not want to mess around with it. I paid 3k, not 500, so I do not want to destroy it by a mistake.
After receiving this refund, I will go for a low noise one, maybe a 7050, or wait until Keithley releases 8050. I do not need one urgently, since I have access to a 3458.

Oh now I understand it. 8050 or 8510 (join venture with Danaher/Fluke 8508A) seems to be in Volt nuts dreams more and more often

[plesa]

One of our members uploaded Keithley 2002 A10 firmware, thanks! I did not see it anywhere before, so far.

I need to get new UV EPROMs to try it, as mine meter came with OTP ROMs without UV window .
ROM binaries available for grabs, as always

P.S. Also added internal photos of 2440 SMU.

Sucessfully upgraded from A9 to A10 firmware. Used EPROMS were UV erasable 27C4001 and not OTP.

[TiN]

Did you get 9th digit now?
I wish we could know what's changed. My UV chips are on their way in postal system.

[TiN]

Updated ROMs to A10. Everything works OK, no recal needed.

Btw, I compared A06 and A09 firmwares, and here's what I found:

Firmware version A09 added next:
* Support for Type N thermocouples
* Support for SPRTD which was absent in A06
* New format setting : HPR (HPRECISION)
* Unknown command ISN?
* Coefficient settings for SPRTD: A4, B4, A7, B7, C7
* RTD custom user setting RZERO, RZER, to set resistance at 0°C
* It also seems that A06 have also only 9 settings profiles (SAV0 to SAV8), while A09 and later have 10 (SAV0 to SAV9).

No strings difference between A09 and A10