Repair : Old Keithley 2000 teardown and fix

[Macbeth]

TiN,
Be careful not to rub the silkscreen writing or logos after soaking in the hydrogen peroxide. They won't get bleached, but they will be softened to the extent that any rubbing can remove them (don't ask me how I know  ). After drying, they will harden up again.

Oh I dunno, it seems like a golden opportunity to rid the display bezel of it's '7001 Switch System' provenance. 

[TiN]

I got 4142's cover out on the sun, will see how it goes. If it work out well, next will be this 2000 and few 2001 panels.

[macboy]

TiN,
Here's another trick I have learned. If the ABS (grey) plastic has dents in it, you can remove them with a hot air gun. ABS likes to return to it's original shape if you heat it to a just-plastic state. I set my hot air to about 140^o but I don't 100% trust the calibration so YMMV. Use a small nozzle pointed directly at the ding and make sure to shield things like the display lens and overlay, they don't react the same way! Don't overheat either, as then the plastic will lose its 'memory' and won't return to its original shape.

My 4 year old once chewed up a Lego man's contruction hat, it was all squished and the plastic had whitened from the stresses. I put the heat on it (just heating the entire thing), first the white-ish marks disappeared, then it just went back to 98% of it original shape. It fits on the Lego man's head again, and you wouldn't know what happened to it. Lego is also ABS. I don't know any other plastics that will do this, most will just get soft and limp when heated rather than return to original shape.

[TiN]

Anyone have extra K2000 transformer for sale? It's P/N TR-299. Mine ceased working out of no obvious reason on another powers during debug (no any output on secondaries, no heating, nothing at all)..

[TiN]

No need for transformer, unit got fixed, calibrated and send to other volt-nut member

Calibration video:



Performance test:



Used calibration path:

DC Voltage - Agilent 3458A calibrated summer 2015 -> A9 HP3458A PCBA with LTZ1000A -> my fixed HP 3458A -> Time Electronics 9823 -> Keithley 2000. This experiment was covered here.
Resistance - Keithley 2001 calibrated Feb 2014 -> Resistor Vishay PG foil 10K -> HP 3458A -> Time Electronics 9823 -> Keithley 2000.

Also few comparisons were made using 3 DMMs together : 3458A-2002-2000:

10VDC, 3458A + 2002L + 2000 after 10V cal
1mADC, 3458A + 2002L + 2000
10mADC, 3 DMMs on same loop
100mADC, 3 DMMs on same loop

[Dr. Frank]

Hello Illya,

I'm not d'accord with your calibration procedure.

At first, there is no estimate of your basic 10Vdc and 10kOhm uncertainty in your lab, for evaluation of the overall 3458A uncertainty.

Then, in your performance test list, you did not add the 3458A uncertainties for the different ranges and modes, neither relative to your basic 10V/10k uncertainty, nor absolute uncertainties.
Therefore, all the deviation numbers of your K2000 are flapping in the breeze.

Additionally, as the KEI2000 has been adjusted just before this verification, its 24h limits should be taken as test limits, not the 1yr. limits.
That would detect any further systematic calibration errors.

Then, 10mVdc and 10 Ohm are at least already outside specification after your calibration / adjustment.
Obviously, there seem to be systematic errors in your setup.

You might do a better offset voltage nulling, to get better agreement.
Maybe you should better use a precise 10:1 / 100:1 Hamon divider to calibrate 10mV and 100mV ranges.

If the KEI2000 does not support offset compensation, it is necessary to do a better estimation of possible errors for 10 Ohm.
Same goes for the 3458A, usage of OCOMP is mandatory.

100mAac probably is also outside specification, as the 3458A already has 800ppm uncertainty for 60Hz.

1000Vdc is also very questionable, as the 3458A itself is not very accurate here, at least +/- 15ppm systematic error, according to specification.

For 10VAC, I assume you used ACV SYNC?
I saw the "MORE INFO" annunciator briefly, but I'm not sure, how you really made this measurement... generally, you should document all 3458A modes in a detailed manner.

Anyhow, for ACV SYNC,  you should better use 7 1/2 digit resolution to get more stable and more precise reading on the 3458A, or at best should use Swerleins algorithm.

Then you may have used silicone or similar cables for Ohm calibration, and twisted plus and minus together.. so you will get big errors for 10k, 100k, 1M, and 10M, due to leakage currents.
I don't know, if the shielded 4 pole cable, you used during verification, is insulated with PTFE, or better material. That would improve measurement precision.
It's also not visible, whether you used appropriate delays on both instruments for 10k and 100kOhm ranges, when using OCOMP feature on the 3458A.

In summary, your calibration process and uncertainty estimations need to be improved.

Frank

[HighVoltage]

I don't know, if the shielded 4 pole cable, you used during verification, is insulated with PTFE, or better material. That would improve measurement precision.
Frank

Where do you buy such high quality PTFE based test cables, if I may ask?

[TiN]

First of all, thank you for feedback. I agree on your valid points, and I intend to address them accordingly.

I'd like to note fact that I had this K2000 sent to voltnut just for a shipping cost, basically free. Ofc it does not justify crude and bodged up calibration procedure revealed here, but will hopefully help to understand why it was done this way. Meter already spent half a year sitting in the pile, waiting for "it's time to come", and I thought getting it sorta calibrated and giving out to less equipped hobbyist would do more good than harm.

Now excuses away, few details on 10V/10k. As input standard for DCV i used as is A9 module (unmeasurable TC), which was measured by last year calibrated HPAK 3458A/F732B (annual spec). More details are covered in short article above, how I transferred DCV to my 3458A. I did not use 10V for CAL, but direct 7.1V CAL. I expect that worsen uncertainty by 40%.
Cross check with K2002 (cal'd in 2007) and second K2002 (cal'd in 2009, but fixed psu issues) got agreement in 2.1ppm window.

10Kohm situation is much worse, as only reference was K2001 calibrated by local Tek.

Now for setup, OCOMP was not used due lack of time and no low-noise cabling either. Even worse, AC unit was blowing air just overhead , so temperature variation was also far from desired +/-0.1C.

100mA on 3458A is rather bad, thanks to 50ppm specced 1 ohm shunt used, revealed by 22ppm/K TC, while K2002 is at least twice better. (ACAL joker not involved). High probability that 1 ohm DCI shunt on A1 will get replaced with some decent VPG foil prior to final cal.

I do plan to spend some qual time writing (bodging) python calibration app, to properly setup instruments and support all required features, as well as generating friendly HTML reports with all data covered.

On good side, I have a project coming soon to do a full calibration of both my K2002, foil resistance standards and LTZ-modules to known and verified absolute uncertainty. After reverse transfer back to my 3458(which will stay home, operating 24/7 as usual) this will give us answer how off I am today. I expect numbers about 10ppm DCV and 90ppm 10Kohm. Hence the rush to get K2000 shipped, as I still working to prepare everything I'd take for calibration.

Also Time 9823 is not calibrated, waiting for same thing to get adjustments after good uncertainty is known.

Hope this train of thoughts is inline of what actually need to be done. Oh, and last thing - I'll be testing code and more similar stuff with K2001. I'll use 24h spec this time, and will be happy to hear you out on that as well. I'll post in my 2001 thread once I have something.

[Macbeth]

Great job as always TiN 

As for Teflon coated twisted pair, etc. Really? I would much sooner trust TiN's (uncertified) FREE calibration than any other typical ebay offering K2000/2015 pulled from a rack and sold as working, proof of calibration if any a worn out sticker 5 or 10 years old. Let's get things in perspective!

But... between the beauty treatment, dodgy transformer, and calibration I seem to have missed the actual fix? 

[TiN]

Tbh actual fixing slipped even for me.

I'd say it was more system repair, than individual parts troubleshooting. That's the excuse I have, not to be able reporting "it was bad opamp U153 which caused problem A, and bad resistor R272 which cause problem B" 

As received, box was throwing all the possible errors, reading jibberish. Cleaned mess and replaced few parts in area where electrolyte from capacitors spread.
That fixed few self-test items, but still were bad reading. Then I had it sitting few months, forgot what was done already.
Switch replacement - got conditions when it sometimes work, sometimes not. Then sitting few months, again, forgot what was done already.
Recently issue was found to be one of infamous 2512 resistor on input FE circuitry, which got cracked due to PCB stress and barely making contact.
Replaced all of them with 25K PTF56 25ppm/K ones, no randomness anymore. But new issue - meter started to hang from time to time and reset itself from no reason.
Replaced SRAM chips, replaced firmware ROMs to 39SF020 Flash with A20 FW. Hangs/reboots seem to be gone.
Again sitting few months..
Vref voltage sometimes drop to some low voltage, causing selftest errors from 200.x and so on. So I started to check circuitry around VREF. That's when "Transformer died". Actually it was OK, just mains fuse got blown quietly without any magic smoke, as I messed up on power polarity while replacing LT1124 opamp near reference  . Don't know how that happen. New LT1124 installed, all back to normal (sorta).
Again sitting a month..
After assembly meter pass self-test once it's cold, but after warmup it fails undocumented 306.2/306.4 tests (which were added since A19 or around that fw?).
Since it seem not to affect actual measurements, and I ran out of time, I let it be and calibrated box as is.

Violated my own rule of documenting everything, and here's the result in it's glory. So people, don't skimp on documentation, it's often more important than actual fixing.
Now it's almost arrive to new user, which has no equipment to debug or test it, so I'd expect no further work to be done. 

[alouko]

No, this is not 7001.
It's broken 2000 hooked up to front panel from 7001.
Obviously, technically it's possible to use 7001 or 2001 front panel in 2000, since it's same interconnect.
Both instruments series (2000,2700,2750,2790, name it, all single-line Keithley stuff) and dual-line using UART to send data from main DMM 68000 to front panel MC68HC705C8.

Issue is only different protocol commands, so someone persistive enough with plenty free time could bodge a little MCU in between to translate 2000's data/readings into 2001/7001's format and enjoy bigger, nicer VFD on 2000 instrument. I'm not gonna go that road. My 2000 native VFD panel is very dim, so I was wondering if 7001's display would work as is.
As result is no, I will go forward and desolder NEWHAVEN DISPLAY DD-51C glass from 7001 board and solder on 2001-112 front panel PCB, since I lack of one 2001 FP.

Is there any documentation on the protocol used between DMM 68000 and front panel MC68HC705C8?  If yes, dimmed VFD could be replaced with a newer display interpreting display commands.  Of course, it would look totally different and certainly not as nice as the original VFD.  But still better than a dim VFD, though.

Perhaps I have to reverse engineer the protocol.  Interesting to know if the main 68k controls individual VFD segment or is the protocol in a higher level.

[Macbeth]

Is there any documentation on the protocol used between DMM 68000 and front panel MC68HC705C8?  If yes, dimmed VFD could be replaced with a newer display interpreting display commands.  Of course, it would look totally different and certainly not as nice as the original VFD.  But still better than a dim VFD, though.

Perhaps I have to reverse engineer the protocol.  Interesting to know if the main 68k controls individual VFD segment or is the protocol in a higher level.

For sure the front panel has its own controller, it's not done by the mainboard 68k. When you power on you will be given the version of the mainboard firmware (e.g. A19) followed by the display firmware (e.g. A02) as 'REV:A19      A02'

Nice little reverse engineering project - it should be quite simple (he says!).

[TiN]

Its fairly simple, binary UART with 9600 speed if I recall correctly. Thats for 2001/2 and similar dual line VFDs, but i expect single line follow same idea. I had posted sample data captures using usual UART-USB dongle somewhere in my site years ago

[afx]

Hello everybody,
I just got a broken Keithley 2700 for free and I'm trying to fix it. I'm not a professional so i hoped to be something simple.
According to the PCB it's made in year 2000.

Symptoms:
In DCV, 2WR and 4WR it shows overflow "OV.RFLW" in all ranges.
In ACV, DCI, ACI it seems to work (shows something near zero, good enough for now)

DCV in detail:
When I short the Inputs, it shows (only) in 10V-range a voltage of -4.6V. In any other Range it shows overflow.
When I let the inputs open, it shows overflow in all ranges plus following behavior:
- in 100V and 1000V ranges the inputs have about -4 to -5V (depending on the meter I'm using and its impedance)
- in the ranges 10V and below the voltage starts at 0V and goes up to about -12.6V. Even with a 10K resistor (on the inputs) the voltage is still about 1.8V, so it's not just a static charge)

What I've done so far:
- According to the troubleshooting in the service manual (found in other thread: https://www.element14.com/community/docs/DOC-52772/l/keithley-service-manual-for-model-2700-multimeter-data-acquisition-system):
  - checked Power Supply voltages - all ok
  - analog signal switching states for DCV (Q101 Q102 Q114 Q136 Q109 K101 Q113 Q105 Q104 Q108) - all ok
- checked for blown caps and measured some of them on board - all look ok
- checked 8 pole switch - ok
- Relays look and sound ok
- all components look ok
- checked U115 LM339 Quad comparator - outputs seem to have the correct state - ok

Known issues:
- it looks like there was some moisture at the back of the board but it does not look very bad and I think that any problem in that area may not affect the analog part of the meter (hmm...)
- the battery is dead, so it sometimes displays "No Comm Link" but here the same, I don't think it's the main problem.

I will attach a thermal image of an area that cought my attention.
There are two little transistors Q111 and Q144 (both PNP MMBT3906L) getting about 80°C
May be unusual, I don't know.

Would be happy to get some suggestions! Thanks!

[TiN]

Replace all electrolytic capacitors, clean or replace front panel switch, check or replace large SMD resistors chain that pass input signal, due to big size they are easy to crack and have loose contact.

[3roomlab]

I will attach a thermal image of an area that cought my attention.
There are two little transistors Q111 and Q144 (both PNP MMBT3906L) getting about 80°C
May be unusual, I don't know.

Would be happy to get some suggestions! Thanks!

K2000 schematic is similar to K2700/K2015, the component numbering is different
https://doc.xdevs.com/doc/Keithley/2000/K2000.pdf

i suspect the 2 Q are the powering section to a LTC1050, iirc it is responsible for the "zero" reference. in my old K2015 which i fiddled with, the Qs around that schematic section are also hot, and prone to damage (iirc, if you tap the SOT23, the readout could move around). so this was the section that i repaired, i replaced the hot SOT23 with TO-92 (iirc it became 40C or something?). *edit i found out the SOT23 i had was operating at about 70C.

if i understand K2700, the 2 opamps leading to the final zero buffer output from LTC1050 are powered from 2 different rails, the guard opamp (pre-LTC1050 AD822) is powered from 18v? 15v, and the LTC1050 (*update U113?)is powered from 5v dropped thru the group of SOT23 "up there" supplied by 20v, quite a distance away.  iirc the K2000 is diff from the K2700 in the 20v rail, the K2000 20v is unregulated (so it could be 21v? 22v? straight from the xfmr)

but a hot SOT23 may not mean it is broken, mine was working till i fiddled it to death haha. you could try to open the switch to clean the insides. or in my case, i removed the switch completely as i no longer wish to deal with the intermittent problem anymore.

the K2000 schematic was reversed by some guy in bbshot, it is not 100% accurate, i tried to do some small updates to the "K2015" schematic, it could be found here around post #19, it could be K2700 have further newer changes https://www.eevblog.com/forum/repair/repaired-keithley-2000/msg814292/#msg814292
i think i do have a 99% working K2700 lying around somewhere ...

[rob040]

And as usual, all my data and photos are always available on related project page here.

Hi TiN,

Unfortunately a lot of the data is no longer available at your site. Will it be possible to restore them, in order to keep this thread complete?

BR, Rob.