First time on Keithley 2000

[Nuno_pt]

Hi all,

Yesterday I receive my Keithley 2000, I buy it for 170€ shipped, the seller said no errors, so when it arrive I've plug the DMM and no errors on screen, good.

Cosmetic aspect:

- The bottom of the unit has some scratchs.
- It has no bench kit (so the bottom scratchs)
- The front face plate is all yellow
- The red Keithley has some scratchs
- The with 2000 Multimeter has some scratchs
- The display lens has some minor not deep scratchs
- The Local word is almost gone
- The rear cover of the Scan card is missing

Inside aspect:

- Very clean unit
- Very clean PCB
- No signal of soldering that I can see
- No rust

One thing I noted is there is no ATMEL chip inside, so this is what I've inside.

U165 - Orbit 6715A

The Electrolytic Capacitors are:
C104 - H9555 85º
C131 - H9644 105º
C146 - H9633 105º
C149 - H9644 105º
C156 - H9633 85º (The big 6800uF 16V)

The Main Board shows 2000-102-02F, guess this is the revision of the board.

Firmware Rev. A06 A02
Calibration date - 03/20/03
Calibration Date Due - 03/20/04
Count - 4

Pictures will come in shortly, no battery on the camera.

The repair log will be this

1 - Changing the 5 Electrolytic Capacitors for new ones (maybe Panasonic, Chemi-Con KMG series) ? Any suggestion here.
2 - Cleaning the front plate with hydrogen peroxide (10 Volumes)
3 - Remove the scratchs from the display lens
4 - Buy some SST39SF020A-70 FLASH ROMs to burn the new A19 firmware (the ones I can find end with 70-4I-NHE or 70-4C-NHE) what is good for this?

After this maybe start looking for a Cal lab here in Portugal to send the meter to Calibration and upgrade to A20 firmware.

Pictures link < https://1drv.ms/f/s!ACbdz8jXIGKi3k0 >

[TiN]

Sounds good unit for decent price.

Your meter made in 1996, so capacitor replacement is very needed step .
Chemicon KMGs are ok, I use them often, had no issues so far.
4I ROMs are industrial temperature grade, 4C - commercial. It does not matter in this case for you, either will work.

[Kjelt]

There is one chip that can become very hot so place a small heatsink on it when you are at it.
Unfortunately I do not remember the number/function but it had to do with the communications

[Nuno_pt]

Thanks TiN for the explaining about the ROM's.
 
Kjelt I'll let TiN see if he knows what chip it is, since he's the Keithley Guru, if it worth the trouble since I'm on it I'll place the heatsink, also on the two IC's between the three black caps, thinking about arrange some way of putting one heatsink for both of the IC's, both connect to an alluminium L and the heatsink on top of the L, if it fits the spare space when the cover is put in.

The first link is update with new photos, one of the display showing some burn on the right side.


I did some cleaning with one product that I use to take the yellow of the plastic headlights from the cars, try it and I can see some good results, put the cream with the finger across all the display, the scrub with a toothbrush, no silkscreen gone or with less color, the buttons was scrub with water and soap and the toothbrush also, some results here https://1drv.ms/f/s!Aibdz8jXIGKi3lb0OYrGbaD-lyGh .


Now to order some:
Chemi-Con KMG caps, all 105º, should I keep the same voltage of the caps or put some high voltage, the originals are 16V, 50V and 63V?

Some LCC32 ROM's http://pt.farnell.com/microchip/sst39sf020a-70-4i-nhe/memory-flash-2mbit-parl-32plcc/dp/1829978?selectedCategoryId=&exaMfpn=true&categoryId=&searchRef=SearchLookAhead&searchView=table&iscrfnonsku=false

One TL866CS version http://www.ebay.com/itm/High-Performance-USB-TL866CS-Universal-Programmer-with-9-Adapters-for-12000-ICs-/121405177087?hash=item1c444ffcff:g:3ygAAOSw-vlVi~W1


One Bench Kit support.

   

[TiN]

Newer versions of 2000 (and some other Keithley gear) are using custom ASIC instead of ALTERA CPLD. Functionally both are same.

The hottest chip should be LM399 reference (one with white round plastic cover on it, in center of the board), which is due it's internal 90° oven. It's a bad idea to put heatsink on it, it supposed to be running hot to remove impact from ambient temperature change to reference output.

LDOs are usually around 60-65c in closed case. Everything else is in 50ish region. I had some thermal images of 2000 here.

I like Nuno_pt's way of thinking. Installing larger heatsinks on LDOs might help, but just a bit (since no forced airflow in the unit). Don't combine them or short to each though, as tab's are not always ground, so connecting all tabs together will cause a short!

[Kjelt]

No it was not the reference chip, it was some communications/glue chip at the end of the board.
I had two of these for work, now at home and one failed on this chip, when replaced we noticed it became rather warm.
Since it was quite expensive ($70 orso incl shipping) we have put a heatsink on it as precaution.

Perhaps I will open it up someday and make a picture.

[Nuno_pt]

Hi TiN and all,

I've take another look at the 3 LDO's (U110, U124, U125), but on U110 & U125 been so close about 3mm between the two and with the 3 caps (2200uF & 1000uF) so close two of them above and one below, there is no good way to fit some heatsink on both of them, or individually, without some metal work.

With U124 being alone is different, I can take out the small heatsink that he has, and fit an larger heatsink, it's worth the trouble?

Seems that Keithley face the same problem that I'm facing regarding the heatsink's in U110 & U125.

I don't have a thermal cam so can't see if the large heatsink will make some large difference, like TiN said no airflow inside the meter.

Also after TiN suggestion that the tab of the LDO's are not always ground, I've check that with continuity test from another DMM, and none of the tabs of the 3 LDO's beep relative to the 2000 ground of the PCB, so this would require separate heatsinks for U110 & U125.

- For the ROM's I order the 4I version Farnell 1829978.

The caps are same values as the original ones (or I should have look at the next voltage for all?), but I look at some better ones, like more hours at 105º or at 125º, this are the ones I found with stock in EU, basically Chemi-Con doesn't have stock in EU or I would do like TiN and would go to the KMG series road, and the ones on Epay are from unknown provenience.

- For the 6800uF 16V the Farnell P/N are:
2472006 (Vishay 6000h 125º)
1744933 (Pana 5000h 105º)

- For the 2200uF 16V the Farnell P/N are:
8126313 (Rubycon 10000h 105º)
2346454 (Rubycon 10000h 105º)
1800641 (Panasonic 10000h 105º)
2479889 ( Panasonic 10000h 105º)
2079157 ( Panasonic 10000h 105º)

- For the 1000uF 50V the Farnell P/N are:
1855192 (Panasonic 5000h 105º)

- For the 100uF 63V the Farnell P/N are:
2079295 (Panasonic 6000h 105º)
1144638 (Rubycon 7000h 105º)

I've order the Vishay, and the rest all from Panasonic, was this the good choice?
This is to leave some ideas for the new ones repairing the K2000, what caps they could look for.

Let's see when the capacitors arrive, I'll change them, but I'm off on vacations in 1 week time, and only after when I get home will order the TL866CS (that usually takes 15 to 20 days to get here) and burn the new A19 firmware, meanwhile will take another look at the 3 LDO's and the heatsinks.

[Kjelt]

If I am not mistaken , using a newer firmware will invalidate the calibration correct? What is the gain of using a newer firmware?

[Nuno_pt]

Kjelt from what I've read in other post's about 2000 you can upgrade firmware till version A19, for the last version witch is A20you need a new calibration since the A20 firmware has one more step in the calibration processe, so if you upgrade the 2000 to A20 and don't perform a new calibration you lose all the data.

I'll let TiN explain what one gain in upgrading to new firmware.

[Kjelt]

Ah that is cool, I might do that also if I knew what I would gain with the latest firmware ? 
One of mine is A05 02 the other A06 02.

[Kjelt]

So to come back to that "hot" ic, I opened the Keithley and it is the one in the upper right corner marked U165 (see picture).
My roms are ST 27C2001 that is different from yours or do you want to replace them with flash roms on purpose?
My elco's still look ok, nothing wrong with them, you advise to replace them just in case?
thanks.

[Nuno_pt]

Kjelt, I want to replace my U156 (EVEN) and U157 (ODD), in PLCC32, that are old EEPROM's style, with the newer FLASH EEPROM's SST39SF020A-70-4I-NHE.

See here on TiN's website https://xdevs.com/fix/kei2000/#firmware.

See also here post #20 https://www.eevblog.com/forum/testgear/keithley-2000-rs-232-wtf/

See also here on the Keithley forum https://forum.tek.com/viewtopic.php?f=32&t=6600

About your Elco caps, see the date code, if they are like my Nichicon '95 and '96 year, the reference is YYWW, the first two are the year of the caps and the last two are the week of that year, so for my caps with 20 years I sure don't want that they leak on the board causing damage.

Look here to what the caps can do https://www.eevblog.com/forum/repair/repair-old-keithley-2000-teardown-and-fix/

My U165 is mark 2000-802A02 ORBIT 6715A - 9629 (Year 96, Week 29) - 9512G476G, instead of the old ALTERA like here on TiN 's board http://dev.xdevs.com/projects/kei2000/repository/entry/photos/k2000_assy.jpg

[Kjelt]

Thanks there is a lot to read 

[Nuno_pt]

No problem Kjelt, lot's of reading here on EEVblog and also on TiN xdevs website (The Keithley Guru) and also on the Tektronix forum.

FYI also:
U156 & U157 are both EEPROM's PLCC32 (this are the ones I will replace)
U151 & U152 are both the RAM's
U136 is an NVRAM

- Page 21 of the repair manual states that:

"Memory circuits
- ROMs U156 and U157 store the firmware code for instrument operation.
U157 stores the D0-D7 bits of each data word, and U156 stores the D8-D15 bits.

- RAMS U151 and U152 provide temporary operating storage.
U152 stores the D0-D7 bits of each data word, and U151 stores the D8-D15 bits.

- Semi-permanent storage facilities include NVRAM U136.
This IC stores such information as instrument setup and calibration constants.
Data transmission from this device is done in a serial fashion."

[Kjelt]

Ok might be worthwhile to retrieve and store the calibration info as well?

[Nuno_pt]

That is way I'll by new Flash EEPROM's, and burn them with A19 firmware that is on TiN's website.
I'll save the old ones if something go wrong I'll still have the old EEPROM's with A06, to put the DMM back to work.

I forgot about U165, I'll see if I can measure the temperature in some way, and if I can I'll post the running temperature after leaving the meter on for 1 or 2 hours.

[TiN]

That's good idea to save calibration constants from NVRAM U136 (which is simple I2C EEPROM, TL866 supports it, but you will need to carefully desolder it from the board, read and then put back).
This way you can always revert original ROMs/CAL back.

Replacing firmware ROMs with FLASH is just to make firmware upgrade easy and quick, without need of getting new OTP ROM or waiting with UV light to erase UVEPROM.
I'm not quiet familiar with different 2000's version, as I got 2000 just to be complete in Keithley DMM family. My main focus was/is on 2001/2002 meters which are different beasts with different hardware.

Replace capacitors, time wasted for discussion if doing it or not is not worth a risk. Spend 10$ and you good to go.

U165 is mark 2000-802A02 ORBIT 6715A

As I mentioned, functionally is equally same custom gate-array. Earlier meters use ALTERA CPLD, newer - smaller ASIC. This chip is digital part of multislope integrating ADC with switch control logic, SPI interface, triggering, mux control and edge detectors/counters. Heatsinking it not required, nor helping.

[Kjelt]

Thanks TiN and Nuno_pt, I agree replacing the elcos now is way better than cleaning up in a couple of years time 

I wish there was a way to calibrate this meter myself, that is, adjusting it a tiny bit against another calibrated meter.

About the firmware, I see in the schematics that there are many revisions, so the pcb's or layouts have changed a bit over time, is it certain that the latest A19 firmware will run perfectly on the older revision pcb's?
I mean if it are only software bugfixes this would be good, but if it is due to different components or layout we might have trouble with the older revisions/pcb's, is there any documentation available that states which firmwares are the latest for certain pcb versions?
Found the answer: https://www.eevblog.com/forum/repair/keithley-2000-firmware-compatibility/

[Nuno_pt]

Kjelt, yes it's better replace them now, then have a mess later on, I agree with TiN on this for the value of the order.

2 - 1829978 = 3.52€
2 - 2472006 = 4.16€
2 - 2346454 = 2.10€
2 - 1855192 = 3.92€
2 - 1144638 = 1.47€

Total order with 23%VAT is 18.66€, I've just bought 2 of each.

I'm afraid to become Keithley addicted like TiN :-)

TiN I'll take a closer look at the NVRAM U136, and see if I can desolder it.

TiN I'm also looking at the 2001, but much more expensive, and like you said different beast with very different hardware.

I should have my caps and the flash eeprom's in the beginning of next week, the caps will be replace then but the eprom's only I get back from vacations.

[TiN]

Kjelt
With 2000 it's probably worth to get unit fully functional, with latest firmware (A19 or A20), all cleaned, caps replaced and fully assembled, leave it running few days to make sure all is okayish and no self-test errors or any suspicious readings, then take it to local cal facility and ask them nicely to calibrate it. Should be doable for something in 100-150$ range or less, if you lucky. Making standards for your own calibration will be way more expensive (2-10 times than the meter at least).

Nuno_pt
Beware this road, it's dark and full of terrors. You will find addiction to seeing that last 9th digit wobble faster than you expect...
If you patient, it's possible to get bad broken 2001 in 150-200$ range, just keep watching for offerings.
And I'm pretty sure in 95% of broken old 2001 (almost each one with capacitor electrolyte damage) all problems are fixable with enough dedication.
Here is one for example: http://www.ebay.com/itm/Keithley-2001-7-5-digit-digital-multimeter-/152146004454?hash=item236c9bc9e6:g:~WwAAOSwmtJXaE3O with megavolt writing on top 
It's likely to go for sale in 400-500$ region though, as still 2d 20h till auction end. Also official calibration of these are in range of 500$.

[Nuno_pt]

TiN I'm looking at it also, shipping is not expensive since it's near, let's see for how much it goes.

I know it's a little narrow and very dark road, leading to 8.5 and praying for 9.5 or 10.5 digits.

Or I can stay on 6.5 and flipping some till I've raise founds to get an 7.5. :-)

[Kjelt]

Hi TiN, I was not suggesting building or buying my own standard.
I was just thinking on my work they have calibrated meters so if I bring my meter and see what the difference is for instance in the DC volts range, it would be nice if by changing the calibration values in the nvram I could tweak it so it is more accurate. But then I need to know what the exact locations and contents of the NVRAM are, I do not think such a table exists?

[TiN]

praying for 9.5 or 10.5 digits.

Unless you've got 50K$-200K$ for JJA voltage standard and unlimited liquid helium supply (15-30$USD/liter) , prayers will not help much 

Precision gear for hobbyist, is rarely an one time buy tool, it's the way of thinking. When you get habit to move slowly and smoothly in the room to minimize airflow around probes and meters to reduce thermal EMF noise  .

Kjelt
The problem is not finding calibrated meter, but transferring it's value into your undertest test meter. You need to have source stable enough to do such a transfer. And to get decent TUR at 6.5 digits, with all those ranges which K2000 wants for calibration - it's usually something big, heavy and expensive like Fluke/Datron calibrators or array of sources. Of course there are darker ways to get something suitable, but that will take years of usual hobby-time to test and validate. So sending 6.5digit meter out for 200$ calibration is cheapest option, unless you have access to calibrator.

Your idea is interesting, but it have caveats down the road.
A. You need to ensure that source does not drift away between your reference measurement and adjustment. It's near impossible with just one calibrated meter to be certain.
B. Making table is easy task, most of calibration values are stored in standard floating point format with fixed offsets. You will need just to figure out checksum algorithm and make a map. It will take you a day with hex-editor even without experience. I did this for 2001 while ago.

[saturnin]

it would be nice if by changing the calibration values in the nvram I could tweak it so it is more accurate. But then I need to know what the exact locations and contents of the NVRAM are, I do not think such a table exists?

Actually, you can use :cal:prot:data? GPIB command to get calibration data from your K2000 too (if I remember correctly, it does not work with RS-232 interface). I attached dumped cal data from my K2000 unit.

I also think it should be possible to change calibration constants via GPIB too. Why should it be possible? I found out K2000 firmware contains secret commands that were most likely used by Keithley's engineers during development as backdoors (see attached part of K2000 FW - A07 version). By comparing known commands from documentation with strings in FW, I found two groups of secret commands:

1/ :diag:keit: group

:diag:keit:warmboot
-> causes restart of the unit via GPIB

:diag:keit:ehraccess
-> the command is accepted by multimeter, but it freezes, does not respond then and must be restarted (is it some kind of enhanced read access to multimeter memory??)

:diag:keit:isn
-> not sure it is correct form of command, I have not tried it yet

2/ :cal:unlock: group
-> I suspect that commands from this group can be used to define cal constants via GPIB
-> their syntax should be something like:

:cal:unlock:dc:<const_name>:<value>
:cal:unlock:ac:<const_name>:<value>

I have not tried them yet too since my unit has good calibration and I don't want to spoil it if I don't know exactly what I am doing...

!WARNING! Commands mentioned above are not documented in the official documentation, may be inaccurate, misleading and you use them at your own risk 

[Kjelt]

2 - 1829978 = 3.52€
2 - 2472006 = 4.16€
2 - 2346454 = 2.10€
2 - 1855192 = 3.92€
2 - 1144638 = 1.47€ 

Thanks for the list 
Where did you get these prices?
For instance the 1829978 costs me € 1,96 ex VAT a piece at Farnell.

[Nuno_pt]

The prices are from Farnell when I order them, today the 1829978 is showing 1.89€ + VAT each.

The prices that I put was for 2 units of each + VAT, the final price was 18.66€ with VAT  for the order

[Kjelt]

Ok thanks, our dutch Farnell is probably a bit more expensive 

[Kjelt]

Last question, I am looking through the list you provided but they are not all Panasonic like you ordered, some are Rubycon, you changed your mind or why?

[Nuno_pt]

Kjelt, I choose that because it was what the dealer from Farnell was having here on stock, so instead of waiting three days for the order to came directly from Farnell.es, I choose the Rubycon, very concept brand of Electrolytics.
It was only more convenient that way, and also the Rubycon that I choose have more hours at 105C then the Panasonic ones, so all for all I choose more hours at 105C or 125C, the rest are the same, Voltage and +-20%.

[dramoto]

Please....
When you have price in Portugal for calibration, share it hier.
I don't find a cheap option for calibration in Spain for hobby purpose.

After this maybe start looking for a Cal lab here in Portugal to send the meter to Calibration and upgrade to A20 firmware.

[Nuno_pt]

Today I got some free time and change the electrolytic caps, all work well, but some careful is needed when de-solder the age caps, like TiN already said in other Keithley treads, the place where the caps are solder are very thin, so they will peal of if you put to much heat on them, this are not polygons, just a very thin circle of copper, so be very careful here.

After clean all the flux, and then clean the PCB with IPA to remove all the residues.

Pictures will follow soon.

Now just replace the EEPROMS when I'm back from holydays, the TL866CS with the adaptors is already in order, it should arrive in the next 20 days, if it doesn't stop in customs, for the VAT.

Then will see the price of the Calibration price, if it's an acceptable price will upgrade to A20 and send it to the Cal Lab.

Will post here the price of the Calibration price and the webpage of the entity lab.

[Kjelt]

Thanks for the heads up for desoldering. No large groundplanes, then around 300^oC will be needed?

[Nuno_pt]

TiN on some other treads said around 250/260, I was ok with 300/320.

An ZD-985 de-solder station will be very useful to take the caps out, the problem is de-solder them with such small circle of copper.

For me was more trouble to de-solder them without the ZD-985, I was using an manual sucking de-solder, the solder part was easy.

Next follow up will be the EEPROMS.

And maybe next the U136 NVRAM, maybe here burn it with the code of some 2000 in calibration specs, the DMM would be in cal without send it to a Cal Lab.

Pictures link : https://1drv.ms/f/s!Aibdz8jXIGKi3nBBkZ_aGJdciGTp

[Kjelt]

I heard a lot of the 78xx are overheating on the Keithley, perhaps an idea to upgrade them to the new Maruta sm 78xx replacements. I do not let my meters run for more than an hour, i switch them on /off when needed just like a multimeter so no need for upgrading them but if you are a long time user that switches on in the morning, and off in the night it might be an interesting upgrade.

[Kjelt]

Today my caps arrived, those 1000uF/50V are incredibly tall 
Looking at your pictures they are the same. They are twice as high as the original ones, oh well if they fit in the casing and last twice as long I am happy , it is not a portable equipment after all.
Thanks for the list, will do the recapping this weekend.

[Kleinstein]

A linear regulator like the 78xx will dissipate essentially the same amount of heat, not matter what type. So not much to save there. It would be better cooling to reduce the temperature or a modified raw supply (e.g. inductor or resistor in series with rectifier if there is ample voltage) that could help to reduce power dissipation, by lowering the voltage before the regulator.

Such a DMM should be one for something like 1 hour before the readings are really stable. So i't no unusual to even have them running 24/7, when used in a critical application.

[Kjelt]

Today my caps arrived, those 1000uF/50V are incredibly tall 

It looks to me that they do not fit or barely fit, Nuno do you already have the cover replaced , does it fit? 

A linear regulator like the 78xx will dissipate essentially the same amount of heat, not matter what type. So not much to save there.

Those Murata's are 78xx pincompatible SMPS replacements, so a lot to save  But they are not in the 15V available, so no solution in this case.
http://power.murata.com/data/power/dms-78xxsr.pdf 

[Macbeth]

I can't see how those tall 40mm 1000uF caps can fit without squeezing them in and dragging the case on top of them every time its opened & closed. Not to mention just pressing on the cover will end up forcing them against the PCB. Also heat expansion? I wouldn't select any capacitor taller than the large 6800uF which barely leaves a gap and is 35mm high.

Farnell 1973521 fits the 12.5 dia 25mm high dimensions of the original but is only rated 2000 hours at 105C. I think I will take half the life vs much greater risk of mechanical damage

I also have a 2015 with a newer revision board. I notice in the Keithley 2015 Service Manual the same caps (C131, 148) are rated at 2200uF and 35V instead of 1000uF 50V. However the board has room for the wider diameter 16x25 vs the 12.5x25 in the 2000. Also they have uprated C146 from 2200uF to 3300uF in the parts list (though mine is fitted with 2200uF from the factory). There are also some small (6.3 x 11) unlisted electrolytics on this board, C260 and C261 both 47uF 50V. It seems another user, jpb, ran into this as well!.

[Kjelt]

I wouldn't select any capacitor taller than the large 6800uF which barely leaves a gap and is 35mm high.
Farnell 1973521 fits the 12.5 dia 25mm high dimensions of the original but is only rated 2000 hours at 105C. I think I will take half the life vs much greater risk of mechanical damage

I agree, I will use those instead, the originals were also that spec so they lasted 20 years, no reason why the new ones shouldn't either.

[Kleinstein]

There analog part is somewhat noise sensitive, so one should avoid switched mode regulators. If really needed one might improve cooling.

I would be careful about using larger capacitance, as this will increase the current load to the transformer and give a slightly higher voltage and thus more loss at the regulators.

[Nuno_pt]

Back from Holliday yesterday.

Kjelt yes all my caps fit inside with the cover put in place, there is a small gap, but they fit, meter is close and open to see if there is some scratch on the caps, but nothing.

The TL866 haven't arrive yet, will've to see if it stops in customs, last time something stop there took 2.5 months to be deliver to my home.

I got the price of calibration from an Lab here near me, they quote 390€ + 23%VAT.

Next step will be to burn the new A19 firmware on the new EEPROMS, when the TL866 get here.

[TiN]

Why not A20, if you decide to go calibration by lab?

[Nuno_pt]

Hi TiN, yes that is my first bet, if I can find some lab that make a better price, I see on some of your K2001 topic that Tek in Germany charge about 500€ with VAT, so this here seems on the higher side, will've to ask in other labs and even on UK, Germany, etc. to see what is the best price, if I find a cheaper one will burn the A20 and sent it for full calibration.

It has also cross my mind to send it to the US for calibration since there is a lot cheaper, some post here has values of $75 for 6.5 digits DMM, but how would the Cal hold with the trip to this side of the pond?

[electricar]

Hey folks,

thank you for the Farnell BOM. It was so much easier for me to get the right BOM for those who cannot buy at Farnell and made instead a BOM for Mouser. I searched only for stocked parts, so the caps on Farnell and Mouser are mostly different parts (other partnumber).
I replaced the old caps yesterday and everything is working well.

Mouser partnumber:

804-39SF020A7INHE
661-EKYA630E101MJC5S
667-EEU-HD1H102
661-EKZN160E222MK20S
661-EKYB250E682MMN3S

Regards

[Kjelt]

Today I finally got my device programmer working so I just modded my first of two Keithleys.
First I placed an extra black heatsink on U119 already on the before picture.
Then I removed the elco's.
I wanted to place also a heatsink on U126 but this would not fit , it would touch U119 and the elco in front of it.
So i made a custom small heatsink from 4mm aluminium, a 3mm tap in it, put a small thermal compound sticker on it to make the heat transfer better and screwed it with a 3mmx6mm conical screw.
Works for me.
Then I removed the I2C calibration eeprom U136. Read it out and copied it to a brand new chip, resoldered the new chip.
Replaced the OTP eprom's with firmware A05 with the SST 39F020A flash eproms with firmware A19 (thanks to TiN and his brilliant website )
Recapped , i put the two elco's that are close to the voltage regulator a bit in an angle so there is more distance between them.
Then came the final moment, power on.
Beep..... A19  YES!   Three errors  Oh no... couldn't write them down quickly enough but from memory one was 812.
Shock.... power off.. wait three seconds... power on.... A19... no errors  uhhhhh 
I do not get any errors any more, so is this normal after changing from A05 to A19 firmware you get this or whats up?
Anyway checked quickly with a 10V source if the calibration is still ok, and it is 
Any advice, or close her up and start using it again?

[Nuno_pt]

Good work there Kjelt.

I'm still in backlog with mine, been working on some projects more old, that I need to finish first.

Since you remove and read the U136 that contains the calibration data, I'm guessing that reading one with a fresh calibration, and burn the data on a new one and put it on another DMM, that DMM would be calibrate?

If the above would work, we can burn A20, and if we find one DMM with A20 and in CAL, it's just read and burn the new U136.

New firmware and new CAL.

[Kjelt]

I might miss something but the Cal data is unique per dmm since each one has a different "error" in its reading. So yes you could switch to a working A20 version but then you need to know exactly the layout of the cal file and what parameters have changed and checksum etc.
Else you would probably end up with a wrong calibrated device.
So i have the data of my cal device and can post it tomorrow if you are interested but it makes not much sense without knowing what offset holds which data or you should make the chip in a zif socket and experiment with changing some data and see what happens.

[saturnin]

I wanted to place also a heatsink on U126 but this would not fit , it would touch U119 and the elco in front of it.
So i made a custom small heatsink from 4mm aluminium, a 3mm tap in it, put a small thermal compound sticker on it to make the heat transfer better and screwed it with a 3mmx6mm conical screw.

Sorry to be blunt, but that heatsink is totally useless. I will not have any significant effect. Your unit is more than 20 years old and LM340T15 regulator has survived it without any troubles. Even without a heatsink. Keithley designed it well...

[Macbeth]

There is no need to desolder U136 EEPROM to read or write to it. I know with other Keithleys (2001, 2002) you can hold the RESET line to stop the MCU working and can play about with the EEPROM at will, but with the 2000/2015 that doesn't work. (Something to do with Tri State?). Thankfully all you have to do is remove the firmware PROMs U156/U157 from their PLCC sockets and switch on and the U136 is your oyster.

Using a typical SOIC8 clip on U136 with your TL866 is problematical because of the big electrolytic in the way, so just wire a couple of micro grabbers for the data pins and croc clip to chassis for ground. No need for Vcc as that is provided when you switch the K2000 on.

Now - firmware change from A19 or A20 does destroy your calibration. Or rather it doesn't, because if you go back to A19 you will find it is OK.

Some investigation is needed to find the memory spaces that have changed so a simple reflash of the EEPROM could be "good enough" for an A20 firmware upgrade without a full recal. Not having to desolder and resolder the EEPROM will go a long way towards that goal! 

[Kjelt]

Sorry to be blunt, but that heatsink is totally useless. I will not have any significant effect. Your unit is more than 20 years old and LM340T15 regulator has survived it without any troubles. Even without a heatsink. Keithley designed it well...

Yes for Keithley the unit should last 5 to 10 years and than it is written off and if the design withstands it, that is fine. From that POV it is build well.
If you look at the defect list for this type you see that most issues are bad elco's and bad regulators (could be that the first is the cause of the second, that I do not know).
If you then look at the layout of these components you see that the regulators are almost put against the elco's and there is no active airflow in the device.
I measured the elco's that came out and they still are ok ESR and capacitance wise, so no need to replace them as well ?
I do not agree. I think that if the unit is open and you can make it better and last another 10-15 years with an half hour extra work than I will do that.
The reason and use of the heatsink is to spread the heat a bit more and drop the temperature of the regulator a couple of degrees, giving the elco a little more headroom.

[Kjelt]

There is no need to desolder U136 EEPROM to read or write to it. I know with other Keithleys (2001, 2002) you can hold the RESET line to stop the MCU working and can play about with the EEPROM at will, but with the 2000/2015 that doesn't work. (Something to do with Tri State?). Thankfully all you have to do is remove the firmware PROMs U156/U157 from their PLCC sockets and switch on and the U136 is your oyster.

Thanks for this tip, for backing up the calibration data that is a very good workaround. 
For experimenting with the data in the eeprom that would not be so handy (for me that is).
You have to pull out the proms, flash the chip, put back in the proms and see what happened over and over again.
I rather put wires on the board than to remove those plcc's each time from their socket, I hate those plcc tools to remove them, I was glad I ordered 5 SST39F020 because I damaged one in  pulling it out the programmers socket (not zif obviously) since my expensive device programmer with multi plcc zif socket malfunctions:
https://www.eevblog.com/forum/reviews/anyone-experience-with-bad-contacts-on-universal-plcc-zif-socket/msg1014319/#msg1014319

What I actually would have loved is a small emulator with second individually program input, than you can fast and easy experiment with all the data.
Since the ic's are so cheap I think I might build a small breadboard with switches for I2C lines and Vcc. Oh well no time for now, will be on my todo list as future project 

Now - firmware change from A19 or A20 does destroy your calibration.

Honestly I do not know the advantages for upgrading any further. The only problem I had was psychologically that both 2000 units had a different firmware on power up A05 and A06, that bothered me.
If they are both A19 I am pretty happy as is, if someone can explain me why the A20 would be superior or have advantages I would like to know that. But as you said you have to recalibrate it and then figure out what changed and how to go back to the correct calibration.

Paying twice $200 for calibration is not for me, I rather spend it on new gear. Manually tweaking the calibration so it gets better (compared to a dmm on work or known chemical voltage cells) would be awesome, esp. if I get both units read exactly the same.

[jorgemef]

Hello.

I am on this process now. Just acquired a used Keithley 2000 with version A06 02. The caps are "good", but thinking of changing anyway.
Last calibration date is 96 but still within calibration limits against AD584LH which was measured with both University Agilent 34401 and Keithley 2000 recently calibrated. In my keithley 10.00127V measured on these two measures as 10.00140V.

Anyway the ask is: Where in europe we can buy the caps and eeprom? Mouser does not have all. TEM does have but not EEPROM. Digikey has it but bigger shipping cost with extra taxes. Farnell is not selling B2C in Portugal. Local retailers only have "strange" caps bands.

Other question, I remember seeing somewhere that intermediate versions until A19 damage calibration for AC or current. Can you confirm if is safe to upgrade? Anyway I took backup of the calibration EEPROM.

Just to add that TL866II requires VCC active even if not connected when reading 24LC16 on board without the PLCC EEPROM or else will read all FF. A long one to discover and see it mentioned nowhere.

Cheers,
Jorge

[alm]

Anyway the ask is: Where in europe we can buy the caps and eeprom? Mouser does not have all. TEM does have but not EEPROM. Digikey has it but bigger shipping cost with extra taxes. Farnell is not selling B2C in Portugal. Local retailers only have "strange" caps bands.

I find it hard to believe Mouser would not sell caps. Digikey has free shipping and duty paid to some European countries. I don't know about Portugal. https://www.reichelt.com/ is another fairly big European distributor. Instead of the EEPROMs, try more modern flash memory as described here.

Other question, I remember seeing somewhere that intermediate versions until A19 damage calibration for AC or current. Can you confirm if is safe to upgrade? Anyway I took backup of the calibration EEPROM.

Do you have a source for this? A changelog maybe? The only thing I'm aware of is that upgrading from version A19 or older to A20 requires a full calibration adjustment.

[jorgemef]

Anyway the ask is: Where in europe we can buy the caps and eeprom? Mouser does not have all. TEM does have but not EEPROM. Digikey has it but bigger shipping cost with extra taxes. Farnell is not selling B2C in Portugal. Local retailers only have "strange" caps bands.

I find it hard to believe Mouser would not sell caps. Digikey has free shipping and duty paid to some European countries. I don't know about Portugal. https://www.reichelt.com/ is another fairly big European distributor. Instead of the EEPROMs, try more modern flash memory as described here.

Mouser has the CAPS but some on order (27 weeks).
Do you know which newer EEPROM will work? Don't mind using different ones if they work. Else was looking for a pack of 10 for 10 euros from China.

Other question, I remember seeing somewhere that intermediate versions until A19 damage calibration for AC or current. Can you confirm if is safe to upgrade? Anyway I took backup of the calibration EEPROM.

Do you have a source for this? A changelog maybe? The only thing I'm aware of is that upgrading from version A19 or older to A20 requires a full calibration adjustment.

It was reply #13 here: https://www.eevblog.com/forum/repair/keithley-2000-firmware-update/ . Not sure was some other sporadic issue or if there is a problem.

[alm]

try more modern flash memory as described here.

Do you know which newer EEPROM will work? Don't mind using different ones if they work. Else was looking for a pack of 10 for 10 euros from China.

Try following the link that I posted, or the link you've posted yourself.

It was reply #13 here: https://www.eevblog.com/forum/repair/keithley-2000-firmware-update/ . Not sure was some other sporadic issue or if there is a problem.

If you read the thread further it gives a straight-forward workaround (upgrade to A15 before A19). And either way with a backup of the calibration memory you can always recover.