Repair : 94' Keithley 2002 8.5-digit DMM , good ol' cap leaks destuction derby..

[plesa]

My newer 2002 has the vents holes in cover and air deflector for airflow. Older 2002 does not have that, so i was under impression that only newer units got this "tweak". Which maybe still true, perhaps someone with unit made in other dates can confirm it?

I did not see DCV performance difference on 20V range either with vent hole or without though (tested within 0.6ppm uncertainty vs 3458A).

Rest of questions I'll reply later, waiting for boarding in LAX atm

My 2002 is made in roughly in 2008 (serial number starts 09). And it does not have any vent holes. I did not noticed from your pictures this difference when we compare units in past 

[TiN]

Well, then it's random . What's firmware your unit running?

[plesa]

Well, then it's random . What's firmware your unit running?

It has A09, but I upgraded it to A10.

[smoothVTer]

Past week or so I've been steadily working on replacing components on K2002 main board:   super-paranoid careful to not touch/contaminate the front or back of the PCB, especially areas around the guarded traces.   While replacing C247,250,253  I saw clear, sticky goop around that sponge which TiN also removed in another K2002 post & I just had to share this because I burst out laughing holding the soldering iron in my hand.

TiN :  Check out how Mr. Evil Spongebob Squarepants poops all over the nice guard traces on the rear of the K2002 mainboard:



Who decided to place Spongebob on top of these guarded traces?  I'm wondering about the engineer who decided its time to melt marshmallows inside an 8.5 digit DMM!  Its like the sponge half-melted into ectoplasm and spilled all over. This residue is tacky and sticky to the touch.   I'm thinking years of heat inside the plastic shield maybe caused the glue/polymers of the sponge to semi-liquify. 

Does anybody have an opinion on whether or not this sponge-from-hell is a necessity?   Can we do without it?   Should I clean with IPA?   T

[TiN]

That's why guard traces are always should be exposed metal.
You can clean it, your calibration already corrupted anyway by big enough impact
I did not clean that spot in my units, decided to live with it, hoping Keithley people know what they are doing.

[smoothVTer]

It is difficult to imagine this was some mistake by Keithley ... perhaps this was a designed-in feature, as somehow to defeat this guarding?

Hard to believe.  If I find more information I will post it here.

[Le_Bassiste]

Who decided to place Spongebob on top of these guarded traces?  I'm wondering about the engineer who decided its time to melt marshmallows inside an 8.5 digit DMM!  Its like the sponge half-melted into ectoplasm and spilled all over. This residue is tacky and sticky to the touch.   I'm thinking years of heat inside the plastic shield maybe caused the glue/polymers of the sponge to semi-liquify. 

Does anybody have an opinion on whether or not this sponge-from-hell is a necessity?   Can we do without it?   Should I clean with IPA?   T

 what you have there is a deliberately applied conformal coating area under the differential JFET pair. it isolates the thermal pad from the circuitry.  i concur with you that the whole contraption doesn't look too trustworthy, but it's probably put in there to improve thermal equalization of the two JFETs. if you look on top of pcb, you will find the small "silver" box in proximity of the differential pair. that's a burr brown DCDC converter, which gets really hot (at least on my specimen). i'm suspecting that the DCDC converter does really spoil the temperature profile on the pcb, so they decided in an afterthought to implement the thermal pad and the coating to improve thermal equalization. i'm just guessing here based on my findings, though. https://www.eevblog.com/forum/repair/repair-94'-keithley-2002-8-5-digit-dmm-good-ol'-cap-leaks-destuction-derby/msg958245/#msg958245

them voltnut folx may be able to better explain this makeshift "improvement". i also noticed, that without the coating (but with pad), the meter initially reads some wild xx mV in 200mV range, inputs shorted, and takes roughly an hour to stabilize to a repeatable reading of a couple of 100 nV (meter is not calibrated). maybe, i contaminated the meter while replacing caps, or the conformal coating simply needs to be there, or i should remove the pad entirely. i don't know. any help welcome.

[smoothVTer]

    Thanks Bassiste,

I failed to notice the jfet diff pair in that area ... good eyes man.   My 2002 does not appear to have that conformal coating on it ... it was just bare PCB / green LPI / gooey mess covering it all.   Just like the user in the post you referenced ...    But your explanation strikes me as the most plausible explanation I've read so far.   I'll sit on this idea for a while and try to think of a hack or mod to achieve the same effect but without the goopy, viscous gluey mess.   

Ideas welcome ...

[smoothVTer]

Success!

Checked all replaced components for orientation / type 3x.   Partially re-assembled K2002, pushed power button on. 

Meter springs to life.   Added bonus:  all self-tests pass.   Seems like no errors.  I let it run ~10 hours over night, re-ran self tests again and all self-tests pass.  Preliminary check done on 10V range:



The 34401A  in the photo was last calibrated 6/2014.   Ignore the K2700 above ... seems  out of cal ...

Well I like the fact that the K2002 I restored ( last cal:  2004! ) matches my most recently cal'd 34401A, at least to 6.5 digits.   Especially since I removed the thermal sponge and then cleaned with IPA the area left with all the goo from the sponge/adhesive.

Some observations:
-- U105 (5V reg) gets uncomfortably hot to the touch.  I can see now why the cap next to it tends to leak/explode.  I am going to put a low profile heat sink on U105, it is in the air channel anyway.
-- Testing 200mV range:   shorted HI to LO with bare copper wire, pushed into contact.   Tried all combinations of filtering, A/D sampling window, Line Sync etc    but  meter seems very noisy.   Readings jumping around between 100uV - 800uV.  I have the metal case cover off and I didn't put the ferrite beads back on the input wires yet ... not sure if that would make this much noise.    I can post a video soon of this behavior soon.




 

[plesa]

The 5V regulator is hot on all 2002, but when covered the airflow will decrease temperature little bit.
For noise you can check the DMM noise by ( EZGPIB) script.

[Kleinstein]

Cleaning of the input section may have an influence on input currents, not so much on the calibration / DC accuracy. The suitable test would be a 10 M resistor or a capacitor to measure the bias.