Keithley 2000 repair/bodge

[manicdoc]

Hi,

Appreciate some help, I have a Keithley 2000 that is showing 400.2 401.2, and 402.2 errors. At rest, the volts count up to 1.1 and then cycle back to zero and continues. Short the inputs and it stops... Similar small biasing on AC.

I've got the service manual and I'll start exploring, looks like possibly a stuck relay or failed test path prior - happy to trace.

What caught my eye is what's in the pictures, a bodge complete with slightly melted components nearby...   Just wondering if this is 'normal' or should I consider it as suspect and therefore might be contributing to the errors?

Just want to line up my ducks before I dive in.  Thanks

[Kleinstein]

The slightly melted capacitors don't look good, but chances are this is only cosmetic damage.
The bodge wire looks like it relaces a supply connection (assuming the right chip is als a DG.... switch chip)  - so more like a fix for a lifted pad.
The soldering at U106 looks really ugly.

This looks like a repair attempt, that did not turn out that well.

Much of the AC part look like it should be repairable and relatively easy to follow the signal path.

[Electro Detective]

Looks like some hot rod wielder thought those pinkies were marshmallows   

Who knows what else they did to try fix it and created more issues along the way..

Maybe they got close and gave up (realizing they are not marshmallows after a taste test)

I wish you luck and hope it's an obvious easy fix likely with some reverse bodgeworx 

[Kleinstein]

The fist suspects are solder bridges around the reworked chips - though this would be nearly too easy.
Following the AC signal should still be easy.  A relative slow counting up could be a missing DC biasing path for an AC coupled amplifier.

[manicdoc]

The fist suspects are solder bridges around the reworked chips - though this would be nearly too easy.
Following the AC signal should still be easy.  A relative slow counting up could be a missing DC biasing path for an AC coupled amplifier.

I was planning on taking the board out and treating the bodge, put some cap tape around the nearby components and carefully use some hot air on low speed with a bit of flux. I also suspect some bridging going on.

I'll check it out first with a microscope so I know what I'm letting myself in for.

Yep, does look like someone went in with a branding iron...

[manicdoc]

Quick update, took the board out and gave the bodge wire the lightest touch and it disconnected itself, quality work, a free range bodge!

Then gave a quick heat across U103 and U105 (DG211) to seat them properly. Then checked continuity. the 5v pin on U103 wasn't going anywhere...  Looks like the pad might have ripped under that as well, scrapped back a little bit of track, and soldered it direct, good contact.

I'll check all pins on these two, then put a jumper wire back for the V- pins between U103 and U105. I'll check if anything near also needs that as it might have been run through the missing pad, same with the 5v... The switches all show open and no shorts to the switch pins.

[manicdoc]

Quick Update.

Finished checking the switching chips, and all good. Put a proper bodge in and reassembled.

Well, the failed 400.2 401.2 and 402.2 tests have gone. But the 0.0v to 1.1v cycling on DCV is still there.

I'll do some probing around and see what gives.

[macboy]

Slow climb from 0 to 1.1 V is actually normal. The lowest voltage ranges have high input impedance (>> Gohm) so if the input is left floating, the tiny bias and/or leakage currents in the input circuitry slowly charge the tiny input capacitance. This causes the reading to climb. When it reaches the end of the range (1.1 V on your case) the meter auto-ranges up to the next (10 V) range. This range has normal 10 Mohm input impedance which is plenty low enough to discharge the tiny input capacitance to zero. Then the meter auto-ranges down again, and the cycle repeats.

[Kleinstein]

With the Keithley 2000 the 10 V range can also be high impedance. Usually there should be no big difference in the input current between the 1 V and 10 V range.
In the Keithley meter the input is isolated from the amplifier that changes the gain by an extra buffer stage. So I don't see an effect of changing from the 1 V to 10 V range. This would be more like a unintended solder bridge to that more happens than just the gain change.

So open input drifting to some 10 or 11 V until the meter changes to the 100 V range would be normal, though some units may also find a voltage with near zero bias and than float around that point.

[manicdoc]

I tried going up one on the range, and it carried on counting up to around 11v, then an 'out of range' error came in.

The 0v to 1.1v cycling is 2 to 3 seconds.

Dirty board perhaps? Could well be a whisker...

Good to know I'm close.

There are signs of other other work around the board, the MUX chip certainly looks like it has been replaced. Again might be a dirty board. I was thinking of a ISA spray and hold the board upside down so it drips off and takes any residual with it.

[Kleinstein]

The open input case is not that relevant.
The fast dift suggest that there is some current to the input charging some 200-300 pF in 2-3 seconds to 1 V needs some 100 pA. This is more (about 2-3 x) than it should be.
It could be a dirty board, but also a leaky (damages) JFET or similar. What is strange is the voltage resetting. This is hard to explain by just a slightly dirty PCB. One could try forcing an external 1.1 - 1.2 V through a resistor for protection, to see if the "reset" is strong or more weak.

A point to check is the supplies. There are quite some reports of leaking failing capacitors in the Keithley meters. So these should be checked and if original planed for a replacemen

AFAIK the self test mainly checks parts behind the initial buffer amplfier and not very much the very input.
Another point would be checking a few more other ranges, like 100 mV, 100 V and the resistance ranges if they works to some degree.

[manicdoc]

The open input case is not that relevant.
The fast dift suggest that there is some current to the input charging some 200-300 pF in 2-3 seconds to 1 V needs some 100 pA. This is more (about 2-3 x) than it should be.
It could be a dirty board, but also a leaky (damages) JFET or similar. What is strange is the voltage resetting. This is hard to explain by just a slightly dirty PCB. One could try forcing an external 1.1 - 1.2 V through a resistor for protection, to see if the "reset" is strong or more weak.

A point to check is the supplies. There are quite some reports of leaking failing capacitors in the Keithley meters. So these should be checked and if original planed for a replacemen

AFAIK the self test mainly checks parts behind the initial buffer amplfier and not very much the very input.
Another point would be checking a few more other ranges, like 100 mV, 100 V and the resistance ranges if they works to some degree.

I've got an evening without meetings tonight. I'll check the power rails and ESR the caps first, then do what you suggest.  Thank you.

Also the 1.1v cycle suggests a diode junction forward voltage drop somewhere

[manicdoc]

Interesting.  The power rails are solid. I'll ESR check the CAPS next time I have the board out.

TP104 has the wiggle on it, then probed back and found U114, which looks to be a unity gain op-amp arrangement, pin 3 +IN fixed pretty much on -3mV + 0 to 0.1mV. Pin 1 OUT doing the cycle in the same range and time as seen previously.

Bad Op AMP? It should be holding it or is something on the PROTECT RING is playing up big time... I checked the FETs in the section prior - all I see is the wiggle via the unnamed net fluttering the Gates a little via the 2x 49.9k resistors.

No sign the op-amp has been touched previously.

I've added a screenshot of the circuit in question.

[Kleinstein]

The protect ring should be only the guard traces on the PCB. So ideally only a moderate capacitve load.

A bad OP-amp is possible. The amplifier is a little exposed, relatively close to the input.
One could check the supply to U113 - just in case there is something wrong there and this loads U114 too much.

[manicdoc]

The power to U113 was stable. Time for a Mouser order...

[manicdoc]

Quick Update:

- replaced one Elec Cap, was well out on the ESR - the others measured fine.
- replaced U114

cleaned around the board.

The cycling has slowed by about half. Did some further reading around on this unit and the front end is known to pick up 'strays' given very high impedance. So I'm taking this as a win for now, as it appears this can be a deep rabbit hole to fall into. Did some voltage measurements with it and seems on the dime. I'll compare against my old Fluke 8500 tonight.