Hi Everyone,
Around 5 months ago, I purchased a pair of K196 DMMs on eBay, both were ostensibly "working" but sold as "for parts" so I knew I could be up for anything
One of the units was working perfectly fine, and agreed reasonably well with my Keithley 2015. The other didn't seem to be able to measure anything, and I believe someone had attempted to calibrate it and failed. I tried to do a quick calibration and also wasn't able to, so off comes the lid...
It has taken me the last 5 months to get the busted one going, but I really have probably put about 10 hours into it, just too much else happening.
The first thing I did was order and fit replacement electrolytics (some where bad, and I just replaced them all on both units), this helped bring the voltage rails back to spec on the good unit, but the other one still had the +/-15V rails to the analogue board were being pulled down a lot (can't quite remember how much), and they would recover if I unplugged the analogue board.
Ok, looking around the board, and my accidental brush of a finger on some MOSFETs told be where to start.. hot.. These MOSFETs were the BUZ71s and 2N7000s used to switch the current ranges through R28 (thick film network) /R29. The resistors happily where
ok. I wasn't able to track down (quickly) any BUZ71 (seem to be discontinued) so I substituted STP16NF06 MOSFETs as they had basically the same specs and I thought would be fine in this role.
The associated LM399 quad comparator was also also toast as was the nearby Resistor network R18, DG211 analogue switch (U23) and LF353 JFET Op Amp (U26). These were all diagnosed by heat and double checking voltages against the known working unit (very handy to have).
Fixing all this allowed the unit to "kind of" measure voltages.. But only to about 2 significant digits so more was wrong.. I put it away for a while to await more time.
Once I had some more time, I stared comparing the analogue signals between the good and bad units (with both units' inputs shorted). This lead me to eventually discover the "reference" 2.1V was being dragged down during one part of the measurement cycle and I discovered that U22 (another DG211 analogue switch) was also a bit warmer than its counter part.. This is were I was glad I bought several spares
After replacing it the unit worked a little better, but the last 3-4 digits were essentially noise..
More time passes, then I get a chance to come back and do some more digging around comparing the 2 units. I found that the integrator (U19) output on the bad unit looks like it is uncontrolled during one part of the measurement with a nice triangle up to about 10V and back to ground. This was not seen on the good unit. Some digging lead to to find that pings 14 and 15 of U16, a 4053 3 way analogue switch was being pulled down from a nice 5V square wave to about a 3V square wave..
After sourcing and replacing this, it didn't really make a difference.. bugger. Looked around for what else was on that line.. various resistor doing charge injection, all fine.. eventually realised U8, a 74HC74 dual D-Type flipflop was also driving that line. Ok, lets source one of those and replace it.. U16 was probably innocent...
JOY!! it lives!!!
So one thing I found was the solder mask on this unit was aweful... the good unit was almost pristine like we expect from a modern solder mask. They must have been using an older method. It was "powdery" or "chalky" and was melted/broken off when I tried to use my vacuum de-soldering gun. I resorted to clipping the parts out and removing the pins one at a time with a normal iron to avoid as much damage as I could. I also seems to hold onto some of the flux and makes it difficult to clean.. did the best I could. Watch out for this stuff in older gear I guess.
I have to do a complete calibration on it yet. But that will come shortly. I have enough gear to do most of the ranges, but probably not the higher voltage AC ones.. (I can make a good approximation using a variac + isolation transformer I guess). I'll be "transferring" the calibration from my Keithley 2015 since that is the best meter I currently have. Eventually I'll use some of the references from the Aussie Calibration club to check this and my EDC521 voltage/current standard.
As for what happened to the poor thing, I'm guessing high voltage event, probably on the current input? Who knows, but it certainly was a high silicon body count event..
So I didn't take many photos of this process, but here are some attached.
I really enjoyed the challenge of tracking this down, and learnt lots doing it.. good fun
Ash.