Keithley 2000 TF-245

[Nogtail]

I managed to pick up 3 "broken" Keithley 2000's for $600 including shipping to Australia. So far two have arrived, and one of them repeatably passes self tests and spits out some believable looking numbers. The other is a bit more interesting: it fails self tests initially but after running for 10-20 minutes it passes everything. When it first starts up there's also a lot of noise on all ranges, so I assumed there was a problem close to the ADC.

I first checked the voltages and found +15 and -15 to be +15.9 and -16.0, so I assumed this was the problem. I checked the other meter and found the voltages there were also high, so I assume this must be common and not an issue.

I had a poke around with some freezer spray and a heat gun and found TF-245 pins 3 and 14 are open circuit when they're supposed to be 15K. After the chip warms up the resistance drops to around 10K which also isn't exactly great.

I've since found the TF-245 is notorious for failing, but doesn't seem to be easily available. I've contacted Keithley to see if they have the part available, but if I'm not able to get it from them are there any replacements? I could probably cram the resistors on a small PCB with castellated edges, but I doubt I'd be able to find the exact values the chip has.

[picburner]

About 6 years ago I had the same problem as you with the TF-245 (mine were perhaps in worse condition), I wrote to Tektronix/Keithley support who redirected me to their distributor here in Italy who sent me 3pcs TF-245 and other cosmetic parts for some Keithley instruments .
I hope that things have not changed in the meantime and you can have your spare parts too.

[Kleinstein]

The resistors to pin14 are 2 of the least critical ones. They are for the fine scale part, that is only used for a relatively small part. I would expect the required accuray at a round 0.5 to 1 % (mainly the ratio matters).  So the chances are good one could replace the 2 involved resistors externally and still keep the rest of TF247.

An error in the 2 resistors would lead to a little higher DNL error, especially for shorter intergration time (e.g. like 1 PLC or less).

Some 10 K is about what is expected for a in circuit measurement, as there would be other resistors / paths in parallel.

[Nogtail]

I haven't measured the other resistors so I'm not sure what else if anything is broken so I'd rather a replacement if possible.

I've tried contacting Keithley (Tektronix) to see if they had any spares. Their US office said they had the part in stock for around $20 each but they weren't able to arrange shipping to Australia. I tried their Singapore office but as far as they're concerned it isn't a saleable part. They put me on to the Australian distributor who seemed very knowledgeable and is looking into finding it for me.

If I can't manage to find a replacement I might try and lift the pin and replace the affected resistors.

Does anyone know what causes the chip to fail so often?

[Kleinstein]

A possible failure more would be damage to a bond wire, so the link from the chip to the lead frame. In this case only one pin may be effected and an external  replacement should be OK. Temperature sensitivity would support the idea of a bond wire problem.
Another possible failure would be a corrosion damage to the resistor film. In this case multiple resistors may be effected and the damage could get worse over time, but should not be very temperature sensitive.

High heat form soldering may accelerate failure - so I would be careful unsodering the resistor array, unless absolutely needed. "Lifting" the pin may be more like cutting the pin.
Mechanical stress, or hot / cold cycles may also cause chip failure. In theory high current could also cause failure from electromigration. However pin 14 should not at all see a high current with a 1.2 M resistor.

Not sure why pin 14 is so prone to failure - AFAIK it is one of the corners, so maybe of the point with higher stress or a weak point during production (e.g. not fully hitting the bond pads, or to little pressure).

[Nogtail]

I finally managed to get my hands on a TF-245 and swapped it out. I recapped and replaced the voltage regulators too, although the +15 and -15 rails still read quite high. The meter seems to work fine on most ranges and passes self-test, although some of the higher resistance ranges read noticeably low - I assume because it needs a calibration.

[Kleinstein]

After change of the TF245 the meter would need a recalibraion with adjustment. There is little need for before data, as the current state is not accurate anyway.

AFAIK the +-15 V are high by disign, not by accident.