Keithley 2000 DMM Repair
I’ve been shopping around for a nice bench DMM for a while now, so I bought a busted Keithley 2000 off Ebay thinking that it would be an easy repair. I assumed that all the input protection in a modern DMM should fail first to protect anything important, and all I would need to do is replace it (maybe a few MOVs, a fusible resistor etc).
What’s interesting is all the input protection appears fine, but the meter is still damaged!
Keithley is kind enough to publish a repair manual, but it seems they have gone through great lengths to obfuscate the circuit as much as possible, while still offering a glimmer of hope to a third-party repair person. There are no schematics, but rough descriptions of the circuit operation. There is a parts list, but a few of the parts (in particular the JFET SNJ132199 – more later) are obfuscated behind in-house part numbers – even though they are standard parts.
When the meter arrived, I first checked the fuses (power fuse fine, amps fuse vaporized) and cracked it open to see if there was any visible damage (burnt components, vaporized traces etc.). The unit seemed pretty good inside, so I plugged it in, and with nothing connected to any of the terminals, none of the functions appeared to work. Actually, some of them were working, but because of the meter’s auto-ranging, they all appeared broken. Also, at this point, I didn’t try to connect anything to the meter, because it all appeared broken – if I had, I would have discovered that in fact, some of the functions were operational.
One nice thing, is the Keithley 2000 has a whole suite of built in test functions accessed by hitting the <TEST> button. I ran the built in tests, and I got the following error codes;
- 302.1 – 2W Sense
- 302.2 – 2W Sense
- 400.2 – Non Inv Path
- 402.2 – Non Inv /10
- 403.2 – Non Inv Bex2
Ok, that’s a lot fewer error codes than I thought there would be, but it’s a start. Let’s look at the manual which describes each of the tests and the components involved (I interpreted the manual incorrectly – but more later). I made a list of the components involved in each test so I can work through which ones might be bad.
The two 302 codes both involve the same circuitry, and from other tests, and probing around, I got the following list.
- R272 – ok, 301.1 passed
- Q109 – ok, 301.1 passed
- R117 9.9M – ok, 301.1 passed
- R115 | R324 | L109 – ok, could not really test L109
- R113 –
- R107 – ok
- R103 – ok
- R108 – ok
- K101 – ok, 301.1 passed
- K102 – dead, hole in side of package!?! charring inside
- Q105 –
- U113 – ok, 301.1 passed
Wait, what! There’s a hole in the side of K102, a small signal relay!
Wut…. So I desoldered the relay and tested it out on the bench. The manual reminds me over and over not to be confused by its operation – it is a latching relay.
In the off state, it is in the “set” position – at power off it should be “reset”. Clearly broken. So I gave it a pulse to the reset pin, hear a muted clicking, but the relay remains “set”. Busted. Time to take it apart.
I jumped the pads for the relay so it was “reset” powered on the meter and ran the self tests again. My reasoning was, even if there were new errors, if the old tests passed, the relay was the only culprit.
Re-ran the tests, failed test 401.2 and 402.2. Good. Now I only need to troubleshoot error 402.2 Non Inv /10. Time for bed.
So, I left the lab far too late last night and left the meter on overnight. It was in the 10VDC scale, and leveled out at around 3.079V. This is with the K102 relay removed, no jumpers installed.
WTF, the DCV tests passed. What is going on here? So, I jumped K102 to the “reset” position and applied 1V to the DMM input. The meter reads 1V no problem. Hmmm…
What if I mess around with the DCV scale? With the voltage source removed, in the 100V and 1000V scale, the meter reads 0V. The lower scales are all wonky.
Looking at the “Analog Signal Switching States” table in the repair manual, I narrowed down the devices in the signal chain which are different between the 10V and 100V ranges. I traced the 1V signal through the chain, and Q108 does not turn off completely. But it’s a mystery JFET.
Let’s have a look at the 2 wire Ω signal. Now, ohms ranges 1MΩ and below are fine, 10MΩ are totally off (see title photo). Tracing the signal through, Q109 does not turn all the way on.
Ok, no problem. I’ll replace the JFETs. According to the parts list, Q108 and Q109 are both listed as “TRANS, N CHANNEL JFET, SNJ132199(SOT-23)”. So Google only produces this thread. According to this post, Keithley also suggests
- Interfet SJN132280
- Fairchild Semiconductor 5149B51374A
- Fairchild Semiconductor 5149_NB51374B
Which all don’t exist. But according to that thread, the top mark was “132” while the top mark on my JFETs is “6G”. Google says that top mark “6G” on a N JFET is a MMBF4393LT1. Bingo. FETs and relays ordered.