Keithley 2100 repair - JFET issue?

[mbrej]

Hi all,

I recently got a 2100 from ebay, which appeared to be working fine, although after playing with it for a while I noticed some of the readings seemed off. In particular, the 2w-res would read 10M, when nothing was connected to the input. I thought it may just be a very dodgy cal, however with another meter the input impedance read 10M even with >10G input impedance selected.

I opened it up and located the 100:1 10M divider used for the 100V & 1000V ranges (marked on the photo below). There are two JFETs (Q203/Q204, MMBF4393L) which when enabled connect the lower side of the potential divider to gnd. The gate of these JFETs (and all the others) are >= 0V when on, and -18V when off (driven from LM339s (open-collector comparator)). I then measured the gate voltage of Q203/4 when they were supposed to be off, and measured -3.5V rather than -18V. (The power rails were fine, the other JFETs had gate voltages of -18V.

I decided to short the gate of Q203/4 to -18V to see what would happen, whilst measuring the current on another meter, which read ~1mA or so. However, after this, the meter's failure mode changed slightly. Now, weirdly, the gate of Q203/4 would be driven to -18V when appropriate, however the -18V gate voltage appears to be leaking into the potential divider. DCV mode reads overload with no input, 10Gohm mode on the <=10V ranges, and another meter can measure -15V or so on the input terminals. It seems that these two parts need replacing (parts in the post...)

While playing with the meter some more, I noticed the DCI reading seemed wrong, more specifically the current value displayed was offset by an amount, and so would not read close to 0A with no input. I probed the meter, and traced the voltage produced by the current shunt resistor which is routed to the input selector. In reference to the attached image, the input signals A (RMS converter); B (DCV, low range); C (DCV high range via 100:1 resistor and relay); D (DCI) and E (??) are all switched using JFETs, before being input to the unity gain buffer X. The current signal D connects to the current shunts via a (presumably) reasonably large valued resistor. After probing this area, I could see that the voltage from the current shunts was being offset by a voltage from somewhere else, when measuring X. Seeing as this offset was negative I suspected it was coming from the dodgy Q203/204 via the relay (which I don't see why they used this as well as the JFET switches...). This was confirmed when I shorted point C to gnd, and the offset on the current measurement disappeared. This also leads me to suspect Q202 is faulty, since it is leaking current between D and S when it is supposed to be off (gate measured at -18V).

My next step will be to remove the Q202, and then confirm that everything else is working correctly, and see if anything else needs replacing. (Other odd behaviour includes the current offset changing when DC volts is applied to the DCV terminals - may be related to issues already discussed).

This seems an unusual failure and I cant really work out how it would have happened. If excessive voltage had been applied these don't seem like the parts I would expect would fail. Anyone have any ideas on this, or anyone seen anything similar in the past?

Thanks, Matt

[Mickle T.]

I have (almost) the same handwritten schematics 

[mbrej]

So I have finally taken a soldering iron to it. After removing Q202 (which connects the 100:1 resistor to the ADC buffer), the offset on DCI had gone, which was promising. I then removed Q203 (leaving the in parallel Q204), and observed the behaviour. How the -15V on the input terminals has also been removed - it seems only one of the parallel JFETs was dodgy.

I replaced Q202 with a MMBF4392 (not the correct MMBF4393 which was out of stock), and now the meter seems to be working correctly for all measurements. Once the correct MMBF4393 parts arrive I'll fit one of these for Q203, and perhaps replace Q202 with the correct part, although I cant see what difference it will make.

For the removed JFETs, out of circuit I applied -10V to the gate, and measured across D-S, which was indeed leaking. As a comparison, a new device did not leak when I did the same test.

Im still not sure why these two JFETs would fail, but if anyone else has this issue, hopefully this post may be of some use.

Nice Mickle T - yours is much neater than mine 

Matt

[Kleinstein]

The MMBF4393 and 4392 are different in on resistance and threshold voltage. There is still quite some scattering and even some overlap (some jets might qualify for both types). For Q202 there should be little effect of using the higher on resistance version. The two parallel ones might be more sensitive to the on resistance. They used 2 in parallel to lower the resistance already, so resistance seems to matter. The lower threshold by itself should not be a problem.

Defects of the input JFETs is not that rare with Keithley DMMs. I am not sure about the reason, but chances are that ESD could be an issue to cause defects.