Keithley 2700 repair part deux

[jp430bb]

After getting some great advice on this forum and replacing U115 (an LM339, no effect) and U209 (an LM393) with a new TI LM393B, which fixed the issue, back in November, my 2005-vintage Keithley 2700 started working again on DCV and 2-wire resistance.   

I should've tried out 4-wire resistance then, because this week, when I tried to use it, I found that 4-wire resistance only works on the 10Mohm and 100Mohm ranges.  Below that, even a 100-ohm resistor reads over-range.  I suspect that there is another failure in the signal-switching circuit that wasn't resolved before. 

There are three more LM393s, and I have more new LM393Bs.  I've tried to trace which comparator outputs drive the JEFT and bipolars involved in signal switching, but some of them I haven't been able to match up without access to a schematic.  Any advice?

Thanks! 

[jp430bb]

According to the service manual, JFET Q113 should be ON in the 4-wire resistance 1M range.  However, I'm seeing -15.6V or so at the gate, same as in the 10M range.  If I short Q113 source to drain, the 4W resistance 1M range works OK.   

The gate of Q113 is driven by U120, the LM339 I replaced, pin 14 a.k.a. 3OUT, via a 49.9K resistor.  That resistor explains why I couldn't find the connection with continuity checks, by the way. U120 3IN+ reads 0.084V in the 1M range and 0.071V in the 10M range, so it's not getting driven high in the 1M range like it should be.  Time to look at U121, an MC14094B.

[jp430bb]

Unfortunately, replacing U121 (M14094B) had no effect.  With the system powered off, I'm still seeing only about 2 ohms from pin 11 (Q8) to pin 8 (VSS).  Anyone have a suggestion for something else to check?

[Kleinstein]

With the 4094 replaced, it the LM339 getting the right signal ?
Is the OE signal at the 4094 correct (clould very well be just hard wired and no issue) ?
There is a chance a broken LM339 may pull down it's input - so it may be the 339 or 4094.

Singnal integrety of the clock signal is a possible cause of trouble, though the slow 14094 version should be not as sensitive as HC versions.

[jp430bb]

With the 4094 replaced, it the LM339 getting the right signal ?
Is the OE signal at the 4094 correct (clould very well be just hard wired and no issue) ?
There is a chance a broken LM339 may pull down it's input - so it may be the 339 or 4094.

Singnal integrety of the clock signal is a possible cause of trouble, though the slow 14094 version should be not as sensitive as HC versions.

After replacing the 4094, the U115 LM339 3IN+ (pin 9) is still not getting the right signal from the 4094 bit 8 (pin 11).  It should be 0V or 5V, but instead, it is consistently getting 84mV on the 1M range and 72mV on the 10M range.  The other 7 outputs from the 4094 are very close to 0V and 5V.  Switching between the 1M and 10M ranges, bits 1, 2, 4, 6, and 7 all transition nicely, either 0V to 5V or 5V to 0V. 

The 4094 OE signal is tied to VDD. 

The U115 LM339 is the one I replaced in my first repair on this 2700, although it was replacing U209 (an LM 393) that fixed DCV and 2-wire resistance then.  I'm hesitant to replace U115 again---if it is bad again, then it seems like something else is damaging it. 

[Kleinstein]

If U115 is bad, the other outputs may also show failure, though the defect can be limited to output 3.
If the LM339 is pulling the input low, this would mean the 4094 would see more current - though the 4094 has rather low maximum current. So it can survive a short and the extra heat may not be easy to detect.

It does happen that LM339 used to switch JFETs die - this is not so rare, though at least I don't know why - it is a bit of a mistery. The normal operation should be well in specs. The LM339 seems to be OK with a voltage at the output higher than the supply. The specs are not very specific there but it more implies that it should be OK.

A critical phase may be the turn on phase, e.g. via a unusual power up sequence.  At the output I would expect the switch jfet to be less robust than the LM339 output. A breakdown of the gate would likely also damage the JFET.  A damaged input is more like a problem from the input side.

I am afraid the LM339 is the prime candidate for the defect. It would be nice to know the reason for the defect (not just for this meter), but It would need quite some luck to find the reason.

[jp430bb]

To see where the weird 84mV on 1M range and 71mV on 10M range was coming from, I cut the 3IN+ pin of LM339 U115.  After confirming no continuity between the stub of the pin entering the chip and the PCB pad, I powered on the unit. 

At the PCB pad and at Bit 8 of the 4094B, I still see 84mV and 71mV in the two 4-wire resistance ranges.  With a scope, I looked at the clock, data, strobe, and Bit 8 of the 4049B.  The inputs all look sharp at least, but I couldn't confirm that they were correct.  The Bit 8 output was a steady 71mV or 84mV, not bouncing around.

In this state, the meter actually reads reasonable values on 1M and lower 4-wire resistance ranges, but I suspect it's not functioning correctly on the 10M and 100M ranges.

[Kleinstein]

This sounds like the signal is used somewere else too, e.g. at the the ohms current source ?

[jp430bb]

This sounds like the signal is used somewere else too, e.g. at the the ohms current source ?

With the system powered off, the resistance between the 4094B Bit 8 line and digital ground reads only 2.1 ohms.  This is a bit higher than my meter lead resistance, which is down around 0.2 ohms, so it's not a totally dead short.  Any ideas for finding out what else may be connected to that trace without removing the board from the meter? 

[Kleinstein]

The signal may also be used to enable the Ohms sense low - it kind of makes sense to use the 2 togehter.  The enable signal to the Ohms current soure would be anther part, and one that could be damaged from externally.  The K2700 seems to be a modern, low cost version of the K2000 in most areas.
AFAIK the signal is not used a 2nd time in the K2000.

A posibility would be to inject some AC current ( eg. 1-5 mA, 1 kHz, 0.4 V max. -> limit by diodes)  and than follow the current with a suitable probe (pic up coil and suitable amplifier).