Keithley 2700 DCV suddenly failed

[jp430bb]

After replacing the U115 LM339 with a new one, unfortunately there is no change in behavior for the Keithley 2700.

In my K2700, U120 is an LM393D dual comparator.  This differs from the unofficial K2000 schematic, where U120 is a quad.  There are more LM393Ds: U131, U209, U169, and U186.  Among U120, U131, and U209, maybe one is the failed comparator and is pulling down the 2.5 V divider.  I'm less suspicious of U169 and U186 based on where they are on the PCB.  U131 has 0 V GND and 5 V VCC, so it does not seem like a good candidate for pulling the 2.5 V signal to -0.54 V.  That leaves U120 and U209. 

In a post from several years ago by @Leuams, U209 had failed spectacularly in a K2700.  In my unit, U209 is driven with +15.7 V and -15.7 V.  It's not shown on the K2000 schematic, but since the schematic has U120 as a quad comparator, and the K2700 has only a dual, is it fair to guess that U209 is doing what the other half of U120 did in the K2000?

Is there any way to work out which LM393D is bad?  Or even if it's something else?  None of the comparators are hot to the touch.

[Kleinstein]

With the 2.5 V ref signal at -0.5 V, it is normal that the comparators would not react to the logic signal. So those wired with the 2.5 V at the neg input should all be high impedance at the output.  Getting a highly negative output would be suspicious.   

Something is drawing down the 2.5 V ref. level.  One could try to inject some additional current to the 2.5 V level (e.g. another switchable resistor of some 1 K towards the 5 V (or a 0/5 V square signal).  When switching this extra resistor on/off, one may see a different reaction on the positive side of the bad comparator. How much would be seen depends on how the bad one behaves.
Maybe the additional current could be enough to bring the level to at least +0.1 V or so, so that the other comparators could work and react to the controls. So one may be able to identify a few good ones.

[jp430bb]

Thanks for the suggestion, but I'm afraid I could fool myself trying it.  Since replacing the LM339D went pretty well, I think I'll order some replacement LM393s and swap out U209. 

TI has a drop-in replacement for the LM393 called the LM393B.  They claim the B has (quoting the datasheet): "lower offset voltage, higher supply voltage capability [up to 38V], lower supply current, lower input bias current, lower propagation delay, and improved 2 kV ESD performance and input ruggedness through dedicated ESD clamps."  Sounds good.

Next Mouser order, with some shift registers and JFETs just in case: LM393BIDR (5 ea), MC14094BDG (5 ea), MMBF4393LT1G (10 ea). 

[jp430bb]

I'm back with good news!  After replacing U209 (an LM393) with a new TI LM393B, all DCV ranges work great, input impedance is off-the-scale for my Fluke 77 on the 3 lower DCV ranges, and 2-wire Ohms is working again. 

Thanks for all the advice!   

[E-Design]

Great job! 

doktor pyta called it!

[doktor pyta]

I'm glad to hear that!

It would be very interesting to investigate the real cause of this issue with comparators.
Keep in mind that all input JFETs and the circuitry are OK.
I presume this might be related to some latch-up event because few times I saw little holes in the blown IC.