[fixed] HP 3456a, front end problems

[FrodeM]

I am in the process of repairing a very early HP 3456a meter, and I'm having one last issue before I can call this more or less completely fixed. (well OK, maybe I need to clean the front panel switches and lights as well.)

It gives a reading, but it gets an offset that after some time settles at around 7uV. Needles to say, it also fails self-test 4 due to this. I did troubleshoot a little, and using a working 3456a at the local hackerspace I can verify that the working unit is able to measure this offset to the point at TP104. This is the output of the front end gate-bias, which means that the A/D itself is in good shape. The problem may therefore be the gate-bias itself, or I rather think it is something broken in the auto-zero circuit or other front-end switching/relays. If I manually enable auto-zero after a cold start, the reading can jump quite a bit erratically every few seconds before stabilizing at the 7uV reading.

First a summary of the two things I already have fixed:

• The old MOSTEK mask ROMs were having bit-rot. this was fixed with writing the latest version of the firmware to a set of new EPROMs, using three MCM68766 to have things look nice.
• A broken LM311 in the overload-detect

The problem is that I am having some problems deducing exactly which part of the front-end is to blame. I guess it might be something which is marginally working, which makes it hard to debug any further. Not fully having a clear picture on how the auto-zero is supposed to work (I'm more into digital than analog) does not help either.

So far, I got a suggestion to probe around with freeze-spray, but is there any basic measurements I can do before resorting to that?

[Kleinstein]

The auto zero mode works by switching between the JFET Q102 Q103 and Q104:
1) Q104 for a zero reading
2) Q102 for a short precharge phase, so that the amplifier input gets to about the right votlage
3) Q103 for the actual input reading.
4) maybe another precharge phase for less effect of switching off Q103

So the problem could be from the normal reading or from the zero reading (e.g. Q104 , maybe R150 ).
The relays for the front / rear switching could also cause problems - so a few cycels or using a short at the back may be worth a try.

An alernate input path could be used with the 4 Wire ohm mode in the 100 or 1000 ohm range (would use Q105 amnd Q115 instead of Q104 and Q103, otherwise similar to AZ mode voltage).
Another point to check could be the input bias current. As 7 µV could be the resut of some 70 pA of current through about 108 K of resistance for input protection. A 10 M resistor at the input would give a reasonable idea of the current.

[FrodeM]

Thanks for the response!

I did check the transistors now, and would you look at that. Pin 7 of U108 is giving a solid -18V to the gate of Q103. Pin 3 and 6 is getting the expected squarewave in, as well as the other gates have expected waveforms, so I think I have a chip to replace.

*Edit*
Nevermind about the replace chip part. I checked the input from P21, and it was between 2.2V and 3V. A wipe of the connectors later and everything's fine. Looks almost like a previous owner has had some oily lubricant or something on the fed-through connectors at some time, Sheesh!

[lowimpedance]

 Just to note though I have repaired a couple of these that had bad comparators that drive the FETS, which would give quite weird symptoms. 
Good to see it was a simple enough fix, these are nice DVM's enjoy.