HP 3456A need some help for A/D repair

[velik_kazakov]

Hi guys and Merry Christmas to all of you!

Start before some days with restoring my new HP 3456A (I have one repaired and working). Get this one from Ebay UK in non working condition.
First exchange all caps in the power supply and all is OK with it.
It start with -3.0000 self test and I find a bad U13 is in inguard logic board. Good power supply to it, good clock but no any signatures on the board. Make a MCS-48 adapter and find my old Willem Programmer (lose some hours to find the programmer in the old cartoon boxes with old toys). Thanks to MadTux for this post (https://www.eevblog.com/forum/repair/hp-3456a-inguard-8048-up-firmware/msg1136747/#msg1136747) and firmware sharing. Flash 2 new 8748 for testing and install it. Both work fine and now all signatures are OK.
Now test go from -3 to -4 and I continue to find the problems. Following the service manual I come to the A/D analog part (digital looks good - chekced all Signatures to it and output of the digital part to analog A/D part). I replace U409 with new one (have some doubt for it). Putting both (working and not working) 3456a on my desk to self test #4 (test + 4 +store + 5) and grounding the A/D input (J29 to GND position) make some pictures of the TP402 oscilograme and post it here. The integrator going always up and I can't find the problem for that. Will accept all ideas before to start to exchange the opamps U405 and U406 or some fets. I do some checks and I think U405 is bad but ask here before to exchange it - there have some mV difference in the input.  Something else, can I exchange LF351 with some modern opamp for testing - I have no this one on hands.
In this test mod left one is bad and right is good one. It is with J35 to non Run position or it show only OL. The top one oscilogram is for bad unit (the output of the intergator) and the bottom one is for good unit. Just for reference I add the picture of the AD schematic.

Thank you for all ides in advance!

EDIT: Some update - I exchange U405 and U406 - it is the same. Same as before, so U405 and U406 I think is good. Will Exchange both later with what I have on hands to confirm for it.

[Kleinstein]

The sope traces don't look that bad. The integrator seem to work at least for the main part. There is the triangular waveform during run-up, just with an extra bias towards drifting  up. The extra bias / leakage would confuse the run-down. For the run-up it just looks like a slight negative input voltage.

So much of the integrator looks OK. The analog auto zero / Integrator reset also seem to work at least partially. So I don't think there is a problem in this part.

I would check the outputs of U402, especially the fine slopes. I stuck bit there for one of the fine slope could cause such a problem.

gate leakage in Q402-Q405 could be another possibility.

I modern replacement for an LF351 is possible, it depends were it is used: U405/U406 are probably OK with a slightly faster OP.

[velik_kazakov]

I check U402 before to write here. All looks good. Have stable zero/-12 when is closed/open, check too S-1 to U407b - good too.
Do some checks following SM 8-B-20 m. for checking A/D analog part but will continue tomorrow. Have no on hands 5v Power Supply for some checks .

[Kleinstein]

A possible error source could also be if the +12 V reference level is not exactly opposite of the -12 V.

Worst case there would be resistor drift or similar inside U500.

[velik_kazakov]

The reference voltage are -11.99V and 12.05V.

[Kleinstein]

From the scope picture the slope due to the unwanted background is rougthly 1/100 of the slopes during run-up. The asymmetry in the reference is about 0.5% and thus comparable size. 

However I think this may still not be the cause of the drift: looking at the zeroing phase of the integrator again, the offset compensation should also compensate for that asymmetry, as it uses the same state as in the drifting phase to establish the zero point. So it looks like the offset compensation phase (combined with integrator reset) does not work that well.
Normally this should charge C411 to a level that the integrator should show essentially no drift.  As the drift rate is still constant over time the voltage in C411 seems to stay constant, so no leakage there. It could still be something like to much offset from Q406a+b or U500 (e.g. +-12 V asymmetry) so that the voltage in C411 would need to be too high. If too negative a voltage is needed, Q408 may start conducting at the gate. Also the output of U406 is limited by the diodes to some +-0.6 V -  so about a maximum of some 20 mV of offset to compensate.
The voltage at TP403 during the zero phase should tell if this gets too large. I would expect this to be about in the -300 mV to +600 mV range. Outside this range could be a problem.
For a temporary quick fix, one could try a resistor of some 50-100 K in parallel to R414, so having less divider and thus a larger compensation range (though slightly more drift from Q407).

Too much offset of U406 maybe could also cause trouble, reducing the range for offset compensation.

[velik_kazakov]

Parallel resistor to R414 lowering the the ramp, but still out of limits. Do some checks now and I think Q406 is bad. Will try to find a replasmrent for it this week and will write back when test it.

[velik_kazakov]

Finaly i find the problem with A/D. It is a drift on the U500 resistor 98.19K (pin 15 to pin9). It is now 100.75K. All other seems good. With 6M resistor in parallel pin 9 come to 0mV and A/D start to work well (it work well from -20 to +20mV - the limits of the AZ). Will find a good resistor to solder here and to do an internal adjustment procedure to test the multimeter.
Kleinstein, thank you for the help!