HP3455A - How to select a cal constant for troubleshooting?

[intabits]

I have two HP3455A DVMs, one is good, the other is faulty.
It shows "OL" at startup, and the Self-test shows Cal Constant errors: 11,10,9,8,6,5,3,2,1,0

I made a "service cable" to do the "half-split" procedure with the good unit, and this indicates the inguard section is where the trouble lies.

All power supply voltages are good and steady with no ripple.
But the +10V reference voltage is bad at 7.8V (TP8). (with auto-cal off, when on, it jumps about quite a bit).
I powered up the reference module separately, and it produces 9.99972V, so it's good but something in circuit must be loading it down. I suspect shorted (or misdriven) FET(s).
It's not the ADC or AC converter boards - when removed, the reference voltage is still bad.
I believe the -10V reference inverter (U7) is good as TP7 is -7.8V, the negative of TP8.

To troubleshoot this, the easiest first step for me is to check the FET driver voltages shown in table 8-B-4, which may uncover a bad driver (LM339), or possibly a bad FET.
So I need to get the machine into a known cal test state to check these voltages, and the manual gives the following procedure to do this - But I'm having no luck with it.

B-B-11. Cal Constants Service Procedure.
8-B-12. When the 34S5A is in the Auto-Cal mode, the instrument measures one or more cal constants
between each sample. The number of cal constants measured depends on the sample rate. In order to
reach a certain cal constant measurement, use the following procedure.
a. Press the DCV and HOLD/MANUAL buttons and then the AUTO CAL button of the 3455A.
The instrument should now be stopped at a certain cal constant.       
b. Make sure the 34SSA is out of the Auto-Cal mode. Press the AUTO CAL button again, if
necessary (the light in the AUTO CAL button should be ofOc.
To locate the desired cal constant or to go through the cal constants completely, briefly press the
AUTO CAL button twice to turn Auto-Cal on and off. Each time Auto Cal is turned on and off, the
Auto-Cal circuitry will attempt to decrement through the cal constants from 13 to0, and the return to
constant 13.
NOTE
The AUTO CAL button should not be pressed on and off too fast or too
slow, because the 3455A may remain in the same cal constant or advance
past more than one cal constant. A few tries may be necessary to decrement
one cal constant step each time.

Questions
(from a.) What "certain" cal constant?

Should the display be showing the cal constant number (13..0) while doing this?
I just get the same "OL" display no matter what I do with those buttons (sequentially, simultaneously, etc). Is there a trick to this that I'm missing?

The only way I can get the cal number display is in self-test mode, and then of course, only the failing numbers. Pressing TEST or AUTO CAL moves to the next lower faulty constant. I can get a given cal number with auto-cal on or off only by cycling through the whole sequence, as each time through, the auto-cal state is reversed.   

Is the above method of using the self-test button instead, actually putting the unit into the indicated cal constant state, at least for the error constants that I'm getting? (so that I can measure the voltages in table 8-B-4)

I can see about 5 places where the +10V can be pulled about by a FET, but I don't want to just remove all of them if avoidable. (I'm hoping to find one that should be off, but with equal source and drain voltage, and just pull that one)

Any other experiences/suggestions/ideas for fixing this?

[Kleinstein]

A bad LM339 should still not pull down the 10 V via the FET, but just turn the FET off.  The more probably path is a FET that is essentially shorted and than instead of turning off the FET the LM339 would pull down on the 10 V.
One may be able to test the FETs (Gate to source / drain) with diode test in circuit. Chances are a bad fet would show up as a near short. The suspect FET switches are those that are off.

[intabits]

Thanks for the reply. Your response prompted me to do some study & research.

I've had almost no experience with JFETs, and most of what little I did know was wrong! I've used MOSFETs and just thought JFETs were similar. Because of my misapprehensions, the switching section of this unit seemed rather mysterious, so I didn't try to understand it in detail.  Now I know that they are depletion mode devices, and are normally conducting until the gate is pulled more negative than the source. (But S & D are often said to be interchangeable, so how does that work?)  You never stop learning!

Anyway, that explains the "Gate Bias" ("G.B.") supply from U1 at TP3 that is used to make the gates of Q1-4, Q13-16 the same voltage as the "multiplexing node" which is the sources of Q1-4 and the drains of Q13-16, making these FETs all on.  (Is the mixed use of S & D here because of the interchangeability?) There's also a "C.B." bias supply connected to some FETs, but I can't see where that's coming from (typo?).

Your comment about the LM339 output going low to turn the switch off, rather than on now makes sense to me (what prompted me to go study).
But I was still confused as table 8-B-4 shows the LM339 outputs should be 0V for the FETs that should be ON in a given cal state, and -24v otherwise. I was wondering where that -24V was coming from, as I had assumed the LM339 GNDs were at 0v, but I just checked and they are at -24V, so a "triggered" comparator is pulling the FET gate to -24v, and so turning it off.  As I delve further into this, things are beginning to become clearer to me...

For reference, and for anyone interested, these tables are on pages 140 & 141 of "HP-3455A-Digital-Voltmeter-1979-WW.pdf" found in many places, eg: http://manuals.chudov.com/HP/HP-3455A-Digital-Voltmeter-1979-WW.pdf
Schematic for the input switching section is on p217.

Next step I suppose is working through the troubleshooting flow charts, measuring voltages and hoping things become clearer.
I'd still like to know to step through the cal tests.

[Kleinstein]

For the JFET Drain and source are usually exchangable, the parameters like input capacitance can change a little if the pins are swapped, but usulally not much. For a N channel the more negative one is source and the more positive is drain.