I finally got around to working on this HP3455A DVM, which I got for cheap at a hamfest because the seller said it emitted smoke when he turned it on that same morning. I did a teardown video of it back in 2018, and at that time could not find any obvious source of smoke.
The fault was showing "0L" on startup, and the self-test produced errors on cal constant tests 11 to 0, except 7 and 4. The power supplies were all good, and no ripple was evident using a scope.
I made a "service cable" and used my good 3455 to do the "half-split" procedure, which connects the inguard of one unit to the outguard of the other. This identified the inguard section as being at fault.
Also, the 10V reference voltage was about 7V and not steady (jumping about I assume due to auto-cal switching). I removed the reference module and powered it up stand-alone, and it produced 10V. So something was loading it down.
The service manual has troubleshooting flowcharts specifically for each of the cal constant test error codes, but explicitly says not to use self-test mode to get into a particular cal constant test state for troubleshooting. It gives a procedure to select one of these states without using test mode, but I couldn't get that to work, on either of my 3455s.
So I posted a query in the Repair section of the forum asking how to do this cal state selection. I didn't get an answer on that, but Kleinstein posted a response that made me realize that my understanding of the JFET operation in the input switching section was all screwed up. After some study, I had a much better grasp on how the switching operated.
I tried using the flowchart for error state 11, but abandoned it because the voltages I measured were a bit off from what the flowcharts specified, which I blamed on the bad 10V reference voltage.
Trying another tack, and dreading a bad JFET, which my research showed would be hard and/or expensive to replace, I pinned my hopes on a bad comparator, which apparently is a common fault.
So I set about measuring voltages on the inputs and outputs of all the comparators (LM339) that control the JFETs. I found three were not pulling the JFET gates to -24V when their +input was 0V (logic 0 from the inguard controller) while their -input was at the 1.2V of the logic reference rail. All three were in A10U9. (The 4th element of U9 was used elsewhere and I didn't test it)
After replacing U9 (socketed), the unit came good!
A few crude tests showed it agreed closely with the good 3455 on DCV, ACV and Ohms. I was a happy chappy!