Okay, so last night I finally got around to troubleshooting this meter. I went to some trouble to mount cameras facing the oscilloscope, the test meters and pointing down onto the HP3478A because I wanted to record a video and share with others. Well...
First, a little side-note, there's an error in the service manual, but not a big one. At the beginning of "Service Group A, DC Volts and DC Current Troubleshooting", section 7-A-4, you find the following text:
b. Check and make sure the Reference Supplies are at the correct level and quiet. The +10V supply can be checked at U405 pin 6, the -10V supply at U404 pin 6, and the buffered +10V supply at JM201...
Maybe there's another way to interpret the text, but U405 pin 6 (U405 is an opamp with output at pin 6) should be at
-10V,
not +10V. As I was happily measuring voltages and writing down the results in my logbook, when I came to U405, I paused... Then I went and looked at the schematic and it's very clear that it should be at -10V, not withstanding that the supply rails for U405 are Ground and -15V! I downloaded the service manual directly from Agilent's website.
Okay,
that's not the interesting part. After I'd measured nine planned voltages with my meter (and numerous others I thought would be useful), and then probed each point with my oscilloscope to ensure there wasn't any noise; apart from something I saw on the scope which I'll add at the end of this post, I saw nothing unusual.
So, since I figured I'd have surely seen *something* amiss by this point, I decided to hook the 3478A to my bench supply at 12V through the back panel, and you know what, it !&@&$ worked and matched exactly the reading on my Fluke 27/FM (which is fairly recently calibrated). Disconnecting the power supply, I looked to see if the weird upward/downward voltage drift with no input was still happening; NOPE!
Stunned, I reconnected the bench supply to the front panel to see if maybe the front/back-panel switching mechanism was flakey. Nope, it !@%#$ worked there too! (to those who'd asked before, yes, I did test both the front and back panels when I originally tested the meter; see my first post)
To see if auto-ranging was working as well as the millivolt range, I switched back to the back panel (more convenient the way I had things on the bench), hooked up the bench supply at 12V and toggled off the output of the bench supply to let it's output caps slowly drain (I know, maybe not the best behaviour for this bench meter, but at almost no current drain it takes a while for the voltage to drop to zero after toggling off the output, and hell, it makes for a convenient test), the HP3478A and the Fluke 27 metering this in parallel.
They tracked each other perfectly all the way down to 0.1 millivolts (as low as the Fluke goes). So, it seems my meter "magically" repaired itself. Well, there's no such thing as magic, so here's my best explanation of what happened, please let me know if this sounds reasonable:
During shipping, some kind of shock or vibration caused an already dubious connection between some component and the PCB to become flakey. Because I was probing many points, and of course pressing the probe firmly either against or down on the pins, I probably restored a flakey connection somewhere, thus restoring the meter to working condition.
Does that make any sense? Can anyone think of an alternate explanation?
P.S. some additional notes/questions:
1) Apart from a faint 5-15 mV ripple at roughly 100Mhz which appeared everywhere, I saw nothing unusual, I'm not sure where this noise came from, I put all my wireless devices into airplane mode and shut down the lab computer, but I figured the noise was some oscillation either in the probe, the scope, or the combination of all driven by the environment. My best guess is that what I was seeing was the whole combo filtering environmental RF noise, as when I had my probe out of the semi-shielded enclosure of the meter, the scope showed high frequency broader-band noise (at 2mv/div and the intensity cranked up) which attenuated when I dipped the probe into the semi-shielded 3478A's enclosure. The 100Mhz signal (which looked like it had a third harmonic on top of it) would appear when probing any point within the meter. I'm still learning RF electronics and am still lacking a lot of experience, but does my thought that the whole setup was simply resonating at ~= 100Mhz driven by external RF noise make any sense at all? If not, what else might I have been seeing?
2) The voltage reference of the meter was actually 10.18V instead of 10.00V, but I believe this is fine as long as it's stable and the meter has good calibration data. Is that correct?
3) When I was setting up my camera to film the scope (Tektronix 475, 400Mhz), I noticed that channel 2 *just* started acting flakey, it was fine earlier in the week :-(, it seems to overshoot on the internal calibration signal and jump around in amplitude, i.e. flakey, this only occurs at more sensitive inputs, iirc 0.2V/div or less, at larger attenuations, it's perfectly fine :-(. Why does one piece of equipment have to break when I'm trying to fix another... I'll probably be posting for help trying to fix the scope now, which, for me is kind of intimidating, but I have been reading about scope front-end design, so hopefully I'll be up for the challenge.