Fluke 8505A - persistent offset condition on shorted inputs

[Rax]

More empirical data below:

• Warming this up takes as much as bdunham7 says, apparently. After taking the DC SC module out and cleaning its board header contacts (IPA) - suspecting that replacing the MB with all the contacts involved may have to do with the issues - I watched it starting at about 8.5uV on zero (stone cold) to inch closer to the expected zero at about 1.8-1.4uV by the 3rd hour of warmup.
• Some more time later, it settles at around 2.2-2.4uV, so no, this isn't sorted out

I think I need to start swapping modules to see where trouble lies.

What my data (these "three zeroes" pursuits) seems to indicate is that, while the ADC and AF modules seem to reasonably maintain their zero points (though, notably, their check points are at 10V, so an "easier" bar), the DC SC seems to have a harder time to stay near zero. So my main suspect currently is the DC SC module.

[Kleinstein]

The amplifiers in the DC input and low pass filter secion can have offset drift. The offset adjustment also effects the temperature drift. Ideally (moderately good transistor models) zero offset also comes with zero drift (e.g. the offset proportional to the absolute temperature). However details like mechanical stress can still lead to some residual drift (maby in the 0.2 µV/K range) even if the offset is zero. In addition there can be thermal EMF effects leading to a more or less stable offset.
So on warmup it is not surprising to see some 5 µV of offset drift, as these old meters consume quite some power and thus run relatively warm.
In therory one could consider doing the adjustment of the input stage for low drift and not for lowest offset.  However with gain at the front, there is no simple compensation with the offset in the filter stage.
Ideally one would adjust the front end offset in the range with highest gain (200 mV ?) and than adjust the filter part with the front end in the 10 V range).

I would not really expect to have dirty contacts to have much effect on the drift. It is more just dust or variations in the power supply and even the mains voltage (migher mains leads to more heat in the meter).

The filter module contains quite a lot of switching FETs and there are reports of some of these failing and cause extra leakage current. This can also cause some offset and possible drift.

[Rax]

So on warmup it is not surprising to see some 5 µV of offset drift, as these old meters consume quite some power and thus run relatively warm.

I am seeing so much time needed for full warmup and so much drift during that, that I almost wonder if this heat management issue should not be addressed by itself first to make this a more convenient in use instrument. Otherwise, this would need to be an instrument running 24/7 to be useful and then I'd like a way to turn off the screen to preserve it.

[Kleinstein]

The rather large and quite full case and no fan makes the warmup slow. A fan could make the warm-up faster (and a lower temperature), though it can also help collect dust.

For the fiter part one could try a test with some 5-10 K in parallel to R3 and this way make leakage of switching JFETs considerably less relevant.
A 0 ohm link may allow to much current flow during transients / power on. So the resistor is to be on the safe side with the current.
An alternative test would be with some 5-10 K at the amplifier input to ground.

[bdunham7]

I am seeing so much time needed for full warmup and so much drift during that, that I almost wonder if this heat management issue should not be addressed by itself first to make this a more convenient in use instrument. Otherwise, this would need to be an instrument running 24/7 to be useful and then I'd like a way to turn off the screen to preserve it.

Usually they're good enough that you can zero them during warmup and get an OK reading, you just need the full warmup for the last 1-2 digits (not counting the 7.5 mode) and the best stability.  They are probably 'in spec' well before they are perfect, or at least as good as they can be.  I think these were mainly intended to be powered on 24/7 and I think the red LED displays last nearly forever.  They run at a pretty low power level.  I thought about trying a fan, but it would require some real effort.  If you removed all the module cases and ran bare boards, you could use an internal fan or two to reduce temperature gradients--that might be a big help.

[Rax]

I'm still not sure what to conclude, but it seems now, after a few days running idle, being also off for a couple of days, and then being turned back on and let warm up for a few hours the offset seems to have stabilized very nearly dead on zero (all three). The DC SG zero is actually almost dead on .0000(-3)V, with a bit of fluctuation between the + and the - polarity of that (and when straying slightly from it, it'd occasionally go to up to -.0002(-3)V, which I have no issues with).

The other two "zeroes" are less than 5-7 counts off (.000005-.000007V).

If this stays this way, I think I can go ahead and readjust the ADC ladders and the rest of the adjustments. Not sure what I should be doing if after that I get an offset in the uV range again.

[Rax]

Maybe I just did something very stupid.

The unit was on the 1V range, and was measuring either 1V or 2V from my calibrator. Then, without thinking, I switched the calibrator to 10V output. Now I seem to see one order of magnitude higher divergencies from the expected value than before on all values I'm looking at. After the last adjustments, it was showing so close to the actual values I wasn't even sure it'd benefit from software calibration.

I read in the SM that during cal mode, some overvoltage protection mechanisms are disabled. I am not sure what the AVG mode was switched to, but when I first pressed it after this has happened, the first value shown was "Err off." So I assume error correction (overvoltage protection?) was on beforehand, and during the excessive voltage exposure?

Did I just break something?

[Rax]

Did I just break something?

Maybe I didn't break anything (though I'd still like to hear experience with the effects of breaching events against the warning at 7-8.3). I think I'm fighting some competing offsets from both the 8505A and the DP8200 used as the source for the reference voltages.

I just (re)adjusted both and the voltages seems to fall closer in line to before. I guess it's what happens where all instruments one's using are "work in progress..." (or most, as my 8502A does have brand spanking new stickers...).

[Kleinstein]

I don't know the details of the cal procedure, but the normal input should still be save up to some 50 V. So I don't hink it shold do some real damage - maybe stress some the the BJTs used as a Zener for clamping. Things could be different if the voltage is applied to some more internal nodes.

The calibrator offset should not be such a big issue, as a zero should be a true zero and not from the calibrator.
There is already a chances to get a few µV of offset from handling cables and stacking instruments.

[Rax]

I don't know the details of the cal procedure, but the normal input should still be save up to some 50 V. So I don't hink it shold do some real damage - maybe stress some the the BJTs used as a Zener for clamping. Things could be different if the voltage is applied to some more internal nodes.

The calibrator offset should not be such a big issue, as a zero should be a true zero and not from the calibrator.
There is already a chances to get a few µV of offset from handling cables and stacking instruments.

Thank you very much, Kleinstein, for providing some peace of mind... There's these points during a project where uncertainty (to borrow a term...) is so high, and thinking on it with the... anxious metrologist is extremely useful and appreciated. Thank you.

I am pretty certain (that again!...) there has been no damage to the instrument, but it's been a pretty weird 24 hrs of monitoring and assessing.

[Rax]

There is already a chances to get a few µV of offset from handling cables and stacking instruments.

I am starting to consider shielded cables and paying close attention to the GUARD wiring and other finicky setup aspects. I am looking at ppms, literally, so likely my "noise" and weird shifts and offsets need to be managed at that level.

[Rax]

Back to the original topic of this - reading through the the datasheet, I see that it mentions that "Zero Stability" - though not sure if it only has to do with software zero or all zeroes - is "less than 5uV," so that's not so off from what I've seen through this.