Fluke 8505A - persistent offset condition on shorted inputs

[Rax]

I am working on an 8505A that's exhibiting some persistent offset voltage displayed on the condition of shorted inputs.

DC Signal conditioner adjustment of R53(ZERO) restores a zero reading on the display (it always does). Later (hours later, maybe a day, etc.) a persistent reading of a few uV on a shorted inputs condition would come back. This fluctuates pretty wildly, from 6-7uV to less than 4-500nV and sometimes "spontaneously" going back to a 0.0000(-3)V display reading. The average faulty condition seems to be around 2-2.6uV.

Not sure this is at all relevant, but the unit had completely riot behaviors at the beginning (random and truncated display readings, rendering it completely useless and making it constantly reboot) which was traced back to a faulty motherboard. Replacing that restored operation of all functions. Units behaves very well to calibration adjustments - everything is within range and easily adjustable to compliance.

I can't figure the cause of the offset issue though. C15 on the A/D module was removed to eliminate the possibility of its leakage causing this.

[Rax]

My current working hypotheses:

• Leaky FET switches in the DC Signal Conditioner. I have some 2N4391s, 2N4393s, and maybe some J111s that may (?) work if it gets there.
• Some contact or board/switch leak in the input module. I'll likely pull this out and see if anything needs some TLC. This is because when I switch between F/R (to clarify: the F has the short, the R does not), I am not getting quite the consistency of switching in the display I'd like to see (so the switch may not quite be working 100%)

Looking forward to other thoughts and input.

[bdunham7]

So you are seeing up to 60-70 counts and typically ~25 counts of variation up and down at random over an extended period?

Typically you would think that leakage/bias/tempco/etc would be found in the sensitive input circuitry found in the vicinity of Q18/Q19 and R52/R53, but remember that the 100mV range has 100X gain as well.  So if you leave it shorted and switch up to the 1V range, do you see the same absolute variations, which at 10X gain would be 6-7 counts max and 2-3 counts typical?  Also, have you checked the input bias current when you have adjusted R53?  Can you put your 1M + capacitor across it for some time and see if the bias current is relatively stable or not?

I've attached the schematic in case anyone wants to comment.  You can't easily replace a lot of the parts in the center because they are all factory matched for tempco.  As for leaky FETS, there's really only  Q14/Q15/Q16 to worry about and at zero volts and zero gate voltage on all except the active one, I think you would only see the issue in the range controlled by the leaker.  I could be wrong.  I suppose you could remove them and jumper the pads for Q14 source and drain to eliminate that as a cause.

[Rax]

So you are seeing up to 60-70 counts and typically ~25 counts of variation up and down at random over an extended period?

Typically you would think that leakage/bias/tempco/etc would be found in the sensitive input circuitry found in the vicinity of Q18/Q19 and R52/R53, but remember that the 100mV range has 100X gain as well.  So if you leave it shorted and switch up to the 1V range, do you see the same absolute variations, which at 10X gain would be 6-7 counts max and 2-3 counts typical?

If I switch to the 1V range (one up), I get a measurement fluctuating between 0.000006-7V. A little weird, because in the mV range I see something like .0033-.0035(-3), so not sure how that matches.

Also, have you checked the input bias current when you have adjusted R53?  Can you put your 1M + capacitor across it for some time and see if the bias current is relatively stable or not?

Good point, I'll double check. I have not readjusted the bias every time I've adjusted the zero (not quite every time, but most times). Namely, I have not adjusted it when it was still within the specified range, so there was no adjustment needed.

[bdunham7]

If I switch to the 1V range (one up), I get a measurement fluctuating between 0.000006-7V. A little weird, because in the mV range I see something like .0033-.0035(-3), so not sure how that matches.

The offset could vary a tiny bit between ranges or perhaps there is actually an issue there, but what I was thinking was leaving it in the 1V range for an extended period of time and seeing if it wanders around a proportional amount.  Also, the actual gains of the physical amplifier are X8 (1V range) and X64 (100mV range), that's why the max voltages for those ranges are not 200mV and 2V, but rather ~312mV (20/64) and 2.5V (20/8).  That gets sorted out in software obviously, so it shouldn't affect transfers between ranges, but there probably is some room for a little weirdness.

[Rax]

If I switch to the 1V range (one up), I get a measurement fluctuating between 0.000006-7V. A little weird, because in the mV range I see something like .0033-.0035(-3), so not sure how that matches.

The offset could vary a tiny bit between ranges or perhaps there is actually an issue there, but what I was thinking was leaving it in the 1V range for an extended period of time and seeing if it wanders around a proportional amount.

I will do that next, but for now, I removed the short and placed the RC across the input and so far I am not seeing it compliant (about twice the top of the acceptable range at about .0060-.0080(-3)), but it doesn't seem to fluctuate very much either. Of course, to compare the graph with the short readings, a much longer plot is needed.

[Rax]

leaving it in the 1V range for an extended period of time and seeing if it wanders around a proportional amount.

Not an extended period of time yet, but while the 100mV range shows an offset of about 2.2-2.6uV, the 1V seems to dwell at around .000005. It seems pretty proportional, given the datapoints from yesterday.

Going up the range, on the 10V I see .000036-.000038V of offset.

I think the above may rule out the FET switches, as the effect persists on all ranges.

I wonder if I should just go ahead and re-null the ZERO and adjust the BIAS, and then watch it for a while and see if it drifts off.

[m k]

Do you have other modes available?

Maybe checking what's happening in RT6 can give something out.
RT5 is another point but there you may need an external trigger, different part of the filter may be in the middle every now and then.

[Rax]

Do you have other modes available?

Not sure what you mean by that. Other than DCV? I have the ohms and RMS options installed, but not the current shunts.

Maybe checking what's happening in RT6 can give something out.
RT5 is another point but there you may need an external trigger, different part of the filter may be in the middle every now and then.

Assuming both are in the DC Conditioner, I was unable to find RT6 (just RT5). Even so, I'm not sure I can easily probe this without an extender.

On further observation, I am becoming increasingly weary of the inputs board and its switches. On a 2uV offset condition with shorted inputs, if I switch to rear from the front knob, the offset would go high (which I assume to be fine, as I have no short on the rear input), but if I go back to the front by switch, it would not necessarily go back to the 2uV indication. Not immediately, and not every time, and sometimes I feel it'd even improve the reading bringing it closer to 0.0000(-3)). So there seems to be some inconsistency of switching (mechanical contact issues?), or there's some grounding issue (hopefully not some MB PCB leak) related to the GUARD circuitry which triggers this behavior. Not confirming this is the case just yet, it may just rock the boat enough when looking at low level signals that it seems to be pertinent.

I just did (another) comprehensive clear to make sure no constants (correct or less so) are messing with the hardware zero point. These have been cleared recently, in case anyone's wondering.

[Rax]

I am currently doing another readjustment of DC Cond ZERO (R53) and BIAS (R52). After all, I haven't readjusted these after eliminating (temporarily, for troubleshooting purposes) C15 from the ADC module. In doing this, I am using a cardboard sheet to put on top of the unit between tweaks to emulate conditions which would occur when the case is on (warmed up regime).

In a prior run through this, it has not disturbed my ADC ladder/linearity adjustments - but rather helped. Which tells me the DC Conditioner is just playing with the zero level, shifting it around.

[1audio]

I would try jumpering the hi to low on the board to remove issues from connectors and switches. Also check the dcr from connector to the board to look for funky connections. You could be fighting some thermocouple that's not obvious.
Those are not easy to work on.

[m k]

All RTs are available through backplane but it's still not easy.
Maybe underside is finally the easiest way.
FR4 proto boards China is selling have edge connectors, but you must modify them with heavy hand, luckily only few contacts are needed.
Old PC ISA boards have better connectors, if you can find any, and you must modify them also.
I've made few KiCad protos, but their possible probe connection quality is totally unknown.

By selecting Ohms you can change some DC SC relay inputs to RT1, Ohms output, guarded bus 23.
I guess at least uV situation should be visible somehow.

By selecting RMS you can bypass the whole DC SC.
Then input is going to RMS board before DC SC relays.
DC SC output RT6, guarded bus 5, is also blocked and RMS output RT6 is activated.

Next in line is Active Filter, from RT6 to RT5, and then A/D.

You can also try shorting input and backplane.
Taking out Ohms and RMS modules is also a difference.

If nothing is not changing anything it must be pretty close to input connectors.

[Rax]

All RTs are available through backplane but it's still not easy.
Maybe underside is finally the easiest way.
FR4 proto boards China is selling have edge connectors, but you must modify them with heavy hand, luckily only few contacts are needed.
Old PC ISA boards have better connectors, if you can find any, and you must modify them also.
I've made few KiCad protos, but their possible probe connection quality is totally unknown.

By selecting Ohms you can change some DC SC relay inputs to RT1, Ohms output, guarded bus 23.
I guess at least uV situation should be visible somehow.

By selecting RMS you can bypass the whole DC SC.
Then input is going to RMS board before DC SC relays.
DC SC output RT6, guarded bus 5, is also blocked and RMS output RT6 is activated.

Next in line is Active Filter, from RT6 to RT5, and then A/D.

You can also try shorting input and backplane.
Taking out Ohms and RMS modules is also a difference.

If nothing is not changing anything it must be pretty close to input connectors.

Thank you, mk, very thoughtful input. Will explore what you're pointing out.

For now, as I said, I've went through the R53/52 adjustment one more time, very methodically, lid on top of the unit, long pauses between each adjustment. I left it running overnight and it seems to be holding its adjustment on the 100mV range, at least. BIAS was at the top of its allowed range, on average, factoring its variations.

Not sure how solid this .0000V +-.0002(-3) is supposed to be. In my case, it's very rare it'd stick within that range with its fluctuations (it's more like .0003-.0004V plus and minus fluctuations). And it'd drift up and down the range, within variations of just the last count (say, centered around .0005, some other times around .0002, and maybe then slide down to the nominal .0000V) - is this normal?...

Also, the bias has bigger fluctuations, may go well beyond .0030 on either side (to even three digits occasionally), though I've been watching for a center of the variations to be at zero.

These fluctuations seem to indicate to me there's something wrong with the DC Conditioner input circuitry. Maybe I should just be looking for a spare DC Conditioner and see if that makes it behave differently.

[m k]

I have no idea what is typical with these machines, even less how back in the day typicals differ from typicals around these days.
Guess is that more or less problematic relays must be a some sort of a latter typical, mechanical switches also.

How to check if RMS is or isn't showing the DCV symptom is also a bit low of ideas.
Maybe a longer time average of range/balance zero can show some notable differences, but is it a mark if there is none.

My meters have also been out of regular use for a longer time so no idea after how long they can be treated as any kind of typicals.

[bdunham7]

Not sure how solid this .0000V +-.0002(-3) is supposed to be. In my case, it's very rare it'd stick within that range with its fluctuations (it's more like .0003-.0004V plus and minus fluctuations). And it'd drift up and down the range, within variations of just the last count (say, centered around .0005, some other times around .0002, and maybe then slide down to the nominal .0000V) - is this normal?...

I pulled out my 8505A, which I consider to be more or less working, and let it run overnight.  This morning, with a short on the inputs, a towel over that area, AVERAGE turned on and the DC zero turned off, I get about 0.0005 (-3) with a few counts of fluctuation.  Turning on the PEAK function and letting that run for some time (maybe an hour?) I recorded high and low peaks of 0.0009 and 0.0002, so 4 counts either way. 

The calibration routine is not a spec, those are allowed to be 'aspirational'.  The spec for the DC100MV range states that the floor is 8 counts provided AVG is on and a DC zero has been done within one hour.  Mine seems to meet that so I leave it be.  Remember that these do not have any autozeroing or ACAL, so you are going to clearly see any drift in any point in the signal chain--it doesn't hide any of that.

[bdunham7]

Assuming both are in the DC Conditioner, I was unable to find RT6 (just RT5). Even so, I'm not sure I can easily probe this without an extender.

RT5 is the output of the Active Filter module.  So this brings up something that is probably a shortcoming in the procedures in the manual, which is that there are three zeroes (DC Signal, Active Filter and ADC) and it is possible to get them pretty out of whack without anything in the calibration process catching it--at least until you run a pot to the limit.  I have not had this issue, but it might be worth a check and is easy to do.

Open the 8505A, flip up the DC Signal Conditioner and Active Filter tops.  Turn the meter on, set to 100mVDC range and install an input short.  Take another DMM set for mVDC and as much filtering as possible and connect the negative lead to TP1 in the Active Filter.  Then with the postive lead check the voltages at TP3 of the Active Filter and then TP4 of the DC Signal Conditioner.  Both should be reasonably close to zero.  I have no idea what 'reasonably close' would be though!  If they are out by more than you like, adjust R53 for as close to zero at TP4 (DC Sig) as you can get, then adjust R14 for as close to zero at TP3 (Active Filter) as you can get.  These signals might be a bit noisy. 

Once you've done that, go back to and do 4-58 and 4-63 in the manual.  Hopefully redoing 4-63 won't alter all your ladder adjustments, but I'm actually not sure.  In any case, it shouldn't matter as you are just looking for a stable popcorn-free zero at this point.

[Rax]

I pulled out my 8505A, which I consider to be more or less working, and let it run overnight.  This morning, with a short on the inputs, a towel over that area, AVERAGE turned on and the DC zero turned off, I get about 0.0005 (-3) with a few counts of fluctuation.  Turning on the PEAK function and letting that run for some time (maybe an hour?) I recorded high and low peaks of 0.0009 and 0.0002, so 4 counts either way. 

The calibration routine is not a spec, those are allowed to be 'aspirational'.  The spec for the DC100MV range states that the floor is 8 counts provided AVG is on and a DC zero has been done within one hour.  Mine seems to meet that so I leave it be.  Remember that these do not have any autozeroing or ACAL, so you are going to clearly see any drift in any point in the signal chain--it doesn't hide any of that.

Thank you for providing that data and overall information. A whole (work)day later I came back to seeing the offset at less than 8 counts (most of time within 2-3 of zero, or on zero). Not to jinx it, but maybe this confirms I may be in good shape and maybe my worry is not confirmed. I'll have to monitor if this gets back into the uV redzone during shorted input, but I feel this controlled stage may be my most reliable data.

[Rax]

Assuming both are in the DC Conditioner, I was unable to find RT6 (just RT5). Even so, I'm not sure I can easily probe this without an extender.

RT5 is the output of the Active Filter module.  So this brings up something that is probably a shortcoming in the procedures in the manual, which is that there are three zeroes (DC Signal, Active Filter and ADC) and it is possible to get them pretty out of whack without anything in the calibration process catching it--at least until you run a pot to the limit.  I have not had this issue, but it might be worth a check and is easy to do.

Open the 8505A, flip up the DC Signal Conditioner and Active Filter tops.  Turn the meter on, set to 100mVDC range and install an input short.  Take another DMM set for mVDC and as much filtering as possible and connect the negative lead to TP1 in the Active Filter.  Then with the postive lead check the voltages at TP3 of the Active Filter and then TP4 of the DC Signal Conditioner.  Both should be reasonably close to zero.  I have no idea what 'reasonably close' would be though!  If they are out by more than you like, adjust R53 for as close to zero at TP4 (DC Sig) as you can get, then adjust R14 for as close to zero at TP3 (Active Filter) as you can get.  These signals might be a bit noisy. 

Once you've done that, go back to and do 4-58 and 4-63 in the manual.  Hopefully redoing 4-63 won't alter all your ladder adjustments, but I'm actually not sure.  In any case, it shouldn't matter as you are just looking for a stable popcorn-free zero at this point.

I redid the DC Cond ZERO and BIAS, then the ADC ZERO (which seems to be a prerequisite of the ACTIVE FILTER ZERO per p.4-22), then attempted the ACTIVE FILTER ZERO.  But once I tried that, I think I understand why they require the AF input be shorted by itself (p.4-22) - I think the TP3(OUT) level tracks, and is higher than the TP4(OUT) - differently put, unless the AF is zeroed independently of the stage before it, one's chasing a fluctuating point driven by the slightly fluctuating ZERO of the DC Cond (which is how the "main zero" point from the shorted UUT inputs is "seen" down the line to the AF).

In that spirit, I think I should be able to achieve the shorting of the AF input (the shorting of RT6 and RT2 on p.4-22, which I don't think I can easily achieve - where's this "Bus Interconnect Monitor?") by jumpering TP1/AF with TP4/DC SG?

It's true that TP4/OUT on the DC SG is quite a bit a mislabel, as that's not the output of the module (RT6 is), but unless I grab the upstream leg of R1 on the AF module and ground it, not sure how else I can short its input (or what better options I have).

[bdunham7]

TP3 (AF) is it's output to RT5 that goes to the ADC.  TP4 (DC SC) is separated only by Q8, which should be conducting, from RT6 which is the DC SC output to the AF.  The bus interconnect is a device that you install in place of the isolator for maintenance purposes.  I think setting TP4 (DC SC) to as close to zero as possible relative to TP1 (AF) is probably good enough to then set the AF zero, but if you want to short it as in the manual I don't see why a jumper from TP1 (AF) to TP4 (DC SC). wouldn't be good enough.  Or, you can peel the AF out of it's shell and solder a small wire to R1 and run it upwards so you have access to it.  Soldering in some handy test wires is how I deal with not having an extender card on these and the 5100-series calibrators. 

[Rax]

TP3 (AF) is it's output to RT5 that goes to the ADC.  TP4 (DC SC) is separated only by Q8, which should be conducting, from RT6 which is the DC SC output to the AF.  The bus interconnect is a device that you install in place of the isolator for maintenance purposes.  I think setting TP4 (DC SC) to as close to zero as possible relative to TP1 (AF) is probably good enough to then set the AF zero, but if you want to short it as in the manual I don't see why a jumper from TP1 (AF) to TP4 (DC SC). wouldn't be good enough.  Or, you can peel the AF out of it's shell and solder a small wire to R1 and run it upwards so you have access to it.  Soldering in some handy test wires is how I deal with not having an extender card on these and the 5100-series calibrators.

I was able to short the upstream leg of R1 to TP1 - one of those moments you congratulate yourself for getting those "witches hats" jumpers for a couple of bucks from the evil retailer - and the adjustment of R14 goes very well, except that the 0.000000V point on the screen (per p.4-22) and the lowest voltage on TP3 (I had to whip out my 3478A, which has the most sensitive DCuV range of anything I have around here) are not the same point... I see about 15uV on TP3 with zero on screen, while I can go below 1uV if I dial in the lowest point I can on TP3.

Shall I adjust per indication of the SM (which is based upon indications on the screen), or to the lowest DC out of the AF TP3/OUT?... [scratching head vigorously]
__________________________________________________________________________________________________
[Later] I lean on doing it per SM, in case the adjustment is intended to compensate for other offsets elsewhere and is setting a relative zero point. Or other adjustments during calibration shift the zero point intentionally.

[bdunham7]

You'll see that step 3 of that procedure has a statement that the ADC is required to be "working correctly".  It may not actually matter, but my idea was to set the AF output to zero as measured by your external meter and then set the ADC zero.  Then all the zeroes add up to zero vs. 3 random values trimmed to zero in total.

I did some stability comparisons for you.  Last night I grabbed a spare AF and DC SC, plugged them into my 8505A, did a very quick zero and let the meter warm up a while.  I set it for 10V, zeroed it and then let it run overnight with the PEAK and AVG enabled.  My heat goes down overnight so it saw a temperature cycle of 4-5C, and that resulted in the lowest value overnight of -10 counts and a high of 0 counts.  As the room warmed up in the morning it went up towards 0 counts again.  So the 24H stability with a 4-5C temperature change is 10 counts or 1ppm, the short-term stability appears to be 1 or 2 counts.  I'll do the same thing on the 100mV range tonight.  So far it looks like the 1H stability is 3-4 counts or so with no real continuous drift either way.

[Rax]

You'll see that step 3 of that procedure has a statement that the ADC is required to be "working correctly". 

Which is why I started this last go around with a fresh ZERO-ing of the ADC. Not sure that wholly qualifies as "working correctly" but WTW.

It may not actually matter, but my idea was to set the AF output to zero as measured by your external meter and then set the ADC zero.  Then all the zeroes add up to zero vs. 3 random values trimmed to zero in total.

I understand your approach with this. I'll think on this some more. Something tells me your approach may serve the linearity adjustments better, including all zeroes being in a more objective place. The inherent non-linearity of these units vs others may be part of it.

I did some stability comparisons for you.  Last night I grabbed a spare AF and DC SC, plugged them into my 8505A, did a very quick zero and let the meter warm up a while.  I set it for 10V, zeroed it and then let it run overnight with the PEAK and AVG enabled.  My heat goes down overnight so it saw a temperature cycle of 4-5C, and that resulted in the lowest value overnight of -10 counts and a high of 0 counts.  As the room warmed up in the morning it went up towards 0 counts again.  So the 24H stability with a 4-5C temperature change is 10 counts or 1ppm, the short-term stability appears to be 1 or 2 counts.  I'll do the same thing on the 100mV range tonight.  So far it looks like the 1H stability is 3-4 counts or so with no real continuous drift either way.

This is very useful as a comp, thank you. My unit seems to be at that point by now - though still collecting data (do I ever stop?...).

To give you an example, after I readjusted the AF ZERO, and the ADC zero prior to that - both open case - the zero point was 2.2-2.0uV off. After having the lid on for a couple of hours, my zero has gone back to apparently less than 3 counts from zero (stays between -.00001(-3) to .00002(-3)). Which is the best it's even been. The effort last night to adjust accounting for the "case on" thermal regime may be paying off. So apparently it almost qualifies for the adjustment +-2 counts criteria! I think we've both observed very long warmup times for these units vs. others from the series, so that seems to match.

Given yours has gone through a 4-5C cycle may throw off my point here, but maybe I have a unit that's a bit more sensitive to varying environmental conditions?

But maybe this is now back on a good track. I'm hopeful.

[bdunham7]

Given yours has gone through a 4-5C cycle may throw off my point here, but maybe I have a unit that's a bit more sensitive to varying environmental conditions?
I'm hopeful.

So last night's thermal cycle was in the 100mV range, shorted, etc.  I got a superficially similar result--I started at 0.0000mV on a fully warmed up meter, it appeared to drop to -0.0010 overnight and now that it is warming up, it's back near 0.0000.  The PEAK LO was 0.0012 and the PEAK HI was 0.0000.  Short-term stability looks like 3-4 counts.  I'm pretty sure it would be within the 8-count/hour limit even with the temperature cycling.  I say superficially similar because the ~10 counts look similar to the previous 10V test, but while the gain difference is 100X, the actual difference in the excursion is ~10X, ~1uV vs ~10uV.  It's not obvious that any single source can be identified as the cause--not that we need to since it is within specs.

I'm actually reconsidering how I did my original 8505A, which IIRC was not 'triple-zeroed' in the manner I suggest.  It works fine, but needs a loooooooong warmup, like 5 hours, to be golden.  I put my original AF and DC SC back in and set the newly characterized ones aside in case they are needed.  I'll try the same drill again tonight with these original parts for comparison.

[Rax]

During one more day or running, I think I'm probably within .0012(-3) and .0000(-3)V. Occasional jumps over .0012 may be explained by air movement in the room (I am covering this with a piece of cardboard, so far from enclosed), but overall seems pretty stable. Short terms stability seemed to be within 4-5 counts.

This morning I tried a "double-zero" procedure (based on screen readings):

• 10V range, UUT input short - ADC R8
• 10V range, UUT input short, short on AF input - AF R14

And they seem to chase each other a bit - meaning if I try for a perfect zero, I'll run out of pot range - which I find pretty odd. But the divergence between the two seemed to become smaller with every adjustment, so I essentially settled for under .000001 on both.

After that, I re-trimmed the "third zero" (100mV, UUT input short, DC SC R53) seems to stay within .0006(-3), though mostly closer to zero than that sounds like. I'll let it run for a while again and see if it drifts either side.

[Rax]

Not to jinx it, and not really thinking this will stay exactly so within the next 24hrs+, but after all these tweaks, and maybe 30 minutes for things to settle, I am pretty much seeing zero on all "three zeroes."

Will keep on observing and reporting. Hopefully this is a good adjustment point. I'd wager in saying is this doesn't satisfactorily stay within reasonable fluctuations (8 counts on the DC SC zero), I probably need to investigate each module and see where the issue may reside.