At what point would you adjust?

[mendip_discovery]

I had our lab meter sent out for calibration the other day and it came back all ok. Only issue is the spec of +/-0.012 has a measured +0.011. Considering it was -0.004 last year. So my fear that within the next 6 months there is a risk of the meter going out of spec.

As a lab I tend not to do adjustments as our calibrator isn't up to do it for anything over 5.5digits so never really thought about at what point I would adjust.

But what would others say would be a good way to set the limits before adjustment.
% of spec ie 75%?

The lab I used only adjust if its outside man 1yr spec so I would like to make it a special requirement that they adjust when it hits a certain limit.

[bdunham7]

Could you give specifics on which meter, what range and what device was used to calibrate it?  Obviously if the difference between last year and this time is mostly drift, your fears are justified.  Also, if you measure 110ppm and the limit is 120ppm, then you would need to have a TUR >10 to not adjust, otherwise you are in the grey zone.  I'm assuming your 120ppm is the all-factors annual limit, not something else.  As for when you adjust, part of the equation is the uncertainty of the calibration equipment.  For normal use equipment, if you have a TUR >10, you might as well always adjust unless you have some specific reason not to.

[mendip_discovery]

34401A, on the 1, 10 and 100 meg ohm ranges and it was a external lab doing to calibration for me and they have a good TUR so can do adjustments happily.

I don't normally do adjustment at my lab as the TUR is a low for doing a great job so it's not something I have looked at before.

[strawberry]

34401A  is within month spec (5.5 all 0) after 1.5y and repair of ohms range . waiting calibration
without frequent data logging it is hard to tell if that adjustment will do anything (continue drift or settle)
100Meg is not adjustable

[bdunham7]

34401A, on the 1, 10 and 100 meg ohm ranges and it was a external lab doing to calibration for me and they have a good TUR so can do adjustments happily.

The spec for the 1M range appears to be 110ppm total (100ppm reading + 10ppm range) @ 23C +/-5C.   So if you test it at a controlled temperature of 23C +/- 1C and you get 110ppm, IMO it is not in spec since the spec is 110ppm within a 5C deviation and  you haven't accounted for 4C worth of temperature deviation.  This would be a no-go for me, I would expect that meter and range to be adjusted to within 30ppm, the 24-hour 1C spec.  Then you might be confident that you would be within the larger 1-year 5C spec for a while afterwards.

The other issue is the guardbanding I was referring to.  That meter isn't really guaranteed to be in spec according to what you are telling us.

[alm]

Without any information about drift, and assuming this is not for a metrology lab where trending is more important than absolute accuracy, I'd say adjust if the as received value plus expanded uncertainty of the calibration is not within the limits of your calibration cycle with 95% confidence (aka if the value plus guard banding is outside the limits). If you are tracking drift, then adjusting if as received + expanded uncertainty + forecasted drift over calibration interval is outside the limits would make sense.

If the meter is drifting so fast, then you might want to consider shortening the calibration interval, until you can better estimate its drift, might be warranted.

See this article about guardbanding: A Pragmatic Method for Pass/Fail Conformance Reporting that Complies with ANSI/NCSL Z540.3, ISO/IEC 17025, and ILAC-G8

[mendip_discovery]

The lab I sent it to only adjust if its out of spec. My issue is that I am at 83%, 58% & 77% of the specs. I have noticed a trend in the measurements I do monthly (ish) in the lab that its drifting outside of the spec and this is causing my a concern I might be quoting a UoM that isn't true due to the amount of drift I have. So I need to work out at what point do I want the external lab to adjust my kit to bring it back to nominal.

Here are the yearly drift graphs and a table of the data from the yearly calibrations.






2020 was a bad year. after seeing that I added the 1, 10 and 100 meg ohms to my monthly checks.




As you can see this is getting me concerned.

I checked it against another unit I have,




I think I will ask the lab to just adjust back to nominal on these three ranges. But I would like to know what is causing the massive drifts. The meter may leave the lab every now and then but its treated well.

[bdunham7]

What do the error bars on the last half of the diagrams represent? 

I'm not sure you have enough data to characterize that result as 'drift', as opposed to random walk, a failure or a difference in calibration standards used.  What are the uncertainties of the calibration standards used in each of these tests?  What happened in 2020 and was it adjusted then?

[mendip_discovery]

  Sorry, the 1 M ohm ranges aren't as bad as I thought. I applied the wrong UoM to the graphs when I did them. 

What do the error bars on the last half of the diagrams represent?

The 34401 UoM. The limits are those of the Transmilles it is measuring.

What happened in 2020 and was it adjusted then?

It was adjusted. I wasn't aware of an issue as we don't do much in the way of >1M ohm measurements.

[strawberry]

how old is multimeter?

[mendip_discovery]

It's a Agilent so old enough. Other than the odd tweak here and there its never that far out, year on year. We have had it since 2015.

I want a 8.5digit meter but budget is limited at the moment.

[alm]

Looking at the data, I don't think a linear fit is useful. It's clearly not a constant, linear drift.

I don't think thinking about X% of the specs makes any sense. Rather look at the uncertainty at the time of calibration (see the paper I linked to earlier). The larger this uncertainty, the further you need to stay away from the limits to guarantee with 95% confidence that you are within the limits (the more guardbanding you need).

Looking at the last graph, assuming the red lines are the limits according to the specifications, it looks like the uncertainty of the measurement is way too high to tell if the meter is in it's 1y specs on 1 MOhm. This could either indicate excessive noise in the 34401A, a bad calibration procedure or a too low TUR. As is, it's pretty much equally likely that the unit was out of spec 100% of the time as that it was in spec.

[mendip_discovery]

Alm, the red lines are the spec for my own calibrator so they are meant to be rather good. The uncertainty of measurment is high as the spec for the 34401A is a big part of it.

What I am concerned about is that the 34401A tends to jump about. This isnt great and the closer it is to the nominal the better I feel. As soon as I go out of spec my budget is wrong and that can cause major headaches.

The lab that did the calibration has to have a decision rule and that was in place so as soon as the reading + UoM touches the spec they go into pass but possible fail so would adjust the meter.

[strawberry]

mine is similar age, I guess it was kept in sunny place (worn/dim VFD compared to my much older HP33120A)
after repair all reading jumped up (I guess, someone calibrated it and failed again and dumped it on ebay as a good instrument)
before  repair didnt stop drifting long after normal warmup time

[alm]

The lab that did the calibration has to have a decision rule and that was in place so as soon as the reading + UoM touches the spec they go into pass but possible fail so would adjust the meter.

That rule makes sense to guarantee (with 95% confidence) that the meter is in spec at the time of calibration. Normally drift is much smaller than the accuracy specification, so it should still be in spec at the end of the interval. In this case this may not be the case, and in that case asking for tighter limits makes sense to me. But if the drift is really that high, then I wonder if it is a defect in the 34401A/ I'm not familiar enough with the instrument to give specifics. Looking at a calibration report of 250 calibrations over a span of 8 years, which I found on the PMEL forum a long time ago, the 1 MOhm was never found out of tolerance in 250 calibrations that roughly had a 1y interval but some up to 2 years. For the 10 MOhm range 1 out of 250 calibrations found this point out of tolerance. So it does not seem like the average 34401A is particularly drifty on this range.

Sorry, the 1 M ohm ranges aren't as bad as I thought. I applied the wrong UoM to the graphs when I did them. 

The 1y accuracy of the 34401A on the 1 MOhm range measuring 1 MOhm is 0.011%, or 0.00011 MOhm, right? So it seems to me your original error bars were correct, and either the 34401A or your calibrator is drifting all over the place.

Do you also have as-received data from previous calibration certificates that allows you to estimate drift of the 34401A?

[Kleinstein]

A damage (e.g. from ESD) that effects mainly the higher ohms ranges is possible and not unseen for the 34401. There are a few reports around here for one of the OP-amps in the Ohms source going bad with an an increasing bias current. This gives an error to the higher ohms ranges. I don't know a good way to test the OP-amp in circuit.
Another source of drift in the high ohms ranges is dirt / contamination of the PCB to cause extra leakage.
At least the leakage from the normal voltage input can be tested, by looking at the meters own bias.

[strawberry]

less than ~1nA leakage could cause errors
sensitive components in ohms range Q202 ,Q211 ,U201-B

Another source of drift in the high ohms ranges is dirt / contamination of the PCB to cause extra leakage.
At least the leakage from the normal voltage input can be tested, by looking at the meters own bias.

it is not electrometer so fingerprints shouldnt do much
input amplifier should drift voltage ranges

[mendip_discovery]

alm,



Is the measurements from external labs. Only once it has been adjusted, in 2020.

I am thinking of some sort of thermal shock or vibration damage. It lives in the lab most of the time and gets taken onsite from time to time where it sits in the boot of a van on a rubber mat for the drive and then in the customers until the return.

[alm]

The huge jump of in all MOhm ranges in 2020 is suspect. It's by roughly the same amount in all three ranges (except 10x higher in 100 MOhm), and for the 1 MOhm range is more 23x its 1y specs, which is more than the average unit will drift in its lifetime. My guess is that some defect occurred back then, and that's causing the increased drift since. In cases of such a huge change I'd expect the cal lab to offer to investigate and repair before adjusting.

By noting which ranges it affects and does not affect, you might be able to narrow down the part of the circuit the fault is likely to be in.