UUT Accuracy as part of Uncertainty Budget

[itsbiodiversity]

Can someone please explain to me why I would use the UUT tolerances in the uncertainty budget? Specifically for a 5.5digit DMM.

I am currently including:

• fluke 1281 uncertainty
• resolution of uut
• repeatability

I'm not too concerned with smaller effects, but any guidance would be great.

I know I come across some who put tolerances in the budget and others that do not. My issue is this...

If the tolerance is say 1V, and 1V enters my equation divided by the sqroot3, then multiplied by k=2, am I not 100% of the time going to be above my tolerance limits for MU vs Tolerance of the UUT?

Thank you so much, as I have picked up a lot through this forum through the years.

 

[BradM]

I haven't seen the tolerance of the UUT included in the budget. Typically the budget you are estimating is to determine the error contributing to the calibration of the UUT. Thus, I would see the UUT error as independent of the standard uncertainty.

 

[dgriffith]

UUT tolerance specification is not normally part of the calibration uncertainty budget. The tolerance is the limit(s) against which the uncertainty of the calibration system is measured against to make a decision, such as pass/fail. etc.
The tolerance specification would be used when the meter becomes part of a calibration/test system used on other equipment.
Check with the mfr. of the meter--you may find that their spec is already at k=2 or 3, in which case you would not divide by sqr3.

 

[itsbiodiversity]

Do you include both the spec and uncertainty of the standard or just the uncertainty? I have seen it stated I believe that if you have the uncertainty of the standard then you do not include the tolerance of the standard in the budget.

 

[dgriffith]

Do you include both the spec and uncertainty of the standard or just the uncertainty? I have seen it stated I believe that if you have the uncertainty of the standard then you do not include the tolerance of the standard in the budget.

Confused. Are you now talking about after cal when you are using it, or still during calibration? If after, then do one or the other. Sometimes you don't know what the uncertainty of the calibration for your MTE was, so you use the specification for the measurement, e.g. ppm range, ppm reading, floor, if any.
Many meter manuals give you an example of the unc/error/tol for a given measurement. That result would be divided by the k factor as mentioned previously to get the standard uncertainty, which is then added to the other budget contributors in the usual way (GUM, et al).

If you know what the uncertainty for the calibration is/was, you can use it instead, but be careful. What kind of MTE are you talking about? If a DMM, that might be more difficult. If a resistor standard with a single assigned value, then use the uncertainty value stated on the cert. My opinion.

 

[itsbiodiversity]

It was a DMM. I believe I need both the accuracy of my standard and the uncertainty of the standard along with the resolution of the UUT/Std, and repeatability.

 

[dgriffith]

It was a DMM. I believe I need both the accuracy of my standard and the uncertainty of the standard along with the resolution of the UUT/Std, and repeatability.

You do not need both the accuracy and the uncertainty, nor do I think you can have both, of a standard when performing a calibration--much too conservative, leading to increased workload due to false reject. It's either a tolerance spec or an uncertainty.

The manufacturer will usually state a tolerance (acceptable performance limits (accuracy)) or an uncertainty (estimated error for a parameter) at some coverage factor.

Example: A 100psia pressure controller tolerance might be ?0.01% Full Scale at k=2. This becomes 0.0001*100=0.01 psi/k=?0.005psi standard uncertainty for every reading on the scale. As you can see, the lower the reading, the greater the potential error--0.005 is a much larger part of 1 than it is of 100.

Example: Fluke 8508A ref. dmm; uncertainties are listed for k=2 and k=3 for range and reading of each parameter; 10Vdc on 20V range might be [(50ppm rdg+35ppm rng)/k]=?[(0.0005+0.0007)/k]Vdc standard uncertainty for the specific parameter.

Only errors reduced to standard uncertainties can be summed together before expanding with coverage factor, whether performance specifications, resolution, repeatability, et al.