Test Uncertainty Ratio TUR (UUT Tolerance / STD Uncertainty) Question

[jfgunn]

I thought this was a topic I understood well but hav begun to doubt myself when I look at some other calibration certificates.

Take the calibration of a 1" micrometer at 1" that reads by 0.001"

The uncertainty of this calibration using gage blocks is roughly 0.0006". 1/2 the resolution of the micromter is the dominant factor.

The accuracy of a 1" mictometer is 0.001". I would say that I do not meet a 4:1 TUR because 0.001"/0.0006" is 1.67. My TUR is 1.67.

When I read the definition from the Metrology Handbook it says that TUR = (UUT Tolerance / STD Uncertainty).

This might lead me to believe that the TUR would be 0.001"/0.000010" = 100 where 0.000010" is roughly the accuracy of a 1" gage block.

This question comes up when I read calibration certificates from accredited labs that says " uness otherwise noted, we meet a 4:1 TUR....".

By my definition, any calibration certificate with this wording would have to report the uncertainty on every certficate where the accuract of the unit under test is equal to 1 division. (ie most weighing scales, micrometers, etc..)

For certificatges I issue, I have used a more stingent definition of the term which would lead me to provide a good calibration certificate to my cusromer either way.

I would apprecaite any thoughs on the topic.

[Benjamin28]

TUR
"Uncertainty of the measurement standards shall not exceed 25% of the
acceptable tolerance." To elaborate, this statement is somewhat misleading, as the actual MU of the calibration process should not exceed 25% of the acceptable tolerance, not just the unc of the master standard. The idea is to take into consideration the major sources of uncertainty, including environmental and operator.

So you would need to know the accuracy of your instrument...so in the case of your micrometer lets say you have an accuracy of +/- .0001" in this case you would need a calibration process which has an uncertainty less than .000025". At this point you would need to know the calibration uncertainty of your calibration process and ensure it is less than the allowable .000025" if you want to meet the 4:1 TUR.
The two items that come into play are your instruments' accuracy, and the uncertainty of your calibration process.

Please discuss further if this doesn't sound correct to you, I don't deal with TUR on any significant basis so I may not be completely on par when it comes to it's practice....

http://www.transcat.com/PDF/TUR.pdf

Seems a handy description of the TUR idea. Easily confused with TAR...accuracy and uncertainty are two different things though.

[Hershal]

Generally, anytime there is a question of 4:1, you should ask your accredited cal lab for the calculations.....NOT their BMC.....BMC is after all often unrealistic.....

If in fact they can support the 4:1 then OK.....although there is a trap.....your default uncertainty now becomes 25% of the rated accuracy of your instrument for that measurement because you don't have any other number.....and often that will be higher than the actual calibration uncertainty.....a good thing actually as that is generally a more conservative number.....

However, if they can't support the 4:1 they should report the specific uncertainty.....

[Mike Linn]

This is a scenario I brought up on this forum many moons ago. 1999 to be exact.

In that thread I asked the question about calibrating a Mitutoyo 293 series digital mic. The specified tolerance is ± 0.00005 inches.. however the resolution is also 0.00005 inches.... assuming your using the beast blocks available, in a perfect 20 °C environment and perfect techniques, the BEST you can do is about ± 30 µin. or a 1.67:1 TUR. Using 4:1 TUR requirement no one in the world should be able to calibrated these devices.

Strangely enough no one was able to address this conundrum. ILAC G:8 states "When a specification describes an interval with an upper and lower limit, a statement of compliance or non-compliance should only be made where the ratio of the expanded uncertainty interval to the specified interval is reasonably small and fit for purpose (meaning that the laboratory should be able to meet the needs of the customer)." This says nothing about a 4:1 anything. (FYI - The 4:1 TUR is a hold over from the old Mil-Spec 45662A 4:1 Test Accuracy Ratio. Which was a poor mans way of ensuring the standard was accurate enough to provide some semblance of acceptable false accept/false reject percentages.). Really its up to the customer to decide if the "TUR" is acceptable or not.

In a perfect world we labs would not be making these pass/fail opinions. That would be left up to the user of the equipment. In fact if you send a standard to NIST.. they will not give you a pass or fail. They give you a measured value and its associated uncertainty. Its up to you to decide if it meets your needs. But 95.45 % of the calibration consumers don't really understand uncertainties and what accreditation requires a labs to do. The customer should be looking at the published laboratory scopes and determining if the lab in question has the bona fides to perform in a way acceptable to their process requirements.

[dv8shane]

In Reply to Parent Post by Mike Linn

This is a scenario I brought up on this forum many moons ago. 1999 to be exact.

In that thread I asked the question about calibrating a Mitutoyo 293 series digital mic. The specified tolerance is ± 0.00005 inches.. however the resolution is also 0.00005 inches.... assuming your using the beast blocks available, in a perfect 20 °C environment and perfect techniques, the BEST you can do is about ± 30 µin. or a 1.67:1 TUR. Using 4:1 TUR requirement no one in the world should be able to calibrated these devices.

Interesting thing about the Mitutoyo digital micrometers is they quote the accuracy less the quantization error. From reading several articles I found it could account for as much as 0.3 of the resolution. So the actual accuracy could be as high as 0.000065. I think this is why the Gidep general procedures call for a tolerance of 0.00010 for micrometers and calipers of the same resolution

As to calibrating when the TUR is less than 4 there are several methods of Guard Banding (reducing the specification) mathematically. This done to contain the probable spread of values within a 2 sigma confidence interval. I did a quick review and this article covers it decently. http://www.metrologyindia.org/ebooks1/UM_001.pdf

[Mike Linn]

Guard banding is fairly straight forward for metrologists but VERY few calibration consumers (in the US particularly) understand this and will have a conniption if you start changing their specifications. They've bought into the concept of "salesman's specs" that may or may not have any basis in reality.

The other issue of the manufacturers published specification and the resolution being the same is not solved by guard banding as you can resolve anything below the resolution. Therefore using the accepted practices, if the UUT deviation is ) then your can make the call for 'In Tolerance' , but anything else will be 'unable to determine'.

[dv8shane]

In Reply to Parent Post by Mike Linn

Guard banding is fairly straight forward for metrologists but VERY few calibration consumers (in the US particularly) understand this and will have a conniption if you start changing their specifications. They've bought into the concept of "salesman's specs" that may or may not have any basis in reality.

The other issue of the manufacturers published specification and the resolution being the same is not solved by guard banding as you can resolve anything below the resolution. Therefore using the accepted practices, if the UUT deviation is ) then your can make the call for 'In Tolerance' , but anything else will be 'unable to determine'.

I think if you explain to the customer in the event it is needed they would understand tightening the tolerances in order to issue a proper statement of compliance is good for them.

As to the specification it does not contribute to the system uncertainty, the system uncertainty is dominated by the resolution causing the approximate 1.67 TUR. By simply guard banding the specification of the measurement at approximately 80% you can issue a statement of compliance.

In theory 80% of the 0.000065 specification (assuming quantization error) is 0.00005 so if it measured within 1 resolution 0.00005 or less it could be claimed as conforming. if it measures out by 0.0010 2 resolution counts it would be definitely out of specification. If you went with the accuracy spec only it would have to measure perfect as 1 resolution would be indeterminate.

[stefanhg]

Correction: TUR = (UUT Tolerance / Expanded Uncertainty)

In the standards ISO 3611 and DIN 863 the maximum permissible error (tolerance) for micrometer 0-1”/0.00005” is 0.0002”.

With this tolerance and uncertainty 0.00003" the TUR will be 6.67:1

My opinion is that some manufacturers define increased accuracy, respectively too small permissible errors. Other manufacturers, which micrometers are not worse than those mentioned above, use in their specifications the recommended tolerances from the standards. I also find these error limits more realistic.