Calculating TUR to include measurment error

[civvychris]

I have a instrument with a OEM stated tolerance of ±2 that has returned from calibration with an as-received error of +0.5

The uncertainty of the standard used for calibration as stated on the certificate is ±0.02

The original TUR;  OEM tolerance ÷ uncertainty of standard was calculated as 2 ÷ 0.2 = 10:1.

Question - how can I determine a new TUR that takes account of the +0.5 error?   

[CalLabSolutions]

You are trying to comparing apple to oranges. 

If the test limits where 2% and the Unc was .2 then the TUR was 10 to 1.

If you send the item off to be calibrated and the test limits were +/-2% and the lab that tested it reads +.5 % with a stated uncertainty of .02% then the value on the calibration would read +0.5% +/- .02%.  That means the lab's calibration data is stating they they know the true value is somewhere between 0.48% and 0.52%.
*
AND NOTE. That does not mean that the Unit Tested's new accuracy is not +.5 +/- .02%. (I see people trying this bs all the time, and do not have the data to back it up.)

Mike

[RFCAL]

Hey Mike!! you're correct!!

[beadwork]

So when you use this instrument part of the Uc budget for it is the OEM tolerance of ±2.

You can't use it as having a new tolerance of +.5.  That was the error at the time it was measured during the recent calibration.

[CalLabSolutions]

I don't want to say you can use the +.5 as a calibrated traceable value.  But you have to be able to prove your instrument's long term stability is better than the manufacturer's +/-2% spec.  And you can't do that with one calibration.

I know of one company that has years of data on a plane jane 3458A that our performs their 3458A Option 002. But you have to have data to back up your claims to better accuracy.

Mike

[OlDave]

It's a lot easier if we are talking about a physical standard like a gage block or a mass standard.

Say this is a 1kg weight that measured 0.5mg heavy. 1000.0005g becomes your new known value for that artifact. In this example it would then have an uncertainty of 0.02mg., though the original post didn't really state a level of confidence on that value.