Calibration certificates avowing accuracy beyond the uncertainty limits?

M

mlee97

Team,

I have noticed that it is a practice within my area calibration shops to calibrate and attest beyond the resolution of both their stated uncertainty and their ISO/IEC 17025 or A2LA accredited uncertainties.

For example: Their calibration certificate may post an uncertainty of +/- 0.00072 inch, but they call a gauge good when the gauge tolerance posted on the cert is +/- 0.00005 and they measure 0.000036".

To me, if you aren't 95% certain of a measurement at a 0.00071" then you shouldn't be attesting to the gauge's accuracy.

Am I misunderstanding something?
 
Last edited by a moderator:

SMcNearney

Registered
I want to ensure I am saying this correctly and not be confusing...
To clarify, the calibration uncertainty is the gold versus what the manufacturer states is the tolerance. Honestly, the manufacturer can state whatever they want, and it's up to you, the end user, to determine if its accuracy is good enough for what you need to do. I also want to ensure we are determining the difference between measurement uncertainty and expanded uncertainty. The calibration laboratory's CMC (old term of BMC) is not going to be what's expressed on the certificate, but rather the expanded uncertainty of the instrument under test's measurement capability is. ie: yes, manufacturer states the tolerance is 0.00005", however due to the calibration process, the measurement uncertainties from the laboratory, the resolution and repeatability of the tool, it could very well put the expanded uncertainty over a tenth of a thou. The expanded uncertainty SHOULD always be higher than the manufacturer stated tolerance. In the event it's not, someone's doing their math wrong. This is very similar to say, a Mitutoyo's electronic micrometer that has the resolution of 0.00005" and Mitutoyo says the resolution is the accuracy (a typical 0-1" micrometer). The mere thought of putting a gauge block to it, to verify, is going to make the measurement uncertainty go higher because now we have to include the uncertainty of the standard, the temperature coefficient, repeatability tests against the resolution, etc. In other words, the Pass/Fail criteria really needs to fall on you, not the calibration laboratory. A lot of standards are heading in this direction, notably Z540.3, the new 17025-2015, and AS9100. I am not that familiar with the other ISOs but this is the direction the world is heading in. I hope this answers your question,
-Sean McNearney, CCT (ASQ)
 

dwperron

Trusted Information Resource
First, when you choose a lab part of your due diligence is to assure that their best measurement uncertainty (CMC) is adequate for your requirements. If their CMC is adequate but the actual uncertainty for your calibration isn't then tell them to do it again.

Next, part of your contract review is to accept their "Decision Rule". This is how they define their determination of pass of fail, conformance, etc. If they claim to only report measured results and not include measurement uncertainty in determining conformance then you have allowed them to report the measured value and their uncertainty. You now accept all risk as to whether or not this is pass, fail, or indeterminate when the uncertainty is included. A good document describing this process is ILAC-G8:03/2009. This one is free, and an A2LA lab like the one you used follows ILAC rules.

Also, check the "boilerplate" on your calibration documents. It is very common for the lab to claim a test accuracy ratio (TAR) or test uncertainty ratio (TUR) of > 4:1 unless noted. If this is on your certificate and they don't note that they are under a 4:1 ratio then they failed to comply with their own requirement.

I don't know what your agreement with you calibration provider included. It is certainly fair to question the adequacy of their calibration, and a respectable lab would admit that what they provided was woefully inadequate. If their CMC indicates that they have the capability to make the measurement with a lower uncertainty then request that they recalibrate it with better equipment. Otherwise, demand your money back and look for another lab.

Unfortunately, buyer beware!
 
M

mlee97

Thank you guys for your replies. I did get a reply from one of the calibration companies who basically stated that their TUR was 2.81:1 and that it is ok.

I'm not buying that yet. First, the TUR sounds pretty low, but what are the Tol's used to get that TUR? The gauge in question was a tapered thread ring and the tolerances I was referencing was the one's stated on the cert. I suppose they got those from a thread standard. and I "assumed" that their tolerance as reported on the cert, STD 1.822340, Tol - 1.822040, Tol + 1.822640 was for the gauge, NOT the measurement. They used the object gauge Tol of .0006 (+/- .0003) for the TUR ratio not their gauge's tolerance.
I thought it would have to be able to actually measure the last stated digit of the upper and lower tolerance.
my last non-zero digit is a 1.82204 so to be able to differentiate between a -1.82203 good reading and a -1.82205 bad reading my Tolerance for the TUR would be +/- 0.00001 not +/- 0.0003.

Who is correct?
 
Last edited by a moderator:

dwperron

Trusted Information Resource
You have piqued my curiosity.

You say this is a tapered ring gauge (NPT?).
I wonder what they used to make these measurements.

Could you supply a copy of the certificate?
 
M

mlee97

The calibration company wouldn't like me sharing their masterpiece.

They claim to have used a "bore set master" and a SIP 550mm measurement center as well as a set of t-sphere probe heads.

Does that sound adequate?
 

dwperron

Trusted Information Resource
My understanding is that for tapered (NPT) rings you use a calibrated master plug, You insert and turn until it bottoms out. The ring is acceptable if measured result is +/- one third of a turn from the NPT Master Plug Gauge Basic Notch.

Thread ring measurements are not typically direct measurements, they are comparisons to a master. The readings, and the uncertainties, are based on the calibration results for the master plugs. The equipment listed in the certificate are probably those used un measuring the plug, not the ring.
 
Last edited:

dwperron

Trusted Information Resource
Concerning the uncertainties provided by the A2LA lab being higher that the tolerances, I attended the A2LA technical forum this weekend, and this issue actually was raised in discussions. A2LA has a requirements document, R205, that your lab must adhere to.

In section 6.4.5), Statements of Compliance, it says:

"when parameters are verified to be within specified tolerance, the calibration laboratory shall determine the uncertainty and take that uncertainty into account when issuing the statement of compliance. In addition, no claim of compliance shall be made when the measurement uncertainty is larger than the tolerance being evaluated.

An exception can be made only in cases where the laboratory indicates in the
contract with the client that the calibration results will be reported without
factoring in the effect of uncertainty on the assessment of compliance, and the client agrees to the contract.

In this case the uncertainty can be excluded when making that statement of
compliance on the calibration certificate. In effect, both parties share the risk that the results may or may not meet the specification since the uncertainty was not included when the results were determined. In these cases, the measurement uncertainty shall still be determined and shall be reported on the calibration certificate."

This gets back to the "contract review" between you and your calibration supplier. Unless you agreed with them not including the uncertainty in the pass/fail determination (the decision rule) they cannot claim that anything passes or fails if their uncertainty is larger than the tolerance. If you did agree then the calibration results they provided are adequate, because you agreed that they are.

If they refuse to address this with you then you could report them to A2LA for this violation.
 
Top Bottom