Uncertainty and TUR

OlDave · 07-09-2011 -- 06:41:50

I'm sure a lot of you guys are more knowledgeable than I am so help me understand some things I'm noticing more and more on calibrations certificates.

I am getting cal certs that state the uncertainty, but the uncertainty exceeds the accuracy tolerance of the UUT. In other words, the accuracy of the item may be ±0.01, but the stated uncertainty of measurement is 0.05 at that test point. On another certificate they list the TUR at each test point and the TUR was 0.18 at one value.

Is this valid and acceptable? Isn't it telling me their standards and/or processes aren't accurate enough to calibrate my item?

CalibratorJ · 07-09-2011 -- 07:24:22

I'm pretty sure that as long a they are listing the Unc and the TUR, it is accepted practice.

The question is, are those Unc acceptable by YOU? And if they quoted the work for you, did you request a specific Unc needed, ie mfrs specs/unc......

This is one of the MAIN problems in the calibration world.... there is no standard.... you can get away with murder as long as you state that your standards suck on the cal reports. And a fancy logo means absolutely nothing until you actually look at the scope that "earns" them that fancy logo.......

Sorry, ok, off my soapbox now.....

PMEL · 07-09-2011 -- 09:03:17

CalibratorJ if the lab in question cannot make the measurement better then the specifications according to their scope of accreditation then yes the uncertainties will indeed be more then the specifications. A2LA is going to make it mandatory very soon to list uncertainties on all measurements. This type of calibration is going to be the future. Could you be a little more specific like what piece of equipment you seen this on. Was it a primary standard? If it was something like a Fluke 87 then I would for sure question this. I do not like this practice for a customer that has no idea what a specification or uncertainties are.

CalibratorJ · 07-09-2011 -- 15:21:50

PMEL, I am very aware of how the whole A2LA uncertainty reporting process works, all too well, unfortunately.

"Is this valid and acceptable? Isn't it telling me their standards and/or processes aren't accurate enough to calibrate my item?"

To answer your questions, YES it is valid; DUNNO, is it acceptable to you, meaning are those uncertainties good enough for you to use in your lab on a daily basis without compromising your measurement ability; and YES it is telling you that they ain't that good (especially the TUR, not so much the uncertainty). If they are accredited, you might want to pull their scope from the website and see if they can even get close to those uncertainties.

One thing to keep in mind when looking at a cal lab's scope, those are absolute best case. Most labs won't always hit those uncertainties, and it's (unfortunately) perfectly fine, just so long as you never, ever, ever, list an uncertainty better than what is on your scope.

Kalrock · 07-09-2011 -- 22:46:23

I'm with Dave in the fact that I probably don't know as much as some people, but the best thing to do in this case is call the lab in question.

One of the other possibilities is that the lab is giving you an uncertainty for the TI. In this case the uncertainty would be greater than the accuracy since it would be RSS'd with the accuracy of the TI and the uncertainty of the standards.

Also you could check the standards on the cert and see what their accuracy is then you could figure out the Type B uncertainty which would give you an idea of what they did, but that is a lot of work when a phone call would do the trick. There isn't any sense in jumping to conclusions. Calling the lab is the best option.

calibr8tor · 07-10-2011 -- 07:48:34

With the increased accuracies of test equipment, 4:1 TUR's are getting harder and harder to meet. Getting1:1 TUR's are fairly common especially when you are doing primary standards. Anything less than 1:1 is not good practice, and the company should have contacted you prior to calibrating this piece to let you know that they could not give you better than 1:1. Perhaps they could have used guarbanding to increase the TUR's. I would not accept anything with a TUR of less than 1:1. That's just my opinion, and you know what they say about opinions.

darkness63 · 07-10-2011 -- 13:37:57

I agree with Kalrock, I have seen that many people are putting their uncertainty budgets with the TI in it. Therefore, the uncertainty would be greater than the accuracy of the TI.

This is OK as long as it is stated that this is what has been done. However, if this is done and then the TI is checked to it's accuracy spec and states an uncertainty spec, in my opinion this is wrong.In this case that uncertainty spec would be that of the measurement applied through the TI to something else not of the measurement being applied to the TI.

calibr8tor · 07-10-2011 -- 14:47:01

If it is listed as measurement uncertainty or uncertainty of measurement, then it shouldn't have anything to do with the accuracy of the test instrument other than the standard deviations. However, if it is listed as Test Instrument uncertainty, then accuracy will come into the picture.

OlDave · 07-11-2011 -- 12:24:09

Well a quick update. After a couple of phone calls, a few emails and 2 (yes two) revisions of the certificate, the uncertainty is now less than the mfg. specs of the instrument at all test points.

Not something I should really have to do with a NVLAP accredited cal facility wouldn't ya think?

Haven't started on the one that reported in TUR.

RFCAL · 07-11-2011 -- 18:09:58

it is up to the customer to state what would be an acceptable level of uncertainty. If that is not made clear in the beginning,then the certificate may state whatever.
I would question any unc that is greater than the spec.That is just too crazy and I would question that lab.Would you care to name the lab? A fly by nighter I bet.

PMEL Whore · 07-12-2011 -- 06:15:39

This is what's wrong with commercial cal, very limited standardization, so evrybody does it different and you end up comparing apples and oranges. BTW does anyone know the conversion factor from apples to oranges?

OlDave · 07-12-2011 -- 06:21:31

Here is what the lab supervisor said about the errors..."I think that my review and the techs setup were very sloppy on this one....the uncertainty issue did surprise me, since it is programmed in I rarely check the math."

This place has been around for awhile and I never really thought of them as a "fly by night" outfit.

If you are really interested in who, send me a PM request. I don't want to name names in the open forum.

Lasario · 09-24-2011 -- 09:13:39

Quote from: PMEL Whore on 07-12-2011 -- 06:15:39
This is what's wrong with commercial cal, very limited standardization, so evrybody does it different and you end up comparing apples and oranges. BTW does anyone know the conversion factor from apples to oranges?

This sums up the problem for most labs. They are too many ways to report uncertainties and all labs do it differently. When reporting uncertainties, the resolution of the TI has to be considered. This can cause some strange uncertainty reports even though the standard is better than 4:1. I think ILAC is trying to standardized the reporting of uncertainties , but from what I have seen so far there is no true golden rule.

WestCoastCal · 09-24-2011 -- 16:44:19

If a lab is able to provide uncertainties and test ratios it demonstrates they can calculate them (however accurately). Don't all labs use the GUM? Providing the uncertainties/test ratios is certainly better than what has been done in the past. How many calibrations have I seen where the ratio is <4:1 and it's not reported? However large the uncertainty is, we at least have the information necessary to make decisions on the specification that we can use for our intruments. I'm not a fan of derating specifications, but sometimes it may be the only choice.
I heard that ILAC is making it a mandatory requirement (not sure if they have the authority to do so) that all 17025 accredited test points include measurement uncertainty. Anyone??
ILAC's home page does a really good job explaining their objective, what accreditation is, what its purpose is and what its goal is. With that being said, it seems to be that accredited labs are on a level playing field with each other with standardization being the goal. How well labs implement the quality standards/standardization is certainly open to interpretation.

RFCAL · 09-25-2011 -- 11:22:28

Uncertainty of measurement should never include the spec of the TI,but should include resolution,drift,stress,environmental factors,or anything that affects the measurement of the instrument you are trying to calculate for. For example, the subject parameter is a Wavetek 4808. The Meaurement parameter uses a 3458A/002 to measure the 4808 1VDDC. You will need the unc from the last 3458A cal, A bias from the 3458A measuring 1VDC,Resolution of both instruments,A drift of past 3458A 1VDC cal data,cables, operator bias,Environment affect, and anything else that will affect the measurement--you do not include any specs from the 4808.You can calculate them, but you do not include them. If you need any more info,send me a private message and I will discuss further.