UNcertainty Queries !!!PLease reply !

[neo_inst]

Hello everybody,

I accidentaly came across your web site while searching for some material on Uncertainty calculation which is an integral part for 17025 certification and i beleive i stumbled upon a GOLD MINE with such experienced people exchanging their views on the topic.

I am working on this project as a student on 17025 certification of a calibration lab. I personally feel its like Quick sand ! ..the more i dig into it the more deeper it gets.

First of all !!...i have a very trivial query

While Calculating uncertainty in a measurement we first classify the factors into type A and B . Now , i have gone thru the examples given in the NIST and EAL guides wherein they have taken eg like dead weight tester and Volume of cylinder wherein the input -output relationship is known . and we can find the affect of input uncertainty on output
eg : In volume , uncertainty in diameter and height dimensions affect the output.

BUT for complex electronic equipment these equations are not given by the manufacturer...then in this case do i just neglect this factor ??

Also i would be really greatful (if possible) whether anyone can tell me the procedure they followed since the eg. given in the books really dont give any insight into the procedure esp for electrical equipments eg Calibrators..

Hope someone would reply.

I would really appreciate any advice regarding the procedure.


NEO
:

[Jerry Eldred]

I can give a very basic answer.

When calculating uncertainties of complex electronic equipment from a particular manufacturer, how much detail they give you (in my experience) is a variable (sorry to say) depending on the manufacturer, and the certificates of calibration you receive.

You won't typically find uncertainties in a manufacturer's manual. You will find their claimed specifications. This is usually based on them characterizing the equipment operation under a particular set of variables, and to some Sigma confidence level, they will combine all data into their claimed specifications.

You will more typically find uncertainties when you have the equipment calibrated by a competent calibration lab (which may include the original manufacturer -- sometimes). A high accuracy standard resistor for example, may have on the certificate of calibration, a certified value, a tolerance limit, and a statement of uncertainty. It is not very typical for Type A and B to be subdivided out. That is something more common to when you are doing your own uncertainty calculations.

I unfortunately don't spend a lot of time in doing uncertainty calculations. I am a generalist (that is, I know some about a lot of areas of calibration, but I am not a leading expert in a particular area). So please take my reply from that context.

When you are calculating your uncertainties, if you are using a certified value standard such as in the example, you would calculate based on that certified value, plus all the appropriate uncertainties. If you are using a complex standard such as an HP 3458A high accuracy DMM, you would more typically use the manufacturer's stated tolerances, as the meter is being used based on tolerance limits rather than a specific value.

Don't know if this has helped or not. Perhaps some of the uncertainty calculating experts could render some more detailed opinions on this one.

[Atul Khandekar]

Have you looked at NIST Engineering Statistics Handbook?
http://www.itl.nist.gov/div898/handbook/

You may find some help there.
-Atul.

[metrologyguy]

I understand your fustration over this issue having gone through it for my lab. Uncertainty calculations for calibrations are a relativly new concept for a lot of us that have been in the business for a long time.

I can make two recommendations for your situations:

1) Take a class in uncertainty calculations. I know this is not a great answer but, trying to learn this stuff by reading is just not that practical. I recommend a class taught by Dr. Hendrick Neilsen, He has taught many A2LA assessors. We had him as our assessor one year and I was pleased to see he applies common sense and can relate to us folks that don't have a Dr. in our title.

2) Take your best effort and go with it. Do your homework and document your logic. Chances are your assessor will want to make changes to what you do anyway. Every assessor we have had wanted our uncertainty budgets recalculated to their satisfaction. Make a spreadsheet with your best guesses and don't get hurt if and when you are requested to make changes. Consider it a learning experience.

I spent months on my spreadsheets, met with friends at NIST who made improvements with them, then I took them to an ASME/ANSI meeting and met with the Chairman of their uncertainty committee who made additional improvements.

Each year my uncertainty budgets get better and I learn more.

It's only tough until you know it.

Keith Kokal

[Ryan Wilde]

Neo,

The first point that Keith makes is very correct - find and take a class. I took the same one the Keith took, and while it doesn't tell you how to do uncertainty for every scenario, it teaches you what to look for and add to your budget. Uncertainty software is also extremely valuable, and I know of a great FREEWARE uncertainty calculator that also has some examples provided.

Jerry is also correct in that most complex equipment will have a "specification sheet", and that sheet SHOULD also tell the environmental factors that the specifications require to meet it. Then you wind up with a fairly simple budget, with the specification being a Type B component, and your own data will fill the rest of the data. Basically, in my experience no valuable uncertainty budget can be developed without some testing of parameters, such as repeatability, drift over time, lead effects, etc.


I have run into equipment with no useful information provided, in which case a very exhaustive study must be accomplished. I have one running right now with a unit in a thick cardboard box which retains heat, just so I can find the effects of temperature on the accuracy, stability, etc.


Ryan

[lhiga]

Neo,

I know what you are going through, as I was you about 6 mos. ago. I have taken Dr. Hendrick Neilsens class (given by Mitutoyo) and I gotta say, he teaches a common sense approach to determining uncertainty. Be advised, though he teaches it with respect to mechanical calibration devices...but, the principles are easily transferred to electronic devices.

Now for your problem...for me, I work in a electronic calibration lab. but the only devices that I need to determine uncertainty for is the lab standards that we actually calibrate. For us that would be resistors, shunts, DC voltage, and thermal voltage converters.

Basically, for type A data I would set up the procedure to make 3 to 5 measurements then do a standard deviation and apply the safety factor for the limited amount of samples.

For the type B uncertainty, it would be all the other uncertainties associated with the measurement. For example, meter readabilities, standard uncertainties, temperature, ancillary equipment uncertainties, thermal affects uncertainties. Just make a list of everything you believe could affect the measurement then try to consolidate these uncertainties as best you can. Then look at the affect of these uncertainties on the devices final outcome. The uncertainties that have little effect can be eliminated, keeping the major contributors.

And thats it....... hope this helps...