eeeeeek!
Charles Wathen said:
Hehe, for a headache, look at this uncertainty analysis:
http://www.thunderscientific.com/reflibrary/25rhuncert.pdf
I doubt most of us could even follow this. ... I was fortunate that I took Dr. Howard [Castrup]'s class for a full week, ...
We need a new Smiley -- "Pulling Hair Out!" (For those of us who still have any, of course.)
I agree with Charles, that example is a bit difficult :mg: to follow, especially at 2230 when I should be shutting down the confuser instead of visiting the Cove! On the other hand, if you skim over the calculus that paper does give a very good example of the type of analysis that often needs to be done. Now, we have to keep in mind that
(a) a two-pressure humidity chamber
is a primary standard,
(b) relative humidity does have very complex interactions that only a physicist could love, and
(c) the author certainly has a higher degree than
I do!
Also the theory is - or so I have been told - that for "a measurement system" this only has do be fully done once and then reduced to either a single number or software code, and re-verified every now and then.
The problem is, I (most of us?) am too short-staffed and spend too much time firefighting to do it consistently. Also I don't have enough equipment to make a
dedicated system for each type of measurement! For example, I have one HP 8902A: is is used about 50% for calibrating communications and navigation system test sets, 30% for calibrating signal generators, and the rest of the time for attenuation and other miscellaneous calibrations. That's three different workstations and three different "measurement systems" and so would require three different uncertainty analyses. (Or, I
really need two more 8902A units and then the three analyses - but the lab is a department of a
major airline and you can see in the papers what the money situation is.)
I've taken one uncertainty course myself but it was over ten years ago. I'm planning on taking a half-day intro that Dr. Castup is doing after the NCSL conference this summer, and if I can ever get someone else to pay for it I will go through one of his full courses and get the software.
Actually, my customers don't even want a certificate - all they want is a new sticker on the instrument
right now. The only ones we print (they're all in the computer - it's automatic) are where they
must have a report of value to use the instrument. As for uncertainty - I evaluate it when writing a cal procedure and apply guardbands if needed but we don't report it - and if we did the customers would not know what to do with it. So for me doing uncertainty the way 17025 wants it is
not feasible or economic and to that extent I agree with CalGuy's assessment of the standard. On the other hand, it is also an idealized goal to strive for ... a benchmark of some kind.