New ILAC Requirements Regarding Repeatability - Your thoughts on this please

[stefanhg]

Hello adamt,

Regarding the measurement uncertainty of calibration of vernier caliper:
- The uncertainty of the calibration cannot be less than resolution of the caliper.
- The uncertainty of the calibration is reported normally with two significant digits; this is not the measurement uncertainty when you use the caliper in the workshop.
- The uncertainty of the first lab is quite small, but possible - see the first example below.
- The uncertainty of the second lab is too large - they cannot make any conformity assessment.
- If the MPE (maximum permissible error) of the vernier caliper is 0.0015, for the first lab TUR=1:2.7, for the second lab TUR=1:0.9

Method 1: use of indicator for the applied measurement force, gauge block grade 1 with MPE = 0.000008in. acc. to NIST
Uncertainty contributions:
- Gauge block - rectangular distribution with half-width of limits =0,000008in; standard uncertainty = 0.0000046in
- Resolution of the caliper - rectangular distribution with half-width of limits =0,00025in; standard uncertainty = 0.00014in
- Calibration method (Mechanical effects, temperature differences) - normal distribution with pooled standard uncertainty = 0.00025in
Combined standard uncertainty: u=0.00029
Expanded uncertainty with normal distribution: U=2*u = 0.00057in
Expanded uncertainty with trapezoidal distribution (Formfactor 0.3): U=1.9*u = 0.00054in

Method 2: without indicator of the measurement force, gauge block grade 3 with MPE = 0.000048in.
Uncertainty contributions:
- Gauge block - rectangular distribution with half-width of limits =0,000048in; standard uncertainty = 0.0000277in
- Resolution of the caliper - rectangular distribution with half-width of limits =0,00025in; standard uncertainty = 0.00014in
- Calibration method (Mechanical effects, temperature differences) - normal distribution with pooled standard uncertainty = 0.0005in
Combined standard uncertainty: u=0.00052
Expanded uncertainty with normal distribution: U=2*u = 0.0010in

[dv8shane]

B

In Reply to Parent Post by Jerry Eldred

On the one hand, it requires including repeatability and suggests reproducibility. But on the other hand, it says to exclude significant contributions from the unit under test (UUT). In things such as digital multi-meters, when calculating uncertainties using best available UUT, and lets say for example the accredited standard is a Fluke 5720A, and a best available UUT might be something like an Agilent 34401A (not quite 4:1 on everything).

And let me throw in the caviat that we have Agilent 3458A's and Fluke 8508A's. But since the 5720A is not good enough to calibrate a 3458A or 8508A, it doesn't seem reasonable to use them for doing repeatability experiements.

All that said, if I then use a 34401A for the repeatability tests, I believe the 34401A may contribute the majority of the repeatability (??). So if I have what ever amount of non-repeatability... lets say for example I do a set of connect/disconnect/STBY/OPER at 10 VDC with the 5720A connected to the 34401A using good 5440 cables. I do a set of 20 readings, and for the sake of discussion (I'm making up numbers for this discussion) I have a 2 Sigma value of 0.00003 VDC. So if this number is likely instability/non-repeatability of the 34401A multi-meter, it should not be included, and should be so stated in the uncertainty budget.

Jerry

In my opinion not only do you include the repeatability but the UUT resolution as well. The 3458A or 8508A should be used even though you can not issue a conformance statement under the 4:1 premise. The reporting will just be the measurement and the uncertainty of the entire measuring system instead. The results in your budget should be better than using a 34401A

Shane

[adamt]

Hello stefanhg,

I have been meaning to reply to your post sooner but you know everyone is busy. Your comment that the uncertainty is to high I agree with.

Do you have any resources that I can read to say that the measurment uncertainty can not be larger than the specificaion on a certification? The reason I ask is that my opinion dose not hold a lot of weight with the person I need to have a discussion with.

[blueicecube2]

If you wanna make your life easier (i.e one MU to serve them all), for the Type A repeatability, use your best of all UUT Spec for the Type A in the CMC calc.

Always, CMC shall be bigger than the reported MU.

(I've always used top bottom method, means calculated MU for every point. But some people likes bottom top, once and for all MU)