Why does Annex C label repeatability Type A but use a rectangular distribution?

Check the conversions within the same table

Locate w_re in Table C.1 and compare its “Type A” and rectangular-distribution labels with the Type B entries for w_r, w_z, and w_rep. Although the underlying variation quantities differ, these conversions all multiply a range by 0.5 and divide by √3. We are comparing the core conversion to standard uncertainty here; obtaining a relative percentage still requires the conversion based on the mean specified in the original table.

Next, check the definition of b_re in C.6: b_re = X1 − X2, the difference between readings taken in an unchanged mounting position. It is a range-type quantity here, not a standard deviation calculated from a series of readings. Identifying the input quantity first makes the issue between label and method visible.

Why this is an interpretation issue, not a distribution-name error

In this method, b_re × 0.5／√3 takes the half-width of the range and converts it using a rectangular distribution. The other Type B entries in the same table follow that approach, yet this entry retains a Type A label. Our finding concerns the inconsistency between the label and the evaluation actually used.

Table 3 for tool calibration is different: for w_re, b_re is a standard deviation, converted using b_re／√n and labeled Type A with a normal distribution. Range-based components are labeled Type B with a rectangular distribution. That table is internally consistent. The annex issue does not justify treating it as incorrect or substituting the tool repeatability method for the device method.

Evaluation type and distribution model are separate concepts; “Type A must be normal” is not a general rule. Nor does having measurement data automatically require Type A evaluation. Taking the half-width of a range and converting it with a rectangular distribution is common in Type B evaluation, but it is a distribution conversion and cannot by itself settle whether an entry should be labeled Type A or Type B.

Preserve both the label and the calculation basis

Record the standard edition, clause or table, original entry, and adopted reading together. Retain the original Type A label and separately explain the use of a range, half-width, and distribution. Keep the original readings, their difference, and the conversion traceable in the calculation record instead of recording only a classification letter.

Continue to use the device’s tabulated formula. A classification query alone is not a reason to substitute a standard deviation or another distribution. Relabeling does not change the numerical result of an unchanged formula. Changing the input quantity or calculation, however, changes the method and cannot be described as a wording correction. Preserving this distinction leaves a clear basis for technical review.

This is our reading, not an official ISO corrigendum. ISO is revising Part 1 and Part 2 through committee drafts ISO/CD 6789-1.3 and ISO/CD 6789-2.3. We will update this page and recheck the issue when the revised standards are published.