| Résumé | The subrange inconsistency (SRI) of a large ensemble of long-stem standard platinum resistance thermometers (SPRTs), representative of worldwide production, has been investigated for all pairs of ITS-90 overlapping subranges between 83.8058 K and 933.473 K. The results are reported in terms of various statistical parameters to facilitate their comparison with the results of different authors, although a statistical test, applied to one specific pair of subranges, supported a Gaussian distribution of the results and justified the subsequent use of the mean and standard deviation as statistical parameters. Depending on the pair of overlapping subranges, the mean SRI as calculated varied from -1.23 mK to +0.21 mK and the SRI standard deviation varied from 0.04 mK to 0.62 mK. These numbers generally increased with the upper temperature limit of the pair of subranges, especially where the lower subrange requires a point which is not included in the upper subrange. The contribution to SRI from the fixed-point uncertainty propagation (PoU) was evaluated. The results showed that, although the effect of PoU on SRI largely cancels out for points common to both subranges, PoU still amounts to 59 % to 130 % of the differences between overlapping pairs of subranges. This means that the differences as calculated are probably a substantial overestimate of the true SRI. It is suggested that this effect is taken into account in making recommendations for typical uncertainties due to non-uniqueness. |
|---|
