A statistical analysis of interpolation errors in precision electronic thermometer calibrations: lessons from 263 calibrations at NMI

摘要

Precision electronic thermometers are now commonly used as reference standards for calibration laboratories and other precision applications. These devices can often perform better than the manufacturer's specifications, however the industrial-grade PRT sensors, particularly if used over a wide temperature range, are often inadequately modelled by the interpolation equations available to the instrument. The estimated interpolation error of the calibrated thermometers is dependent on the number of calibration points used, and since most calibration laboratories perform calibrations manually, there is a trade-off between uncertainty and cost. This paper considers 263 calibrations performed by NMI on high-precision electronic thermometers. At NMI calibrations are performed at a large number of temperatures, with a calibration-system uncertainty of 3 mK (5 mK above 250 deg C). We can thus examine the interpolation error introduced by calibration at coarser temperature intervals. We find that use of 20 deg C intervals below 0 deg C and 50 deg C intervals above 0 deg C is sufficient to achieve interpolation errors of below 5 mK in most common precision electronic thermometers.