Calibration and self-validation of thermistors for high-precision temperature measurements

• Variables affecting the uncertainty of precision NTC thermistors were investigated.
• The proper choice of model parameters strongly influences the interpolation error.
• The Steinhart–Hart equation shows a poor performance and should not be used.
• Drift should be quantified by single-point validations at the triple-point of water.

In this paper we discuss influencing variables affecting the uncertainty of high-precision temperature measurements (u < 1 mK) by means of NTC thermistors. These are proper instrument settings, a suitable choice (distribution) of calibration temperatures in due consideration of uncertainties, self heating and the number of parameters in the calibration equation. Within this work we used 4 wire measurements to eliminate the influence of lead resistances, switched dc current to reduce errors by thermoelectric effects and amplifier offsets, a reference resistor with a nominal value close to the thermometer resistance to maximize the resolution and a maximum current consistent with the input voltage range and self heating of the thermistor.We present results of high-precision calibrations of a so called super-stable thermistor and demonstrate the influence of changes of the calibration equation on the interpolation error. Our results confirm previous findings that the number of parameters in the interpolation equation can have a considerable influence on the interpolation error. It was confirmed that the Steinhart–Hart equation shows a poor performance and should be replaced by the more suitable models recommended in White et al. (2014). For the quantification of the long term stability of a calibration we recommend repeated single-point validations at the triple-point of water. If these are supplemented by measurements at the gallium fixed-point possible changes of the curvature of the characteristics can be detected.