Readouts and bridges for reference thermometers

A reference thermometer's certificate is normally for the sensor alone, expressed in resistance or via ITS-90 coefficients. Converting a resistance into a temperature requires a readout whose own accuracy, resolution and stability are known. Precision thermometry readouts and resistance bridges span a range of performance from 0.01 °C to below 0.001 °C equivalent.

What to check

• Calibration of the readout's resistance measurement, or system calibration with the probe attached.
• Resolution and noise at the resistance of interest.
• Linearity across the range if calibrated at few points.
• Measurement current and whether the readout can halve it to check self-heating.
• Coefficient handling: ITS-90 or Callendar–Van Dusen coefficients entered and verified.
• Drift of the readout between calibrations.

A common error is to enter certificate coefficients into the readout and then treat the displayed temperature as if the readout contributed nothing.

What is actually reading the probe

Three classes of instrument turn a reference PRT into a temperature, and they differ by roughly two orders of magnitude in what they contribute to the budget.

The conversion to remember is that a Pt100 changes by about 0.385 Ω per °C near 0 °C, so 1 mΩ of resistance error is roughly 2.6 mK. At 100 Ω, 10 ppm is 1 mΩ. That single line lets you read any readout specification straight into the temperature budget.

Self-heating, and how to measure it rather than assume it

The measurement current warms the element it is measuring. A 1 mA current through a 100 Ω sensor dissipates 0.1 mW, which in stirred water might raise the element by a millikelvin and in still air by tens of millikelvin. Because the power goes as the square of the current, the check is easy: read at the normal current, then at half it, and the difference in the two temperatures is three quarters of the full self-heating error. Multiply by 4/3 to get the correction at normal current.

A readout that cannot change its measurement current cannot support this check, which is a real argument for the better instrument. It also matters that the current used in service matches the current used at calibration — a certificate taken at 1 mA does not describe the sensor at 2 mA.

Wiring, and the error that hides in it

A four-wire connection carries current on one pair and senses voltage on the other, so the lead resistance drops out and only the element is measured. Three-wire compensation assumes the two current legs are equal and subtracts one from the other; when they are not — different lengths, a corroded terminal, a junction box repair — the residual appears directly as a temperature offset. A 0.1 Ω imbalance is about 0.26 °C. For reference work, use four wires and verify continuity of all four.

Checklist before the readout enters a budget

• Is the readout calibrated for resistance, or is the probe-plus-readout calibrated as a system? A system calibration is simpler and usually smaller, but the pair must then stay together.
• Are the ITS-90 or Callendar–Van Dusen coefficients from the certificate entered and read back? Transposed digits here produce a confident, wrong temperature.
• Is the resolution adequate? A display rounding to 0.01 °C contributes 0.0029 °C as a rectangular half-interval, which can outweigh the reference itself.
• What is the drift specification between calibrations, and has it been checked against the last two certificates rather than taken from the brochure?
• Has the noise at the working resistance been observed, rather than assumed from the last digit?