Thermistors: high sensitivity over a narrow range

A thermistor's resistance falls roughly exponentially with temperature. Around room temperature a 10 kΩ device changes by about 4 %/K — ten times the relative sensitivity of a Pt100 — so small temperature changes produce easily measured resistance changes. This makes thermistors attractive for cold-chain loggers, stability chambers and medical thermometry.

Limitations

• Non-linearity: the resistance–temperature relation must be characterised per device or per batch.
• Self-heating: high resistance limits current, but small beads still heat in still air.
• Range: exposure above the rated temperature causes permanent shift.
• Interchangeability: precision types are matched to ±0.1 °C; others vary widely.

Well-aged precision thermistors can be more stable than industrial PRTs in the 0–70 °C range; their calibration history tells you whether yours are.

Sensitivity is the whole argument

Around room temperature a 10 kΩ NTC thermistor changes by roughly 4 % per kelvin — about 400 Ω/K — against a Pt100's 0.39 Ω/K. In relative terms that is ten times the sensitivity, and in absolute resistance it is a thousand times, which means the readout's resistance accuracy stops mattering almost entirely.

That is why precision loggers and monitoring systems in the 0–70 °C band use thermistors: a modest measurement circuit gives millikelvin resolution, lead resistance is negligible, and no four-wire connection is required. Over a narrow range around ambient, a well-aged precision thermistor can be more stable than an industrial PRT.

The costs that come with it

• Non-linearity. The resistance falls roughly exponentially, so the relation has to be characterised per device or per batch — usually with the Steinhart–Hart equation, which needs three calibration points, or a beta value over a narrow span. Two points will not do.
• Range. Useful span is roughly −50 to 150 °C, and much narrower for a given device. Beyond the rated maximum the resistance shifts permanently, and unlike a PRT it does not recover.
• Self-heating. The high resistance limits the current, which helps, but small glass beads have very little thermal mass and dissipate poorly in still air. Check it by halving the current, exactly as for a PRT.
• Interchangeability. Devices sold as interchangeable to ± 0.1 °C are, within their stated band; outside it they are not, and treating them as drop-in replacements without recalibration is a common error.
• Ageing. Precision thermistors are aged by the manufacturer to stabilise them. Cheap ones are not, and drift for their first year.

Calibrating one

Three points minimum, spanning the range of use, fitted with Steinhart–Hart, and the coefficients recorded with the device rather than in a spreadsheet somewhere. Verify the fit by measuring at a fourth point that was not used in it — a fit that reproduces its own inputs perfectly and misses an independent point by 0.05 °C is telling you something the residuals will not.

A thermistor's calibration history is the thing to look at before trusting one for precision work. The good ones hold millikelvins for years; the poor ones move, and the specification sheet will not tell you which you bought.