Pt100 tolerance classes explained (IEC 60751)

IEC 60751 defines a single resistance-versus-temperature curve that every platinum resistance thermometer is meant to follow, then sets out how far an individual sensor is permitted to deviate from it. The tolerance classes are those permitted deviations. They are a property of the manufactured part, not of a measurement you make with it.

The four classes

Each is the fixed term plus a fraction of the absolute temperature, so the band is tightest at the ice point and widens in both directions.

The expressions are ±(0.10 + 0.0017|t|) for AA, ±(0.15 + 0.002|t|) for A, ±(0.30 + 0.005|t|) for B and ±(0.60 + 0.01|t|) for C. Class C is exactly twice class B at every temperature, because both terms double. Class B is not twice class A: they share a doubled fixed term but the proportional terms are 0.005 against 0.002, so B pulls further ahead as the temperature rises, from twice A at 0 °C to about 2.4 times A at 450 °C.

Ranges matter as much as the numbers

• Class AA: −50 to +250 °C for wire-wound elements, extended to +600 °C for thin film.
• Class A: −100 to +450 °C.
• Class B and C: −196 to +600 °C.

This catches people out in cold-chain work. A ULT freezer running at −80 °C is outside the defined range of both class AA and class A, so a supplier quoting "class A" for a −80 °C application is quoting a figure the standard does not define there. Class B is the only class defined that low, at ±0.70 °C for −80 °C.

What the tolerance is worth in ohms

Converting through the sensitivity makes the size of these bands concrete. At 0 °C a Pt100 changes 0.3908 Ω/°C, so class A's ±0.15 °C is ±0.059 Ω. At 300 °C the sensitivity has fallen to 0.3562 Ω/°C and class A's ±0.75 °C is ±0.267 Ω. If your readout resolves 0.01 Ω, it resolves roughly a fortieth of the class A band at the ice point: the sensor, not the instrument, is the limit.

Tolerance is not uncertainty

This is the point the classes are most often used to dodge. A class A sensor is guaranteed to lie somewhere inside ±0.15 °C of the standard curve at 0 °C. It does not tell you where inside, it says nothing about the instrument reading it, the lead resistance, self-heating or the immersion, and it says nothing about drift after a year in service.

If you treat the class as an uncertainty you are making a rectangular assumption over the band, which gives u = 0.15/√3 = 0.087 °C at 0 °C for class A, before any other component. That is defensible as a bound for an uncalibrated sensor, and it is usually much worse than the same sensor's uncertainty once calibrated, which is the whole argument for calibrating.

A certificate supersedes the class entirely. Once a sensor has been calibrated, use the certificate's correction and uncertainty. Quoting both a class and a certificate for the same sensor, and then choosing whichever is more convenient, is the wrong way round.