Choosing the right model

The best model is the one that represents the documented characteristic or calibration evidence you actually have. Do not choose a model family merely because its class name sounds familiar.

Use a built-in model when

Use a built-in such as pt100, pt500, pt1000, ni1000, ni1000_tk5000, or ni120 when the physical sensor is intended to follow that verified nominal characteristic and you do not need an individually characterized model.

See Built-in RTDs.

Use IEC60751RTDModel when

You have an IEC 60751 PT-385 platinum sensor with a characterized resistance at 0 °C, often written R0, but still want to use the standard PT-385 curve shape.

Example: a calibration record says the probe's R0 is 100.017 Ω.

Characterized IEC 60751 models

Use CallendarVanDusenRTDModel when

A calibration certificate, manufacturer, or technical source gives you an explicit Callendar–Van Dusen coefficient set and valid temperature range.

Callendar–Van Dusen models

Use PolynomialRTDModel when

The authoritative source publishes one polynomial over the complete declared range, or when you intentionally fit a global polynomial from calibration data.

Polynomial models

Use PiecewisePolynomialRTDModel when

The source publishes different polynomial coefficients over different temperature intervals. Keeping those intervals preserves the source model more faithfully than forcing everything into one global polynomial.

Piecewise polynomial models

Use TabulatedRTDModel when

The authoritative source is a resistance-temperature table and linear interpolation between the published points is the behavior you want to retain.

Tabulated models

Use calibration fitting when

You have your own measured (temperature, resistance) observations and want to fit a validated polynomial model while retaining evidence such as residuals, RMS error, fit range, and conditioning diagnostics.

Calibration fitting

A practical decision rule

Prefer the representation closest to the authoritative source:

verified nominal characteristic → built-in model
standard PT-385 + characterized R0 → IEC60751RTDModel
published CVD coefficients → CallendarVanDusenRTDModel
one published polynomial → PolynomialRTDModel
published interval polynomials → PiecewisePolynomialRTDModel
published R/T table → TabulatedRTDModel
measured calibration observations → fit_polynomial()

Do not identify an RTD by resistance alone

A nominal 1000 Ω sensor could be Pt1000, Ni1000 6180, Ni1000 TK5000, or another characteristic entirely. The resistance at 0 °C is only one part of model identity.