Skip to content

Sensitivity

Sensitivity describes how strongly one quantity changes when another quantity changes. For RTDs, two forms are useful:

  • resistance sensitivity, dR/dT, in ohms per degree Celsius (Ω/°C);
  • temperature sensitivity, dT/dR, in degrees Celsius per ohm (°C/Ω).

They are local values: the sensitivity can change with temperature because an RTD characteristic is not necessarily perfectly linear.

Resistance sensitivity

from rtd_sensor import pt100

d_r_d_t = pt100.resistance_sensitivity_ohms_per_celsius(100.0)
print(d_r_d_t)

This answers: near 100 °C, how many ohms does this Pt100 model change for a one-degree change in temperature?

Temperature sensitivity

from rtd_sensor import pt100

d_t_d_r = pt100.temperature_sensitivity_celsius_per_ohm(100.0)
print(d_t_d_r)

This answers the inverse question: near 100 °C, how much temperature change does one ohm represent?

Why this matters

Suppose two measurement systems each have a resistance uncertainty of 0.01 Ω. The corresponding temperature contribution depends on the RTD model's local dT/dR. That is why rtd-sensor uses exact model sensitivity in its resistance uncertainty propagation.

Sensitivity also helps explain why a Pt1000 gives a larger absolute resistance change per degree than a Pt100 while sharing the same normalized platinum characteristic.

Compare Pt100 and Pt1000

from rtd_sensor import pt100, pt1000

pt100_sensitivity = pt100.resistance_sensitivity_ohms_per_celsius(25.0)
pt1000_sensitivity = pt1000.resistance_sensitivity_ohms_per_celsius(25.0)

Because the Pt1000 has ten times the reference resistance, its absolute resistance sensitivity is also roughly ten times larger for the same normalized characteristic.

How rtd-sensor calculates it

Built-in platinum and custom Callendar–Van Dusen models use the analytical derivative of their active characteristic. Polynomial and piecewise polynomial models differentiate their supplied polynomials analytically. Tabulated models use the slope of the active linear-interpolation interval.

The package does not estimate these values by taking arbitrary finite differences around the requested temperature.

Common mistake: treating sensitivity as constant

A quick rule of thumb may use one nominal sensitivity value, but professional uncertainty or resolution work should use sensitivity at the relevant operating temperature when the model provides it.