Skip to content

The hardware/acquisition boundary

rtd-sensor starts after the acquisition layer has produced the best available estimate of the RTD sensing element's resistance in ohms.

physical RTD
    ↓
acquisition hardware + corrections
    ↓
compensated resistance in ohms
    ↓
rtd-sensor model
    ↓
temperature in °C

What belongs in the acquisition layer

Depending on the hardware, that may include:

  • ADC or converter register handling;
  • SPI, I²C, GPIO, or another bus;
  • excitation-current behavior;
  • reference-resistor calculations;
  • two-, three-, or four-wire topology handling;
  • lead-wire compensation;
  • converter fault detection; and
  • hardware-specific calibration.

For example, a MAX31865 integration should first use the converter's information and circuit configuration to obtain an RTD resistance estimate. That resistance can then be passed to rtd-sensor.

What rtd-sensor owns

rtd-sensor owns the mathematical RTD model layer:

  • resistance ↔ temperature conversion;
  • RTD model range and validation;
  • model sensitivity;
  • custom/calibrated model behavior;
  • RTD-specific uncertainty propagation;
  • model identity and conformance semantics.

Why keep the boundary clean?

A Pt100 model should behave the same whether the resistance came from an industrial bridge, a MAX31865, a laboratory DMM, or a simulator. Keeping hardware out of the model layer makes that possible and allows independent testing of each responsibility.

Hardware faults versus RTD model errors

An open circuit, SPI failure, stale ADC sample, or converter fault is an acquisition problem. A resistance outside the model's valid range is an RTD model problem. Applications may report both, but the package does not collapse them into one exception vocabulary.