0
\$\begingroup\$

How would you design a replaceable temperature sensor probe such that end users do not have to send the entire unit (processor plus probe) away to be re-calibrated? So that you could just send them a new calibrated probe that they could plug in and it would "just work" without having to program in new calibration parameters (e.g. offsets, polynomial coefficients, or what have you)?

\$\endgroup\$
9
  • 3
    \$\begingroup\$ Have the a ROM onboard the probe. The problem is...what's going to calibrate the electronics reading the probe? The probe isn't the only thing that can go out of calibration. But maybe you don't care about that. \$\endgroup\$
    – DKNguyen
    Dec 31, 2022 at 2:59
  • 2
    \$\begingroup\$ include the analog section with the sensor ... have only digital signals at the interface \$\endgroup\$
    – jsotola
    Dec 31, 2022 at 3:04
  • 1
    \$\begingroup\$ You are right @jsotola I'll withdraw the question \$\endgroup\$
    – vicatcu
    Dec 31, 2022 at 3:09
  • 1
    \$\begingroup\$ Using an interchangeable sensor is the most popular way in industry. Eg. RTD, thermocouple, thermistor. \$\endgroup\$ Dec 31, 2022 at 3:25
  • 2
    \$\begingroup\$ The first and most important thing is what are the tolerances at what temperature(s)? That has a big impact on the answers to your question. A tolerance of +-0.0001 degree is much different then one that is +- 1.0 degree. \$\endgroup\$
    – Gil
    Dec 31, 2022 at 3:34

2 Answers 2

2
\$\begingroup\$

Interchangeable probes are the most common way in industry to do this. Only if the sensor is very unusual in some way and it's impractical to make them interchangeable would one want to accept individually calibrated sensors. Sometimes a simple zero shift (single point) user calibration is provided which allows a small correction at a single temperature.

For example, a Class AA RTD has an interchangeability of 0.1°C + T*0.0017 where T is the temperature in °C. So at 40°C the sensor error will contribute less than +/-0.17°C. That's more than good enough for most purposes. The sensors are manufactured to be that accurate (via laser trimming and control of the metal and substrate). Thermocouples have more like 1° + 0.4% using special limits of error wire (though the error at the cold junction temperature will be zero, from thermodynamics). Those errors refer to the matching of the actual sensor to the (nonlinear) curves for that sensor type.

On the other hand, some sensors for extreme accuracy and extreme temperatures are supplied with individual polynomial coefficients matching serial numbers and those values have to be loaded into the measurement system. That's very, very expensive (especially the testing of the sensors) and potentially error prone so it's best avoided if possible.

\$\endgroup\$
0
\$\begingroup\$

The most foolproof way is to build the ADC into the probe, along with a microcontroller that contains the calibration.

The probe transmits the data digitally to the processor after applying the calibration.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.