I'm working with an electronic temperature logger that is being affected by heat generated internally.

How does one come up with a calibration equation to calculate a more accurate reading of ambient temperature based on what the temperature sensor reads, taking into account its own power consumption?


After a few hours and in equilibrium, the sensor reports values that are actually 1 degree Celsius higher than the ambient room temperature (22C) measured by a calibrated device. The sensor is accurate to 0.1 degree C at reporting the temperature of the device itself (which due to heat generated by the electronics has gotten warmer)

The device consumes ~0.1 watts of power, weighs about 200g and has an average specific heat capacity of 1.0 j/g (weighted mix of glass, abs, fr-4, copper). Dimensions are 1"x 3"x 4".

What I've got so far is this heating calculation: 200g * 1c * 1.0j/g / 0.1w / 60s = ~33 minutes to heat up 1 degree.

I'm assuming what we need is to figure out Sensor value - Heat-generated + Heat-dissipated to arrive at actual temperature. Which will require measure the K in newton's law? then what?

I'd really appreciate you help here.

  • $\begingroup$ Are you sure that the operating heat is not affecting the immediate environment to that degree? Simulation suggests that if some of the sensor packages we installed in the Double Chooz far detector were left running full time they could affect the immediate environment to that level. (Of course that is in a very static environment, and packages that combine several instruments.) Accordingly we only power up the most power-hungry parts of the boards for a few seconds every five minutes or so. $\endgroup$ – dmckee Oct 18 '12 at 18:09
  • $\begingroup$ I make a point of measuring ambient temp a few cm away from the device. knowing the heat generation rate (measure and computed) and measuring the dead device cooling rate, I'd like to be able to infer what the ambient temperature was if the sensor reports a given T. $\endgroup$ – MandoMando Oct 18 '12 at 18:38

Typically you would attempt to measure rather than calculate the effect.

Perhaps by having a second, calibrated device with a long probe that provides an independent measurement of the temperature. You do this in situ if possible or in some reasonable test stand (which might be as simple as disposable cooler filled with you working fluid).

The only real alternative is to read the data sheet; either for the whole device (if is a off-the-shelf instrument), or for the particular chip (if it is something that you manufactured to spec).

As a desperate fall-back position you might be able to find a rule-of-thumb for devices in the same class, but those are unlikely to be centrally tabulated. Ask around is my best suggestion.

  • $\begingroup$ It would have been nice to be able to calculate even rough calibration. In theory, it's just jules-in minus jules-dissipated. I'm just not able to work out the jules-dissipated part :) I'd hope the measurement route remains consistent given that anything anomaly (say wind) that would affect the measured data, would also affect the calculated estimate. $\endgroup$ – MandoMando Oct 23 '12 at 15:19

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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