I need to measure the performance of some sensors with a strong temperature dependency. Later on, I will do some calibration for the entire temperature range, but for the moment I just want to do a test over 24 hours at a constant temperature.

So I have a styrofoam box and inside a heating element attached to a power supply. I set the power supply to a constant voltage, so that after a few hours it reaches a steady state about 55 degrees C.

Now, the problem is that even with the insulation and heating, there is an effect of the ambient temperature over the 24-hour cycle. It's only about 0.7 degrees C, but that's enough to screw up the results.

Of course, I could set up a PID controller for the temperature, but that introduces a few problems. First, it would be a pain to set up and tune, and second, I would be afraid of introducing oscillations which I was hoping to avoid by using a constant voltage power supply.

The other idea would be to try and improve the insulation or maybe introduce some thermal mass like rocks or those freezer packs with water inside. That would avoid the oscillations of an active PID controller, but would it be enough to suppress the daily temperature fluctuation?

It seems like a really stubborn problem for something so basic. Is there something I'm not seeing here? Is there some easy setup or principle I could exploit to get a decently constant temperature? I don't have to hit any specific target temperature range, just something that would settle at some temperature and not fluctuate over 24 hours.


"heater with PID control" sounds like an unreasonable engineering effort, given that we have devices designed for this exact purpose: thermostats.

But the classical solution is to keep the temperature stable at the melting temperature of ice. This is still going to be a quite stable 273.15K, independent of the variations in atmospheric pressure. Rocks don't have quite the same heat capacity, plus the exact temperature would be less well known.

How much ice you'll need depends on the heat leak into your container, but this leak will be reasonable constant over time. You'll have about 20K difference, so the +/-0.7K won't affect the heat leak too much (and remember, this only affects the melting speed of the ice, not the temperature of the water/ice mix)

Engineering wise, you'll want to have the sensor affixed to a metal heat conductor, whose other end is submerged in an ice/water mix. You need sufficient water to make a good contact with the heat conductor. Don't try to optimize the "ice fraction". The ice might actually be too cold initially, but that's not a big problem. The heat conductor will mostly be in contact with the liquid phase, and that will be 273K. A 24 hours setup should be easy to achieve without needing excessive isolation.

  • $\begingroup$ One of the problems I've had has been moisture and condensation. I suppose I could also stick the circuit board in a plastic bag or something to keep the moist air out. I think if I go this route then I best get a metal box that can be partially submerged in the ice water because the air circulation near the surface could affect the temperature of the device, or give me a non-homogonous temperature. Unfortunately, this will be a bit lower than the usual operating temperature of 20-40 degrees C. But I suppose there aren't many phase changes that are as safe and as easy as water and ice. $\endgroup$ – W4uoe9A Dec 13 '17 at 17:39

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.