Detecting very small temperature changes with a thermal camera

[AndyC_772]

Hi all,

I have a board which is battery powered and which is supposed to draw an absolute minimum of current when it's in its standby mode. It's drawing more than expected, though, and I'd like to know where it's all going. The standby current is in the order of 0.5mA, and it's a complex board that would take a long time to tear down and measure conventionally.

When the board is fully active, a TIC would easily be able to highlight the components which are dissipating significant power, but the standby current alone is not enough to cause easily measurable heating. Assuming, for example, a 10k resistor with 100uA flowing through it, and 250K/W thermal resistance to ambient, the temperature rise works out to 25mK... likely to be buried in the noise, but that's not to say it couldn't be detected. The actual dissipation is likely to be distributed more widely than this, so the temperature differences are likely to be small indeed.

The experiment I had in mind was something along the lines of:

- set up the board inside a sealed, well insulated box
- point a TIC at the area which remains powered up in standby, and secure it rigidly in place
- switch on, put the board it in standby, and leave for a couple of hours for the temperature to stabilise
- capture a video stream from the TIC, to record a large number of frames of the board all precisely aligned
- switch off power
- capture a second video stream from the TIC

Theoretically I should then be able to stack and average the frames taken before / after switching off the supply, to average out the random noise and allow smaller temperature differences to be resolved. The parts which cool down in between the two videos are the ones responsible for the unwanted dissipation, of course.

Has anyone ever used this technique with thermal images, and more importantly, is there any software (preferably free, of course!) out there which can extract the individual frames from a video sequence and average them without losing my 'signal' to a rounding error?

[dfmischler]

I would think you would be better off trying to keep the temperature as low as possible for standby operation so that the components that are radiating the most heat (i.e. consuming the most power) will stand out from the background on your thermal camera.  Am I wrong?

[Rerouter]

another thing you can do to increase the contrast is to spray it with freeze spray, the items generating heat will more rapidly return to ambient,

[madshaman]

Idea:

Keep the camera idea but instead spray the board with isopropanol or something similar.  The alcohol will evaporate first from the hottest components.

WARNING, make sure there are no spark sources and/or submerge the thing in CO2 (easily produced) or nitrogen first.

[Dago]

This is why it's a good practice to put 0 ohm jumpers on the power traces for different "blocks" of the circuit, you can always solder them off and measure which area of the circuit is consuming the power. Also handy when you're testing step-up/down modules that are feeding the rest of the circuit, so you can make sure you won't break anything if the voltage is too high for example.

[madshaman]

P.S. Obviously no need for a TIC, just a normal camera angled to highlight the liquid

[AndreasF]

In principle this should work. Almost the same principle is often used for digital astrophotography (e.g. http://keithwiley.com/astroPhotography/imageStacking.shtml).

[AndyC_772]

I would think you would be better off trying to keep the temperature as low as possible for standby operation so that the components that are radiating the most heat (i.e. consuming the most power) will stand out from the background on your thermal camera.  Am I wrong?

The temperature difference between a device and its surroundings will stabilise at a more or less constant level regardless of the actual temperatures involved. If the board's at 20 deg C and the part is at 20.1 deg C, then cooling the board to (say) 15 deg C would mean the part ends up at 15.1 deg C. There's also the practical difficulty of cooling the board and keeping it at a very constant, uniform temperature.

It may be that a TIC can more clearly perceive small differences at lower temperatures, but there wouldn't actually be a larger difference to measure in absolute terms. So, I think the best bet is to seal the board into a box with very still air (at whatever the ambient temperature in the lab ends up being), give it plenty of time to stabilise, then cut the power and observe what happens.

another thing you can do to increase the contrast is to spray it with freeze spray, the items generating heat will more rapidly return to ambient,

Nice idea - I've seen the effect - but in this case I think the total amount of power involved is too small. The temperature differences which exist across the board as a result of being blasted with freezer spray would absolutely swamp the tiny amount of heating I'm expecting to see.

Keep the camera idea but instead spray the board with isopropanol or something similar.  The alcohol will evaporate first from the hottest components.

I'm looking for about one surplus milliwatt... I doubt it'll evaporate anything much.

This is why it's a good practice to put 0 ohm jumpers on the power traces for different "blocks" of the circuit, you can always solder them off and measure which area of the circuit is consuming the power. Also handy when you're testing step-up/down modules that are feeding the rest of the circuit, so you can make sure you won't break anything if the voltage is too high for example.

Agreed; most of the power domains on the board are switched under software control anyway.

At the opposite end of the scale in power terms, I did once learn a very valuable lesson about 0R links, namely: they're not 0R, or anywhere close. If you have a CPU core drawing several amps at, say, 1.2V, you can't put a 0R link in the way... I did, once, and the core voltage dropped to about 1.05V. Amazingly it kept working just fine, but a thick, solid copper wire link soldered in place of the 0R resistor made a huge difference to the voltage drop.

In principle this should work. Almost the same principle is often used for digital astrophotography (e.g. http://keithwiley.com/astroPhotography/imageStacking.shtml).

That was my first thought too, and thanks for the link - it confirms that software intended for astrophotography should, in theory, be able to pick out the subtle variations between frames that I'm looking for.

I don't suppose there's any particular tool which you can recommend for this purpose...?

0.5mA isn't that much, and could just be due to either the battery self-draining or the resistance of the switch circuitry. It's possible it could be seen with a TIC, but unlikely.

It's definitely a real current draw, I'm supplying the board from my HP 6632B which has a usefully accurate meter built in. I expect it to be a challenge to see the effect with a TIC, but that's half the fun!

[AndreasF]

...
I don't suppose there's any particular tool which you can recommend for this purpose...?
...

I'm afraid not. I had read a bit about these techniques some time ago, but never actually done them myself, but as far as I remember there were plenty of freely available options to be found. DeepSkyStacker looks pretty good (also see here).

One issue might be that your camera probably doesn't allow you to read the raw imaging sensor data, so the tiny signal may already get obliterated in some image processing algorithm inside the camera.