How linear are the output of microbolometers?

[Ben321]

Once their own thermal signal has been compensated for by dark-frame subtraction (capture an image of a flat object kept at a constant 0K in an advanced laboratory, after allowing the thermal camera to warm up for several minutes to a constant operating temperature), so that a signal output of 0 means the object is at absolute 0 temperature how linear is its response? Obviously you need to apply a curve to convert linear radient energy readings to temperature, but that curve itself must have an input that is linear with regard to radient energy. So with an average microbolometer array (whether VOx, or amorphous silicon), once the dark frame has been subtracted, how linear is the actual signal? Is it quite linear with respect to radient energy? Or is it closer to linear with respect to temperature? Or is it completely non-linear (following some advanced transfer function regarding radiant energy input to signal output)?

[Vipitis]

constant 0K

not possible.

this is really complicated. But you may find some hints here: https://www.eevblog.com/forum/thermal-imaging/yet-another-cheap-thermal-imager-incoming/msg779709/#msg779709

[IwuzBornanerd]

constant 0K

not possible.

this is really complicated. But you may find some hints here: https://www.eevblog.com/forum/thermal-imaging/yet-another-cheap-thermal-imager-incoming/msg779709/#msg779709

Cool!  I don't recall seeing that post before.  I find it very interesting that his spectral radiance curve there looks very similar to the curve in Seek's frame with ID of 9, which I use for converting sensor data to temperature.    Maybe someday I'll finally get things tweaked into a good calibration.

[Bill W]

Forget the 0°K bit, but relative pixel outputs are fairly linear in radiant energy (integral of Planck curve over the active spectrum).

Say the radiant energy of 20, 50 and 200°C are 5, 15 and 305 nominal units.

Then if the pixel signal at 20°C and 50°C are 200 and 300 ADC count then the signal at 200°C will be about 3200 ADC count.
(ie 100 ADC counts is 10 units of radiance, 3000 ADC counts for the additional 300 units of radiance going from 20 to 200°C)

Real life is not quite so kind as pixels heat up, lenses change transmission etc so you need calibration for accuracy, and tighter more extensive calibration for thermographic accuracy.

Bill

[Ben321]

constant 0K

not possible.

this is really complicated. But you may find some hints here: https://www.eevblog.com/forum/thermal-imaging/yet-another-cheap-thermal-imager-incoming/msg779709/#msg779709

Of course EXACTLY 0K is not possible. But I meant approximately 0K, which is possible. I've heard of labs where they have gotten to within 1-millionth of a degree of 0K. This would be close enough for what I'm thinking of. In fact, I believe it would be a good reference at the factory, for initial 0K calibration of every microbolometer manufactured.

[bap2703]

There is no point in using 0K. The sensor works relatively to its own temperature. Think of a strain gauge measuring the length relative to its calibration length thanks to a Wheatstone bridge.