Bolometer Head Completed

No doubt regular readers are sick of hearing about the bolometer work, but the base project is drawing to a conclusion (for now at least): I've finally built something resembling a finalised sensor head for it.

The new Bolometer Sensor Head

An old Pentium 1 CPU heatsink was extracted from a dead server carcass, then drilled and tapped for M3 hardware to accept the sensor assembly and its draft exclusion shield. This more mechanically robust assembly helped the short-term baseline noise enormously, but the smaller thermal mass of the CPU heatsink compared to the previous rather large unit has made the long-term thermal drift more obvious.

I've also confirmed that ambient temperature is indeed responsible for the sawtooth changes in baseline bias power. This was strongly suspected and kinda obvious when you consider how the device works, but there is nothing like hard data to support a hypothesis. Temperature resolution is not very good (I am using the raw signal from an LM335 temperature sensor which has a sensitivity of only 10 mV/K), but the correlation is obvious.

Bolometer Bias Power vrs Ambient Air Temperature

Here is a longer capture of the baseline drift over several nights:

Bolometer Bias Power over Several Nights

The effect of solar forcing on the air condition cycles is quite obvious.

Also interesting, convectional changes inside the sensor unit are now observable with the robust sensor head. Here is some data collected with the head in different orientations compared to the local gravity field. The two small dips after the initial baseline rise are where I applied the red and green laser pointer beams, in fact this is how I noticed the effect, I turned the sensor head on its side so I could more easily aim at the sensor plate through the hole in the draft exclusion box.

Bolometer Bias Power Variation with Sensor Head Attitude

The sensor is a small brass plate with SMDs sticking off the back of it. Naturally when the "fin" of the brass plate is vertical it is cooled by convection more efficiently. This can clearly be seen in the data as larger bias power consumption to hold the same sensor plate temperature. Reduction of this effect would require a redesign of the sensor housing to minimise the opportunity for convectional cooling. I might try a sensor buried in A/B foam, but I wonder about its time constant. Running the sensor at a smaller bias (so its temperature differential with respect to the ambient is smaller) would reduce the effect, but in practice all RF power measurements are performed fairly quickly and are always "differential" in nature so the effect isn't of huge consequence now that I am aware of it.


I need to add a shutter to the hole in the draft exclusion cover. This would allow rudimentary "Dicke Switching" of radiant energy (say like Sun light)... I'd like to try measuring solar flux directly with the bolometer (inspired by Langley's original work), and perhaps with the help of a stepper-driven Heliostat a full day's data showing cloud and atmospheric extinction effects.

Another use for this basic technology occurred to me while I was playing with the sensor head convection effect. It should be practical to resolve differences in an gas flow's thermal conductivity. This would mean a similar device could operate as a universal chromatograph detector. It would not be very sensitive, but it would respond to just about everything unlike other detectors (like FIDs). Building my own gas chromatograph now seems fairly do-able... Time to find something suitable for the stationary phase, I figure air can be used as an initial mobile phase and a simple Peltier oven can regulate column temperature.

Leave a comment on this article.

Parent article: RF and Optical Bolometer.