> How do they measure temperature, though? It seems to me that the
> temperature would govern the wavelength directly, not the intensity of
the
> IR radiation. The intensity is related to the quantity of radiant heat,
> which would be a function of the surface that it was being radiated from,
> etc. So how does one get temperature from measuring the magnitude of the
> voltage generated on the sensor? It seems to me that this is like saying
> that you can sense color by measuring the amount of voltage generated by
a
> solar cell.

The pyroelectric sensor doesn't actually measure the wavelength. The
voltage it produces is proportional to the amount of (IR) light that falls
on it. It just happens that its response is fairly flat between 1 and 20
microns. (A silicon photodiode response is < 1 micron)
The radiation isn't a single wavelength for a particular temperature, but a
broad emission with a peak value that moves acording to the temperature.
The radiant intensity also increases and is more pronounced than the shift
in peak wavelength. If you heat a piece of metal, it is brighter when it is
white hot, than it was when it was red hot.

  |____________________
  |    __________________\ 400
  |   /     ________________\ 300
  | /      /   _______________\ 200
  |/___/__/________________\___
 A     B  C     D    E              F

In the graph above, the y axis is radiant energy, the x axis is wavelength.
Each of the lines represents the radiation at a particular temperature
(200K, 300K, 400K). These should really look more like skewed, inverted
parabolas, so please think of them as curves. At 200K, the light emitted is
between wavelengths C and F. So the midpoint (peak) is at about E.
At 400K, the intensity is higher and the emission is between A and F, with
a midpoint at D. So as the temperature rises, the peak intensity gets
higher, the bandwidth increases and the peak moves to the left.
This is Planck's Radiation Law. I'm afraid (I think) I know enough to
understand it, but probably not enough to explain it very clearly
If the sensor has a flat response, the bandwidth and peak movement have no
effect and the temperature can be derived from the intensity.

You do have to compensate for the surface that is doing the emitting. The
curves of temperature vs. peak wavelength and radiated energy are for a
theoretical "black body". To compensate for the material, a fiddle factor
called the "emissivity" must be applied.

Steve.