> I'd like to learn more about your power distribution technique for the RGB
> peaks, and to try to implement that on SpectralWorkbench.org. How do you
> know what it "should" look like -- i.e. a true reference spectrum of a
> known source like a halogen bulb? It's a chicken and egg problem but we
> could borrow a commercial spectrometer to get us started?
An Incandescent light bulb approximates ideal black body radiator and you
can generate a black body spectrum algorithmically with wanted kelvin as
input. This will give you a smooth curve that your app should be able to
reproduce - if not then calibration/gain filter is needed. The hard part is
knowing exactly what kelvin the bulb actually have. But I guess you could
iterate/backtrack like Robertson's method to find best fit and then you
make your gain filter for each band to make the power distrubution from
your camera look exactly as the best fit. As a side effect your app can
show what kelvin the bulb have.
If several different devices with your invention show radically different
Kelvin values you must be able to input the exact kelvin the bulb have. But
i *guess* they should be in the same ball park at least.
> The IR mod has only been done on the webcam "desktop" version of the
> spectrometer. I made a halfhearted attempt to remove the filter from an
> Android phone and will try again, but i don't imagine most people will be
> doing that. So I think the mobile version will mostly be in the RGB 400-700
> nanometer range, no infrared. It will still be useful for many things!