That’s the main issue, as it will heavily alter the spectrum based on geographic position, time of the year, weather condition etc…
I’ll meet with koef3 soon, and buy a halogen light with a spectrum he measured from him to use as a calibration source, as well as 2 or so LEDs for double-checking after.
Central europe, so no, it will never be perfectly overhead.
Are you in astrophysics? I had a lecture about it at uni, was pretty fun, but I ended up going into a different field eventually. Still led me to work with spectrometers, but sadly I have no more access to the labs ^^
A known, calibrated lamp is probably a sensible way to go if you have a really good stable power supply to drive it. I would always be worrying though if it was the sensor, the lamp or the power supply that was drifting whenever I saw a change. Over long time periods, I do not know if the lamp bulb or the spectrometer is likely to be more stable. Maybe by using another external photometer (need not be a spectrometer) you can convince yourself that the lamp itself either is or is not stable.
You mentioned starting with halogen bulbs. I wondered if a really simple old-style incandescent would be a better black body. (That is if you can even find one! Maybe they still sell “appliance bulbs” meant for inside an oven that are simple incandescents?) I had a very quick look using my XRite ColorMunki. I had two simple 60W incandescents and a 40W halogen that claimed to be “60W equivalent” output. They are pretty much the same and all gave black body fits in the range 2500-2600K. But then the ColorMunki is limited in wavelength range to that part of the spectrum that a decent black body in your ThorLabs plot, so this is probably not telling me much new or useful. Here are my three different bulbs. If nothing else it illustrates the variation to be expected between bulbs.
For the solar spectrum, there are very good models available to simulate the solar spectrum at whatever Sun position and altitude you like. For example SMARTS from the US NREL could be worth a try. https://nrel.gov/grid/solar-resource/smarts.html
Ultimately though, this is just a really difficult job to get right. Absolute spectrophotometry is hard work, even with lab grade equipment. We still have some 1-2% bumps and wiggles in one of the instruments I work on that I have never been able to fully calibrate out.
From the plots you show, this CMOS detector clearly has an ugly spectral response. Calibrating that is going to be a challenge. I confess that I have only ever done this sort of thing on far more expensive sCMOS detectors. I found a really nice plot showing that, here:
It is clear that some have very complicated QE curves, like you are seeing. By comparison, something like the back illuminated pco.edge 4.2BI has a beautifully smooth QE curve. I have used that detector and calibrating it was quite easy!
Anyway, I am glad you are putting in the effort on this. I am tempted to buy one of these devices myself, so am interested in how you get on.
Hello, I want to know what is the “Bayer filter”, actually I can check it out from wikipedia.
you mean the hobbyist used the same sensor as those used in cameras(CCD & CMOS)? he just removed the Bayer layer to make it more sensible to the lights? then why the cameras need the Bayer layers? actually I know there is a startup in Europe that is developing camera sensors which can sense the whole visible range of spectrum for EACH pixel, that is horrible. I think it’s intended to achieve better quality of photographs.
the website is https://www.spectricity.com/
the hobbyist used a camera level sensor, so the “Little Garden” device’s main cost is for the sensor part right?
The Bayer filter is basically the difference between a monochrome and a color CMOS - a color-filter layer that filters all but the wavelength range for one color from the incoming light per subpixel.
By removing it, all 3 sub pixels are exposed to the entire spectrum, instead of just red/green/blue.