One thing to bear in mind with these kind of DIY spectrometers is that typically you only perform a wavelength calibration (e.g. you can transform from pixel nu

One thing to bear in mind with these kind of DIY spectrometers is that typically you only perform a wavelength calibration (e.g. you can transform from pixel number on the detector to wavelength). That's reasonably easy to do with a CFL bulb and some known spectral peaks. What's much more challenging is accurately measuring colours, or the relative intensity of certain spectral features. The problem is that your image sensor has its own spectral sensitivity which you usually don't know. This also means that you can't really quantitatively compare spectra between spectrometers because the sensors will be different. Most of this time this isn't much of an issue, but suppose you want to measure the spectrum of some tungsten halogen lights around your house (or the sun). You won't see the characteristic blackbody curve. You'll see the spectral response of your detector with some absorption lines on it. It shouldn't matter much for e.g. differentiating between two species because you'll have the same instrumental reponse for each. A fairly simple way to do this is to use a tungsten bulb with a known colour temperature. You can use the blackbody equation to generate the expected spectrum and work out the transform between your spectrometer and the calibration lamp. If you want to do spectroradiometry (measure the absolute intensity per wavelength), you need to know the power spectral distribution of the lamp too. In the real world you'd need to buy a NIST-traceable (or equivalent) calibrated lamp.