– thanks very much for these links! I actually found out in the mean time that I already had digitized the Kodachrome curves:
@PM490: your internet find is amazing - very interesting! It’s the first data on Agfachrome film I get my hands on.
No, not at all. That is quite a general issue and would be present with any sensor, including any monochrome one. The thing is quite simple: with a narrowband illumination setup, you are sampling only in very narrow ranges - ignoring all other available data. That is, with a narrowband setup, you are blinding your sensor in certain bands were not enough illumination is available. That’s actually more a feature than a drawback with multi-spectral cameras. They can see signal variations normal cameras cannot see. But a three-LED setup can hardly be called a multi-spectral setup. Again, I think that narrowband scanning works because the dye spectra are well-behaved and rather broad. Also, the LEDs one is using might have a larger bandwidth than specified (I worked in the examples above with 20 nm bandwidth), broadening the curves. The operating temperature of the LEDs in a long-running scanning task might also help with a little bit of broadening. If that happens, a narrowband 3 LED setup comes closer to a broadband illumination - albeit a very funny one.
A small hint about the challenges of narrowband scanning might come also from the fact that some commerical film scanners have two spectral-wise differently placed red LEDs - allowing the operator to adapt the illumination towards the film to be scanned. And the Swiss people have also drifted since the publication of the DIASTOR paper towards a multi-spectral camera setup, covering the visual spectrum with a whole bank of filters/LEDs. Something like that is great for archival scanning, but way to much effort for digitizing S8 consumer footage.
When I tried out the narrowband illumination setup, the HQ camera was not available and there was no way of saving raw files with these cameras. I also tried out an industrial USB3 camera for a while. I do have HQ cameras were the IR-blockfilter is removed and HQ cameras were the stock IR-blockfilter is replaced by a professional one. While all this improves gain/noise issues, it did not help with the colors in the narrowband setup. The major improvement I encountered was switching to broadband illumination and being able to fine-tune a sensor (the HQ one) with a color science of my own.
Given, we talked about the noise issues of the HQ camera as well as the challenging dynamic range of classical S8-footage which is hard to capture with a 12 bit camera. There are better sensor options available. The intial Kinograph used a DSLR. Others have used anything from USB-webcameras over Iphones to machine vision cameras. That’s a personal choice, I suppose.
@cpixip In the context of the above charts in your postings, the kodachrome are depicting spectral absorption, unlike the above which represent passthrough response.
That is the reason I inverted the datasheet curves in the above representation.
That government document is wild (and a bit of a treasure trove). Why on earth would they have ever needed a Top Secret clearance over film dyes?!
In any event, good find!
Well, the tags “Keyhole” and “Itek” in the document give a hint. The early birds of the Keyhole satellites used conventional film material for the purpose of taking imagery. Which was ejected and caught in mid-air by airplanes. Obviously, there was an interest in using the best film available for the purpose. And obviously, you would not want to share this information with your adversary. Funny that Agfachrome was taken into account - however, in these times, Agfa did produce a lot of commercially used film stock, including x-ray film, for example. And the company was located in West-Germany. So that might make sense.
Well that’s one of those sentences you’d never have expected to read in your whole life. 
