Ok, that’s what I wanted to be spelled out. So you basically use camera as spectrophotometer in the sense that backlight color adjustment is monitored via camera set at specific WB. Fair enough. Another important point that you synchronize RGB peaks at 75% of max intensity. Overall it was not obvious to me as I personally would do this as close to right side of histogram as possible. In reality I actually was trying to use Spectromaster to get objective reading of blank orange mask without camera doing all this RAW to jpeg transformations for me. The only trouble is that with Spectromaster the whole procedure is very labor intensive as you have taken a reading, adjust lights, take reading again and so on. Yes if done properly , NLP conversion is very good and the step with white balance simply no longer need. Thanks for details!
Yes, NLP does a good job with most of my negatives. And as far as white balance goes, we could use custom WB as taken from an unexposed part of the film, but again, RGB backlights with mask compensation aims at converting with tone curve inversion only.
@seklerek’s “technicolor” approach with separate takes with R, G and B light and combining the takes in software gave me the following idea:
- Scan with long exposure time and a backlight that switches colours accordingly, e.g. one time unit of G, 2 time units of R and 4 time units of B light. This could help to
a) get rid of having to combine separate takes in software and
b) make implementation easier.
Interesting idea, could be fun to try out with a pc-controllable light source. My next prototype of the RGB panel will have full control over USB/serial so I can give it a go when it’s ready 
I am failing to grasp why we need 3 light exposures in sequence instead of having each RGB component to run at specific brightness. I understand the idea coming from movie production with their “timing” exposing positive film with three RGB lights for proper color.
…it’s to run the LEDs with a simple interface, no flicker, no RFI, high efficiency etc.
We could also put the LEDs in a long tube and set distance mechanically for the right mix. There are many possibilities and I tend to low tech 
Oh I see. That’s why I resorted to RGB lamp house from Philips PCS130 enlarger - all analog
It’s to avoid crosstalk between the RGB channels on the camera sensor, if you look at the histograms of my example raw files you can see that the red file is not strictly red pixels, same with green and blue. I think that by extracting them this way and recombining we can eliminate all this cross contamination.
This is of course valid point, but if I were to go for three shot scanning I would probably employ monochrome industrial camera and thus would increase resolution by eliminating Bayer array and demosaicking process. I know certain folks do that, but I myself cannot be tempted by that route.
Get a Pentax or Leica Monochrome camera…and there is always PhaseOne.
I’d be very tempted if someone provided the camera, preferably the 150 Mpixel variant.
Check out my proof of concept: Scanning Colour Negatives With a Monochrome Camera
dxomark.com publishes data on camera sensors. check out the data, e.g. for the Nikon Z7ii
How was your experience with the Focotar @davidS ?
I’m considering buying one to test it for my scans. For fun. They are quite cheap..
Trained in professional silver printing in one of the best Parisian B&W laboratories in the 60s, I sought to achieve the best possible results when digitizing 135 negative films. However, I was disappointed by the results I obtained with the various processes I tested. Digitization with a DSLR seemed the most interesting, yet something wasn’t right…
My research on the internet helped me understand that the cause of my disappointment was the diffuse light, which is nonetheless used, more or less discreetly, by all processes. Collimated point light thus became the only possible recourse, even though it is well established that its use is not without drawbacks. Would I be able to master it?
Spoiler alert: yes! But it’s anything but a simple and quick solution. On the other hand, what a pleasure to obtain A3+ prints that I am not ashamed of
I set up an ugly monster for a very small budget: less than 250 USD with second-hand equipment (excluding the DSLR, which is also second-hand).
I have just put online the complete description of my battle:
https://bw-film-scanning.oguse.fr
(Also available in PDF: [https://bw-film-scanning.oguse.fr/document/bw-film-scanning-excerpt-en.pdf])
I hope this can help or give ideas to some.
And my apologies for my poor English.
@ Alain_Oguse This is truly brilliant research! You’re making all the important points. I had been thinking about exactly this setup — a camera mounted to an enlarger — but always assumed it would be impossible to achieve light that’s both collimated and uniform enough to make it viable for camera scanning. The same goes for demosaicing: I’ve always seen a kind of conundrum in the fact that film “grain” is often around 5 microns, while many 24MP full-frame sensors have a pixel pitch of about 5 microns as well. I’ve long suspected that a proper demosaicing algorithm plays a crucial role, and I even experimented with Darktable, where the algorithm can be selected — but I never followed through on the research. Fantastic articles — thank you! Your article deserves its own thread on this forum - so it will have better visibility.
Alain_Oguse I’d also agree that you might consider creating a separate topic. This particular one, on DSLR scanning setups, is indeed a unique and valuable resource for quickly seeing the many different approaches to copying film with a camera but (imho) it doesn’t lend itself to in-depth discussion of any particular aspect, mainly because of its linear structure. The same is probably true of the equally excellent related topic on here of " Suggested backlight sources for scanning film with DSLR."
Some other examples of collimated v non-collimated here:
@Alain_Oguse : In 2021, when I started my home scanning journeys, I started the thread mentioned above “Collimated light for more sharpness“. Frankly, I gave up, and resorted to the ease of a diffuse light source (with a custom trichromatic RGB source to properly do color though). Anyway, I was super intrigued and impressed by your write up, and thank you for the efforts to do so. It motivates me to do yet another attempt, especially since you clearly point out that there is so much more to extract from a small 35mm negative specifically. I’m a physicist by training and profession and have the added luxury of my girlfriend being an optics professor, but miss your valuable experience in how things should look in analog-only world, so I’ll try and contact you privately. Perhaps we can work together?
Hello,
Frankly, I gave up, and resorted to the ease of a diffuse light source
(with a custom trichromatic RGB source to properly do color though).
That sounds like a good direction to me. For two reasons:
- A few years ago, before my technique seemed really usable, I did some
tests with tabular B&W film. And the results were really ugly. Now, as I
understand it, all color film is tabular grain. Perhaps this is a clue
as to why collimated light has such a bad reputation among those who
work with color - Je suppose votre source RVB is better than all the high CRI LED. Pour
moi, mon inclinaison aurait été d’utiliser tout de même une LED. Car ce
que je sais de la photo couleur date de l’époque où ce que nous pouvions
obtenir de mieux des films qui existaient étaient de toute façon encore
moins précis et moins stable que ces high CRI LED. Et j’aurai tenté de
récupérer les choses, tant bien que mal, en postproduction. Even a DIY
job like that would still be more qualitative than the 20x25 cm
Ektachrome film shots I used to develop by hand, in the 60s, in 12-liter
tanks in 12-film baskets…
For you, I can see two problems:
- That I run tabular B&W tests again to check whether these tabular
grains really don’t like collimated light. If so, that’s it, long live
diffused light for color. That would be a shame, if I’m to believe this
study you’re probably familiar with:
https://www.researchgate.net/publication/327937633_Chromatic_Callier_Effect_and_its_Repercussions_on_the_Digitization
_of_Early_Film_Colors - If not, then how do you make a good-quality point RGB light? And at
this point, you probably know a lot more than I do.
Anyway, I was super intrigued and impressed by your write up, and thank
you for the efforts to do so.
Thanks a lot !
I’m a physicist by
training and profession and have the added luxury of my girlfriend being
an optics professor, but miss your valuable experience in how things
should look in analog-only world, so I’ll try and contact you privately.
Perhaps we can work together?
Oh yes, please. It would be a great pleasure.
Best regards.
Sorry for my bad English and thanks to AI.
Thanks for your thoughts! I have described my custom RBG source somewhere on the discussions here, I believe in an optical sphere threat, but I also describe it on my blog:
A point source that has programmable R, G, and B is easy: It is one of these LED’s out of the 8x8 array.
Regarding your comment on tabular grains: Are you saying that a point source will yield ugly results also for modern B&W film that has tabular grain, such as Kodak’s Tmax and Ilford Delta?
Deserves its own thread, indeed!
Yes, the tests I mentioned that gave ugly results were made with the Tmax. But I don’t fully trust these tests because my process wasn’t yet fully developed. I’ll have to redo those tests to confirm.