Hi everyone, I have been thinking for a while about how I want to digitize my photos and negatives and I wanted to bounce some ideas around (partially inspired by the recent thread posted testing different LEDs)
What I was planning on was “dye separation” type setup where there are three illumination wavelengths that are supposed to be close to maximal absorption for one of the dyes relative to the others two.
My concerns include:
I don’t really have a clear picture of how it would work in practice, in a simple ideal world maybe each wavelength would be absorbed by only one of the dyes and you could immediately distinguish them but in the real world this is not the case, so I worry that I am underestimating how complicated this might be or overestimating how well it would work.
even if I do roughly have a digital image that models the concentration of each dye at every pixel it doesn’t really say how the dye responds to broad spectrum light in real life
On the other hand, rather than trying to digitize the individual dyes, I could try to better how the images and negatives actually appear, for example by using a broad spectrum illumination source and a camera with a bayer filter or by using additional LED colors. But I feel like (at least for the first option) in my mind the “fundamental data” in some sense is like the concentration of the dyes rather than how they appear to the eye and I don’t think it captures that as well (although I suppose you could do like a white + RGB).
Basically as a summary I guess what I ideally want is something like a “raw image” for prints and negatives, but on the other hand I feel like if I had a print and photographed it with a regular bayer sensor camera under an RGB LED it would look dead compared to a broad spectrum source, there’s a part of me that feels like I’m basically doing that with more steps
I did just look back in the backlight thread again and I see the topic has been discussed already but I would still be interested in any advice / suggestions
I think that it boils down to two different choices:
Color-reversal: you have broad dyes in your film stock, designed to give good colors when viewed in a projection setting. Projection-setting equals broad spectrum illumination, ideally similar to a Halogen lamp. Your digital camera should be tuned to “view” the colors as similar as a human observer does.
Negative: this type of material used to be projected onto positive film material or paper. The sensitive dyes in the receivung medium are usually tuned more narrowly than the broadband “receivers” of a human eye. So it’s better here to use narrow-band illumination. Color science is weird here, as it depends on both the negative and positive material’s characteristic. You have to choose your sampling wavelengths carefully, otherwise you might end up with weird color shifts especially for colors located at spectral places you do not sample with your narrow-band LEDs.
Some people have tested whitelight illumination combined with narrow-band RGB-LEDs, some even implemented multi-spectral narrow-band setups. You might enjoy looking at their results.
While a narrow-band or even multi-spectral setup enables you to raise the separation of color channels, you generally end up with not having a good color science available. That is: there is a chance that the colors you get from your footage might be off. Compare the two results here- the color of the lamp post is especially different in both scans. The question is: which one is the correct one? Probably it boils down to personal taste ….
The first thing I wanted to do actually was print photos, just to be a bit more precise, although in my mind I was thinking whatever lighting method I used I would just adapt for negatives and 8mm film.
I think I would probably not bother with white + rgb or “multispectral” LED, I feel like I’m probably either going to go with 3 LED (or maybe like one or two more if it would help with separation) and a monochrome camera or incandescent lights and a bayer filter camera. I guess I’ll probably end up testing both and seeing how they turn out (honestly it’s probably not even really a hard decision but there’s a part of me that doesn’t like the idea of digitizing with a bayer filter and I should get over it)
The most important thing really is that they look good and normal
If that is addressed to me I haven’t really done much yet. I am working on soldering some LEDs I got to wires so I can test them but some of the ones I got are really too small for me to work with, so a few of them I can’t use.
In general, in case anyone has comments, my idea was (ignoring the specifics of how to actually do it in software) to take a picture of a print photo with a bayer sensor camera under broad spectrum illumination as a “reference” (maybe a color checker too), then take 3 pictures under LED illumination with a monochrome camera and try to “match” the 3 channel LED image to the reference image.
Basically my thought process is, three narrow band illuminated images aren’t good enough to properly reproduce the appearance of real things, but since the print photo is basically made up of three absorptive dyes the actual spectrum it can reproduce is limited, so maybe three LEDs is good enough to reproduce the color.
But this is also ignoring practical concerns (like color tints if the uniformity is bad)
I do plan to test “blue vs royal blue” tonight or maybe tomorrow, but this has already been done I think so nothing really new. I unfortunately killed my one photo red led, I had wanted to test that vs a lower wavelength red. Green I don’t know if I killed but I don’t think I’ll be able to solder wires to it
I did just compare two now but of course they were both actually royal blue so there was basically no difference, I’ll have to solder the regular blue tomorrow and try again
Well I broke the blue LED, I’m not really sure what I’m going to do next. I’m not sure if I want to order something preassembled or try to solder myself. I was originally thinking I would order some blank PCBs and get the LEDs separately but now I’m thinking that might not work so well. But in addition to preassembled being more expensive I don’t know if the LED selection would even be good enough
In general, the datasheet of SMD LEDs gives some insight as to what maximum temperatures to use. SMD paste and an SMD hot air welder rework station are your friend. I bought one for US$30 in Amazon, and it did the work.
I think the package is maybe called “3030” for the ones I like, here’s an example https://look.ams-osram.com/m/553368862235ef0f/original/GD-PSLM31-14.pdf I don’t have a rework station, I was originally thinking I could just use a soldering iron but then I kind of realized the pads are on the bottom of the package so I’m not really sure how that would work, another option is a hot air gun but I didn’t really want to buy one and I kind of feel like it would blow the LEDs off
I am planning to use boards for the actual things, I was just soldering to wires because I wanted to test a few different types of LEDs first. I’ll think about the air gun,
and I should get over it)