I just saw this thread; I posted about this topic in a different thread a few days ago.
My experience so far has been very different. Shooting with the Big Scanlight in RGB mode, I am getting results that are so far off that I can’t even figure out the correction to get them to look right (at least, within NLP itself; I do think I could get it there with Lightroom adjustments). Principally, the saturation is unnaturally high and the colors are skewed magenta. I also did not find that I could skip the white balance step, though I do agree that the shift needed is smaller.
I’m curious if there’s something different about my setup. Is there some particular film stock you’re using that you’re getting great results with? Most of the negs I’ve tried have been Portra/Ektacolor 400, but I have tried some older stuff from my archives and had similar problems. I’ve gone back to just using it in white mode rather than RGB mode for now, because at least that gives closed to natural results.
I’d like to revise what I said in my post, because this stuck with me and I went back into my office to do some more testing. What I found is that, while the RGB light does not completely cancel out the orange mask (doing a white balance still creates a noticeable shift), it does work to run NLP on the negs without doing a white balance first. And the result might be a bit better (though, at least on the couple of sample frames I tried, it wasn’t radically different). So that might be something I will play with some more.
Using NLP with an RGB source can result in weirdness. I find it is much better to do straight inversions and color corrections with the Scanlight scans without any special inversion algorithms. I use C1, but I’m sure a fast and efficient workflow could be built in LR as well.
Alright, this was a great tip. Just white balancing and manually inverting without involving NLP at all works very well. Thank you for suggesting this!
Hello,
This is a DIY project I’ve been interested in as well, so your information has been very helpful. Thank you.
I’m planning to make a 4x5 light box in a slightly larger case. As you mentioned, do you think a 9x9 arrangement would be sufficient, or would you recommend using more?
I’m also considering using the FilmGrade Flicker-Free LED Dimmer. If you think that would be a good choice, I’d really appreciate hearing your opinion.
I’d consider a deeper box for better uniformity of lighting and ventilation. Total dissipated power!
Arrange as 8x10 and check if you want/need uneven spacing. Smaller spacing near edges and corners. From a mechanical/usage point of view, I’d prefer both cables got out at the same side/back of the dimmer.
Thank you for the advice. I’ll look for a deeper box on Amazon. Hopefully, I can find one that would work well for this project.
Also, could you please explain a little more about the 8x10 arrangement? Do you mean the number of LEDs and the number of LED strip sections, or are you referring to something else?
You want a 4x5 capable scan light, if you arrange the LEDs in a 4x5 (or 8x10) pattern, you match the dimensions more homogeneously, at least that’s the idea.
The strips seem to come with groups of 3, therefore, a 9x9 arrangement (or 8x9) is as close as you get. Getting more LEDs near the edges helps to reduce the falloff (see here) but again, the strips might not cooperate. Due to the falloff of LED panels, a bigger scan light reduces the differences between central and peripheral illumination. BTW, the calculation is for a row of 7 LEDs, not quite your project, but close.
One row was scanned using a Kaiser Plano backlight, the other with an iPad displaying a kind of violet that I selected to be the screen background. This (provided by Apple) violet does a) compensate for most of the orange mask and b) provide almost equal levels of R, G and B for less noisy highlights. Camera was set to UniWB to see the best approximation of a RAW histogram.
The 1970s negative is very thin and taken in dark, mixed light. The two candles were actually brightening the view on the wooden 3D puzzle that the math teacher struggled with. In short, not the best negative for colour conversion - B&W is easily within reach though. Scanning exposure ranged from +0 to +2 in thirds, darkest scans at left.
The conversion was made with Frontier colour model, default pre-saturation and my standard preset…and again, differences can be found in tonality and colours - if one has to find them at all cost. Handling this lousy negative’s conversion is easiest with a positive copy, e.g. with AI background and person presets. Sticking to RAW did not really work, Lightroom 14’s AI did not recognise the person or the background.
My take regarding thin negatives:
NLP can easily handle different light and exposure and still provide reasonable starting points…which does NOT mean that results can have wider variations depending on conditions.
ade few pieces of RGB+W light called Shoebox-light RGB which is good for good for up to 4x5 ( I own one of his boxes btw). Darrel published few pictures of his device, I think I will not be in wrong if I republish it here. The point is that device is quite big and much more voluminous then you envision. The heat LED will produce in confined space will not be trivial so you have to provide means for heat to dissipate. I think you need really large metal box and sort of white diffuser more resembling integrating sphere to have light uniform enough. I personally believe that you need styrofoam half sphere ( avail on Amazon and art stores ) and have all light coming from above and mixed in sphere. 4x5 is not trivial size and unless you achieve very high uniformity of light the whole thing will simply be wast of time and money. Certain things need to be done fully analog as fixing things digitally will be unsustainable
Hello,
Thank you for the photos and the detailed information.
I was thinking of buying an aluminum box from AliExpress, approximately 340 x 235 x 95 mm, arranging the LED strips appropriately through testing, and then placing a Perspex Spectrum Opal 1TL2 Acrylic Plastic Sheet(3 mm or 5 mm) on top. If necessary, I was also thinking of adding a fan to the aluminum box.
Excuse me for asking, I don’t mean to pry, but back in June I see that you ordered a Godox P260C Pro Bi-Color Light Panel (CRI Average≥98, TLCI Average≥98). It’s pretty large but otherwise I would have thought it was about as good as it gets as a film copying light source, I’ve seen it used in a very high end system. Did you not get on with it? I suppose it’s the fact that it’s not RGB.
Regarding the 5" x 4" light source you could I suppose decide that for final copying of 5"x 4" sheet film you would always stitch using the long side of the sensor across the short side of the film sheet, in which case you only need a 100mm wide section of even illumination. You need a very high MP sensor to get all the information from 5" x 4" otherwise.
Darrell Davis gave me an option on a cancellation for his Mk.2 light box. I couldn’t take it up, the timing was wrong and I’m in the UK so I wasn’t sure what the final cost was going to be, taxes, tariffs etc. Still for me it looks like the best option out there, though I don’t think it’s an option now unless he does another batch. I like the fact that he has a tunable amount of white light as well as RGB which he found works very well with NLP, I think he only used NLP. He published a lot about it on the FB forum as I’m sure you know. It’s a pity he didn’t post it all on here as well.
a backlight that compensates the orange mask makes NLP almost completely immune to exposure variations in a range of 0 to + 2EV
Using “very low” pre-saturation helps to provide conversions with “cinematic” look
The background I used to mostly compensate the film mask has sRGB 177-139-254
A higher value for red would have fully compensated the mask, but as a proof of concept, anything near the given values is good enough. Again, backlight was an iPad set for violet background colour. Even though iDevices aren’t good at CRI, the colours they provide are good for scanning.
