Interesting that there is some inconsistency. All I can say is that when scanning, the only time the color vignetting issue became detectable in my images is when I scanned very close to the light source. Up until that point it wasn’t even in the radar (can be seen in my example above).
I’ve had issues with color shifts in unintended areas of my frames when using FFC with the color correction enabled, so instead I change the color profile of my FFC calibration frames to monochrome so that only light falloff is corrected. So no color correction applied. I will add that with the Tone Carrier the distance to the film is significant. I need to measure it but I’d estimate it is in the ballpark of 30mm.
I used a Valoi Easy35 with the light source removed (with and without the diffusion panel) and scanned at least 10-15cm up from the panel, and also tried a Negative Supply carrier right on top of the light. Same pattern in all cases. That’s why I think it has to do with the sensor physics. I did get a less intense and differently colored pattern when trying out an old 55 Micro Nikkor, so optics are definitely part of the issue. No issue with white light at all.
Here is a crazy idea: if anyone has those Cinestill collimating sheets handy, maybe try a comparison with and without with the Scanlight? If it does have something to do with diverging rays, this may provide some insight. They are not available in my country, so it would cost a lot in shipping for me to try it myself.
I have access to them, but I do remember reading on Jack’s site saying that the sheets will actually cause some vignetting (the ones he sells at least). So I can’t imagine they would make it better. I’ll think about trying at some point but like I said, for whatever reason this is not an issue I’m noticing in my setup until the film is close to the light. The collimating sheets will certainly increase brightness.
Also, just FYI, the collimating sheets Jack sells have a strong blue cast to them which was having significant effects on red color reproduction in my early scans. Removing them was a big improvement. The Cinestill sheets are clear however.
Perhaps, but if he is not getting vignetting without them, then his lens and sensor are likely different. I get vignetting at all distances, have tried adding diffusion, and still get the same pattern, so it makes sense that going the other way might work.
I ran a couple of tests yesterday, and discovered that if i shoot the light source by itself, with no extraneous light able to get near it at all (Neg Supply hood AND Sigma lens hood overlapping), and I set all the RGB levels to max (I previously was neutralizing the mask at the RGB controls), I do not get any visible color shift at the edges, even with the contrast jacked. If I insert a blank piece of film, the color pattern is back. Underexposed images still require FFC/LCC.
FWIW I leave all RGB channels at max. I don’t believe Jack recommends attempting to neutralize the orange mask with adjustments. I’ve taken test shots to ensure no RGB channels clip, and have found if I expose at +1/3 stop, this is maintained generally. When converting, I follow standard NLP workflow. I use FFC (with no color correction, changing calibration frame to monochrome profile), white balance off the film border, then convert. I’ve been pretty happy with the results. Sometimes I need to do some white balance correction. But again, I haven’t notice weird color shift vignetting, just the once time I scanned up close. Wish I could say why, but that’s my workflow.
Here are some recent 35mm conversions, Porta 400 & 800.
1 Like
Withe an unexposed and developed piece of film, the pattern might be caused by reflections between the light panel and the film base or between the film and front lens.
What happens if you put the film directly on to the panel without any holder or space?
Placing it directly on the panel without ay stray light hitting it would be difficult. I will test it next time both emulsion side up and also down. With the nested lens hood/carrier hood setup, though, I don’t think any stray light is getting in.
Maybe it’s not about stray light, but about reflections. Other than that, the lens that took the negtive could have had some falloff and that is why an unexposed part of the film should be used.
I am only testing with unexposed film. I tried the neg both face up and face down, and it has the pattern either way, so it is less likely to be reflection. I also noticed that in my previous test of the panel without any film (the one that looked even good under high contrast), if I start skewing the white balance sliders, the pattern re-emerges.
Ok, my head is still spinning a bit from a lot of loose testing and general frustration, but I think I have identified that my Sigma lens is the problem, but cannot definitively say why. AI models into which I fed all the details have a couple of opinions…
So I remembered that I have an old 55mm Micro Nikkor and PK-1 that I bought a long time ago to try out camera scanning before I got the Sigma（first the 70, then the 105), but I did not actually end up using them much. I decided to try a few test shots on recent negs, a couple of which required FFC in LR with the Sigma lens, with the Nikkor lens and quickly discovered that the issues with color vignettes were largely gone. In fact the results were quite good, edging out the Sigma for sharpness and contrast at first glance (this was only a few frames. I will continue to compare). Blank film with the contrast jacked showed far less color pattern (not zero, but way, way less), and there seemed to be less color cast overall.
Why? Cannot say. I don’t believe most of the nonsense that AI tells me, but I ask anyway. ChatGPT thinks that modern lenses like the Sigma are so ultra-corrected that an alternate light source like an RGB light create anomalous ray angles. Makes sense? The Google AI says that modern macro lenses like the Sigma rely on software correction profiles as an inherent part of the lens design, so that in reality, they are actually not that well-corrected optically. Also makes some sense.
At the end of the day, it seems that my vintage Nikkor has circumvented the issues of RGB light scanning, so something in the above may be correct. I will probably be trading the Sigma as part of an upgrade to a fancier camera body, perhaps. But after a few more comparisons…
Just as a disclaimer…this is the rabbit hole I went down, and these were the results. Please don’t take my word for any of this, as it may be a giant misunderstanding. Do your own tests.
…and changing every link in the chain one by one and in all combinations can lead to something that we can live with. Starting with “good” gear is no guarantee that it will cooperate nicely in each and every setup.
1 Like
I just got one of the new batch of Big Scanlight, and wow. The results I’m getting in narrowband RGB mode with NLP are awful. Saturation seems way too high, too much magenta, and I I have not been able to figure out a correction that makes it look decent. In white mode, it seems ok, but I’m wondering if anyone has advice on using narrowband RGB (or should I just ignore that mode for now?). I’m still pretty new to NLP, so it’s entirely possible it’s user error!
Try setting NLP to linear.
Honestly, you can just skip NLP entirely and invert manually until Nate makes an RGB-ready version. NLP corrects for the deficiencies of white light, so now that you fixed that on the hardware side, it is over-correcting.
Another thing can try is white balancing for your favorite film stock on the light itself, and saving that as a preset. It may or not help with NLP. I honestly don’t remember.
