Lenses for Scanning (film digitizing)

[Peter1111]

Looking for a new lens for scanning 35mm film on a full frame Sony. Since I started using an RGB light source, my Sigma 105 Art lens sometimes shows color vignetting, and generally requires a lot of software correction, so I’m looking for something more optically compatible. I tried a 55/2.8 Micro Nikkor and a 60mm f4 Rodagon (non-WA). The latter was flatter, but neither is as snappy as the Sigma.

Ideally, I would grab an Apo Rodagon D or Apo Componon, but since those are unobtainium where I live, I’m looking for alternatives. Has anyone used the Voigtlander 65mm Apo Lanthar for scanning? I know it is reputed to be excellent in general, but scanning is a special set of constraints. I know it needs some extension to get to 1:1, but that’s no problem. How is the flatness? Any quirks? Other ideas?

Canon MPE 65mm Macro lens. This is the most affordable route with very high quality. This lens is amazing!

 

[xkaes]

Canon MPE 65mm Macro lens. This is the most affordable route with very high quality. This lens is amazing!

It's surprising that in this auto-everything world, that they would produce a lens that lacks auto-focus. For me, macro and auto-focus are not very compatible, but this NON-auto-focusing lens goes for $1,000. You can get a lot of great, non-auto-focusing, macro glass for a 5% of that.

 

[silvergelatin]

The simple answer to everything is to buy or already own a 1+1 macro lens, job done, adapter is in the post, light source sorted and a copy stand enabled. How to make it more complicated probably just read the next post.

That worked fine until I added the RGB light. I cannot claim to fully understand the cause, but it must have something to do with how the different wavelengths are projected into the sensor. There is another thread about it here where a different user with the same light has similar color vignetting. But the RGB light is such a massive and fundamental upgrade in image quality, I think it’s worth going down the lens rabbit hole for a bit.

 

[xkaes]

I hadn't thought about the light source as being a confounding factor. I've never had to deal with that. Just lucky? I only use halogen for continuous light (3200K) and electronic flash (5500K).

 

[silvergelatin]

Canon MPE 65mm Macro lens. This is the most affordable route with very high quality. This lens is amazing!

I was curious about this lens, but am concerned that you can’t back up a little from 1:1. That means you have to frame the entire image perfectly, which seems like a pain. I like to leave a little space for white balancing off the border, etc.

 

[silvergelatin]

I hadn't thought about the light source as being a confounding factor. I've never had to deal with that. Just lucky? I only use halogen for continuous light (3200K) and electronic flash (5500K).

Flash was my gold standard until I got the Big Scanlight. Colors are just more separated and intense, and manual inversion became much easier. Not there aren’t still difficult images, but overall, it is cleaner and allowed to me to stop using inversion programs like NLP.

 

[250swb]

A couple of points.

If the RGB light source is the important link in chain the lens has to be exactly matched and neutral in colour reproduction or it will not render accurately, equally the camera sensor needs to be exactly matched and have none of it's own 'characteristics' that can impinge on accuracy, and the photographer has to put their own opinion to the side (an even more difficult task). That is an almost impossible chain to create. If any one thing is 'out' the quest for ultimate accuracy becomes irrelevant.

And more important absolutely nobody will be able to tell the difference between a scan using an RGB light source or something like a Cinestill CS-Lite given post processing software and the photographers opinion overrides everything that has gone before.

With any 1+1 macro lens you can have much space around the image as you like, you just move the camera further away on the copy stand and re-focus.

 

[koraks]

It's surprising that in this auto-everything world, that they would produce a lens that lacks auto-focus. For me, macro and auto-focus are not very compatible, but this NON-auto-focusing lens goes for $1,000. You can get a lot of great, non-auto-focusing, macro glass for a 5% of that.

The MP-E 65 is unique in that it's optimized for 1-5x reproduction rate without any additional optics, extension tubes etc. AFAIK it's the only such lens on the market for a regular camera system. It's also been around for quite a while. And like you said, it has no AF because it doesn't make sense in the application, so no surprise that the feature is missing.

Due to its emphasis on very high magnification ratios it's not the most suitable lens for this particular application unless you only do 24x36 and smaller film recording, and are digitizing it with a FF35 sensor.

Also, I'll include this link here that was posted in a similar thread recently as well: https://www.closeuphotography.com/2x-lens-test-2018
While it doesn't necessarily result in a clear answer to the question in post #1, it may be helpful in scouting options and determining which parameters to optimize for.

If the RGB light source is the important link in chain the lens has to be exactly matched and neutral in colour reproduction or it will not render accurately, equally the camera sensor needs to be exactly matched and have none of it's own 'characteristics' that can impinge on accuracy

What does this mean in practice, though? What is a 'matched' lens in terms of color reproduction? What is a 'matched' sensor in color reproduction? Do you know of any such system used in amateur or pro film digitization? I think in reality, every aspect of the image chain has its own set of peculiarities and the key is to manage this combination in such a way that the end result conforms to the requirements set to it. That doesn't automatically mean that everything is 'matched', 'neutral' etc. It just means that the process is optimized and where necessary, steps are included to compensate for one thing or another.

 

[originalwinslow]

There is another thread about it here where a different user with the same light has similar color vignetting.

This might've been me. I'm still battling this issue but I've made the calculus that simply adding a radial mask to the frames satisfactorily resolves the color vignetting in 95% of images without introducing ugly color shifts. Bonus is that it's free and easy if you have LR or equivalent software.

It's unclear if a different lens will actually get rid of this problem with the RGB light. The only thing I have heard that works is to elevate the film plane 3+ cm from the surface of the light with NR glass or equivalent. I bought some for this purpose but so far cannot find a reliable way to elevate the glass sufficiently and maintain stability. A 3D printed holder would be great. I haven't gotten around to running a rudimentary test to confirm this though, so grain of salt.

And more important absolutely nobody will be able to tell the difference between a scan using an RGB light source or something like a Cinestill CS-Lite given post processing software and the photographers opinion overrides everything that has gone before.

I disagree. There are clearly visible tonality and saturation improvements that cannot be recreated with simple post production. I also see improved dynamic range in some scans and overall more accurate skin tones particularly in the reds. I was using the CS light and won't be going back.

 

[silvergelatin]

This might've been me. I'm still battling this issue but I've made the calculus that simply adding a radial mask to the frames satisfactorily resolves the color vignetting in 95% of images without introducing ugly color shifts. Bonus is that it's free and easy if you have LR or equivalent software.

It's unclear if a different lens will actually get rid of this problem with the RGB light. The only thing I have heard that works is to elevate the film plane 3+ cm from the surface of the light with NR glass or equivalent. I bought some for this purpose but so far cannot find a reliable way to elevate the glass sufficiently and maintain stability. A 3D printed holder would be great. I

Yes, I was thinking of your situation with the 60mm. I was able to get rid of the vignette by using the 55mm Micro, so I’m surprised the 60 is giving you problems. I tried different film heights as well and it made no difference for me.

One thing I can recommend is to try out Capture One, as the flat field correction is much easier and doesn’t create extra files. You can even save it as a preset if your height/aperture/alignment don’t change. With an RGB light, I found I don’t really need NLP anymore.

 

[Ardpatrick]

I know nothing about RGB scanning apart from reading various conversations on this forum. It seems to be a lot of trial & error, but folks who do it tend to persist, which is interesting.

Could someone who is working with the process post or otherwise link to some scans that show a comparison of RGB vs single exposure scanning? I’d love to get a visual sense of the benefits.

 

[originalwinslow]

Yes, I was thinking of your situation with the 60mm. I was able to get rid of the vignette by using the 55mm Micro, so I’m surprised the 60 is giving you problems. I tried different film heights as well and it made no difference for me.

That is strange considering the 55 and the 60 are essentially the same lens design as far as I understand. I tried shooting blank frames on a couple other (non-macro 50mm AF-D and 35mm AF-D) lenses and got similar vignetting.

Could someone who is working with the process post or otherwise link to some scans that show a comparison of RGB vs single exposure scanning? I’d love to get a visual sense of the benefits.

Here is a side by side example that shows significant improvement in tonality and saturation in the greens and reds. I find the difference is more apparent in certain emulsions. This is 500T developed ECN-2 which looks much better with the RGB IMO and in a way that I could not have replicated without heavy-handed post-processing. It also looks more accurate to the real-world color of the wall and skin tones . These are NLP conversions with no adjustments other than a slight white balance tweak. I bring the contrast/saturation down a bit more in the RGB conversions and they come out looking quite nice.

These are both single exposure, however. I haven't tried much stacking of three separate RGB exposures as there isn't a simple workflow currently available. It's too time consuming to make it viable for me at the moment and the results I have seen were not blow-me-out-of-the-water improved.

@silvergelatin uses a technique of balancing individual RGB channels to negate the color mask and doing manual inversions. With this method they are able to eliminate NLP or other conversion tools from the workflow. The RGB light needs to be adjusted manually for each emulsion. I'd love to see some results from this.

 

[silvergelatin]

@silvergelatin uses a technique of balancing individual RGB channels to negate the color mask and doing manual inversions. With this method they are able to eliminate NLP or other conversion tools from the workflow. The RGB light needs to be adjusted manually for each emulsion. I'd love to see some results from this.

I don't bother with that anymore, since I didn't really see a big difference between that and just white balancing. Maybe there is a difference, but I'm too lazy to do a proper test. Even without the individual adjustments, I am still getting better results in Capture One than in NLP.

 

[_T_]

If you look at the histograms for the color channels of a digitized frame done with white light before processing you’ll notice that the information for the red channel is nearer the center of the histogram and takes up a significant portion of the available spectrum (from 0 to 255), the green channel is pushed lower on the spectrum and takes up less of its bandwidth, and the blue channel is a tiny little spike all the way at the bottom of the spectrum into which all the blue information is jammed.

When you process these channels you end up stretching out the information so that each covers approximately the same amount of the spectrum, so you get a good red channel, a worse green channel, and a blue channel with very little information.

Normally you can’t really tell because the information in the other channels masks the ugly blue channel, but if the frame needs significant editing it can become a serious problem in the form of inaccurate color, banding, noise or other artifacts. The situation can be compounded by errors in exposure.

So the idea of using a separate red blue and green light to make the exposure, or even using a separate red blue and green exposure is that you can adjust the exposure for each channel before the information is captured thus using as much of the spectra (again I mean spectra of brightness values) as possible giving you a file with a lot more gradiations of color in the green and especially blue channels.

This can translate to slightly more accurate colors and fewer artifacts in the final image. And for some frames you could end up with much better color if there happens to be a lot of areas of blue or yellow in which the extra information could improve the image. Stuff like faces or bluish objects in a poorly lit or poorly exposed scene.

I’ve only had a couple of situations myself in which I would have benefited from this kind of process but if you have the shot of a lifetime and miss it because your blue channel is ugly and can’t be corrected, making your subject look sickly and pallid no matter how you adjust the image then you might wish you had it.

I probably won’t get into personally. Seems cool though

 

[koraks]

So the idea of using a separate red blue and green light to make the exposure, or even using a separate red blue and green exposure is that you can adjust the exposure for each channel before the information is captured thus using as much of the spectra

This can be one part of the argument, but it's not the full story. The main rationale why esp. in the professional domain the shift was made to multi-spectral capture is one of reducing crosstalk between the color channels.

To match the color channels to the sensitivity of the capture sensor (e.g. a dSLR/mirrorless digital camera), it would suffice to use filtration; some indeed have used a regular enlarger dichroic filter head as the light source, balancing it so that the red, blue and green curves overlap on the digital image. From a viewpoint of digital image quality (esp. s/n ratio) this can make sense, although I personally find it doubtful that it yields all that much better quality. This leaves the issue of spectral purity that does demonstrably make a difference, although also here, one might argue that a single white-light capture is in practice for the most parts plenty good enough. But we often want to optimize, even if there's not necessarily a rational argument to do so. Just, because we can - that sort of thing.

 

[_T_]

I didn’t realize that cross talk could be such a problem as to be the primary reason for using RGB. Maybe some of what I’ve attributed to low color depth in the few files I’ve had problems with could actually be attributed to this? Like I say the rare occasions where a file gives me trouble it’s either a poorly lit subject or an incorrect exposure so the starting point is already one of salvaging something from a problem negative.

 

[koraks]

Oh, to be clear, I understand and underwrite your analysis w.r.t. histogram compression and the potential issues it brings. I just think it's not the primary reason why the industry and at least some amateurs are moving towards narrow-wavelength exposures. However, that approach does allow two birds to be killed with the same stone. My experience is similar to yours in that only some problematic negatives really present challenges and if your exposure and development are somewhat reasonably in control and you're using reasonably fresh film, these problems can be avoided entirely.

 

[silvergelatin]

Do we need a rationale or explanation? I don't get into the theory or thought-experiment end of this at all. I tried the Scanlight and it beat everything I had tried up to that point (95 CRI light panel, flash), by quite a bit. The difference is not subtle. I have not compared to an expensive 99CRI Neg Supply light or a discontinued Solux, so I cannot say I have tried everything, though. The difference is not just in the finished image (which is almost always better so far), but how easily I got there.

 

[koraks]

Do we need a rationale or explanation?

If the question is "do I need to face the additional complication or can I also get there with the materials I already have", then yes.
I don't contest that you got easily where you wanted to be with a trichromatic capture approach, but that's a different question from whether you could have gotten to more or less the same place with the white light source you used before. Odds are you could get pretty close, and that most of the problem was related to how to do the color balancing.

PS: I acknowledge that for you personally the question is probably moot: you tried this, it worked for you and you're happy with it. My comment above refers more to the question of generalization, which I think is relevant due to your question why 'we' need a rationale. The answer, I think, is that some of us do, while acknowledging the some won't need one.

 

[originalwinslow]

All I'll say is that I have many images scanned using a 95 CRI that I thought looked great. Then I ran them over the RGB and realized how much better they could have been. The change was significant enough to take the time and effort to re-scan quite a few negatives. Yes, you could probably match the tonality of the RGB using heavy-handed editing in photoshop if you have a reference image of the "better" scan.

But probably not unless you know exactly what to edit for.

I will agree that the trichromatic stacking approach is more effort than it's likely worth. I can get 99% of the way there with single capture based on the examples I have seen.

My biggest RGB issue (besides the vignette) is unnatural saturation. Noritsu scans, for example, can come out quite pleasingly neutral and desaturated. While I appreciate the color depth of the RGB, images can sometimes take on a "candy" color if that makes sense.

 

[MattKing]

All I'll say is that I have many images scanned using a 95 CRI that I thought looked great. Then I ran them over the RGB and realized how much better they could have been.

That may be more to do with how inadequate the CRI specification is for these purposes as anything else.