Dedicated camera just for scanning

[Alan Edward Klein]

What do governments and museums use to archive their film?

 

[macfred]

What do governments and museums use to archive their film?

Ted Dearberg of Imperial War Museum London and Andrew Bruce of The Postal Museum talking about their glass plate and negative scanning workflow https://blog.townswebarchiving.com/...itising-glass-plate-negatives-scanner-vs-dslr
This article is about storing negatives in a museum archive https://sydneylivingmuseums.com.au/stories/archives-negatives

 

[Alan Edward Klein]

Ted Dearberg of Imperial War Museum London and Andrew Bruce of The Postal Museum talking about their glass plate and negative scanning workflow https://blog.townswebarchiving.com/...itising-glass-plate-negatives-scanner-vs-dslr
This article is about storing negatives in a museum archive https://sydneylivingmuseums.com.au/stories/archives-negatives

https://www.flickr.com/photos/igel_rupert/

From the first link was this comment: "Over the years we have dabbled with camera capture but it has proved problematic in terms of quality and consistency particularly for damaged or over or underexposed negatives; where the camera accuracy suffers because it compensates for this by adjusting its exposure in a way that scanners don’t do."

Have those who use cameras run into this problem? How do you deal with it?

 

[Les Sarile]

Les, you "forgot" to get yourself a modern camera-based scanning setup to get familiar what you're trying to compete with here. In auto mode it will do a full 36-exp roll in under 5 minutes with better results that you've shown above. And on top of that, people usually spend minutes per shot, to get to yet another level, just like they do with RAW files when shooting digitally.

I have the full Nikon autobellows and slide film copier setup and can make copies as fast as I can spool it up with the fullframe D800. Speed and resolution are not an issue. Admittedly, I lack incentive to get better with post work on color negatives due to having the Coolscans.

 

[Les Sarile]

I've bought and used many scanners before finally decidng on the Coolscan 5000 when it was first reelased in early 2000 after testing some film at the PMA show in Vegas. Picked up a 9000 just after Nikon discontinued it and before they skyrocketed in price. Picked up the V just recently from a yard sale just down the street from me. I continue to to try post work on DSLR copies of film just in case I no longer have access to my Coolscans but they're looking like they will outlast me just like the rest of my manual cameras . . .

Here is how a Coolscan's 4000dpi compares to my Pentax 14MP K20D as well as Nikon 36MP D800 when it comes to scanning 35mm Kodak Techpan film.

You can see that even though the D800 (7360 x 4912) applies more pixels then the Coolscan (5600 X 3600), they are almost equal in actual detail resolved with a slight edge to the Coolscan.

Full res -> http://www.fototime.com/8372250EA44CB06/orig.jpg

But of course none of these methods of scanning can fully resolve this particular film as you can see from the optically magnified crop on the right all the detail that was not resolved.

I'm good with what the Coolscan can get out of my film in terms of detail but really more with the color/contrast accuracy. Color negative inversion is the biggest obstacle for me and spotting dust and scratches is a real PIA specially since ICE is so much better for anything else except b&w.

 

[Adrian Bacon]

Les, you "forgot" to get yourself a modern camera-based scanning setup to get familiar what you're trying to compete with here. In auto mode it will do a full 36-exp roll in under 5 minutes with better results that you've shown above. And on top of that, people usually spend minutes per shot, to get to yet another level, just like they do with RAW files when shooting digitally.

@MattKing you are absolutely right, except I would elevate software to be of uttermost importance on that list. The problem is that assembling a complete kit requires above average understanding of how these components work and interact, so having them pre-assembled with a big red "Easy" button on top would be nice. In fact, I was toying with an idea of making a prototype and putting it on Kickstarter a while ago. I started by purchasing different variations of existing components and putting them together. In the end, I lost interest. Here's why:

There's already a company that sells you the complete hardware kit: Negative Supply. But instead of selling a complete package, they chose to break it into components and charge a hefty premium for each. If you go through the exercise to determine the total addressable market for this product, you'll quickly understand why: the market is tiny. Integrating NS-like hardware with a camera, a lens and good software will require additional R&D headcount which, combined with the need to reduce price, leaves little room for good margins.

If you want best-money-can-buy solution for scanning film, it is already on the market: get a complete set (stand, light pad, negative holder) from Negative Supply, a mirrorless pixel-shift camera, high-quality auto-focus macro lens and Negative Lab Pro (Negmaster is also a good option). The end user experience and results will be amazing. Putting these parts together into a single package is just an exercise in cost-cutting with limited upside from a business perspective. In other words, the volume isn't high enough to justify upfront costs of integrating existing components into a lower-priced package.

This is exactly why I've been so reticent to make my code publicly available. I have serious doubts that the market is large enough, and making it available for the general public means I have to spend a significant amount of effort supporting other cameras that other people are going to insist on using instead of buying supported hardware because they don't understand how much effort it is on my end to support a whole pile of hardware so that they get generally good results with the same or less effort than what is currently available on the market.

 

[Adrian Bacon]

From the first link was this comment: "Over the years we have dabbled with camera capture but it has proved problematic in terms of quality and consistency particularly for damaged or over or underexposed negatives; where the camera accuracy suffers because it compensates for this by adjusting its exposure in a way that scanners don’t do."

Have those who use cameras run into this problem? How do you deal with it?

They're doing something wrong if they're having that problem. If you shoot raw and run in manual mode, there is no such thing as the camera compensating for over or under exposed negatives.

 

[fmueller]

Get a mirrorless camera with a pixel shift function. You'll get great true RGB files with a very low noise floor. There are ways to get super high res images but if you're not printing big most APS-C cameras are just fine for this purpose. If you do want super high res, I recommend a Panasonic S1R or the current Sony A7R body.

The other thing I can recommend is an AF macro lens. Why risk an out of focus scan that you have to re-do later? AF will generally nail it every time, especially on a mirrorless camera that can AF seek on the whole frame. I think Sony may have an APSC mirrorless with pixel shift and probably more than a few Sigma or Tamron macro choices.

This is THE answer. Used S1R's for less than $2k. L mount Sigma macro. Get a good negative holder and light source. A cheap copy stand is nice but you can start out with camera inverted on a tripod.

This is a lightning fast setup compared to any flatbed or dedicated negative scanner...with superior results.

 

[locutus]

S1R With the Sigma 70mm macro on a copy stand in 196MP high resolution mode, you'll be amazed by what you suddenly see in your negatives >

 

[Les Sarile]

This is THE answer. Used S1R's for less than $2k. L mount Sigma macro. Get a good negative holder and light source. A cheap copy stand is nice but you can start out with camera inverted on a tripod.

This is a lightning fast setup compared to any flatbed or dedicated negative scanner...with superior results.

Can you share some results in full res?

 

[JWMster]

Software: Adrian's right as rain - speaking as a guy who does some coding. The great thing about software you put together yourself is you have EXACTLY what you need for your workflow. The bad thing about selling it is all that support for folks who want to pull it in other directions.

Personal Perspective: Quality vs. Speed: For my bit, the time and difficulty involved in scanning is a factor that has to be weighed as well. I have a Nikon LS8000 and its fine. Les's camera scans look a decade or so old to my eyes, and whether it's true or not, I sense something of a false comparison. Nikon scans are "okay" but nothing I'd brag about. I've seen Epsons that look MUCH better, and indeed many DSLR scans that do as well. But both of these can be VERY slow. From what I read, a pixelshift DSLR scan isn't necessarily faster either. The point is that DSLR scans CAN BE VERY QUICKLY done. There's a positive there at least as a starting point.

Workflow Objective: Parse the workflow: Get initial scans as quickly as possible. Bank time for working on the SELECT images by speeding the cull process. This involves BOTH being more selective in shooting (than spray and pray and its variations - which is harder to do even with great light than it sounds), and speeding through negatives to identify the SELECTs from the simply "good but not great" in the first place. Maybe you can judge negatives without scanning? I'm not that guy. Even 4X5's I like to "see" the conversion. With 35mm? There's really not much option unless you shoot chromes. Scanners that overwork EVERY image 'cause that's what they do? Scanners where even the PREVIEW slugs along? These slow the process down and at best might want to be what you use for SELECTs rather than for the prelims.

For my bit having done the latter, and I am literally... "DONE with THAT!" Sticking with film means solving some of these issues best as I can. But if film is only a small part of what you do, and if you're not already spending time developing negs, too, then maybe you've got time to spare and it's not a problem. Or maybe the patience of Job is just in the blood stream? Me? Folks tell me I'm a patient guy.... but I'm not crazy.

 

[macfred]

Does the quality of a photo increase with the resolution?
Irony-modus off

For what do I need approximately 200MP large files? For sharing on the net? Surely not ... For printing ? What do you have for printers at home in operation?
I have access to a pretty good A3 printer; the prints are quite excellent - even with much lower resolution.
'For the wall' I prefer a hand enlargement anyway.

 

[runswithsizzers]

I've bought and used many scanners before finally decidng on the Coolscan 5000 when it was first reelased in early 2000 after testing some film at the PMA show in Vegas. Picked up a 9000 just after Nikon discontinued it and before they skyrocketed in price. Picked up the V just recently from a yard sale just down the street from me. I continue to to try post work on DSLR copies of film just in case I no longer have access to my Coolscans but they're looking like they will outlast me just like the rest of my manual cameras . . .

Here is how a Coolscan's 4000dpi compares to my Pentax 14MP K20D as well as Nikon 36MP D800 when it comes to scanning 35mm Kodak Techpan film.

You can see that even though the D800 (7360 x 4912) applies more pixels then the Coolscan (5600 X 3600), they are almost equal in actual detail resolved with a slight edge to the Coolscan.

Full res -> http://www.fototime.com/8372250EA44CB06/orig.jpg

But of course none of these methods of scanning can fully resolve this particular film as you can see from the optically magnified crop on the right all the detail that was not resolved.

I'm good with what the Coolscan can get out of my film in terms of detail but really more with the color/contrast accuracy. Color negative inversion is the biggest obstacle for me and spotting dust and scratches is a real PIA specially since ICE is so much better for anything else except b&w.

Although I have seen this chart a few times before today, I am still having trouble understanding what it is trying to tell me. On the right, the enlarged chart says it is for the "K20D+[bellows, etc]+SMC Pentax-M 50mm F4 Macro" The red arrows are pointing to a value between 11 and 12. But on the left, the target for the K20D has red arrows pointing to values closer to 5-1/2. Why the difference? And why even discuss the Pentax K20D as representative for digital resolution when you have results from the far more capable Nikon D800?

And also, I wonder if @Old Gregg might have a valid point about the digital equipment you use in your comparison as being something less than 'modern' ?

The K20D (2008) was early days for digital, and the SMC Pentax-M 50mm F4 Macro was designed in the 1970s to be used with film cameras, not digital. The Nikon D800 from (2012?) has a more recent sensor than the Pentax, but I don't see where you say what lens was used with the Nikon. For our purposes - copying film, or in your case, copying test charts - I think the lens is probably more significant than the camera, right? The main point of your posted chart is a comparision between the resolution of the Nikon D800 and the Coolscan - so, at the very least, the viewer needs to know what lens was used on the D800.

The aperture used can visibly affect the resolution, as can vibrations from a less-than-solid setup. What shutter speed was used? Did you press the shutter release with your finger, or did you use a cable release or time delay? You have gone to a lot of work in preparing your test chart, but without knowing more about your test conditions, it is difficult for me to interpret the results.

 

[runswithsizzers]

@runswithsizzers your comment is hinting at the elephant in the room: the major weakness of camera scanning is its high dependence on operator skills. This is true with every technical hobby, and also plays a role with scanners, but to a much lesser extent.

So true! There are challenges to both approaches.

But, "The fool who persists in his folly will become wise." ~William Blake. That is, whether one chooses the path of buying an off-the-shelf film scanner - or puts together a rig built around a digital camera, a certain amount of effort and education are going to be required to achieve success (not to mention money). But assuming the major pitfalls of each method have been avoided, can anyone really say that a film scanner produces higher image quality than a competent camera-scan?

 

[Les Sarile]

Although I have seen this chart a few times before today, I am still having trouble understanding what it is trying to tell me. On the right, the enlarged chart says it is for the "K20D+[bellows, etc]+SMC Pentax-M 50mm F4 Macro" The red arrows are pointing to a value between 11 and 12. But on the left, the target for the K20D has red arrows pointing to values closer to 5-1/2. Why the difference? And why even discuss the Pentax K20D as representative for digital resolution when you have results from the far more capable Nikon D800?

And also, I wonder if @Old Gregg might have a valid point about the digital equipment you use in your comparison as being something less than 'modern' ?

The K20D (2008) was early days for digital, and the SMC Pentax-M 50mm F4 Macro was designed in the 1970s to be used with film cameras, not digital. The Nikon D800 from (2012?) has a more recent sensor than the Pentax, but I don't see where you say what lens was used with the Nikon. For our purposes - copying film, or in your case, copying test charts - I think the lens is probably more significant than the camera, right? The main point of your posted chart is a comparision between the resolution of the Nikon D800 and the Coolscan - so, at the very least, the viewer needs to know what lens was used on the D800.

The aperture used can visibly affect the resolution, as can vibrations from a less-than-solid setup. What shutter speed was used? Did you press the shutter release with your finger, or did you use a cable release or time delay? You have gone to a lot of work in preparing your test chart, but without knowing more about your test conditions, it is difficult for me to interpret the results.

Good questions and hopefully these provide clarification.

I have no idea what Old Gregg means by a more "modern" as he is only using a 30MP 2016 Canon 5D Mk4 while I show a 36MP 2012 Nikon D800. I am also not sure why anyone would consider it having any validity? Older camera with more MP is more applicable in terms of resolution testing.

To the left bottom is just the full view of the test target as captured on the frame of 35mm Kodak Techpan. The center area outlined in red will be the area shown as 100% crop.
Above it is 100% crop of the center area taken with K20D + Pentax M 50mm macro + bellows to achieve 1:1 magnification.
Above that is 100% crop of the center area with the Coolscan 4000dpi.
Above that is 100% crop of the center area with the D800 + Nikon 50mm f2.8 macro.
The bigger 100% crop on the right is the Pentax K20D + autobellows optically magnified to about 4.5X to show details not resolved by the previous methods. Of course it is important to show that the film target contains more detail then can be resolved, otherwise the test would be invalid.

Of course dslr copying this frame of film is trivial as it takes seconds on my Pentax and Nikon autobellows. I took many shots of it at all apertures and only used the best. Also as you can see, I am also only using the center part which is the optimal performance of any lens. Since the camera and film are attached on the autobellows, there is no vibration issue but nonetheless I used mirror lockup, timer and remote software to run the cameras.

For myself, resolution testing is very simple to do and verify. The real elephant in the room is not the speed and resolution of dslr scanning but the post work required for color negatives specially. I've tried it myself as well as reviewed many who have published their procedures and results and it is nowhere near simple and much more time consuming only to get less then optimal results. Then you still have to manually spot it for dust and scratches! Old Gregg claims with all his experience that it takes him at least 5 minutes per frame and the procedure is not suitable across all the frames in the roll.

 

[Les Sarile]

@runswithsizzers your comment is hinting at the elephant in the room: the major weakness of camera scanning is its high dependence on operator skills. This is true with every technical hobby, and also plays a role with scanners, but to a much lesser extent.

Do you have any comparisons from scanner and dslr copying you can share?

 

[runswithsizzers]

Do you have any comparisons from scanner and dslr copying you can share?

Not sure if that request is directed towards Old Greg or me?

Thanks for providing more details about your testing process - although I still don't know what lens was used on the D800.

When you say, "Older camera with more MP is more applicable in terms of resolution testing," I agree. But I am less certain about older lenses. I have seen it mentioned on a macro-centric website that film-era macro lenses are pretty good, but modern macro lenses may be a little bit better? In <this review> of the Pentax 100mm F4 bellows lens, the author says, "This is a bellows lens that performs well, but less than typical modern 100 mm or 105 mm macro lenses - likely a symptom of the age and simplicity of the design." The same author liked the Nikon 55 mm f/3.5 AI micro-Nikkor better than the Pentax, but again concludes, "The sharpness and resolution are extremely good although modern macro lenses in the same focal length range will very slightly outperform it. The field is moderately flat below 0.5:1 and OK above that, again, not as good as a modern macro lenses in the same focal length range." That author has tested many lenses for close-up work and has posted a list that may be of interest <here>. Notice that he has tested those lenses at various magnifications, so you can select a magnification to see the test results most applicable to your needs.

I do agree with your observation that the biggest problem to be solved when camera-scanning color negative film is the inversion and color correction part. I tried to make some rough comparisons of my early efforts using Vuescan with my Minolta film scanner vs. manual color adjuistments in Photoshop. The results <here> are not at all scientific. Since then, I have started using the Negative Lab Pro plugin in Lightroom. On <this page> I did some more unscientific comparisons between my early attempts with Negative Lab Pro, and the scans provided with processing of a roll of Kodak Portra 160 done by Simple Photography Services (Photrio member, Adrian Bacon). As explained on that webpage, I made some adjustments to the scans, but not a lot. I believe my NLP results were starting to improve a bit towards the end of the roll, but hiring out my color negs to Simple Photography Services is the more time-efficient solution.

 

[Les Sarile]

Gary, regarding the fixed focal lenght nikkor lens used, it is inconsequential since I factored any quality issues by only using the tiny center portion. I don't believe there is a fixed focal lenght Nikkor/Pentax/Minolta/Canon macro lens where that central area is so bad across all apertures. However, I am open to the possibility that I used the one lens that is and would be open to your evaluation of that frame of film. PM me and we can work out the logistics if you are so inclined.

Your results of color negative inversion looks reasonable to me. I have done a few just to get familiar with the process but am afraid I lack the incentive to pursue it any further given the time it takes my Coolscans to provide "better" and cleaner results. Unlike detail and resolution, "better" color results from color negatives is completely subjective unless you are working with known color items as opposed to something in nature or something you remember it to be. If you're interested, we can also try some color tests particularly if you have a known poor result. I've done it every now and then for others on photonet back when I first got my Coolscan 5000 as we used to do a lot of color comparisons across many films, scanners and software. On the first page of this discussion I posted some of the more obvious color failures I've encountered.

 

[NortheastPhotographic]

Does the quality of a photo increase with the resolution?
Irony-modus off

For what do I need approximately 200MP large files? For sharing on the net? Surely not ... For printing ? What do you have for printers at home in operation?
I have access to a pretty good A3 printer; the prints are quite excellent - even with much lower resolution.
'For the wall' I prefer a hand enlargement anyway.

I do wish the camera had a 4 shot native sensor res mode. However if you take the large file produced and downsample it to the native size you'll see a huge improvement over a single bayer capture of the same image. This will only become more apparent if you need to do shadow recovery in a dense piece of film. I've recovered shocking amounts of detail from very dark chromes.

But generally the high res capture is useful because it lessens the need for stitching multiple images to achieve a large file. If you have a 6x6 piece of film you're using very little of the sensor and thus losing the most resolution. If you can create a very large file you have pixels to lose, so to speak, before it becomes an issue.

 

[macfred]

^
@NortheastPhotographic : Thank you for the detailed remarks!

 

[Adrian Bacon]

I do agree with your observation that the biggest problem to be solved when camera-scanning color negative film is the inversion and color correction part. I tried to make some rough comparisons of my early efforts using Vuescan with my Minolta film scanner vs. manual color adjuistments in Photoshop. The results <here> are not at all scientific. Since then, I have started using the Negative Lab Pro plugin in Lightroom. On <this page> I did some more unscientific comparisons between my early attempts with Negative Lab Pro, and the scans provided with processing of a roll of Kodak Portra 160 done by Simple Photography Services (Photrio member, Adrian Bacon). As explained on that webpage, I made some adjustments to the scans, but not a lot. I believe my NLP results were starting to improve a bit towards the end of the roll, but hiring out my color negs to Simple Photography Services is the more time-efficient solution.

ive had a number revisions to my color since you last sent film to me. Btw, if the skies are too magenta for your liking, it’s a pretty easy fix to ever so slightly either increase the green or reduce the red using the LR tone curve for the highlights to nudge it more towards greenish blue and less towards magenta/purple.

interesting comparisons. NLP seems to be too aggressive when it comes to pushing highlights to clipping.

 

[Adrian Bacon]

What remains unsolved IMO, is batch/unsupervised scanning. The tools I have support it, but the results always require additional corrections. For starters, I would love to see a plugin which can reliably auto-crop & remove film rebate from scans.

I think that is a personal preference thing. I have many clients that request that I leave the rebate as a border around their image. I actually had custom scanning masks 3D printed for my setup for exactly this purpose, and my default scan is to include the rebate as part of the scan and put in a non-destructive crop in the generated DNG to exclude it. That way, for those that want it, simply go into the crop tool in the LR develop module and change the crop to include it. It does come at a slight cost to total image resolution, but at 32MP, even after the crop, you still have plenty of resolution for all but the most demanding uses.

and yes, my scans are meant to be starting points. I fully recognize that color is a totally subjective thing. This is why I deliver 16 bit floating point DNG files with a relatively neutral zeroed out look. Maximum flexibility and precision.

 

[runswithsizzers]

@runswithsizzers thank you for a nice comparison! Always fun to look at those, but let's not forget that Adrian's tool and NLP both provide just a starting point for further tweaking. In both cases you're dealing with a DNG with nothing being clipped. NLP offers configurable offsets for white/black points, plus several presets (I recommend "Lab Soft") that are more conservative. Recently I started using Negmaster, and this tool also allows for any final look you want. Basically all of them work just fine, the difference is about personal preferences and user experience. In my mind color inversion for negative scans is a solved problem.

What remains unsolved IMO, is batch/unsupervised scanning. The tools I have support it, but the results always require additional corrections. For starters, I would love to see a plugin which can reliably auto-crop & remove film rebate from scans.

Just to be clear, all the examples shown were post-processed, to taste. That was true for both Adrian's Basic Scans, and for the camera-RAW scans I adjusted with NLP. So my results are highly subjective and limited by my skills. I'm sure there are many users with more capable editing skills that could have produded better results than mine.

I can't say I tried every combination offered by the NLP software, but I did try many of them, and I posted my best effort, as a new user of that software. I often use Lightroom's Gradient and Brush tools to apply local color and exposure adjustments. So for me, the biggest drawback to NLP is not being able to use those tools without first creating a positive image (and the drawbacks that go with duplicate versions of the same photo).

I am looking forward to trying Negmaster.

 

[Alan Edward Klein]

I think that is a personal preference thing. I have many clients that request that I leave the rebate as a border around their image. I actually had custom scanning masks 3D printed for my setup for exactly this purpose, and my default scan is to include the rebate as part of the scan and put in a non-destructive crop in the generated DNG to exclude it. That way, for those that want it, simply go into the crop tool in the LR develop module and change the crop to include it. It does come at a slight cost to total image resolution, but at 32MP, even after the crop, you still have plenty of resolution for all but the most demanding uses.

and yes, my scans are meant to be starting points. I fully recognize that color is a totally subjective thing. This is why I deliver 16 bit floating point DNG files with a relatively neutral zeroed out look. Maximum flexibility and precision.

Why DNG rather than Tiff?

 

[Adrian Bacon]

Why DNG rather than Tiff?

because DNG is Adobe Lightroom’s native format. DNG also lets you store the image as a CFA array instead of full RGB, which means smaller files. DNG also stores all of your edits in the file non-destructively, so you can pass just the DNG file around and all the edits go with it. AND, DNG is in fact a TIFF file.