DIY DSLR Scanner vs Drum Scanner

[JasonF]

Hello,

I really love the idea of the DSLR scanner. But, I'm really struggling with something ...

Film Grain and Aliasing:

http://www.photoscientia.co.uk/Grain.htm

The above link pretty much sums up the issues I am having with my own DSLR scanner. Thus, the idea of using a DSLR scanner sounds great at first. But, this method seems to suffer from the problem of scanner aliasing. And to be honest, I'm not sure what to do about it. And, I'm also still trying to understand it...so bare with me.

From that perspective, I can see the attraction to drum scans for the best of scans. However, my hope, was that a DSLR scanner could at the very least match a drum scanner if not surpass it. Why? Drum scanners are highly expensive (easily a few thousands dollars), bulky, have fewer companies producing them today then in years past, and all around not cost effective for the average Joe. Not all film merits that of being drummed scanned, but when I do, I'd like to be able to have them done in a feasible way. And this is where, I thought, the DSLR scanner would come in and help the average person out. For me, I already have a DSLR and lenses....and that's why I'm testing the waters with it.

What makes the drum scanner most effective over that of a flat bad scanner, DSLR scanner, or otherwise is the PMT and the ability to control the aperture. This is discussed here:

http://www.karimsahai.com/2010/08/scanning-photographic-film-with-a-drum-scanner/



Drum scanners fulfill the same role as any other film scanning device, while relying on an entirely different technology: photomultipliers, or PMTs. In essence, PMTs are extremely sensitive particle detectors tuned to respond to uv, infrared and visible light (photons). The principles at the heart of a drum scanner were derived from research in the field of nuclear physics, in which scientists observe and measure the results and effects of particle collisions. In the case of drum scanners, several photomultipliers are combined to offer the very best scanning device for photographic film. PMTs act as photon collectors and bypass the methodology used in CCD scanners which is to amplify an electrical current. Drum scanners do not suffer from the CCD side-effects that we know: noise & limited dynamic range, specially in shadow areas of your film. The ability for drum scanners to vary their built-in aperture to match the grain size of the film you’re scanning; their ability to scan all film formats including very large exotic sizes and their ability to offer superior batch scanning capability (you can fit a LOT of images on a single drum. Some drum scanners are sold with multiple drums) make them the very best scanner there exists for quality-conscious photographers who want to capture all the range that is encapsulated in their film, at very high resolution.

This ability of the drum scanner to adjust the aperture to match that of film grain seems to be the overriding advantage that it would have over flat bed scanners, DSLR scanners, or otherwise.

My question out there who may know this better than I, is if there is a way for DSLR scanners to overcome the problem of scanner film grain aliasing that drum scanners seem to have the ability to overcome? Using a DSLR scanner, is there any techniques that can be employed that would help to minimize the scanner aliasing issue and therefore pull out more of the picture/image from the film?

Wiki has a section on this called "Grain effect with film and digital":

http://en.wikipedia.org/wiki/Film_grain

In there are two images:

http://en.wikipedia.org/wiki/File:MartinIversenNorway1991.jpg
http://en.wikipedia.org/wiki/File:MartinIversenNorway1991Grain.jpg

The zoomed in detailed one is pretty much the result of what 'I see' in my own images scanning with the DSLR scanner. You get what looks like colorful noise in the image with a lot of grain. Again, this is what I also see and suffer from.

Apparently, if a drum scanner were used, it would be able to pull out more of the image from the picture than the grain. But, I have yet to see some comparisons of this.

For example, controlling the aperture on a drum scanner to match the film grain, what would the resulting image look like over that of a slide/negative scanned by a flat bad or a DSLR scanner?

Anyway, this is what I am struggling with: how to overcome the scanner aliasing issue that drum scanners seem to have over DSLR scanners? Or, if there are any techniques that DSRL scanner users can apply to effectively match the same quality that a drum scanner can perform? Or, for all I know, DSLR scanners by their very nature just can't compete with the drum scanner regardless of how many megapixels the DSLR scanner has? For example, can I conclude that even a Hasalblad H5D-60 with a 60 Mpixels (8956 x 6708) sensor and a good lens could not pull out as much detail from film than that of a drum scanner due to scanner aliasing issues?

Here's a blog that touts the ability of the DSLR scanner:

http://jamiemphoto.com/blog/2013/12/22/yes-your-dslr-really-is-the-best-film-scanner

The title is, ""Yes, Your DSLR Really Is the Best Film Scanner". Hmm, but is a DSLR really the best film scanner? I'm not entirely sure if DSLR scanners suffer from scanner aliasing issues as that is what I have run into as my problem with DSLR scanning approach. Or, maybe the author of this article is applying techniques that are well know by others to overcome this but that I am not aware of. If so, I'd love to hear what those techniques are.

Thoughts?

[Noah A]

I've never tried the dslr scanning method, but I followed the last link in your post, and the author is using a central square to shoot Hasselblad 6x6cm negs with a nikon DSLR, resulting in a 12-16mp image. I can tell you with certainty that drum scanning a Hasselblad 6x6 neg will give you a much higher resolution. It's unclear from the post if the author has ever actually tried a drum scan. He does mention a Nikon Coolscan. I used to scan on a Coolscan and my medium format scans on that machine were also much better than 16mp. The author seems more interested in speed and convenience than in ultimate image quality. That's fine, of course, but since you're concerned about grain aliasing it sounds like you may be more interested in image quality.

Of course folks who are really trying to maximize the quality of dslr scanning are probably stitching multiple frames, but then of course the dslr scanning process won't be quite as fast or convenient. There are some detailed posts on DSLR scanning on this forum somewhere, so you should search them out if you haven't already done so.

The drum scanner has a few real advantages. First, the fact that the film is mounted to a drum ensures proper focus across the frame. Second, the mounting fluid helps fill in small scratches on the film, making them mostly invisible in the scan. Most importantly, as you mentioned, the scanner's aperture adjustment can be used to change the look of the film's grain. It's possible to use a smaller aperture to get extremely sharp results that emphasize grain, or a larger aperture to really minimize the grain. I choose an aperture somewhere in the middle, and the results look very much like the grain I'd get from an analog print.

I'm extremely happy with the results I get from my Howtek HR8000, but again I've never tried dslr scanning so I can't really comment on the comparison. I wouldn't say that drum scanning is difficult, but there is a learning curve to be sure, and it's a skill that could take some time to master. It also can be expensive and time-consuming. But if you're really after the ultimate image quality, it's a proven technology that produces incredible results.

I use an Epson V750 to do preview scans and the drum scanner for my final scans. Depending on how much you shoot (and how many final scans you need to make), it may make sense to do preview scans yourself on a flatbed then outsource the drum scans.

[mdarnton]

I use a camera for scanning. Mine is a 12Mp D300, with micro-Nikkor. For 35mm it's great, much better than a flatbed. At 6x6, it's just a hair behind the scanner because the camera's image is less smeary than the scanner's slightly higher res smeary version), and with the writer's D600, the camera is still probably better than the scanner. For anything larger, the flatbed wins, for my D300. I don't know where the crossover point would be for a D610, or better yet, a D810--somewhere around 4x5, probably. I haven't had any reason to use a drum scanner, but sure, a drum scanner should beat any of the above. Lots of people use flatbeds, but they are not all that great, just cheaper than a drum. However, I rarely make prints larger than 8x10 or the next size up, where I doubt there'd be a gain.

The grain aliasing problem depends a lot on exact grain size vs scanning dot size. It's more of an interference pattern than anything, and so it's dependent on several things. If a flatbed were sharp enough, it would probably have the same problem, but flatbeds just aren't that sharp compared with a camera in a 1:1 faceoff. The reason the scanner wins on larger sizes is simply that my D300 is limited to 2800x4200 pixels, no matter what it's scanning. A scanner, which tops out (actually) around a smeary-looking 2200ppi can't beat my camera on a one inch scan (2200 vs 2800) but it can certainly beat it scanning 4x5 at 2200 per INCH (yielding 8800 total pixels per width on the 5" width) when my camera is limited to 2800 TOTAL, regardless of the width of the neg.

What you do depends on what you need, right? I'm never going to make a 30x40 print, probably. . . .

[Peter De Smidt]

What camera and lens are you using?

Aliasing, an interference problem, can be minimized by increasing the resolution of your scan. With stitching and focus stacking, you can get as much resolution as you'd like. I use a Rodagon D 75mm F4 at 1x for all of my scans, regardless of the format. (At 1x I don't bother with focus stacking. The film is wet-mounted to a very flat plate of glass.) With BW film, even Technical Pan, I found moving up to 2 to 3 times was not worth the effort. I don't have very much experience scanning color with the Dslr scanner, though.

With my Cezanne, a pro flatbed, scanning at 6000 spi gives the best rendition of 35mm grain, i.e. the grain is the smallest and most even it can be. Lower resolutions scans gave noticeably coarser results.

Regarding http://jamiemphoto.com/blog/2013/12/...t-film-scanner, I would be wary of any article that claims that a given technique/system is the "best" without presenting very carefully done comparisons.

[JasonF]

Hello Peter,

Thanks for your input on this. Believe me, I would love to make this work.


What camera and lens are you using?

Currently, I have a bellows setup with my Canon DSLR and a Canon EF 50MM 1:1.8 lens. At first, keep things simple right and start with what I already have an own. With this lens, I am able to get some good magnification. The biggest question, is how much magnification am I getting? This I am not sure yet. And, how do I know when I am pulling too much magnification out of the lens before perhaps the lens isn't resolving as sharply anymore? Not sure just yet about that question either. I might have to do some lens calculations to better understand what magnification I am getting. But that should answer your question as to what type of lens I am using.

I'm of the understand that a lower MM lens will pull a beter magnification then that of greater MM lens. I see you are using a 75MM lens. Wouldn't that just give you more of a working distance? For film, it doesn't hurt to be close to the film. But, I need to first understand the pro's and con's there too.


Aliasing, an interference problem, can be minimized by increasing the resolution of your scan. With stitching and focus stacking, you can get as much resolution as you'd like. I use a Rodagon D 75mm F4 at 1x for all of my scans, regardless of the format. (At 1x I don't bother with focus stacking. The film is wet-mounted to a very flat plate of glass.) With BW film, even Technical Pan, I found moving up to 2 to 3 times was not worth the effort. I don't have very much experience scanning color with the Dslr scanner, though.

Great suggestions Peter. Do you have a collection of scans posted anywhere on the net of this scan-stich technique? Right now, I have only begun to explore the possibilities of DSLR scanning, but I have spent some time first doing some reading to see how others have been approaching this issue before attempting this path. But, I have not been satisfied as to what those techniques are to improve and reduce film grain ailising. And, also, because I seem to suffer from seeing it highly in my DSLR scans.

For me, my interest is in blank and white film, positive color slides, and negatives. And among these, I have various formats from 35mm and smaller to large format film.

For film, I am very much interested in stitching, but don't see much use for focus stacking considering that film lies along a plane. In this case, that simplifies things to where at the very least I can take focus stacking out of the picture.

You said that you found moving up to 2 to 3 times was not worth the effort? When you say 2 to 3 times are you talking about a 2x or 3x magnification?

You also made another good point that increasing the resolution of the scan would help better deal with the aliasing issue. Here's what John Striver said over at http://photo.net/film-and-processing-forum/00VDBX



The reason to use higher resolution in scanning is so that the lens serves as the required optical low pass filter. The grain on Ektar 100 may be so fine it's not a risk of grain aliasing. The risk comes when the grain is about the same "dot pitch" as the scan is at, that's when you get the "beats" that cause grain aliasing.

I've read this before that higher magnification would help which you mention here too. Recently spoke with a friend. He has a better lens that should be able to achieve a higher magnification. But, I'm not entirely convinced yet that would help much. But, I will borrow and pick it up and give it a try.

I've also read that it's better to scan at a higher resolution and then down-sample by about 50% to sharpen up the image. I will try these two approaches here shortly.


With my Cezanne, a pro flatbed, scanning at 6000 spi gives the best rendition of 35mm grain, i.e. the grain is the smallest and most even it can be. Lower resolutions scans gave noticeably coarser results.

So what are your thoughts on a higher magnification and then stitching since also having used a Cezanne? I'd like to try 1x, 5x, and 10x if possible and see if higher magnification and then down-sampling would help much? Perhaps you have a better perspective on this or may have already tried?


Regarding http://jamiemphoto.com/blog/2013/12/...t-film-scanner, I would be wary of any article that claims that a given technique/system is the "best" without presenting very carefully done comparisons.

I am wary yes. It's like "Hey trust me, use a DSLR. But I won't show you any of my techniques as to how I am pulling high quality scans using a DSLR". And, from my initial tests, I'm excited but at the same time a bit disappointed. I find hope in scanning and stitching, but in my scan test I suffer from film grain aliasing.

But all talk and no photo's make it less fun right? I'll have to try to get some captures together of my aliasing issues and post them on here. Then you can see for yourself what I suffer with. Then, I can try a few different lenses too to see if that helps to resolve the film grain aliasing issue.

[Peter De Smidt]

The easiest way to find out your magnification is to photograph a mm ruler parallel to the long dimension of your sensor. Compare that to your sensor size. In my case, that's 35.9 x 24mm. So when I have 35.9 mm when the image is sharply focused, then I'm at 1x magnification. I use the APO Rodagon D75mm F4, as it's the best lens for 1x magnification that I can afford. (http://www.bhphotovideo.com/c/product/44495-USA/Rodenstock_452345_75mm_f_4_APO_Rodagon_D.html). It was specifically designed as a slide duplication lens at 1x. (There's also an f4.5 version, which is optimized for 1/2x forward and 2x reversed.) The only lenses that are obviously better that I know of are the rare Printing Nikkors, and they cost thousands of dollars.

It'll be interesting to see what magnification your lens is at. It's a standard lens, right? Non-macro? If so, are you using it reversed?

Since you have Canon, check out Charles Krebs EFSC article at: http://www.krebsmicro.com/Canon_EFSC/

Regarding higher resolution and stitching, I stitch all of my medium and LF dslr film scans. Focus Stacking would add a lot of complication to the setup for large films. It would require another stepper axis on my machine, increased programming complexity for the robotics, and more complicated post processing. As I said earlier, with my system, I didn't see much, if any, advantage going higher than 1x, but that's with a very high quality lens at f4. At 1x, the depth of field of the lens is about 6 thousandths of an inch at f4. As you go up in magnification, that gets smaller. Down-sampling would help quality at the cost of a smaller file. I may go higher than 1x with 35mm in the future, but only if I can see a real increase in quality. I bet that people regularly overestimate the resolution of their film images.

[SergeiR]

Something to think about. IMHO.

- your dSLR will at best be 1/3-1/4th or mp it claims (unless you got Fuji, or true achromatic camera but its other kettle of fish) with infamous RGGB pattern of sensors. so you got whopping 4mp worth of resolution out of 12.
- even if you will get 12mp, its barely 3000x4000px. if you scan film @2400 dpi on flatbed it equals tooooo… right. barely over 1inch to barely under 2 inch scanning area. Considering that downsampling is possible we can say that you will have to go 4-8 times down on each side (4x5 film , 8x10 film judged). And we do talking true RGB pattern sensor btw (i.e array resolves single pixel for all 3 color values, not 4 pixels as in dSLR). Doesn't matter how smudgy - they still will be better. Tonal gradations still will not resolve fully in dSLR (12/14bit vs 24-48 bit will still loose).

film flatness, lighting, reflections, actually good micro-lens (which, by the way will detiorate quality at f16+ on your typical 35mm and sub - dSLR ).

If you can afford drum - get drum. If you can't - short of speed (but then if you going to do sliding back shots and glue panoramic results aka stitching - won't be much faster), even flatbed wins.

[Peter De Smidt]

My experience differs from Sergei's, at least with BW film. I get significantly better scans with a dslr than with an Epson flatbed, both in resolution and dynamic range. (I measured about 1000 dpi more resolution and at least 1 stop more dynamic range.) Comparisons, including those with an Edmund Scientific high resolution chrome on glass target, were posted in the dslr scanner thread years ago now. Some of the files aren't up anymore, alas. This winter I hope to have more time to do some tinkering and comparisons, but for the moment I don't have the time.

Btw. a good comparison of top scanners, by someone who owns and uses them, is message #42 at: http://www.largeformatphotography.info/forum/showthread.php?106507-Recommend-a-Good-Flatbed-Scanner/page5

[SergeiR]

My experience differs from Sergei's, at least with BW film. I get significantly better scans with a dslr than with an Epson flatbed, both in resolution and dynamic range. (I measured about 1000 dpi more resolution and at least 1 stop more dynamic range.) Comparisons, including those with an Edmund Scientific high resolution chrome on glass target, were posted in the dslr scanner thread years ago now. Some of the files aren't up anymore, alas. This winter I hope to have more time to do some tinkering and comparisons, but for the moment I don't have the time.

Btw. a good comparison of top scanners, by someone who owns and uses them, is message #42 at: http://www.largeformatphotography.info/forum/showthread.php?106507-Recommend-a-Good-Flatbed-Scanner/page5

Could be just b/c i am far less industrious than you :) I never attempted to build fully devoted station and stuff for such scanning.

[onnect17]

I would say the resolution of any CCD based system should be listed as a range or a "not less than" number. Also they should be tested with different levels of contrast.

The tricky part with the DSLR scanning is the assembly of the images. No lens is perfect and it's very likely there's some barrel distortion bigger than half pixel in most of the image so the only way to get a decent image is to either deal with a good amount of math or significantly overlap the shots. On the top of that chromatic aberration and vignette are issues to deal with.

Not sure if the current SLRs are including software to calibrate those parameters with non manufacturer lens.

[Peter De Smidt]

Using a template in PTGui gives very good stitches.

Resolution is complex. In what direction? At what contrast? And so on. For instance, my Cezanne has higher resolution in one direction than in the other. I don't have the time or inclination to run tests to counter every conceivable objection.

The results of my dslr BW scans are significantly better that I achieved with a consumer flatbed, and I spent a lot of time optimizing those flatbed scans. The dslr scans are almost as good as I can do on my Cezanne. The limitation is my D600. Daniel got over 5000 dpi results with a D7000, Rodenstock Magnagon and a high res target. The targets are available. Interested parties can buy one and test their own scanner. What do you get when you set the scan size to 4x5"?

[onnect17]

Resolution in CCD based systems should be higher in the center and lower and equal on the corners. If there is any difference between corners at the same distance of the center then there's an alignment problem.

I own both, the Magnagon 75mm f5.6 (fixed) and the Apo-Rodagon 75mm f/4 D. At 1:1 and f/8 the later should perform as good as if not better than the Magnagon for color applications. The only advantage of the Magnagon is the bigger circle of image, less falloff, and better performance at multiple magnification (contrary to the 0.8X to 1.2X range in the Apo-Rodagon). In any case, both lens can resolve a lot more than the capabilities of the sensors in those cameras so the resolution itself will be limited by the magnification.

I know this does not apply to your BW scanning (assuming you are using an white light source, of course) but others scanning color via DSLR should keep in mind the sensors in question are designed using the Bayer pattern so the real sampling is half and 3/4 of the megapixels are product of interpolation. So in order to sample for real an area in color at 5000 dpi with the D7000 sensor it's necessary to use a magnification of almost 2x. (209 pixels/mm divided by 2 ==> ~105 pixels/mm ==> ~2650 pixels/inch).
So let's say we want scan a 35mm negative with the D7000. Half of the sensor size is 23.6mm/2= 11.8mm ==> 36/11.8mm = ~3. Then 3x3=9 so it will be necessary to take at least 9 shots without overlaping to complete the frame.

This is not an issue with tricolor or monochromatic linear CCDs used in some professional scanners. Not familiar with the Cezanne but it seems to use the same Kodak chip as the Imacon.

[Peter De Smidt]

And this discussion helps the original poster in what way? All we know is that he is getting more "grain" in his scan than he'd like. If it truly is due to grain aliasing, we know that increasing the resolution of his scan will help. Neither Daniel nor I have noticed this problem with our dslr scans. While I mainly scan BW, I have scanned color slides and negatives, and Daniel mostly scans color. A priori complaints about Bayer sensors just muddy up the issue. Why not post them in the appropriate dslr scanner thread? It would be: http://www.largeformatphotography.info/forum/showthread.php?87535-DSLR-Scanner-Negativity-post-negative-thoughts-about-the-DIYS-Project-here&highlight=dslr+scanner Or better yet, why not hold off on that until you have some empirical tests that support your theorizing?

Both Daniel and I have posted the scans of our resolution targets. They give clear results in two directions. That is the basis for our claims. We have checked both the center and all corners of the frame.

[onnect17]

I don't think there's any "theorizing" in my post. The numbers are taken directly from Nikon specs.

But talking about theory, how this could help the original post? Very simple.

He has to know the real sampling frequency to avoid getting close to the size of (or a multiple of) the grain in the emulsion. This is very important with Tgrain type emulsions (most of the modern emulsions) where the grain size is very even. A very common approach to address the problem is to reduce or increase the sampling resolution by 50% where possible.

He should know every single aspect of the technology in the camera that could impact the resolution in the image, Bayer, antialiasing filter, etch.

He should be able to recognize the grain. Many times there is some degree of sharpening applied to the image without the user being aware. The grain in the color film is red, green and blue. If you see anything else then check for unintended sharpening or processing.

I am glad to hear you and others are happy with the results being obtained with the DSLR based scanners. However it was not being totally happy with the results the reason why I explored many scanning technologies until I settled with drum scanners, and the cost for most of them was between $500-$1000, which I guess is lower than the cost of a D600.

[Peter De Smidt]

"He should know every single aspect of the technology in the camera that could impact the resolution in the image, Bayer, antialiasing filter, etch."

So the same applies to you before you do anything? You understand _everything_ before doing anything? So you get a lot done?

The theorizing isn't in the regurgitation of Nikon info. The theorizing is in how that information impacts scan quality, and in particular how that info is a problem (or leads to a solution) in this case.

Again, to summarize:
1) The poster has a problem with excessive "grain."
2) Other people with dslr scanners with Bayer pattern sensors don't have this problem.
3) Therefore the problem isn't with Bayer pattern sensors in general.

The proper way to approach this problem is clear. First, we need to find out how the original poster is doing his scans. For instance, it doesn't look like he's using a good lens or magnification for this task, but more info will clear that up. With that info, the answer might be obvious. If it's not, he can send some film to me. I'll scan it on both a dslr scanner and on a Cezanne at 6000 spi. That would tell him whether he wants to pursue dslr scanning or not. If someone else wants to offer up their valuable time and scan the film on a drum scanner, well, that'd be helpful.

The dslr scanner project has taken years to get to this point. All along the way, critics would pop up repeating the same old complaints. Then we showed samples, and they went away for a little bit.... Does this mean that dslr scanners are without flaws? Absolutely not. But the scans achievable are clearly better than what's possible with an Epson flatbed, which isn't to say that they can't give good enough scans for some purposes.

It's similar to the situation with drum scanners and pro flatbeds. Some people dismiss the flatbed scans, as PMT scanners, they hold, are inherently superior. But then you have someone like Tim Parkin acquire a large number of the scanners and use them. He finds out the some of the flatbeds are better in some regards than the drum scanners.

Here's another example. I bought a D200 when they came out. Over the next few years, lot's of theorists bemoaned increasing resolution, mostly because they enjoyed bitching about something, but in part because the sensor sizes were getting smaller, and it was "a law of physics" that smaller sensors would lead to less sensitivity, less dynamic range and more noise. Well, a while ago I bought a d600. Despite have much more resolution than the D200, it has significantly better sensitivity, dynamic range and freedom from noise. Obviously there's more to the equation than the sensor size. Experience trumps theory.

[onnect17]

On the contrary. My approach to scanning devices is that I never know enough, so I try to learn as much as possible about the technology behind, sometimes bringing the equipment near the danger zone.

There is not such a thing as a perfect scanning device. The more we learn about the shortcomings the easier it's to judge the resulting image, and many times theory and heuristics challenging each other. It also helps setting expectations. I use it positively and have a lot of fun.

Perhaps Jason is confusing large size grain with grain aliasing. The first link shows a high speed film with the expected large size grain. I am unable to see any aliasing issue in the image. The lens is not the issue, for sure. The 50mm f/1.8 is not a macro lens but it’s good enough for the tests. He’s more than welcome to get in touch via PM.

[Peter De Smidt]

Jason,

What light source are you using? At one time I had a Coolscan V, which had a fairly small LED light source. With fine-grained films, it thoroughly trounced my Canonscan 9950F flatbed, which had a large and diffuse light source, but with course grain film the Nikon exaggerated the grain tremendously. So with things like Kodak HIE, a very coarse-grained film, the Canonscan did a much better job.

[JasonF]

A lot of great questions! When I get a better chance (which I don't have at the moment), I'll answer all these questions. I plan on posting a few scans so you can see what I see. And, I'll be provide a lot more detail on this soon as to what I'm using too. Then you can help me determine what the issue is that I seem to face.

[Bruce Watson]

...my hope, was that a DSLR scanner could at the very least match a drum scanner if not surpass it.

ROFLMAO! Thanks, I needed that!

There's always hope. My hope is that a rich relative will leave me millions, that my curious genetics will suddenly become "normal", that politicians and corporations will come to their senses and decide that "endless war" is a really stupid policy, and ... well, there's always hope.

I guess you're hoping for a change in the laws of physics. Good for you; there's always hope.

[Peter De Smidt]

Yeah, I have to agree with Bruce on that. My goal was to better an Epson scanner. No problem. But the dslr scanner is not as good as my Cezanne, let alone as good as a drum scanner when it comes to d-max.

[JasonF]

Yeah, I have to agree with Bruce on that. My goal was to better an Epson scanner. No problem. But the dslr scanner is not as good as my Cezanne, let alone as good as a drum scanner when it comes to d-max.

Oh well, so much for the encouragement.

[Peter De Smidt]

I'm not sure what the problem is. So you can make a very good scanner with a dslr, but you might not be able to better every other scanner in existence, including ones that cost 10's of thousandths of dollars. Sure, used pro scanners don't have to cost that much, but they also have their issues, such as requiring old computers to run, the cost of parts and maintenance, assuming they can be fixed at all, the use of software from the 1990s...... The older pre-press drum scanners, the big Heils and similar, were meant to be serviced regularly. So, yes, you can spend $1000, $2000....$x000 on an old pro scanner, and you might get a good scanner. (Even if you spend $100,000 on a new Heil or $35,000 on a Premier, assuming you can still buy either of these new, you won't be getting a scanner that's the best at everything, as Tim Parkin has demonstrated.)

In any case your results depend on the maximum density of your film. If you're using Velvia 50 or some other high density film, then it's likely that a good drum scan will produce the highest quality. If you're using BW negative film with standard density, then most scanners, including dslr scanners, will have more than enough range. In addition, it seems that people tend to over estimate how much detail they capture on film.

I prefer my dslr scanner to the consumer flatbed and Nikon Coolscan V that I used to have. It cost me about the same to build as a new Epson. And doing so was fun, but it's not perfect.

[Daniel Moore]

I'm a little late to the party.

Jason, please post an image of your problematic DSLR scan results and let us know your magnification when you can.

As for Dmax/resolution, DSLR's improve with each passing year. Only time will tell how things will turn out comparison wise from one year to the next.

I'll reiterate one of my favorite aspects of DSLR scanning, the ease with which I can arrive at (finally!) pleasing color in color neg. Slide film still throws me a bit when the Dmax is extreme, I'm not sure if that's down to my monitor though (Eizo Nanao CG241W), which is slightly deficient in the deepest blues where I have trouble. One day we'll be able to actually see on screen all the colors a DSLR can capture, maybe. : )

Epson's releasing a V850 soon, it's been many years since the V750, it'll be very interesting to see what they have come up with in that time.

I want to make one small correction to Peter's reference to a D7000 scan I did, it was with rather with a D7100, 16 mp vs. 24 mp.

Jason, you should try and mock up a station where you can manually stitch at higher magnification, my hunch is your problem will be mitigated to where you won't notice it anymore. Short of that, prove to yourself that shooting at 1:1 M reduces the issue.

Keep us updated. There's a lot of room for development in DSLR scanning and we all benefit from it's study.

[Peter De Smidt]

I"m glad that Daniel chimed in. I'll add another benefit with dslr scanners, and that's the easy ability to tailor image capture to each negative. With most scanning software it's very difficult to see what adjustments change the data captured by the scanner as opposed to what adjustments are just software manipulations of that data. With a dslr scanner, there's no ambiguity. One can tailor the capture via light source adjustments and shutter times. In addition, you can upgrade your scanner easily. Perhaps you have a low budget starting out, then something like a Micro Nikkor 55mm could do the trick for very little. Later on you can always add a more expensive lens, a different camera, ... Heck, you could add a stepper on the Z-axis and do focus stacking, which would allow you to work at 5-10x magnification, although that would be overkill in my opinion.

[sergiob]

Judging just by naked eye, no need for scientific tests, my Pentax 645D + Pentax 120 f4 macro lens scanning rig beats any flatbed I 've used. And it is much faster to use to. I can scan a full roll of 35mm in around 15 mins. The only glitch to my workflow is in Lightroom, where I have to invert the curves to see the photos as positives. Once inverted, the sliders will also be inverted, and they are counterintuitive and clumsy to use so. If I chose the Photoshop route then I guess that wouldn't be a problem.

[Will Frostmill]

Judging just by naked eye, no need for scientific tests, my Pentax 645D + Pentax 120 f4 macro lens scanning rig beats any flatbed I 've used. And it is much faster to use to. I can scan a full roll of 35mm in around 15 mins. The only glitch to my workflow is in Lightroom, where I have to invert the curves to see the photos as positives. Once inverted, the sliders will also be inverted, and they are counterintuitive and clumsy to use so. If I chose the Photoshop route then I guess that wouldn't be a problem.

Over on Luminous Landscape, Mark D Segal & Todd R Shaner have an article on DSLR scanning, with two different methods for dealing with inverted curves. There's a pdf and a downloadable photoshop action. Link: http://www.luminous-landscape.com/essays/scannerless_digital_capture_and_processing_of_negative_film_photographs.shtml

[Daniel Moore]

Thanks for that link Will, it's a great write up. Page 34 of their pdf has a processing method comparison I found especially interesting, showing subtle but significant tonality differences.

Marks capture setup, a dangling camera really, is gutsy and is an example of how far one can push the idea.

[timparkin]

Interestingly this shows the problem with flat beds and dslrs as scanners - if you boost the shadows you see the amount of halation or flare from the light to dark areas 125983

Drum scanners get around this by effectively masking the whole frame apart from the one pixel they are analysing.

The bayer array should be thought about too - effectively the colour resolution is less. You may not notice this unless you see a side by side shot between a drum scanner and a dslr. What you'll see is a loss of fine colour detail (i.e. you might lose fine specks of reds such as berries in trees etc).

Metameric failure is also an issue with DSLRs (the filters they use are 'compromised' because of the need to get high iso response and also the subselection of materials you can use that can be transformed into pixel size filters. This may cause relative colour shifts of certain dyes that may be difficult to correct (I can't say I've tested but I know flatbed flourescent light sources can cause all sorts of issues with fuji film stock).

Not saying you can't get acceptable results from a DSLR and you can probably exceed most flatbeds but a good drum scan has many advantages - some more subtle than others.

Tim

[timparkin]

Happy to scan a test on my Howtek 4500, Screen Cezanne Elite Pro and Heidelberg Primscan.

It's interesting that I always thought Portra 400 had more grain than Portra 160 until I scanned them both on a Screen Cezanne. Turns out that it was probably an aliasing issue (which happens with drum scans and flat beds etc). The higher resolution of the Screen Cezanne combined with whatever other design decisions they made make a very smooth scan from Portra 400 - just about the same as Portra 160. So I think you need 5500dpi + to over sample grain (at least on new Portra which is quite a large grained film to begin with).

Tim

[Peter De Smidt]

Hi Tim,

Thanks for chiming in!

You're comments are spot on. The Stouffer step wedge was photographed with a 40 year old lens and without a lens shade, but even a modern lens with an application specific lens shade there will still be some flare.

I have a Cezanne, and I agree that scanning at a high resolution, approximately 6000 spi, gives very smooth scans with regards to grain.

Daniel and I haven't done much with the dslr scanners lately, as that's more of a winter activity. After some refinements, we'd be happy to scan some negatives and then send them off to you for a comparison.

[timparkin]

Hi Tim,

Thanks for chiming in!

You're comments are spot on. The Stouffer step wedge was photographed with a 40 year old lens and without a lens shade, but even a modern lens with an application specific lens shade there will still be some flare.

I have a Cezanne, and I agree that scanning at a high resolution, approximately 6000 spi, gives very smooth scans with regards to grain.

Daniel and I haven't done much with the dslr scanners lately, as that's more of a winter activity. After some refinements, we'd be happy to scan some negatives and then send them off to you for a comparison.

Sounds good - I can send you a couple to scan as well (from the Big Camera Comparison test we did)

It seems flare is inevitable - clear parts of the film itself diffuse light slightly and hence cause a 'glow' around dark edges. You get the same with flat bed scanners too (but only in one dimension). My Fuji Lanovia is stunning at getting shadow detail out of chromes but ruined by flare causing variations in dmax across a frame that are really hard to even out.

Tim

[Kirk Gittings]

http://www.luminous-landscape.com/essays/scannerless_digital_capture_and_processing_of_negative_film_photographs.shtml

[Daniel Moore]

Is anyone making, or planning to make drum scanners?

Imagine. A digital capture device using a shifting sensor to deal with the interpolation required by Bayer sensors.
Thank you, Peter for sharing this (albeit rumored):

http://www.dpreview.com/articles/2323618506/rumored-olympus-om-d-e-m5-successor-may-offer-resolution-boosting-sensor-shift

Time marches on.

What's next.

And yes, I'm making assumptions, but you can bet they're coming, can't you. - Improvements. They're coming.

[Will Frostmill]

Sounds good - I can send you a couple to scan as well (from the Big Camera Comparison test we did)

It seems flare is inevitable - clear parts of the film itself diffuse light slightly and hence cause a 'glow' around dark edges. You get the same with flat bed scanners too (but only in one dimension). My Fuji Lanovia is stunning at getting shadow detail out of chromes but ruined by flare causing variations in dmax across a frame that are really hard to even out.

Tim

What if the light is polarized? Would you get less flare in one direction? With a flatbed you can scan textured photographs twice, rotating 90 degrees between scans, and sum the images in Photoshop, knocking out the differences. If we got less flare in one direction, the same technique could be employed, yes?

[Peter De Smidt]

Interesting idea, Will.

[Will Frostmill]

Thanks, Peter! :)

[mongole]

I know, onnect17s remark (in post 10) about barrel distortion is some months ago, but maybe this is interesting as well.

Adobe Lightroom offers a very good barresl distorsion correction and also correction of light fall off to the corners. It's available for all "known" lenses, but you can also create your own correction profiles with their software: http://www.adobe.com/support/downloads/detail.jsp?ftpID=5490

[Daniel Moore]

That would be helpful for single captures.

As for captures intended for stitching, distortion is actually not an issue since the very act of stitching requires that the lens distortion be characterized. In essence each time control points are generated these factors are determined and accounted for in order to be able to stitch them. Algorithms handle the vignetting or every stitch made would look ghastly. Where there are not enough details to allow for this profiling (large expanse of sky for example) a template can be use with known valid data from a previously saved, highly detailed stitch solution.

The lenses Peter and I are using are of very high quality and I have not seen any chromatic aberration. Even if there were some it can be dealt with in the Raw processing stage by many converters. It's possible that stitching programs can handle CA as well but I would prefer to handle it at the Raw stage if present.

[mongole]

I see, sound's reasonable. Otherwise one would need to "fix" the images on every panorama and the like...

[8x10 user]

Ideally the size of the grain cloud times reproduction ratio should equal the sensor size. Multshot cameras should be good as it gets rid of the bayer grid issue and you can sub sample between the pixels which helps due to the stochastic nature of film. I could see a bayer grid causing issues. If your magnification ratio is too high for the film/sensor then you can get grain aliasing. Grain aliasing can be removed digital by programs like noise where however this negates the benefit to the increased resolution unless the resolution is many times higher then the grain clump so that the clump appears more like a halftone pattern in need of descreening. With negative film grain aliasing and natural noise/grain can be extenuated taking a small section of the cameras raw histogram and expanding it so much. This causes a combed histogram and can posterize the grain or noise.