DSLR Scanner: Scans and Comparisons

[Peter De Smidt]

The head of our European research branch has requested a thread to post scans concerning dslr scanning, which includes comparisons to other methods, and so here it is.

[cabbiinc]

When I considered scanning film with a camera a few thoughts came to mind.

a) When scanning with a dSLR vs a dedicated scanner you have to take into account lens distortions. That isn't to say that a scanner doesn't have lens distortions as well, but it's a much more controlled environment. This is something that can be corrected for, but would be one more step that may be missed or more likely done improperly.
b) Size limitation. Most dSLRs are under 20mp. But I can scan 4x5 film at far more pixels than the current crop of typical dSLRs can afford.
c) Dust removal. Even with diligent dust control measures you'll still find dust. Having a dedicated scanner with proper dust removal is, in my opinion, worth the extra cost and time it takes to scan.
d) Size. Even a medium or large sized scanner will likely take up less room than a copy stand type setup realistically.

On the other hand...
1) Time. It takes much longer to use a flatbed or dedicated film scanner to scan even 35mm film than it would to just *click*. Is time worth more than the reasons stated above?

These are just my opinions. Sorry, I don't have any examples of one vs the other.

[Peter De Smidt]

This is not the proper thread for this sort of discussion.

[Old-N-Feeble]

Huh?

[Peter De Smidt]

This thread is for showing scans.
Such as:
http://i955.photobucket.com/albums/ae37/peterdesmidt/Lighthouse_Cezanne.jpg

Versus this:

http://i955.photobucket.com/albums/ae37/peterdesmidt/LightHousePTGui.jpg

The first image is from a Cezanne scan. The second is from my dslr scanner.

The original thread is the place for general discussion of the project. It is: http://www.largeformatphotography.info/forum/showthread.php?84769-Making-a-scanner-with-a-DSLR

[Peter York]

Peter, what were the scan settings for the Cezanne? The DSLR scan is excellent. Looks very promising!

[Peter De Smidt]

Hi Peter,

4000 spi.

To me, it looks like the AN holder held the Cezanne back.

[Peter York]

I think you are right. To my eye its most apparent in the rounded bits, but it shows everywhere (ugh, back to testing). I wonder if the differences are visible in print (and at what size).

A DSLR scanner that reaches 4000 spi would be fantastic. Have you "scanned" an entire 4x5? I'm wondering what the time investment per image is.

[Peter De Smidt]

I used the Cezanne clam shell holder for the dslr at first. A close-up view of the grain showed smearing, and so I ditched the holder and taped the negative emulsion up to the thick glass plate. That's what gave the scan above. The Cezanne scan shows the same smearing. Soon I'll scan a negative on the Cezanne without going through a sheet of AN acrylic. It should perform better.

Note that these prints would be huge, on the order of 50" prints from a 6x7 negative. With a computer and the ease of looking at a 100% pixel view, it's so easy to become obsessed about very minor things.

I haven't scanned an entire 4x5. So far I"ve stuck to 6x7cm negatives, just to make things easier. Joseph did a 4x5" negative, if I remember rightly. Right now I'm concentrating on automation. The team made some good progress today. We've narrowed down stepper motors, drivers and software candidates.

[Kirk Gittings]

I sure appreciate your efforts Peter.

[sully75]

yeah me too. Not sure how much energy I would put into fiddling with building one, but if it were something that were a sure thing, more or less, I'd be all over it.

[Peter De Smidt]

Thanks Kirk!

Sully, the goal is to come up with a fairly easy to replicate system. In particular, I'm trying to stay away from a design that requires expensive tools. For instance, to make my current prototype, you'd need a decent and well-aligned table saw, but it'd be fairly easy to make something similar out of alternate materials, such as threaded steel pipe, with only wrenches and a drill, at least you could for the camera support stage. Once we get a fully working prototype we can do some sample scans for people. That way you should be able to see exactly what is involved and what the results are. It certainly won't be for everyone.

[jb7]

They look very clean Peter, and remarkably grain free-

These are filthy by comparison, but they're tests of the process, and not to be seen as final output-
from the DSLR Scanner No.7 thread-

Details are screenshots from V750 4000dpi scans overlaid on the DSLR Stitched output, about 600kb each-
I re-stitchsd the original; I found a few anomalies, on closer inspection-


71015


http://img209.imageshack.us/img209/1681/v750jb7dslrcomparison1.jpg

http://img220.imageshack.us/img220/8463/v750jb7dslrcomparison2.jpg

http://img717.imageshack.us/img717/5971/v750jb7dslrcomparison3.jpg

http://img714.imageshack.us/img714/1904/v750jb7dslrcomparison4.jpg

[Peter De Smidt]

That's a very nice comparison, Joseph.

[cabbiinc]

This is not the proper thread for this sort of discussion.
Sorry, I didn't see the other thread and you hadn't received a reply in a few days. Thought I'd make a comment.

[Adrian Pybus]

These comparisons look really promising.

[Peter De Smidt]

https://www.dropbox.com/s/4mgibsf4k7o651w/Mill_test.tif

Compare the two layers. Thoughts?

[Jody_S]

They look very clean Peter, and remarkably grain free-

These are filthy by comparison, but they're tests of the process, and not to be seen as final output-
from the DSLR Scanner No.7 thread-

Details are screenshots from V750 4000dpi scans overlaid on the DSLR Stitched output, about 600kb each-
I re-stitchsd the original; I found a few anomalies, on closer inspection-


71015


http://img209.imageshack.us/img209/1681/v750jb7dslrcomparison1.jpg

http://img220.imageshack.us/img220/8463/v750jb7dslrcomparison2.jpg

http://img717.imageshack.us/img717/5971/v750jb7dslrcomparison3.jpg

http://img714.imageshack.us/img714/1904/v750jb7dslrcomparison4.jpg

This looks very promising indeed. I've been considering switching to dSLR since I keep having trouble with the density of my negs with my cheap HP scanner. I was surprised to find less dust with my dSLR setup than I was getting with the HP. So if I'm going to invest $1K for a new V750, I want to know if I'm better off building a rig for stitching with my dSLR. If the premise of this thread is to let the images speak for themselves, then it is a very helpful idea.

[VictoriaPerelet]

Peter, I haven't visited this web site for awhile. Here's crop from 24000DPI DLSL scan that I made by "reverting" DeVere Enlarger:
Original image (and number 50 outlined in red square at lower part of chrome):

http://www.victoriasphoto.com/Notes/Enlarger_Scan/scan.jpg

Here's 1x1 (direct from camera sensor) detail:
http://www.victoriasphoto.com/Notes/Enlarger_Scan/1x1_enlargement.jpg

You can see film artifacts. Described in this thread (http://www.largeformatphotography.info/forum/showthread.php?48057-scanning-with-dslr-part-III-Extracting-gigapixes-in-your-darkroom) few years ago.

Victoria

[Peter De Smidt]

Interesting, Victoria. Thanks for posting. I don't think I saw that thread before.

[Peter De Smidt]

So here are some comparisons.

First, we have a scan from the dslr scanner. The layers are from different lenses.

https://dl.dropboxusercontent.com/u/3595413/Lens_Comparison_1x_Bergger_200.tif

This is a full-sized 8-bit Gamma 2.2 layered file. The sample area was the center of a 4x5" of Bergger 200 film developed in PMK. It was taped to the glass. With one lens, the exposure was 1/100th of a second, and with the other one it was 1/60th. I used the light source to roughly equalize the channels.

The files were brought into Lightroom 5, where I inverted them. I then used the curve feature to stretch each channel until it almost clipped, which should be pretty representative of actual use. Default Lightroom sharpening was applied to both. (Each could stand quite a bit more sharpening.)

After bringing the files into Photoshop, I converted to Grayscale 2.2 using the green channels, which showed a little more detail.

Second, we have a scan of the same area on my Cezanne:

https://dl.dropboxusercontent.com/u/3595413/Cezanne_Bergger_test-0109.tif

I scanned this version at 8000 spi and then down-sampled to roughly the size of the dslr scan in Photoshop using Bicubric sharper. No other sharpening was applied.

[Ken Lee]

I'm late to this thread and no expert with real scanners but the two are not dramatically different to my amateur eyes.

I can't tell if we're seeing grain or digital noise in some of the darkest areas, but both scans seem to getting right down to the practical limit. Some parts look better in one file than the other, much like the example you already posted: it's hard to tell whether what we're seeing is due to sharpening or actual resolution.

I'd call it a draw myself. Is the DSLR version much cheaper and easier to make ?

Would it be more revealing to scan something with finer grain, like TMX ?

[Peter De Smidt]

I appreciate your opinion, Ken.

With the dslr scans, digital noise would appear first in the brightest tones, as those are the densest areas of the negative. There was a very small amount of sharpening applied to the raw conversion in Light room, using it's standard default setting.

Easier to make? That's a hard one to quantify. If you're scanning 35mm, then the dslr scanner is much faster than the old pro flatbed.

My dslr scanner takes up about 1/4th the space of my Cezanne. Over time, the big flatbeds are going to get even harder to find and more expensive to keep running. Minus the dslr, my DIY scanner costs less than new bulbs for my Cezanne.

I used the Bergger film, the grainiest I've used in large format, for a couple of reasons. It lays flat, it makes a very good optical print, and I wanted to see how the systems handle grain. I don't want to get rid of grain. For my small format work, I often like grain, but I prefer it to appear equally sharp and close to what it does on the negative. Many scannessr exacerbate the grain dramatically, which was my experience with a Coolscan film scanner and grainy film. With larger format, I generally don't want to see grain in a print, but if the grain is well imaged in the scan, I expect that the overall quality of details in the negative should look good. I'll do some scans of finer-grained film down the road.

[Ken Lee]

Perhaps the scanners which exaggerate grain (like the Nikon Coolscan) use a collimated beam of light, like a condensor enlarger. They give an impression of greater sharpness but introduce the Callier Effect: dense areas rendered disproportionately dense, often to the point where the tones are lost.

I don't have a DSLR, but if I did (or some other camera with a good sensor and a good macro lens) I'd try your approach. It would be nice indeed to get all the data that the film has to offer !

[Kirk Gittings]

Damn dude I think you've done it!

[Peter De Smidt]

Perhaps the scanners which exaggerate the grain (like the Nikon Coolscan) use a collimated beam of light, like a condensor enlarger. They give an impression of greater sharpness but introduce the Callier Effect: dense areas rendered disproportionately dense, often to the point where the tones are lost.


I think you're right. Grain aliasing might also be involved.


Damn dude I think you've done it!

Thank you, Kirk! Richard Iles, our Arduino programer, and Daniel Moore were at least as responsible for where we are now as I was, and a number of other people here contributed greatly to the endeavor. I still have some refinements to make, but once that's done, I'd be happy to scan a negative for each of you so that you can compare. Daniel will probably get even better results with his D800e.

[SURF]

Nice done, Peter!

It would be interesting to scan resolution target with both scanners. Geometrically the images are good. I think you should never use bicubic sharper cause it is for those who do not want to bother with fine tuning the sharpening. It applies rather big radius.

What is the calculated resolution of the DSLR setup?

Lens R looks better for the first sight.

Still wonder what is the real resolution of the Cezanne scanning different sizes. Never played with it. I will state that the resolution race is finished with DSLR setup when real 6000 dpi will be reached. Of cause it is for 35 mm only. OK, 5000 dpi will be good too. :)

To my eyes the DSLR scan looks better.

[Peter De Smidt]

The Cezanne should be getting at least a true 6000 spi. I've seen that confirmed by a couple of sources. I'd love to scan a good resolution target, but I don't have one, and they're pretty pricey. My conclusion is that the dslr scans are very close in quality to one of the best bw negative scanners. And that's at 1:1. I did scan a 35mm Technical Pan negative yesterday, and the dslr resolved the grain just fine. I had a slight alignment issue with my extension tube,though, and so I'll have to redo it. It's an open question as to whether higher magnification would be useful. It might be that the depth of field gets too small, and I'd have to use focus stacking. That might be ok for 35mm negatives, but I doubt it'd be useful for LF ones.

[Peter De Smidt]

50mm Componon-s Reversed 2x F5.6 4x5 Bergger 200 PMK:

https://dl.dropboxusercontent.com/u/3595413/50mm_Componon_s_2x_f5_6_reversed_upload.tif

[Peter De Smidt]

To give some perspective, here's the complete image:
102639

[Peter York]

Wonderful Peter! Simply wonderful! Your efforts, along with those of Richard Iles and Daniel Moore, have done a great service to the film community.

[jb7]

My conclusion is that the dslr scans are very close in quality to one of the best bw negative scanners. And that's at 1:1.

That's pretty much what I would have hoped for from the start, and my own investigations showed a definite improvement on a V750 scan, so I'm not surprised that more was available.

I hope that people take you up on your offer to make comparison scans, and look forward to the results from that...

They said it couldn't be done, but to be Frank, anything's possible if you put your mind to it... As long as it's an open mind-

[Peter De Smidt]

Thanks, Guys. It's been a lot of fun, mainly because the people involved were great to work with. At least it was fun until I started testing lenses. Yawn!

[SURF]

The Cezanne should be getting at least a true 6000 spi. I've seen that confirmed by a couple of sources. I'd love to scan a good resolution target, but I don't have one, and they're pretty pricey.

Better to say: you don't really want one. :)

OK, let's do some maths. You DSLR scan in my estimate is 4000 dpi. If you have scaned the whole 4x5" on a Cezanne in one take - it will be 2000 dpi. That is because Cezanne CCD is 8000 pixels and if one will scan 4" so there will be 2000 pixels for one inch.

2000 dpi is the level of Epsons and while it is OK for LF why bother with DSLR scan then? It is better to buy an Epson.

Shall we see those 5300 dpi Cezanne scans compared to DSLR scan? 5300 dpi - is the stated max optical resolution of the Cezanne scanner. It happens when you scan rows and every row will have no more than 8000 pixels.

[Darin Boville]

Peter,

I was following these threads a year ago but have missed much since--you've made a lot of progress. Two thoughts/questions:

1) Many users are looking at what you are doing from underneath--that is, they have or are considering buying a V700/750 and are wondering if what you are doing is significantly better, does it justify the cost/hassle. With that in mind would you consider adding a third comparison (the v700/750) to give us a direct sense of how much better this is?

2) Any thoughts on how soon other users might be able to build/buy their own rig? ...and are we allowed to mention kickstarter here? :)

--Darin

[Peter De Smidt]

Better to say: you don't really want one. :)

OK, let's do some maths. You DSLR scan in my estimate is 4000 dpi. If you have scaned the whole 4x5" on a Cezanne in one take - it will be 2000 dpi. That is because Cezanne CCD is 8000 pixels and if one will scan 4" so there will be 2000 pixels for one inch.

2000 dpi is the level of Epsons and while it is OK for LF why bother with DSLR scan then? It is better to buy an Epson.

Shall we see those 5300 dpi Cezanne scans compared to DSLR scan? 5300 dpi - is the stated max optical resolution of the Cezanne scanner. It happens when you scan rows and every row will have no more than 8000 pixels.

I scan 4x5 film with 3 rows with the Cezanne. According the the Seybold Report, in their scanner test they used a resolution target. The Cezanne they tested resolved about 6000 spi, the limit of their target. Tim Parkin thinks it capable of even higher. See: http://www.largeformatphotography.info/forum/showthread.php?106507-Recommend-a-Good-Flatbed-Scanner/page5&highlight=tim+parkin

[Peter De Smidt]

Hi Darin,

I'll try to find someone with a V700/V750 locally, but currently I don't know of anyone.

I love your enthusiasm. People could build their own rigs right now. The control system design and programming is open source and available, complete with Fritzing diagrams and completed Arduino sketches. I'm not sure what we'd offer with a Kickstarter campaign. I don't want to get into the scanner manufacturing business.

[Peter De Smidt]

Btw, here's a partial scan of the 35mm Technical Pan Negative at 3x magnification: https://dl.dropboxusercontent.com/u/3595413/Componon-s%203x%20TP.tif
I used a 50mm Componon-s reversed.

My conclusion, and I'm very happy to hear other opinions, is that it does show a tiny bit more image information than a good 1x scan, but I doubt that it'd be visible in anything but a mammoth print, and sharpening and post processing in general would have to be outstanding in order to preserve it. For me, I doubt that such magnification would be anything more than a waste of time and disk space. We can already see the grain with 35mm TP at 1x, which indicates that we should be fine with any regular film, especially as we move up in format.

[Darin Boville]

People could build their own rigs right now. The control system design and programming is open source and available, complete with Fritzing diagrams and completed Arduino sketches.

I would need a set of step-by-step instruction written at the level of a third grader :) Just kidding--but I did have to google "Fritzing diagram" and "Arduino."

--Darin

[Ken Lee]

Btw, here's a partial scan of the 35mm Technical Pan Negative at 3x magnification: https://dl.dropboxusercontent.com/u/3595413/Componon-s%203x%20TP.tif
I used a 50mm Componon-s reversed.

My conclusion, and I'm very happy to hear other opinions, is that it does show a tiny bit more image information than a good 1x scan, but I doubt that it'd be visible in anything but a mammoth print, and sharpening and post processing in general would have to be outstanding in order to preserve it. For me, I doubt that such magnification would be anything more than a waste of time and disk space. We can already see the grain with 35mm TP at 1x, which indicates that we should be fine with any regular film, especially as we move up in format.

What do you mean by a 1x scan please ?

[Peter De Smidt]

Hi Ken, I'm refering to magnification. A 1x scan is at 1:1 sensor size to subject size, a 2x is 2:1, a 3x is 3:1.....

[Jody_S]

Hi Ken, I'm refering to magnification. A 1x scan is at 1:1 sensor size to subject size, a 2x is 2:1, a 3x is 3:1.....

I had deduced that, and I gathered that your 35mm Tech Pan scan shows that film can definitely out-resolve a state-of-the-art digital camera. But thanks for making it clear. Again, this is very helpful to those of us with crap scanners who are trying to decide which way to go.

[Nathan Potter]

Btw, here's a partial scan of the 35mm Technical Pan Negative at 3x magnification: https://dl.dropboxusercontent.com/u/3595413/Componon-s%203x%20TP.tif
I used a 50mm Componon-s reversed.

My conclusion, and I'm very happy to hear other opinions, is that it does show a tiny bit more image information than a good 1x scan, but I doubt that it'd be visible in anything but a mammoth print, and sharpening and post processing in general would have to be outstanding in order to preserve it. For me, I doubt that such magnification would be anything more than a waste of time and disk space. We can already see the grain with 35mm TP at 1x, which indicates that we should be fine with any regular film, especially as we move up in format.

Peter, I agree with your comment. I don't know the grain size typical of TP so I'm guessing a bit, but I've used it for photographing integrated circuit chips at 1X on 35mm TP and can't really see any grain well imaged on a 20X enlargement. The reason is that the taking lens resolution is the limiting factor. I was using an older 90mm Vivitar series one macro with the 1:1 adapter, an extreme;y high resolving power lens considering its age.

Looking at it another way TP grain size is in the range of 0.5µm to about 5µm (I'm estimating a bit) when developed using a fine grain developer. There are no 1:1 lenses that will resolve those dimensions; microscope objectives would be required. Are you sure you can see clear grain in TP at 1X using your 50 mm Componon S? I guess your TP neg. could be grainier than some I have done in the past. With an NA of around 0.1 the componon is limited in resolving power to the 5 to 10µm range (5µm is 100 lp/mm.) Clearly there would be no advantage in going to 3X for any practical purpose.

Earlier work with my test setup showed really adequate scans at 0.5X with the 90mm Vivitar macro. I would tend scan at 1X for a print up to about 30 X 40 inches and that exceeds what my Epson 750 can do by a very decent margin.

The real caveat in this scanning business is to find a high quality LF film where the image resolution on film is in the sub 20µm range to begin with. I tend to work with a COC of 20µm limit in setting a hyperfocal distance, but it is common to use 50µm or even 100 µm as a limit. The 20µm limit yields 80 µm on a 4X print for 16 X 20 from a 4X5.

BTW some great work there Peter, thanks.

Nate Potter, Austin TX.

[Daniel Moore]

I would need a set of step-by-step instruction written at the level of a third grader Just kidding--but I did have to google "Fritzing diagram" and "Arduino."

--Darin

Those searches are probably still in Peter's and my browser history. : )

I'm considering a dedicated D5200 with it's 3.9µ pixel pitch which would result in a potential 6512 dpi.

Here's a sample of scanned 4x5 Tech Pan, shot with a tripod mounted and cable released Crown Graphic and Xenar 135mm, technidol, captured at 1:1 with the D800E, using a Schneider Apo Macro Digitar 80mm f6.7 (donated to the project by a well regarded and unnamed benefactor : ) with nothing but default sharpening in Capture One, 100% of what that looks like after stitching with Sinc36 algorithm in PTGui (Run on sentence? Nah.):

102781

The offer to scan for comparison stands for me as well, it's critical to the project. Tonality in various media and many other comparisons are still to be explored.

Warning, V700 rant:

I've owned a V700 and was content to toss each and every result of scanning small and medium format as they were crap. That's not to say I didn't loathe the 4x5 scans as well, but for smaller prints, yada yada yada.
Unless you are radical and earn your living by handheld shooting of your own fart bubbles in the bathtub, medium format simply deserves better than V700 resolution, IMVHO. I've revisited some of those earlier V700 scans and am content with the now very good if not excellent results.

Peter and I have taken entirely different approaches mechanically. There are still others waiting to be invented. As Peter's mentioned, the software and electronic aspects are worked out enough to allow for individual implementation, right now.

I also want to keep reiterating the colossal contribution of Richard Iles, without whom.. well, I don't want to think about it. : )

[Peter De Smidt]

Nice work, Daniel!

One of the difficulties we face is that building our scanners requires a bit of equipment and skill. To build mine, for example, a well-aligned table saw and a drill press we're really helpful. If you have those, and know how to use them, then building a similar scanner shouldn't be a problem. If you don't have those tools or skills, then I'm not sure what I can do to help. I could work on building another scanner using parts that don't need those tools, perhaps using aluminum extrusions or threaded steel pipe, but I've already spent a tremendous amount of time and a fair bit of money on this project, and I have other things to do.

Daniel's more elegant design requires a higher level of skill to build.

[Daniel Moore]

Borrowing from an existing support structure may be the fastest way to DSLR scanning. Though not quite yet 'existing', I think the frame from a Rigidbot 3D printer (http://inventapart.com/rigidbot.php) may work as a foundation for example.

From their website:

The RigidBot 3D design is fully expandable in the X, Y and Z directions. Using the injection molded plastic joints, you can adjust the size by swaping out the metal bars with longer or shorter ones (available at any hardware store or on our future online store).

[marfa boomboom tx]

Borrowing from an existing support structure may be the fastest way to DSLR scanning. Though not quite yet 'existing', I think the frame from a Rigidbot 3D printer (http://inventapart.com/rigidbot.php) may work as a foundation for example.

From their website:

The RigidBot 3D design is fully expandable in the X, Y and Z directions. Using the injection molded plastic joints, you can adjust the size by swaping out the metal bars with longer or shorter ones (available at any hardware store or on our future online store).

Agreed. and reiterating my suggestion when this conversation was just a single thread... "rewrap reprap"

http://reprap.org/wiki/Main_Page

original box: darwin
http://reprap.org/wiki/Darwin

alternate: eiffel
http://reprap.org/wiki/Eiffel

the family tree:
http://reprap.org/wiki/RepRap_Family_Tree

[Peter De Smidt]

How many people would build a kit? For instance, suppose there was something along the lines of: http://www.adafruit.com/products/1169, with some additional parts. If there was enough demand, we might be able to talk Adafruit (or similar) into putting together a kit. They already have most of the stuff needed. But my gut feeling is very few people would buy it. It would be too much of a pain for most people to build. The film plane has to be flat and parallel to the image sensor within a few thousandths of an inch. I have a precision square, straightedge, dial indicator, and a laser alignment device, and even with these tools, getting everything just right takes a lot of effort. If someone has those types of things, and is a machinist or wood worker, then those are the type of people who like to come up with stuff on their own. For those people, the biggest hurdle would likely be the control system, which Daniel and Richard have worked out and made public.

Perhaps a Kickstarter campaign would make sense. The goal would be for us to develop a kit from easily sourced materials, such as aluminum extrusions.... If we set a low goal, one that's just enough for us to buy all of the materials needed to figure this out, and we don't get funded, well, that would be our answer. If we do get funded, that might show Adafruit (or others) that there's a big enough demand.

[marfa boomboom tx]

How many people would build a kit? .... If we do get funded, that might show Adafruit (or others) that there's a big enough demand.

How many? slim.
anecdote: last summer ('12) I held a maker camp, along the lines of a build camp from the homebrew days of computing. It was held in Presidio county (out of the way ranch land), so that may explain the low numbers. High interest. A few dozen signups. At the end of the build, we had 4 units built.
This past summer, they built 3D printers and automatic watering systems. Those sessions were full.

[suggestion: kit for some of the mechanical; made, including connectors for any motion drive and control sections (sparkfun).. the alignment stage is the problem in mechanical]

[suggestion2: get one of the "photo workshop" centers to check interest for a summer workshop]

[Struan Gray]

How about a kit based upon a popular enlarger. The enlarger light source provides illumination, and the existing structure (and adjustments) provide the parallelism. Best of all, there is an existing film holder apparatus, with lots of options for glass/glassless etc according to taste.

In an ideal world, the scanning frame could be mounted in two positions. Either on the enlarger head (perhaps with the head inverted and a camera mounted above it) so that the negative is moved through a spot of light created by the head. Or on the baseboard, and the camera is moved about photographing the projected image from a regular enlarging lens.

The first allows tighter control of the optics, but at a cost of complexity in the negative stage. It allows you to adjust focus on the fly if so needed. The second relies upon a good enlarging lens, and adequate alignment to maintain focus, but is a much simpler thing to build if you already have a working, well-aligned enlarger.

Anyway, congrats on the progress and success so far. Impressive stuff.

[Peter De Smidt]

I've acquired an Edmund's USAF Hi_Res resolution target.

Here is the result with a D600, Rodagon D ARD1 at 1:1, f4, No Sharpening:

https://www.dropbox.com/s/1pqnh221b7abz1u/Rodagon_D_ARD1_F4_D600_USAF_HIRES.tif

I"m not experienced reading these charts. What do you think?

This page should help: http://www.aig-imaging.com/mm5/PDF/USAF%201951%20Test%20Target%20T-20_v1-04.pdf

My tentative analysis is that my current system gives 3650 spi horizontally and 2900 spi vertically. I expect that these numbers are limited by the camera sensor. (Rik Littlefield on the Photomacrography forum agrees.) I'll be very interested to see what Daniel gets with his D800E.

For comparison, my Cezanne achieves 5800 spi.

[Nathan Potter]

These seem to be nice results Peter. It's hard to draw accurate conclusions from your image of the target here due to the web quality and the inability to load a webarchive image into PS to be analysed for contrast.

I think I'm seeing a decent linewidth pair though on the target at the 5/6 site and possibly around a 50% contrast. I've looked at a lot of these but can only estimate from your image. 5/6 is 8.77 µm lines so about 57 lp/mm and would be really decent for 1:1 reproduction. Given your D600 is 6016 X 4016 pixels (5.98 X 6.32 µm pitch) you may not quite be limited by the sensor.

In any case a resolution of 57 lp/mm at 50% contrast would be nearly double what my Epson V750 would provide (30 lp/mm at 50% contrast), see plot below. This would confirm somewhat analytically what I've been thinking should be possible using a DSLR scanning technique when carried out with precision. 57 lp/mm would equate to about 3000 spi at 50% contrast.

http://farm7.staticflickr.com/6171/6154033421_f5c72e4169_z.jpg (http://www.flickr.com/photos/argiolus/6154033421/)
EPSONcont-web-1 (http://www.flickr.com/photos/argiolus/6154033421/) by hypolimnas (http://www.flickr.com/people/argiolus/), on Flickr

Nate Potter, Austin TX., Holliston MA.

[Ken Lee]

Nathan -

Thanks for posting your most excellent graph of resolution versus contrast with the Epson V700 scanner today - which possibly explains why different testers report different "results" with that scanner.

Whenever I've scanned film (not targets) with my Epson I've never seen much improvement over 1600, but scan at 2400 anyhow just to be on the safe side. This matches your results at 50% contrast, a number to which you seem to return as a standard of some kind.

Does 50% contrast correspond to a practical limit of vision or some other rule of thumb ?

[Peter De Smidt]

Nate,

Thank you for your opinion and the graph. I agree (and a couple of other people have in correspondence) that 3000 spi is a good estimate of what my system is providing. It'll be interesting to see what can be achieved with other sensors/lenses.

I'm curious about the dropbox file. Downloading it doesn't give you the original file? Is there a better way of sharing tifs?

[Peter De Smidt]

With a reversed 50mm Componon-s at 2x magnification and f5.6 on lens, I get at least 5792 spi. (Group 6 Element 6). In one direction it's at least 7298 spi. (Group 7 Element 2). That should be sufficient for most purposes :)

[Nathan Potter]

Ken, a 50% contrast value is purely arbitrary and simply a convenience for comparison. One can clearly extract a workable image from even lower contrast values but the tonal details become increasingly lost; or perhaps compressed is a better description.

Contrast can be pulled greatly when scanning film due to the degree of light source collimation and only secondarily due to the contrast inherent in the macro lens used. I think this is just a matter of personnel taste coupled with the objectives of the artist.

It is interesting to think about the three basic scanning systems and their optical paths. The flatbed, Epson V750 for example, uses a diffuse light source and a strip of discrete sensors to record an image. The DSLR can use either a diffuse or collimated light source with a fixed sensor array which yields a system a bit more akin to a photomultiplier drum scanner, like a Tango. The drum scanner uses a three beam point RGB focused light source in the micron range which yields in effect, the ultimate in collimation and high contrast and potentially preserves the most accurate tonal detail from a high quality film.

Peter, Your dropbox file yields a webarchive file which cannot be opened in Photoshop. I'm not the sharpest light on computer stuff but I would guess that one could send a tiff file through dropbox and that could be opened in PS.

I think you are about right with a 2X magnification yielding about double the spi from a 1:1 image. All other variables being about equal, I'd expect 3000 spi to jump to around 6000 spi. However a 2X scan will have a vanishingly small DOFocus and would require a really large number of stitches. 7000+ spi is certainly well beyond what I would ever need from any of my chromes and B&W films.

Nate Potter, Austin TX. Holliston MA.

[Daniel Moore]

If you do not use -right click and choose 'save target as'- Peter's link should open up a dropbox page with a download button which worked for me.

[Peter De Smidt]

Nate, I agree that 7000 spi is overkill for any but the sharpest small format film. A true 3000 spi should be enough for the great majority of LF negatives. My issue is that my other scanner does better...and so why use the dslr scanner? I'm going to check out what a few of my lenses do at 1.5x.
.
.
.
1.5x with the Rodagon D gave at least 3649spi. (Group 6, Element 2).

[Nathan Potter]

If you do not use -right click and choose 'save target as'- Peter's link should open up a dropbox page with a download button which worked for me.

Quite right Daniel - it worked for me also. Thanks

Nate Potter, Austin TX., Holliston MA.

[Drew Bedo]

How Long does it take to scan a 4x5 . . .an 8x10?

[Peter De Smidt]

Drew, I'll time a scan sometime soon.

Here's a d600 shot of a Stouffer step wedge. I've listed the densities of the steps, along with the green sRGB channel and LAB L values when the file is taken into Photoshop.

[Daniel Moore]

Here's a res chart shot with the D800E using a Schneider 80mm f/6.7 Apo Macro. Magnification is 1:1 give or take a few pixels.

103614

It looks to resolve the same whether it's a .tif, .jpg or .gif.

[Peter De Smidt]

It looks like Group 6 Element 2 is resolved. That's 3650 spi.

[Daniel Moore]

Drew, using a 3 second exposure delay I captured 25 frames of a 4x5 scan in three minutes 8 seconds. I'm moving the stages at very nearly 30mm per sec.

[Daniel Moore]

Here's a resolution test using a D7100 and Magnagon 75mm f/5.6:

112275

Looks to me like group 6 element 5 is resolved putting it at 5160 spi. Yikes!

[Peter De Smidt]

An outstanding result. Thank you for doing the test!

[rbultman]

I'm currently in the middle of designing a scanner using the Actobotics (http://www.servocity.com/html/actobotics.html) system. Its rather like a precision Erector set. The design will use 4 of their slider kits (http://www.servocity.com/html/channel_slider_kit_a__637108_.html#.UzQJOl5RE5M), two for each side of each stage, with stiffening elements and joining plates where necessary. I plan on using a stepper motor from Spark Fun to drive each stage. I plan to use a shaft to connect the two sides of a stage together so that they can be driven together. I contacted them a short time ago regarding stepper motor mounts and they said that the mounts were being finalized and would be available soon. This should make is possible to mount a NEMA face motor to the slider kit. There would be one motor per stage.

Since the rails are on the sides, this makes it possible to make a platform with an aperture through with light would shine for the "exposure". The light source could then be very small, only slightly larger than the capture area. The camera and light source would be stationary. A very high quality light source could be used. This should result in the most even illumination across all sections of the negative. This also removes the weight of the light table from the stages so the stages would not care much additional weight. This idea might require negative carriers with glass in order to ensure that the negatives remain flat. If flatness or other technical problems make the aperture concept unworkable, the "moving light table" concept can always be used in its place.

A slider kit is $120, so the cost of the scanner will be more than $500, excluding the camera, lens, etc. I think it should be less than a new flat bed scanner, which is why I started looking at this concept, as well as resulting in higher quality scans than a flat bed scanner could produce, I assume. I have a Sketchup model put together of the concept and will put that on my Dropbox if people are interested. I appreciate comments and feedback regarding the suitability of this approach. I have yet to consider how the camera would be mounted but assume a gantry could be constructed to hold it.

Regards,
Rob

[rbultman]

Slight correction to my other post: I meant two slider kits per stage, one per side of each stage.

[Daniel Moore]

I'd love to see the sketchup file, sounds fascinating. I'm now working on a circuit board for Scanduino which should be adaptable to many systems such as this.

[Peter De Smidt]

Sounds like a good system. My only reservation concerns the " The light source could then be very small, only slightly larger than the capture area." In my experience that will lead to significant light falloff at the edges of the sampled area.

[rbultman]

Yes, you are right. I guess the light source needs to be sufficiently large to avoid light falloff. I was thinking of an array of white LED's behind frosted glass acting as a diffuser. I suppose if the stage was solid (no aperture) and of frosted/milk glass, the negative would have a better chance of being flat and the light could still be stationary.

Here is a link (https://dl.dropboxusercontent.com/u/8213345/scanner1.skp)to the sketchup model.

[Daniel Moore]

Could you save that again for a previous version, I can't open it as it's newer than my Sketchup version, 2013.
As for the fixed lamp approach, if you can get a consistent exposure and 1 or 0 points out of 255 variation across the capture zone you might get away with it. The acid test would be to create a scan with plenty of sky, stitch the images and reduce the result image size to 1" in height and look for a grid pattern artifact. That's pushing the system very hard but will reveal a weakness in the illumination method very well.

Film flatness will make or break the scan, DOF (at 1:1 at least) is that shallow. I've gotten great resolution from a Rodenstock Magnagon f/5.6 but the DOF is ~.006", less than the film's thickness requiring the emulsion be face down and runout of the travel over the film area held to very nearly nothing.

[Peter De Smidt]

Are you planing on scanning black and white negatives, color negatives or slides? If it's the latter two, you might consider an rgb led setup.

[rbultman]

I am planning on using a 100mm macro lens (Canon EF 100mm f/2.8 Macro USM Lens) with my Canon Rebel for starters. The specs say the lens is capable of 1:1 but I'm not sure if that is with an APS-C sensor or full frame. I'm hoping that the lens will result in minimum light falloff given that the lens is meant for a full frame sensor but will be used with an APS-C sensor. Given the testing of the lens at Photozone, it looks like f5.6 or f8 will be the best place to start for testing.

Now I'm worried about DOF if you say that it is 0.006 at 1:1. I suppose that the film and sensor need to be absolutely parallel over the capture area of the negative. How difficult was that set up? It would seem that everything is critical in alignment here and even variations in thickness of the stage material would impact this. Lots of variables to control. What gotcha's did you guys encounter?

How critical is wet scanning to this? What fluid did Dan use in his video? Does Peter also do wet scanning? It didn't look like it from his YouTube video, but that was some time ago and maybe he has moved on to that. I have not done wet scanning.

I plan on most scanning B&W negatives but may scan color negs as well. Why would the RGB light source aid in scanning color material? I'll use this almost exclusively for large format film as I use a Nikon LS-8000 for 120. If the Nikon breaks, I might use the DSLR scanner for 120 also.

[Daniel Moore]

Some general notes, the DOF I stated was an estimate based on the specific lens and it's fixed f5/6 aperture I'm using. At f/8 and 100mm you will have a lot more in relative terms. It's such a nebulous subject it's hard to say exactly, too many factors to consider. I leave it to the experts to clarify further. One of the best ways to see what's happening I've found is to pick up a set of .002" plastic shims and cut and stagger them. Then carefully photograph the edge of the stack and see for yourself what sort of acceptable range you're getting.

I can't speak to other systems but in my it wasn't as hard as I anticipated since I made as perfect a lens mounting ring as I know how on high end machinery, the rest was maintaining extremely consistent stage thicknesses. Finally, allowing for shimming the glass carrier for final tuning with a dial indicator.

In my opinion, wet mounting is important for best results. Film curl can easily exceed DOF. I considered it a black art before I started doing it. As you saw in the video, it's very straight forward. I did get some bubbles in my first attempts but gauging the speed of laying down the film and mylar and how much fluid to use is a simple matter with a little practice and I only get bubbles on trouble film like that with bends due to sprocket hole bending or clip dents on 4x5 sheets, which I cut off now. I'm using Lumina from ScanScience. Also, let's not forget that wet mounting results in higher contrast and less spotting, scratch removal so it's worth the extra time IMO.

The RGB LED's we've been using allow to adjust the light color to establish a more neutral capture for each color film type, leaving only to invert and optimize the color channels to arrive at exceptionally good color starting points. Not doing so meant a more work trying to get to a good starting point. I'm going to be testing another LED source, noted for it's high CRI at some point.

Also, I'd worry about vignetting last as it's easily corrected in Raw processing.

[Daniel Moore]

I'd be remiss if I didn't mention one big 'gotcha'. Peter discovered there was an unfavorable interaction between the sensor and the LED PWM pulse frequency which resulted in a rolling banding of uneven color on the result image. This artifact is visible at faster shutter speeds and not at slower ones. Regardless, it's undesirable and the solution is to use a technique proposed by Ludvig, turn on each color for a specific duration. This allows to eliminate the PWM frequency of the color controller entirely. Duration of on time for each color determines the color balance relative to the exposure itself. That new circuit is ready to go and the code ironed out, it just needs squirting into the Scanduino code base which I don't expect to be much trouble.

[rbultman]

I'm having a hard time visualizing how turning on each color for a specific duration works relative to PWM? Did you try different PWM frequencies? What about using a low-pass filter after the PWM output to reduce ripple or using a DAC output? How many effective bits do you need to acheive the color range? Are you using a custom LED setup of your own design or a commercial one? Are you just looking for a change in color temperture of the light source?

Sorry about the rapid-fire questions. I design and program embedded systems for a living for consumer products so I'm hoping to be able to lend a hand here. Although, given that you are designing your own circuits, I'm guessing you don't need any help.

Regards,
Rob

[rbultman]

Could you save that again for a previous version, I can't open it as it's newer than my Sketchup version, 2013.

I saved this (https://dl.dropboxusercontent.com/u/8213345/scanner1_sketup_8.skp)to Sketchup Version 8. Hopefully you will be able to open it.

[Peter De Smidt]

Regarding wet-mounting, I agree with Daniel that it is needed for the best results, but the results without it can be very good. Wet-mounting eliminates Newton's rings, helps with film flatness, increases contrast, minimizes some physical problems on the negative. Note, though, that if you film is really curly, wet-mounting here might not solve the flatness issue. In a drum scan, the negative is pulled down, since as the cover sheet is pulled, it exerts downward pressure along the curved drum. In out situation the mounting platform is very flat. The cover sheet can be pulled tight, but it's like a drum skin, with a minimal amount of downward pull. In addition, the extra surfaces of the mylar sheet, which is a dust magnet, can add to the dust problems. I've used Lumina, Kami, and Prazio Oil. Lumina is my favorite.

112931


There's much more info about our light sources thread: http://www.largeformatphotography.info/forum/showthread.php?87536-DSLR-Scanner-Light-Sources&highlight=dslr+scanner+lightsource

[Daniel Moore]

Rob, let me quote your questions and answer them in the light sources thread Peter linked to.

[Daniel Moore]

I had a look at your 3D file and it looks very interesting. The butt joints on the upper carriage rails leaves me wondering how well machined those ends will be but you won't know until you try it, as I did with my approach. I'm not clear on whether the belt is direct drive or motors are mounted at 90 degrees (gear driven). The pulleys look like they are not GT2 type. The precision of GT2 is much finer than other types which I attribute to obtaining an insane level of repeatability with my system. This can greatly impact the ability to use templates to stitch featureless areas. The fewer the teeth the higher the torque I've read. I'm using 20 tpi pulleys with Nema 17 motors rated at 60 oz/in.

If you haven't already, here's a great quick read on the subject of belts/pulleys pertinent to our application:
http://reprap.org/wiki/Choosing_Belts_and_Pulleys

[rbultman]

I'm moving the conversation to the negative stage thread.

[colorfulbliss]

Hi all! I have spent the last month and a half experimenting with DSLR scanning and have some samples to show. I currently own a PrimeFilm 7200 for 35mm, and an HP G4050 for 6x7 and 4x5. I am fairly satisfied with the results from the PF7200, but the G4050 is pretty disappointing from a sharpness standpoint (not to mention the frequent battles with streaks). So the goal was to get a scanner that could match the output of the PF7200 with the larger formats. They aren't many choices at a price level that a (non-wealthy) hobbyist photographer can afford. I considered a dedicated medium format scanner, but they are fairly expensive and would orphan the 4x5. I tossed around the idea of getting a v700, but I've read so many mixed opinions on its performance that I hesitated to shell out the cash for it. Then I stumbled upon the DSLR scanning threads here on LFF and thought I might give it a whirl. So far I only have a cobbled together proof-of-concept rig, but after some agonizing hours playing with it I do believe it is a worthwhile project.

Here's what I'm working with so far: Samsung NX100 (14.6MP APS-C), 55/3.5 Micro-Nikkor-P, Porta-Trace box with original lamps removed and replaced with RGB LED ribbon light strips, camera and light source stationary, hand panning the film holders out of my two current scanners (not very accurately, I might add).

Workflow: Raw > MakeTiff > Microsoft ICE > ColorPerfect > Lightroom. sRGB colorspace maintained throughout processing.

This sample is Ektar 35mm shot with a Maxxum 35-105 3.5-4.5 and self-developed. Scan is at 1:1, 3-shot stitch

https://farm3.staticflickr.com/2938/14410753804_33828c3f35_h_d.jpg

Comparison. DSLR on the left PF7200 on the right

https://farm4.staticflickr.com/3869/14410753714_41f5d9f16a_o_d.jpg

I have a sample from the G4050 if anyone really wants to see it. It's a blurry mess as you might imagine.

I haven't yet sorted out how I am going to approach building a negative stage, but I probably will stick with something manual as I don't scan a large volume of films. That could all change though!

Thanks for looking!

[Peter De Smidt]

Nice job! Thank you for posting.