Page 3 of 3 FirstFirst 123
Results 21 to 29 of 29

Thread: DSLR SCANNER No.7

  1. #21

    Join Date
    Jul 2007
    Location
    Austin TX
    Posts
    2,049

    Re: DSLR SCANNER No.7

    Joseph, your last comment is what interests me. DSLR imaging at 0.5X allows a four fold increase in the field of view. A 24 X 36 mm field at 1:1 becomes a 48 X 72 mm. field at 0.5X. The detail referred to the sensor of course is reduced by 0.5X but a higher full frame pixel count can offset this - although a factor of 2 is a lot to compensate for. But full frame at 0.5X will nearly cover a 4X5 with four frames. The catch is probably in the quality of the lens. To make use of the intrinsic resolution of a 7 um pixel pitch sensor at 0.5X one needs a lens capability of close to 100 lp/mm at some reasonable contrast.

    The only item I have to get started with is a 90mm, f/2.8, Vivitar Series 1 macro with 1:1 adaptor; although still highly regarded for its unusual resolution capability approaching the 100 lp/mm level. It is of notably low contrast however. I should try to find some resolution data on it. These are still available in the $400 to $500 range I believe.

    On the other hand 36 MP X 4 will yield a 140 MP count on a 4X5, and GB+ files. I think most 4X5 images don't contain anywhere near that kind of detail. BTW that would be about 4 times the resolution I would normally get from my Epson 750 where I see a spot size equivalent of between 20 and 25 um. at reasonably high contrast. The full frame 20+ MP Canons seem ideal for this application as well as a new Nikon D800 - although very pricey - for me anyway.

    But this would be my direction if financially feasible.

    Nate Potter, Austin TX.

  2. #22
    Peter De Smidt's Avatar
    Join Date
    Jan 2001
    Location
    Fond du Lac, WI, USA
    Posts
    8,954

    Re: DSLR SCANNER No.7

    There's some info regarding all of this at: http://www.photomacrography.net/foru...=101164#101164
    “You often feel tired, not because you've done too much, but because you've done too little of what sparks a light in you.”
    ― Alexander Den Heijer, Nothing You Don't Already Know

  3. #23
    joseph
    Join Date
    Jan 2007
    Location
    Chapel Hill NC
    Posts
    1,401

    Re: DSLR SCANNER No.7

    Interesting testing over there Peter, looks like there's a lot more resolution in the aerial image-
    but there are perhaps a few flaws in his reasoning, and methodology-
    Finer sensor pixel pitches will still be limited by diffraction themselves, and a more accurate test might have seen him use sine test patterns, rather than the slightly slanted linear ones.

    Also, all his testing seems to be done pretty much on-axis, and there's no corresponding off-axis results- and that's an area we're going to need to make use of. The smallest amount of field curvature will eat all the resolution away from the centre, in our case, making the centre correspondingly useless, on its own.

    Nathan, you're correct- a high resolution lens, used at an aperture that isn't going to limit it, coupled with a high resolution sensor, would be a good start. Assuming of course, that lens also has a flat field, and that the film is flat. I measured a piece of transparency film in a holder last night, and the deviation from edge to centre was 1mm (actually 0.99mm, but nobody is going to believe that) A piece of b+w negative, drum processed without clips, was much flatter, but still had some curl.

    You're correct also, in that it's main use might be in being able to make less frames to perform a complete scan.

    As I mentioned earlier in the thread, because these scans are made up of any number of frames you choose, using any magnification ratio, they are potentially limitless- you don't need the best lens, or the the highest resolution sensor- you just need to put in the effort with the exposures and stitching. True, the pixels are not going to be very efficient, but downsampling will help that- perhaps even before you get to the stitching stage, so using the D800 will definitely be more economical in terms of time spent. Though I think, if you're going to be using a program like PTGui to stitch, the magnification have to be more like 0.4x, to allow for overlap.

    Regarding the sensor, I think a long range of sensitivity is going to be important- the D700 did get into the shadows on this rather dense transparency, but bringing detail out of them might need local attention- just as bringing detail out of the low range on a drum scan might need local attention- all the processing on the scan in this test involved global adjustments.

    And regarding the lens- if you're going to scan a 4x5 in 4 shots, then flare might become more of an issue- depending on how resistant to flare your lens is-
    photographing small parts of the transparency means that flare is controlled locally, in 56 locations, in this case- It might become an issue for some scans of larger areas, and some lenses...

    I'd be in complete agreement regarding the level of detail contained in a 4x5- which, of course, is going to be limited by the taking lens and the film. It's difficult to see how it would be necessary to present more than 140 million pixels (actually, more like 100-110 million, taking account of overlap)- unless you wanted to scan at a microscopic level, where the image is made up of the grain structure of the film, which is a kind of a quantum leap beyond the resolution of the taking lens...

  4. #24

    Join Date
    Jan 2010
    Location
    Barcelona/Spain
    Posts
    1,403

    Re: DSLR SCANNER No.7

    Damn... I didn't realize there was so much porn in here. Amazing work as always Joseph. Great design. I shot some negatives against the window before I got my V700 using a D300 with both a 50mm 1.4 and a 105mmVR 2.5. Very badly arranged but image quality was better than expected.

    What are your thoughts on the compared images?

  5. #25
    joseph
    Join Date
    Jan 2007
    Location
    Chapel Hill NC
    Posts
    1,401

    Re: DSLR SCANNER No.7

    Thank you Ramiro-

    Yes, I copied some film with a point and shoot some years ago, and made a nice Blurb Album for the family for Christmas- so I was well primed for this exercise-

    The compared images-

    Well, they're much too large, but that stems from the DSLR Capture, and being able to compare at the native resolution.
    The magnification of the taking lens was fixed, being a reversed prime- I could have gone for more magnification, but not less, at the time.

    The V750 doesn't resolve at 4000dpi, and the reversed lens doesn't resolve at the the pixel pitch of the camera sensor, and the taking lens (72XL with centre filter) doesn't resolve to that level of magnification on film either. ( I think...)

    But no matter, it might be instructive to see the captured information, knowing that improvements can be made-

    I think it's quite clearly shown that the DSLR scan and stitch out-resolves the V750, but not by a whole lot.
    It shows that the V750 employs a lot of smoothing, even when Digital ICE is turned completely off. I suppose this could also be a function of its uprezzing algorithms-

    The DSLR stitch shows better native colour, I think- because I was able to edit individual frames in Aperture, then stamp the remaining frames in the set, allowing far greater control, at a greater level of detail. The Silverfast Software, (or the Epson) by comparison, is a very broad brush, and adjustments made post scanning are a lot more processor intensive. Scanning at a ridiculous 4000 dpi took around 35 minutes too, much longer than making the DSLR images, though using an Epson at 4000 dpi is not something you'd do often. Ever.

    The DSLR Scan shows all the dust, the Epson removes most of it- though I'm going to have to try a more diffuse light source on the next run of tests.

    Next time, I plan on making a dedicated test negative, with sine charts at centre and edge, using a valuable piece of T55, which is apparently capable of 180 lp/mm, according to some sources- I've got a new scanner lens to test, which looks promising, though I can't mount it properly yet, without extra components. It's designed for a much smaller sensor, but at the increased magnification I'm using it at, it should be fine. It's specified to have 100 lp/mm in the centre, 60 lp/mm at the edge, on an image circle of 22mm at 1:1, so I think that will work out just fine using the usual stitch overlap. I might have to go just beyond 1:1, or use a crop sensor camera, but initial tests, just holding the lens up to the camera, show a very highly resolved image, and signs of a very flat field.

    I plan on making the test image using a Nikkor-M 300mm- it's probably the sharpest lens I have, though other options are 150 Symmar-S, 450 Nikkor-M, or some Super Angulon XL's - thoughts, anyone?

    I have no doubt that a DSLR scanner is a viable option, and would hope that the next run of tests gets it into Drum Scanner territory-
    based on this prototype, it might not take long to put the 4x5 version together, which will be whole a lot smaller-

    I think the stitching doesn't have to be very onerous, and a slightly more generous overlap would allow the software to make a comprehensive assignment of control points, leading to a far less labour intensive output.

    All in all, not a bad outcome.

    If my V750 failed tomorrow, (which it won't) I don't think I'd replace it, but for most normal scans, including all proofing, it's just too convenient. If I had a need for higher res scans, I think the next generation of this thing will be able to provide them-

  6. #26

    Join Date
    Mar 2006
    Location
    Pittsfield, MA
    Posts
    784

    Re: DSLR SCANNER No.7

    Quote Originally Posted by jb7 View Post
    Thank you, two is good, three including Peter...
    Four, but until the semester is over I can only passively contribute. I have done similar scans before, and using a Beseler Dual Mode Duplicator to scan slides for a local studio. What I envision for my scanner is an x-y table using linear guides and 17 frame steppers and ballscrews or leadscrews, an 8x10 glass holder, light pipe or light box type light source with strobe or MH for a light source and a smallish PIC controller to sequence the steppers and fire the release.

    Eminently "doable" and I probably have most of the stuff sitting in the garage, sans the PIC stuff, but for the time being I can use a PC with a paralell port and Mach 3 to drive the table and release the shutter as I already have a machine set up with it as a controller.

  7. #27
    Moderator
    Join Date
    Apr 2009
    Location
    Northern Virginia
    Posts
    5,614

    Re: DSLR SCANNER No.7

    Quote Originally Posted by Nathan Potter View Post
    On the other hand 36 MP X 4 will yield a 140 MP count on a 4X5, and GB+ files. I think most 4X5 images don't contain anywhere near that kind of detail. BTW that would be about 4 times the resolution I would normally get from my Epson 750 where I see a spot size equivalent of between 20 and 25 um. at reasonably high contrast. The full frame 20+ MP Canons seem ideal for this application as well as a new Nikon D800 - although very pricey - for me anyway.

    But this would be my direction if financially feasible.
    4x5 negatives contain one level of detail that has proven itself to me as worth pursuing--grain. If we can resolve grain, then we can restore a high level of MTF in the DSLR through what would normally be excessive sharpening. That does not create artifacts, because at that scale, it's all noise on the negative being scanned. But it means we get full use of the sensor in the DSLR. It also means tonal separations at larger scales will be based on grain dithering, rather than on pixel color depth. This should greatly improve dynamic range, though I have not yet really proved that to myself. This seems to me a very useful outcome.

    For my 12.5MP Canon 5D and the lenses I've tested so far, I needed greater than 1x magnification--more like 1.8x--to be able to resolve grain well enough to test the theory. And that was marginal, but it did work. Peter was able to resolve grain (I think) using his 55mm Micro-Nikkor at 1:1, so I've picked up one of those in hopes I can improve on what I have and not have to use the higher magnification. I think the sensel size should be small enough in the Canon to render grain at 1:1 if the lens can, though the AA filter is a wild card in that assumption. The most convenient answer seems to me the lowest magnification that allows us to resolve grain well enough to oversharpen at the grain level enough to sharply resolve grain. If the sensel density and lens quality is good enough to do that at 0.5X, then that will be easier to stitch than doing it at 1x or 1.8x. If we ever get to the point of being able to do the whole 4x5 film with a single frame in the DSLR, 4x5 film will probably vanish from this earth soon after. But it's likely I won't be able to afford it--either the camera or the glass.

    (I do not know how this thinking will translate to color film, though I think it will do okay with negative materials and be more challenging with transparencies. A system that is good enough to resolve grain in black and white film, though, will surely do well enough with color film, though perhaps without the benefit of using grain dithering to help with the dynamic range issue.)

    Rick "whose 5D came as a result of a bonus, and who no longer gets bonuses" Denney

  8. #28
    Peter De Smidt's Avatar
    Join Date
    Jan 2001
    Location
    Fond du Lac, WI, USA
    Posts
    8,954

    Re: DSLR SCANNER No.7

    One thing to note, my example was with Fuji Acros developed in Pyrocat MC to a slightly lower than normal density, the equivalent of n-1. Most other film will be grainer. And if you use Bergger 200, for example, grain shouldn't be too hard to find.

    Rick, our Linos lenses are probably optimized for 1:2 normally positioned and 2:1 reversed. They should be pretty good.
    “You often feel tired, not because you've done too much, but because you've done too little of what sparks a light in you.”
    ― Alexander Den Heijer, Nothing You Don't Already Know

  9. #29
    joseph
    Join Date
    Jan 2007
    Location
    Chapel Hill NC
    Posts
    1,401

    Re: DSLR SCANNER No.7

    Yes Peter, perhaps choosing your grain size will be an important factor? I have some Bergger 200 stored, haven't used any yet, but had always expected that it would be for contact printing. Polaroid 55, which I mentioned earlier, that might turn out to be a no grainer...

    Anyway, not particularly relevant, I'm sure the right size grain has been imaged before with a drum scanner. Perhaps it deserves another one of your scanning thread sub-sections? I see one potential issue, the stitching of such images, so it deserves to be studied in greater depth-

    Erie, it sounds like you know what you're doing- so good luck with that-
    As I might have mentioned earlier, an xy stage would be desirable, but not essential- however, that's the position of someone for whom the electronics and the software control would be quite alien; I'm sure if I had access to a stage, I would use it.

    Quite early in the main thread I suggested using the motion stage from the RepRap 3d printer system. It uses threaded rod, nuts, standard bearings, and 3d printed parts for components, and should be the least expensive way to achieve a moving bed. Positional accuracy and consistency is not particularly important if you're using a stitching application which involves the assignment and comparison of control points for individual images, but might be overwhelmingly important if you choose to combine tiles using another method, and I'll be using a stitching application- it's actually impossible to manually position images in PTGui without having control points assigned, although you can replace the images in an existing project, which might work.

    The RepRap bed uses open source design and software, and there's a lot of information available, so if I chose to automate it, I think that's the route I'd take. It's accurate enough to provide the 3d modelers with resolution down to 0.1mm, which is 30 times more accurate than I managed to achieve with some of my positioning in the test scan in this thread. I see the benefit of such a system, I just don't see it as an essential part of this particular project, which has more fundamental issues to deal with- such as optical performance, and lighting.

    Good luck with your stage, look forward to seeing it- Maybe I'll change my mind when you come up with a foolproof and inexpensive solution, and I'll wonder why I was so ambivalent to it in the first place- I can change my mind easily, if I'm presented with something shinier...

Similar Threads

  1. DSLR Scanner: Lenses
    By Peter De Smidt in forum LF DIY (Do It Yourself)
    Replies: 120
    Last Post: 12-Nov-2018, 13:19
  2. Making a scanner with a DSLR
    By Frank Petronio in forum LF DIY (Do It Yourself)
    Replies: 616
    Last Post: 9-Jan-2018, 03:06
  3. DSLR Scanner: Lenses
    By Peter De Smidt in forum Digital Hardware
    Replies: 15
    Last Post: 6-Mar-2012, 20:43
  4. DSLR Scanner: Workflow
    By Peter De Smidt in forum Digital Hardware
    Replies: 0
    Last Post: 24-Feb-2012, 13:26

Tags for this Thread

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •