Making a scanner with a DSLR

[Struan Gray]

Lenses can be telecentric on the object side, which gives you the constant magnification and eliminates parallax. Or they can be telecentric on the image side, which for regular photography is usually presented as a way to get the light square on to the sensor. Or they can be telecentric on both sides, like the Xenoplan.

One advantage of a bilateral telecentric for our application is that the images they produce are less sensitive to tilts or misplacements of the object, lens or sensor. This technical brief from Schneider has more details than most will want to know:

http://www.schneiderkreuznach.com/pd...ric_lenses.pdf

The ones that are bilateral are usually specced for a fixed repro ratio and working distance (distance from front lens to object). The good ones *only* work well under these conditions and perform poorly or not at all at other magnifications and distances. For example, a pure bilateral telecentric optic is actually afocal for infinity light - it does not focus it at all, but simply expands or contracts the parallel beam. It's more like a rifle sight than a photographic lens.

So you have to be careful with eBay oddballs (even if they do look supremely cool :-). This Xenoplan lens will surely perform very well on the intended format, but that's quite a bit smaller than FF 35 mm and there is no guaranteed way to make it cover a larger sensor except for turning it round and taking pictures at 3:1 instead of 1:3. There is probably some wiggle room, but you would need to have the actual lens for testing, or a conversation with the Schneider application engineers, to know for sure.

[rdenney]

So, what about long, fast medium-format lenses? I have a Pentax 200/4 that covers 6x7 and a CZJ Sonnar 180/2.8 that covers at least that. Is the idea that these could be arranged nose to nose with a stop in between them? That's sort of an inside-out Super Angulon, which is really a pair of reversed telephotos arranged tail to tail. Something about the concept is not clicking in my brain.

Rick "noting the large exit pupil of these lenses" Denney

[peter ramm]

Straun, the Harvard system is a rara avis. Good for them. I've seen so many one-offs sitting in back rooms and ultimately cannibalized for parts.

Given the level of expertise that participants here are showing, why not think outside the camera box a bit? Sure, working up a flying spot or rail-driven PMT would be a project but linear arrays are simple and reasonably priced. Is there anyone here with the skill to replace the crappy array/lens element in an existing Epson with a better one? You'd wind up with something like one of the higher performance flat beds. This is very much a question, not a suggestion. I do not have any expertise in this type of integration. Everything we did was with cameras - which is why I distrust them.

[Peter J. De Smidt]

And you can still buy them new. See: http://www.edmundoptics.com/products...productid=2904
They do have pretty impressive effective apertures. I have a Mitutoyo 2x infinity objective, but it's effective aperture at 2x is 18, for instance, and a Rodagon D would have an effective aperture of 11 at 1:1.
But the Schneiders are only meant to cover a 1/2" sensor, which isn't ideal. Plus mounting that thing with a c-mount seems a bit silly. It would definitely have to be braced.
Scitex, Screen and Imacon didn't use telecentric lenses for their scanners, (I have some of them) but perhaps it was simply for cost reasons.
I'll be very interested to see what Struan comes up with.

Note that there are two Linos machine inspection lenses that are supposed to be better than the Rodagon Ds. See: http://tinyurl.com/72un7ra.
But, just because I know that Joseph is watching, a reversed Nikon 55mm P Nikkor gives very good performance, including low distortion, and they're inexpensive and useful for other things.

[Peter J. De Smidt]

Peter, can you point us to the type of linear sensors that you are suggesting?

I'm curious, though, as to why you think we won't accomplish our goal with a dslr. We've had three people run tests now, all with non-optimized systems, and in each case the dslr setups have out-performed an Epson, and in some cases out-performed a Cezanne in some respects. With better lenses and light blocking we ought to be able to do even better.
In addition the gigamacro people are using a dslr to do some pretty impressive stuff. See: http://gigamacro.com/2011/film-grain...ro-photograph/ And that's with a lens, the Canon 65mm, that we can easily do better than at 1:1 magnification. Plus, there's the huge advantage of being able to use your dslr to take pictures in situations where you don't want to use your film cameras. (For instance, I photograph a lot of indoor judo matches, a situation certainly not suited to LF photography.) So what's the problem again?

[Nathan Potter]

As Struan points out I inaccurately suggested that there was a depth of field advantage to a telecentric setup. The advantage is in the constant magnification ratio achievable while varying the object to lens distance. This property ostensibly allows for more error free and faster stitching of individual frames. Although it appears that restricting off axis rays by use of a supplementary aperture would lead to some increase in of DOF at a cost of resolution. The equivalent NA of the system will decrease.

Nate Potter, Austin TX.

[peter ramm]

Peter, can you point us to the type of linear sensors that you are suggesting?

I'm curious, though, as to why you think we won't accomplish our goal with a dslr. We've had three people run tests now, all with non-optimized systems, and in each case the dslr setups have out-performed an Epson, and in some cases out-performed a Cezanne in some respects. With better lenses and light blocking we ought to be able to do even better.
In addition the gigamacro people are using a dslr to do some pretty impressive stuff. See: http://gigamacro.com/2011/film-grain...ro-photograph/ And that's with a lens, the Canon 65mm, that we can easily do better than at 1:1 magnification. Plus, there's the huge advantage of being able to use your dslr to take pictures in situations where you don't want to use your film cameras. (For instance, I photograph a lot of indoor judo matches, a situation certainly not suited to LF photography.) So what's the problem again?

Peter, I think this is the package that Epson is using. Next step would to contact suppliers like Dalsa, Kodak, Fairchild etc. and get suggestions. Not my area of expertise but there are other savvy people here. Ideally, you would get something with better optical performance that can be inserted into the Epson mechanism.

http://pdf1.alldatasheet.com/datashe...C/UPD8870.html

As to commenting, my viewpoint is based on experience, but it is still only my own view. I don't want to discourage anyone elses' views and I really want you to succeed. I would love to turf my Epson. That said, i believe a DSLR scanner will easily best the resolution of an Epson, that its dmax will be less - though the Epson is pretty crappy as well - and that the the work needed to do even small plates (4 x 5) will mean the systems that are built will not always get used. However, the ability to get really high resolution scans from low cost equipment may be enough to justify the work for some. YMMV

If this were LULA or some other digital site, I might be more optimistic. However, people shoot LF because they love the look of film - at least I think that's why. That look is subtle, with its own set of artifacts and peculiarities. Just think how obsessed we get with how a lens "draws". Replicating that sort of thing takes a pretty good scanning system - with tonality as its primary criterion. I haven't seen anything camera-based that rings my bell, and that includes systems like Gigamacro.

[Peter J. De Smidt]

Peter,

Thank you for the explanation. In my scans of bw film, I haven't noticed a problem with dmax, but that could be a completely different story with something like Velvia. It's true that HDR techniques can significantly increase dmax, as demonstrated in a test presented earlier in the thread, but that's at the cost of double (or more) the exposures of an already exposure happy system.

I'm curious about the real range of information required when talking about high res scans of bw film.

An online guru recently said, "Film has so much noise (grain) that it's self-dithering, meaning that at high resolutions, bit depth is almost completely irrelevant." I don't know if that is true, or even germane, but it sounds like it might be.

In any case with a project like this I'm a try-it-and-see type of guy.

[rdenney]

An online guru recently said, "Film has so much noise (grain) that it's self-dithering, meaning that at high resolutions, bit depth is almost completely irrelevant." I don't know if that is true, or even germane, but it sounds like it might be.

This has been demonstrated even on this forum, I think by Ben Syverson, pretty persuasively.

But that means we have to scan at high enough resolution to resolve the grains, and not just to notice that grain exists. I don't think I'm going to be able to do that with my 5D at 1:1. Maybe it's possible with a sensor that has much higher sensel density. Of course, going to higher magnifications make a lot of things work better except for the level of effort.

Rick "still pondering a collimated light source" Denney

[Ben Syverson]

Rick, your 5D at 1:1 will resolve grain, just not as finely as some of the newer cameras. It should be enough to break up any banding, so I don't think you'd see any benefit from processing in 16 bit...