Making a scanner with a DSLR

[Peter De Smidt]

Thanks Kirk. A brief skim showed that the template idea might be very helpful. I'll look into it.

[rdenney]

Rick,
As much as I'd like to sell you my Olympus E-520* (with live view! can mount Nikon lenses!), you have the tethering software you need on the disk that came with the 5D. Here's what google told me:
http://www.ehow.com/how_7330275_can-...computer_.html So, it looks like you can park your laptop next to the setup and check focus instantly.

You can't check focus instantly. You can make a preview exposure and see that quickly. Live View, of course, gives you a real-time moving display. I actually find I can check focus pretty well just zooming in on a test exposure using the LCD on the back of the camera.

But I can also see through the viewfinder well enough to focus, as long as I can get my eyeball to the viewfinder window without having to perch precariously on one toe.

Live View on a camera with an LCD that can be twisted around to point forward is an excellent solution, but I don't have one of those.

Rick "whose skunkworks is not well funded" Denney

[rdenney]

Thanks Kirk. A brief skim showed that the template idea might be very helpful. I'll look into it.

We've been operating under the assumption that two images can't be stitched easily if they cover flat, featureless subject matter. Some have argued that grain and dust will be enough to allow that, but I think I'll see if I can get Photoshop to stitch two images across a featureless stitch boundary to see if it's possible. Of course, I might not know if it fails, which seems telling in its own right.

Rick "thought booked for the next few weekends and limited to talk for a while" Denney

[Dave Jeffery]

Wow, 23 pages. I can't read it all but perhaps someone (Frank?) wants to summarize what is a very interesting topic.

We used to make camera-based scanners that created step and stitch images at any resolution you wish. These were mounted on microscopes and did a nice job, but we never claimed to have the DR of PMT-based scanners. Never happen except with hi-bit cryogenic cameras viewing self-luminous targets. Works OK (only OK for complex reasons) at about 4D of dynamic range under those conditions. No resemblance to a film scanner, though.

The core problem (not the only problem) is flare. As soon as you are dealing with a detector array and an illuminating beam, you have flare. That flare limits you to much less than 4D of DR, even at low precisions.

In contrast, a scanning PMT limits flare because the beam is tightly constrained (especially if it is a confocal scanner). That's why drums are better than any array-based scanners (including linear arrays), and why camera-based scanners cannot match the performance of scanning PMTs. You can be fairly clever in limiting the flare component in a linear array scanner, but that only goes so far.

Making a crappy scanning PMT device is easy. Making one with the tolerances to deliver both spatial and bit-depth precision is very demanding (= expensive). So, how many of us LFers would shell out for a new product? Aztec would know. The numbers are far too small.

If we were wanting to build something anyway just to try and do better than an Epson V700 / 750 what would you recommend for a set up using a DSLR?

I was raised in Grimsby, Ontario so can you throw an old fellow Canuck a bone?

[argos33]

Rick,
I agree the problem of the whole unit becoming too high is an issue, however I am going to stick with designing a vertical setup for now. I think for most normal use you will be using the DSLR close enough to the film to stitch that it is low and it shouldn't be an issue. For those wanting to do a quick 1 or 2 frame capture it might start to get high, but perhaps some of the aforementioned solutions would be acceptable (tethered, live view, etc). As a side note I remember seeing a neat video online about tethering a cheap portable DVD player to your DSLR. If the camera didn't have live view you could at least see the photo on a bigger screen to asses focus and exposure. The technology is getting better and better too - it's just a matter of time before you can plug your DSLR into a cheap tablet I'm guessing. Another thought might be to make the whole column pivot where it attaches to the base, so that you could use it in either orientation depending on the height, etc. Seems to me that with some slight modifications (resting points behind the column, reinforced lockable joint etc) it wouldn't actually be that much more work.

As far as the light source, I too was thinking LEDs with diffusion - as mentioned before the light output on a regulated DC power supply should be very consistent, and you could always add a gel filter (between two diffusion layers for example) to adjust the color if it is way off. The rest could be dealt with post-capture. I am thinking LED light banks similar to this one except maybe smaller depending on format:

http://www.amazon.com/LED-Wholesaler.../dp/B001NDU9XG

I also don't see why you couldn't make such a light source dimmable so you could adjust for the density of the negative. This style of light source would also save a lot of room which would help to keep the whole thing from getting to big/high up.

Peter, as a friend of mine likes to quote: "Don't let the bastards grind you down!" This is a great idea and we all appreciate you sharing your photos and hands-on experience thus far. I think it is great that we can talk about an idea like this and develop it together in an open-source manner. Nice job on the prototype - as Rick pointed out it doesn't have to be pretty - just able to test the viability of the design, you can always improve the materials and look of it. I like the idea of using a track for general positioning like yours and on the Aristophot (thanks Frank), and then having a much more precise platform for actually moving the camera up and down to focus. That would help a lot for folks using process, enlarging, and microscope lenses.

I am in a similar boat as Rick in that I am pretty busy right now, but very interested in making my own prototype and have already started planning and setting some materials aside in the shop.

Evan

[Kirk Gittings]

We've been operating under the assumption that two images can't be stitched easily if they cover flat, featureless subject matter. Some have argued that grain and dust will be enough to allow that, but I think I'll see if I can get Photoshop to stitch two images across a featureless stitch boundary to see if it's possible. Of course, I might not know if it fails, which seems telling in its own right.

Rick "thought booked for the next few weekends and limited to talk for a while" Denney

I make a large part of living stitching images in PS for clients needs. The only problem arises when there is little detail in the overlap areas such as a featureless wall in the overlap area-then I have to go to a more sophisticated program. I know this would be a nightmare in many of my 4x5 negatives with allot of sky. I don't even need to test it on film as I basically test it every week in other ways. I don't think grain will matter any more that texture in stucco does. I think the "template" feature in Auto Pano Pro might overcome this issue because you can "tell" the software where the tiles go.

[Nathan Potter]

http://www.ebay.com/itm/Leitz-Aristo...t_11180wt_1348

Here's a nice hunk of hardware, a Leitz Aristophot Macro Photography stand with Linhof 4x5 back. Tinker with that.

Heck through a 45 scan back up there and do like Paul Graham's book "Films.": http://www.photoeye.com/bookstore/ci...?Catalog=ze536

I use one of these Leitz Aristophot units in my darkroom for precision copying. It is essentially a vertical optical bench which is highly configurable. I'd consider it for a scanning DSLR copy setup with X/Y stage at the base and bottom illumination. Well built from steel parts, carefully machined and virtually an antique. The original ring light with 12 tungsten bulbs and opal diffuser is a wonder of uniform illumination.

Nate Potter, Austin TX.

[Nathan Potter]

Wow, 23 pages. I can't read it all but perhaps someone (Frank?) wants to summarize what is a very interesting topic.

We used to make camera-based scanners that created step and stitch images at any resolution you wish. These were mounted on microscopes and did a nice job, but we never claimed to have the DR of PMT-based scanners. Never happen except with hi-bit cryogenic cameras viewing self-luminous targets. Works OK (only OK for complex reasons) at about 4D of dynamic range under those conditions. No resemblance to a film scanner, though.

The core problem (not the only problem) is flare. As soon as you are dealing with a detector array and an illuminating beam, you have flare. That flare limits you to much less than 4D of DR, even at low precisions.

In contrast, a scanning PMT limits flare because the beam is tightly constrained (especially if it is a confocal scanner). That's why drums are better than any array-based scanners (including linear arrays), and why camera-based scanners cannot match the performance of scanning PMTs. You can be fairly clever in limiting the flare component in a linear array scanner, but that only goes so far.

Making a crappy scanning PMT device is easy. Making one with the tolerances to deliver both spatial and bit-depth precision is very demanding (= expensive). So, how many of us LFers would shell out for a new product? Aztec would know. The numbers are far too small.

Peter, you are correct about flare really limiting the Dmax capability of sensor array capture. I think I had suggested earlier in this blossoming thread that using a highly collimated light source would help to greatly increase contrast. Actual tests may show this is highly desirable.

Nate Potter, Austin TX.

[Peter De Smidt]

I like the idea of LED lights. Bright, compact, low power and low heat (for their brightness). They can be dimmed, and there's no reason to use super high bin parts. (I'm also a salt water aquarist, and LEDs are all the rage in that hobby, but light output is critical for a reef tank. I expect that getting enough light won't be too hard for us. One 3w Cree LED would probably be sufficient.)

Another benefit over flash is that you wouldn't need a separate viewing/focusing light. My one worry with LEDs, especially for color, is the discontinuity of the spectrum. The only way to know if this is an actual, opposed to a purely theoretical, problem will be to try it.. I'll be testing flash and halogen lighting for a start, simply because I have everything on hand.

[Brian C. Miller]

The only problem arises when there is little detail in the overlap areas such as a featureless wall in the overlap area-then I have to go to a more sophisticated program.

This is not an issue! The software already knows the size and position and overlap of each segment. You don't have to worry about syncing up with grain, dust, whatever. That is only needed when the software is assembling a jigsaw puzzle made by a delusional epileptic monkey on acid. The whole thing is computer controlled, so each capture is a known segment on a grid.

This isn't the problem you're looking for. Move along.

-- Brian "No Jedi mind trick required" Miller