Making a scanner with a DSLR

[peter ramm]

Right now there's a Multiphot camera offered on the bay for $6,5K, but I had the chance to buy one for a fraction of this, down here in Brazil.
I haven't tested it yet, but it needs a new bellows and I need to adapt a power supply for their 6w bulbs.

I understand there are some stages used on microscope work wich moves the samples on X and Y axis, but I never worked with those things and I guess someone might know if they could be adapted as scanning devices.

Cesar, microscope stages are made for small movements. The largest I am aware of handle microwell plates (about 10 x 7 cm). You might find one that handles 4 x 5" scanning, but the retail on something like that would be upwards of $10K. Never mind the software that runs it.

There have been various stand-mounted macro photo rigs, all very pricey and not really suited to film scans (the illumination path is not optimal). All the dedicated macro systems I know of predate desktop imaging (no stitching correction) and I don't know how well modern ones solve stitching. I have seen modern stereo microscopes used in this way but can't remember how well the software did. Really, macro systems are designed to image single fields at low power, not stitch at higher power.

It would be fun to build something like this starting with an optical bench, scanning mirror system, laser and a rail-mounted PMT package. However, I think the goal here is to make something about the same cost as an Epson, so the fellows are limited in their parts selection.

[rdenney]

The thread has grown so long people are unlikely to read all of it. So, while we gather parts and build prototypes, which will lead to posts that can provide a summary of what was done, it might be useful to reiterate the requirements.

Needs:

The user will scan 4x5 or (possibly) 8x10 negatives, using the DSLR they already own, and additional equipment that can be sourced surplus or used for less than the cost of an Epson V750. The user will use the scan to make prints up to 8x enlargement (it's just a number larger than 4--which the Epson can do). The user will make use of the apparatus for at least 10 years.

The user is not a software programmer and will use products generally available in the market or openly provided from a source that will maintain it.

Some users may add this: The user will start the scanning process and return when the final images is scanned and ready for editing. (I don't add this, because I can live without it and I think it leads to requirements I'm unable to fulfill within my budget, or software requirements that are incompatible with other requirements, particularly longevity, except for those who are software programmers.)

Requirements:

The system shall provide an MTF of at least 50% at 2880 spi (based on 8x enlargement printed at 360 ppi).

The system shall use a standard DSLR, with a standard macro lens arrangement.
--The DSLR shall provide a minimum sensel spacing of 8.8 microns.
--The macro arrangement shall provide a 1:1 magnification.
--The lens shall provide 50% MTF at 60 lines/mm over the entire frame.
--The lens shall provide distortion less than 0.04% (less than one sensel across the frame)

The system shall provide a camera fixing and focusing apparatus rigid enough to prevent more than 4.4 microns movement during the exposure.

The system shall provide a focusing adequate to fulfill other requirements.
--The focusing system shall not alter the magnification

The system shall provide a movable negative stage.
--The negative stage shall allow sufficient movement to scan the entire 4x5 (or 8x10) film
--The negative stage shall hold the film flat enough to fulfill other requirements
--The negative stage shall provide movement precision within 5 microns over the range of movement
--The negative stage shall provide manual movement OR The negative stage shall be moved by an automated system (which will entail other requirements I'll let someone else write)
--The negative stage shall be parallel to the sensor plane within 0.04%

The system shall provide a light source.
--The light source (or correction system) shall provide even illumination within 1% across the frame
--The light source (or correction system) shall provide even illumination within 1% from frame to frame.
--The light source shall support MTF requirements.

I'm sure I've missed a few.

In reviewing these, it seems to me that negative stage has to be precise indeed to avoid imposing geometric distortion that would complicate the stitching process. And the specified lens may be unobtainable for a reasonable price. But if we don't write requirements, we won't fulfill them.

Based on these, I've bought another pair of the precision slides, because I think I'm going to need two slides each for x and y to minimize any cantilevered flexibility in order to fulfill this requirement. That is going to put a strain on my construction skills, to be sure. And I'm now way over budget. But a design is forming in my mind.

Rick "practicing his systems engineering beliefs" Denney

[William McEwen]

The thread has grown so long people are unlikely to read all of it.
Rick "practicing his systems engineering beliefs" Denney

That's me.

Forgive me if this has been mentioned here...

I saw a video of Douglas Kirkland, who shoots some 8x10.

He places his negative on a light table and shoots a picture of it with his DSLR.

Seems to me we should invest in light tables and copystands, make hybrids, and go that route...

[rdenney]

That's me.

Forgive me if this has been mentioned here...

I saw a video of Douglas Kirkland, who shoots some 8x10.

He places his negative on a light table and shoots a picture of it with his DSLR.

Seems to me we should invest in light tables and copystands, make hybrids, and go that route...

So, do you think that would fulfill those requirements? The point is to provide a durable replacement for high-end flatbeds (at least) into the future. It's probably gonna take more than a plain light table and a copy stand, though I have no doubt that's fine for Internet display. Then again, the fixtures we are contemplating are just copy stands and light tables, with two exceptions: Adequate precision to fulfill the requirements, and the ability to move the negative precisely to allow stitching multiple images.

Rick "preparing for a film-scannerless future" Denney

[Peter De Smidt]

In general, I agree with Rick's summary, although I'm taking his word for it on some of the specific requirements.

One where I'm not so sure about, though, is:

"--The negative stage shall provide movement precision within 5 microns over the range of movement"

If you mean x-y positioning of the negative, we might not need to be so precise.
A DX sensor measures 23.6 x 15.8 mm. An overlap of about 25% should be enough for stitching. This would mean moving the negative 11.85mm moving across the y-axis for the second exposure. A 4x5 negative is roughly 95.25mm wide, which will take 9 frames to cover the width of the negative. 9 times 11.85mm equals 106.65mm. A total distance of 106.65 mm minus 95.25 equals 11.4 mm. So the total slop in the system needs to be less than 11.4mm for the width. Doesn't this mean that + or - 1mm per step would be ok, at least for 4x5" negatives?

That's with an automated movement. With manual movement to reference marks, wouldn't + or - approximately 3mm to each reference mark be enough? Take exposure. Slide carrier to mark. Make exposure...slide...

[rdenney]

If you mean x-y positioning of the negative, we might not need to be so precise.

Yes, you are right. I really mean in terms of distance to the camera, but also in terms of rotation to avoid the stitching having to rotate the pixels (which is lossy). But lateral position doesn't need to be precise, if we use software to find the stitch line, as I hope we can.

In the design I was visualizing, my negative holder would be like a flag mounted on one end to the slider. Precision is good in all axes except what I can only describe as the flag rotating around the flagpole. Doing so would cause perspective distortion, not mention loss of focus. Writing the requirement in that way, and being clear about what sort of precision might be needed for planarity to avoid distortion, forced me to put another slider on the other end--the flag is too big and catches too much wind to be that securely supported from one end.

I sure hope the stitching works because I've spent too much on the machinery already.

Rick "probably chasing a pig in a poke" Denney

[marfa boomboom tx]

Some thoughts on the Software:

My best guess for long term users not interested in rolling their own,
even from source code (whatever that is-- 1,2,3 have source and active community of coders.

1- Hugin
http://hugin.sourceforge.net/tutorials/index.shtml

2- Or panotools for a panoramic approach. (the stitcher isn't sourced,so ?risk??)
Panoramic approach
http://wiki.panotools.org/


3- For those who like command lines and think a Stitch approach can be solved in your situation, this is worth your time:
http://www.imagemagick.org/script/index.php
see: 'montage' examples


4- Finally, there is the GigaPan thing. It costs, but gigatools may be "free", I don't know. The gigamicroscope stuff depends upon the software from GP.
Of course, that whole thing is only as stable as the University System and Carnegie Mellon is. Though, it seems the gigapixel imaging phase is picking up advocates, not losing them.

//

[mortensen]

Thanks for the resume, Rick - a good sum up for those of us who didn't have patience to read 42 (!) pages in here.

Now, post some blueprints and prototype pictures

[Nathan Potter]

X/Y stages have the advantage of precision in planarity over a large area (and of course also precision in X/Y positioning accuracy). It has been suggested that the X/Y positioning accuracy is not needed as long as rotational errors are not introduced during translation and I think this is true depending on the completeness of the stitch algorithms. So the planarity in the Z direction is what we want and should not want to pay for precision in X/Y unless that is unavoidable - and it may be.

In fact microscope stages come now in sizes up to 14 X 14 inches or more for studying IC wafers of 12 inch diameter. But we couldn't afford to pay for those and generally they don't come with a very large aperture for scanning over a light source. Dennys' approach of just buying the slides drastically reduces the cost but requires mounting a transmissive plate and a mechanical contrivance to paired slides (paired to maintain planarity and magnification factor as he explains) while scanning. The slides seem like a good experimental approach but need mechanical design and fabrication elements.

Ricks' Requirements write up is a quite proper way to establish performance objectives, in this case chosen to exceed the performance of an Epson V750 by a substantial margin. Certainly I would choose to achieve a nominal 10 um resolution at say 50% contrast but would be overjoyed to see 5 um resolution at 50% contrast as might be obtained using a new Nikon D800E just for an experiment. Budget requirements would keep the DSLR to a minimum of cost based on Ricks requirements.

Nate Potter, Austin TX.

[rdenney]

In fact microscope stages come now in sizes up to 14 X 14 inches or more for studying IC wafers of 12 inch diameter. But we couldn't afford to pay for those and generally they don't come with a very large aperture for scanning over a light source. Dennys' approach of just buying the slides drastically reduces the cost but requires mounting a transmissive plate and a mechanical contrivance to paired slides (paired to maintain planarity and magnification factor as he explains) while scanning. The slides seem like a good experimental approach but need mechanical design and fabrication elements.

Yes, that will be the tricky bit. But I have some ideas. It involves buying some precision carpenter's squares.

The problem I had with X-Y tables was that 1.) I couldn't find one with an adequate range of motion in the surplus market (read: ebay plus only a couple other places) and new ones were prohibitively expensive, and 2.) none that I saw had an aperture for allowing the negative to be backlit. The only solution to that would be to put the light source on the table, which makes the part being moved heavy again.

I don't think using individual slides will be that difficult with the approach I have in mind. But we'll see. The question will be their general availability to those who want to follow along--they are expensive new. But I had no idea what was out there last week, and am designing based on what I find. That may well be the price of entry into this game, until someone can figure a way to productize it.

Rick "making it up as he goes" Denney