DSLR Scanner: Camera Supports and Positioning

[ludvig friberg]

I made a sketch based on aluminum profiles and drylin N 27mm sliders. Its quite cheap maybe around 200 bucks. Here is a small animation and a picture of drylin sliders



If you want to fullscreen go to vimeo an watch it there


BOM
With norwegiean prices(everything is much more expensive here... in us probably half the price if you shop around)
SLIDE CARRIAGE 27MM 39kr x 6 = 234
igus rail 300mm 85kr x 4 = 340
##belt 650mm 2.5 50kr x 2 = 100
##pulley 88kr x 2 = 176

steppers 140 x 2 = 240
driver, easy driver 129 x 2 = 258
arduino 200

20x20 profile 3meter 150
extra profile stuff fasteners 200
glass 100

=
1998

[Peter De Smidt]

Ludvig, that's a nice looking system. Thanks for sharing it!

[ludvig friberg]

I looked around on some american vendors and updated the BOM with american prices. Its around $250. Here in norway it costs me around 320 but I already have most of the stuff anyway. I look forward to building this.

SLIDE CARRIAGE 27MM 6 x $5 = 30
igus rail 300mm 4 x $13 = 52
##belt 650mm 2 x $5 = 10
##pulley 2 x $7 = 14

steppers 2 x $15 = 30
driver, easy driver 2 x $15 = 30
arduino $25

20x20 profile 3meter $30
extra profile stuff fasteners $12
glass $10

=
243

[mortensen]

... gets more and more interesting. Again, your work is greatly appreciated.
I still wonder, though, whether the workflow of this dslr-scanner - with mounting, stitching and spotting - will be faster than that of a drum scanner. I know this hasn't been the primary design criteria, but it is quite significant.

[Old-N-Feeble]

For me it's not about work flow because I'll be scanning very few images. It's more about attaining maximum quality for minimal cost. Having stated that, I'm fairly sure I'd be better off financially by paying to have drum scans done for me. However, I'm a bit of a control freak so this idea interests me.

[Peter De Smidt]

Speed depends mostly on how many frames you're going to take. For instance, suppose someone is scanning 35mm film with a D800. One frame should be plenty and will be incredibly fast. Even exposure bracketing, if needed, would take only a few seconds. If you want to get super high res, you could always stitch two or three frames. That will add significantly to the time involved, but it will still be only a couple of minutes. For this use, a manual slide system would be ideal. For 35mm, I don't know of a faster way to get high quality.

With larger film, the dslr system will take longer, just as it would with another scanner. As long as you keep the magnification reasonable, say 1:1 or less, it shouldn't take all that long to do a film. Even with my manual slide system, it only took me about 2 minutes, max, to take the 25 frames for a 6x6cm negative. Currently, the stitching takes longer, although that depends a lot on the image. Manually adding control points is the big killer. I'm hoping that the precision of an automated stage will greatly lessen the stitching time, but we don't know about that yet.

[ludvig friberg]

Like I said before I intend to project a laser pattern on the film and run one full cycle like that first when I start each batch, I will the stitch first using the laser pattern to make a perfect stitch. Then just replace the source image sequence with the real negs, hopefully I can do this with scripting so each set of tiles will lay in different folders and the script just churns through them based on that first laser pattern. I assume this only needs to be done once a day, unless the temperature in the room changes a lot or something bumps the setup out of alignment.

I have done some theoretical tests with computergenerated tiles and it looks neat. What stitching software do you guys prefer? I will most likely look for the one with the best scripting.

[Peter De Smidt]

Microsoft Ice is free, fast, and it supports change of position (as opposed to pivoting around the entrance pupil of the lens) stitches. On the down side, it's not very customizable, and there's no way to add control points. So either it works on an image, or it doesn't. I have the non-pro version of Pt-gui. It's very powerful. Autopano Giga is also quite good. I've not been able to get Hugin to work consistently on my system.

[ludvig friberg]

I was in a large hardwarestore today picking up some supplies for my house and I had a look at the drawer slides they had just for fun. I bought a small ball bearing sliding unit 310mm long. It was very cheap, $8 for a pair. I just tried it out with a dial indicator and it is not bad! Its going +- 0.03 mm or so in Z. I think most of that is due to the non polished/milled surface. That is good enough for a prototype and probably for the final unit too. I think its more fun trying to make this a cheap project.

here is a picture and a video of the device.

[Old-N-Feeble]

You could buy a dozen drawer slides, cherry pick the most accurate ones and return the others.

Cheap is good.
Simple is good.