PDA

View Full Version : DIY Robotic Scanner Back?



maxr
28-Jun-2011, 17:14
A long, tortuous story. For the past year or so I've been building robotic camera stuff using Arduino boards and inexpensive servos. You can read some of it here: http://www.maxotics.com First a thermal imager, then a robotic panorama head like the gigapan (only built from the ground up). A couple of weeks ago a friend of a friend gave me some photographic equipment to sell (from her late husband), including some (whoa this is an expensive hobby!) 4x5 equipment and lenses. I'll probably sell before my 30 days is up here, but just in case I don't I'll post in a month.

While I'm here, I'm curious is how many people have tried to build scanning backs. The cheapest one seems to be $6,500. I think it can be done cheaper. That's coming from someone who spent thousands to build his own robotic pano head when he could have bought a gigapan for a few hundred :)

I've also built some crude DOF adapters for my video camera, so I figured I'd do the same thing. Have the view camera show the image on some sort of ground glass and then raster image using a macro lens on an X/Y thing.

If you can point me to any people/sites that have done, or are working in this area, I'd be interested. I'm also happy to answer any questions, of course.

My other somewhat related question, if you're not going to go the chemical route, why not just do stitching to get the same LF effect. For example I did this test of my daughter using 27 photos stitched together from DSLR shots.

http://farm4.static.flickr.com/3599/5830375518_2c1c29ae9b_z.jpg

Greg Lockrey
28-Jun-2011, 19:02
How about this one: I call it my Poorman's Scanner Mod II

I am able to get 575 mp files with this set up just moving the rear standard as I don't have mirror box issues to contend with.

Sinar P rear standard... Studio Tool adapter and bellows.... Olympus E-P1 12 mp M4/3rds camera..... and with the Sinaron Macro 180mm f/5.6 lens

I am able to scan very large art pieces rivaling my flat bed scanner (Epson 10,000xl) in sharpness and color fidelity.

maxr
28-Jun-2011, 19:11
Interesting. Are you taking images using a common nodal point (like a panorama), or are you sliding the camera top/down, left/right? I also take it you taking about 50 captures? If so, you do those all my hand?

Greg Lockrey
28-Jun-2011, 19:26
Interesting. Are you taking images using a common nodal point (like a panorama), or are you sliding the camera top/down, left/right? I also take it you taking about 50 captures? If so, you do those all my hand?

In this instance it is a parallel movement of the rear standard to the work (up/down- left/right). The limits of the camera rear standard allow for about 50 separate 1/2 frame 35mm images. I can double that if I move the front standard but for purposes of copying flat art work and due to using cross polarization lighting I get weird effects. So I limit the process to just the rear standard. It would be very cool to motorize that function btw.

I did try using panoramic method (nodal point) and stitching in PS but since I use polarizers on both camera and lights it gets weird effects copying flat pieces at such a close distance. There is a "sweet" spot using cross polarized lighting and any movement off that spot shows up in the stitch I have found.

maxr
28-Jun-2011, 19:38
I don't understand much of the view camera terminology. What the front and back standards are, etc. I'll do some research.

The right solution would probably use stepper motors, but I would use servos, which I'm used to, are cheap, and easy to use. What has really bedeviled me in robotic photography is gearing and tracking.

If you were doing only left/right it would be a faily simple solution. You'd put your camera on a rail (you can use U-shape aluminum from your local hardware store and skake bearings) and a winch servo with a string that wraps between camera and motor. On a servo it's easy to send it a command and tell it to move a certain number of degrees. So on a 180 degree servo, you might find that each position is 5 degrees. So you'd program it to do 5 degrees, wait, fire the shutter (many solutions for this), etc.

The 4x5 view camera I have (which I'm supposed to sell, but could test with) is an Osaka, mahagony 4x5. Do you have any idea if that would work. If not, in theory, could I just test with a 4x5 print? Are you focusing on a screen, or some other way?

Greg Lockrey
28-Jun-2011, 19:47
The rear standard is a geared mechanism rise/fall, left/right shift, fore/aft tilt and port/starboard yaw where the film is placed and in this case the camera body. Front standard is for the lens while it has movements up/down, left/right and fore/aft etc. it is more manual. I can focus using the camera's screen in real time and it is done on the rear standard. I would imagine that a simple gearing system could replace the rise/fall knob and left/right shift thumb screw for using stepping motors. This camera is so sturdy I can easily take the 50 shots moving the standard to all positions in about a minute by hand. Your camera is a typical "field" camera and all movements are pretty much by hand. The focus is probably geared.

maxr
28-Jun-2011, 20:00
I'm noodling some ideas. My big question is about any image degredation by taking an image on the ground glass instead of the sensor getting directly exposed. Or put another way, how much loss of quality do you get between your method and shooting film and having that professionally scanned? Thanks!

Greg Lockrey
28-Jun-2011, 20:04
The sensor is "open" to the lens anyway.... I don't get the question. I don't use a ground glass, the digital camera took it's place. As for quality to film? My clients sometimes complain that the images I make are too sharp and show too many of their imperfections on their canvasses. :) The only thing film has is a larger D-max compared to digital but when necessary I can use HDR imaging to make up for that short coming. I haven't had to do that in the 12 years that I been making digital copies of art work. Perhaps if I shoot a landscape scene where the lighting range is beyond the camera's sensor I would need it. I'm not that fussy in that regard.

maxr
28-Jun-2011, 20:08
OH! So you take the lens off the EP-1? Or never put one on ;) NOW your photo of your setup is starting to make a lot of sense to me.

Greg Lockrey
28-Jun-2011, 20:23
OH! So you take the lens off the EP-1? Or never put one on ;) NOW your photo of your setup is starting to make a lot of sense to me.

Oh YEAH.... I'm using no lens on the digital camera..... :) it replaces the film. But what you thought IS how those old Polaroid digital to film scanners worked. They were merely small film cameras taking an image off of a CRT screen.

The Studio Tool kit is how to marry the camera systems. Cambo makes a camera specifically for digital cameras and there are other manufactures out there with medium format adapters etc. Studio Tool allowed Canon to Sinar or Horseman adaption and was relatively inexpensive. I used this for a long time but got an adapter to use the Olympus as there is no mirror box getting in the way. The mirror box left me to only 80 mp files. I don't think they are available any longer. The other systems are very expensive. There are cheap adapters that have sliding backs but they are primitive. Your camera could take one of those but they aren't geared movements and depending on the camera lens selection is limited to certain sizes.

maxr
29-Jun-2011, 07:00
If I'm going to do it, all I really need is a large format lens, of which I have 3. I also have an enlarger lens, not worth much, I could probably test with. I'd then just build a crude box/bellows thing.

Does that sound right to you?

My big problem is a robotic X/Y axis camera mover. There are plenty of kits/ideas in the CNC world, but I want to make something cheap that you, or anyone could build.

What I've done with my panobots is use counter-weights to make the camera virtually weightless. In that way, cheap servos can move it up/down, left/right.

1. I get/build an open-end box that will take a large-format lens in the front.
2. I build a sliding open-end box for the camera movement plane on the back.
3. I build an X/Y camera mover that will take X number of images from the front lens projection.

I may not need a large sensor. The 4/3rds are after better clarity and depth of field, but for what you're doing you don't really care if you're getting 100 images from a small sensor or 20 from a large one, do you?

BTW, I will probably use a Sigma SD14, which has a large sensor and, I believe, has the most color accurate sensor available.

Any thoughts on this plan? THANKS for all your insight so far!

Greg Lockrey
29-Jun-2011, 08:38
Basically you have the idea... Large format lens to give you covering area for sure. The Sigma 14 is a DSLR, which means the mirror box will get in the way at about 6-8 exposures verses 50 with a sensor with no mirror in the way. Big sensor or small doesn't make any difference just as long as you can get the mega pixels for the area your working in. 575 mp in a 4x4" area is pretty decent. Most color fidelity can be corrected in PS. You won't beat a Sinar P rear standard for stability and would be the easiest to adapt to stepping and/or syncro motors. Everything you need to make is already there. The next thing is to hook up your camera to a bag bellows and to the Sinar frame. All kinds of adapters are out there and can be engineered to work. Sinar parts on the used market are dirt cheap compared to buying new. My set up cost about $2,000.00 to make and that was $600 for the Sinar F with P rear standard and other parts, The lens was $700... but it is top of the line Macro and the E-P1 can be had for under $300. The Studio Tool was $400... which is a lot for a bellows and hardware but considering the hassles it eliminated in camera movement it paid for itself.

engl
29-Jun-2011, 12:05
Sigma cameras with Foveon sensors have rather poor color accuracy, I'd avoid using them for this application. Reds are all over the place, as a result of being at the bottom of the sensor stack. The layered design has inherent color metamerism issues that can not be corrected.

They do however have comparatively high resolution in single-color areas compared to Bayer array cameras. They also do not have an anti-aliasing filter. They create false detail (as any sampling without filtering will do), but unlike sensors with Bayer arrays do not create "color moiré".

My guess is that the best camera to use is something like a Panasonic GH2, G3 or GF2, all with ~16 megapixels and rather weak AA filters (which is good, since diffraction will work as a AA filter anyway when using LF lenses).

maxr
29-Jun-2011, 13:07
Thanks Engl, the Sigma camera lost me at "the mirror will go back in 30 seconds no matter what you do." :) The challenge here, I take it, is a precise X/Y movement mechanism. The sensor will be the least of my problems if I take this on. Right now I'm leaning towards a cheap webcam, which is easy to move and control through the PC. Obviously, I'd love to go out and buy tons of equipment but like many people here, I'd imagine, I reached my marriage limit a while ago! Another possibility is any Canon that can be hacked through CHDK. I have a lens-dead G5, but I don't think it will take a photo with an error displayed. Something to research. If the webcam works, the Agent V webcam uses a CCD instead of CMOS. I have one. Going to check how easy to take apart. If you have any cheap sensor ideas I'd love to hear it.

boris
29-Jun-2011, 13:55
a short movie: http://www.youtube.com/watch?v=RkdP5OdYwjo&feature=related
and a forum: http://www.kolor.com/forum/f20-merlin,orion-and-papywizard?id=20
i use this system with good results. total cost without dslr are around 400€. in about 10 minutes you can scan the scene with 300 megapixel.
boris

engl
29-Jun-2011, 13:57
I'm not sure, but I think there is a risk a small sensor (from a compact or web camera or such) will have strange color artifacts when used far from the center of the frame, especially if using a wide lens. Sensors have a limit to the incident angle before the rays start going through one color filter and hitting the sensel supposed to go with another color filter. Microlens arrays on the sensor might also have issues with high angles in incidence. This is more of a problem as pixels get smaller, since the thickness of the color filters and microlens arrays can not be arbitrarily thin.

In short, before building anything, start by making sure that the small sensor still provides an image without crazy colors when used to capture areas far from the center of the image circle. If possible, try with a wide lens.

maxr
29-Jun-2011, 14:17
Hi Boris, I'm all too familiar with pano technologies. I even built a robotic head from the ground up. I don't know if there are any benefits to scanning a large format image vs doing a non-nodal point stitch. I found this forum because I'm selling some view camera stuff for a friend and got side-tracked. Perhaps Engl can answer the benefits of large format vs stitch photography.

On the small sensor, sounds like you're saying if you're going to use a small sensor you'd probably have to scan it in a slight parabola, which is sounding complicated. I'll try to go with a large sensor then, if I succumb to the challenge.

engl
29-Jun-2011, 14:40
It might also work, it really depends on the specifics of the sensor in your web camera. Resolution is relatively low so the pixels are not as tiny as for example 16MP cell phone cameras. Butchering a web camera sounds like fun so I'd give it a try, just make sure to test it far from the center of the image circle as well as the center.

I'm not quite sure what you mean by the "benefits of large format vs. stitch photography" though. Large format used how? Film? Stitching how?

maxr
30-Jun-2011, 05:41
The pros/cons of taking 10 full-frame digital photos with a non nodal point and stitching them together vs taking a single Large Format 4x5, for example.

Greg Lockrey
30-Jun-2011, 07:07
The pros/cons of taking 10 full-frame digital photos with a non nodal point and stitching them together vs taking a single Large Format 4x5, for example.

Digital is faster with more work and a cleaner process.
Film is easier to take the shot but takes longer to process and is messy.

JohnnyV
30-Jun-2011, 14:35
Check out the following pages... I'm sure there's some info to assist you:

http://people.rit.edu/andpph/text-better-scanner-cam.html
http://people.rit.edu/andpph/articles.html

Michael E
30-Jun-2011, 16:02
I have read about other people who based their LF scan back on a cheap desktop scanner. Unfortunately, I don't have any links to offer. But I thought the concept was very interesting, especially for users of 8x10" cameras.

Michael

maxr
30-Jun-2011, 16:29
Thanks! Awesome stuff. Will take me some time!


Check out the following pages... I'm sure there's some info to assist you:

http://people.rit.edu/andpph/text-better-scanner-cam.html
http://people.rit.edu/andpph/articles.html