453
I modded a scanner into a camera. This is ‘drake V’
(lemmy.world)
Taken on a small group of Islands in the Oslo fjord, called Hvasser. A 15 meter peice of fabric playing in the wind, scanned right to left in 21 seconds. Got really lucky with the clouds this time, allowing a single beam of sunlight in as a highlight.
I love your post, I've been thinking of ways to capture images from large format lenses, and you inspired me to start experimenting with flatbed scanners.
I'm using a Canon N 650U because I had one lying around. I haven't modified it, so the light source still exists.
I'm genuinely surprised how easy it is to get an image. I experimented by putting the scanner directly under the ceiling light in my kitchen, and I can hold the lens over the flatbed and see the image forming well enough that I can get it into focus. Then just hit the scan button and tada, one image of a ceiling lamp. I didn't need to dismantle the scanner, or use a gg screen or anything. It just works with the scanner and the lens.
Since then I've rigged up a system to hold the lens in place, and a slider to move it up and down to focus. And I drape a cloth over everything to keep stray light off the sensor.
Exposure is difficult. My large format lenses don't have any aperture control, so I either have to build one, or use an ND filter. Or if I don't have enough light, I add more continuous light sources.
I saw your comment about vignetting, but I didn't understand it at the time. I've mostly been experimenting with a lens that creates a 100mm image circle, and it works fine. But with another lens, I've just tried to scan the entire A4 surface, and the corners are black with a small image in the centre, even though I can peer under the curtain and see a nice bright image from the lens that covers the whole flatbed. I concluded that the sensor will only accept light from one direction, and can't see light arriving from an angle. So, yeah I'm considering surgery to fix the aperture slit in front of the sensor. I don't really know what to expect when I open it up. Did you find it easy to remove and modify? Can you offer any advice before I get elbow deep in scanner parts?
Also, my images all come out in black and white, which is weird. Objects sitting on the glass come out in colour, but the image formed by the lens scans in black and white only.
The whole rig is too heavy to pick up right now, so I mounted a little mirror on top, so I can point it anywhere, lol.
Thank you!
The n650u looks very similar to the lide 30, so I’ll assume that it’s more or less the same inside.
The sensor assembly is fairly straightforward. A plastic housing that gets pulled by a cord. Once you liberate the assembly from its track/cord/whatever is holding it and moving it around, you get to the good part.
The sensor itself, and the RGB LED that supplies the light is on a PCB that’s mounted to the plastic carriage from the underside, it’s held together by plastic tabs (it’s a few years since I did this, so the details might be off, but it wasn’t a difficult task). Use a sharp utility knife to cut the tabs and remove the sensor PCB. I’ve used strips of gaffers tape to put it all together again. The LED looks like a small white rectangle stood perpendicuar to the PCB, I simply broke it off. This makes the canon drivers throw an error, but VueScan doesn’t care. If you’re squeamish about actively destroying your scanner, you can probably figure out a way to cover it up with some tape or foil.
The circular vignetting you’re seeing is due to the pinhole array, it’s a thin black plastic strip with a bunch of tiny holes set in front of a slit, just rip it off, its either glued or welded in place, but I’ve never had a problem getting it out with some pliers. While you’re messing about in that area, get rid of the prism that spreads the light from LED as well, it probably won’t do much of anything if left in, but it feels better to remove it.
Once you run the camera in this state, you will discover that you get uneven exposure from edge to edge, a linear vignette perpendicular to the axis which the sensor moves. This happens because the plastic housing of the sensor is throwing shade. So use any available abrasive machine, and cut that housing down to its essential functions of holding the sensor in its place, and as a pressure against the glass plate.
Unfortunately I don’t have any large format lenses around, so I’ve stuck a magnifying glass to the front of my camera (you know, gotta keep it punk rock), but the way I handle aperture is by cutting holes into black cardboard (very thin cardboard) at about the size I think will work, and taping the hole across the lens. One of my goals with this project is to try make photography into a tactile, direct and intuitive process so I try to avoid unnecessary machinery.
The reason you’re only getting grayscale is due to how a CIS scanner renders color. During a normal scan, the RGB LED only flashes one channel at a time, and the driver figures out if it’s looking at the red, green or blue signal all by itself, and at the end it interpolates the data to render a full color image file for you to enjoy. I’ve been playing around with the idea of tapping the signal path and use it to trigger an external RGB lighting rig in a studio, to get full color images of models. But so far it’s only a funny thought!
And lastly, the sensor is exceptionally sensitive to infrared light, to the point that the heat from the internal electronics of the scanner causes streaking. I’m using a few pieces of carefully placed aluminum foil to shield the sensor in select areas, and that reduces the issue to a tolerable amount.
Good luck on your build! This project has given me a boatload of insights into both photography, electronics and computing, and I hope you will have a similarly awesome journey.
Edit: im happy to answer any further questions you might have. it would be awesome to see some shots once you get the contraption up and running!
@Leavingoldhabits Hi again, can I ask you some technical questions? I'm struggling modify my scanner. I've removed the clips holding the PCB, and I even made a test scan in that state, and it still worked. Then I spent hours milling out the pinhole array. But when I assembled it again, I get weird scans.
I have modified two different scanners. LiDe 110 - I get a tall narrow PNG file with some digital noise. I assumed I damaged the sensor, or a ribbon cable or something, so I shelved it, and looked for another scanner.
N650U - I took more care modifying this one. I get a full width PNG file, but it has a narrow strip of white on one side, and black everywhere else. I wonder if it is related to the initial calibration that it does before scanning. I left the LED in tact, but I removed the prism, so maybe it's lighting up just that one end of the calibration strip?
Did you have any calibration problems? Have you encountered anything like this? Do you have any advice?
Thanks
Hi!
I’ve only ever managed to get the lide30 to play nice. I have destroyed one 110 and two 220-scanners in my efforts to build a more versatile rig.
What happens if you drop the prism back in?
I’ve also speculated about there being som calibration and automatic gain control going on in these newer models. I believe it could be used to normalize the values coming off of the individual sensor segments.
On my latest attempt at a 220 scanner, I actually built a small dimmable LED circuit that I attached externally to try and influence the calibration I thought was happening, but there was a disappointing lack of results. Come to think of it, I still have that rig laying around somewhere, and the experiment may have been flawed, I’ll have a look next time I’m at my workshop. I’ll let you know what I find.
I know that rig is at least functional, as objects placed directly onto the glass renders crude shadows on the scans.
That’s the long answer, the short answer is no, I don’t have any tips, maybe aside from working on fooling the possible calibration somehow.
Ok, thank you.
I managed to change the image by just shining a torch into it during the calibration and the scan. This is exciting, I might be making progress :-)
Thats really Cool! Thanks for letting me know, this might be the impetus I needed to go back to working on the 220 camera!
So, when I say "change the image", I mean the torch does affect the calibration, in the sense that I get different patterns of stripes based on the position of the torch, but it's still just outputting stripes.
Here are two scans I made by waving the torch around randomly during the calibration, then resting the torch on the glass.
For some reason each pixel is just outputing the same brightness for the whole duration of the scan, except for that black spot where the torch is, which is weird.
So, I re-installed the prism, luckily it just slots back in. I'm not sure if it helped at all, I still get the much the same result most of the time.
I did manage to get this result. It's black at the top because the lid was closed. I opened the scanner half way thru, and the scan turned white. Then I waved the torch over the sensor and got a definite zigzag. And there's a hint of grey in the middle, which is encouraging.
So, the sensor is still working, it's responding to light, just not in a usable way.