Properly photographing a 3.5" floppy disk for archival is annoyingly complicated. The label has THREE sides!
I've already built an automated system to take a picture of the front of a disk, but really I need to take THREE photos if I want to get the whole thing.
That means either three cameras or I need to rotate the disk 90° and then 180°, which is going to really stress the limits of my mechanical engineering skills.
So the front is easy. The disk slides down a slide, it's stopped by a servo, I take a picture with a camera aimed down at it.
The back... Either I flip the disk, or I have a camera under the disk which takes a picture aimed up.
And the edge is the worst. I can't have a camera aimed at it unless I either move the camera out of the way of the disk, or I make the disk move in an L shape
How about this: I stick with the "stop disk and aim down" method, but I do it on a transparent surface, and I add some mirrors.
Then the disk can be photographed from three sides at once.
If I can rotate the disk sideways I might be able to solve the edge problem. Then I could just have a second camera that aims at the edge.
I could rotate it sideways with some static obstacles, but I may need a servo mechanism or something to do a 180 flip to get the back, unless I do the transparent glass thing.
The disk comes out, it whacks into a bumper and rotates 90° sideways, and three separate cameras photograph it at once, then something ejects it?
The problem is that all possible answers for this problem (other than "don't worry about the edge and back") are either going to be tricky to design and fiddly to operate, and expensive in terms of cameras.
And some solutions are both.
The best designed option would probably be a sort of carrier that rotates. It'd have to grab the disk by the edges, but if it could move the disk from the original orientation, to 90° for the edge pic, then 180° for the back pic, it could then rotate to like 270° to drop the disk.
That might be the most reasonable option.
It would require variable focus on the camera, since the surface it's photographing is closer for the edge picture.
Or just a second camera mounted roughly parallel that is on fixed focus for the edge picture. That might be easier/cheaper to do, actually.
(since multiple fixed focus cameras for raspis can be cheaper than one camera with a motorized focus)
How many floppies are you photographing that you need to automate the process?
Okay yeah that's a lot.
@foone Would a big enough lens be able to image all three sides at once?
...also, destructively, couldn't you break open the diskette and lay the whole label flat?
@kerio A big lens? I talk about setting up a mirror system earlier in the thread. I'm not sure what you mean by a big lens, though. I need to photograph three sides of the object at once.
And I'm not gonna do this destructively, especially because I have thousands of disks like this. I don't want to destroy them all!
I'd look into UHMW (https://en.m.wikipedia.org/wiki/Ultra-high-molecular-weight_polyethylene) it's what industrial grade systems would likely use here.
You can buy strips and probably secure it like your thinking the metal.
@foone That's awesome! I'm wondering if you can avoid having to focus the camera if you rotate the floppy at the edge and center the camera on the edge?
I made a little cardboard model: 1. Floppy goes in, picture taken of back. 2. Rotate 90 degrees, picture of spine. 3. Rotate 90 degrees, picture of front. 4. Rotate a little more and dump the floppy.
The camera distance remains the same without distortion and gravity does the alignment. Hopefully. I have no idea how finicky this would be. 😅
Have you considered painting it green, and putting it on a green background, so you can easily remove everything in the image except the disk?
@mctwist I intend to release my designs once this is more finished, yeah.
The big problem is that it depends on an existing floppy copier which you need to retrofit: so it's not a from-scratch solution. I don't know if I'm going to get to a design I can build from scratch, but it'd be nice
If you let the disk fall/slide into a slot shutter-side-first (and stop it there for a while, presumably with another servo) you could probably use the same camera for front and side.
Alice Averlong🏳️⚧️
Frederic Barthelemy
mmu_man
vfig
Developing Stacy 
Σ(i³) = (Σi)²
Thorsted
James Grimmelmann
1.3.6.1.4.1.61513
Ud😷 Sauer
kerio
ftg
keen456
keen456
Jayne
Jo
ratfactor
David Machaj
Slim Amamou
Fabian Schlenz
Tracy 🏳️⚧️
Mathias
bifo
Wayne Werner
McTwist has moved
I love this, so I
robryk
Esteban Küber 
Whine Spectator
Nire Bryce
Whovian NineThreeSixNine
QCATOPR 
Riley S. Faelan