cross-posted from: https://lemm.ee/post/44155947
I’d think they’d handle this with calibration. It doesn’t need to be as sexy as commercial, it just needs to have a reasonably easy process to fix it.
Something like when you get a new lense, you aim it at a laser difraction pattern on a clean wall.
Now you don’t worry about minors differences in body or lenses.
It would be super cheap to make a laser difraction grid. You could map the lense deformation because you know the lines on the grid are straight. This would be solely for mapping the properties of the lens / mount and how to handle defamation profiles. Once you dial in the lens you probably wouldn’t need to run it again assuming it can id the lens when you mount it.
I would say you could use red green and blue lasers and look at convergence, But I’m not sure in any decent hardware that that would actually be off
Edit: you should note, iPhone already does this for face ID. It’s not really that much of a stretch to make it go the other way.
“It would be” so you haven’t done this but speak confidently about it being cheap and accessible?
