I’ve been experimenting with the Pi Cameras ever since they appeared for my Super8-focused telecine project. In fact it started to seem realistic just when the cameras were released. The best thing about them is they’re using reasonably small sensors so achieving the required magnification of the tiny Super8 film frame was easy.
I experimented with both Pi Cameras - v1.3 (5Mpix OmniVision OV5467 module) and v2.1 (8Mpix Sony IMX219 module) in both cases adapted to the M42 mount allowing me to use my favorite lens for this kind of work - Flektogon 35/2.4. This set up doesn’t even require a reverse mount, just some macro rings. Now, the extra resolution doesn’t seem to make much difference. The amount of details is roughly the same, perhaps it’s just the sample films I used, but my impression was that I couldn’t really get more resolution from the tiny frames anyway.
What I hoped the newer camera would be better at was dynamic range, noise, white balance and especially the uneven color cast, which seemed to be a common issue with the v1.3 camera. This is much better with the v2.1 camera, but there’s some uneven shading, presumably from the built-in lens correction algorithms. There’s also a bit of uneven red-colored noise patterns across the image, somewhat similar to what older compact cameras shown when pushed to the limits. I have a bunch of sample photos but haven’t had the time to put these together as comparisons.
My personal conclusion was that I’d probably want to get a better machine vision camera for production quality, or perhaps give some mirrorless system a try (Nikon 1 is promising due to its smaller sensor size). But Pi Cameras are both (5Mpix and 8Mpix) good enough for “proof of concept” and for fine tuning, and the newer camera may be sufficient for preserving home videos.