I should probably add that in my case there are specific types of rotation: no small rotations, but larger sweeping arcs 40, 60, 90 deg etc (sort of like steering wheel in a bus). The direction of motion isn't important, just the final rest position. Having said that, the previous 2-ring design has some flaws, so I came up with a new 2-ring idea: the outer ring calibrates black, white and mid-gray values for the current lighting conditions. The inner ring just reads a white-to-black gradient (in the same lighting conditions) to get the current angle. The mid-gray is to account for non-linear light response (gamma). I think light sensors can detect 256 colors (360/256=~2 deg), it should be more than enough for my 20 deg accuracy
