Into the sub-pixel realm: Optical simulation

Hi Marianne,

I studied this question too (with similiar results in some, but not all cases). I looked at effects of sampling, AA-filtering, sharpening, defocus, diffraction and lens aberrations and a way to separate the effects. You find parts of the work at

http://www.falklumo.com/lumolabs/articles/sharpness/index.html

http://www.falklumo.com/lumolabs/articles/D800AA/D800AAFilter.html  (an article you didn't like)

The more general formula is r^p = SUM_i (r_i^p) with p=1,2 or another value. p=2 holds true in an exact way if the MTF(f) ~ exp(-a/f^2) between f=0 ... f=MTF50. That's an S-shaped curve. I found that sharpening is an important ingredient to make MTFs S-shaped enough to make the above formula apply. Diffraction in particular is too flat to make the formula apply in a straight-forward manner. When combining two flat MTFs, the choice of p=1 may be a better one.

Moreover, I developed a way to make the formula with p=2 apply to the combination of diffraction and defocus. This method starts from the result of a wave-optical treatment of the combined effect.

Enough said, I really like your work on this simulator. Congratulations.