High resolution and diffraction limit

From 史家胡同:

不太相信高像素的分辨率优势这么快就完全消失了，衍射会减少分辨率，但是什么时候完全抵消？是不是跟镜头和眼神儿有关。。。我只相信真图，要是有人直接拿一个７Ｄ和３５０Ｄ在ｆ１１的图来对比就好了。

Ｐｈｏｔｏｚｏｎｅ有同一个镜头在３５０Ｄ（８ＭＰ）和５０Ｄ（１５ＭＰ）上的评测，即使在ｆ１１，５０Ｄ的１５ＭＰ也是在ｒｅｓｏｌｕｔｉｏｎ上胜出一筹，比如Ｃａｎｏｎ　１０－２２，在ｆ１１，１４ｍｍ焦距，在３５０Ｄ上中心ｒｅｓｏｌｕｔｉｏｎ是１９０９，在５０Ｄ上就是２２０６．５，明显高出很多。其他镜头也一样。

In other words, even at small aperture, beyond diffraction limit, lab tests show higher camera resolution still helps with picture resolution. Why?

The reason:

The final resolution (picture resolution) is a convolution of the lens resolution and the sensor resolution. Rigorously speaking, this is an integration. For simple cases (either zero or one), you can approximate the picture resolution qualitatively (note, only qualitatively, but the trend is correct) in the following way:

1/P = 1/L + 1/C

where P is the final picture resolution in lines per mm, L is the lens resolution, C is the camera resolution.

Case 1: If L=3, C=3 (camera resolution matches lens resolution), then P=1.5

Case 2: Increase the camera resolution to C=4 (beyond lens resolution), then, P=1.7, resulting in better picture resolution than Case 1.

Case 3: Increase C to infinity, then, P=L, resulting in picture resolution equal to the lens resolution. This is the maximum resolution you can achieve classically (by human eyes).

Case 4: Rayleigh criterion is meant for human eyes: Rayleigh decided that human eyes cannot see intensity variations less than 20%. But a CCD can detect 20% intensity variations without any problem. So numerical analysis can extract additional resolution beyond the lens resolution. This has been done in the Hubble Space Telescope case, about 15 years ago.