I think what Ray meant is that the effect is not some absolute limit like a cliff but more of a slope. Your post with the text image clearly shows a worsening of the sharpness just as the theory predicts, but not a 'wall' past which everything is immediately dreadful.
[a href=\"index.php?act=findpost&pid=217504\"][{POST_SNAPBACK}][/a]
Thanks for the clarification, Nick. That is exacly what I meant .
We should not forget the Rayleigh's derived formula which describes the resolution limits of a lens at a particular F stop, a limit which is of great interest to astronomers.
As I understand, in it's simplest form for those who are not mathematicians, as I am not, the resolution limit of a lens at a particular F stop, in terms of line pairs per mm, is given by the formula 1600/F stop.
For example, at F16 a lens can deliver as much (and actually more, I believe) as 1600/16 lp/mm, or 100 lp/mm.
On the basis of approximately 3 pixels per line, the new 50D is capable of only 73 or so lp/mm.
Now I know, even if the the 50D were a foveon sensor without AA filter and were capable of recording over 100 lp/mm, it still wouldn't be able to record those faint lines at F16 because the MTF of those lines is only about 9 or 10%. They've lost about 90% of their original contrast as a result of diffraction. Camera system noise, read noise and insufficient quantum efficiency would all conspire to bury such faint signals.
But what about detail at 70 lp/mm at F16 and say 25% MTF? If such detail is of sufficiently high contrast, glittering black specs in the sunshine, I wouldn't be surprised if the 50D were able to record them in circumstances where the 40D would not be able to, simply because the 40D does not have sufficient pixels and resolving power.
Below is what Norman Koren has to say on the subject. The bold italics are mine.
Lenses are sharpest between about two stops down from maximum aperture and the aperture where diffraction, an unavoidable consequence of physics, starts to dominate. For 35mm lenses, this is typically between f/5.6 (f/8 for slow zooms) and f/11. At large apertures, resolution is limited by aberrations (astigmatism, coma, etc.), which lens designers work valiantly to overcome. MTF wide open is almost always poorer than MTF at f/8.
Diffraction worsens as the lens is stopped down (the f-stop is increased). The equation for the Rayleigh diffraction limit, adapted from R. N. Clark's scanner detail page, is,
Rayleigh limit (line pairs per mm) = 1/(1.22 Nω)
N is the f-stop setting and ω = the wavelength of light in mm = 0.0005 mm for a typical daylight spectrum. (0.00055 mm is the wavelength of green light, where the eye is most sensitive, but 0.0005 mm may be more representative of daylight situations.) I've seen a simple rule of thumb, Rayleigh limit = 1600/N, which corresponds to ω = 0.000512 mm. The light circle formed by diffraction, known as the Airy disk, has a radius equal to1/(Rayleigh limit).
The MTF at the Rayleigh limit is about 9%. Significant Rayleigh limits are 149 lp/mm @ f/11, 102 lp/mm @ f/16, 74 lp/mm @ f/22, and 51 lp/mm @ f/32. Larry, an experienced lens designer, finds these numbers to be somewhat conservative because the Rayleigh limit is based on a spot, which has lower resolution than a band. His numbers of 125 lp/mm @ f/16 and 64 lp/mm @ f/32 are derived from a Kodak chart he contributed to Robert Monaghan's Lens Resolution Testing page.
Most lenses are aberration-limited (relatively unaffected by diffraction) at f/8 and below. The OTF (optical transfer function) curve in David Jacobson's Lens Tutorial shows how MTF (the magnitude of OTF) varies with spatial frequency for a purely diffraction-limited lens at f/22.
