Quote Originally Posted by DAdeGroot View Post
...

On an APS-C body, diffraction creeps in from f/11 onwards. You can usually get away with f/16 and not notice too much, but beyond that, and with a sharp lens you'll see softness compared to a wider aperture.

.....
in many cases this may be 'correct' but from what I've noticed, diffraction is more about lens design(design brief??) and less about the size of the sensor.

Tammy 70-200/2.8 is sharp at f/2.8 and starts to diffract at about f/8. Not very noticeable, but on a resolution chart(the cheapo downloadable version with warnings not to print it up on your home printer version ) I've noticed a slight loss of very fine detail with this lens at f/8, compared to f/5.6 at least, and it looks closer in sharpness/detail at f/8 as it is at f/2.8.
Not an overly concerning issue tho, as the loss is very slight and not noticeable in a real image anyhow.

But I think Steve is on the right track, as someone(can't remember who now) pointed out to me a while back. When I use my old manual Tamron 300/2.8 it gets sharper from f/4 to f/8 and maintains that sharpness level at f/8 up to f/16(there goes the f/11 rule ) but when I add the Tamron specific teleconverters, there's very good sharpness right up to f/32 where I simply ran out of light!(at 30sec exposure, and I was too lazy to delve any deeper into the curiosity). So as the TC's are added and stacked, it was mentioned that the iris is being moved further away from the film/sensor plane, and the 'diffracted image is now being relayed to the sensor via more lenses(hopefully not diffracting or altering the image any more).. but the images produced were definitely sharper at f/22 than at f/11, using various combinations of TC's on this lens. Subject to lens only decreased by a few inches at most, maybe 75mm, over a focus distance of 5-6 meters, so the subject to lens difference in distance could be regarded as irrelevant (or 0.015% difference). But the iris to sensor plane difference was huge with the addition of the TC's.

Then there's the design of the lens to be taken into account too.
Diffraction doesn't necessarily have to exist at smaller apertures, but I think must be a design compromise that the engineers have to deal with as best as they can
If the lens is of a smaller aperture design(such as those found in the MF/LF format, where a 135mm f/8 or 300mm f/8 prime lens is not uncommon, then the engineers don't have as much of a compromise to deal with as they would if the lens was an f/2.8, or f/1.4 type. There's obviously a very good reason why that 50mm f/1.4 lens only stops down to f/16, and not f/22 or f/29 as most f/2.8 lenses do.
I doubt very much that a fast lens couldn't be designed to delay the onset of diffraction till much later than the normal f/11 or whatever, but at what cost? Who's willing to pay $10k for a fast prime that's still as sharp at f/22 or f/29? I suspect no one really, as the lens's primary purpose would be for it's speed.