Sensor size and diffraction.

Tom Axford wrote:

ProfHankD wrote:

No, Bob's right: you need a well-specified resolution reference to give a CoC (circle of confusion), and that can be taken either by pixel dimensions or by approximating human visual acuity for a specific print size and viewing distance.

Well, you seem to want to re-open that argument. I'm not biting, I stick by what I said. We have been through it all before!

What? Yeah, Bob's post was a bit inflamatory, but I'm not opening an argument, I'm ending one with the formal definition (and even gave some useful implications which you handily edited out in your response).

DoF is computed based on an assumed allowable CoC. That's all there is to it; even wikipedia gets it right. You can argue that a particular sensor format should be associated with a particular CoC based on the magnification needed to create a specific print size to be viewed from a particular distance, and the wikipedia article even quotes some common CoC assumptions for various film formats and explains some of the subtleties of CoC choice, but go back to the OP:

"There is a critical size "hole" for visible light - approximately 3mm (1/8 inch)."

That's completely wrong. There is only a particular diffraction limit once you pick an allowable CoC. The exception to that rule would be an aperture close to the wavelength of the light, which would be a lot smaller than 3mm. 3mm wavelength would be well beyond what cameras can see (deeply into the IR band)... and yes, CoC does vary with wavelength -- diffraction comes earlier for reds than for blues. (Actually, there's also the recently discovered extraordinary optical transmission (EOT)  that sort-of breaks all the rules and allows apertures smaller than the wavelength of the light, but that doesn't really apply to refractive lenses.)

Anyway, as I posted earlier, most engineers designing cameras are not complete idiots and hence don't bother putting an aperture iris in that can go way beyond any reasonable diffraction limit -- this is why so many Canon PowerShots have an ND filter insertion mechanism rather than an iris. On properly-designed cameras with permanently attached lenses, the only thing that can get you in trouble is wide-angle use of small apertures (large f/numbers) that were intended to accommodate the long end of the zoom range. That is not what the OP said, but what he said is useful in that it roughly corresponds to application of this fact.