Ok, I'm sorry if I offended anyone. The "police" comment was intended as humour.
Hi Michael,
No problem. I do understand your reaction, which even triggered your response to double check what it was that you were seeing different from what physics tells us. One might ask why your example still looks reasonably good?
Well, there are a few things that can help to understand why your images still look acceptable, especially with a bit of increased sharpening.
Physics tells us that for monochromatic green light, say with a wavelength of 555nm, a circular f/22 aperture will have a MTF=0% cut-off frequency at 1/(22 * 0.000555) = 81.9 cycles/mm. The physical (Nyquist) limit for the 5.3 micron sensel pitch of the 645Z sensor is at 1 / (2 * 0.0053) = 94.3 cycles/mm. So at 87% of Nyquist and higher spatial frequencies there will be 0% MTF response, that's what physics dictates. Depending on the chromatic corrections of the lens, red will blur more, and blue will blur less, but the majority of Luminance contribution comes from Green, and Luminance dominates our senses to detect sharpness.
This loss of resolution is of course mostly affecting the details in the focus plane, other detail will already be somewhat blurred, increasingly more as one gets further to the edges of the acceptable DOF zone. It also becomes more destructive for lower contrast micro-detail which vanishes beyond repair. Your high contrast bars and text input will still look relatively unaffected, because most of it is larger than the cut-off frequency (which is helpful for sharpening, but won't restore what is lost).
So Ray is correct in pointing out that 3D subjects
appear to suffer less (because they are already blurred), but they do also suffer. And because the highest resolution detail is lost, it kind of blends in with the already blurred OOF detail, at a more comparable (lower) resolution level.
But as Bill showed, when we want to analyze the amount of loss, a better target should be used. Not to negate what we see in real life scenes, but to determine the best trade-off between our artistic DOF requirements, and loss of quality. Nothing wrong with making a trade-off that hurts technical quality, if it improves the overall image, as long as we make it a conscious trade-off. The image and its intended use should dictate our choices.
As I said before, for the 645Z, f/16 is about the narrowest aperture you can get away with without compromising technical quality too much. It will have lower resolution (especially low contrast structures with micro-detail) than at the lens' optimum resolution aperture, but in the balance of things there is an acceptable loss when it improves the image creatively. Hopefully the viewfinder/live view will allow you to judge that correctly, and I assume the camera offers a stopped down preview option.
Starting at f/18, you will start destroying detail (in the plane of best focus) beyond recovery, so this is where the better creative DOF must outweigh the technical loss of resolution/realism. Again, if the situation provides a creatively better image, by all means go for it, but do realize that it won't enlarge as crisp as wider aperture shots, and it will need a lot more advanced post-processing help to (realistically) fake that impression of sharpness.
I have respect for science, I have respect for facts.
I have greater respect for achieving an image that achieves what I want it to creatively in spite of breaking the rules. The rules and facts need to be understood, then they can be ignored as one wishes.
Agreed, although 'ignored' is a bit strong. Understanding the trade-offs as dictated by science helps us to make better informed judgement calls, especially if we have some wiggle room (maybe f/18 with better technical quality is also acceptable instead of f/22). Sometimes we need to accept lower resolution, if the image (DOF) dictates it (which assumes an informed technical vs creative trade-off).
Cheers,
Bart
P.S. The 81.9 cycles/mm at f/22, also roughly suggests a potential maximum enlargement factor of 16x if we want to achieve a 5 lp/mm printed resolution, thus the 43.8 x 32.8 mm sensor would produce 717 x 537mm (28 x 21 inch) output with the required resolution for uncompromised close inspection, while f/18 would allow 15% larger output with the same resolution in the plane of best focus.