Erik,
I did take a look at Tim's article and found it provocative in a helpful way. He maintains that humans can resolve 117 cy/mm (lp/mm) rather than the usually quoted 73 cy/mm as the criterion for the diffraction limit. In the old days, lens testers used the USAF line bars in determining the resolution of lenses, but more modern thinking is that one must take the contrast into account as well as the resolution figure. When interpreting lens tests reported as MTF, contrast at lower resolutions is given more weight than high resolution at low MTF.
Hi Bill,
While true, one needs to recognize that these MTFs are usually measured before (!) sharpening. As illustrated by your sharpened versions from shots of my test target, the higher spatial frequencies get boosted to levels that approach those of lower spatial frequencies. The bi-tonal USAF target and similar, are un-fit (due to aliasing and ringing artifacts from sharp edges that may fall in alignment with the pixel grid, or not, i.e. phase errors) for discrete sampling systems like our sensors (but then they were designed for film, and are perfectly suitable for that).
This is taken into account by SQF (subjective quality analysis). Bob Atkins gives a simplified explanation of SQF. He notes that human perception peaks at about 6 cy/degree, which corresponds to about 1 cy/mm on a print viewed at 34 cm (about 13.5 inches). With these facts in mind, 117 cy/mm at 2-5% contrast may not contribute that much to perceived sharpness and we should place more emphasis on resolution at 50% contrast (MTF of 50%). Indeed, Norman Koren points out that it does not make sense to measure resolution where it is extinguished (at low MTF).
Before sharpening. This is where additional tools like Topaz Labs Detail make a difference. They additionally address these lower spatial frequencies and can apply more or less contrast (even guided by luminance of color contrast) by changing the amplitude of that detail level, assuming it is there is sufficient quality to begin with. Deconvolution sharpening of the highest spatial frequencies also boosts lower frequency quality to a certain degree.
Our MTF 50 results at f/22 do indicate a severe loss of resolution.
And the spatial frequencies for those MTF50 scores are already pretty low, especially when enlarged for output.
Perceived resolution is a complex psycho-physical process, and therefore Michael's impressions based on shots of subject matter that's familiar to him, are interesting (although hard to quantify). And he is of course right that there is more to a successful image than resolution alone. It will take some objective testing to see where the perceived resolution is coming from. Based on objective facts, it's going to be easier to make balanced decisions between DOF (how much is enough, e.g. f/16 or f/22, in some situations) and resolution (how much do we limit magnification potential by our choices).
My
DOF planning tool attempts to address several of these issues, and also reports if diffraction affects the desired COC limits. Section 4 of my
Slanted Edge tool web page, allows to independently (does not require the prior 3 steps but will use them if available) play with different scenarios, such as that diffraction adjusted COC limit. It reports the estimated MTF50 spatial frequency for adjustable sensel pitches and allows to copy and (e.g. with Excel) plot estimated MTF curves for diffraction combined with different levels of COC blur.
Cheers,
Bart
