Hi Bart...
How do you go about determining best radius setting? I, for example, use a G9 m43 20mb sensor with variety of lenses.
Hi John,
Long story short, I take shots of a Slanted Edge (actually two edges, one near horizontal and one near vertical) and based on the edge profile(s) I fit a blur function that tells me how much blur radius is required for a perfect edge to achieve the same blur as observed in the edge profile. The Slanted Edge shots are from a series taken on a focus rail that allows me to find the perfect focus for the chosen aperture (the focus rail shot with the smallest blur in the series is used for analysis).
After analyzing a number of lenses, the same pattern emerged, the best lenses come close to 0.7 radius blur in the image center for the optimal aperture, and the blur radius (to compensate for in Capture Sharpening postprocessing) grows on either side of the Aperture range, but with a good lens diffraction takes a heavier toll with narrower apertures than residual aberrations do at wider apertures, similar to the example chart I've shown. There's more to the story, like close focus changing the magnification factor, and thus the blur radius to compensate for, but I don't want to complicate things too much.
I've since become better at visually judging edge sharpening effects (like halo artifacts) to allow visual correction that correlates with the numerical analysis. I also use Capture One Pro as my main Raw converter, and they've introduced an automatic (deconvolution based) Diffraction control that's very effective for improving narrower aperture shots. So the diffraction blur induced loss of sharpness is much less in my Raw conversions, and I can use Capture sharpening in postprocessing that's pretty much the same for any Aperture. FocusMagic usually only requires a blur radius setting of 1 pixel for ultimate sharpening.
On detail, do you determine artifacting using ‘alt’ key masking while adjusting detail?
No, in Lightroom I look for exaggerated / ugly artifacts in predominantly low detail areas, but I have to admit that I've grown an allergy for such artifacts (because I now know what good sharpening looks like), so it has become easier to spot them. Lightroom's implementation of deconvolution is pretty quick and dirty and when the detail control is set too high, the 'dirty' part kicks in. Only then, if there is noise amplification in e.g. mostly featureless sky regions, I reduce (not eliminate) the noise amplification by masking (which admittedly is a nice feature as it is implemented in LR, unlike the sharpening workflow as a whole).
Of course, Capture One also has a similar feature nowadays, although implemented differently and more geared at suppressing halos from post Raw conversion sharpening with a more USM-like type of sharpening (which I only use when in a rush, or replace by its output recipe sharpening after resampling, for which it offers a proofing preview). For better quality output I take the additional time and effort to post Capture sharpen with FocusMagic, followed by Creative Sharpening using TopazLabs Detail at the intended output resolution.
Cheers,
Bart