Now that I have the Sony A7rII it would be a good thing to find adequate sharpening settings. The aims were:
- Avoid/limit haloing
- Sharpening that allows large prints
- Having a reasonable starting point
The method I used was to shot a test chart and analyse with Imatest and tune parameters until I got a good level of sharpening, without MTF going excessively high (say below 105% on the sharpened image), little haloing and high SQF.
I came up with three methods that show good promise:
- Using "scene/landscape" sharpening in Lightroom. (Keeps parametric workflow)
- Not doing any sharpening in Lightroom and do all sharpening in FocusMagic. Default parameters and 75% sharpening, in order of avoiding oversharpening on low frequencies. (Breaks parametric workflow)
- Not doing any sharpening in Capture One and doing all sharpening in FocusMagic. (Breaks parametric workflow)
These candidates remained after some "math based" evaluation. From the three candidates I made small crops printed on A4 size paper, corresponding to 95x145 cm prints. That is around 37"x57". These prints were pretty similar. I would say that the C1/FocusMagic combo was probably the best, doing pixel peeping on those cropped prints.
The lens I used here was the Sony 90/2.8G at f/4. Results may vary.
So what I have found was that the "scene" sharpening in Lightroom is pretty OK with the A7rII. Capture One had a small benefit, I was able to pick it from the other two.
Using the scene setting in LR didn't break the parametric workflow, no TIFFS and no Photoshop involved.
All images has shown some colour demosaic artefacts, just a fact of life with non OLP-filtered images. These artefacts were definitively visible in prints. The lens needed to be stopped down to f/16 to remove all artefacts. I did not test moiré removal tools. These normally are quite good at reducing colour moiré but don't really help with monochrome artefacts.
So what is my take? I guess that I am quite happy with the "landscape/scenic" setting in LR 6. It has been developed by Jeff Schewe. It is truly parametric. The other methods give marginally better results.
This is all for now.