Is diffraction really a big deal in all situations at f/16?
Resolution is probably the most tested and discussed property of camera-lens combinations, yet usually either the test, the conclusion, the discussion, or even all of them is flawed in many ways. To give food for though I offer three images. I took two photographs of the test chart presented by Bart van der Wolf in this small article. I added “Nyquist circles” to show the theoretical resolving limit. I used the very sharp Cosina Voigtländer 50mm f/1,5 lens on Sony A7 and took one of the images at f/2,8 and then another at f/16 which is deep in the zone where diffraction blurs the image significantly. I made a rather quick alternate version of the latter with some (non-optimal) deconvolution as sharpening.
Make sure that you look at the images without any web browser scaling – clicking on an image may reveal a larger, full size image, which is the one that is of interest.
f/2,8 – the unidirectional anti aliasing filter allows colourful aliasing to present itself.
f/16 – contrast is clearly lower and diffraction has certainly paid a visit.
f/16 with some deconvolution and sharpening – I am sure I could get this to look better, but without putting much effort at all the results is not bad at all I think.
One should remember that especially deconvolution may increase noise in the image, so in light starved situations it might not be ideal, but looking at the image above, shooting at f/16 isn’t necessarily the disaster many think it is.
For easy comparison, here’s an image with all above samples side by side at 300% enlargement. You may need to click it once or twice to see the whole image.
To exapand the topic to the question of proper sampling, below is a sample picture which is made from one of the colour channels (green – I should have used blue, but green’s ok) as filtered by the Bayer colour filter array on the image sensor. The camera used was 24MP full frame camera – the crops were taken at f/32, f/45 and f/64. It is clear how even at f/32 diffraction is not nearly enough to prevent aliasing and that even at f/45 slight aliasing exists – f/64 seems to provide enough blur for good sampling. In other words, much much more resolution would be needed. More on this subject is here.
Abe R. Ration's lens and camera blog 