Ignoring the brightness falloff towards the corners, which is easy to deal with in image editing and may prove to be esthetically appealing, the most important measure of the purple-corner effect is the change in chrominance (that portion of color that is unrelated to luminance). CIELab provides a formula for calculating chrominance: (a^2 + b^2)^.5. Measuring the maximum chrominance error of the four corners of the test images will give us a pretty good idea of the level of the problem.
|WC chrominance error|
|Sony 16mm f/2.8 E mount||5.0|
|Leica Tri-Elmar 16/18/21mm f/4 @ 16mm||6.0|
|Leica Tri-Elmar 16/18/21mm f/4 @ 18mm||6.0|
|Leica Tri-Elmar 16/18/21mm f/4 @ 21mm||5.0|
|Leica Super-Elmar 18mm f/3.8 ASPH||7.8|
|Leica Elmar 24mm f/3.8 ASPH||9.4|
|Leica Elmarit 24mm f/2.8 ASPH||5.1|
|Zeiss Biogon 35mm f/2||3.6|
|Leica Summilux ASPH 50mm f/1.4||2.2|
So, if you think the Sony 16mm is adequately corrected for the purple corner effect, you’d definitely be happy with the Zeiss 35mm and the Summilux 50, which are better. The Leica Elmarit 24 mm is the same, and the Tri-Elmar is the same at 21mm and a bit worse at 18mm and 16mm. The 18mm and 24mm Elmars are clearly worse.
One problem with photographic measurements is that you can identify problems that you were happy ignoring. Some people, after they’ve see the numbers for the Sony 16mm, will look critically at pictures taken with it and want to fix the corners.
Another thing to consider is, although I’ve been calling it the purple corner effect, the errors aren’t always the same hue. Look at the results in the previous post; you can see hues from red to blue, and you can have different hues at different parts of the image.
Great work on this! Just wondering what f-stops you used to determine the WC chrominance error figures.
Thanks for the support. And thanks for telling me about Cornerfix. It’s really useful.
Short answer: f/8.
Long answer: complete testing methodology is at http://blog.kasson.com/?p=1233
As you’ll see in later posts, wide apertures make the purple corner effect worse.