Background: I have been around for years a 17-40L, which I then "screened out" of newly purchased specimens, but that is unfortunately only really good at f11. With increasing camera resolution, the edge blurring disturbed unfortunately becoming more and were larger prints have long been an annoyance. Unfortunately, there was never a Canon clearly better alternative. My envious eyes fell on the long 14-24 Nikon, I thought even more to the adapter solution, because I always usually Scan manually, but have the discarded due to lack of (or impractical) solution for Grauverlauffilter.
I have now compared the 16-35 II and the Tokina 16-24 / 2.8.
The Tokina is "build like a tank" and much sharper than the Canon, but the Tokina also much bigger and heavier has a strongly curved front lens. This creates an extreme Streulichtempflindlichkeit with strong ghosting that restrict use. A filter solution - and me is a gray gradient at large angles of importance - is possible only with difficulty and with very large filters [...]. Although the Tokina is the worst corners better than the Canon at the best, I have therefore decided for the Canon and avoid the 16mm because at 19mm it is more balanced.
I almost chose the 21er Zeiss, have then but the test results in [...] compared and seen that the 16-35 at 20mm and my relevant field of application (F8) is only slightly worse than the Zeiss, but stop flexible usable.
A word to the filter insert: When you think of a Cokin P system should quickly forget the vignetiert clearly (also sawn-off holder). I have a Lee Holder (100mm) with 82mm wide-angle bracket and Hightec filters in use. At 16mm also vignetiert still sooooo easy, but is acceptable. Moving from a good 17-40L worthwhile only if you need larger prints and quality wirkich plays a role.
But who photographed landscapes and less need no filters, which I put to test the Tokina his heart - as you can see, what actually still so contagious quality in the camera wide angle ...
The expected from Tokina decentralization, incidentally, was no worse than that of the Canons ... tested within each 2 copies ...