01-09-2014, 06:35 AM
Jabez02, it's not only the very narrow DOF which makes it difficult for the AF module to focus - given this module is carefully adjusted and the lens reacts always the same on various distances between camera and object.
It's also not only the tolerances between AF module and sensor itself which ideally should be exactly parallel. In reality I've seen the planes not being parallel which lead to slightly different focus results of the left and right outer focus point. And if I have 51 focus points, I'll use them. But with high resolving sensors and just normal manufacturing tolerances tiny differences suddenly start to matter. It's very demanding to manufacture these devices.
A decenterred lens is another failure source.
And at least, in reality we rarely have sensor parallel walls to focus at - this happens just for reproduction tasks and there we close the aperture to the best resolution two or three stops. In reality we do have DOF of what we're aiming at. Look a the topography of an eye - the distance between eyeball and eyebrow is visible with f/1.4, with f/1.2 even more shallow. When aiming the little AF point's frame, to which part of the eye it is referring? The eyelash, the little wrinkles on the skin? If there's some depth in space within the AF point's frame, the AF-module starts to guess - at least, I can't predict, on which point the best focus will be.
With a fast wide angle in combination with Nikon's comparatively big AF point frames (in FX it's better, in DX the frames cover too big areas), almost every focus action becomes a guessing. When Klaus starts to test the Sigma 18-35/1.8, he will notice that, too. I think those fast lenses deserve the best AF and I'm pretty envy on the EOS 70D's modern AF design. DPreview found significantly more keepers because of good sharpness than with other, older AF modules.
I can't blame Nikon - the "fastest widest" wideangle they offer is 24/1.4*, for them there's no need to nourish a strong and partly better competitor. But this is short thinking. If I knew, I like this lens for whatever reasons and would start to buy APS-C DSLR without any or only few lenses and I would know the AF of this particular camera is the best for it - I'd be stupid to go to the Nikon system.
* which costs 260% of the Sigma zoom.
It's also not only the tolerances between AF module and sensor itself which ideally should be exactly parallel. In reality I've seen the planes not being parallel which lead to slightly different focus results of the left and right outer focus point. And if I have 51 focus points, I'll use them. But with high resolving sensors and just normal manufacturing tolerances tiny differences suddenly start to matter. It's very demanding to manufacture these devices.
A decenterred lens is another failure source.
And at least, in reality we rarely have sensor parallel walls to focus at - this happens just for reproduction tasks and there we close the aperture to the best resolution two or three stops. In reality we do have DOF of what we're aiming at. Look a the topography of an eye - the distance between eyeball and eyebrow is visible with f/1.4, with f/1.2 even more shallow. When aiming the little AF point's frame, to which part of the eye it is referring? The eyelash, the little wrinkles on the skin? If there's some depth in space within the AF point's frame, the AF-module starts to guess - at least, I can't predict, on which point the best focus will be.
With a fast wide angle in combination with Nikon's comparatively big AF point frames (in FX it's better, in DX the frames cover too big areas), almost every focus action becomes a guessing. When Klaus starts to test the Sigma 18-35/1.8, he will notice that, too. I think those fast lenses deserve the best AF and I'm pretty envy on the EOS 70D's modern AF design. DPreview found significantly more keepers because of good sharpness than with other, older AF modules.
I can't blame Nikon - the "fastest widest" wideangle they offer is 24/1.4*, for them there's no need to nourish a strong and partly better competitor. But this is short thinking. If I knew, I like this lens for whatever reasons and would start to buy APS-C DSLR without any or only few lenses and I would know the AF of this particular camera is the best for it - I'd be stupid to go to the Nikon system.
* which costs 260% of the Sigma zoom.