I think this whole AF situation illustrates how deep are the problems with the design of Canon's EOS AF system. They seem to adjust lenses to compensate for the wayward tolerances in their bodies. Then the lenses don't work right on other bodies which have different tolerance design failures. Canon thought it could use its lens calibration system to relax tolerances and save money, to the point where tolerances aren't accurate enough for f1.4, or 21 megapixels, or, more importantly, a 5-7 µm pixel pitch. So now we have factory fresh and used lenses and cameras which don't focus accurately with a great variety of Canon's own gear in the market. If you buy a used lens it may or may not work accurately with your existing body, or some future body - and the same goes for a new lens too. They have tried to offload the burden of endless re-calibration onto the consumer with the +- adjustment system in the new pro bodies, but the AF system itself isn't reliably accurate anyway, and off-centre AF points aren't even calibrated well with the centre one, whether the centre one is correctly calibrated or not. Users are often not very methodical in their use of the adjustment system, so it produces as many mismatches as improvements.
Basically, EOS AF is hosed in the new era of wide apertures on 5µm, and the only way Canon is going to emerge from this mess is to start producing bodies with more accurate tolerances, and lenses which are accurately calibrated to this new higher tolerance. Old gear will then gradually settle down to the more accurate tolerance, and Canon service centres will eventually have time to breathe again. If only more people would complain about this crap and demand accurate calibration of their lenses WITH MULTIPLE BAD-TOLERANCE BODIES, then Canon would wake up and realise this demands either the old-fashioned solution of better body tolerances, or a much cleverer self-adjusting addition to new bodies' AF system which will quietly learn to optimise focus better during use.
Meanwhile, I'm using manual focus, Liveview, and small apertures whenever I can.
brainiac wrote:
I think this whole AF situation illustrates how deep are the problems with the design of Canon's EOS AF system. They seem to adjust lenses to compensate for the wayward tolerances in their bodies. .
Interesting point, but it maybe it's not fault?
1: When an engineer designs an AF system, he has to make choices & trade-offs. There's no way to connect the electronics directly to our brains (yet.)
So, the way Canon weighs the various AF points may be different from how Nikon does it, and this may differ even more as you go from near-focus to far--focus.
Enter a third-party company, trying to make one lens for both systems. Tough.
2: All mechanical things change with time, and tolerances exist within any given price-point. If you want ultimate precision, you can pay 4X the Canon price for a Leica lens. (Of course, that argument doesn't quite defend the 1Ds III.) - But even a Leica can change over time. I read a post by one pro who said he has all his lenses & camera calibrated to each other at least once a year. Maybe that's the cost of being picky, I dunno.
Maybe in the near future we will see intelligent electronics that can auto-compensate for a lot of these variations, and even be reset every 6 months or so, to deal with time-changes in the metal & plastic. The onboard chip would make variations to the AF for each lens, at many separate distances and apertures. (and FL's for zoom lenses) Ideally, it could even analyze multiple distanced pics of, say, the 1951 focus chart, and do all calculations automatically.
-------------
Anyway, I think that, given the price / performance ration of this Sigma 50, one might just forget about AF completely. Look at it as a manual lens (with the added benefit of auto-exposure.)
This hasn't stopped folks from taking great portrait shots with the Rokkor 58, after all.
brainiac wrote:
Basically, EOS AF is hosed in the new era of wide apertures on 5µm, and the only way Canon is going to emerge from this mess is to start producing bodies with more accurate tolerances, and lenses which are accurately calibrated to this new higher tolerance.
Or maybe its time to include an optional "open-loop" AF feedback system. The new contrast-based live-view AF is telling...
I guess I'll send this in for calibration and see what happens then. I've got 30 days to return it to Amazon, so I'm hoping that will be enough time for Sigma to get it back to me.
Or, I might just let this go and pick up a 35L instead. Not sure yet.
brainiac wrote:
The Leica 180 f2 APO doesn't seem to do bokeh CA.
The longer the lens, the less the effect, it seems to me. But it must also depend on which of a whole heap of optical compromises the designers have made for a particular wide-aperture lens.
But, in any event, I'm interested to see the horrible magenta fringes in the second example above. They seem as bad as the fringing I've seen in samples from the Canon 50/1.2L.
It's a reminder that this Sigma does have failings other than the AF. Still, if it focused properly for everyone, I'd buy one like a shot.
Any owners other than Chez Wimpy reporting flawless AF yet?
I did notice the CA, of course. But as Glassbottle said, I've also seen it in some of the most highly regarded Canon Ls such as the 35L and 50L. I'm not so concerned about it, as it doesn't often arise in the conditions I shoot in regularly.
What does concern me is the AF issue. I just got an email from Sigma and they said they would have it back to me in 5-10 business days after they receive it. I think I'll send it in and see what happens. They want me to send in my 5D too.
One week to go, I'm really hoping the Sigma service centre can fix my lens. I also hope it comes back clean and pristine as I managed to get a UV filter on it with 0 dust between filter and front element
I'll let you guys know what the results are as soon as it arrives back! I got some good news though, I was the #1 salesperson for cameras for my company nationwide! (Australia) hehe.
badlydrawnboy wrote:
I did notice the CA, of course. But as Glassbottle said, I've also seen it in some of the most highly regarded Canon Ls such as the 35L and 50L. I'm not so concerned about it, as it doesn't often arise in the conditions I shoot in regularly.
What does concern me is the AF issue. I just got an email from Sigma and they said they would have it back to me in 5-10 business days after they receive it. I think I'll send it in and see what happens. They want me to send in my 5D too.
I don't know about sending my Canon camera to Sigma, I wouldn't want them to modify it and risk maybe both my warranty and the function with my other lenses. All they need to do is get the lens to conform to specification, they have no business with the camera as I see it.
What is rather interesting and kudos to the people on this thread, unlike with the 50L people seem to be able to admit that the lens could have a problem without having to resort to brand war fistfights. I suppose by dint of already being on the alt forum people are able to see past the 'it must be bad technique' nonsense that dominates the canon forum.
FWIW: I tested one in-store today and after seeing in the viewfinder (using EE-S on 5D) that the lens clearly focused on one spot as the shot fired, the image on the LCD afterwards was focused about 30cm behind. The shot was taken with about 5 meters distance to subject.
There is no point in posting the images as the test was quick and possibly flawed, but it convinced me there is something to this, and it made me not purchase the lens, sadly, as I really wanted to.
Even if I can return it and get it fixed, what do I do with the misfocused images I shoot in the meanwhile...? The store clerk had not heared of this problem and was going to let the managers know.
