Interesting. I was under the impression that the floating RE of the 85 / 1.2L offset the shift.
The RE of the 85L is not a floating element; all but the rear end are floating, AFAIK. The RE is static, and AFAIK it serves to eliminate vignetting, not focus shift. At least, AFAIK, that is what these fixed elements at the back of a lens are used for normally. Note that the literature mentions a floating construction, not a floating element, and the 1 element I am sure about that it is not floating is the rear one. It is conceivable that elements 5 & 6 (which is a group), or elements 5, 6 and 7 move independently of the rest, thus constituting the floating construction. I haven't been able to verify that, however. To me it looks like the front 7 elements move when focusing.
Anyway, 1/4 of DoF at close distances only at F/2 and less at the rest of the apertures from F/1.4-F/1.6 to F/2.5 is well within tolerances for a fast lens, so nothing to worry about. I was amazed by this, however, because my 50L was calibrated to 1/6 of DoF at F/2.
The 85L is a modified Double Gaussian slightly telecentric design if I read the schema correctly, and the 50L is a modified Double Gaussian design as well. These designs are known to have strong SA especially at very large apertures. Used with deliberate care, this is what introduces great bokeh, at the expense of a focus shift potential at large apertures.
In a way, this is what happens, just not in-camera. I understand from the Canon technician who calibrated both my 85L and 50L that these are the hardest lenses to calibrate, which is done with software on an optical test bench. Hardest of the two to calibrate, actually of all Canon lenses, is the 85L apparently.
It is not only calibrated for wide open performance, at least, it shouldn't be. If that was the case it would be a 5-minute job rather than the hour and a half it took, by a trained technician.
It should always be back focus, not front focus, when caused purely by SA. You'll find that the copies that have front focus generally have the worst problems: they tend to get focus wrong at almost any aperture and any distance. At least that is the conclusion I have drawn from all the problems I have read about here, on other forums and have spoken about with fellow shooters.
That would be rather extreme softness, in my view . I couldn't live with that .
After the second round of calibration, when I personally delivered the 50L and spoke with the technician to explain my problem, and what I thought could cause it, it is fine, and well within spec as mentioned above. I did find and do find that focus shift was always worst at F/2 with all 5 copies I tested, both on 40D and 5D, although others had different results. These were all copies from different batches with different date codes.
I would expect it to work better than the classic 5D's AF. It seems the 5D II system takes different wavelengths into account. I expect there to be less problems with AF in fluorescent lighting, just like the 50D, as compared to the 40D .
This is hard with a FF body, due to the angle at which the light comes in. I hope they'll manage to do something about this sooner rather than later, however.
Yes please, although I generally am fine with center point AF - used to this from the old analog days after all .
I would agree. I am also aware that the AF sytem employed by cameras is better and faster than anything I could achieve in 10X the time. I can only do as good, with luck, or aided by Live View or an Angle Finder in extremely slow mode.
Perfect? Of course not. But not too shabby, actually. The 50 / 1.2L has less difficulty nailing it wide open than the 50 / 1.4 (don't know why), but I prefer how it renders the image regardless (most likely contrast/bokeh). Still, if Canon made a 50 / 1.4 II that nailed the focus more consistently below f/2, I think a lot of people would call it a killer lens, even if the optics remained unchanged.
My main problem with the batch of 50 F/1.4s I tested was focusing close-up, combined with the (lack of) bite below F/2.8 at other distances, except for one which was ok at F/2. By that time I had already seen how the 50L rendered beyond the "problem zone", so I knew it was just a matter of finding the right one. When I found one that just backfocused by 2 cm at all distances and apertures, i.e., focus shift that was negibly small and masked by "normal" backfocus, I got that and had it calibrated, twice, to get it to work as it should .
Kind regards, Wim