What most posters want to know is why calibration is needed.
I'm guessing the front- or back-focus happens due to the distance difference between the AF sensor and the Image sensor. When there is any difference, the AF sensor makes a correction - effectively telling the lens to focus in front or behind the target (from the AF sensor's viewpoint) so the image on the Image sensor is properly in focus. This is the software "fix" for front- or back-focusing bodies. But the amount by which the AF sensor thinks the image should be in front or behind will depend on the lens, focal length, and focus distance. There must be communication protocols for this that can get screwed up (e.g., the lens not communicating the focus distance properly).
I'll call that Theory 1. Fixing this type of FF/BF issue would require both lens and body.
Theory 2 is more relevant to variation between lenses of the same type:
The AF system works at the maximum aperture of the lens. Many lenses are quite soft at their maximum aperture. This softness occurs because light from a target point takes all paths available to the image sensor, but the paths vary in their focal length. I.e., light hits the entire lens (just considering it as one element), but light that travels through the edges of the lens may be focused in front of the IS while light traveling through the center of the lens may be focused behind. The result is a soft image, with the lens displaying a continuum of focal lengths, rather than just one.
Not all lenses behave the same on all bodies. Some bodies are MUCH MORE accurate with certain lenses, than others. The EOS 350D is/was a very notorious problematic AF-body, with a lot of lenses. However, it could also AF GREAT with other lenses.
So, this is not the answer for all AF "problems"
Now, the AF system takes the light from 2 paths near the edge of the lens, and compares them to determine the correct focal distance. In the above example, the AF sensor would tell the lens to shorten focal length so the 2 paths become in phase - focusing properly on the AF sensor. But then most of the light from the image (especially if stopped down) would be coming through the center of the lens and would focus behind the IS. The lens would appear to be front focusing. This could be corrected by calibrating the lens without the body being present.
I believe these theories fit is with Wiba's general argument in the thread.
Cheers,
Nigel
The AF does not take light from 2 paths near the edge of the lens. As you may (or may not) know, the AF sensor has different AF points, all of which get fed different parts of the image. The idea that there actually are two paths near the edge of the lens comes from that internet paper by kerr (or k-something else).
But, just spend some time looking at the sensor optical path openings on the bottom of your camera. You will notice a pattern which somewhat resembles the AF point pattern in the view finder. For the 450D with its diamond shaped AF point pattern, you can clearly observe 3 not totally rectangular openings, with shapes indicating the basic square of the "middle" 5 points, and extensions in those shapes for right most, left most, top most and bottom most points.
I am not prepared to open up my camera to study the separate optical lens systems/paths, though
.
The AF sensor optics use an OOF image to determine focus. In a way, it is looking at what we see as "bokeh" as end product, the widening doubleness we start to see in OOF areas is basically the "left" and "right" image the AF sensor compares (or top and bottom).
What the AF will never do either, is "determining the the correct focal distance". It can't do that. It only can see a phase difference, but does not know what focal distance that translates to with the lens system, the lens system is unknown by the AF system.
Lenses that show a focus shift when closing down are not just calibrate-able just like that either. Else lenses like the 50mm f1.2 L would not show that behavior.
That focus shifts CAN be a cause for inaccurate focussing is evident, though. They will then show that behavior on different camera bodies, of course. Where it CAN vary, depending on the AF sensor, and used AF point.
We already know for instance that on a 5D using the outer points with the 50mm f1.2 can resolve some of the focus shift problems.
Again, this is not the explanation for all AF problems.
