privatebydesign said:
First off, all lenses focus when they are wide open, however that isn't the end of the story. AF modules have an effective aperture, Canon AF modules have been said to have an aperture of f2.8, just like the standard focusing screens have an aperture of f2.8 or so.
Canon's (and any other company's FWIW) phase detect AF sensors don't introduce this F/2.8 limit for fun and to create pain. Quite to the contrary: this limit aperture is an essential element of such an AF system.
This document gives a good explanation how such systems work. The AF sensor "sees" even less than the full F/2.8 aperture, it "sees" just two small sections of the outer ring of this aperture.
privatebydesign said:
Ever tried manually focusing an f1.2 through the viewfinder? It is very difficult because you are seeing a dimmer and deeper dof view than the lens actually is.
This is a property of the ground glass used in most modern cameras and only marginally related to the AF system. The AF system does not look at the ground glass. In fact there are replacement parts for your camera's ground glass which will "see" larger apertures than F/2.8, yet the AF system will remain the same.
privatebydesign said:
Once you get to slow lenses wide open at f8 the view for the AF module becomes so dark it can't reliably attain accurate focus, it is looking for contrast but the lights are so dark there isn't enough, hence the slowdown in AF speed when we use TC's, and why some people find f8 easy to use (bright sunny day with a 100-400 and 1.4 TC with a high contrast bird in the sky), and some people find it totally useless (same lens but very late afternoon trying to focus on a bear in the woods, lower light levels and much lower contrast).
The problem with F/16 lenses is NOT the darkness of the image, even very bright subject matter won't AF correctly at F/16. The problem is that the AF sensor doesn't receive
any light from such a lens, as it looks only in the direction where an F/2.8 lens would send light from but an F/16 lens doesn't. Replace the term F/2.8 with F/4, F/5.6 and F/8 for the respective AF sensors.
privatebydesign said:
5/ Focus shift has nothing to do with AF.
6/ There isn't an offset value to which lenses are focused after AF has been achieved to allow for focus shift.
I have a Sigma F/2.8 28-70 EX DF lens which proves you wrong: even in AI servo mode it will reliably and consistently front focus on the center AF point of my EOS 3, and that front focus is strong enough to be very visible even in the view finder. This would be utterly impossible without stored correction values which tell the camera "if your AF sensor is happy, move x amount in direction y anyway".
Doug Kerr calls this focus correction BFCV in
this posting. The problem with Canon's 50L is that these BFCVs are tabulated for focal length and subject distance, but not for selected aperture. Evidently Canon did not consider this issue when they initially laid out their AF algos, and later AF algo revisions put more emphasis at making third party lenses incompatible than at making the 50L work correctly.
The posting also states clearly, that "The AF subdetectors each work with rays through a small subaperture, essentially all rays passing through the overall aperture at a small range of distances from its center.". I hope this finally puts the "the AF sensor uses the whole F/1.2 area to focus the 50L" myth to rest.
