The following was posted on the EOS3 user group, it was very good info that I though to post this to the EOS group, the contributor is [EMAIL PROTECTED] > > > > > > As I go through the technical details of both, I noticed that EOS3's > > > central cross sensor AF point works only with f4 or larger. Why is > > > that? What's the point of this? I clearly don't understand why is it > > > limited. Is it supposed to be faster than other focusing points with > > > lenses with aperture of f4 or larger ? > > > No one really answered this initail question: why do Canon's > professional AF SLRs use a cross-type AF sensor that requires use of a > fast lens to take advantage of both the horizontal and vertical cross- > type AF coverage provided? To understand why, you have to understand > what happens on the actual AF sensor iself during AF operation. > > All current AF SLRs focus by using at least one pair of light- > sensitive pixels, arranged in a straight line and separated from each > other (something like "--") on the actual AF sensor or "chip". Cross- > type sensors use four such lines, two arranged 90-degrees perpendicular > to the others. > > The camera's AF optics, located under the mirror in the base of the > camera body, split the incoming light into two separate beams for each > light pair using a very precise and tiny optical system. The pair of > light beams is projected sharply on the line of pixels, and by > detecting where on each line pair the light beams are focused, the AF > system can detect whether the subject is in-focus, front-focused, or > back-focused, and in either of the latter two cases send a signal to > drive the lens's AF system in the proper direction. > > Here's where it starts to get interesting: just as you have depth- > of-field at the subject itself, within the camera at the AF sensor you > have "depth-of-focus" -- a designed-in range of acceptable focus within > which the camera gives thumbs-up and declares the subject to be > properly focused-upon, at which point the lens drive is stopped. It's > a tolerance designed into any AF system -- they all have it in one way > or another. > > It's a fact that the farther apart a pair of line sensors are, the > more exact and the less "wiggle room" there is in its decision as to > whether a subject is truly in-focus or not. This is where the concept > of "high-precision" AF comes in. By using line sensors on their pro > SLRs (starting with the first EOS-1) that are spaced further apart on > the actual AF sensor, there is a significant increase in precision of > AF evaluation and much smaller tolerance for the system to determine > what is truly in-focus. > > This issue of precision is really a moot point when using a > moderate-aperture lens at normal focusing distances, like a 28-80 f/4- > 5.6 zoom lens. But it becomes relevant in terms of guaranteeing tack- > sharp results when using fast professional lenses like the 400mm f/2.8, > 200mm f/1.8, 85mm f/1.2, or even wide-angle lenses like the 24mm f/1.4. > With the closely-spaced line pairs of a consumer level camera like an > Elan or Rebel (EOS 30 or 300), it's possible with these fast lenses and > their tissue-thin depth-of-field to run into shots which are OK but > upon close inspection are not tack-sharp, because of the range of > acceptable focus settings that the closely-spaced line sensors have. > > The problem is this: as the maximum aperture of the lens in use > gets smaller (such as, from an f/2.8 lens to an f/5.6 lens), the > diameter of the twin beams of light being projected onto the AF chip by > the AF system's optics becomes smaller as well. It has nothing to do > with either the amount of light you're shooting in, nor the actual f/ > stop you're going to use -- it's only related to the lens's widest > maximum aperture. > > Thus, as you move the line pairs farther apart to increase their > accuracy and precision, you require a wider beam of light than some > lenses can provide. Knowing this, when an EOS-3, or any version of the > EOS-1 detects a lens is mounted with an aperture slower than f/2.8 (f/ > 4.0 at the center-most point only for EOS-3 and 1v), it simply turns > off the sensor pairs that are located far apart, knowing that such a > lens won't fully cover all the pixels on the line pair(s) on the AF > chip. > > You can see this if you look at a picture of the AF sensor in the > brochures of a camera like the 1N, EOS-3 or 1v -- some sensor pairs on > the chip are located far apart, while those arranged at a 90-degree > angle are much closer together. The latter guarantees operation with > the entire range of lenses down to f/5.6 (f/8 at the center point only > with EOS-3 and 1v), while the sensors further apart provide much more > precise information and focusing accuracy for those fast lenses that > need it -- IF you're using one of those lenses. > > Cross-type sensors are a good thing, because in general you're > giving the AF system twice as much information to work with to read the > subject's detail and assess proper focus. On their pro bodies like the > EOS-3 and 1-series, they've taken advantage of the cross-type design to > basically give dual types of coverage, optimizing some of the > information for truly professional lenses that can really benefit from > it. With slower lenses than f/2.8, all 45 AF points in the Area AF > provide the same level of coverage (using vertical line sensors only, > closely-spaced so they're fine with all lenses down to f/5.6), reacting > to horizontal and diagonal detail, as mentioned in earlier posts. > > Bodies like the Elan and Rebel series, and the previous EOS A2/EOS-5 > series, used a center point with cross-type coverage, but both > horizontal and vertical pairs of line sensors on their AF chips were > closely spaced. There was no high-precision AF claimed on these > cameras. The benefit? True cross-type coverage -- twice as much AF > information, very useful when shooting subjects without much detail -- > was available with any lens down to f/5.6 (thus, pretty much any lens > in the system). The problem? For those using very high-speed lenses, > if you were really critical, sometimes there could possibly be little > frame-to-frame focusing inconsistencies, even shooting a static subject > multiple times from the same camera position. > > I know this is long and sort of involved, but it's necessary to > answer the original question which is "why", and what are the benefits > of this system as its currently constituted. I hope this is helpful to > some of you. > * **** ******* *********************************************************** * For list instructions, including unsubscribe, see: * http://www.a1.nl/phomepag/markerink/eos_list.htm ***********************************************************
