You camera actually has two completely different focusing systems. One is significantly faster than the other.
When you use the "live view" mode, the focusing system works just like a point & shoot camera. These use something called "contrast detection" auto-focus. Here's the idea...
Imagine you are going to take a photograph of a simple UPC code. These have plain white backgrounds with black stripes. If the image is sharply in focus then what you'll have are some pixels that are completely black adjacent to some pixels that are completely white.
But if the barcode is poorly focused, then you'll have some black pixels, which transition into slightly less-black / grayish pixels, which eventually transition to white pixels. In other words as we inspect each pixel in a row of pixels, each pixel is only slightly different than the last one... there's not a lot of difference between adjacenet pixels in an out-of-focus image. That's "poor contrast". If you focus the image, then you'll have several pixels which are the same (e.g. a series of black pixels) and then suddenly you'll have a huge contrast difference when you leave the edge of stripe and get into the white background... the contrast change is huge. That indicates strong focus.
The computer in the camera is searching for contrast changes and then it fiddles with the focus to see if it can improve the contrast change.
The OTHER focus system is the one DSLR shooters prefer to use the vast majority of time. It's called the "phase detect" system.
This system does not use the camera image sensor to focus... instead it uses dedicated focus systems which live on the floor of your camera ... directly below the reflex mirror. As light enters the camera, the mirror bounces some of that light down into these focus sensors. It passes through a beam splitter (a prism) which separates the light into phases.
An analogy would be to take a photo, print it, cut the print into two halves, and then slightly mis-align the two halves. It would be obvious to use looking at the image that the two halves were mis-aligned. Also... you would be able to tell if the right half was too high... or too low. You'd also quickly be able to tell if it was just a little bit too hight... perhaps by just a millimeter... or extremely mis-matched by several centimeters.
The phase-detection system is basically doing this with the phases. If the focus point is in focus then there will be no difference between the two phases. If it is out of focus then not only will there be a difference, but the camera can actually tell how much of a difference exists and in which direction it needs to go to correct the focus.
Given the spaces between each of the 9 auto-focus points, when you point the camera at a subject, each point is probably over a subject in your scene which has a different distance to the camera. If you allow the camera to use any of the 9 points, then the point it will always choose is the point which can lock focus with the nearest subject. But due to the way the phase-detect auto-focus system works, the camera can instantly know which of those 9 points can lock focus at the nearest focusing distance AND in which direction it needs to adjust focus to improve AND how much it needs to adjust focus.
The MAJOR difference to you in these two systems is that one of them -- the phase-detect system -- is very smart... it knows exactly what it needs to do to correct focus. It does not guess. It does not "hunt" by trying different adjustments to see if it is making things better or worse. This makes for a very snappy focus system.
The other is the contrast-detect system and that uses the sensor itself. It does this because in live-view mode, the reflex mirror has to swing up and clear of the sensor -- meaning there's no mirror to bounce some light down into the phase-detect focus system. The contrast-detect system has to "guess" it's way to better focus... it hunts and zeroes in correct focus and this takes a bit longer to do.
There is a slightly cheat mode that you can enable. This allows the camera in live-view mode to momentarily drop the mirror so that it can use the phase-detect sensor to correct the focus... then swing the mirror back up to take the shot.
The Canon 70D is a bit of an exception to all of this (you have a T3i but I should mention this exception because I have a feeling we'll be seeing more cameras with this feature.) Canon developed a special sensor technology which is albe to do a version of "phase detect" auto-focus directly on the sensor itself. As such, the 70D (and it's currently the only DSLR camera in the line-up offering this feature) is able to do very fast on-sensor auto-focus and can do continuous auto-focus even when shooting video without ever having to do "focus hunt" and guess its way to better focus.
For the rest of us... we get MUCH better performance by putting our eye up to the viewfinder to frame, focus, and shoot.
I should also mention that the focus motors are in the lens and not all lenses focus at the same speeds... some are faster than others. Canon lenses with "USM" focusing motors tend to be the fastest (there's some variance even among all the USM lenses). The STM lenses are next... fairly quick but not as quick as the USM lenses. The lenses which are neither USM nor STM are the slowest to auto-focus. It's also easier for the camera to focus when it has plenty of light (it's hard to focus in the dark.)
Tim Campbell
5D III, 5D IV, 60Da
