After reading the other new thread on DSLR focusing, I had some questions.
First, and correct me if I'm wrong here, but a camera's autofocus is limited in range?
If so, and you need to focus on something 200 yards in the distance, how does autofocus focus, and it does?
Second, how accurate is autofocus when it actually gets a lock? Can we as humans manual focus as accurately as autofocus? Let's assume autofocus is set to ONE point, not several to avoid averaging.
[a href=\"index.php?act=findpost&pid=146313\"][{POST_SNAPBACK}][/a]
DSLR autofocus uses phase detection. That's a fancy way of saying "rangefinder."
Basically, an AF sensor simultaneously looks through your lens out of two opposite sides of the lens' exit pupil. Its sensors are long thin strips, so it sees a pattern on two strips. It analyzes these patterns and then calculates how far their offset ffrom each other. That lets the camera know how far to move the lens in order to get them to line up perfectly - which equates to good focus. The AF makes that calculation and sends to order to move the lens the appropriat amount and voila - you have focus. Some AF systems might do a double-check and either re-check the AF sensors (not likely) or more likely verify that the lens was actually moved where it was supposed to go. I notice that my Olympus DSLRs will sometimes make a little micro adjustment after the main focus movement. It is this basic mode of look once and then move the lens that makes phase detect faster than the contrast detect in compact cameras. Contrast detect uses a form of trial and error. Phase detect makes one reading and ideally calculates how much and in what direction to move the lens.
Each AF sensor looks out the exit pupil at positions that correlate to different f-numbers. The smaller the f-number, the greater the precision because your "rangefinder" is working with an effectively larger base to its triangle. The more expensive DSLRs have higher precision f/2.8 sensors, as well as f/5.6 sensors. This, BTW, is why AF doesn't usually work much past f/5.6. The AF sensors are trying to look out an f/5.6 window that isn't there. Likewise, the higher precision f/2.8 sensors are of no benefit when an f/4 lens is used. FWIW, the Canon 1D-MIII has four different precisions in its AF sensors: f/2.8, f/4, f/5.6, and one f/8 sensor.
So, given you are using lenses that aren't faster than the sensors in your camera, you should have pretty accurate AF. At least that's true if your camera and lenses are calibrated well. Remember that one big AF move? Part of that includes some camera/lens intelligence on the inertial overshoot to expect with a given lens and degree of movement. There is also the bit about spherical aberration. If you are shooting with a lens that has a lot of spherical aberration what looks like good focus at the f/2.8 may not be good focus at f/4. So errors can creep in there too.
With faster lenses shot at fast apertures, the spherical aberration bit can be a much bigger problem if you are shooting at an aperture significantly faster than the camera's AF sensors. If you pop a fast f/1.4 lens on a budget DSLR that only has f/5.6 precision sensors, you could be in trouble. Maybe not if the lens has minimal spherical aberration. But if it has lots, then you may never be happy with the focus.
My experience with Olympus DSLRs is that the most accurate focus comes from using 10x Liveview manual focus, AF with Olympus ZD lenses is next, split prism manual focus after than, then comes AF confirm with manual lenses (using a Panasonic DMC-L1, and last in line is manual focus using the standard focus screen. Different users with different eyes, lenses and cameras may find a somewhat different order.