Originally Posted by
DavidEccleston
It has also occurred to me that just turning on AI Servo isn't enough. You need to know how to use AI Servo.
When you press the shutter half-way, the camera starts tracking motion. It will need a second before it's figured out speed, acceleration, etc. If you fully depress the shutter before it knows how to track, you'll get a mis-focused image.
While the camera is tracking, you must keep the AF point (or points) over the subject.
Once you're tracking and you fully press the shutter, two things can happen. Older cameras will do what I describe first, and newer ones, I believe, let you choose to skip part one.
1. The camera takes the shot at the current focus distance. You said, 'the the picture, NOW!' and the camera did it. By the time shutter lag, etc are factored in, the subject may have moved out of your depth of field (in focus area).
2. The camera is in continuous burst mode, or a newer camera lets you choose 'focus' over 'shutter' priority. In this case, the camera is not only tracking, but predicting. Keep your focus point on the subject as the burst fires. This can be tricky. There may be a delay between shots as the camera recalibrates, re-predicts, and refocuses. Unlike the first shot, these shots will be taken when the camera has figured out where the subject WILL BE when the shutter opens, not where it was during tracking. These should be sharper than image 1 in the burst.
I found out about this two-step process from an ancient EOS-1 film camera white paper I found online. I've never seen anybody or anything else mention it (aside from me, here on TDP). From my experience, it holds true on my T1i. I think I remember seeing a setting on my 7D to choose the priority, such that the first image will be taken when the camera is predicting, not when you press the shutter, but I'm not sure. I haven't done extensive testing on my 1Ds, but I imagine it follows this two step procedure as well.