What is the maximum velocity possible for autofocus (focusing speed) on a Canon 5D Mark II

Question

This question is very specifically worded, because I lack the knowledge to formulate it in general and also proper terms. I am an amateur, so don't hesitate to improve this question with better wording.

I am interested in optimizing my position (point-of-view) when shooting (thriathlon) bike races. I have found that the autofocus works better (not so surprisingly) when the riders are passing by at wider angles in contrast to when coming up from sharper angles. In other words, shooting from the opposite side of the street works better than from the side where the riders pass by more close. I am pointing at a single rider at a time, half-pressing the trigger, then turning myself with the rider, until they are in a position I want to actually shoot. This is most of the time somewhere where they are nearest.

Some pictures are very sharp, other out of focus. The autofocus will need to keep up with the dwindling distance, and I am interested what is the max speed that a rider can have, relative to my position, while the continuous autofocus still works.

I also tried manual focus, with a guessed average distance, but this was very unreliable, as not all riders passed at the very same distance of course.

I am shooting with a Canon 100-300mm f/4.5-5.6 lens, at approx 100mm when on the same street side, and 200mm when from the opposite side, at the moment when the riders actually pass by. I have a Canon 5D Mark II. I used the AI-Servo mode, the single active focus point being the one that is most to the "back" of the rider.

What is the minimum or optimum distance (how can I calculate it) to the rider, that my autofocus still works reliably, given a steady hand.

You are also welcome to challenge my technique:

• maybe a one-shot autofocus would be better?
• Should I position myself at a slight turn?
• I could try a more inclining spot of course, with slower speeds, but this question was with a flat terrain in mind.
• Would a better lens work better in general (i.e. a larger aperture)

Answer 1

The lens you are using is has a ring type USM focus motor... these are very fast.

The latest top of the line cameras do 120 AF calculations per second; the 5DII does not, although I do not know the actual rate.

In most situations phase detection AF does not predict *focus distance. Focus tracking predicts the next location of the subject (2D) but not the distance. The required focus shift is then determined by the phase offset of the AF images at that location. When the shutter is activated AF is then locked and shifted slightly based upon the focus distance change rate history, in order to account for distance change during shutter delay.

(*Some Nikons have a "3D focus" mode that does add in distance prediction in certain situations.)

The point of that is that AF is almost always slightly behind in distance calculation/determination; no matter how fast it is. So as things become more demanding with shorter distances/higher relative speeds you need to make room for more error. You can do that by reducing the relative speed (longer distance, shorter FL, slower subject), and by increasing the depth of focus (smaller aperture). Note that reducing the FL and increasing the subject distance will not make a difference if you crop in post to compensate.

Additionally, you probably want to use AF point expansion which allows the camera to use surrounding focus points to assist in tracking... IDT the 5DII has zone AF. I would also select a focus point towards the front of the cyclist, not the back. Note that AF point expansion tends to focus on the nearest point. Also, only the center focus point is of the cross type which is generally more accurate.

Finally, panning only fully assists with focus and motion on one axis; where the two paths line up. So optimally your images need to be timed to be taken in that zone. Panning also cannot compensate for simultaneous motion in multiple directions; e.g. a cyclist pumping up and down. It is very hard to differentiate small amounts of motion blur vs slight defocus, but I think motion blur is affecting both example images.

If the cyclist will pass at a known distance, it is also possible to use manual *"zone focus" instead in order to eliminate AF variability entirely. When combined with timing the shot appropriately, zone focus eliminates all variables of error.

*zone focus is simply manual pre-focus with an adequate depth of field allowance.

Answer 2

Disclaimer: I never used Canon and I never tracked high speed objects.

1. AF sensor is an image sensor albeit having low resolution. If you track moving target with camera the background starts moving and gets smudges on the AF sensor therefore making it easier for camera to decide what to focus on.

2. I do not expect larger aperture lens to work better because brightness is far from being the limitation IMO (and because DoF is already too small).

3. AF motor speed depends on how far the target is and (as you pointed out) at which angle it's moving relatively to the eyesight. If a bicycle is riding towards you at constant speed the motor will have to work faster over time. There's very simple test you can do: ask somebody to ride at constant speed towards you and do slow burst of photos. You will then see with your own eyes the limitation of AF motor (the bigger the speed the further away the rider on the last sharp photo will be). I bet it will be significantly harder to find this information online because there's no standard testing methodology for this.

You did not explicitly ask about anything other than AF and your pictures do not have good resolution but IMO you should shorten your exposure (the scarf looks smudged in first example and the print on the shirt looks smudged in the second example even though it's apparently inside sharp field) and close down the aperture (the driver does not even fully fit inside depth of field on second example). This will inevitably increase noise but you need to compromise to get sharp images.