• Please use real names.

    Greetings to all who have registered to OPF and those guests taking a look around. Please use real names. Registrations with fictitious names will not be processed. REAL NAMES ONLY will be processed

    Firstname Lastname

    Register

    We are a courteous and supportive community. No need to hide behind an alia. If you have a genuine need for privacy/secrecy then let me know!
  • Welcome to the new site. Here's a thread about the update where you can post your feedback, ask questions or spot those nasty bugs!

Canon autofocus acquisition behavior

Greg Dunn

New member
I've used a variety of Canon bodies over the years, and since the vast majority of work I do is event or sports related, trying to understand and adapt to the focus acquisition has always been a goal. Of course Canon has given us an increasingly flexible set of parameters to adjust in order to optimize performance, but some of the details they have presented seem to be slightly at odds with real world behavior.

The majority of sports work I do is for regional roller derby teams. There are two common situations represented during an event: (1) a solo skater which needs to be picked out from the background or pack of other skaters; (2) a pack of skaters interacting, often shuffling so that any one of them could be frontmost in fractions of a second. I've come to realize that these frequently provide conflicting requirements.

Background: I know that the AF "point" shown in the viewfinder is quite a bit smaller than the actual sensor on the PDAF array, and I always try to allow for this when picking a subject and AF point. I also know from experience that AI Servo is not guaranteed to work well unless I accurately track a subject for a second or so during motion. Whether tracking a single subject or a small group, I always strive to put the AF point(s) in one place and minimize my own movement, to give the AF point(s) a better chance to do their job. I've also found by trial and error that an IS lens providing stabilization of the image generally improves AF performance. I always verify the chosen AF point (when appropriate) during review of the photos. The testing has been done with either a 1Dx or a 7D mk II, and both seem to behave similarly.

Examples: I have occasionally tried to use the "expanded" or "small zone" AF patterns in order to help track a single moving subject. For the expanded AF point, Canon says "It’s particularly well-suited to subjects that either move rapidly (or unpredictably), and for subjects that don’t have a lot of inherent detail"; and that "if it loses detail or falls off the actual subject, the surrounding AF points are instantly activated". Well, much of my work involves shooting teams, often in dark venues, which have varying detail on their uniforms. Initially I thought that using expanded 4 or 8 point AF mode might help. What seems to happen, however, is that if the center AF point loses a subject, whatever is within the array bounds with sufficiently high contrast captures the focus, even if it's a background (many backgrounds in sport have colorful and contrasty advertisements, it appears). I had to compensate by lowering tracking sensitivity and AF point switching speed, but even when they're at minimum, when the primary AF point slips off the subject, it's likely to jump to the background. I've had to finally resort to single point only, and pay rigorous attention to keeping it on a contrasty part of the subject or it will still focus on something else. Often "pumping" the AF-ON button is necessary to force re-acquisition of the subject anyway - and often that's really close to the shot so I'm not giving the AI Servo much time to do its thing.

For zone or small zone AF points, Canon specifically says "the camera will always focus on the nearest
subject, or part of a subject, within the zone." This is clearly false. Initially I tried small zone mode to ensure that the camera didn't jump to the background while tracking a lone athlete, even when they moved near the edge of the AF point. With the sensitivities at minimum, it appears that if a sufficiently contrasty bit of subject creeps into the AF array, the algorithm still prefers it and switches focus; even if that bit is further away, or (again) the background. In fact, using minimum sensitivity for focus tracking appears to work against the algorithm when shooting small groups, because if the subject you originally selected is obscured by motion of the other athletes, it will stick to the first subject and not move to the closer ones, even though the focus point which finally is shown as active when the shot is taken is much closer. It appears that the only way small zone will attempt to focus on the closest subject is if you increase tracking and AF point switching sensitivity - which goes against the need to lower them to keep from focusing on the background. I'm still testing the validity of this approach for grouped subjects.

I've tried every imaginable combination of rigorously tracking the subject, changing AF point modes, and racking the sensitivities up and down while testing. But it seems like the best setup so far is (and this only works with a few bodies) to have one set of parameters for focusing using the AF-ON button, and another set assigned to the * button next to it. That is, have two setups for the most common situations - single subject and group. Switching in real time is less difficult and hopefully will give better results overall (I'm still testing that as well).

Does anyone care to comment on these findings or share their procedures for best AF of moving subjects?
 

Asher Kelman

OPF Owner/Editor-in-Chief
Greg,

Without yet addressing the details of your questions, can you provide b.g. on the focal length, aperture and range of shutter speeds necessary. The first two gives us an idea of the depth of field whilst the shutter speeds needed provides a sense of the light levels your are working in.

Asher.
 

Greg Dunn

New member
I've seen the issue with the following lenses:

Canon 24-70L f/2.8 mk I
Canon 24-105L
Canon 70-200 f/2.8 IS
Canon 70-200 f/2.8 IS II
Sigma 50-100 f/1.8

The shorter lenses get used with the 7D mk II and the others with the 1Dx. All focal lengths get used during an event.

Typically I shoot from about f/4 to f/5 depending on the body - I have to balance between enough DoF to keep a small group in acceptable focus and a desire to minimize details in the background as much as I can. It gets too cluttered when I stop down further. I shot one event (4 days) at f/3.5 with the 70-200 because ambient light was marginal but strobes were not allowed. That was borderline acceptable DoF on the full frame sensor body.

Almost all my work is done at 1/250 shutter speed because I use strobes to get my desired lighting and to stop motion. I have tracked ambient light levels as part of my investigation; the absolute worst case is in our practice space where we get about EV4 (ref. ISO100) from old fluorescents. That's just enough for AF but mandates strobes for actual shooting. Other events range from about EV6 to EV8, where at the high end I was just able to shoot at ISO3200 without strobes and at 1/500 (the lowest I could get away with and still stop motion blur). AF is more troublesome near EV4, to the point where I have been using the Sigma and 70-200 (depending on which body I'm using) to ensure enough light gets to the PDAF sensor. The Sigma does a little better due to its huge max aperture, but does not eliminate the issue.

The AF behavior is better at higher light levels but still exhibits the issues described above.
 
Top