Real-world AF performance - shooting soccer

by Carey Rose

Processed to taste from Raw. Canon 70-200mm F2.8L @ 200mm | F2.8 | 1/1000 sec | ISO 1600. Photo by Carey Rose

In a similar vein to what we photographed for the Nikon D5 review, I headed to West Seattle's Delridge neighborhood to shoot a friendly soccer match. It presents a similar number of challenges to what we saw with the D5 - that is, low (and mixed) lighting, fast, unpredictably moving subjects with ample distractions (but not so fast moving that I couldn't follow the action), and plenty of opportunities to try both Canon's recommended settings and the updated iTR system.

Before we get to that, though, it's worth noting that the 1D X II comes with an 'Anti-Flicker' setting that aims to sync up the shutter during bursts with the brightest point in the pulsing of some forms of stadium lighting. Even though it drops the burst rate a bit, I found it to be highly effective and left it on for almost the whole duration of the shoot.

Now then, let's start out with the Canon-recommended autofocus setup.

AF Point Expansion

Both the user manual and built-in autofocus guide on the 1D X II suggest 'Case 4' as a starting point for a soccer match. Indeed, the camera labels 'Case 4' as being 'for subjects that accelerate or decelerate quickly,' but with the caveat that it's not best for subjects moving at a constant speed. The camera also recommends adjusting 'Accel./Decel. tracking' to +2 (instead of the default +1) if the AF system can't keep up. This makes sense - soccer players tend to move erratically, so the AF system needs to be responsive enough that if a subject stops or suddenly moves, focus is updated appropriately. Increasing the 'Accel./Decel. tracking' setting does exactly this.

In addition, the Canon Professional Network (here's their guide to the original 1D X) suggests using AF Point Expansion with four additional points - above, below, left and right - to track the action in case your chosen point slips off your subject. In practice, this setup works well. Very well.

Note: These are cropped screen grabs from Canon's DPP 4 software. They are not representative of the 1D X Mark II's image quality. They also do not show the 'helper' points being utilized in addition to the highlighted point.

1 2 3 4 5 6 7 8 9 10 11 12

As you can see, leaving the camera on default settings worked well, tracking a player moving toward the camera while not refocusing on a momentary distraction. This is a good representative sample of how the camera would repeatedly behave with these settings. So overall, thumbs up.

Just out of curiosity, though, I went ahead and bumped 'Accel./Decel. Tracking' up one notch, as the camera recommends may be useful. The difference in behavior, even though it was just one parameter bumped one notch, was pronounced.

1 2 3 4 5 6 7 8 9 10 11 12

So, despite this adjustment being meant to affect the acceleration or deceleration of a specific subject, the camera interprets this by rapidly re-focusing on a foreground distraction. Presumably, this is because the camera is mistakenly recognizing that distraction as the original subject rapidly accelerating toward the camera, and then rapidly accelerating away again (and this was repeatable, not an isolated incident). This is exactly how Nikon's 25-point dynamic area AF works with subject movement set to erratic, and was something I was expecting to happen. But then, with these same settings, there was this:

1 2 3 4 5 6 7 8 9 10 11 12

Count 'em - that's six frames out of focus because of a distraction before the camera snapped back to my original subject, even though I was following him perfectly well with the center point. What's more, I completely missed that player's kick. Not really what I was looking for. So, I switched 'Accel./Decel. Tracking' back to its default value for Case 4, and all was well again.

The moral of the story, for me, is this: I honestly don't buy the argument that it takes years to get the most out of the autofocus system on this Canon (and even if it was true, that's not a good thing). But you do have to be very careful with your adjustments, and be willing and able to adapt and adjust quickly, sometimes during an event, to come away with the best results.

Now, let's take a look at how Canon's iTR fared.


As stated previously, Canon's iTR technology (Intelligent Tracking and Recognition) works a little differently than Nikon's analogous 3D Tracking. Instead of using one point, like the Nikon, it keeps a 'cloud' of AF points on your subject, and isn't as 'sticky.' In my experience, the system as a whole seems to lean a lot more on depth information than pattern and color recognition from the metering sensor, whereas the Nikon seems to treat both sets of information more equally. In general, though, a system that weights phase (distance) information more heavily than pattern recognition is more likely to be distracted by other objects at similar planes.

But in any case, why would you want to use iTR? Well, in addition to the greater compositional freedom it gets you (as discussed on the previous page), it's also a good go-to mode if you aren't terribly familiar with the game in front of you - so rather than having to follow unpredictable action during a crowded scene yourself, you can let the camera take over. It also removes the requirement of strictly keeping on point centered over your subject, which can be difficult for a fast moving subject at telephoto focal lengths. Much of the time, it worked pretty well.

1 2 3 4 5 6 7 8 9 10

This isn't a terribly taxing example, admittedly, but it's a good illustration of the general performance of iTR - it works just as well for tracking a single distinct subject moving downfield as it does with a subject surrounded by possible distractions. On the face of it, this looks pretty good. But this also tended to happen with alarming regularity:

1 2 3 4 5 6 7 8 9 10 11 12

As you can see, the 1D X II's iTR system had no problem following its subject with a number of autofocus points - but it did have a problem maintaining focus. Inexplicably, the camera racked focus toward the camera (i.e. front-focused), rendering the entire scene blurry, before re-focusing properly on the original subject.

Like the system's (good) behavior in other modes, this wasn't an isolated incident, but I could never find a particular rhyme or reason to its occurrence. It just seemed to happen randomly, which is also something we've seen during testing in our studio.

Summing up

The 1D X II's built-in guide was a great starting point for shooting soccer, as was the suggested autofocus area mode. As long as I could follow the action myself, the system performed well despite all it had working against it (and, I should note, while shooting wide-open at F2.8). That said, increasing the system's responsiveness to deal with erratic motion came at the cost of jumpiness that could (and occasionally did) lead to sequences of out-of-focus shots. This is an unfortunate tradeoff of this system - we noted similar behavior with the Canon 5DS/R.

Canon's iTR technology has certainly been improving with each new camera the company releases, and the 1D X II is no exception. Unfortunately, we've found it to be generally less effective than Nikon's 3D Tracking, possibly because it appears to still rely heavily on phase information, as opposed to image analysis and pattern matching off the metering sensor. It's not pinpoint precise enough to track the eye of a face (similar to our experience with the 5DS/R), and it often exhibits some odd re-focusing behavior - possibly because of confusion over what the actual subject is. That said, it is still a good general-purpose setting if you don't like to set your own focus point or if you're having trouble following the action yourself.

Now that we've covered team sports, let's see how the camera does at the rodeo.