RobGalbraith.com
Go to advertiser website.
     Home
     RSS
     CF/SD/XQD
     About
     Contact
Go
Go to advertiser website.
 
An analysis of EOS-1D Mark III autofocus performance - Continued
(This is part 2 of the August 1, 2008 article update. Part 1 is on the previous page, part 3 is on the next page)

eos_1ds_mark_iii_sunrise.jpg
Morning Glow: Canon EOS-1Ds Mark III (firmware v1.1.2) + EF 16-35mm f/2.8L II, One Shot AF, ISO 100, 0.5 sec, f/8 (Photo by Rob Galbraith/Little Guy Media)

Q. It sounds like you keep the autofocus on AI Servo even when shooting static subjects. Have you tried One Shot also?


When the new firmware came out, we started and then soon after stopped doing One Shot testing. We started because the goal for this article was to include some thoughts about One Shot AF. We stopped because I'd personally forgotten how much I dislike this autofocus mode. That's because for what I shoot and how I shoot, putting the camera in a mode where it can't track movement, even the small movement of a bubbly portrait subject, is not smart (and yes, you guessed it, I messed up part of an assignment because the EOS-1Ds Mark III was set to One Shot when it shouldn't have been).

Just enough testing was done to reveal a few things. Whereas the older firmware can introduce a change in focus distance with each One Shot activation, even if the AF point is aimed at the same subject at the same distance each time, the newer firmware seems to give more repeatable and accurate focus results under the same conditions. But if the test is done slightly differently - setting the focus to some other distance, then focusing one time on the subject using One Shot, autofocus variability seems about the same as before: there's too much.

To give yourself a reasonable chance with One Shot, it looks like you'll want to pump the autofocus two or three times with it pointed at the same area of the subject. Alternatively, switch the camera to AI Servo, configure it to autofocus only when the rear AF-ON button is pressed, then "lock" the focus by releasing AF-ON. Not only did AI Servo produce more consistent stationary subject autofocus than One Shot (albeit in limited testing), it also means you don't have to switch AF settings when the subject starts to move. Just press AF-ON.

Obviously, there are situations and shooting styles that are well-suited to One Shot. But ultimately, we're not the photographers to do the One Shot assessment, and we haven't spent nearly enough quality time with this autofocus mode to say more.

Q. Does the new firmware improve the camera's tracking of fast moving subjects?

This is perhaps the easiest question to answer, because in all but backlit or shadowy conditions, the EOS-1D Mark III with firmware v1.2.3 and EOS-1Ds Mark III with firmware 1.1.2 act much the same as before when pointed at a subject that's moving rapidly.

If you recall the grading system we used in the December 10, 2007 installment, you might remember we assigned an A if the camera's autofocus really got the job done, a B if it mostly got the job done but produced a few too many blurry frames, and Cs, Ds and Fs to indicate that the AF system wasn't up to the task. The lower the grade the more out of focus pictures you could expect, and B was the threshold; a B- or lower meant the autofocus performance was unacceptable for a pro-class camera.

The EOS-1D Mark III with firmware v1.1.3 earned several B- grades, all of them for tracking in fast-moving outdoor sports photography, and there's nothing about v1.2.3 that suggests these grades should be raised. With the newer firmware loaded, focus doesn't shift as far out in front when it shifts, though it shifts about as often as before, and it throws in a few more instances of backfocus error too. But the net effect is the same: the camera captures too many photos that are unusably blurry.

Firmware v1.1.3 earned a B for basketball action back in December. For this article we've only been able to simulate the game by having players run fast action drills, but there's nothing in those pictures that suggests the EOS-1D Mark III with firmware v1.2.3 will be other than a B going forward. Which is acceptable, though not an improvement.

There are two caveats, the first one minor, the second one not so minor:
  • First, EOS-1D Mark III autofocus with firmware v1.2.3 loaded is very slightly slower off the mark. This may affect its ability to bring into focus a volleyball player executing a dig, as noted earlier, though we'll have to wait for volleyball season to see if this is so. We have felt the slight lag, however, when trying to quickly compose and shoot players leaping up to head the ball in soccer, and during the frequent passing of rugby. The EOS-1D Mark III still feels like the quickest camera around when the AF system is first activated, it's just not as quick as before.

  • Second, the EOS-1D Mark III with firmware v1.2.3 gets a headache when it has to photograph backlit or really shadowy subjects that are moving quickly. Here are three examples:

    • During a late afternoon soccer game, we tried alternating between frontlit and backlit shooting positions and, while the difference wasn't quite Dr. Jekyll and Mr. Hyde, it was headed that way. When frontlit, the EOS-1D Mark III was responsive, fired at 10 fps steadily when needed and produced a B- effort when game action was at its most intense. From the backlit position, the EOS-1D Mark III was similarly responsive until the play got going moderately fast, and then the camera's autofocus would briefly lag or stop altogether, along with the frame rate.

    • Similarly, shooting triple jump (a field event) with young athletes coming straight at the camera but not moving too quickly, the EOS-1D Mark III with v1.2.3 did a reasonable tracking job. When older and much faster athletes showed up, the camera once again began to spurt, hiccup and briefly stop firing. Subject speed, the low contrast nature of backlit shooting plus the erratic subject movement that is part of the lead up to the sand pit in triple jump, all this together caused the camera to struggle.

    • While photographing soccer goalie tryouts, the camera repeatedly displayed a slowness bringing the goalie with the darkest uniform into focus, a slowness we'd not experienced before with this camera. Even in the reduced-resolution animated sequence below, you can see that the EOS-1D Mark III's AF doesn't grab hold of the subject until frame 4, and it's not until frame 6 that the goalie is truly in focus.
Molasses: Canon EOS-1D Mark III (firmware v1.2.3) + EF 300mm f/2.8L IS, ISO 200, 1/4000-5000, f/2.8. Click photo to view a larger version of frame 6 (Photo by Rob Galbraith/Little Guy Media)

Note that at this time of year here in Calgary, the sun only hugs the horizon around sunrise and sunset. The backlit situations were all shot between mid morning and late afternoon, a period in which the sun stays fairly high in the sky. This means lens flare wasn't contributing to the poor backlit effort of the EOS-1D Mark III with firmware v1.2.3.

Under the same backlit conditions, the EOS-1D Mark III with v1.1.3 doesn't exhibit this same autofocus performance drop, not to this degree anyway. The EOS-1D Mark III with v1.2.3 loaded, when used to photograph fast moving backlit or shadowy subjects, drops to a C grade or lower at times.

We've photographed soccer, rugby, track and more with both the EOS-1D Mark III and EOS-1Ds Mark III, enough to say that almost everything here applies to both models with the newest firmware installed. Their characteristics and overall performance for shooting fast moving things can fairly be called similar but not exactly the same. Looking over all the sequences we've fired off with both cameras, the EOS-1Ds Mark III with v1.1.2 tends to frontfocus further in front than the EOS-1D Mark III with v1.2.3. This is true whether the EOS-1D Mark III is set to shoot at 10 fps, or to match the EOS-1Ds Mark III at 5 fps, in our testing.

We've not shot enough with a v1.1.2-equipped EOS-1Ds Mark III in backlit or shadowy situations to say if the same autofocus degradation occurs with it as it does with the EOS-1D Mark III. It probably does, but we don't have the pictures to confirm that.

Below are a dozen downloadable sequences from soccer, rugby, basketball plus track and field , shot mostly with the EOS-1D Mark III and firmware v1.2.3. There are comparison sequences taken with the same camera and firmware v1.1.3, plus one batch of photos shot with the EOS-1Ds Mark III (firmware v1.1.2). The brief descriptions below of each ZIP archive will give you an idea of what the download contains:
  • GoodBad In the over 14,000 track and field photos we've shot with the EOS-1D Mark III (firmware v1.2.3), a long lens and a wide aperture, there has been a grand total of one extended sequence with perfect or near-perfect focus throughout. That's the Good. The Bad is that only about three minutes later, during the next 100m heat, the camera coughed up a terrible sequence filled with multiple groupings of blurred frames and an instance of the camera's AF system collapsing temporarily.

    The challenge of selecting sports sequences for inclusion here is to try and give a sense of the autofocus performance we've typically experienced, and neither of the sequences in this download does that. But they do show the range, from the very rare great sequence to the fairly rare AF meltdown. The remainder of the downloads in this section are more indicative of the EOS-1D Mark III (and EOS-1Ds Mark III) AF system's capabilities, most of the time.

  • Carry1, Drill1, Carry2, Drill2 These sequences were photographed first with the EOS-1D Mark III with firmware v1.1.3 installed, and then repeated with v1.2.3.

  • FastTrack Frontfocus, plus a little backfocus, is what you'll find sprinkled throughout the pictures in this download.

  • SlowTrack When the athlete is moving not to fast, focus can be decent. For some reason, this often includes a fast-moving athlete that's slowing down, such as when they've crossed the finish line. An example is in this download.

  • FrontShift Spontaneous and often inexplicable frontfocus are an inescapable part of EOS-1D Mark III AF when photographing things that move.

  • Backlit When viewing the final soccer sequence in this download, note that after Backlit_36 the camera stopped firing momentarily, and when it resumed the autofocus trailed the subject for subsequent frames.

  • Jump1 The long jump photos in this download are typical of the overall sports autofocus performance of the EOS-1D Mark III with firmware v1.2.3.

  • Layup As with v1.1.3, the EOS-1D Mark III with firmware v1.2.3 installed does a reasonable job of tracking basketball players driving to the hoop.

  • TrackField1 A 100m race plus long jump with the EOS-1Ds Mark III (firmware v1.1.2).
Click on a thumbnail to download a ZIP file containing all the pictures in that sequence (processing notes).

GoodBad: Canon EOS-1D Mark III (firmware v1.2.3) + EF 300mm f/2.8L IS, AI Servo AF. Click thumbnail to download full-resolution sequence, 48 Pictures, 70.2MB (Photos by Rob Galbraith and Mike Sturk/Little Guy Media) Carry1: Canon EOS-1D Mark III (firmware v1.1.3) + EF 300mm f/2.8L IS, AI Servo AF, ISO 200, 1/4000-5000, f/2.8. Click thumbnail to download full-resolution sequence, 16 Pictures, 24.1MB (Photos by Rob Galbraith/Little Guy Media) Carry2: Canon EOS-1D Mark III (firmware v1.2.3) + EF 300mm f/2.8L IS, AI Servo AF, ISO 200, 1/5000, f/2.8. Click thumbnail to download full-resolution sequence, 11 Pictures, 16.2MB (Photos by Rob Galbraith/Little Guy Media)
Drill1: Canon EOS-1D Mark III (firmware v1.1.3) + EF 300mm f/2.8L IS, AI Servo AF, ISO 200, 1/3200-5000, f/2.8. Click thumbnail to download full-resolution sequence, 31 Pictures, 52.3MB (Photos by Rob Galbraith/Little Guy Media) Drill2: Canon EOS-1D Mark III (firmware v1.2.3) + EF 300mm f/2.8L IS, AI Servo AF, ISO 200, 1/3200-6400, f/2.8. Click thumbnail to download full-resolution sequence, 59 Pictures, 99.8MB (Photos by Rob Galbraith/Little Guy Media) FastTrack: Canon EOS-1D Mark III (firmware v1.2.3) + EF 300mm f/2.8L IS and EF 500mm f/4L IS, AI Servo AF. Click thumbnail to download full-resolution sequence, 97 Pictures, 179.8MB (Photos by Rob Galbraith/Little Guy Media)
SlowTrack: Canon EOS-1D Mark III (firmware v1.2.3) + EF 300mm f/2.8L IS, AI Servo AF. Click thumbnail to download full-resolution sequence, 61 Pictures, 97.4MB (Photos by Rob Galbraith/Little Guy Media) FrontShift: Canon EOS-1D Mark III (firmware v1.2.3) + EF 70-200mm f/2.8L IS and EF 300mm f/2.8L IS, AI Servo AF. Click thumbnail to download full-resolution sequence, 14 Pictures, 29.9MB (Photos by Rob Galbraith/Little Guy Media) Backlit: Canon EOS-1D Mark III (firmware v1.2.3) + EF 300mm f/2.8L IS, AI Servo AF. Click thumbnail to download full-resolution sequence, 38 Pictures, 66.5MB (Photos by Rob Galbraith/Little Guy Media)
Jump1: Canon EOS-1D Mark III (firmware v1.2.3) + EF 300mm f/2.8L IS, AI Servo AF. Click thumbnail to download full-resolution sequence, 99 Pictures, 153.7MB (Photos by Rob Galbraith and Mike Sturk/Little Guy Media) Layup: Canon EOS-1D Mark III (firmware v1.2.3) + EF 70-200mm f/2.8L IS, AI Servo AF, ISO 6400, 1/1000, f/2.8. Click thumbnail to download full-resolution sequence, 13 Pictures, 81.5MB (Photos by Rob Galbraith/Little Guy Media) TrackField1: Canon EOS-1Ds Mark III (firmware v1.1.2) + EF 300mm f/2.8L IS and EF 500mm f/4L IS, AI Servo AF. Click thumbnail to download full-resolution sequence, 32 Pictures, 103.4MB (Photos by Rob Galbraith/Little Guy Media)

Q. Can you say a bit more about what you've shot and how you've shot it?

Earlier, we listed the cameras and lenses tested. We used that collection of gear to rattle off about 27,000 frames on the EOS-1D Mark III bodies and about 4500 frames on the EOS-1Ds Mark III. Almost all shooting was done with the cameras on CR2, though we did set the EOS-1Ds Mark III to JPEG when more frames in a sequence were needed (the EOS-1Ds Mark III tops out at 13-14 RAW photos in succession).

We shot so much primarily because the testing became about more than just trying to figure out whether the new firmware made the AF system in the EOS-1D Mark III and EOS-1Ds Mark III all better. It also became about figuring out how it was different, once it was apparent that Canon had done something other than just give the AF a minor tweak, and about trying to find a way to coax better autofocus performance from these cameras when shooting outdoor sports.

canon_af_expansion_anim.gif
Expanding AF: An animation showing single and Assist AF options in Mark III cameras (Graphics courtesy Canon)
Most photography was done with the cameras configured to use either a manually selected single AF point (usually the centre point) by itself, or in conjuction with two or six surrounding points (set with C.Fn III-8 [AF Expansion With Selected Point]).

Canon calls these additional points Assist AF points, and the naming is spot-on. That's because the Assist AF points seem to be invoked only when the selected AF point is drifting well away from the subject; otherwise, single subject autofocus performance with the Assist AF Points enabled or disabled is essentially identical (in almost all cases we couldn't tell which sequences were shot with Assist AF points active without looking at the sequence's metadata).

Okay, that's not the whole story: the Assist AF points will also kick in when the selected AF point is giving confusing readings to the system. And while you might think this is happening frequently in outdoor sports photography, given the EOS-1D Mark III and EOS-1Ds Mark III's middling tracking performance, the cameras don't seem to think so; the only instances in which we could see or feel that the Assist AF points were being called on was when we deliberately moved the selected AF point off the subject (this is when the Assist AF points can and do help keep the AF from jumping to the background), when we aimed the selected AF point through a narrow gap between two closer objects to a subject further away or when there were multiple subjects at different distances all crowded around the selected AF point.

Overall, though, basic single subject tracking performance of the EOS-1D Mark III with v1.2.3 and EOS-1Ds Mark III with v1.1.2 doesn't seem to change whether Assist AF points are enabled or not. Stationary subject focus stability also appears to be the same, with or without Assist AF points turned on. The only situation in which Assist AF points can most definitely cause trouble is when shooting multiple subjects clustered together. Then, it's clear that one manually-selected AF point is superior.

We shot some action with all 45 AF points active, but this configuration isn't really intended for shooting sports with rapid and ever changing camera-to-subject distance. For shooting things on a consistent path and not moving too too fast, it seems to work alright - the frame rate stays consistent and the AF system continues to adjust the lens' focus distance throughout extended sequences.

But with 45-point AF (Canon calls it Automatic AF Point Selection) turned on briefly during soccer, rugby and the faster track events such as the 100m, the EOS-1D Mark III locked up several times, each time we tried it. We mention this not really to criticize the camera. It's to explain why this time, as in months previous, most shooting and testing has been done with fewer points active. Trying to use all the points to shoot fast sports, with the EOS-1D Mark III, EOS-1Ds Mark III or any other camera we've used in years past, doesn't work out, because it's usually the wrong AF configuration for the job. The 45-point configuration is for other shooting situations.

We also experimented with Custom Settings C.Fn III-4 [AI Servo AF Tracking Method] and C.Fn III-3 [AI Servo Tracking Sensitivity), to see if fiddling with the former would bump up the number of in focus frames when Assist AF points were enabled, and the latter to see if it would help to reduce the slight AF system lag introduced by the new firmware as well as reduce the cameras' tendency to shift focus when a focus shift trigger - such as a swinging arm - would come through the frame. The answer was no in all cases.

Finally, we spent far less time shooting the runner and soccer player drill tests that have figured prominently in several previous article installments. It's hard to escape doing these kinds of tests, because they make it practical to try out different camera settings as the athlete does the same motion repeatedly, and because they produce photos that can be bundled into clean, easy-to-interpret comparisons between one camera and another (they have served as the best way we could come up with to show the autofocus differences between the EOS-1D Mark III and EOS-1D Mark II N, for example).

They also greatly compress the time it takes to test autofocus, which is the primary reason we started doing these tests in the middle of last year: we needed a quicker method for figuring out if a different setting or body or firmware update was going to appreciably improve EOS-1D Mark III autofocus, quicker than getting out and shooting a series of games, over and over.

But these tests aren't the be-all, end-all, and based on what we've seen in a few Internet forums and on other websites, they're also open to significant over-interpretation and misinterpretation. We've included some example sequences from runner and soccer player tests in this article, because of their effectiveness as demonstrators of autofocus characteristics, but the majority of the pictures shown in this installment, and the pictures on which we've based most assessments, were shot the old-fashioned way: at actual games and events.

Q. You haven't said anything about the EOS-1D Mark II N. Is it still a better autofocus camera than the EOS-1D Mark III?

Yes.

Because the EOS-1D Mark II N has been discontinued for over a year, because the majority of questions, comments and pleas for help in recent months have been from Canon EOS-1D Mark III and EOS-1Ds Mark III owners rather than photographers considering replacing a Mark II camera, and because trying to simultaneously cover both Mark III cameras with both old and new firmware had the potential to be confusing enough, we opted to steer clear of most comparisons to the older camera in this update. We'll break with that plan briefly now.

Compared to the Mark II lineup, the Mark III models (with the new firmware) are slightly better at holding focus on static subjects, plus they may still have the advantage in raw autofocus speed (though the new firmware appears to have achieved more stable focus on stationary subjects in part by sacrificing a small measure of that speediness). But the EOS-1D Mark II N, the EOS-1D Mark II and EOS-1Ds Mark II do an overall better job of autofocusing in a broader range of shooting situations than the EOS-1D Mark III and EOS-1Ds Mark III, and at times the performance difference isn't subtle.

For instance, the older cameras excel at shooting outdoor activities like track, rugby, soccer and football. When I first got an EOS-1D Mark II back in the summer of 2004, I immediately became a better sports photographer. While I was still the same guy looking through the viewfinder, the camera's ability to get so many of the key frames in focus, and get so many frames in extended sequences in focus too, meant I was immediately producing better moments, and more of them. That continued with the EOS-1D Mark II N.

Site co-editor Mike Sturk has had the same experience, and it's reflected in what he still shoots with today: his primary sports body is an EOS-1D Mark II and, for assignments requiring more pixels, he opted to steer clear of the Mark III series and instead purchased a used EOS-1Ds Mark II earlier this year.

Ouch: Canon EOS-1D Mark II (firmware v1.2.6) + EF 300mm f/2.8L w/Extender EF 1.4x, ISO 400, 1/2500, f/4.5. Click photo to enlarge (Photo by Mike Sturk/Little Guy Media)

By comparison, the EOS-1D Mark III, even with all of Canon's focus fixes to date, has consistently delivered outstanding sports action autofocus really only with volleyball. It has elevated the level of peak moments I'm able to capture of that sport. For the outdoor sports listed in the previous paragraph, the EOS-1D Mark III does the opposite: its autofocus is good enough to enable some good photo to be gotten, but its tendency to overestimate forward motion too often means the camera's autofocus is at times more hindrance than help to the photographer using it.

Here are two more ways to think about the differences between the current and previous generations of 1-series Canons.
  • Making the grade The Mark III cameras earn about an A now in slow moving or static subject photography, and they might still be at that level for volleyball action. For most everything else, results hover in the B- to C- range: certain indoor shooting situations are a B, while it's possible to drop down to the D level when the subject is moving fast towards the camera and is backlit or otherwise in shadow. But, a whole lot of what we've shot lands somewhere in-between. In other words, the AF results are variable, and there are more than a few environments in which the results will be on the wrong side of the line of acceptability. All of this applies to the EOS-1D Mark III with firmware v1.2.3, and most if not all applies also to the EOS-1Ds Mark III with firmware 1.1.2.

    The Mark II cameras can autofocus most things well, some things exceptionally so, while there's really nothing we've tried to shoot with an EOS-1D Mark II, EOS-1D Mark II N or EOS-1Ds Mark II that has resulted in serious autofocus screwups. These models get As and Bs all day long, and only occasionally lower than that in our experience.

  • Testing, schmesting If you don't care about tests or grades but you just want to know what will help you make crisp pictures of things that move, here's another way to think about the difference. Because we set out to not look at the EOS-1D Mark II N in this article, we also planned from the outset to spend less time using the older camera. After a few weeks of shooting with the Mark III cameras and the new firmware, poring over the photos, it was becoming apparent that I was losing my frame of reference, losing my sense of what good action autofocus is. This prompted two changes: additional side-by-side shooting involving the EOS-1D Mark II N and, more importantly, reviewing a portion of the tens of thousands of sports photos we've shot with Mark II cameras in the last four years.

    This helped set things straight. If you put aside all the this-camera-does-this and that-camera-does-that, if you forget about trying to microanalyse autofocus behaviour and just look over a few dozen assignments shot for no reason other than to try and make good pictures, the Mark IIs show themselves to be strong autofocus cameras and powerful sports picture making tools, particularly outdoors. Conversely, the Mark III cameras get the focus wrong too much of the time in the same situations in which Mark II cameras excel. The EOS-1D Mark II, EOS-1D Mark II N and EOS-1Ds Mark II don't autofocus perfectly, but on balance they got the job done well.
We've published numerous EOS-1D Mark II N sequences in previous installments of this article; you might download a few if you're unfamiliar with its sport autofocus capabilities. Below are two more sequences from Canon's previous generation, one shot with the EOS-1D Mark II N, the other with the EOS-1D Mark II (several of the same events and athletes can be found in sequences from other cameras in this article update).

Click on a thumbnail to download a ZIP file containing all the pictures in that sequence (processing notes).

TrackField2: Canon EOS-1D Mark II N (firmware v1.1.2) + EF 300mm f/2.8L IS, AI Servo AF. Click thumbnail to download full-resolution sequence, 82 Pictures, 97.7MB (Photos by Rob Galbraith/Little Guy Media) Track: Canon EOS-1D Mark II (firmware v1.2.6) + EF 400mm f/5.6L, AI Servo AF, ISO 400, 1/4000-5000, f/5.6. Click thumbnail to download full-resolution sequence, 16 Pictures, 24.6MB (Photos by Mike Sturk/Little Guy Media)

Q. What do you think is causing the so-so sports tracking performance of the EOS-1D Mark III and EOS-1Ds Mark III?

We don't have any idea: it could be something that's not quite right in the autofocus hardware, firmware, or a combination of the two. While testing and shooting with the Mark III cameras, though, we have become familiar with some of the things that trigger the onset of blurry frames. With the EOS-1D Mark III (firmware v1.2.3) and EOS-1Ds Mark III (v1.1.2), they include:
  • Abrupt changes in subject contrast or brightness If the active AF point is aimed at, for example, a moving athlete's torso, and the athlete's motion is such that their torso comes in and out of the sunlight repeatedly (as is almost always the case in midday sunlight), the EOS-1D Mark III and EOS-1Ds Mark III often misinterpret this as a speed change and shift the lens to the wrong focus distance. The Mark II cameras don't don't do this, at least not nearly to the same extent.

  • Abrupt changes in subject colour Similar to the above paragraph, if, while tracking a moving subject, the active AF point drifts from dark-coloured pants to a light-coloured shirt, the Mark III cameras can misinterpret this and produce two or more blurry frames before regaining correct focus. And again, the Mark II cameras don't don't do this, not to the same degree.

  • eos_1d_mark_iii_af_size.jpg
    Interference: The EOS-1D Mark III (firmware v1.2.3) appears to have picked up focus on the soccer ball, even though its arc was such that it didn't come into the active AF point's rectangle (Photo by Rob Galbraith/Little Guy Media)
    Interference from non-interfering objects In sports, swinging arms, flying balls and other objects come close to the active AF point all the time. The AF system in the Mark III cameras, regardless of how C.Fn III-2 [AI Servo Tracking Sensitivity] is set, is too often thrown off balance by this. Even with only the centre AF point active and Assist AF points disabled, a hand or soccer ball simply passing near the AF point is sometimes enough to introduce a frontfocus shift.

    The incoming object doesn't have to actually pass through the AF point's area, and it doesn't have to hang around long, for the autofocus to jump frontwards. Sometimes, the change in focus distance is such that it's clear the camera was trying to track the closer object, even though it never entered the AF point's red frame, other times it looks like the closer object's brief presence simply confused the AF system. In either instance, the subject is no longer in focus, and it can take two or more frames in a sequence for the AF system to regain its footing. The Mark II cameras don't do this at all, and almost entirely ignore things like swinging arms passing through the AF frame too.

    Part of the explanation for the Mark III cameras' behaviour is the fact that the centre AF point's sensor is considerably larger on all four sides than is indicated by its rectangle on the viewfinder screen, and that at times the AF system is switching focus to or being tripped up by objects that are briefly within the actual detection area, even if no part of the object comes within the AF point's red frame. We've only noticed this behaviour since loading the newest firmware, but it may have been there in earlier firmware versions too.

    Points other than the centre one may also act larger than indicated in the viewfinder, but we've only tested the centre point thoroughly enough to say that its total detection area is definitely larger than indicated - a lot larger. The fact that the AF point is actually bigger than shown, however, doesn't on its own explain why the Mark III cameras are so quick to stumble when an object passes through.

  • Same AF point, different distances As mentioned, the frame rate will slow, as will the speed of the AF, when the AF point is targeting a backlit or shadowy subject. For the AF system to feel like it has stopped, and for the camera to briefly not fire at all, an additional element is usually or perhaps always required: the active AF point has to be overlapping both the subject and background. If you combine a low-contrast subject with an AF point that isn't fully on its mark, the EOS-1D Mark III (and probably the EOS-1Ds Mark III also) can experience a momentary, jarring brain freeze. Whatever logic is built into the AF algorithm to handle this situation seems to have trouble when the subject is backlit or in shadow, and the result is no frames, or blurry ones.

  • No contrast When the AF point is targeting an area on a fast moving subject that is fairly dark and featureless, there comes a point where the AF system tosses up a hail mary pass and hopes for the best. Unlike the example of the previous paragraph, where the camera stops firing, in this instance the EOS-1D Mark III (and probably EOS-1Ds Mark III too) keeps on shooting at or near full speed, but focus is shifted to the front about as far as the lens can muster between frames. The EOS-1D Mark II N we've seen shift focus frontwards in the same scenario, but only slightly.
This is a partial list of things that can trigger an AF system error in the Mark III cameras, but as to why this is so, that's a mystery that can only be solved by Canon engineers, given that the Mark II cameras are either less affected by or ignore completely the same triggers. You'll see some of the triggers at work in the various downloadable sequences found throughout this article update. The sequences below do as well:
  • Trigger1 As the injured athlete's body shifts slightly, the AF system responds by pushing the focus frontwards. We've said it before, and we'll say it again: it doesn't take much provocation for the EOS-1D Mark III to shift focus in front of the subject being tracked.

  • Trigger2 With only the centre AF point active and Assist AF points disabled, it's hard to know exactly what the camera is thinking about as the AF point is moved gradually from one hurdle to another. The only thing that's certain is that the AF point's active area is well outside the area marked by the viewfinder's red rectangle.

  • Trigger3 Mix together a speedy subject, backlighting, a dark target and a swinging hand and the AF system goes off the rails for several frames.
Click on a thumbnail to download a ZIP file containing all the pictures in that sequence (processing notes).

Trigger1: Canon EOS-1D Mark III (firmware v1.2.3) + EF 500mm f/4L IS, AI Servo AF, ISO 320, 1/4000-5000, f/4. Click thumbnail to download full-resolution sequence, 4 Pictures, 7.1MB (Photos by Rob Galbraith/Little Guy Media) Trigger2: Canon EOS-1D Mark III (firmware v1.2.3) + EF 300mm f/2.8L IS, AI Servo AF, ISO 200, 1/8000, f/2.8. Click thumbnail to download full-resolution sequence, 18 Pictures, 26.1MB (Photos by Rob Galbraith/Little Guy Media) Trigger3: Canon EOS-1D Mark III (firmware v1.2.3) + EF 300mm f/2.8L IS, AI Servo AF, ISO 200, 1/2500-3200, f/2.8. Click thumbnail to download full-resolution sequence, 6 Pictures, 7.5MB (Photos by Rob Galbraith/Little Guy Media)

(This is part 2 of the August 1, 2008 article update. It continues in part 3 on the next page)
Next Page: August 1, 2008 update (Part 3)
Go to Page: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 
Go to advertiser website.
2000-2013 Little Guy Media. Not to be reproduced without written permission.