Note: We added a new section to this article on May 29, 2009. This and other updates are accessible from the popup menu at right.
In our April 22, 2007 first look at the Canon EOS-1D Mark III
, we described the camera's autofocus as being blazingly fast at acquiring initial focus. So fast that covering women's volleyball over several days was pure joy, thanks to an incredibly responsive autofocus system in the preproduction body on loan from Canon. We also found its autofocus to be quicker off the line in dim light than any camera we'd used before.
High Jump: Canon EOS-1D Mark III production model (firmware v1.0.8) + EF 500mm f/4L IS, ISO 640, 1/3200, f/4 (Photo by Rob Galbraith/Little Guy Media)
But that wasn't the end of the autofocus story. We went on to say:
After that, [autofocus in] our preproduction EOS-1D Mark III is a mess. It can't hold focus on static subjects very well and it can't track moving subjects very well. While Canon didn't provide any details about the autofocus limitations we would encounter in the preproduction body, we hope this is what they were referring to and this is what engineers have been solving since.
Do the autofocus weaknesses (and strengths) of the preproduction body carry over to shipping EOS-1D Mark IIIs
? Does the new model match or exceed the autofocus performance of recent 1-series Canons like the EOS-1D Mark II N
? Over the last three weeks, we've been shooting extensively with two full production bodies in an attempt to answer these very questions. Here, in FAQ form, is what we've figured out about EOS-1D Mark III autofocus:
Q. Is the autofocus better in the production EOS-1D Mark III than it was in the preproduction unit?
No. The autofocus problems of the preproduction unit are mirrored in production EOS-1D Mark IIIs.
Q. What are the camera's autofocus problems?
It has four, all of which are related to the camera's autofocus performance:
Under certain conditions, the EOS-1D Mark III has difficulty acquiring focus initially. In a multi-frame burst, the camera will sometimes shoot three to five frames before a moving subject comes into focus, and occasionally a moving subject will not actually snap into focus before the burst is completed.
Under certain conditions, the camera is unable to properly track a moving subject. We've shot numerous sequences of 20+ frames where no more than five or six frames are in focus, even when the AF point has been on the subject throughout.
Focus can shift slightly but constantly at times when the subject isn't moving. Under certain conditions, the subject may not actually come into focus through a sequence of frames, even though the point of focus can be seen to be shifting throughout the sequence. This is true whether the camera is set to AI Servo and focus is active throughout the sequence, or when it's set to One Shot and focus is activated between each frame.
When tracking a subject that's moving somewhat erratically, the camera is far too quick to shift focus elsewhere - to the background or, with a field sport like soccer, to a player passing through in the foreground. With the first three problems, autofocus settings changes don't make things better or worse. With this problem, Custom Function III-2, AI Servo Tracking Sensitivity, does have an impact. But regardless of how this Custom Function is set, it's not possible to make the camera's tracking sensitivity be right. There's more on this ahead in the article.
Q. You say the camera's autofocus difficulties occur or are worse "under certain conditions" - what does that mean?
It means that when the light is especially bright and the temperature is warm, the camera's autofocus performance drops like a stone. Yes, you read that correctly. On sunny, warm, beautiful days - the sort of conditions in which autofocus usually thrives - the EOS-1D Mark III's ability to make in-focus pictures of still or moving subjects is greatly reduced.
The very first thing we shot with a production EOS-1D Mark III was a rugby match on an overcast, soon-to-be-rainy evening. Browsing through the pictures, we could see quite a few that were properly focused, though perhaps not as many as the EOS-1D Mark II N would have captured (owing mostly to problem #4 - again, more on that ahead).
Nose Job: Canon EOS-1D Mark III production model (firmware v1.0.8) + EF 500mm f/4L IS, ISO 1600, 1/2000, f/4 (Photo by Rob Galbraith/Little Guy Media)
A few days later we were back shooting at the same club from about the same field position, but now it was full sun, little wind and about 25° Celsius (77° Fahrenheit). The result? So many pictures out of focus: first frames, frames mid-sequence, frames of players on the bench, you name it, it was a whole lot of blurriness.
Fast forward three weeks and we've now shot multiple games of soccer and rugby plus track, indoor volleyball, moving cars, grip and grins, an executive portrait and personal stuff. It's added up to over 15,000 frames across two production EOS-1D Mark IIIs, and the most obvious autofocus theme to emerge after poring over all these pictures is that when the lighting is full sunlight and it's a nice day outside, the autofocus is unusably poor. The camera will still produce in-focus pictures (several are included in this article in fact), but the ratio of crisp to blurry frames is well below an acceptable level. Generally speaking, the camera tends to backfocus when it's focusing incorrectly.
Here's another example: the last thing we shot before writing this article was a series of test photos of an athlete running at different speeds on a track, plus practicing soccer drills. It was approaching 28°C (about 82°F) and the sun was beating down on a mostly windless early afternoon. Whether shooting her frontlit or backlit, the results were the same: in most sequences, well over half of the frames were out of focus; the in-focus rate was particularly low when shooting her head-on (you'll see an example sequence later in the article).
As with other tests, we repeated several sequences with the athlete wearing a different shirt, to see if the camera's autofocus sensor was somehow reacting badly to a particular colour or pattern. From these tests and real-world shooting, we can see that whether the clothing is black, grey, white, blue, red, yellow, logo, no logo, it doesn't seem to matter: the autofocus problems are the same.
By comparison, we shot head-on a different track athlete jumping hurdles on a heavily overcast and cool late afternoon. We rattled off one extended sequence where every frame is either in focus or just the tiniest bit out, plus numerous other sequences where the vast majority of frames were perfectly focused. An example sequence is below.
Snappy: Canon EOS-1D Mark III production model (firmware v1.0.8) + EF 300mm f/2.8L IS, ISO 800, 1/3200, f/2.8. Press the right arrow to advance through the sequence. A green bar in the corner of the photo means it's in focus, orange means it's very slightly out of focus and red means it's significantly out of focus. The active AF point is indicated by a red rectangle. Because the photos are displayed at less than full resolution you may not be able to see much difference between the sharp and soft frames above. Let the colour coding be your guide to image focus. (Photos by Rob Galbraith/Little Guy Media)
Focus on static subjects is equally affected. Even in dim, cool conditions, the EOS-1D Mark III tends to shift focus slightly too much when pointed at a person or object that's motionless. But the problem isn't severe, and we've not shot with
any camera that can hold focus with a long lens on a static subject as well as we think a camera ought to. Given that, the EOS-1D Mark III may not be particularly worse than any other when the light isn't bright and the temperature isn't too warm. But when the sun is out and the mercury is rising, it can be a challenge to make an in-focus picture of somebody or something that's not moving. An example sequence is below.
Shifty: Canon EOS-1D Mark III production model (firmware v1.0.8) + EF 300mm f/2.8L IS, ISO 200, 1/4000, f/2.8. Press the right arrow to advance through the sequence. A green bar in the corner of the photo means it's in focus, orange means it's very slightly out of focus and red means it's significantly out of focus. The active AF point is indicated by a red rectangle. Because the photos are displayed at less than full resolution you may not be able to see much difference between the sharp and soft frames above. Let the colour coding be your guide to image focus. (Photos by Rob Galbraith/Little Guy Media)
We don't know whether it's the light, the heat or both that's causing the problem we've encountered, but we're leaning towards both being the culprits somehow (and we're quite sure that heat waves rising from the ground aren't a factor in what we've shot to date). We've also shot action in full sunlight on cooler days in the 14°C to 18°C (about 57°F to 64°F) range, and the autofocus, while still not great, hasn't been the wipeout we describe above.
Q. You described the EOS-1D Mark III as being incredibly fast at acquiring focus initially, but you also describe the camera's ability to focus initially as being a problem. Can you explain the contradiction?
In fact, both statements are true.
What the camera's autofocus system does well is rapidly acquire initial focus in heavy overcast daylight, indoor light or stadium light. For example, the speed of initial focus with the EF 70-200mm f/2.8L IS is noticeably quicker than with any other Canon we've ever used. And, under these conditions, the initial focus is also spot-on. Despite the EOS-1D Mark III's various and significant autofocus problems, we'd use it again in a heartbeat to shoot pretty much anything indoors that required ultra-speedy, accurate focus of the first frame.
In bright light and warm temperatures, speed of initial focus remains incredibly fast. But focus accuracy plummets, and is therefore a problem.
Q. Doesn't the EOS-1D Mark III have a feature that allows you to adjust the autofocus so that the camera doesn't shoot out of focus pictures?
Yes it does. It's Custom Function III-7, AF Microadjustment, and its purpose is to enable the end user to compensate for focus calibration errors in the body or combination of the body and attached lens.
With the two production EOS-1D Mark IIIs we've used prior to publishing this article, plus the preproduction body plus a third production body we've just started shooting with, we've checked a minimum of three lenses with each, and in each case focus calibration has been a-ok and no AF Microadjustment has been needed. The three lenses - an EF 70-200mm f/2.8L IS, EF 300mm f/2.8L IS and EF 500mm f/4L IS - are the ones used for almost all of the pictures we've looked at in making judgements about the EOS-1D Mark III's autofocus.
AF Microadjustment can compensate for focus calibration errors - that is, errors that result in the camera focusing behind or in front of the subject with all lenses or a given lens - but it isn't meant to correct for the autofocus problems we've encountered in the EOS-1D Mark III.
Q. How do you know the lenses aren't the problem?
In addition to making sure that no AF Microadjustment tweaks were required with the trio of Canon telephoto lenses we use the most, we also performed another simple test of sorts: we used the same lenses on an EOS-1D Mark II N at several of the same locations and events as the EOS-1D Mark III. They worked just as well as they always do on that camera.
Q. How do you know the autofocus settings dialed into the camera aren't the problem?
First, it's important to understand that most of the EOS-1D Mark III's autofocus options are personal preference settings. They provide ways to configure the camera to your liking rather than fundamentally change the way the system determines subject distance or commands the lens to move to a certain point of focus. In other words, serious autofocus problems aren't likely to be fixed by adjusting a few Custom Functions.
That said, our desire to properly understand EOS-1D Mark III autofocus, combined with a nagging feeling that there must be some way to get the darned thing to focus well, meant we threw logic out the window and began working through the various settings that might possibly help, maybe. To that end we tried enabling, disabling or otherwise fine-tuning these options:
- C. FnIII-2, AI Servo Tracking Sensitivity (normal, Slow and in-between)
- C. FnIII-3, AI Servo 1st/2nd Image Priority (all options)
- C. FnIII-4, AI Servo Tracking Method
- C. FnIII-5, Lens Drive When AF Impossible
- C. FnIII-8, AF Expansion with Selected Point (Disable, Enable Surrounding Assist Points)
- C. FnIII-16, Continuous Shooting Speed (10 fps, 8 fps, 3fps)
We shot most things using the centre AF point only, though at times we switched to a couple of the outer AF points or all 45 AF points. The result? The only setting that had any real impact on the basic character of the autofocus was C. FnIII-2, AI Servo Tracking Sensitivity. We'll talk about it next.
(For an explanation of what these Custom Functions do, you might want to consult the camera's user guide, which is available as a PDF via the
Product / Software Manuals
link on the EOS-1D Mark III product page
Q. How does the EOS-1D Mark III's autofocus compare to the EOS-1D Mark II N?
The EOS-1D Mark II N has the finest autofocus system we've ever used. It's not perfect: it can be slow to focus in dim light and its predictive tracking algorithm is easily tripped up by a subject that's continually speeding up and slowing down. All in all, however, we don't know of another camera that can match its level of autofocus performance. Well, except for the EOS-1D Mark II and EOS-1Ds Mark II, which feature the same system (with minor algorithm changes in the latter model to account for its slower maximum frame rate).
So, the bar is set pretty high for the EOS-1D Mark III, and it was Canon's stated goal that the new model would eclipse the EOS-1D Mark II N's autofocus capabilities. As you've undoubtedly surmised, it's our view that the EOS-1D Mark III's autofocus is not so much inferior as it is broken. Therefore, the EOS-1D Mark II N is the better autofocus camera in almost all respects.
The example sequences below demonstrate this: we shot bursts of this athlete starting at several different points in her run and at three different speeds. In all cases, the EOS-1D Mark II N performed as it always does. Which means it has a bit of trouble predicting focus when the runner is accelerating, but after that it gets most frames in focus. By comparison, the EOS-1D Mark III under warm, bright conditions struggles to get more than a handful of frames properly focused.
As we noted earlier, the EOS-1D Mark III isn't without its autofocus charms:
The EOS-1D Mark III's ability to focus quickly and accurately in low light, prior to squeezing off the first frame, is unparalleled. Whether the focus will remain true as the camera tracks movement - like someone walking down the aisle - is another matter, and we honestly haven't done enough low light tracking with a production camera to give an opinion about how it would do. The preproduction model's accuracy when tracking suits walking around in ballroom lighting, however, was middling at best.
- If you shoot amateur theatre or dance, where light isn't low it's non-existent, the EOS-1D Mark III's autofocus speed in this environment coupled with its phenomenal high-ISO image quality probably makes it the camera to choose. Even with its autofocus warts.
Overall, though, the EOS-1D Mark II N is going to be the more reliable autofocus companion.
Most of what we've described as EOS-1D Mark III autofocus problems relate to what we presume are unintentional design flaws in the camera. But one - AI Servo Tracking Sensitivity, or rather the delay timings that lurk within - seems more like an intentional change. If you're unsure of what this setting does, then rename it in your mind to New Subject Acquisition Delay. It dictates how long the camera will keep tracking the subject it has been tracking, before shifting to a new point of focus, when the manually-selected active AF point drifts off the subject to the background or something passes through the active AF point in the foreground.
With the EOS-1D Mark III's AI Servo Tracking Sensitivity set to the middle (normal) position, and the thing being photographed is a sport like soccer, there is effectively no delay at all before the focus shifts to the background when the AF point moves off the action or when another player runs through. It makes the autofocus system feel overly jumpy, and it makes the job of getting in-focus pictures harder.
C. FnIII-2: AI Servo Tracking Sensitivity
Shifting AI Servo Tracking Sensitivity one notch towards Slow makes the camera hang onto the focus better when the AF point moves off the subject to the background temporarily, or when a ref is in the way briefly. But, it also has the effect of making the camera feel slightly sluggish at acquiring focus when holding the rear AF-ON button down constantly and moving from subject to subject. The sluggish feeling increases further when the camera is set to Slow.
(All this applies to front-button focusing too; if you keep the autofocus engaged by half-pressing the shutter button while jumping from subject to subject, you'll experience the same sluggishness.)
By comparison, the EOS-1D Mark II N when set to Standard tracking sensitivity gives the best of both worlds: there is often the right amount of delay when you want the camera to keep on tracking the subject, a delay that allows you just enough time to get the AF point back on the subject before the focus shifts to the background or foreground. But when deliberately moving from one subject to another with the autofocus continually engaged, the point of focus shifts fairly rapidly, as you would want.
In other words, it feels like the EOS-1D Mark II N has an intelligent, variable delay length built into the AI Servo Tracking Sensitivity algorithm, with the Standard setting being optimum for a whole lot of assignments. In truth there probably isn't a variable delay, or at least the variability probably isn't intentional. It's more likely the by-product of the time it takes for the slower (but not slow!) autofocus system in the EOS-1D Mark II N to perform certain calculations. But it does accurately sum up how the EOS-1D Mark II N
feels and is part of what makes for a lot of in-focus frames with this camera when shooting sports like soccer, football, basketball, rubgy and hockey.
By comparison, to make the EOS-1D Mark III usable, or at least perform as well as it can, it's critical to turn AI Servo Tracking Sensitivity down one notch, but then at times this means you may have to pump the autofocus button when deliberately switching from one subject to another. We think it's important for Canon, in addition to sorting out the other problems we've described, to make the timings and feel of the EOS-1D Mark III's AI Servo Tracking Sensitivity match that of the EOS-1D Mark II N.
Q. Canon had told you they were still working on autofocus at the time you got the preproduction camera. What did they change prior to the production EOS-1D Mark III's release?
Not a lot. On a recent trip to Canon in Japan
, we learned that one change was implemented, and that this change wasn't likely to impact autofocus behaviour much, if at all, in shipping units. We're not able to say what the change was, because it was to a capability of the autofocus system that Canon considers to be a trade secret.
Q. Do you think all these autofocus problems are because you got a lemon camera?
Former Nikon USA executive Richard LoPinto had a great response for questions like this: "It's possible, but not probable."
We've shot the heck out of two production bodies (with serial numbers about 5000 increments apart, both with firmware v1.0.8) and one preproduction body (with preproduction firmware, but whose autofocus system, it turns out, is 99% the same as shipping cameras) and they all act the same way. In addition, though we're just beginning to photograph with a fourth EOS-1D Mark III (third production body, also firmware v1.0.8), so far its autofocus is giving off the same vibe as the others.
At some point, we think the reasonable conclusion to draw is that these autofocus problems are inherent to the camera model. If we weren't confident of that we wouldn't have written this article.
Q. I don't shoot sports. Will the camera's autofocus limitations affect me?
The answer is a non-answer: it depends on what you shoot. For example, from the autofocus system's perspective the tracking of a large bird in flight can't be all that different from tracking a sprinter moving at full speed. Given that the EOS-1D Mark III on a warm, bright day doesn't track a sprinter very well, it's very likely that it won't track a bird any better. But if you don't shoot much with long lenses, you might find the EOS-1D Mark III's autofocus to be just fine.
Of the three telephoto lenses we've used most, the combo of the EOS-1D Mark III and EF 300mm f/2.8L IS or EF 500mm f/4L IS (plus bright light and warm temperatures) has shown the worst results overall. By comparison, these two lenses with the EOS-1D Mark II N are awesome. The 70-200mm f/2.8L IS on the EOS-1D Mark III has also not been up to par overall, but for shooting subjects that don't move erratically, the autofocus performance of the EOS-1D Mark III hasn't degraded quite as drastically when this zoom is attached, the light is bright and the temperatures warm.
A shorter lens we've used a fair bit on the two production bodies is an EF 24-105mm f/4L IS, and we've been satisfied for the most part with the autofocus when this lens is on the camera. But we've not used it at all outdoors on a warm, sunny day yet.
If you shoot things that don't give the autofocus system much of a workout, like group portraits with about a 50mm or wider lens, chances are you'll be okay. We used the preproduction EOS-1D Mark III with both an EF 50mm f/1.4 and EF 35mm f/2.0 and focus was fine.
Q. Have you talked to Canon about this?
We've been in touch with both Canon USA and Canon Canada to pass on a shorter version of what you're reading here. We've also shipped to Canon a large batch of EOS-1D Mark III and EOS-1D Mark II N files that demonstrate everything described in this article.
Will Canon fix the problems? Can Canon fix the problems?
We can't speak for Canon and don't want to speculate about whether or if they can fix what we think isn't right about EOS-1D Mark III autofocus. Here's what we can say: The EOS-1D Mark III's designers obviously developed it with the sincere belief that its autofocus system was as good or better than anything they've done before, and the EOS-1D Mark III has only been in the hands of paying customers for a few weeks as this is being written. Therefore, if there is to be a solution to these autofocus problems, the only thing that's certain is that these are the early days of that process. Whether the end result is an EOS-1D Mark III hardware revamp, firmware update or no change at all, we don't know.
If you own the camera and its autofocus isn't meeting your expectations, it's important to let your Canon rep know. We recommend giving them lots of specific detail about what's not working. If you've just botched an assignment with the camera you might want to cool down first so that you can deliver the information in as friendly and constructive a manner as possible. You might also offer example files.
Q. Should I return my EOS-1D Mark III or keep it while I wait for Canon to fix the autofocus?
We've received a few emails from EOS-1D Mark III owners claiming autofocus problems. Not a lot - at a glance it looks like fewer than 20 photographers so far - but of those, most asked some variation of this US$4495 question.
While we can't answer this question for you, we can commiserate. After returning one of the bodies to Canon, website co-editor Mike Sturk sent me an email that summed up perfectly what it can be like to have and then part with an EOS-1D Mark III: "I'm glad it's gone back [to Canon,] because the files were so great I found it hard NOT to use the camera."
He had just completed several days shooting show jumping with the EOS-1D Mark III, an experience that was equal parts joy - thanks to the image quality - and frustration, because the EOS-1D Mark III's autofocus was too hit-and-miss when compared to his trusty EOS-1D Mark II.
Q. I'd like to test my own EOS-1D Mark III. Any suggestions?
Make it a real test, by really photographing the things you would actually photograph that require good autofocus. Otherwise, you may find that you've wasted your time performing a test that doesn't ultimately tell you whether the camera's autofocus is up to the challenge of what you normally shoot.
Face Massage: Canon EOS-1D Mark III production model (firmware v1.0.8) + EF 300mm f/2.8L IS, ISO 200, 1/5000, f/2.8 (Photo by Rob Galbraith/Little Guy Media)
If you're photographing moving subjects, get your shutter speed up and try to shoot at or near the lens' maximum aperture. For all our telephoto testing we tried to keep the shutter speed above 1/2000, even in fading light, and for the bright light assignments the camera was almost always in the 1/4000 - 1/6400 range. High shutter speeds help to eliminate camera shake and subject movement as variables. Wide apertures mean that depth of field is less likely to mask point of focus errors.
Shoot on RAW - even if you usually don't - and process the CR2s in Digital Photo Professional (DPP) 3.x with Sharpening set to 3, even if you don't typically use Canon's software. Converting the files to quality level 10 JPEGs is fine, they don't have to be big TIFFs for this purpose.
What you will be left with is a batch of files that are detailed and very slightly oversharpened (assuming very good glass, no camera shake, that kind of thing). Examine a few hundred or a few thousand files at full resolution that have been processed this way, and you'll soon be able to differentiate which frames are perfectly in focus, which ones are just slightly out and which ones are more than slightly out. The combination of DPP's detailed conversions from CR2s coupled with slightly too much sharpening is important, because it becomes quite a bit simpler to spot the exact point of focus, and easier still when you're trying to assess focus in picture after picture after picture.
You can see the placement of the active autofocus point superimposed over the picture in Canon's ImageBrowser (Mac) or ZoomBrowser EX (Windows)
Q. Has your opinion about aspects of the EOS-1D Mark III other than autofocus changed since the first look article?
Not really: except for autofocus, the EOS-1D Mark III
is a photographer's dream machine. That's the way we felt after shooting the preproduction camera back in April, and this is no less true with the production version in June. Looking back, however, there are two things we would have given better coverage of or emphasis to when we wrote the first look article:
The rear LCD. Canon has done an excellent job of tuning the screen's tonality so that judging exposure on the 3-inch (diagonal) rear display is easier and less error-prone than with any camera we've ever used. Colour calibration is also decent for a screen like this; we were shooting side-by-side with an EOS 5D, for example, and the pronounced blue cast that whites take on in that camera's rear display is completely absent from the EOS-1D Mark III.
But, reviewed images look too soft. This is especially true when looking at a picture with some detail in it, like a group shot or a landscape, and especially in the full screen view, where nothing looks crisp. Zooming in improves things a bit; you can almost tell with some frames whether the focus is correct or not. Switch on Live View, however, and you'll see what the display is capable of. The zoomed view in particular is so much sharper, not in a way that masks focus error but rather one that emphasizes and makes it easy to set the point of focus. The on-the-fly sharpening of the display image in Live View needs to be ported over to image review, along with whatever else Canon can do to make pictures that are sharp actually look sharp on the rear LCD.
Highlight Tone Priority. It works exactly as Canon says it should: you get a bit more highlight detail and a bit more highlight gradation at the expense of a tiny bit more noise. For shooting outdoors on a sunny day at ISO 200 (the lowest ISO possible when Highlight Tone Priority is enabled), the only regret we've had so far is not using this setting more. It doesn't magically solve the problem of overexposure, but it does give properly-exposed files noticeably better highlights in wide dynamic range scenes, with no real downside at lower ISOs: the increase in noise is visible but the noise levels are still low, while shadow gradation remains superb.
Melon Ball: Canon EOS-1D Mark III production model (firmware v1.0.8) + EF 500mm f/4L IS, ISO 200, 1/2500, f/4 (Photo by Rob Galbraith/Little Guy Media)