August 8, 2007 : EOS-1D Mark III autofocus with firmware v1.1.0 installed

Q. Have you tested the camera's autofocus since firmware v1.1.0 was released late last month?

Yes.

Q. Does it fix the autofocus problems?

No. As with firmware v1.0.9 and v1.0.8, the EOS-1D Mark III's ability to acquire focus initially, hold focus on static subjects and track moving subjects remains poor. In bright sunlight on warmer days in particular, the camera delivers an unacceptably low percentage of in-focus frames. The EOS-1D Mark II N's autofocus, by comparison, is awesome.

Firmware v1.0.9 and v1.1.0 include a change to AI Servo autofocus that does reduce the EOS-1D Mark III's tendency to shift focus to the background prematurely when the active AF point moves partly off the subject for a moment. This is what we thought we were seeing with v1.0.9 last month, and after shooting several thousand more frames with v1.1.0 loaded, it's clear that this is the case. This improves the feel of the camera's AI Servo tracking, but it doesn't increase significantly the number of properly-focused pictures in our testing.

Q. Are you sure the problems you've had with EOS-1D Mark III autofocus aren't because your gear needs calibration or repair?

Yes. We're of the firm belief that the EOS-1D Mark III's autofocus problems come from something other than a body or lens that needs adjustment or parts replacement. Using just the firmware v1.1.0 testing as an example, the two EOS-1D Mark III bodies, as well as EF 300mm f/2.8L IS and EF 500mm f/4L IS lenses, were all checked and calibrated by Canon USA service technicians before they were shipped here, so that they could definitively rule out equipment miscalibration or some other malfunction as the cause of the autofocus problems we've been reporting.

Q. Has Canon acknowledged that the EOS-1D Mark III has the autofocus problems you describe?

No. But two Canon USA representatives observed our firmware v1.1.0 testing here in Calgary on July 31 and August 1, then hand-carried a hard drive of files back with them for their own evaluation and for onpass to engineers in Japan (and we've subsequently learned that the hard drive has made it to Tokyo).

The hard drive contained a collection of about 3700 EOS-1D Mark III and EOS-1D Mark II N files shot last week while Canon USA staffers took notes and captured video of the proceedings. We lined up a runner and a soccer player for their visit, and fired off multiple rounds of each athlete in action, on a variety of autofocus settings and wearing different jerseys (the athletes, not us).

The lens in all cases was an EF 300mm f/2.8L IS. Settings changes included slowing the camera to 8 fps, dialing in automatic focus point selection, enabling autofocus point expansion (C. FnIII-8), slowing AI Servo Tracking Sensitivity (C. FnIII-2), switching on Continuous AF Track Priority (C. FnIII-4) and experimenting with AI Servo 1st/2nd Image Priority (C. FnIII-3).

On the first day the sky was somewhat overcast, the temperature was about 16 蚓 /61 蚌 and the EOS-1D Mark III's autofocus performed not terribly. The EOS-1D Mark II N's autofocus easily trumped the EOS-1D Mark III, but the EOS-1D Mark III did make a number of in-focus photos.

On the second day it was a clear blue sky and full sunlight, the temperature was 20-21 蚓 /68-70 蚌 while shooting and the EOS-1D Mark III's autofocus performed terribly: in the majority of sequences, the majority of frames were blurry. By comparison, the EOS-1D Mark II N worked as it always does shooting stuff like this under these conditions: it got most frames perfectly focused or very nearly so.

These two days worth of pictures are what Canon left with, and while the day two shoot especially shows the EOS-1D Mark III's autofocus system performing poorly, we had a nagging feeling that Canon didn't see the camera performing at its very worst, particularly in the soccer player sequences shot on the warmer of the two days.

Because of this we ventured out again, after the Canon USA folks had left, to shoot in the sunny, 27 蚓 /81 蚌 weather that was Calgary on the afternoon of August 3. These photos show the EOS-1D Mark III's autofocus system at times failing to an even greater degree than earlier in the week.

We can't say for certain that it's because of the warmer temperature that day, since it was a different athlete wearing a different jersey too. But because the particularly poor autofocus performance we saw in the August 3 photos is consistent with all previous warm day shooting with other athletes and other uniform colours (not to mention other EOS-1D Mark III bodies), we continue to think that ambient temperature plays a role in the camera's autofocus difficulties.

A selection of photos from the August 3 session has also been sent to Canon. With or without the August 3 photos, however, Canon now has compelling evidence of the type and degree of autofocus problems the EOS-1D Mark III has.

Note, however, that this opinion is still ours alone. Canon USA  isn't ready to go on record with an acknowledgement that the camera's autofocus is deficient. And while Canon engineers at the company's Tokyo headquarters have commenced an investigation based on the files we've supplied and data from Canon USA that describes the shooting conditions, camera settings and more, Canon in Japan hasn't confirmed there is a problem or set a deadline for when they'll complete their investigation.

Put another way, the pictures speak for themselves, but it may be some time yet before Canon speaks about them, pending the result of the process that's underway now.

Q. Didn't you send files and feedback about the EOS-1D Mark III's autofocus problems to Canon awhile ago?

Well, yes, in early June actually. And in a perfect world, those files - which deliver the same message about the camera's autofocus problems as those we shot last week - combined with earlier feedback would have been enough to convince Canon in Japan to dig in and investigate.

But that's the glass half empty view of this situation. The good news is that Canon USA is now tackling this head-on, by dispatching staff to view our testing, by sending carefully-calibrated gear for us to shoot with and by using their influence with Japan to kick-start the investigative process.

The commitment by Canon in Japan to investigate our claims of EOS-1D Mark III autofocus problems doesn't also commit them to a course of action once their investigation is complete (other than sharing the results of that investigation). But the flurry of activity initiated by Canon USA in recent weeks means there's now reason to be hopeful that a fix may come in time.

Q. When will you be writing about EOS-1D Mark III autofocus again?

The moment that Canon passes on the findings of their investigation. Until then, you can assume that if this article hasn't been updated, it's because we think that EOS-1D Mark III autofocus still doesn't work properly.

Q. Can we see some of the sequences you've shot with firmware 1.1.0 loaded?

Yes. Below you will find links to eight different sequences of full resolution pictures. Five were shot with the EOS-1D Mark III, and three were shot with the EOS-1D Mark II N. The cameras were set to capture CR2s. The CR2s were processed to JPEGs in Digital Photo Professional with Sharpening set to 3, then run through a Photoshop scriptxx that superimposed the active AF point on the frame, so you can see where the point of focus should be. (For more information on why these pictures were handled this way, see here.)

Here are some viewing tips:

• As before, the sequences were selected not because they're portfolio-quality photos, but because they're representative of the autofocus traits of the EOS-1D Mark III and EOS-1D Mark II N as described in this article. Note that these pictures haven't been adjusted in any way other than sharpening, and in some cases the colour and brightness would have benefited from more appropriate Digital Photo Professional processing settings in particular. In other words, please judge focus in these pictures, but try not to get hung up on image quality, since you might be seeing the result of less-than-optimal processing settings rather than a camera image quality problem.
• To have a proper look at the photos, view them in a browser like Photo Mechanic that enables you to display them rapidly at full resolution. You can also open the photos one-by-one in Photoshop, magnifying each to 100%, but this will be a cumbersome process. Shooting information is both in the EXIF metadata and in the Caption field.
• The sequence called Static1 was shot with AI Servo AF active throughout the sequence; for the sequence called Static2, the camera was set to One Shot AF, and the autofocus was reactivated between each frame. AI Servo AF was used for all other sequences.
• While our firmware v1.1.0 testing included various autofocus settings combinations, to keep things simple we've included here only sequences shot with the EOS-1D Mark III set to fairly standard settings: 10 fps, centre point AF without expansion, AI Servo Tracking Sensitivity on normal and so on. Put another way, C. FnIII-1,3,4,5,6,7,8 were all set to 0, while C. FnIII-2 was set to the middle position. The settings variations we tried yielded either the same or fewer in-focus frames (reinforcing the point that the most serious of the autofocus problems outlined in this article can't be solved by adjusting the camera's Custom Functions).
• All of the photos below have been taken with an EF 300mm f/2.8L IS. When attached to the EOS-1D Mark II N and various other Canon cameras past and present, this lens is the best autofocus lens we've ever used. It's Canon's autofocus standard-bearer really. We know from shooting the EF 24-105mm f/4L IS, EF 70-200mm f/2.8L IS and EF 135mm f/2L with the EOS-1D Mark III that the negative autofocus characteristics of the EOS-1D Mark III show themselves with these lenses too, but less so.
  
  For example, when shooting runner sequences with an EF 70-200mm f/2.8L IS, the number of in-focus frames captured by the EOS-1D Mark III has approached the okay level, but when using the same camera and lens to shoot an actual soccer game, the autofocus performance has been poor. The EF 135mm f/2.0L is our favourite Canon portrait lens because it's optically phenomenal, but it has been difficult to get the EOS-1D Mark III to autofocus consistently and accurately - even in One Shot AF mode - with this lens and a subject that doesn't fill the frame. By comparison, we've had relatively little problem autofocusing normal to wide lenses like the EF 50mm f/1.4 and EF 35mm f/2.
  
  All of this means that if you don't shoot things that move and you don't own an EF 300mm f/2.8L IS or one of Canon's other supertelephotos, you may not see the EOS-1D Mark III's autofocus problems to the same degree as shown below. But our experience suggests that the autofocus problems carry over to other lenses and other shooting situations, just not to the same extent necessarily.
• The EOS-1D Mark III sequences below show the autofocus system performing poorly. Conversely, the EOS-1D Mark II N sequences demonstrate why we like the older model so much for sports. But these sequences are here so you can decide, for example, whether the EOS-1D Mark III's autofocus is or isn't good enough for your purposes, or so you can decide whether the EOS-1D Mark II N is really as good as we keep saying it is, for the shooting of things like soccer, rugby and football.

Click on a thumbnail to download a ZIP file containing all the pictures in that sequence. The first five sequences were shot with the EOS-1D Mark III, the last three were shot with the EOS-1D Mark II N.

These photos are for personal viewing and printing only. They may not be republished in any form without the permission of the copyright holder. This includes the posting of these photos onto another server.