RobGalbraith.com
Go to advertiser website.
     Home
     RSS
     CF/SD/XQD
     About
     Contact
Go
Go to advertiser website.
 
An analysis of EOS-1D Mark IV autofocus performance - Continued

This page contains links to downloadable collections of photos shot with the EOS-1D Mark IV. There are over 900 full resolution frames of speedskating, basketball, soccer, track and runner tests.

If you're seriously considering the purchase of this camera to shoot peak action sports, then we think it's critical you see with your own two eyes what the camera has been throwing at us, both good and bad. We experimented with a couple of different ways of illustrating the EOS-1D Mark IV's autofocus behaviour, ones that didn't require monster downloads. We concluded that getting a big batch of files onto your hard drive, where you can view them in a proper photo browser at full resolution, was the only effective way to go about this.

Hawks Head: Canon EOS-1D Mark IV (firmware v1.0.4) + EF 300mm f/2.8L IS, ISO 2500, 1/2500, f/2.8. Click photo to enlarge (Photo by Rob Galbraith/Little Guy Media)

Before we get to the links here's some important information about the gear we tested, camera settings and more:

In the downloads are pictures shot with four of the five EOS-1D Mark IV bodies we've used to date. One was supplied by Canon USA, and had been checked out by Canon USA service staff before we received it. Second is a body we purchased, the third was loaned for a period of time by a local shooter, while the fourth and fifth bodies we borrowed at events where there happened to be friendly photographers prepared to let us try their cameras. The vast majority of pictures were shot with the first three cameras, while all the track, soccer and runner tests were shot with the first two.

The lenses were as follows:
  • EF 85mm f/1.8
  • EF 70-200mm f/2.8L IS (focus calibrated by Canon service)
  • EF 300mm f/2.8L IS (focus calibrated by Canon service)
  • EF 400mm f/2.8L IS (new and supplied by Canon USA; before shipping its focus was checked by Canon service)
  • EF 400mm f/2.8L IS (borrowed briefly to compare to Canon USA supplied 400mm)
  • EF 400mm f/2.8L II (used briefly to compare to newer IS version)
Nearly all of the downloadable pictures were captured with the calibrated EF 70-200mm f/2.8L IS, calibrated EF 300mm f/2.8L IS or focus checked EF 400mm f/2.8L IS.

We figured out early on that to get the camera to track as best it can, the EOS-1D Mark IV must be configured to use multiple AF points. This is a departure from any previous Canon 1-series SLR we've used. For example, if the camera is set to use a single AF point alone, the number of in-focus frames can be considerably lower than we've described in this article (for lower light tracking in particular it can be a wipeout). In addition, the frequency with which the camera will frontfocus increases and the camera's 10 fps frame rate will constantly slow, and slow dramatically, as the AF system struggles to detect subject distance.

In our testing, enabling C.Fn III-8-1 (Left/Right AF point), made no discernible difference, C.Fn III-8-3 (All 45 Points Area) was impossible to use for sports because the AF point jumps around continually (in other words, this setting is meant for another purpose) while C.Fn III-8-2 (Surrounding AF Points) brought a noticeable improvement in both AF system feel and the number of in-focus pictures. While the frame rate still slows down a fair bit, it's not as big a drop and it doesn't drop as often as when this option is disabled (this behaviour is linked to the Custom Function discussed next).

All but a handful of the downloadable pictures were taken with C.Fn III-8-2 (Surrounding AF Points) turned on.

For all the photos, C.Fn III-3 (AI Servo 1st/2nd Image Priority) was on its default of C.Fn III-3-0 (AF Priority/Tracking Priority). Set this way, the AF system is to give priority to focus acquisition, rather than drive speed, when it's newly activated just before shooting a picture, and then between frames in a sequence. The result can be a slowdown of the shooting rate when the camera's AF system thinks it needs more time.

With the EOS-1D Mark III set this way, it would continue to plow through at 10 fps almost regardless of scene or subject. In contrast, the EOS-1D Mark IV on the same setting will frequently slow the frame rate. This isn't necessarily a bad thing if it means better autofocus. Since the goal was to wring the best AF system performance out of the EOS-1D Mark IV we left this Custom Function set to C.Fn III-3-0 always.

We fiddled with C.Fn III-2 (AI Servo Tracking Sensitivity) a lot. This setting controls how the AF system handles interruptions such as a ref running through the AF point and temporarily blocking the player. We weren't really needing to experiment with it for that purpose, since right out of the box on C.Fn III-2-0 (standard), it does a nice job of ignoring subjects and objects that appear briefly between camera and subject, at least ones that are significantly defocused. This is how we think it ought to work by default, and it does.

We wanted to see, by setting C.Fn III-2 to slower or slow, if instances of the frontfocus shift that occurs - even with no closer subject within the AF point group - would be reduced or eliminated. No dice. There was zero change in the camera's tendency to immediately and rapidly drive the lens to a much closer focus distance when it shouldn't, since there is in fact nothing closer to focus on. This setting appears to have no bearing on the EOS-1D Mark IV's frontfocus problem, and as such all downloadable photos were shot with C.Fn III-2-0 (standard) dialed in.

Switching C.Fn III-4 (AI Servo Tracking Method) from its default of C.Fn III-4-0 (Main Focus Point Priority) to C.Fn III-4-1 (Continuous AF Track Priority) served to do only one thing: make it more likely the EOS-1D Mark IV's AF system would begin tracking the background at just the wrong moment. With the EOS-1D Mark III, we could never reconcile the user guide description of what this setting is supposed to do and how it actually behaved, at least when shooting fast-paced sports. The EOS-1D Mark IV's C.Fn III-4 is equally enigmatic. As such, all downloadable photos were shot on C.Fn III-4-0.

The photos are a mix of firmware v1.0.4 and firmware v1.0.6. The latter version, which came out a few weeks after the EOS-1D Mark IV first hit the streets, introduced a single change: "AF performance has been enhance[d] for receding subjects and for subjects that are approaching at a low speed." Or so says the release notes for the update.

We photographed basketball, speedskating, soccer, track and runner tests with both firmware v1.0.4 and v1.0.6 and could see no difference in AF tracking. Whatever the tweak is that Canon engineers made, it has been undetectable. All downloadable photos indicate on them what firmware was installed at the time, but it's safe to view the two firmware versions as effectively identical, at least for what we've shot.

Fuzzy Logic: Canon EOS-1D Mark IV (firmware v1.0.4) + EF 70-200mm f/2.8L IS at 120mm, ISO 200, 1/320, f/6.3, Dyna-Lite Arena strobes. Click photo see an enlarged version with AF points superimposed (Photo by Rob Galbraith/Little Guy Media)
All photos have basic shooting information and AF point(s) embedded. The manually selected AF point is in red, while any active surrounding points are marked in blue (click on the thumbnail at right to see an example of the AF point overlay). With some multi-point configurations, including C.Fn III-8-2 (Surrounding AF Points), the EOS-1D Mark IV's metadata doesn't show which specific point(s) in a group were used for the AF calculation for that frame.

The downloadable photos show where the points were positioned, then, but there's no way of knowing which the camera actually chose.

All photos were captured as RAW CR2s and converted in Canon Digital Photo Professional 3.7.3 at full resolution. To get the photos down to a sort of manageable file size for inclusion in the downloads, they've been saved as JPEGs with moderately heavy compression. You can still clearly see where the camera focused, but you'll also see some JPEG artifacting and a slight overall degrading of image quality, relative to the original.

Also on the subject of image quality, some of the speedskating and much of the basketball is pretty grim. That's because, to get to a high enough shutter speed to ensure motion blur wouldn't make it difficult to see if the camera was focusing properly, we shot at up to ISO 10000. Some sequences required brightness boosting software exposure compensation too, which boosts noise also. Ugly mixed lighting and rushed white balance aren't doing the pictures any favours either.

Finally, we went with lower noise reduction than we might otherwise apply under these shooting conditions, so as to not introduce smearing that would make it harder to see where the camera set the focus distance to. In fairness to the camera, then, don't judge the EOS-1D Mark IV's image quality based on the nasty colour and noise you'll find in some of the indoor available light pictures, because it's a byproduct of how we captured and processed the photos to analyse AF. You will, however, get a sense of why we prefer shooting basketball with strobes.

Speaking of strobes, the basketball download contains mostly available light photos and only a handful shot with big flashes. That's because you can't really learn too much about how the camera's AF system is behaving by shooting one frame at a time: the picture is either in focus or out of focus, and if it's out there's no way to really know what the camera's AF system was up to right before it snapped a blurry picture. So, we've gone more with available light sequences.

As mentioned, EOS-1D Mark IV autofocus was at its very worst during soccer, on a bright, sunny morning when the players were frontlit. The light was intense, perhaps too intense somehow for the AF system. We've experienced tracking this completely terrible only one other time with the EOS-1D Mark IV, and that was while shooting informal runner tests with family and friends, in the first couple of days after taking delivery of the camera (a different body and lens than was used for the soccer). Those informal tests also took place in intense, frontlit sunlight, and the results were as described: completely terrible.

Tied into the above is subject colour. Throughout our outdoor shooting, the colour of at least one of the jerseys in many of the worst autofocus frames or sequences has been red. We attempted on two occasions to test this further in a more controlled manner, by having the runner repeat her runs with different coloured tops on, including red.

Two of those tops are shown in the downloadable runner sequences: blue (the PV singlet), and as you'll see, the results are good to great, plus red, where there is more than one instance of the dreaded EOS-1D Mark IV frontfocus shift. We shot other tops too, but it was really only the red that set off any alarm bells. Still, how the camera performed with the runner wearing the red top was not anywhere near as poor as in other more real world situations, such as sunny soccer.

eos_id_mark_iv_runner_a.jpg
eos_id_mark_iv_runner_b.jpg
eos_id_mark_iv_runner_c.jpg
eos_id_mark_iv_runner_d.jpg
Living Colour: Some of the top colours included in EOS-1D Mark IV runner tests (Photos by Rob Galbraith/Little Guy Media)

Downloadable photos

Click on a thumbnail below to start the download of a zip-compressed folder of EOS-1D Mark IV photos. The downloads are huge, ranging in size from about 177MB to about 869MB.

Note that we tried to assemble a much smaller sampler download for those wanting just a taste of EOS-1D Mark IV AF, but after a couple of hours of trying to put together a leaner collection that was still representative, we gave up. The EOS-1D Mark IV's AF system does not summarize easily, either in words or in pictures.

Be sure your viewing of the photos includes some time spent at a magnification level of 100% in your preferred photo browser. If you look at the pictures exclusively at 50% magnification or less then you'll miss seeing the true softness of some of the out of focus files.

Click to begin download
Click to begin download
Click to begin download
Runner: 139 files, 176.5MB
Speedskating: 127 files, 289.4MB
Track: 139 files, 217.3MB
Click to begin download
Click to begin download

Soccer: 344 files, 868.5MB
Basketball: 169 files, 597.4MB


Thanks to David Sparer, Erik Allin, Todd Lehman, Arno Hoogveld, Samantha Culiver and Nicole Acosta for their assistance with this article.
Next Page: February 19, 2010: Canon USA press release on EOS-1D Mark IV autofocus
Go to Page: 1 2 3 
Go to advertiser website.
2000-2013 Little Guy Media. Not to be reproduced without written permission.