The Canon EOS 40D retains the same AF pattern as the 20D and 30D, 9 zones arranged in a diamond pattern as shown below:
The number of zones is the same as the EOS 20D/30D and their layout is the same, but that's where the similarity ends. The EOS 40D uses cross type sensors for all the AF zones which operate with lenses with a maximum apertrure of f5.6 or faster. They respond to both vertical and horizontal detail in the subject. In the 20D/30D the eight outer zones were linear (single axis) sensors.
The center AF zone(shown in red) is also a cross type sensor (diagonal cross) but it also has an enhanced precision mode when used with lenses with a maximum aperture of f2.8 or faster which is capable of greater focus accuracy.
Canon claim that the AF calculation of the EOS 40D is 30% faster than that of the EOS 30D.
Note that 4 of the Af zones (shown in yellow above) are at locations near a "Golden Ratio" point. The golden ratio divides a line into two segments in the ratio of approximately 1.62. Since the days of classical Greece, this has been considered to be a very aesthetically pleasing way of dividing up an image. Positioning a subject at one of these "golden ratio" points often leads to a pleasing composition. It's very similar the the often quoted "rule of thirds" for composition.
Though I have not yet had a chance to fully evaluate the AF performance or compare it directly with that of the EOS 20D, in a brief handing of the camera AF certainly seemed a little faster and more positive with the EOS 40D than the EOS 20D, especially when using one of the outer focus zones.
[Update 09/10/07]
AI Servo Tracking Test
So far I've found the AF of the 40D to be fast, positive and accurate. It's hard to quantify, but it feels better than the EOS 20D, especially when using the outer AF zones. The following images show a test of AF tracking. The target was a car moving away from me at about 30mph, the lens was an EF 300/4L USM set to f4 (for maximum shutter speed and minimum DOF), the EOS 40D was shooting at maximum speed (6.5fps) and focus mode was set to AI servo.
Here's the first shot shown full frame:
The section outlined in red is shown in the following 100% crops as the vehicle moved away:
As you can see, the 40D held pretty accurate focus over these shots. That doesn't mean you'll get perfect focus tracking every time with every lens under every lighting condition of course, but it shows the 40D is certainly capable of focus tracking at 6.5fps, given a fast lens and a decent target.
[end of 09/10/07 update]
[Update 09/18/07]
Low Contrast AF Test
In this test I was interested how well the AF system of the Canon EOS 40D acquired lock on a fairly low
contrast target. The image below shows the test conditions. Both Canon EF 300/4L USM and a Canon EF 70-300/4-5.6IS USM lenses were used, each one set to 300mm. The target was in bright sunlight and exposure was around 1/500s at f5.6 at an ISO 100 setting.
AF on both cameras was set to "One Shot" mode. For those interested in more technical details, he target was actually some low contrast printing on the back of a cereal box!
With the Canon EOS 40D, AF lock was fast and positive, both with the center AF point and any of the 8 outer points covering the low contrast AF target area shown on the right above. There were no examples of misfocusing or failure to achieve a focus lock, i.e. the focus beep and the solid green light both confirmed positive AF and the resulting images were in focus.
With the Canon EOS 20D, no focus lock could be achieved, even with the center AF zone selected. The focus hunted before eventually giving up. With any of the outer 8 zones the same behavior was observed. If the focus zone was moved slightly over to one side or the other so that an area of the box with higher contrast detail fell under the AF zone, then focus could be obtained and was fast and accurate, but over the low contrast area of the target AF was never reliably achieved.
Conclusion: The AF performance of the EOS 40D exceeds
that of the EOS 20D under conditions of a low contrast target. The 40D was remarkably fast and accurate, while the 20D hunted for focus and failed. The EOS 40 was clearly the more capable camera in this test.
Low Light AF Test
In a quick test of low light AF capability I set up a normal contrast target in a darkened room and adjusted the lighting lower and lower until the camera failed to find focus. With the EOS 20D and EOS 40D both set to ISO 100, and using the same EF 300/4L on both, I found that the EOS 20D could usually get focus at a light level where the exposure was 15s at f5.6 (pretty dim!). With the EOS 40D I could go down to a light level one stop less, 30s at f5.6 and still achieve focus. Focus under these conditions was slow, and usually consisted of a few "back and forth" adjustments before locking in, but once locked it was accurate. Based on this single test and under these specific conditions, it looks like the EOS 40D can focus in about 1 stop lower light than the EOS 20D. Note however that the Canon specifications on the AF of the 20D and 40D in low light are identical according to the printed manuals. They are both rated for -0.5 to 18EV. In this case, 30s @ f5.6 and ISO 100 is 0 EV, while 15s @ f5.6 and ISO 100 is 1 EV.
[end of 09/18/07 update]
NEXT-> Part V - Live View