Canon Live View - how much better than using AutoFocus?

In the Fall of 2007, I upgraded from a Canon Rebel XTi to a Canon 40D. One new feature I was excited about was Canon's Live View which raises the mirror and has the CMOS sensor provide a "live view" on the 3" LCD at the back of the camera with 1x, 5x, and 10x zoom capability ... plus it's 30 frames/second!

One touted benefit of this feature is using the CMOS sensor image for focusing. I.e. you can see what it would image, and then manually adjust the focus - note that (unfortunately) Canon does not provide contrast based AF (at least not yet) so the camera can't use the CMOS image for auto-focus. In addition to remote/cramped shooting situations, Live View can be extremely helpful in unusual and/or very low light situations such as Astro-Photography ... or if you have are trying to get that focus really, really dead-on .. especially in Macro work.

I've had some back focus issues so I was curious to see how well the standard autofocus system worked using Canon's new live view system. Turns out this is extremely easy to do since the Canon 40D has an option to toggle the mirror up/down - grabbing autofocus and then going back to Live View.

So I setup a tripod on my 2nd floor balcony. For my initial testing, I shot at night, using my driveway lights and some halloween pumpkins across the street as focus targets. Since this is a high contrast situation, one would expect the best performance by the AF system. It's also easier on my (old) eyes to manually focus! The next day, I did some daytime shots - intentionally under cloudy skies so it would be a more difficult test. There are some slight exposure differences, but focus on the ... focus ... ;-)

Camera settings were full manual for all tests and timed shutter release was used. First, the center focus point was used ... and then the camera was switched to Live View and I tried to get a "better" focus using the 10x view. To keep it simple, in-camera generated JPEG's are used. Lens used (at full zoom and maximum aperature to minimize depth of field) were the 50mm F/1.4, 28-135mm F/5.6, 17-85mm F/5.6, 70-300mm F/5.6 DO. For the night shots, all of the F/5.6's were ISO 1600 (yes, the XTI is slightly darker at that speed) and the F1.4 was at ISO 100. Shutter speed was 1/100 of second. The daytime shots were all ISO 100 with the 50mm at 1/3200 and everything else at 1/400.

Note that I included some shots from my Canon Rebel XTi - I was pleasantly surprised to see it do quite well. BTW, one thing that was VERY obvious is how jittery long telephoto lens are. At 10X Live View, the 300mm lens bounces all over the place ... and took several seconds to stabilize on the tripod after I let go.

To analyze this yourself, simply click on the link below of the image you want to see, then the view, then the lens, and finally which camera/focus you want to see. Javascript is used to "flip" the image to the one selected. Give it a few seconds the first time as the browser caches the image - after that, you can go back-n-forth to easily see differences.

OOOPS - you do NOT have Javascript enabled - this is required for the links below to work.
OOOPS - you do NOT have Javascript enabled - this is required for the links below to work.
OOOPS - you do NOT have Javascript enabled - this is required for the links below to work.

Choose Image: driveway or pumpkins (night shots) or Hulk & Paint Can in daylight
Choose View: original 25% resized ... or crop of the center (all scaled 1x, 2x, or 4x to 972x648):
Choose Lens: 50mm - 85mm - 135mm - 300mm
Choose Camera/Focus: Canon 40D Live View - Canon 40D AutoFocus - Canon Rebel XTi AutoFocus
972x648 non-resized center crop, no sharpening or anything else, saved as JPEG quality=8

Hulk & Paint Can in daylight shot fromCanon 40D using Live View with 300mm lens - center crop

Conclusion: This is a subjective analysis, but I was very impressed how well the Canon autofocus did ... although the night shots are a high contrast situation and best possible case. After I used the camera autofocus, I had a very difficult time improving the focus using Live View - this was also true in the follow-up daytime test. There was a difference between Live View with the 50mm/F1.4 lens ... but there aren't a lot of pixels at that wide angle. I.e. the standard autofocus is pretty darn good! ;-)

Note that the XTi was comparable except for the 50mm/F1.4. Perhaps not surprising since they use similar high-precision auto-focus at the center point ... but the 40D has diagonal sensors that come into play for lens of F2.8 or faster ... and this is more pronounced in the daytime shots (which has less contrast for the autofocus system). I've always felt the 50mm lens wasn't that sharp on the XTi at longer distances - glad to see it improve with the 40D. Note that it's possible that "L" lens may show more of a difference; anyone want to send me some for testing! ;-)

A quick-n-dirty test of the above is to use the non-center focus points and see if there is a difference in the final results using those versus the center focus point on the Canon 40D. I shot some pictures outside and inside ... didn't look that different ... so may be that the autofocus on the 40D is just a bit better in general.

Add an "original" to the URL path if you are interested in seeing the full-res images.
I also did some Canon 40D AI Server autofocus testing and welcome any comments/suggestion on all this.