I can certainly see why this issue doesn't warrant taking a lot of time, and why people might call this a waste of time -- yours and mine.
Well, I took the time -- for once -- to read a bit more of the thread on that other site, and it seems that at least some posters thought of Michael's objections.
The response was, in brief, that it might have something to do with the coating of the lens elements. And that at least one person had a "good" sample.
But still, it's only speculation, and as Michael mentions, this lens isn't exactly rarely used among photographers.
I find the lack of information from some of them disturbing:
- Which f stops did they use?
- Did they use a UV filter?
- If so, of which quality?
- And did they retest without a UV filter?
- Did they take their lens out of the fridge before shooting at 70% relative humidity?
- Have they scrubbed their lenses with sandpaper? Who knows.
Okay, the last two may be out of line.
So I decided to set up a test shot myself (and I certainly don't blame anybody else for not doing this):
- EOS 20D
- EF 24-70mm f/2.8L
- Compact fluorescent bulb (IKEA brand) in desk side lamp
- Poor ambient lighting
- Bookshelf at ~1m distance
- AF set to single point center
To ensure that those ominous IR rays would have a chance at disturbing the autofocus, I didn't use a lens hood. No UV filter, either.
I took four shots with the self-timer and mirror-lockup at ISO 100, 70mm, whereof two at f/2.8 and two at f/5.6, and of course I refocused between each shot. Two of these were with the lamp aimed at the lens at an angle that didn't put the lamp itself in the view:
I used PS CS2 with the most recent ACR, camera white balance, all other adjustments zero (not auto), prophoto RGB. Resized in CS2, 100% crops from CS2. Converted to 8bit, then sRGB with relative colorimetric and dither. Converted to JPEG with save for web, JPEG maximum quality (the 300x200 minis are in medium quality). No, I didn't bother to do anything in CS2 except that.
Image 1 (f/2.8, lamp aimed at camera, 0.3s):
Image 2 (f/2.8, lamp aimed at bookshelf, 1/30s):
Oops, out of focus?
Yes, it appears to have back-focused slightly. The first f/2.8 shot was similar, but to a lesser degree.
Image 3 (f/5.6, lamp aimed at bookshelf, 1/8s):
Image 4 (f/5.6, lamp aimed at camera, 1.3s):
I can't really say that any of these images show any evidence of the problems that people claim in the DPReview forum. But this is my lens and my camera. *shrug* My guess is that it is the photographers having focusing problems.
That being said, I had an interesting experience with my 20D and 24-70L myself yesterday evening. I was out shooting around the abundant forest roads in Maridalen, Oslo, hoping to capture some scenes I saw while on a bicycle trip earlier in the evening. There had just been a rainfall, and I thought "cool, that will finally let me capture how greenery looks just after rain".
The 24-70L consistently back-focused. On subjects 2-3m away, it insisted on focusing at more than twice that distance (close to 10m according to the lens). Weird. Perhaps I'll see a hint when I look at the raw files later.
In other situations, the camera and lens have been behaving in a predictable manner.