Jeff
Offline
• • • • • •
Upload & Sell: On
|
 p.20 #10 · ['NEW Fix' UPDATE!] - MkIII AF still broken | |
Of note, today I overnighted a DVD of 'post-fix' RAW AI Servo images to VA at CITS' request. It will be very interesting to hear the results of their inquiry, as they will need to give me some sort of feedback as to the reason for the rather poor (~55%, on average) hit rate. I still say they are going to try to say the camera focused on the (rather small) snowflakes... 
Anyway, in going back through my images last night, I started thinking some more about my most recent AI Servo tests from my supposedly 'fixed' MkIII, and it prompted me to go back and completely re-read all of Rob's 8500-word December 10 update.
A couple of things that gave me pause in Rob's most recent report on 'Blue Dot' MkIII's, quoted below:
Q. Are there any settings that can help squeeze the best possible AI Servo autofocus performance out of the EOS-1D Mark III?
Yes, one: C. FnIII-7 [AF Microadjustment]. This setting's primary purpose is to enable the photographer to compensate for focus calibration errors in the body or combination of the body and attached lens. Its hidden purpose is to improve the EOS-1D Mark III's tracking performance. The Phoenix testing with Canon included shooting sequences with an AF Microadjustment amount of 0, -1 or -5 applied (via the [Same amount for all lenses] option). The inescapable conclusion was that the -1 sequences contained fewer frames that were completely out of focus than the 0 sequences (the -5 sequences were uniformly poorer all around and this amount was excluded from subsequent testing).
If we'd rattled off only a handful of 0 to -1 comparisons, this result could have been chalked up to a testing fluke. But because we shot and compared dozens of sequences and could see that the same slightly better performance in the -1 frames is consistently there, we've gone on to shoot almost all real-world peak action with -1 set, and recommend that you try doing the same.
To be clear, dialing in -1 for AF Microadjustment in this case is not about compensating for focus calibration error. The bodies and lenses we used to test the impact of AF Microadjustment were all calibrated by Canon USA or Canon in Japan.
Nor is it about making all photos be frontfocused slightly (minus side adjustments give the focus a slight frontward push). Instead, it's about making more properly focused or very nearly properly focused frames when the camera is tracking. The weirdest part of all is that the most notable difference between 0 and -1 sequences is that the -1 sequences show fewer instances of the camera focusing too far in front of the moving subject.
The tracking improvements brought about by this AF Microadjustment change are slight, but they sure seem to be real. We can't explain why this is so, and we also can't take credit for the idea: it came straight from Canon, as part of the testing regimen specified starting with the October Phoenix trip.
The suggestion to dial in a -1 AF Microadjustment raises a few questions:
What do you do if you already have this Custom Function set to compensate for focus calibration error with one or more of your lenses?
Should the -1 AF Microadjustment be set only when shooting things that move?
When photographing static subjects, but with the autofocus mode set to AI Servo, is it better to have the camera set to 0 or -1?
Do those who use One Shot autofocus mode choose 0 or -1?
In the future, will Canon somehow insert this AF Microadjustment trick directly into the EOS-1D Mark III's autofocus algorithm - perhaps by applying -1 AF Microadjustment compensation automatically when the camera senses subject motion?
The answer to all of these questions is the same: we don't know. What we've chosen to do for now is set -1 when shooting action, then switch over to 0 when shooting portraits. We can see that -1 is the way to go for sports, but we haven't done enough comparing of the difference between 0 and -1 when shooting portraits and other static things to know which setting is better or if in fact it makes any real world difference at all....Show more →
I still continue to find this whole '-1 AF Microadjustment' thing very odd, and haven't yet tried it on my camera. It makes me suspect that Canon may have written some sort of software kludge into that single setting, but I have no idea why they'd even need to use such a strategy. My 'bad' AI Servo results most often showed front-focusing of 6" to 1 foot, so perhaps I need to try this non-publicized strategy of Canon's. Either way, I don't know what to think (until I try it, that is).
Otherwise, our experience has been that - like other 1-series Canons before it - the EOS-1D Mark III will deliver the strongest raw autofocus performance in configurations that are more basic. Assuming that the autofocus point can be kept on the subject almost all the time, manual selection of the centre AF point is the way to go. Our confidence about this was bolstered by the fact that Canon's testing regimen for the Phoenix trips called for only centre AF point shooting.
Interesting to note, for those testing their camera for AF oddities.
Finally, the fact that RG has not yet tested a 'fixed' MkIII makes me wonder just a bit. It's too bad they didn't give him that opportunity when testing the Blue Dot versions. It would be very interesting to know if most of the people not experiencing problems with their MkIII have assembly line iterations, or whether just as many 'fixed' versions are performing properly, at least for the subjects that are being shot.
I'd guess it's going to be awhile before RG hears from Canon (if ever), and thus awhile before we hear anything substantive from Rob.
Have a good weekend, y'all...
Jeff
|
