cyberstudio
Offline
Upload & Sell: Off
|
 p.1 #5 · misfocus on a sigma 50-150 ?? | |
The most difficult to get right is the mid distances. Mid distances are used very often, and to compound the problem, at mid distances even slight AF errors are easily discernible in the images.
We have been doing research on AF accuracy over the last 5 years and have taken tens of thousands of test target shots. In the beginning we made the mistake of focusing our testing on minimum focusing distance. We have since then recognized the mistake and put equal emphasis on closest focusing distance, mid focusing distances and even infinity. We are not forgetting infinity, even though there is supposed to be a lot of depth of field. When you critically evaluate 100% crops you realize there are different shades of "infinity".
Our cameras are cameras, not laser measurement devices. Properly calibrated, they do a great job under all real photographic situations. Overly focused on a single test case but ignoring the big picture always does more harm than good. Any autofocus accuracy testing must be comprehensive and cover a wide variety of test conditions. You want 90% of the test cases to be very accurate and the rest of the 10% accurate enough. You do not want 1% of the test cases to be super accurate but the rest of the 99% to be way off. Test targets are still useful tools because if you do test targets well the real photographic subjects will do even better. We have always had pretty gigantic AF test targets for this purpose.
The camera and the lens must be matched for AF accuracy evaluation. That is to say, the same lens may be accurate on one camera but inaccurate on another, and the same camera may be accurate on one lens but inaccurate on another lens.
Bo-Ming of conurus
Edited on Feb 10, 2010 at 12:39 AM · View previous versions
|
