Originally posted by janneman But there there is more to "fast AF, than just the in-lens-AF motor. Yesterday I played a bit with K-7 and Tamron 18-250 in relative low light.... rough focus was about half the AF. That time was almost doubled by the final corrections. Moreover, for example, without moving lens on another sbject, if trying a second shot (same object, same distance) AF aain did a few adjustments. [...]
Also, SR "kick-in' time???
janneman,
you're right and I hope I wrote nothing else in the article.
The problem was that with e.g., a DA*16-50, the focus on both K-5 and K-7 was almost instant in the office-alike Pentax business area lighting (if the lens was only defocussed a bit). It was then impossible to say which one beeped or triggered faster.
Visually, I got the impression that the K-5 started to focus faster and also locked faster. Maybe, the difference I saw was significant actually. But we talk about ms differences jugded with the naked eye here. So, maybe I saw nothing.
So, a real benchmark where focus times can be measured with ms accuracy must be made. And another one in low light. And accuracy should be
measured too as it is often ignored by magazine reports. AF.C is another topic. As is single image AF.S or AF.C
accuracy for a moving subject...
For me, the real difference is the AF sensor measurement integration time. It is unknown (except I saw a few bits about it in the pre-production camera EXIF fields). If this time is small, focus on fast moving subjects is possible (even if it then moves out of focus -- at least with trigger priority and maybe prediction, the result with not too thin DoF should still be usable). If this time is large, focus can't be measured because it acts as if the subject would be blurred or noisy. Like AF becomes impossible freehand at too long focal lengths or in very low light. The AF then stalls and the camera won't trigger anymore.
So, with fixed subject contrast and illumination, there should be a limit for maximum subject speed (magnification change speed actually) at which AF stops to work. If this can be measured (together with the latency between AF measurement and photo taken), we may have a good characterization of an AF system. Much more meaningful than all the common trigger latency tests (200ms vs. 300ms etc.).
Last edited by falconeye; 09-29-2010 at 04:19 AM.