Originally posted by RaduA I still think that the measurement type used by ColorFoto (and blende8 to slaughter the DA*55 btw
) is unrealistic in general usage where much smaller corrections are way more common.
It is reasonable to assume so.
Still, I disagree after having run my own tests.
E.g., look at the 5.4EV sound sample for Pentax K-7 in my blog (
http://www.falklumo.com/downloads/pub/2009.06/afs/K7-03.mp3). You hear three focus operations: First from 35cm to ~1m, from then on staying at 1m.
The initial "long run" of the screw drive to 1m was actually faster (60ms) than the second confirmation long run (90ms). The third run didn't use a long run. By long run I mean a significant shift of focus. In all three cases, the long runs have been followed by two short runs (micro-adjustments) each. It is the total time of short runs which dominate the total focus time (560ms in the first, 430ms in the second case, delay from shutter half press to focus activity included).
Therefore, it doesn't matter too much if the lens was far from final focus in the beginning (for the K-7 and kit lens). It only saves 1 out of 3 or 4 focus operations total. They may be unnoticeable with the K-7 (20ms in 50ms succession) but the times are adding to the dominant term in shutter latency. I think this may be the price to pay for good focus accuracy.
It would be interesting to watch for microadjustment focus operations in AF.C mode. Maybe, it is has a lower shutter delay where initial focus position matters more.