Originally posted by kenyee Thanks for taking the time to this. Looking forward to part II (kids running araound) because that's what I think most people mean by "AF-C", but I suspect they'd be surprised how badly Canikon do at that range as well. If you do a part III, BIF would be good...
Part II was out long ago
->
Falk Lumo: K-7 and its continuous autofocus -- PART II
Maybe, lock at my conslusive review report (latest blog entry) where I listed all review parts I've done.
I asked a friend to move as erratically as possible (so, she came really close as well
). K20D was a complete fail, K-7 went down to 1.5fps and 25% in focus only. BUT ... as the next cited post points out, NO camera out there can do this. It is just not state of the art. According to a little math I did, focussing would have to be as fast as 50ms but no camera out there can focus faster than 200ms (K-7: 300ms).
Originally posted by ManuH Even using a D3 and a fast SDM lens is apparently not enough for "kids running around".
Originally posted by ManuH That's interesting because I can't get much more than 3fps in AF-C even in bright light with stationnary subjects. Maybe your pre-production camera was tweaked differently or there is a bug in the final version?
I don't think so. But you're right, my figure was with a beta firmware. I've seen figures as bad as 1.5 fps too (cf. above). It may be a question of subject speed, available AF contrast or which lens was used, or what aperture. Dunno. My test used a car's radiator grill as mandated by the standardized test setup.
Originally posted by lol101 Then shouldn't the K-7 blast at full 5.2fps in AF-C, no matter if focus is achieved or not...?
No.
AF.C means that refocus is required in between the black out phases. A contrast autofocus system (if fast enough which nobody seems to be able to do right now) could maintain full speed in AF.C. Not a phase autofocus system, from no vendor.
If the K-7 drops from 5.2 to 3.9 fps in AF.C mode, then the time per image increases from 192 ms to 256 ms, so the additional focus is done in only 64 ms per image which already is a very good figure. Maybe, the black out phase on the K-7 is shorter than 192 ms (as would be the case if 5.2 fps is processing-speed limited -- which I don't think as 4 read-out channels can maintain 6 fps). But figures are still rather good.
Note that in my car test, the K-7 did not refocus between every image. It verified focus and in maybe 50% of images, kept it. This is one advantage of fast fps: there is less refocussing needed per image, and if it does refocus, it travels less.
So, whatever be your brand, switch to AF.S and maybe, lock exposure, to achieve full fps.
BTW, the K-7 even outperforms the marketing spec in that it can sustain 5.2 fps until card full with fine-tuned settings (like high JPG compression).