June 15, 2009
K-7 and its continuous autofocus -- PART II
This second part of the article about AF.C in the Pentax K-7 is based on experience in general and from a less controlled test situation.
To speak in general terms, the continous autofocus in the K-7 is significantly improved over, e.g., the K20D. However, so far I know of no autofocus which is fool proof in very difficult situations, like nearby kids running around you wildly. This would require focus latencies well less than 0.1s. I estimate focus latencies for AF.S in the K-7 to be 0.3 to 0.5s, and AF.C may be about 0.2 to 0.3s.
So, I decided to find out just how badly AF.C would fail in such difficult situations. I asked a friend to run "eights" around me, as close and fast as possible. I tried to make as many shots (hi burst) and counted how many good ones and acceptable ones I got. I used the kit lens at 55mm, f/5.6. To tell the truth, I missed to have the subject in center in a couple of shots and I had the AF set to use center point only.
Because of the awkward testing situation, I decided to count as "good" what looked sharp at 1600x1200 resolution. And "acceptable" what looked sharp at 800x600 resolution. "Excellent" would be sharp at full 4650x3100 resolution. The photo at the beginning of the article shows one of the "good" examples from the K-7. I did the experiment with both, a K-7 and a K20D.
Results from random run:
35 images in 23 s (1.5 fps)
4 excellent, 5 good, 12 acceptable, 14 fail.
24 images in 21 s (1.1 fps)
0 excellent, 2 good, 6 acceptable, 16 fail.
Esp. comparing the excellent/good ones we see that the K-7 could actually capture some of the moments which the K20D couldn't. But it is certainly not fool-proof as well.
Results from straight run:
I made another experiment, with the same setting, where I made the person running straight towards me, as fast as possible, starting from 25m away.
3 excellent, 12 good, 1 acceptable, 1 fail.
6 excellent, 2 good, 2 acceptable, 3 fail.
With ~90% good images, the K-7 looks like a real winner. Of course, I found it a bit cumbersome that only 3 out of 15 good images have been tack sharp (less than from the K20D). Seems like the AF.C in the K-7 has a bit more shutter priority than in the K20D. Nevertheless, with only 60% good images, the K20D was no match for the K-7.
This confirms my general observation with the K-7: in most situations, i.e. in situations where it would be straightforward to keep a moving subject in center will most images turn out sharp. If this is difficult and if one would have to rely on automatic AF field selection then the AF.C in the K-7 will most likely run into trouble too.
Overall, the AF.C in the K-7 is a good match for its improved speed which is a ~70% improvement over the previous K20D.