Wed 05-Sep-12 02:08 AM | edited Wed 05-Sep-12 02:14 AM by Ferguson
OK, lacking a clear understanding (at least mine) of the focus modes, I set up a controlled test. I'm sure people will find reason why it is not adequate, but it is informative.
I took an old-style metronome, and put a yellow postit note on the top with a big red + mark on it. The yellow is about twice the size of a focus point.
I then set it at the slowest speed, so it takes about 2 seconds to go over and back. I got a lens suitable to have it traverse about 3 focus points on the lens (starting in the center, cross the 9 point boundary and into the 21 point boundary.
Then I tried bursts at 9, 51, 3D to see what would happen. Set for 10fps, normal (3) for lock-on time, and release+focus (at first).
The 9 point mode clearly had a boundary - when I got outside of it, it reverted back to inside its zone. As expected.
It very clearly remembered where it was on the last exposure, or otherwise would keep finding it much of the time. As it should.
3D tracking (which I gave a bit of an advantage to with color) was better than 51 points.
But frankly the tracking stinks. Not sure how well this shows up, it's a series of 6 shots where it tracking, then (the plus is moving very slowly left) it skips 3 different times focusing, then it is tracking again on the same focus point.
There were a LOT of gaps like this. The good news is during the gaps, it did NOT change focus. I looked at very high magnification, and generally when the focus point display was gone from ViewNX, the focus appeared in exactly the same distance.
The bad news is when it did "notice" it again, it frequently mis-focused for at least one frame.
These are two consecutive frames, the first was one of a series of 6 or so where it lost it, but each remained in good focus. The bottom frame is where it found the card again, and mis-focused.
The next frame after was more or less the same (without focus indicated), the following showed the same focus point but was in good focus again.
In the 3D mode it never grabbed the background -- it did grab the top of the metronome a few times, which I guess is not too unexpected.
In 51 point mode (not 3d) it frequently grabbed the background. in fact it was so erratic as to be questionable whether one could say it tracked at all. And when it grabbed the background it focused firmly on it, the metronome and card went entirely out of focus.
9 was more accurate than 51 until it went out of range -- when it came back in range it took a long time to lock in again, but until it went out of range it was better than 51.
OK, so it's probably not scientific - but then again neither are volleyball players. It's fun to play with. And I do think I may try 3D with sports next time -- I've always assumed it was like "automatic" (which I consider worthless), but maybe it would track better. But as far as it doing well at tracking -- not in this case. Not at all.
PS. The camera was locked down on a tripod and not moved between series. it was a 24-70 @ 48mm, F2.8 in "A" mode, auto-iso ended up at 1800 hence the noise, 1/50th.