>>3544866I mean, I tried to span a decent range of the focus range, from minimum focusing distance to 20m which would be practicality infinity for most lenses other than very long focal lengths.
Testing is when something is moving at, say, 5m/s (18km/h) doesn't say much because in 1 second the subject moved 5m, and if you captured 20 frames, it means you have a frame for every 25cm of your subject's motion. The camera could miss AF every other frame, and you still wouldn't notice because it would be covered by the depth of field.
Even at shortish distances, say 2-3m, an f/1.8 lens won't allow you to notice missed focus.
Think of it like that:
•frame 1: subject at 3m, lens focused at 3m, sharp
•frame 2: subject at 3.25m, lens focused at 3m, still sharp
•frame 3: subject at 3.50m, lens focused at 3.50m, sharp
•frame 4:: subject at 3.75m, lens focused at 3.50m, still sharp
etc etc.
The camera effectively autofocused at a rate of 10fps, but you can't tell.
However you still got 20 frames, and they're reasonably sharp, and the camera autofocused (even if not at all frames), so the manufacturer can say "20fps with autofocusing".
So a way to test against that claim vs reality is to
1. start from minimum focusing distance where the 0.25m off target will be visible. Or
2. move faster (like my suggestion for 20m/s) so for the off frames, the discrepancy between the lens focus point and the subject real distance won't be masked from the depth of field.
Or even better, combine the above.
There should be a distinction made between "20 fps while autofocus is working" and "20 fps, every frame individually autofocused".
It may sound extreme, but 72km/h isn't that huge of a speed, say if you shoot motorsports from up close.
Of course 20fps with autofocus (on every frame) is still overkill and not of great practical use, but if the manufacturers want to claim it in their brochures, they should man the fuck up and prove it.