One of the best interviews I’ve read (in Japanese only, unfortunately) about what matters in contrast-based focus systems for m43’s cameras. This was some press stuff that Panasonic did when it released the gh2 and the 14-140mm lens.
A quick summary of the major points:
- 4/3 systems (pre micro-4/3) were phase detection based. As such the body did much of the focal point discovery, and only had to direct periodically as to what the focus distance should be set to. With contrast optimization, the body needs to constantly read out the current focus point of the lens in order to correlate the current contrast level and the focus distance. That means more information bandwidth, and this lead to two extra data pins between m43 lenses and m43 bodies as compared to normal 4/3. Conclusion? using 4/3 lenses on m4/3 bodies will always be slower to AF.
- In the GH2, the sensor read out rate for focusing was doubled to 120fps from 60fps. Since the body is sampling contrast levels as the focus motor moves, doubling the sample rate means both that the camera can iterate faster when zoning in on a target, but also it can actually move the focus motor faster without fear of undersampling. (In order to make sure you don’t miss contrast peaks, you need to sample at some minimum delta between focus distances, so sampling faster means you can change the focus distance faster.)
- Upping the readout rate means less light per sample. Improvements in the sensor enables faster sampling because you don’t have to spend as much time collecting each sample. The sampling rate also apparently varies depending on light level (makes sense)
- There’s some algorithmic stuff on the body side. The body tries to accelerate the movement of the focus point if it detects that it is far away from the peak contrast point, and only slows it down once it has zoned in on an area.
- Sounds like the 14-140 also was the first to use a ultrasonic linear motor (which I guess is what you find on many dslr lenses?) Being able to move the lens quickly and precisely really matters in a focus system which is essentially a feedback loop between lens control and sensor readout.
Though I originall read this article a long time ago, I ended up reading again after looking at a bunch of new potential lenses for my gh2. I wanted to know if there was some proprietary junk that made the focusing improvements only work for pure panny body/lens combinations. But it doesn’t sound like that’s the case. Faster motors should help with all bodies, and improvements on the body side (sensor read out rate, search algorithm) should help all lenses.
Of course, there could be some other jonx that they didn’t talk about in the article.