Originally posted by falconeye Lenses aren't designed for PDAF (or CDAF). They have (or don't have) an AF motor or screw and a given mechanical accuracy. E.g., my old FA31 focusses in under 1s (using K-5 CDAF).
The necessity of dedicated CDAF lenses is a µFT phenomenon which must be related to technical problems of FT lenses other brands don't have. Maybe, related to their focus by wire concept, I don't know.
I read something about this when Sony came out with an adapter to mount their Alpha lenses on the NEX bodies. Focusing their A mount glass on the NEX was really slow with it. They explained it as thus:
Phase Detect autofocus has the ability to look at the current focus and know the direction of defocus (front or back) and approximately by how much. So it's easier to snap to the focus point directly. However, their Contract Detect autofocus algorithms basically samples the image many times per second as the focus is changed. Then the processor decides if the image is getting more contrast or less. The most contrast point is assumed to be the most in-focus. If you watch your Pentax K-5 do CDAF, you'll notice that it often overshoots the focus point, then comes back. So it might be a similar method. Once it gets close to the focus point, it moves the focus in small increments back and forth until it gets the point just right. In order to optimize the hardware design for this type of algorithm, they need the lens to focus quickly and to be able to change the lens from focusing forwards to backwards quickly. So they designed the lenses such that the elements of the lens that move during focusing are as light as possible. Lighter elements means less inertia, which means they can change position more quickly.