Originally posted by wjjstu CDAF is a computer algorithm not voodoo magic. It takes input from the sensor, does work, and produces output for the AF drive system. In a properly modular design it can be re-used in different devices (and Pentax's camera platform does appear to be modular enough for them to re-use things). What you're saying is like saying a sorting algorithm on a smartphone costs less than a sorting algorithm on a supercomputer because of reasons. Plus as somebody else pointed out, they did end up doing a bastardized version of it for the K-01. The whole Q/K-01 program just seemed like a huge waste in resource allocation, one-offs in trying to be different, and both now abandoned.
No K products in 2011 is irrelevant. They were allowed to launch the entirely new Q mount with cameras and lenses.
Yes, basic CDAF is very simple but it's slow and it hunts a lot. Fast and accurate CDAF is voodoo magic.
Good CDAF depends on choosing a sensor, designing the read-out, and writing data handling algorithms that are much more like those used for video than for still photography because CDAF depends on fast frame rates and fast processing of the framestream. Given that Pentax explicitly picks higher DR over higher framerates for sensors and deprioritizes video in general, they are much less likely to be able to easily create fast CDAF.
BTW, modular design is the last thing you want for fast, real-time CDAF. All the code for sensor operations, frame-stream processing, AF estimation, and lens motor control need to be very tightly coupled to eliminate latency and to choreograph everything with respect to everything else. Modular may be less costly to write, easier to debug, and easier to reuse but it is much slower.