AF systems, whether they are contract detection and phase detection, they work the same way as in many image/signal processing systems. Hardware aside, the key component is the algorithm that computes the result. This is something I'd like to know. Maybe I should talk to a image/signal processing researcher at my University who knows the internals of contrast detection well in order to find out the details.
May be I can help you here... Unlike a Video AF system where several full scans of a subject are required and a DSP algorithm is needed to compute the defocusing info (hence the aliasing error), phase detectors are usually single element of various shapes (rectangular, crosshair... etc) where digitization is sometime not necessary.
This simple experiment
has been taught in many EE labs at numerous university campuses. The phase detector system is so simple in fact that it can drive the focusing mechanism directly with only relatively simple signal conditioning circuit that they are universally used
in all CD and DVD players to lock on their optical tracks!
In early camera applications the only requirement from their processor was just to detect which phase detectors to use (usually the one with the strongest signal, ie the highest contrast) in a multi-position AF elements set-up, and to switch (or user selectable) the corresponding output to drive the lens mechanism. The key point here is that in a phase detector setup, the entire system could stay in the analog domain
; Unlike a Video AF system... where frame rates, quantization, synchronization, and sampling errors from a bursty signal which could get in the way of the signal processing software trying to decide what to part of the AF signal to lock on. My bet is if you vary the scan rates, there'll be frequencies where the A1 will work best (but also a few where other Video AF cameras will fail!)
All in all there's nothing wrong with the A1... It's working as designed and all you did was to find a peculiarity in an otherwise excellent camera which I don't own (yet). 8)