That's not quite how it works. The separation used by the AF is completely fixed by the distance between the (at least two) microlens+aperture+sensor systems th

That's not quite how it works. The separation used by the AF is completely fixed by the distance between the (at least two) microlens+aperture+sensor systems that make up an phase AF sensor. The microlens/aperture system of the AF masks essentially the back plane of the lens (which follows from the exit pupil) for the AF sensor (interacting mainly with the exit telecentrism of the lens, which is why AF speed and acccuracy can widely vary with lens type, even if the same f-stop is used). Also a reason (besides the fact that many digital sensor designs perform better with telecentric designs due to increased scattering and reflections in the sensor itself compared to film) why modern lenses are almost always strongly telecentric designs. A completely separate matter is the circle of confusion which is indeed directly related to the aperture of the lens. (I hope I got most of the optical terms correct, I don't normally discuss this topic in English)