The visuals are modular and have a couple of major entry points: one is "analyse", called with a 4608 byte buffer of sound data, which does the necessary FFTs or whatever is needed for that visual. The second entrypoint is "draw", which actually draws the visual. Some visuals simply don't use the "analyse" (or will survive fine without it) - like the 3d dot-shapes.

The output stage is in the DSP, which can select digital in (from the SA1100), radio or aux input. The DSP applies all the sound processing (bass, treble, fader, balance, EQ) and then puts it out of 4 independent DACs it has onboard.

We will be releasing info about how to do your own visuals, but we are going to change the call interface to them - at the moment, they're not very friendly to write as they are just a binary block which you call the first word of with opcodes (as they were written in the RiscPC's built-in assembler by Toby :) ).

Hugo