On Sun, Jul 13, 2003 at 12:46:53PM -0700, Mark Knecht wrote: > > > > I don't understand this whole 'drift' thing. If I'm able to listen to a > > previously recorded drum track and record a guitar track over it (in > > sync with the recording) with either card, how is it possible for them > > to be out of sync with each other? > > Peter, ... > I hope this helps clarify the physical reasons that this is difficult. > > There is a software reason also. Linux/Alsa will onlt interrface to a > single clock, so clocks being generated by two sound cards are not something > that Alsa is designed to handle. :-/ -> ! Would not a software/driver-solution to synchronize the clocks/adjust the drift be handy here? At least theoretically, somthing similar to ntp. This would need 1: a way to measure the clock-difference exact enough 2: a way to delay the the stream running over the hardware within the driver within very small units. 3: ... ? But I doubt that this idea could get far, probably timing would become a rather statistical than exact procedure in this region. And expensive too. Also, maybe clocks are not even stable over time (temperature, etc.) and would have to be remeasured periodically. Also, I bet this idea has been around earlier and if possible we would already have it realised. The psycho-acoustic boundary for discerning discrete acoustic events/clicks is about 1/20 s, AFAIR, so 50 ms would have to be beaten to satisfy the ear. Björn