On Fri, 2010-01-22 at 18:34 -0500, Tim E. Real wrote: > Good, but that would introduce glitches. > What if the next sync comes at two or three times as long as the last one?
Say the last sync came 20 ms ago. We guess that the next will be something like the same but and devide the next 20 ms seconds in 4 periods of 5 ms each. Three things can now happen: a) The next pulse comes very early, perhaps after 10 ms? Dump out on the interface whatever was supposed to play in the later, in the last periods. b) The next pulse comes very late, perhaps after 30ms? In that case we have have already played everything. There will be slight gap which will go unnoticed given the dramatic change of tempo. c) The next sync comes approximately as expected, perhaps after 19 or 21ms? Well, then there is no problem at all :-D > One simple (slower) tempo change and the rest of the song is out of sync. No! > We should not guess, we cannot assume or look ahead. Yes we can! (Hey, Roland already figured this out back in -85 or so) > The only solution seemed clear to me: Straight ahead linear real time, > no assumptions of what's to come. This is soo 1982 ... > Really the resolution it is not noticeable, even at 20BPM. 20BPM is a resolution of 125 ms?? I am now convinced you do not not record yourself playing keyboard. _______________________________________________ Linux-audio-dev mailing list [email protected] http://lists.linuxaudio.org/listinfo/linux-audio-dev
