On Tue, Dec 09, 2003 at 09:25:42AM +0100, Clemens Ladisch wrote: > > > I seem to recall that it does, but that the packet size is smaller, > > > making the latency problem easier to deal with. > > > > Isoch - 1394 has a timer operating on the bus. This timer happens > > (roughly) every 125uS. > > USB isochronous transfers happen once per millisecond. OK, so its the difference between 44 or 45 samples per packet at 44.1kHz (USB) and 5 or 6 (1394). So, a plausible jack period of 64 samples will be 11 or 12 1395 packets, but always 125uS of jitter of course, (10 or 11 packets at 48k). I dont really have any idea how bad that is, its 8% of the availble time slot for processing the 64 samples. You could lessen the effect of the jitter by having triple-buffering in software, as in a 3x64 PCI soundcard - I dont know wether thats better than just going to 128 samples or not. FWIW, for my needs I rarely go below 256 samples/period anyway, even though my setup is capable of it, but I understand there are people who really need very low latency. At 256 samples/period the jitter is only 2.1% of the time slice, which is unlilkly to matter - cache temperature and so on has a far greater effect than that. - Steve
