E-mail: calvert
Date: 30 Dec 2004
Time: 10:10:51
From: Calvert Dayton, SIA
E-mail: calvert
Date: 30 Dec 2004
Time: 10:10:51
Seems like the most likely cause of stuttering is probably insufficient buffer size. And I believe the driver for the US 122 does allow you to tweak the buffer size it uses, among other things.
Basically, audio comes into the computer in chunks. The smaller the chunks, the lower the latency through the input stage but the faster the chunks need to be picked up and used. If the computer can't pick up the chunks and use them fast enough you end up with dropped buffers, which I'm guessing is what you're describing when you say "stuttering."
Increasing the buffer size up to several times typically has little impact in terms of the amount CPU time required to process the data, since you're typically talking about a difference of a few hundred, to a couple of thousand samples at the extremes. So you really don't save anything on that end by decreasing the buffer size. But the act of picking up those chunks and moving them around is a relatively high overhead operation that remains fairly constant in terms of CPU consumption until you get into buffers many times the sizes we're talking about. So that's something you obviously want the computer to have to do as infrequently as reasonably possible.
The good news is that we don't much care about latency for our application. So if you find that bumping up that buffer size helps with this problem there's really no good reason not to. If you're doing hard disk recording, the difference between 5 and 10 ms of throughput delay can be pretty significant. But for analysis purposes 20 or 30 ms or more is probably no big deal. As long as it's good data to begin with, you should get good answers from it.
Hope this helps