On Thu, 25 Jan 2007, Roman Katzer wrote: > Bill et al, > > On 24/01/07, Bill Unruh <unruh@xxxxxxxxxxxxxx> wrote: >> I may not be up to date, but I would be very doubtful that Linux could >> handle that (or any computer processing). The problem is the latency. Ie, >> sound is read in in chunks, processed and then sent out in chunks. Thus >> there is a time lag between the time the music comes in and goes out. It >> is >> hard to imagine this as being less than 10 time slices ( eg 1/4000 sec for >> a 44100 rate system) That time lag would mean that the music went out that >> much later, and mixing with the direct sound could produce some >> interesting >> effects. Now if the speakers were say sufficiently far from the >> performers, >> then that time lag might be OK. (sound travel time). It would be a tricky >> juggling act however. > > I can run up to 40+ channels through a Linux box in real time with a > lag (latency) of less than 20 ms. 20 ms is the acceptable latency > when it comes to syncing sound to a motion picture, because it's at or > below the JND for humans. > > That said, a 1024 sample chunk at 48 kHz is about 21 ms, so running > ALSA/Jack with a block size of 512 samples would allow you to do this > kind of processing with a small enough latency. > Run all of your audio through the DSP box and there won't be any > interference with "live" sound. 20 ms is 6 meters. 20ms causes interference at multiples of 50Hz-- ie all the way from the bass on up. It is like getting very strong reflections from a wall 3m behind the perfomers (stronger than the original.) Now that may or may not cause problems, I do not know. You do not need to play with 512 samples since most DSP effects can be done on line with 4-20 length buffers. Ie, you do not operate on the chunks. It is that because of the buffered nature, tend to get the first sound back out a buffer time later than it came in. If 20ms is really enough time not to cause acoustic problems, then the latency is irrelevant. > > The monitor speakers won't have to be fed through the DSP system, > they're only for the musicians to hear themselves in time before the > sound from the house system gets to them. Also, the audience should > be primarily fed the output of your house system. > > By the way, 1 ms equals roughly one foot (34 cm) of sound travel. So a > latency of 20 ms would be less than 7 meters. You would have to think > about time-aligning your house system before worrying about the > latency of your DSP system. > > So with that, I'd say you should think about your physical acoustics > (room modes, distance from speakers to listeners, different room > behavior depending on number and location of listeners etc) before > trying to get the latency down. > > I don't see audience distance as a big problem. When you're up front, > you (rightly) expect to hear the instruments or monitor speakers > directly, and you will. When you're further from the stage, you > expect the nicely mixed, mastered and equalized sound. You'll get it, > and the delay shouldn't be a problem. Localiasation of the sound tends to depend on getting the sound from the performers directly first. If the sound from he speakers gets to you first it becomes very tiring on the audience and becomes very hard to localise the sound, from what I understand. But certainly your argument would suggest that the latency is not as big a problem as I would worry that it is. > > Best regards, > Roman > ------------------------------------------------------------------------- Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys - and earn cash http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV _______________________________________________ Alsa-user mailing list Alsa-user@xxxxxxxxxxxxxxxxxxxxx https://lists.sourceforge.net/lists/listinfo/alsa-user
