Michal Seta wrote: > Hi all, > > I am working on a project in which I want 5 musicians to walk around a > building and play. They will carry netbooks (mics and headphones) and > their signals will be sent to a concert hall where it will be > broadcasted. Additionally, a "conductor" will decide which individual > stream will every one of those 5 musicians listen to. So I need a > 2-way communication on each netbook and 5 in and 5 out on the server. > The server will be wired. > > I tried oggcasting but latencies were just unacceptable (10-20 > seconds) so I started experimenting with jack.udp. i'd say you could get <3s roundtrip with careful tuning, but that is still unacceptable for live playing "on a beat". > I am getting very > promising results with one machine on a wired LAN and the other over > WiFi but I get a lot lost packets. I do not mind to loose some > quality and some stuttering and I don't mind some latency (although > jack wants to pump everything in real time) but the netbook that's on > WiFi has a lot of trouble even getting the audio data out or reliably > bringing it in, even in areas where WiFi signal is very strong. > > I also tried NetJack but I cannot even establish a master/slave > relationship between 2 machines (if one is on a WiFi). Perhaps it > demands a reliable bandwidth? I am not sure, I did not spend much > time with it. you want to use the CELT codec over netjack. quite decent packet loss concealment, good sound, almost as good as vorbis. and you need to avoid packet collisions. with 10 (!) real-time audio streams on one wifi, there will be collisions all over the place. the only way around that i can see is to use several wifi cells, and keep each performer on a separate one, if possible. still, risky. it also means you need to hire thugs that jump on every spectator trying to pull her/his iphone or other wifi gadget out. but: you could try vorbis encoding downstream over http (with re-send in case of packet loss) at just under a second, and netjack/CELT for the monitoring, where dropouts are more tolerable. if, as usual in experimental arts, money is not an issue at all ;), use pro-audio uhf transceivers. all in all, that would amount to 5 two-channel in ear sets for the musicians' monitors and one or two beltpack transmitters (depending on whether mono is ok or you want stereo from the netbooks), i.e. up to 10 frequencies in each direction. the "walk around the building" part is tricky, though. if it's an open hallway, all fine, same for lightweight walls. for thick walls, some remote antennae could help, and if you can do without diversity (i.e. you can tolerate the occasional dropout), you can use the A/B antennae to cover two separate areas. renting such a pile of gear will set you back at least 1.k5€... _______________________________________________ Linux-audio-user mailing list Linux-audio-user@xxxxxxxxxxxxxxxxxxxx http://lists.linuxaudio.org/mailman/listinfo/linux-audio-user
