Wouldn't the latency for the group be the latency od slowest group member? I thought the latency of a single hop for a single wireless mic or headset exceeds the maximum tolerable. > -----Original Message----- > From: ietf <ietf-bounces@xxxxxxxx> On Behalf Of Toerless Eckert > Sent: Wednesday, March 18, 2020 6:08 PM > To: Randy Presuhn <randy_presuhn@xxxxxxxxxxxxxxxxxxx> > Cc: ietf@xxxxxxxx > Subject: Re: videoconference technology challenge: choral rehearsals > > Hmm... 130 orchestra members, each in their own home, good > microphones, good headsets. Audio from all members gets centrally mixed > and sent back to members. Likewise composer video. I wouldn't be worried > about sound floor, a bit of adaptive mixing should well be able to take care > of that. > > Depending on distance, I would be worried about latency. > Differential latency is likely a big killer, but also absolute. > > There are i think a bunch of those remote orchestra experiments, but i have > not seen good quantitative number about its impact (differential/absolute). > > There was a nice exhibition 2? years ago at SF MOMA of a (prerecorded) > small 6..8? people orchestra doing this gig within a larger home, each in a > different room. Would love to put a configurable latency mixer into such a > setup and see what happens when i play with latency and jitter ;-) > > Toerless > > On Wed, Mar 18, 2020 at 02:32:17PM -0700, Randy Presuhn wrote: > > Hi - > > > > On 3/18/2020 2:02 PM, Cullen Jennings wrote: > > .... > > > That would be cool. The mainstream systems don???t support that > because the noise floor gets so high as you start mixing lots of participants > but some universities have experimented with simular things. Google > Stanford Laptop Orchestra. > > > > Thanks. > > Totally not my field, but... > > I think the concern about the noise floor makes sense if one were > > trying to get good fidelity for a soloist. But in the normal use case > > for a choral rehearsal, unlike the business meeting scenarios that > > seem to predominate, most of the time at least 10%, and more likely > > 75-100% of the participants would be singing, so there'd an awful lot > > of signal atop whatever the noise floor added up to. (Another reason > > why I think it might make sense to pre-process the divisi separately.) > > Plus, if one is used to performing with an orchestra, one learns to > > filter out an awful lot of noise while taking all the timing cues from > > the conductor. :-) > > > > Randy > > -- > --- > tte@xxxxxxxxx