[AG-TECH] Real Time Music Collaboration

[Markus Buchhorn]

At 12:52 PM 12/03/2002 -0700, Joey Mendoza wrote:
>First, we have a drum major (conductor, click track, etc.), providing a
>zero-latency pulse to everyone.  The people farthest away from the
>audience at the back of the football field (ie. those with the greatest
>latency) play exactly in time with the drum majors pulse.  Those closest
>to the audience (ie. minimal latency) play in time with the sounds being
>produced from behind them.
[...]
>Now the actual logistics of getting something like this to work over the
>AG may be pure madness, but having each node *try* to play in time with
>the nodes with the greatest latency might work.

This is a very good approach, and one I hadn't considered before (I was 
also a former drummer and drum major of a military band :-) but we never 
spread out that far for it to be an issue)

I think there's an additional problem where the audience is distributed - 
i.e. the latency varies between multiple sources and multiple receivers. 
The joys of multicast doing its thing (ideally) means you have a 2D mesh of 
sources and receivers. So while the drummer may be furthest from you, he 
may be closest to me - so the trumpets who are close to you are far from 
me, and the scheme goes south (or north, depending where you sit).

Two options leap to mind. One is to have a mixing node, which is defined as 
the notional audience and hence the coordination target, and it feeds the 
signals back to the rest of the AG - so it's not a single multicast tree, 
but a pair of them. The other is the clicktrack approach as suggested, but 
perhaps do it in a more technical fashion (we are a network research 
project after all :-) ), say using a multicast group to emit a timestamp 
signal, which everybody plays to. Each receiver then muxes/buffers all the 
sources it sees to the multicast timestamp-series before playout. This 
would be very hard on the musicians though, since you're playing alone to a 
metronome, which ain't no fun, and for some pieces just does not work.

This is a neat problem! :-)

Cheers,
         Markus


Markus Buchhorn, Information Infrastructure Services,   | Ph: +61 2 61258810
Markus.Buchhorn at anu.edu.au, mail: CompSci,CSIT Bldg #108|Fax: +61 2 61259805
Australian National University, Canberra 0200, Australia|Mobile: 0417 281429