[LAU] netjack and WiFi

[torbenh]

On Sat, Jan 09, 2010 at 02:30:35PM -0500, Michal Seta wrote:
> Hi all,
> 
> Happy new decade!
> 
> Back on that WiFi jamming issue I wrote about a while ago.  Thank you
> all for the various suggestions.  I have spent most of my time trying
> to get somewhere with netjack as I found it the most promising, not
> only in terms of efficiency but also scripting and control.  However,
> I have hit some brick walls which are certainly due to the nature of
> WiFi networks (packet collisions and such) which make my setup very
> unstable.
> 
> First of all, in my setup, there will be one "server" computer
> collecting 5 signals from the 5 musicians and playing them for the
> audience.  Moreover, the operator (or software) will send back 5 mono
> signals to performers (one signal per performer).  The idea is that
> each performer hears only one instrument at a time.
> 
> Currently the show stopper lies in alsa_in and alsa_out components.
> Whenever I run both on my netbook (Atom 1.6G with Atheros
> Communications Inc. AR5001 Wireless Network Adapter (802.11g)) they
> consume around 70-80%CPU and they peak at around 50% each whenever
> they choke on net over/underruns.  They eventually segfault (and
> sometimes bring the jack server down with them.  This particularly
> true whenever the WiFi signal signal strength is weak (around 50-60%).

i have commit the change to the old deadline algo to svn.
it seems to work. at least it seems to work fine with a netem loopback
that looses 20% of its packets.


-- 
torben Hohn