[pulseaudio-discuss] Network Audio with Pulse

Matt Patterson matt at v8zman.com
Mon Apr 7 16:45:49 PDT 2008

Yeah, sounds like you have the rtp thing. I assume you realize you can 
have multiple multicast addresses so there can be simultaneous streams 
that don't collide/get mixed.

I don't think there is a way you can avoid the mesh (multicast is 
basically a mesh, just for free) unless you designate a server machine. 
In which case you could set up a single tunnel sink to each client 
machine and then have all the switching happen on that machine. I use 
the remap module to split each sink into 4 inputs (could be a tunnel 
sink), then connect each input to a different mpd instance, and control 
what is heard out each device by muting 3 of the 4 inputs. I end up with 
16 remapped sinks in this case (4 output devices * 4 remapped sinks 
each). I will be adding a 5th zone to my whole home audio soon, so that 
will make it 25. The 16 sinks/streams seems to cause no undue load on 
the system (Core 2 Duo 2180), we'll see how 25 does :)

I haven't played with the tunnel sink module.


Jim Duda wrote:
> Matt,
> Thanks for the feedback.
> I understand your point about using the command line interface module. 
> I actually would end up using the socket approach from perl once I had 
> it all working properly.
> I believe I understand how the rtp approach would work. I think what you 
> are doing is as follows.  All stream senders would send on the rtp_send 
> side, connecting the rtp_send.monitor to the default alsa sink (for 
> local sound).  All other machines would have an rtp_recv, and send the 
> output of rtp_recv to the default alsa sink.  Each of these rtp_recv 
> would be muted by default.  If machine B wants to join in, machine B 
> would unmute it's rtp_recv and thereby get the stream.  Do I have a 
> basic understanding of how this approach would work?  I have played with 
> this to some degree, so I think I understand.
> I assume using the tunnels would work in a similar fashion.  However, 
> you need to build a mesh of tunnel connections.  In my case, with 4 
> nodes, the mesh is 3 tunnels for each mode, 4*3=12 tunnels in total. 
> Each receiving node would then mute each tunnel by default, turning on 
> the one it wants.  The annoying part of this approach is that you have 
> to decide which source you want to connect to, whereas, with the rtp 
> approach, you simply join the "collective".
> I have played with the combined_sinks somewhat too.  However, since 
> upgrading to FC8, the pulseaudio server keeps crashing when I attempt to 
> use a combined sink.  I've been trying to get a core dump to Lennart, 
> but I haven't been able to get gdb to help me out, I keep getting some 
> problems with some threading library (or something of that nature).
> I'm now trying to understand how the paprefs gui mechanism works.  I 
> haven't been able to get any of the options to be enabled for operation, 
> all the controls are grayed out, trying to understand why.
> Jim
> Matt Patterson wrote:
>> I played with something similar but my goal was an audio multiplex 
>> switch all on the same machine to the rtp lag issue was less apparent. 
>> As for controlling it, I just wrote a simple python app that connects to 
>> the unix socket (same thing pacmd does) and I issue commands to load 
>> modules, mute inputs, etc so things can be controlled. I then wrote a 
>> php wrapper around the python app so my web based audio control could 
>> come about.
>> To go this route you have to make sure the command line interface is 
>> available either via TCP or Unix socket (I chose unix socket). If you 
>> like I would be happy to send my hacktastic python code to help get 
>> things moving.
>> I believe that using the tunnels allows you to have the sync feature 
>> where rtp doesn't, so maybe play around with getting them working???
>> Matt
>> Jim Duda wrote:
>>> There was a similar thread, back around New Year's regarding Network 
>>> Audio.  I've read the entire thread a few times.  I'm having similar 
>>> problems, yet different.
>>> I'm looking for some advice as to how best to use network audio with pulse.
>>> I have multiple linux computers in my house, four to be specific.  One 
>>> operates as a file server, one as a desktop, and the other two as 
>>> diskless think clients which basically operate as media players.
>>> I use these computers in a home automation network in my house using the 
>>> misterhouse home automation software (misterhouse.net).
>>> All machines are running stock fedora 8.  The two thin clients, are not 
>>> running the full suite of services which a desktop would.  For example, 
>>> they are not currently running avahi or hal (but could if necessary).  I 
>>> can certainly turn on what needs to be running.
>>> I'm hoping to perform the following using pulseaudio.
>>> Let's call my machines A, B, C, D.
>>> Let's assume that some stream is started on machine A, playing in the 
>>> living room.  I would like to be able to have that same stream play on 
>>> machines A and B simultaneously.  I don't care if I have to go to stream 
>>> A and say send to machine B now, or, go to machine B and ask B to fetch 
>>> a stream from machine A.  I can make both work.  I want to be able to 
>>> drop the stream to B at anytime.  I realize that if the source stream 
>>> stops, then all streams would in essence stop too.
>>> I need to be able to access the controls to switch streams using a 
>>> command line application which I can call from perl using the system 
>>> call.  I've seen the stream switch in pavucontrol.  I've seen the 
>>> move-sink-input in pactl (but failed to get it to work, I guess I don't 
>>> understand how the params work as I always get some error message).
>>> At some other time, I may want to have machine C join in the stream with 
>>> machines B, C.
>>> How is this best to accomplish?
>>> 1) Should I use combine_sink on the source machine?
>>> 2) Should I use rtp?
>>> 3) Should I use tunnel_sink?
>>> I've played with rtp.  Although it works, the audio isn't synchronized. 
>>>     Maybe it should be synchronized, but I haven't found that to be 
>>> true.  I can hear latency delay between multiple machines.
>>> I know how to play across the network, using the pulseaudio alsa plugin.
>>> I'm now trying to play with the network options in the paprefs 
>>> application.  On my main server and desktop, all the network audio 
>>> options in paprefs, configure local sound server, are all grayed out.
>>> Each machine has these modules installed from FC8.
>>> sudo yum list '*pulse*'
>>> Installed Packages
>>> akode-pulseaudio.i386             2.0.2-4.fc8            installed
>>> alsa-plugins-pulseaudio.i386      1.0.15-3.fc8.1         installed
>>> pulseaudio.i386                   0.9.8-5.fc8            installed
>>> pulseaudio-core-libs.i386         0.9.8-5.fc8            installed
>>> pulseaudio-esound-compat.i3       0.9.8-5.fc8            installed
>>> pulseaudio-libs.i386              0.9.8-5.fc8            installed
>>> pulseaudio-libs-devel.i386        0.9.8-5.fc8            installed
>>> pulseaudio-libs-glib2.i386        0.9.8-5.fc8            installed
>>> pulseaudio-libs-zeroconf.i386     0.9.8-5.fc8            installed
>>> pulseaudio-module-gconf.i386      0.9.8-5.fc8            installed
>>> pulseaudio-module-jack.i386       0.9.8-5.fc8            installed
>>> pulseaudio-module-x11.i386        0.9.8-5.fc8            installed
>>> pulseaudio-module-zeroconf.i386   0.9.8-5.fc8            installed
>>> pulseaudio-utils.i386            0.9.8-5.fc8             installed
>>> Available Packages
>>> audacious-plugins-pulseaudio.i386 1.3.5-3.fc8            fedora
>>> fluxbox-pulseaudio.i386           1.0.0-2.fc8            updates
>>> gstreamer-plugins-pulse.i386      0.9.5-0.4.svn20070924. fedora
>>> kde-settings-pulseaudio.noarch    3.5-38.fc8             updates
>>> pulseaudio-module-bluetooth.i386  0.9.8-5.fc8            updates
>>> pulseaudio-module-lirc.i386       0.9.8-5.fc8            updates
>>> Both the avahi and gconf modules are loaded as displayed in the Modules 
>>> section of the Paprefs Manager display.  What else is necessary?
>>> I have auth-anonymouns=1 loaded for both native-protocol-unix and native 
>>> -protocol-tcp.
>>> I've read all the documentation on the pulse wiki many times.  I've 
>>> browsed through all the postings on the mailing list over the past 6 months.
>>> I'm just playing now with the server and desktop which have full blown 
>>> stock fc8 installs, just to figure out how all this works, then I'll 
>>> incorporate the thin clients later.
>>> The whole package is rather complicated and I haven't had much success 
>>> in putting it all together.
>>> I've done my homework.  I just cannot get it working ...
>>> Thanks,
>>> Jim
>>> _______________________________________________
>>> pulseaudio-discuss mailing list
>>> pulseaudio-discuss at mail.0pointer.de
>>> https://tango.0pointer.de/mailman/listinfo/pulseaudio-discuss
> _______________________________________________
> pulseaudio-discuss mailing list
> pulseaudio-discuss at mail.0pointer.de
> https://tango.0pointer.de/mailman/listinfo/pulseaudio-discuss
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/pulseaudio-discuss/attachments/20080407/6aa6e0bc/attachment.htm>

More information about the pulseaudio-discuss mailing list