[mythtv-users] Sync'd MythMusic

Alan Snyder ax763 at yahoo.com
Wed Feb 25 09:20:43 EST 2004


> On Tuesday 24 February 2004 6:45 am, Adam Biskobing wrote:
> > I have multiple frontends arrayed around my house, and I 
> was wondering 
> > if anyone had figured out a way to sync the music playing 
> in mythmusic 
> > to all the frontends in the house so that they are not all only 
> > playing the same playlist but at the same time.  Has anyone found a 
> > way to do this?
> 
> My experience with trying to stream to multiple machines in 
> the house or 
> office has convinced me it's not worth it unless all the 
> locations are 
> completely out of earshot of each other.
> 
> There's no practical way to completely synchronize the 
> players, even if they 
> all use a streaming protocol that uses a timebase shared by 
> all players, the 
> local buffers and whatnot seem to always cause enough 
> variation in sync that 
> if you can hear more than one player it's obvious that they 
> are not in sync. 
> Even if you get the sync well within 1 second, it's still 
> amazingly annoying 
> to listen to if it's not perfect.
> 
> Start up an icecast server, serve up a stream and press play 
> on two computers 
> at the same time and you'll see what I mean. Streaming just 
> isn't any good 
> for use as extension speakers. ;)
> 

I'm guessing that none of the current protocols are designed for time
synchronization.  Every protocol buffers in order to deal with network
delays and the delays in the sound systems are variable.

It seems to me that three things would be needed for true sync:

1. Identical (within reason) soundcard buffer delays, both in the sound
system and in hardware.
2. NTP-type time sync among the stations
3. Keyframes in the audio stream that include an absolute timestamp, so that
each system can adjust for clock differences.

Because normal desktops can't be assumed to be time synced, I can't imagine
that any of the existing streaming protocls have the needed time-stamped
keyframes.

A problem, as James points out, is that the total skew to avoid pretty
annoying effects is probably in the tens of milliseconds.  That puts a
pretty tough limit on the errors allowed in (1) and (2).



More information about the mythtv-users mailing list