RAOP streams are delayed

[takkat]

Pulseaudio 0.9.21 streams very well over an attached AirportExpress? in Ubuntu Karmic 9.10 am64. However the streams have a quite long delay of 3 to 6 seconds thus making them unsuitable for video.

Interestingly, volume control from PA/Gnome-Applet reacts immediately as expected but volume controls from applications have the same 3 to 6 s delay.

It would be very nice to have this delay solved in one of the following milestones (if possible).

[coling]

Yeah this is a known limitation of the driving code just now.

Volume changes are passed over a control channel to the device so are acted on immediately even if the device has a full buffer.

If the app in question is clever, it passes on the volume changes to lower down the stack the change will be immediate but if it attenuates the data it's sending out it will have the delay.

I'll be fixing this at some point when I rewrite the timing code. Just finding it hard to find the time for that as it's a rather complex process.

[takkat]

Replying to coling:

If the app in question is clever (...)

all the apps I tested are not ;-)

I'll be fixing this at some point when I rewrite the timing code. Just finding it hard to find the time for that as it's a rather complex process.

Good to hear, that you have the issue in mind - gives us a little chance for the future. As for now I'm so happy that it works at all that I can live well with the delay. Fortunately the volume control works fine even when using media keys.

[bendavis78]

I get about a 7-8 second delay here. Would be awesome if this could be fixed!

[Chris56]

Will there ever be any work done on this problem?

I would rather call it a defect than an enhancement. With the 5 secs of delay that I observe seeking within some music is almost impossible...at least a nuisance.

I must say I have no idea of the internals of Airport Express music handling. However, I discovered, that as soon as I enable the iTunes option, a permanent! stream of 200 KBytes/sec is being transmitted to the Airport Express. Given both wlan connections (computer-router, router-Airport) this chews up 400KBytes of (limited) wlan bandwith for "silence". Is this the way it should be??

Chris

[coling]

What it's labelled as really has no bearing whatsoever so feel free to call it what you want :)

It is fully intensional that the silence is transmitted as without it, the APEX gets into a strange state and we can no longer reconnect to it without restarting PA. Even unloading and reloading the module isn't enough. It's all rather strange. So sending silence keeps the connection open and thus allows us to use it.

Also the timing methodology is totally basic. It uses the buffer of the APEX itself to implement flow control where as we should really be feeding it data in real time. This is all complex stuff.

There is also a much more accurate timing model available in the Airtunes v2 protocol, but it's not fully understood or reverse engineered yet. Some work has been done to get it working but the results still cause choppy sound at present.

It's all work in progress, but as a lot of it is guess work it's a very slow process. If you'd like to help out with reverse engineering the protocol and such like, please let me know and I'll put you in touch with the right people.

[Chris56]

I wasn't aware of the fact, that the protocol for the Airport Express has not been published. What a tedious task. Friends of mine using iTunes on windows switched to Linux and that way the problem arose. Therefore I cannot be of any help. Are there any other - openly documented - protocols and devices that transmit music over wlan? The permanent 400 kBytes/sec over the shared wlan bandwidth (s.a.) is a fairly harsh condition of the RAOP module ...