On Tue, Jan 26, 2010 at 3:22 PM, Anthony Minessale <anthony.minessale@gmail.com> wrote:
We use a conference with energy detection on 12+ hours a day and I don't recall losing any of the audio. The instant you breach the level that same packet is sent, and you have to have several consecutive packets below the level to stop.
Interesting could you post the conferences.xml config for that particular conference? I have yet to find a way of using energy detection that doesn't miss the first few ms of sound, for example the word "Testing" often comes out as "esting" perhaps with the tail end of the T coming across. It's not a problem when everyone is a native english speaker but when on a conference containing people with a poor grasp of english or people who reply with lots of quick utterances such as "yup"
My guess is you are talking to something lame like a Sonus who is resetting its jitterbuffer when you start talking again.
In production it's hitting Cisco gateways but I've experienced this in a lab environment consisting of a mixture of Snom 360, 820 and Polycom IP6000 devices. What would be great is to have start/stop talk events while still always sending audio from participants to the conference