Silence Suppression

Hello,

Does anyone have any experience of silence suppression? I am interested in the potential bandwidth savings and the effect on quality.

I read somewhere that it can be more effective to reduce the bandwidth during silent periods, rather than shutting off the stream altogether. This is because:

1. It sounds more realistic to get some noise from the far end rather than generating comfort noise

2. By keeping the voice stream on, jitter buffers can still assess the delay on the link and payout the voice correctly, rather then having to jump into action when the silence ends.

Any thoughts?