Will they never end ?!?!?!
I was finding that audio was fine up to about 1'50", and then breaking down. So, back to hex dumps.
Although I was re-inserting a new sync frame on transmit, I wasn't taking into account the fact that I'd lost the incomign sync frame, so my reference point was a frame out. This lead to audio being up to 840ms (two frames) out.
Now, this creates a problem. I have to have at least 840ms of buffer to know whether I've lost two syncs, which means a delay between receiving the audio and playing it out. But, I've seen many instances where three syncs get lost, so that's at least 1260ms. So, there is a tradeoff between solid audio and delay. I'd take solid audio every time....
Why oh why oh why did they design a system like this ???? Surely a sequence number would have done much better with a session ID and separate voice, data and control planes ? This is just plain nasty, especially given that the path from the sender is hardly the most reliable transport mechanism.
Subscribe to:
Post Comments (Atom)

0 comments:
Post a Comment