> > I thought it was jitter? I thought the ear could handle dropouts, but > > time-variance turned it all into noise? > > True, jitter can be a problem also. Usually you digitize and send in > 20ms packets or 50 per second. So you may not notice a few dropped > packets, especially if they don't all happen at once. And normally if > packets arrive out of order they are dropped. There might be algorithms > for trying to smooth over dropped packets, I'm not sure. But VOIP > implementations can and often do try to measure latency and implement a > jitter buffer. It adds a bit of latency depending on how big the buffer > is but it works similar to a buffer on a CD player (remember CD players > and skipping?) so that by adding a little delay the buffer can put > packets back in order and play them at a continuous speed so the ear > does not hear any errors. But if latency varies too widely (jitter) the > buffer could run out or overflow and cause drops. >
yep...there's all kinds of forward error correction, but mostly if you lose a packet you try to interpolate it based on the packets before and after using either simple math, or a codebook with statistical probablities to regenerate an approximate set of samples. That's also why most VoIP stuff runs over UDP, less overhead and you don't care about sequence and flow control that you get from TCP. -T -- [email protected] http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list
