Hi,
Ive also asked something similar in the Flac forum, but this question also applies to WavPack (or indeed any lossless stereo codec with a configurable block size!).

Im considering using a stereo lossless codec for realtime transmission of speech in a VoIP system. I know this sounds strange but it is necassary to preserve the original waveform upon decode for my application. So I know flac can use a variable block size, and I assume WavPack can do the same, thus I can control, "blocking/framing delay". My question is, once I have a block of samples (small block for minimal blocking delay), what is the delay caused by the algorithmic complexity of WavPack for coding that block? Also is the decode operation within similar delay bounds as the encode?

Cheers

Chris

This is not what you asked, but it is indeed strange to see a lossless codec in this application. Even assuming an average comp. ratio of 25% you still have to reserve full uncompressed bandwidth for peaks in the data rate. Since there is no buffer, you can't count on those 25%, and might as well transmit uncompressed.

Ok, I tested several sources recorded from a mono mic input, removed the silence (simulating VAD), and filtered each with a binaural HRTF. This results in stereo signal, which I then used Flac to encode. Compression was dependent on the HRTF filter (spatial location) but averaged to around 0.4. I guess what you're saying is that for some blocks coded lossless (whatever the codec), there maybe more or less compression than others? So the data rate is block dependent?