deshead wrote:(ADPCM being the simplest: it uses 4 bits per sample instead of 16 ... So it sounds like crap.)
A 320kbps mp3 stores each sample using only 4 bits on average, but yet it doesn't sound like crap.
ADPCM basically stores each sample as a difference between the sample and what that sample would be predicted to be based on the previous samples. It's actually quite good in most circumstances, and near-lossless (but not actually lossless) as long as the waveform doesn't deviate too widely from the prediction model.
WAV is just a different name for RIFF, which is a sibling to AIFF (both coming from IFF). They are all essentially the same - a generic container which stores multiple chunks of data. They can store images and video and metadata and all that other fun junk too. WAVs can even store mp3 data if you want (the original Fraunhofer CODEC actually put the mp3 stream into a WAV container instead of an MPEG one).
The thing about file names is that they're there for the user's benefit, not for the computer's benefit. There's no structural difference between an mpg, mp2, or mp3 file, or between m2a or, m2v, or between m4v, m4a, m4p, aac, mov, qt, etc., they're just there to tell the user what kind of file it is or which application to use to open it by default (which just happens to usually be taken care of by the file browser these days).
As far as the difference between mu-law (often written u-law as a mu symbol looks like a u) and linear PCM, u-law is stored in a format which gives more precision to low signal levels, which often gives it better quality (very similar to how monitor response is non-linear). It was used a lot back when most sound hardware was only 8-bit and so it was important to make the most of the precision. In practice, 16-bit linear is about as good as it's going to get, at least for final mastering.
In general, 16-bit linear PCM WAV is the most compatible with other tools.
