Audio data is characterized by the following parameters, which correspond to settings of the A/D converter when the data was recorded. Naturally, the same settings must be used to play the data.
- sampling rate (in samples per second), e.g. 8000 or 44100
- number of bits per sample, e.g. 8 or 16
- number of channels (1 for mono, 2 for stereo, etc.)
Approximate sampling rates are often quoted in Hz or kHz ([kilo-] Hertz), however, the politically correct term is samples per second (samples/sec). Sampling rates are always measured per channel, so for stereo data recorded at 8000 samples/sec, there are actually 16000 samples in a second. I will sometimes write 8 k as a shorthand for 8000 samples/sec.
Multi-channel samples are generally interleaved on a frame-by-frame basis: if there are N channels, the data is a sequence of frames, where each frame contains N samples, one from each channel. (Thus, the sampling rate is really the number of *frames* per second.) For stereo, the left channel usually comes first.
The specification of the number of bits for U-LAW (pronounced mu-law -- the u really stands for the Greek letter mu) samples is somewhat problematic. These samples are logarithmically encoded in 8 bits, like a tiny floating point number; however, their dynamic range is that of 12 bit linear data. Source for converting to/from U-LAW (written by Jef Poskanzer) is distributed as part of the SOX package mentioned below; it can easily be ripped apart to serve in other applications. The official definition is the CCITT standard G.711.
There exists another encoding similar to U-LAW, called A-LAW, which is used as a European telephony standard. There is less support for it in UNIX workstations.
(See the Appendix for some formulae describing U-LAW and A-LAW.)