The digital audio coding is a digital electrical signal representing a sound wave. It consists of a sequence of integer values and are obtained from two processes: sampling and quantifying the digital electrical signal.
Sampling a digital audio signal.
Sampling involves setting the amplitude of the electrical signal at regular time intervals (sampling rate). To cover the audible spectrum (20 to 20000 Hz) is usually sufficient sampling rates of more than 40000 Hz (CD-Audio standard uses a 10% higher rate in order to contemplate the use of filters not ideal), with 32000 samples per second width similar to the FM radio or a cassette tape, ie band would enable register components up to 15 kHz, approximately.
To play a particular frequency range sampling rate of slightly more than double (Nyquist-Shannon sampling theorem) is needed. For example, in the CD reproducing up to 20 kHz, employ a sampling rate of 44.1 kHz (Nyquist frequency of 22.05 kHz).
The quantification is to convert the level of the fixed sample in the sampling process, typically a voltage level in a finite integer value and predetermined range. For example, linear quantization using a 8 bit linear coding discriminate between signal 256 equidistant levels (28). You can also make non-linear quantification, such as log-quantifiers as mu-law or A-law, which, for example, still using 8 bits operate perceptually as 10 linear bits for low amplitude signals on average, as eg human voice.
The most used form of linear PCM digital audio is the audio CD: 44.1 kHz sampling rate and 16-bit linear quantization (measuring 65536 different signal levels) and, in practice, can record analog signals with components up to 20 kHz and signal-to-noise over 90 dB.
