Sound Quality - Digital Audio

Digital Audio

Digital audio is stored in many formats. The simplest form is uncompressed PCM, where audio is stored as a series of quantized audio samples spaced at regular intervals in time. As samples are placed closer together in time, higher frequencies can be reproduced. According to the sampling theorem, any signal with bandwidth B can be perfectly described by more than 2B samples per second. For example, for human hearing between 0 and 20kHz, audio must be sampled at above 40kHz. Due to the need for filtering out ultrasonic frequencies resulting from the conversion to an analog signal, in practice slightly higher sample rates are used: 44.1kHz (CD audio) or 48kHz (DVD).

In PCM, each audio sample describes the sound pressure at an instant in time with a limited precision. The limited accuracy results in quantization error, a form of noise that is added to the recording. To reduce quantization error, more precision can be used in each measurement at the expense of larger samples (see audio bit depth). As each additional bit added to a sample, quantization error is reduced by approximately 6 dB. For example, CD audio uses 16 bits per sample, and therefore will have quantization noise approximately 96 dB below the maximum possible sound pressure level.

The amount of space required to store PCM depends on the number of bits per sample, the number of samples per second, and the number of channels. For CD audio, this is 44,100 samples per second, 16 bits per sample, and 2 channels for stereo audio leading to 1,411,200 bits per second. However, this space can be greatly reduced using audio compression. In audio compression, audio samples are processed using an audio codec. In a lossless codec audio samples are processed without discarding information by packing repetitive or redundant samples into a more efficiently stored form. A lossless decoder then reproduces the original PCM with no change in quality. Lossless audio compression typically achieves a 30-50% reduction in file size. Common lossless audio codecs include FLAC, ALAC, Monkey's Audio and others.

If additional compression is required, lossy audio compression such as MP3, Ogg Vorbis or AAC can be used. In these techniques, lossless compression techniques are enhanced by processing audio to remove details that are unlikely or impossible for human hearing to perceive using principles from Psychoacoustics. After the removal of these details, lossless compression can be applied to the remainder to greatly reduce the file size. Lossy audio compression therefore allows a 75-95% reduction in file size, but runs the risk of potentially reducing audio quality if important information is mistakenly discarded.