Sound formats

On of the first application of digital electronics was to convert the analog audio to digital. The simplest way is PCM (Pulse Code Modulation) that samples the analog signals using a sampling frequency. The Nyquist-Shannon theorem say this sampling rate must be double of the highest expectable frequency. Those sampled pulses got then converted in a digital value using a analog to digital convert. In telephony voice got sampled at 8kHz with 8 bit, resulting in 64kbit/s

Looking at this approach shows much possibilities to reduce or compress the data.

First approach was ADPCM (Adaptive Differential Pulse Code) that makes the hight of the steps of the AD converter depending on the previous signal tendency. This way voice usually sampled with 8 bit could be reduced to 3 bits without noticeable quality loss. The sampling rate for voice can be 8kHz * 3 bit= 24kbit/s

The next step in data compression makes use of psychoacoustics. The human brain picks dominant frequencies and ignores adjacent frequencies with a lower amplitude. Using such approaches sampling rates for voice could be reached in the area of 4.8kbit/s. It should be noted that those formates loose data in the original signal, however the data lost is considered to not be noticeable by the human brain.

Most modern audio codecs make use of psychoacoustics. Such codecs are mp3.

Mp3 has also the possibility to add meta data (ID3 tag) next to the sound data. This data is often be shown by media players and hold things as title of the song. To see and edit this data programs as mp3info or using the gui gmp3info from easytag from or id3ed.

Linurs Servernest startpage